|
1. |
Expert systems for structure elucidation of organic molecules by spectral methods |
|
Russian Chemical Reviews,
Volume 68,
Issue 7,
1999,
Page 525-547
Mikhail E. Elyashberg,
Preview
|
|
摘要:
Russian Chemical Reviews 68 (7) 525 ± 547 (1999) Expert systems for structure elucidation of organic molecules by spectral methods ME Elyashberg Contents I. Introduction II. Methods for identification of structural fragments III. Generation of structural formulae of isomers IV. Prediction of spectra of probable structures V. Pathways of development of expert systems VI. Strategy of molecular structure elucidation VII. Conclusion Abstract. The state-of-the-art of the investigations aimed at creating expert systems for establishing the structure of organic molecules from IR, 1H and 13C NMR spectra is analysed. Computer methods used for identification of molecular frag- ments, generation of their structure and spectra prediction are considered. Principles of the creation of modern expert systems and general strategy of solving structural problems are discussed.The bibliography includes 174 references. I. Introduction More than 30 years have elapsed since the publications 1 ±5 in which the problem of using mathematical methods and computers for establishing the structure of organic molecules from their spectra were first considered. Even in these articles it was shown that the structural-group analysis (SGA) can be formalised using methods of mathematical logic and performed on a computer,1, 2 while application of graph theory and combinatorial analysis makes it possible to generate automatically (perform mathemat- ical synthesis) structural formulae from fragment sets.3 In this time, many interdisciplinary sciences have united on the basis of mathematical algorithms as computer programs to form a new line of investigation, application of artificial intelligence (AI) to the determination of molecular structures from spectral data.Since World War II, spectral methods have received wide acceptance in chemistry. It will not be an exaggeration if we say that IR, NMR spectroscopy and mass spectrometry revolution- ised the methodology of chemical investigations and `chemical' thinking and have become one of the main sources of structural information for chemists. Unquestionable are the advantages of spectral methods, viz., their versatility, fast operation, ease of constructing a computer interface and the use of well developed theories (this particularly concerns IR andNMRspectroscopy).It is these features, as well as the accumulation of a great body of information on empirical spectrum ± structure correlations (SSC), ME Elyashberg All-Russian Research Institute of Organic Synthesis, ul. Radio 12, 107005 Moscow, Russian Federation. Fax (7-095) 261 07 77. Tel. (7-095) 261 51 61. E-mail: elyas@molspec.msk.ru Received 25 February 1999 Uspekhi Khimii 68 (7) 579 ± 604 (1999); translated by AMRaevsky #1999 Russian Academy of Sciences and Turpion Ltd UDC 543.42:681.3 525 526 531 532 537 542 545 that have stimulated investigations on the application of AI in chemistry. To formalise the process of molecular structure determina- tion, it was necessary to find mathematical tools adequate to the problem considered.Since a molecule has a discrete nature and its energy levels are discrete too, mathematical logic, graph theory and combinatorial analysis were chosen as appropriate mathe- matical tools. It is discrete mathematics which describes the models of the `black box' type that has become one of the main instruments of mathematical chemistry, which has progressed intensively in the last decades. At first, mathematical synthesis of the structures from the atoms and fragments seemed to be merely a stage in the identification process of organic compounds; however, soon it became clear that its importance goes far beyond the scope of the problem considered. In fact, innumerable sets of organic mole- cules designed by the structure generator can be interpreted as enumeration of logical implications from a system of axioms that form a basis of the theory of molecular structure.At the same time, knowing only the structural formula of a compound, one can predict a number of its important properties. This made it possible to pose a problem of the development of mathematical methods for predicting the structure of the compounds with desired properties and, vice versa, the properties of compounds from a structure given (studies of N S Zefirov's scientific school in this line of investigation have gained world-wide acceptance). Rich experience accumulated by chemists shows that most predictions on the possibility for unknown compounds to exist in nature and to be synthesised, made on the basis of the classical theory of the structure of organic substances, were borne out, the percentage of wrong predictions being relatively small.6 Formalisation of chemical research based on integration of many sciences has become an important result of the studies on the application of AI in molecular spectroscopy.By and large, the development of concepts and methodolog- ical approach in this new area of study was completed by the early 1980s, which was embodied in a number of monographs 7 ±10 and reviews.11 ± 15 The efforts of investigators from many countries have resulted in general principles for creating AI systems [currently, they are called expert systems (ES)] and in developing corresponding algorithms which were tested on a large number of problems.About twenty ES for molecular spectroscopy have been documented. They differ in their possibilities, use different algorithms and instrumental tools and are aimed at the applica-526 tion of different methods of spectral analysis. However, common to these systems is the fact that they simulate reasoning of an expert (spectroscopist) when solving an identification problem. This thinking process involves the following main stages: (1) structural-group analysis of an unknown compound; (2) generation of possible structural formulae from the frag- ment sets found; (3) test for the correlation between structural formulae and experimental spectra (and additional information) using spectra prediction of the assumed structures.Each stage is simulated by a corresponding block of an expert system. The above-mentioned blocks are relatively independent and their functions are usually realised using different mathemat- ical methods. Currently, methods for identification of the frag- ments, structure generation and prediction of their spectra have become almost independent divisions of computer spectroscopy and mathematical chemistry. To come as close to solving actual problems using ES as possible, it was necessary to investigate the nature of the structure determination problem. Peculiarities of this problem were consid- ered 16 as an example of the inverse problem.17 As is known, the inverse problems belong to the so-called ill-posed problems: they do not have a unique solution unless a priori restrictions are imposed.18 Important conclusions were drawn that it is impossible to perform a structure determination without the attention of a chemist and that all problems should be solved in the interactive mode.At that time, it was also appreciated that information on the spectra ± structure interrelations is usually `fuzzy' (ambigu- ous). With this in mind it was suggested 19 that methods of fuzzy mathematics 20 should be used for SGA formalisation and approximate spectra prediction. The results obtained by different research groups up to the early 1990s have been reported in reviews.21 ± 23 At about the same time a considerable interest has been expressed by investigators in algorithms of artificial neural networks (ANN),24 ± 27 the application of which in different areas of computer chemistry seems to be rather promising.The number of studies in this investigation line grows rapidly. It appeared to be so new for chemistry and chemometrics that no mention of ANN was made, e.g., in a `Chemometrics' monograph 28 published in 1986. At the same time it became clear that ES have a chance to become an analytical tool of organic analysis only if they `learn' to operate on large non-additive structures (e.g., the structures of natural compounds) and possess a high degree of user-friend- liness. The efforts of several research groups focussed on this line of investigation have led to the creation of new ES (see, e.g., Refs 29 ± 32).In the present review, the experience mostly accumulated in the last decade in the area of creation of the ES oriented to the molecular structure determination from IR, 1H and 13C NMR spectra is generalised. First, methods for identification of frag- ment sets, structure generation and prediction of their spectra are considered and then strategies of structure determination are analysed taking the most perfect ES as examples. Particular emphasis is placed on discussing the problems that arose in the course of ES operation and analysing possible ways of over- coming these problems. II. Methods for identification of structural fragments To identify the structural fragments, most ES use characteristic features of functional groups and other structural elements in the molecular spectra.Usually, the knowledge base of an ES consists of molecular fragments, ranges of changes in their characteristic features (frequencies in IR spectra and chemical shifts in NMR spectra) and the rules relating the spectra to the structure. In a number of systems, libraries of the structures and corresponding spectra are included in the knowledge bases in addition to molecular fragments. Depending on the type of the knowledge ME Elyashberg base, different methods of structural information retrieval are used. 1. Formal logic method It has been shown 1, 33 that the scheme of relations between the elements of molecular structure and their spectral features can be described by a discrete structural model, therefore application of mathematical logic in the studies of the spectrum ± molecular structure interrelation naturally follows from the nature of the object studied.It is assumed that the molecule is additive or partially additive and each fragment Ai of a set A of fragments related to the problem given is characterised by one or several intervals of changes of characteristic features oi 2 O (O is a set of spectral features). Boolean algebra was used as a mathematical tool of the theory of SGA. This made it possible to derive in the general form logical relations connecting structural elements to their characteristic spectral features and to construct a model of reasoning when identifying structural fragments.Boolean algebra deals with elementary propositions and their combinations, which are called Boolean functions. Combinations of elementary propositions are constructed using `and' (conjunc- tion), `or' (disjunction), negation and implication operations. The latter has a meaning of a complex proposition `if..., then ...' and is denoted by an arrow (?). For instance, the sentence `if a molecule contains a fragment Ai, then feature o is observed in its spectrum' is represented by the Ai?o implication. Elements of Boolean algebra and SGA theory are reported in the monograph.7 In the general case, reasoning of an expert who performs SGA seems to be as follows: if a Boolean function T(A, O) correctly reflects interrelation between the spectrum and molecular struc- ture, then the presence in the spectrum of a combination of spectral features, which is described by a Boolean function R(O), indicates the presence in the molecule under study of at least one combination of the structural elements defined by a Boolean function f(A), i.e.T(A, O) ? [R(O) ? f(A)]. This expression is a general formulation of the problem of qualitative spectral analysis in terms of Boolean algebra, it is a logical equation with respect to function f(A). Solution of the logical equation is reduced to calculating the function f(A) using an algorithm which is based on the concept of designation numbers and reported in the monograph.7 Function f(A) con- tains fragment sets, each of which can explain characteristic features in experimental spectra.The approach described serves as the basis for the SGA block in several generations of the RASTR system 34 ± 36 and is used in the X-PERT,29 CHEMICS,37 EXPIRS 38, 39 and other systems. This makes it possible to select successively possible fragments for producing logically consistent sets appropriate for further struc- ture generation. However, such an approach has a drawback consisting in too `rigid' logical determinacy. For instance, a small deviation of a characteristic feature of the fragment in the experimental spectrum from the limits of a given interval leads to the loss of the corresponding fragment. In addition, the know- ledge base of the algorithm remains to a great extent similar to correlation tables and, hence, vast information presented in terms of natural language in numerous monographs on molecular spectroscopy (regularities, empirical rules, etc.) is not used.In this connection attempts have been undertaken to create such an ES in which SSC are described in more detail.19, 40 2. Methods of predicate logic To create an ES which operates on the knowledge presented in terms of a natural language with professional restrictions, it is necessary to find appropriate mathematical tools. The require- ments placed on such mathematical tools are caused by the character of the knowledge formulated in terms of natural language. On the one hand, descriptive (declarative) knowledge,Expert systems for structure elucidation of organic molecules by spectral methods in particular, in the field of IR spectroscopy (see, e.g., Ref.41), is characterised by some fuzziness and ambiguity. The statement that, usually, stretching vibrations of theC=Cbond are observed in the IR spectrum in the region near 1650 cm71; however, sometimes they are shifted to the region 1685 cm71 can serve as an example.41 On the other hand, logical inferences in the course of solving qualitative problems are also performed in terms of a natural language with inherent ambiguity and hidden estimates of reliability (e.g., `often', `rarely', `sometimes', etc.). The man thinks using fuzzy concepts rather than numbers. Therefore, theoretical tools appropriate for adequate simulation of our knowledge and reasoning must meet the requirement figuratively formulated by L Zadeh who initiated fuzzy mathematics:42 `The theory of nature must reflect the fact that nature paints by arbitrary strokes rather than by a ball pen'. To represent the knowledge on IR spectroscopy, applied logical calculus 19, 40 (ALC) based on the theory of first-order predicates 43 and theory of fuzzy sets 20, 42 was proposed.The basic principles of ALC are as follows. A logical function P(X1, X2, ..., Xn), where Xi [ M(Xi), which takes a `true' or `false' value on substituting the objects belonging to a certain object domain is called an n-place predicate. Most often, the predicate reflects properties of the objects and relations between them.For instance, let predicate P(X) reflect the property of the substances to have a strong IR absorption band near 1700 cm71 and let X1={acetone} and X2={propylene}. Then, obviously, P(X1)=1 (`true') and P(X2)=0 (`false'). If M(Xi) is a set, then a fuzzy subset F of the set M(Xi) is defined as a set of ordered pairs {Xi, mA(Xi)}, where mA(Xi) is a characteristic membership function, which varies from 0 to 1 and indicates to which degree the element Xi is a member of the subset F. For instance, the membership function of the proposition `there is a band near 1710 cm71 in the IR spectrum' will be written as follows: mA(1710 cm71)=1. In this case it is assumed that there exist certain bounds, e.g., 1700 and 1720 cm71, and beyond them the function mA(Xi) is zero.If predicate variables are assigned the meaning of the elements of fuzzy sets, then we get fuzzy predicates. Each n-place predicate symbol is associated with mappingMn into a set of the values of a characteristic membership function, which in the case of ALC is called a truth function mp, 04mp41. A system of predicates describing the IR spectrum ± structure interrelation was developed to describe the spectrochemical knowledge. In particular, predicates characterising the IR spec- trum (`there is a band of vibrations of the type z of a structural group g in the spectrum; the band is characterised by frequency Xi, intensity Yi and halfwidth Zi') and molecular structure (`the molecule contains a fragment R1', `the molecule contains a fragment R1 separated by n1 bonds from m1 fragments R2', `the molecule contains a ring of size j, which includes a fragment R3', etc.) were introduced.Specific predicates characterise the state of aggregation of the substance, solvent type, etc. To describe changes in the data, second-order predicates are introduced, and using them it is possible to indicate the direction of changes in the variables which appeared in the first-order predicates (e.g., `increases', `decreases', `disappears', `becomes split into d compo- nents', etc.). Using such a system of predicates and conventional logical operations, it is possible to formalise declarative knowledge to a great extent. For instance: (1) `stretching vibrations ns of compounds with one double C=C bond and alkyl substituents (Alk) absorb in the region 1640 ± 1680 cm71' STRUCT(C=C, Alk)?SPECTR(ns, C=C, 1640, 1680); (2) `conjugation of double bonds of two alkyl-substituted groups results in splitting of the band of stretching vibrations ns into two bands' 527 Q1[STRUCT, C=C, Alk(C=C)]? ?Q2 [SPECTR, ns, C=C, X1(SPL,2)].The logical formula G and the predicate symbol P are associated with a set of outside values of the truth function me, 04me41. The external value of the function me reflects the knowledge quantification (`often', `sometimes', `in some instan- ces', etc.). For instance, the statement `stretching vibrations of the C=C bond are often observed in the region 1640 ± 1660 cm71' is represented by elementary formula [0.8]SPECTR(ns, C=C, 1640, 1660 {10}), where [0.8] is the external value of the truth function, {10} is the segment length (in cm71) on both sides of the bounds of the region 1640 ± 1660 cm71, on which the trapezoidal member- ship function decreases to zero. To determine the value of the truth function mi appeared in the elementary formula, it is necessary to multiply the external value of me by those of the membership functions of fuzzy constants and variables in this formula. In this case, taking into account the properties of a triangle, we get m(X)=0.5 at X=1665 cm71; at me = 0.8, the mi value will be equal to 0.560.8 = 0.4. Spectrochemical knowledge can be divided into facts (expres- sions containing constants) and regularities (expressions contain- ing at least one logical variable).It is convenient to operate on structurised knowledge. To structurise the knowledge, a class is specified of general regularities that are valid for most organic molecules (e.g., `there are no small rings containing a triple bond in organic chemistry'). Principal functional groups possessing characteristic spectral features (C=C, C:C, C=O, C:N, OH, etc.) are accepted as structurisation units of other knowl- edge. The molecular fragment is described by its core (C=C, C:C, etc.) to which surrounding atoms are successively added, thus respectively forming first-, second-, third-level fragments, etc. The knowledge hierarchy thus formed is a tree graph called a declarative knowledge network.The description of a network node contains the fragment structure, possible types of its vibrations, regularities and facts for each type of vibrations and those characteristic of the given fragment. If declarative knowledge is considered as a system of axioms of the theory used for solving an applied problem, the SGA problem is represented in terms of statements (theorems) that should be proved or rejected. It has been shown 43 that the classical resolution method adapted to the theory of fuzzy predicates can be used for automated proof of the theorems (automated SGA). The formalism suggested makes it possible to solve SGA problems as well as those of structural interpretation of IR spectra.The results of interpretation can be used for self-learning of declarative knowledge, in the course of which the facts and regularities are tested, values of characteristic membership func- tions are changed and new facts and regularities are revealed. The approach described was realised as a research prototype of an ES written in the Prolog algorithmic language.44 The system reveals consistent fragment (network node) sets that explain the spectrum. Simultaneously, interpretation of the IR spectrum is performed: each frequency is assigned to a certain type of vibrations of corresponding fragments with an indication of the value of the truth function mp for the assignment given. Predicate logic and Prolog algorithmic language were also used for structural intepretation of IR spectra in the EXPEC system.45 ± 47 The knowledge base of the system includes a rule interpretation generator.As in the preceding system,44 fragments included in the knowledge base are hierarchically structurised in the form of a network of declarative knowledge. When inter- preting the spectrum of unknown compound, the network is scanned to find the rule for each fragment, according to which the relation between the fragment and spectrum is analysed. Each result obtained is associated with its probability. Explanatory facilities make it possible to trace the reasoning process of the program.528 The knowledge base of the system is formed on the basis of SSC known from the literature and is supplemented using the rule interpretation generator,45 as well as due to input of expert's knowledge.To perform automated rule generation, a library of IR spectra of organic compounds belonging to the CHO class is used. The program generates a set of potential rules, selects the least correlated ones and minimises their number. The presence of the rule generator 45 is an advantage of the EXPEC system. Automated rule generation involves four steps: (1) selection of structural fragments; (2) database retrieval to find all structures containing a Si fragment; (3) search for the intervals of characteristic features for Si and (4) selection of intervals appropriate for using interpretation rules. The problem of the knowledge base formation is of fundamental importance for creating an efficiently operating ES, therefore the above-men- tioned approach to automatisation of this process appears to be rather promising.The interpreter (inference machine) selects only those network fragments for which the presence of groups in the network nodes was validated. For instance, the rules concerning detection of phenol compounds can be used only after the presence of benzene ring and hydroxy group was confirmed. For each substructure, the probability of its presence is calculated using a trapezoidal probability distribution function for a given interval (e.g., this shape was chosen for the band membership function of the absorption region 19, 40). Selected fragments are arranged in order of decreasing the probability of their presence.The investigators from different research groups suggest different rules for selecting possible fragment sets (see, e.g., Ref. 48). It seems that these studies will progress and make it possible to come right up to creating ES for the molecular spectral analysis that interact with a chemist in terms of a natural language with professional restrictions. However, on this way the investigators will inevitably face many difficulties. Overcoming them will require joint efforts of the experts who carries out interdiscipli- nary research such as, e.g., mathematical linguistics, mathematical logic, knowledge technology, etc. 3. Application of artificial neural networks Beginning in the early 1990s, the attention of chemists was drawn to the possibilities of new promising mathematical tools of computer chemistry, viz., artificial neural networks (ANN).Particularly rapid is the increase in the number of studies on the application ofANNto interpretation and classification of spectral data. A neural network is a simplified model of the human brain, consisting of several layers of neurons that send signals to other neurons depending on the input signals received. Such networks function according to the `black box' type and possess a common ability to construct empirical models of the systems for which theoretical dependences between the input and output are too complicated or even unknown. Models are obtained as a result of the network training.In the course of training, the network is represented in the form of input ¡¾ output pairs related by simu- lated transformation. A network trained on these examples is capable of predicting the output signals from the input signals not `seen' at that instant. The training procedure may be time- consuming (tens of hours); however, a network once trained generates the answer or prediction almost instantaneously. For instance, a network can be trained to generate structural informa- tion (output) retrieved from a spectrum (input) or to predict a spectrum (output) from structural information (input). By apply- ing this procedure to the spectra, it is possible to establish which functional groups can be present and absent in the specimen and to complement these data with reliability estimates of the results obtained. Let us consider general principles of designing neural net- works 24, 49 in more detail.Strictly speaking, ANN is a computational structure consist- ing of a number of interconnected elementary units called nodes or neurons. In a simplified form, a neuron functions as follows. The `signal' values si (i=1?m) at the neuron input (we will call them inputs) are multiplied by weighting factors wi and summed up to give the function Net: Net=s1w1+s2w2+.....+ smwm . Then, the Net value is transformed into the output signal of a neuron using a nonlinear transfer function. The sigmoid function sf a 1 a expO¢§siwiU 1 is most often used as a transfer function. This signal is then used as the input signal for other neurons and the organisation of the connections between neurons defines the network architecture.A large number of neurons and interrelations between them makes the network behaviour more complex. The network can be modified by `adjustment' of its parameters, among which weighting factors are the main ones. Adjusting the network parameters in order to obtain a desired behaviour is called the network training. This process is carried out iteratively using a learning algorithm on a set of examples. At each iteration, the network outputs are compared with the reference values (examples) and, if they appear to be different, the program performs correction of weighting factors. The network is trained until coincidence of outputs with references.In the first studies on structural interpretation of IR and NMR spectra, multilayer ANN trained using the back-propaga- tion of errors (BPE) algorithm 24 were most often used. In these networks, each neuron is connected to all neurons in the preceding and succeeding layers. The first (input) layer receives the input data and distributes them over the succeeding layer. The last (output) layer, or decision layer, is responsible for outputs of the entire network. Other layers, if any, are called hidden layers (Fig. 1). Information is propagated from the input layer through hidden layer(s) to the output layer. For instance, an IR spectrum encoded as the intensity values taken at points uniformly dis- tributed along the abscissa axis is used an the input vector.The output vector coordinates are certain numerical values (often, zeros and unities) indicating the presence or absence of particular structural features. The number of points used for encoding the spectrum defines the size (the number of nodes) of the input layer while the number of structural features (fragments) introduced determines the size of the output layer. There is no generally accepted rule for setting the number of nodes in hidden layers: it has to be large enough to provide correct data processing; however, increasing the number of nodes increases both the learning time and risk of the network `over-training'. The latter consists in decreasing the network performance after a certain Input signals Output signals Figure 1.A scheme of neural network. ME Elyashberg Input layer Hidden layer Output layerExpert systems for structure elucidation of organic molecules by spectral methods number of iterations (they are called cycles or epochs) in the course of training. After random initialisation of the initial values of weighting factors wi, the network training is performed on a set of spectra of the compounds containing fragments associated with desired outputs. (This set is called the training set.) There are numerous variants of this algorithm; however, all of them are based on the same principle, a steepest-descent minimisation of the total error, calculated as the sum of differences (or the squared differences) between the network outputs and desired results.To determine the network performance (efficiency), a set of examples (the test set) is used. This set should differ from the training set in order to measure the network capacity for general- isation instead of learning. The size and composition of the training set strongly affects the learning efficiency. For instance, increasing the number of examples increases the overall performance of the network, but only up to a certain point.50 The effect of the composition of the training set appears to be much more complicated. The effect of presents/absents balance for structural features in the molecules, i.e., the proportion of compounds presenting a structural feature to those not presenting it in the training set has been studied.49, 51, 52 In particular, it has been found 49 that increase in the proportion of presents in the training set increases the probability of recognising these fragments in the compound analysed to the detriment of the absents.At the same time, it was found that there is a certain limit of presence under which the network is unable to perform correct classification. The setting of a well-balanced and diversified training set can be difficult, especially if the network to be trained has many outputs. The number of network outputs is of great significance: according to the reported data,50, 53 mono-output networks perform better than the networks with many outputs; moreover, large networks are more difficult to optimise than small ones (networks with smaller number of outputs).Since the best result for each output may occur at a different training time, the ANN training is stopped at a stage where the averaged (but not for each output) network performance will be optimum. The use of many mono-output networks for the determination of a large number of structural features is much more time-consuming than the use of a single network. Therefore, advantages and drawbacks of using a single multi-output network are opposite to those of using several mono-output networks. A hierarchical system for spectrochemical studies, which should retain the best of both approaches, was proposed.54 In such a system, a multi-output top-level network performs a rough classification of the spectra.Depending on the structural feature considered, the hierarchy is extended to deeper levels where further refinement is performed. The top-level network outputs activate the first-level networks and so forth through the whole hierarchy. The design of a hierarchical system of neural networks can be considered as a compromise between one multi-output network and many mono- output networks. Its main characteristics are also intermediate. It was found that it is easier to balance the components of the training set for each network of the hierarchy than for a single large network (in some sense, as simple as for a mono-output network). In addition, optimisation of the learning time is easier for a smaller number of outputs, and the modular conception makes it possible to modify the system without the entire reconstruction.A hierarchical system gives information on structural features on different levels: from general characteristics (`a compound contains a CO group') to more specific ones (`a compound contains an acidic group'), the number of networks to be activated in each particular case remaining small. For instance, there is no need to find out whether a compound is a primary or tertiary alcohol if the OH group has not been recognised by the top-level network. However, mention has been made 49 that the adjustment of a hierarchical network is a complex task and requires a priori knowledge and consideration of the dependence of sub-level networks on upper-level network performance.Let us consider the results of studying a hierarchical system as an analytical tool for determination of the fragments using IR spectra.49 The experiments were carried out using a set of 8620 compounds and their IR spectra taken from the Sadtler Bio-Rad library. Calculations were performed on an IBM RS6000 work- station. The system was trained to recognise 33 structural groups whose hierarchy is shown in Fig. 2. As can be seen, recognition of carbonyl group was carried out using two specially designed networks, one for wide-spread structural features (carboxyl, ketone, ester and amide) and another for rare structural features (acid chloride, anhydride and aldehyde).The following network parameters were used. Input vector. The IR spectrum in the region 500 ± 3596 cm71 was represented as a set of 259 points separated by increments of 12 cm71. The weighted mean of absorbance in the vicinity of each point was scaled to a 0 to 1 range and used as input. Hidden nodes. Networks contained only one hidden layer, and the number of hidden nodes was set to 30 for all networks. Transfer function. The classical sigmoid function was used. Structure encoding: 0 and 1 were respectively used to encode the absence and presence of a structural feature. Learning algorithm. All networks were trained using the BPE algorithm with a learning rate a=0.05 and a momentum m=0.90 (Ref. 24). Decision making. Since the network outputs were real values from the 0 ± 1 range, two thresholds were necessary for decision about the presence or absence of each structural feature f in the molecule: accept level (presence threshold) ALf and reject level (absence threshold) RLf of a structural feature.If the output number is between RLf and ALf, the answer is considered ambiguous and the fragment is reported as `not classified'. These decision thresholds were optimised individually for each network. Training and test sets elaboration. The sets of examples required to train 10 networks constituting a hierarchical system were built by random selection from initial population with the constraints to reach a prescribed proportion of compounds containing preset structural features. For the top-level network, aromatic C6 (6) C7OH (2) C=O (3 ± 4) Top level (1) C7NH (5) C=C (7) Figure 2.Hierarchical organisation of networks. The numbers of acti- vated networks are given in parentheses. 529 ortho meta para primary secondary tertiary amides I amides II CONHCO CONHNH monosubstituted disubstituted (10) trisubstituted phenols alcohols (8) COOH CHO anhydrides Cl7C=O C7O7C=O C7CO7C HNCO (9) amines I amines II pyrroles C=CH2 CH=CH2 C=C7C=C530 the whole collection of spectra was used. Sub-level networks were trained using specially designed protocols. As was mentioned above, in hierarchical structures the net- work dedicated to recognition of the fragments belonging to a specific class is activated only if it recognised the example as a member of this class.For a structural feature, two types of misclassification are possible: (1) the fragment was identified in the compound in which it does not occur; this case is called false present (AdP); and (2) the fragment was not identified in the compound containing it; this case is called false absent (PdA). False absents are not propagated through the hierarchy, since the structural feature is considered as absent. However, false presents raise a new problem: sub-level networks must refine classification of the fragment which actually is not contained in the compound. Special experiments showed that the hierarchical system is robust and bears the change of networks without losing the global capability of the hierarchy.The best training was found to occur for the training sets of 1000 examples and the number of cycles was between 5 and 20. The best results were obtained for C=O and OH groups that are easily recognised in the IR spectra. The following indices were suggested for measuring the performance of the entire hierarchical system:49 Ptf � a general characteristic of the presence of a structural feature (the total present found); it is defined as the ratio of the number of compounds in which the presence of the fragment was correctly identified to the total number of compounds containing the given structural feature; Atf � the same as above, as applied to the absent structural features (the total absent found); Pte � the total error of decision of the presence of the structural feature (the total present error), which is defined as the ratio of the number of compounds containing the feature but identified as absent to the total number of the compounds containing the given structural feature; Ate�the total error of decision of the absence of a structural feature (the total absent error), which is defined analogously to the preceding index.The results of testing the system using the criteria introduced are shown in Table 1. Comparison with the reported data 55 shows that, on the whole, the hierarchical system performs better than multi-output network; however, the results qualitatively rank below those obtained using well optimised mono-output net- works.Currently, a few studies dedicated to application of ANN to the selection of molecular fragments has been reported; however, investigations in this area are gathering force. Thus, analyses of IR spectra using ANN were reported.52 ± 61 An ANN was used to identify amino acid residues in proteins from their `fingerprints' in 2D NMR spectra.62 Adjustment of the networks in the case of joint application of different spectral methods to elucidation of structural fragments was reported.51, 63 This investigation line seems to be of particular interest; therefore, let us consider the results reported in the last two articles in detail. A study 51 was carried out in which IR and 13C NMR spectra and molecular formulae of organic substances served as input signals for a neural network preliminarily trained to recognise molecular fragments.TheANNcontaining 512 inputs, 85 outputs and *50 neurons in the hidden layer was used. The training set comprised 1560 compounds for which IR and 13C NMR spectra were known. A total of 128, 294 and 90 inputs were respectively specified for IR spectra, NMR spectra and molecular formulae. Such a distribution was chosen taking into account the estimate of the amount of information contained in each data type. IR spectra were encoded in the same way as in the study by Cleva et al.49 13C NMR spectra were encoded by present/absent indicators of signals at intervals of 2 ppm. The indicators for signals of different multiplicity were stored in prescribed positions of the input vector (e.g., encoding the chemical shifts of quartets in the spectral range 0 ± 64 ppm was performed using 32 intervals to which positions 128 ± 159 of the input vector corresponded).The molecular Table 1. Performance of global hierarchy. Atf Fragment C7OH CO7OH Ph7OH Csp37OH CH27OH CH7OH Cquatern.7OH C=O CO7OH CO7OR CO7R CO7NH CO7NH2 CO7NHR CO7NH7CO CO7NH7NH CHO OC7O7CO Cl7CO C7NH C7NH2 C7NHR CO7NH Pyrrole Aromatic C6 C67OH monosubstituted disubstituted ortho meta para trisubstituted C=C 93.3 91.2 92.8 91.3 91.1 90.6 91.6 94.3 94.1 93.6 92.9 91.5 93.4 94.9 95.6 95.6 86.8 92.2 94.6 92.0 91.1 93.4 91.6 93.8 79.0 80.9 80.8 70.0 73.9 73.7 71.7 76.4 93.6 93.2 93.3 89.4 C=CH2 CH=CH2 Csp27C=C formula was encoded by the numbers of present types of atoms, the degree of unsaturation and molecular weight.Structural formulae were encoded by 0 or 1 depending on the presence or absence in the molecule of each of 85 fixed structural elements including functional groups (C:N,H2C=CH, etc.) and chemical classes (steroids, terpenes, pyridines, etc.). Optimi- sation of parameters and the network training using the BPE procedure were carried out on HP 9000 RISC and other work- stations. Performance of the system was studied for different combinations of input data. It was established that performance of the system in the case of using NMR spectra only is higher than that reached using IR spectra only; the inclusion of molecular formula increases per- formance of the system for any combination of IR and NMR spectral data; simultaneous use of both types of spectra leads to a substantial increase in performance of the system (i.e., the number of reliably identified groups increases appreciably).These conclusions were not unexpected; however, the impor- tance of this study 51 is that the procedure for identification of fragments by joint use of IR and 13C NMR spectra, based on ANN application was first proposed. This approach can be realised with ease in a fragment selection block of an ES. Application of ANN to joint interpretation of IR and mass spectra (MS) of gases has been described.63 The network was trained to identify 26 functional groups.The proportion of correctly identified present groups was 86.4% when using com- ME Elyashberg Pte Ate Ptf 16.4 18.0 32.3 18.2 27.2 41.9 62.9 5.5 13.7 15.2 39.0 11.8 10.5 35.8 21.4 52.4 22.1 10.0 48.0 15.6 43.0 60.3 29.8 79.8 5.9 29.7 35.0 37.8 62.8 83.3 49.0 59.1 56.9 68.7 75.8 62.0 6.7 1.0 0.8 1.1 0.7 1.2 0.1 3.1 1.3 1.6 1.7 3.0 2.0 0.8 0.1 0.3 6.1 1.2 1.7 6.6 3.0 1.8 1.3 0.4 16.3 2.9 2.1 10.7 0.2 0.3 1.1 6.0 6.3 0.9 0.3 1.8 88.3 79.8 56.4 78.6 71.5 57.2 35.1 91.8 82.8 77.0 50.7 85.5 87.0 55.5 74.2 47.6 74.0 82.8 47.0 82.7 52.4 36.3 66.0 14.3 86.4 69.9 54.0 53.3 34.9 16.7 47.6 34.8 42.6 30.4 23.0 34.4Expert systems for structure elucidation of organic molecules by spectral methods bined data; this index was respectively 88.4% and 78.2% when IR spectroscopy and mass spectrometry data were used separately.Among 26 groups, only for 8 did the use of MS lead to improving the recognition quality; the results of determination for the C7Cl bond have been considerably improved. This confirms the known fact that IR spectroscopy is more appropriate for identification of functional groups than mass spectrometry. Absent functional groups were predicted to an accuracy of 95.5% using combined data and IR spectroscopy data only and to an accuracy of 87% using mass spectrometry data only.The results of this study 63 show that principles of combining the data obtained by different spectral methods, as applied to ANN, require special investiga- tions to find optimum variants. It should be noted that, as a rule, different groups of researchers use different indices of network performance, there- fore comparison of the results obtained in their studies is difficult. In this connection the problem of the selection of generally accepted indices and standard test sets, which would make it possible to easily compare the efficiency of different classification methods, appears to be topical. Despite improved capability of ANN to identify fragments, successful application of the networks is to a great extent depend- ent on the user's skill in adjusting the network parameters (decision thresholds, composition of the training set, etc.) to meet specific requirements when solving any problem.Currently, there is no information on the use of ANN in any actual ES dedicated to identification of the molecular structure and the number of structural features that can be recognised using an ANN is unclear (for comparison, it should be noted that spectral libraries of known ES contain hundreds of fragments). Therefore an estimate of actual possibilities of this approach is still ahead. However, it can be stated with certainty that application of ANN to the determination of individual classes of chemical compounds is rather efficient.4. Methods for recognition of large fragments To solve the problem of identification of large molecules, an ES must have adequate analytical tools for determination of large fragments. Attempts to create libraries containing large fragments and ranges of variations of their characteristic spectral features are of no promise because of imminent combinatorial explosion and the virtual impossibility to predict the ranges. An actual way of solving the problem is using modern databases containing structural formulae and spectra of substances. The basis of the method is a hypothesis that similar spectra (subspectra) corre- spond to similar structures (substructures). Two approaches to identification of large fragments using IR, NMR and mass spectra databases have been described.The algorithm suggested to this end in the first approach is reduced to three main operations:64 ± 72 (1) search for structures whose spectra are similar to a given experimental spectrum and ordering the former according to decrease in similarity; (2) selection of a number of structures in the beginning of the list that, according to the basic hypothesis, have the highest resemblance to the structure of the analysed compound; and (3) selection (among these structures) of largest common fragments that can be plausible constituents of the structure of the unknown compound. The probability of correct identification of the fragment increases if the database retrieval system selects the data using several types of spectra and the selected fragments are tested on the basis of coincidence of the predicted subspectra (the latter can be most easily done using 13C NMR spectra).Realisation of the above algorithm requires reliable quantita- tive criteria for structural and spectral similarity, as well as methods for revealing the largest common subgraph in a set of chemical graphs. Studies on the development of such methods gave rise to an independent investigation line in graph theory and numerous studies on the subject have been reported. The other method for recognising large fragments found a wide application in practice due to the appearance of large 531 13C NMR spectra databases.In these types of databases the structures are described by `spherical' code 73 while the signals in the spectra are assigned to corresponding carbon atoms. This made it possible (see, e.g., Refs 30, 31, 74) to develop algorithms which can automatically create the subspectrum ± substructure correlation libraries used for solving the problem raised. First, the program compares all subspectra of library fragments with the experimental NMR spectrum of the analysed substance. In this case it is necessary that the signal intensity (i.e., the number of carbon atoms) be also indicated in addition to the chemical shifts and signal multiplicities. As a result, only those substructures (fragments) are selected to which subspectra coinciding (within the limits of prescribed deviations) with the query subspectra correspond.Further, the most probable fragments are selected by a detailed analysis of spectral information and ranked in descend- ing order with respect to the number of carbon atoms and skeletal atoms. Thus, the largest fragments appear in the beginning of the list (the use of the described procedure in ES is considered below). III. Generation of structural formulae of isomers No more than two dozen of publications concerning the search for ways of increasing the software performance and methods of more efficient use of structural constraints have appeared after the first generators of isomers 3, 75 ± 78 were developed. This relatively small number indicates stubbornness of the problem rather than a loss of its actuality or interest.The numbers of possible structural formulae and the CPU time grow exponentially as the number of skeletal atoms and fragments involved in the generation process [they are called discrete units of structure (DUS)] 7 increase. Therefore, all the efforts were aimed at using structural con- straints even in the course of rather than after structure generation and at developing new algorithms that make possible improving the efficiency of generation. This interesting and ambitious problem drew the attention of mathematicians, specialists in the field of graph theory and combinatorial analysis, which resulted in the development of new approaches and appearance of efficient software (COCOA,79 MOLGEN,80 ± 82 and GENM83 ± 86 programs). Par- allel to the search for efficient generation algorithms, computer performance has increased swiftly.On the whole, this has led to appreciable increase in dimensionality of the problems which can be solved using modern generators of isomers. Currently, the number of DUS used in the structure generation that takes a `reasonable' time lies in the range from 10 to 16 depending on the valence of the atoms considered as DUS, the number of free bonds in the fragments, the unsaturation degree of the molecule and several other factors (cf. with 7 ± 8 DUS in the 1970s). For instance, it took 3420 s to generate 4347 isomers of composition C15H32 using (the first in the USSR) program generator written by V V Raznikov in codes of an M-20 computer,75 whereas a modern GENM program installed on a personal computer equipped with a Pentium CPU solves this problem in a matter of seconds.One more example can be given: generation of 217 isomers of composition C6H6 using the MAISS program written in Fortran algorithmic language and installed on a Minsk-32 computer took *30 min;77 the GENM program performs this task for a fraction of a second, which means that the actual rate of structure generation has increased by several orders of magnitude. How- ever, since recognition of the structures of large molecules is of prime interest for practice, the generation stage remains a bottle- neck for all ES despite the advances achieved. Basic requirements for the structure generators have been formulated and realised as long ago as 1970s.The results of these studies were generalised.7, 8 The most important requirement is that, irrespective of generation algorithm used, the software suitable for practical use must contain flexible tools for imposing structural constraints among which the following are typical. 1. Fragments�DUS identified at the SGA stage. It is usually assumed that they have no overlapping atoms and the number of532 free valence bonds of terminal atoms in such fragments is fixed. Therefore such fragments are also called macroatoms and they participate in structure generation along with free atoms. The DUS sets are designed of atoms and macroatoms and the total number of DUS in the set determines the dimensionality of the problem. Some programs, e.g., GENOA87 admit partial overlap of the macroatoms; however, this makes the algorithm more complicated and increases the generation time.2. Obligatory and forbidden fragments (GOODLIST and BADLIST, respectively). These are structural constraints imposed by a chemist and each structure must meet these require- ments. This type of constraint is imposed on the basis of a priori knowledge of the compound under study (the origin of the specimen, data of non-spectral experiments, etc.). These frag- ments have a peculiarity consisting in the possibility for their structures to overlap and superimpose on the DUS structure. 3. Constraints associated with the ring size and bond multi- plicities make it possible to formulate general requirements for generated structures with consideration of specificity of the DUS set and a priori information (the presence of multiple bonds in macroatoms, rings, the inclusion of the unsaturation degree of the molecule, etc.). 4.Fragments which are unlikely in organic chemistry. These include small rings containing triple bonds, structures that contra- dict the Bredt rule, certain strained fused structures with small rings, fragments of the7O7O7O7type, etc. Since `construction of generators' has become an individual division of mathematical chemistry and combinatorial mathe- matics, we will restrict ourselves to mentioning some new programs that are of prime interest for the ES creation.Detailed information concerning modern structure generation algorithms can be found in the theses.88, 89 TheCONGENgenerator developed in 1970s 90 is based on the algorithm generation by building up a chemical graph from its centre. As a result of improving this algorithm, the MOL- GRAPH80 and MOLGEN81 ± 83 programs have been developed. The MOLGEN program generates structures using nonoverlap- ping fragments. It has a high performance and is capable of generating stereoisomers. Currently, this generator has received wide acceptance in Europe. A GENM program 84 ± 86, 88 for generation of structural isomers was developed in the Russian Federation. The parame- ters of the program are at least as good as and in some instances superior to those of the MOLGEN program.The algorithm used has a number of unique features that make it possible to rule out `hopeless' branches of the solution in the course of generation in accordance with the constraints imposed. Operating experience of the GENM generator at the Institute of Organic Chemistry (the Siberian Branch of the Russian Academy of Sciences, Novosi- birsk, Russian Federation) and in the RASTR36 and X-PERT 29 systems has shown its high reliability and fast operation. The SMOG program 89, 91 allows a chemist to analyse differ- ent structural hypotheses in the course of investigations of chemical reactions. This program can generate structures con- taining atoms in unusual oxidation states or carrying formal charges, generate the molecules using both non-overlapping and overlapping fragments and recognise different resonance forms of the same aromatic molecule among the results of generation.The SMOG program can serve as a basis for the development of software tools designed to operate on a special-purpose work- station used in the chemical laboratory; however, the program operates not so fast to be used as a `conveyer-type' structure generator in ES. In some instances where the fragments are elucidated inde- pendently by using different methods of investigation (IR, NMR, MS, etc.) it may appear that the structures of particular fragments have common atomic groups (e.g., 7CH=CH7C=O and C=C7CH3). In this connection the GENOA program 87 ME Elyashberg capable of generating the structures using overlapping fragments was developed to extend the potentialities of the CONGEN program.At the same time, structure generation using overlap- ping atoms belonging to different fragments is the major condition for operation of a number of recently developed programs.30, 31, 74 This type of algorithm was developed to provide structure generation using a set of fragments identified by means of 13C NMR spectra (see Section II.4). The program tries to match the atoms belonging to different fragments and predict the subspectrum of a new, enlarged fragment. If the predicted subspectrum does not contradict the experimental spectrum within the limits of prescribed admissible deviation, an attempt is made to add one more fragment, etc.Thus, structure generation is a successive assembly accompa- nied by verification of the correctness of all possible combinations of the fragments. Structure generation using this approach does not require knowledge of an empirical formula and even the molecular weight. However, it should be compensated for the absence of empirical formula by ascribing a certain number of carbon atoms to each signal in the experimental NMR spectrum (this cannot be done in all cases, especially for symmetric molecules). This algorithm implicitly uses the CnH(m7k) part of the empirical formula, where k is the number of hydrogen atoms bonded to hetero-atoms. Information on the latter is retrieved in the course of structure generation.However, if the set of selected fragments is not exhaustive with respect to the structure of the entire molecule, the generation fails. Application of this type of generator is considered in detail below. Mention may be made of a number of studies (e.g., Refs 92 ± 97) in which different principles of structure generation were evaluated. However, their actual efficiency and performance can be assessed only by solving the test problems on clo-type computers, which seems to be difficult. In some instances, structure recognition is not restricted to determination of the structural formula of the compound, but also involves the establishment of the spatial model of the molecule. To this end, algorithms and programs for generation of stereoisom- ers,83, 98, 99 and building of spatial models of the molecules 99 ± 103 were developed.They use structural formulae of organic com- pounds as the input information. Building spatial structures can appear to be necessary for prediction of the spectra of candidate structures, thus becoming an important stage of elucidation of the true structural formula of the compound analysed. It should be noted that one of the most efficient 3D-generators, CORINA,101 can be accessed via the Internet (the webpage URL is http:// www2.ccc.uni-erlangen.de/servies/3d.html). IV. Prediction of spectra of probable structures Prediction of the spectra of isomers to reveal plausible structures is the final stage of the ES operation.In practice, depending on how hard are structural constraints imposed, the answer file may contain tens, hundreds or even thousands of structural formulae that cannot be distinguished taking into account the ranges of changes in characteristic spectral features. This requires predic- tion of complete molecular spectra. Time-consuming quantum- chemical calculations are inapplicable in this case, so only fast approximate spectra prediction using various empirical correla- tions, rules and parameters, as well as spectral databases can be used. Numerous studies concerning this problem have been reported. Because of its prime importance for the methodology of the development of ES, let us consider the approaches to prediction of IR, 1H and 13C NMR spectra.1. Simulation of 13C NMR spectra 13C NMR spectroscopy is one of the most powerful methods for elucidation of the structure of organic molecules and testing structural hypotheses. Among all types of molecular spectra, it is the 13CNMRspectra that can be simulated most easily since in the case of off-resonance they consist of singlets, doublets, triplets andExpert systems for structure elucidation of organic molecules by spectral methods quartets (1H713C spin ± spin coupling through one bond) and each anisochronal (magnetically equivalent) carbon atom is ascribed a signal with specific chemical shift and multiplicity. Usually, 2D DEPT NMR experiments make possible reliable determination of signal multiplicities even in the case of strong overlap.Prediction of 13C NMR spectra can be performed with different degree of detailisation. a. Prediction of the number of signals with different multiplicity Shelley and Munk104 were the first who suggested use of prediction of the number of signals with different multiplicity for selecting candidate structures. The method is based on the topological analysis of a multigraph on which the structural formula of the molecule is mapped. All skeletal atoms are divided into groups of topologically equivalent atoms. Where all carbon atoms in the structure appear to be unique, the predicted number of signals in the spectrum is equal to the number of carbon atoms in the molecule; in this case the number of different types of multiplets is predicted automatically.If the structure contains several groups of equivalent carbon atoms, the expected number of signals in the spectrum is equal to the number of groups. The authors 104 analysed how stereochemical factors affect the correct- ness of prediction and suggested a number of rules which make it possible to take into account splitting of the signals of dimethyl groups located near a chiral centre. A simple comparison of predicted and experimental numbers of different types of multip- lets allows one to reduce appreciably the number of candidates. However, it should be noted that accidental coincidence of inequivalent signals can occur or unexpected splittings can appear in the spectra. To improve this approach, it was sug- gested 29 that the number of signals with different multiplicity in characteristic spectral regions be predicted, which increased the resolving ability of the method used.b. Prediction of chemical shifts Simulation of 13C NMR spectra is based on the prediction of chemical shifts of carbon atoms in the structure under consider- ation. The model spectrum obtained is compared with the experimental one and the structural hypothesis is accepted or rejected on the basis of calculated mean or standard deviations. There are two most widely used procedures for predicting NMR spectra, based on construction of empirical models 105 ± 110 and on application of spectral databases.111 ± 115 Recently developed software 116 ± 118 uses both approaches.Attempts to simulate 13C NMR spectra using ANN are also documented.119 ± 128 Grant and Paul 129 suggested the first additive linear model for calculating the chemical shifts of carbon atoms in aliphatic hydrocarbons using increments with consideration of the effects of environment up to d-atoms. Later, linear models were con- structed for alkanes, cyclohexanes, alcohols, amines, chlorine- containing compounds, etc. The formulae derived and tables of increments were collected in a number of monographs (see, e.g., Refs 130, 131). The approach described provides a rather high quality of prediction, which stimulated the development of software for analysing the candidate structural formula, selecting a corresponding additive model for each carbon atom and calculating the chemical shift.This type of software is of limited utility since chemical shifts can be predicted only for sp3- hybridised carbon atoms and the increment values cannot be modified. FuÈ rst and Pretsch 110 removed these restrictions. Using large databases containing structural information and assigned 13C NMR spectra, the authors of this study succeeded in con- structing linear models which can be applied to many classes of chemical compounds. The models contain configuration- and conformation-dependent parameters and take into consideration the configuration of C=C bonds and the presence of axial and equatorial substituents in the cyclohexane ring. The software developed allows modification of parameters (reference values and increments) and input of new additive rules.A test of the 533 program on the set of 170 000 known chemical shifts showed that the rules derived are fulfilled for 97% of the cases. The standard deviation of predicted chemical shifts from experimental values was 5.5 ppm; our experience has shown that calculations of the spectrum of medium size molecules (C12±C16) on a modern personal computer take no more than a second. Obviously, this program can be used for filtration of the answer files created by an ES. In another approach,111 ± 115 databases containing structural information and chemical shifts of carbon atoms are used. For all carbon atoms of each reference compound contained in the database, atom-centred fragments (ACF) with a prescribed number of concentric layers are generated.These fragments and the corresponding chemical shifts are stored as an ordered list. To predict the spectrum of a candidate structure, the program selects all possible ACF existing in the structure, performs a search for their analogues in the database and ascribes the chemical shifts taken from the reference fragments to carbon atoms. If an ACF was not found in the database, the program usually performs `interpolation' using the available `most similar' structures. The results obtained using this approach are in very good agreement with experimental ones when using a large database containing a diversified set of structures (see, e.g., Ref. 73). Usually, the difference between predicted and experimental chemical shifts lies within the limits of 1 ppm, which means that the prediction accuracy is rather high.However, despite its indubitable advantages, the approach described has some drawbacks. The absence of stereochemical information in the database (only structural formulae are avail- able) may strongly affect the prediction quality. As has been shown,132 neglect of stereochemical effects may increase the difference between predicted and experimental chemical shifts by more than 10 ppm. Spectral properties of the reference fragments used for extrapolation may appear to be not the closest to those of the fragments of the hypothetical structure. Additionally, the fact that the effect of an atom equally distant from another atom can be different in different compounds is ignored at the encoding stage.Finally, prediction of a spectrum takes a rather long time (tens of seconds and even minutes) depending on complexity of the molecule. A method is described,117 which is a combination of both above-mentioned approaches. At the first stage, the environments of carbon atom in the reference and target (verified) structures are represented as vectors and the Euclidean distance between these vectors is calculated. After repeated application of this procedure, a test set of carbon atoms whose chemical environment is similar to that of carbon atoms in the target molecule is generated. If the degree of similarity of the chemical environment of the target atom appears to be high, the chemical shift characteristic of its `twin' in the reference structure is ascribed to this atom in the target structure.Otherwise a linear model is constructed using the test set. To this end, descriptors describing topological, geometric and electronic properties of the environment are introduced. Models are constructed using multiple linear regression (MLR), partial least squares (PLS) method 28, 133 and principal component analysis.28 The test of the combined approach on the set of 38 500 atoms has shown that the mean deviation is 1.69 ppm, whereas this value was 1.85 ppm in the case of straightforward database retrieval procedure. It was mentioned above that ANN can be used not only for structural interpretation of the spectra, but also for prediction of spectral pattern corresponding to a particular structure.Attempts to apply ANN for prediction of chemical shifts in NMR spectra have been undertaken taking methyl-substituted cyclohexanes,121 mono-substituted benzenes,122 monosaccharides,125 ribonucleo- sides,126 sp2- and sp3-hybridised a-atoms in acyclic alkenes 127, 128 etc as examples. Artificial neural networks are trained to predict spectra of compounds belonging to prescribed classes on a training set which includes structures encoded analogously to the procedure534 described in Section II.3. In the course of training, the network uses encoded reference structures as input information and the output signals are compared with the 13C NMR spectra of these structures.Training is completed if the deviations of predicted spectra from the reference ones take appropriate values. Let us consider the results of computational experiments 128 as an example. The authors trained a network to predict the chemical shifts of sp3-hybridised carbon atoms in the a-position with respect to the double bond in acyclic alkenes. Parallel with the training, a MLR- based model was developed. Testing both models showed that the standard deviation lay within the limits from 1.3 to 1.5 ppm for ANNand was 3 ppm for theMLRmodel, which means thatANN can predict the spectra more correctly than linear models. Artificial neural networks can be applied only to predict the spectra of compounds belonging to those classes on which the network was trained, whereas the structures included in the answer (output) files created by the ES may contain diversified and unexpected atomic fragments.We believe that the optimum strategy of applying prediction of 13C NMR spectra must include two steps, viz., screening of structures (using fast approximate prediction methods, e.g., linear models 110) followed by more accurate (oriented to chemical classes) prediction applied to those structures for which the mean or standard deviation lies within the limits of error of the computational method used and, hence, further refinement is required. 2. Prediction of 1H NMR spectra As was mentioned above, relative simplicity of prediction of 13C NMRspectra is associated with the fact that they consist of clearly defined multiplets. It is a much more complicated problem to predict 1H NMR spectra, because they contain both first-order signals and higher-order multiplets.This is also the reason for difficulties encountered when trying to compare automatically experimental and predicted spectra. That is why information on the methods for prediction of 1H NMR spectra is scarce. The Proton Shift program developed 134 to predict the chemical shifts in 1H NMR spectra on the basis of simple linear models 131 automatically determines the substructures for which certain rules are applicable and considers the rest part of the molecule as a set of substituents at each of the substructures selected.To extend the area of application of the program, different extrapolation procedures were used, including division of large substructures and replacement of missing substituents by related substructures. The program takes into account cis ± trans- isomerism of double bonds as well as axial and equatorial positions of substituents in six-membered rings. Linear models are based on a large set of increments and the total number of parameters that can be used in calculations approaches 3000. A test on the set of 583 chemical shifts showed that the mean deviation of predicted values from the measured ones was 0.06 ppm (the standard deviation was 0.18 ppm). The authors of another study 135 on evaluation of the quality of the Proton Shift system using a large test set found that the standard deviation of chemical shifts was 0.3 ppm, which is closer to actual values.Using the ASSEMBLE structure generator,136 24 308 isomers with 111 605 different proton environments were generated. The chemical shifts were assessed in 89% of the cases, which can be considered satisfactory. The mean time taken to predict the chemical shifts for one structure on a Macintosh Quadra 650 computer was 0.30 s. When comparing the predicted and experimental spectra, the input of the latter is performed manually as a set of chemical shifts of CH, CH2 and CH3 groups. The program compares each experimental chemical shift with the closest value of corresponding chemical shift in the predicted spectrum. Then the standard deviations are calculated to rank the isomers generated.The procedure for comparing the chemical shifts in high-order spectra used in the study cited 135 remains unclear. Most likely, the approach described can be used only for ME Elyashberg preliminary rough selection of structures in the answer file ranked after predicting the 13C NMR spectra. The ACD/HNMR Predictor program 137 is designed to calculate automatically the absorbance curve in the 1H NMR spectrum from the structural formula of the molecule. The program has an internal database containing 180 000 experimen- tal chemical shifts and 25 000 experimental spin ± spin coupling constants (SSCC). The algorithm of quantification of spin ± spin interactions in organic compounds used by the system is based on parameters found for more than 3000 structural fragments.Each dH value in the database is assigned to a particular structural fragment and each JH±H constant is assigned to a pair of interacting structural fragments taking into account the skeleton bonding them. The internal database of assigned dH and JH±H values is based on the results of analysis of 48 000 experimental 1H NMR spectra. The dH and JH±H values for fragments not found in the database are calculated using special-purpose algorithms. Calculations of the spectrum are performed for the basic frequency of the spectrometer, which is considered as a parame- ter. Calculated chemical shifts and constants are provided with 95% confidence intervals to indicate the reliability of the calcu- lated values.The spectrum is calculated with consideration of second-order interactions and long-range SSCC, which allows simulation of the spectra of strongly coupled spin systems containing up to eight magnetically inequivalent nuclei (if there are magnetically equivalent nuclei, then a larger number of interacting spins can be taken into account). The program recognises cis ± trans-isomers of alkenes and syn ± anti-isomers of amides, oximes, hydrazones and nitrosamines. After input of the query structure whose spectrum should be predicted, the program automatically `splits' the molecule into a set of unique fragments that are compared with the fragments contained in the internal database.If any structural fragment coincides with a fragment from the database, its experimental dH values are included into the set of chemical shifts of the structure. If some fragments are not found in the database, the chemical shifts are extrapolated using the parameters of similar fragments. After composing the final lists of chemical shifts and SSCC, the program composes and diagonalises the spin Hamiltonian matrix to calculate the number of lines, their positions and intensities and to assign the dH values to the hydrogen atoms of candidate structure. The program generates the spectrum plot (including digitising of positions of the lines and the integrated curve). Calculations of chemical shifts, spectral contour, inte- grated curve and digitising of maxima take from several seconds to tens of seconds CPU time on a computer with a Pentium type CPU.The ACD/HNMR Predictor program was developed to interpret experimental spectra and test hypotheses. However, where preliminary ranking of the structures in the answer file created by the ES was already done on the basis of 13C NMR spectra prediction, only 5 ± 10 structures placed at the beginning of the final list require additional testing. Hence, comparison of the calculated spectra with the experimental ones to select the most probable structures can be done with ease by a chemist. The author of this review has assured himself that the approach used in this program is highly efficient when analysing the answer (output) files for many problems solved using ES.31 In all the cases considered, it has been possible either to elucidate the true structure or to reduce drastically the number of alternative structures.3. Prediction of IR spectra The nature of IR spectra is such that, in principle, they can be used for selecting candidate structures. This is due to the fact that IR spectrum depends not only on which groups constitute a molecule, but also on their arrangement; in other words, the IR spectrum reflects properties of the entire molecule. Therefore, considerableExpert systems for structure elucidation of organic molecules by spectral methods efforts of many investigators were aimed at developing methods for prediction of IR spectra of organic molecules.Semiempirical calculations of an IR spectrum using a com- puter database of molecular parameters (the geometry, force constants, electrooptical parameters) can be considered as an efficient method for predicting the spectra. A corresponding database as well as computational techniques and software were developed under the supervision of L A Gribov.138 The IR spectrum is calculated in the harmonic valence-force field approx- imation using a semiempirical computational procedure for calculating the band intensities.139 The force constants and electrooptical parameters to be stored in the database are found from quantum-chemical calculations followed by solving mechan- ical and electrooptical inverse spectral problems. Calculated spectra are represented as spectral distributions of the absorption coefficient and can be immediately compared with the experimen- tal curves.Calculations of the IR spectrum of a medium-size molecule take several minutes of CPU time; however, preparation of input data may take from 15 min to several hours depending on the problem complexity. Our experience shows that this approach can be successfully used not only to select preferable isomeric structures, but also to elucidate plausible stereoisomers and conformers.140 Figure 3. Experimental (a) and calculated (b ± d) IR spectra of candidate structures.141 graphical representation of the spectrum of the molecule as superposition of the bands on the intervals corresponding to the fragments found. It is obvious that this approach leads to inevitable loss of many bands of skeletal vibrations in the `fingerprint' region, as well as overtones and combination bands, and that the influence of molecular symmetry on the band intensities in the region of functional groups is ignored.It is also clear that the fragment library is far from exhausting all possible atomic combinations. It should be kept in mind that, as a rule, characteristic spectral features of isomers included in the answer (output) file of ES correspond to experimental spectra. However, structurally similar isomers can be recognised first of all using noncharacteristic IR bands containing encoded structural information and observed in the `fingerprint' region'. In practice, the most likely structure is chosen on the basis of careful comparison of frequencies and intensities of no more than five bands in the experimental and predicted IR spectra (see Fig.3). Therefore the approach sug- gested 142 should be considered unacceptable for the analysis of the answer structural files. An attempt to use ANN for predicting the IR absorbance curves of organic compounds from numerically encoded struc- tural formulae has been undertaken.143 It has been possible to predict satisfactorily absorbance in the regions characteristic of functional groups; however, this approach appeared to be unacceptable for prediction of the `fingerprint' region. It is likely that the failure can be explained by application of too rough a coding procedure, which, in particular, completely ignores the spatial structure of the molecules.An analogous conclusion was also drawn by Weigel and Herges 144 who attempted to simulate IR spectra using an ANN trained on a set of spectra obtained from experiments and semiempirical quantum-chemical calculations. regions in the IR spectra was created within the framework of the DARC system. The expected intensities and band shapes are indicated for each interval. The prediction method includes the search for reference fragments in the candidate structure and In contrast to the above-mentioned failures, a new approach 145, 146 seems to be encouraging. Taking into account Transmission However, the advantages of this method for prediction of IR spectra have their counterpart consisting of at least three draw- backs: the method is not so fast as those used for prediction of NMR spectra and, hence, it cannot be used for `conveyer-type' calculations of dozens of structures (to say nothing of hundreds and thousands of them); prediction of the spectrum cannot be performed automatically without attention of a chemist; and the method can be applied to recognise small- and medium-size molecules only.The first two drawbacks are obvious and their influence will gradually decrease with improving the algorithms and increasing the computer performance. The third one is of crucial importance, since it is due to the nature of molecular vibrations rather than to limitations of the software performance imposed by the size of molecules (e.g., a program 138 makes it possible to operate with the structures containing over 100 atoms).In the general case, the larger the molecule, the more rich its IR spectrum. This statement has a consequence consisting in a nearly uniform distribution of absorption bands of rather large molecules over the `fingerprint' region, which makes it senseless to compare the calculated spectra of candidate structures with experimental ones. Therefore, it is appropriate to use straightforward calculations of IR spectra in those cases where it is impossible to make unambiguous decision using fast prediction methods. Prediction of IR spectra of three isomers with closely related structures, for which standard deviations of predicted 13C NMR spectra appeared to be within the limits of admissible errors of computational procedure, is illustrated in Fig.3. The structures are ordered in accordance with decreasing similarity between the predicted and experimental NMR spectra. Comparison of calcu- lated and experimental IR spectra shows that the calculated spectrum of the structure shown in Fig. 3 b is closer to the experimental spectrum than the calculated spectra of the other structures (see the region from 10 to 14 mm). The same result was obtained using the ACD/HNMR software package 137 developed for prediction of 1H NMR spectra. Thus, three methods for prediction of spectra indicate the priority of the first structure (2,4-dimethylanisole).Impossibility of using semiempirical methods for fast auto- mated prediction of IR spectra has stimulated the search for another approaches. Dubois et al.142 suggested the simplest procedure for simulating the IR absorbance curve. A database containing 700 substructures with fixed expected absorbance abcd 12 10 8 6 535 CH3OCH3 H3C CH3OCH3 H3C OCH3 CH3 H3C CH3CH3 H3CO l /mm536 that IR spectra are strongly dependent on the spatial structure of molecules, the authors attempted to train the ANN on a training set composed of three-dimensional models of organic molecules. Since all models should be described by vectors of equal dimen- sionality, the impossibility of using Cartesian atomic coordinates in this case is obvious.To describe the spatial structure of molecules with different numbers of atoms using the same number of descriptors, it was subjected to a transformation based on the equation used for analysing experimental data on electron diffraction. This made it possible to find a procedure for coding any spatial structure by a vector with 32 or 64 coordinates (the 3D MoRSE code). Using the CORINA program 101 for fast transformation of structural formulae into three-dimensional models, the authors studied the spatial structure ¡À IR spectra correlations using ANN. The structural formulae from the train- ing set and the target structures were first transformed into spatial models and then encoded using the 3D MoRSE code. Experiments show that this methodology makes it possible to simulate adequately the IR spectra of specified classes of com- pounds throughout the absorption region including the region of `fingerprints'.The experimental and predicted IR spectra of a quinoline derivative are shown in Fig. 4 as an example (in this case, theANNused for prediction was trained by the BPE method on a set of more than 50 quinoline and isoquinoline derivatives). The similarity of experimental and calculated spectral curves for this non-additive complex molecule shows that the approach has a chance to become an analytical tool for choosing plausible structures. However, it should be kept in mind that in this case successful results were obtained after training the network on the set of related compounds.D Cl 0.8 N O O 1 2 O 0.40 3000 2000 1000 n /cm71 Figure 4. IR spectra of trisubstituted quinoline: experimental (1) and predicted using ANN (2).146 It is noteworthy that, in principle, this approach also makes possible solving the `inverse' problem, viz., construction of the spatial model of a molecule using its IR spectrum. This was illustrated taking 1-phenylbutan-1-ol 146 and cholesterol 147 mole- cules as examples (in the latter case, the model found from the IR spectrum appeared to be fairly close to the structure obtained from semiempirical quantum-chemical calculations). Yet another method 148 for prediction of IR spectra seems to be promising. On the basis of the search for nearest neighbours of a target molecule in the library containing structures and corre- sponding IR spectra the authors proposed to develop a method for simulation of IR spectra in the region 2250 ¡À 550 cm71.The prediction mechanism is based on a reasonable assumption that similar IR spectra correspond to similar structural features. Thus, it is assumed that it is highly probable for compounds assigned the highest similarity rank to have IR spectra similar to that of the target molecule. The predicted spectrum is composed of the spectra of the neighbours (the nearest neighbours in the structure space) if they meet specific requirements for structural similarity. ME Elyashberg The structures and spectra are considered as such, i.e., no substructure ¡À subspectrum correlations are used.This approach is most likely to retain noncharacteristic skeletal vibrations, overtones and combination bands, as well as the bands that are changed depending on the molecular symmetry (usually, these bands are in correlation tables). This method of nearest neigh- bours has also the advantages that it requires no preliminary classification of reference substances and allows easy elucidation of the target molecules that have no analogues in the reference set, thus indicating impossibility to simulate the spectrum. Implementation of this idea required choice of such proce- dures for coding fragments, structures and spectra that would make it possible to introduce quantitative measures of structural and spectral similarity.Two methods of structure representation used in the study cited 148 were the method of ACF with one atomic layer 149 and a method which employs a modified graph- theoretical descriptor designated as the ES vector.150 The ACF code is a reasonable compromise between the undue increase in the number of fragments and comprehensive descrip- tion of the molecule. In this case, each atom is characterised by its own type (C, O, etc.) and by the types of the atoms and bonds of its nearest neighbours (hydrogen atoms are ignored). The fragment size varies from two atoms and one bond (a terminal group) to five atoms and four bonds for the atoms with four valence bonds. The fragments are ordered in the same way as the coordinates of the n- dimensional vector and each molecule is characterised by its own vector.The presence or absence of a fragment in the molecule is represented as a binary 1 or 0, respectively, and the occurrence number of the given fragment in the structure is indicated. The program selected 180 different fragments in the set of about 9000 compounds for which the IR spectra were known. The structures were coded as vectors with n=180. The molecules consisting of C, N, O, Br and Cl atoms and containing ordinary, double, triple and other bonds were selected. On the average, the structure was described by eight different fragments while the reference molecules were polyfunctional. Using the ACF code, it is possible to exactly describe various functional groups and, at the same time, impossible to distinguish the variants of mutual orientation of functional groups within the same molecule (e.g., substitution types in benzene ring).There- fore, the ES vector 150 was used as an additional descriptor to simulate the structure in more detail. It is less sensitive to changes in the atomic groups than the ACFcode; however, both codes well complement each other. The ES vector code is based on the counting statistics of the number of routes in the coloured molecular graph mapped onto the coded compound. It provides a way for performing a detailed (though highly redundant) description; therefore the principal component analysis 28 of encoded data was applied in the next stage. To characterise each molecule, the contributions from the first eight principal compo- nents responsible for 99% of changes of complete code were used.Experiments have shown that the structure encoding system 148 is rather well suited in this case. Various procedures for estimating the degree of structural similarity have been considered.151 In particular, similarity of structures X and Y characterised by vectors X={x1, ... xi, ... , xn} and Y={y1, ..., yi, ..., yn} can be described using the Tanimoto coefficient xiyi i . ST �� xiyi y2i ¡¦ x2i �¢ i i i X XX X The X and Y vectors can be composed of zeros and unities representing the presence or absence of the fragment, respectively, or of integers representing the occurrence number of the fragment in the structure.As applied to IR spectra, the 0/1 representation is more informative than the integer representation, since the absence of any fragment often s the entire IR spectrum, whereas the difference in the number of the same structuralExpert systems for structure elucidation of organic molecules by spectral methods fragments usually results in a gradual change in the spectral pattern. Therefore, the structures to be sorted are selected using 0,1 vectors and ranked in accordance with the similarity criterion calculated on the basis of the occurrence number. Ranking of molecules using ES vectors is based on calculations of Euclidean distances from the vector of the target molecule. Experiments were carried out using the IR spectra recorded in the region 2250 ± 550 cm71 with a resolution of 2 cm71 and represented as sets of 128 points separated by 13.5 cm71 (this corresponded to the same spectral resolution); at each of the 128 points, the value was averaged over seven points in the exper- imental spectrum covering the 13.5 cm71 interval.Spectral intensities were normalised so that the absorbance of the most intense band was taken as 1.0. Reliable estimate of similarity in the IR spectra is by no means easy problem. The spectral pattern seen by a spectroscopist's eye is hard to digitise. To solve this problem, the Pearson rank correla- tion coefficient and Spearmen rank correlation coefficient (the strongest, medium intensity and weakest peaks are respectively characterised by the highest, mean and lowest ranks) were used.148 The spectrum prediction algorithm was as follows.1. Search for the nearest neighbours in the structure space. Compare the target structure with all reference structures in the collection and select k nearest neighbours for each of three models of structural similarity (the Tanimoto coefficients for both the 0,1 and integer vectors and the Euclidean distances for ES vectors). Get three lists of coincidences, containing information on k structures each. 2. Rank the selected molecules in accordance with the three measures of similarity. Search for the same structures in the three lists and sum up the individual ranks. If some molecules were found in one or two lists only, then ascribe rank k+1 to these molecules in those lists in which they were not found.Sort molecules in ascending order of complete ranks and generate the final list. 3. Select those molecules whose characteristics either exceed the structural similarity threshold or are below the distance threshold. 4. If the final list contains information on less than two molecules, then terminate the procedure and display the results. Go to step 6. 5. Check the IR spectra of selected molecules for spectral homogeneity using correlation coefficients. If more than 50% of the spectra (e.g., 2 out of 2, 2 out of 3, 3 out of 4, etc.) are much alike (i.e., the correlation coefficient value exceeds a certain threshold value), then use the averaged spectrum for prediction.If the checking for homogeneity fails, then return control to the chemist. Go to step 6. In any case, display all available informa- tion.6. Reject automated prediction since no or only one molecule is available. Prompt the chemist to assess the situation and decide between continuation and termination of the program. If a decision to predict the spectrum is made, then prompt the chemist to specify the molecules that were ascribed the highest rank and whose spectra should be used. Thus, prediction can be performed either automatically or manually. If the chemist makes final decision, the result of prediction is characterised by higher quality. The CPU time is dependent on the number of fragments in the structure and takes several seconds on a personal computer with a Pentium CPU operated at 75 MHz, which is well suited.Unfortunately, the authors restricted themselves to demon- stration of a unique example (Fig. 5), which shows that the predicted and experimental spectra are similar. However, no attempts were made to test the method by sorting actual answer files of the ES, which makes impossible to obtain a statistically justified estimate. In our opinion, this approach has the advantage that its algorithm is free from the `black box' type elements; therefore, the spectroscopist can easily establish the reasons for change in the prediction quality. 537 D 1 2 2150 1950 1750 1550 1350 1150 950 750 n /cm71 Figure 5. IR spectra: experimental (1) and predicted on the basis of search for the nearest neighbours of the target molecule (2).V. Pathways of development of expert systems Among the expert systems created in 1970 ± 1980s, the best known is the DENDRAL system8 referred to as a `classical' example of ES in numerous publications on chemistry and books on AI. Popularity of the DENDRAL system is, on the one hand, to a great extent associated with the fact that this project was worked out under supervision of the Nobel Prize winner J Lederberg and noted scientists in the field of AI. On the other hand, the authors of the project have reported about 40 articles collectively called `Application of artificial intelligence in chemistry', which could not but drawn the attention of many specialists. However, Gray 152 critically analysed the idea behind the DENDRAL system and concluded that this artificial intelligence system is far from being perfect.This problem was hotly debated 153, 154 with participation of the directors of the project. In contrast to a wide- spread opinion that the CONGEN program has found extensive application, Gray stated that it used rarely (he knows 8 articles in which application of the system was reported). One of the reasons for not-too-high popularity of the system is probably that it is not user-friendly. In this review the emphasis is placed on the modern ES appeared in the 1990s. The potentialities of modern ES and those belonging to earlier generations were compared.155 In the last decade, the efforts of researchers were aimed to improve the methods for retrieval of structural information from the spectra, to search for optimum strategies of computer-assisted structure recognition and to increase the user-friendliness of the systems.Our experience has shown that non-user-friendly ES have no chances to become an analytical tool for the chemist. A user- friendly system must: (1) have convenient and operationally simple tools for information interchange with the user, which allows the chemist control over any step of the solution of the problem; (2) offer the chemist the possibility of getting explanation of each decision made by the program; (3) inform the chemist in case of `emergency' at any solution step and display messages adequate to the situation; (4) have advanced graphical tools for representing spectral and structural information to allow a chemist to quickly assess current status of the program operation and make corresponding decision.In this review we consider certain ES, which, in our opinion, reflect modern tendencies and new approaches. However, let us first define basic concepts to be used later. Let the empirical formula of an unknown compound be determined and the number of isomers corresponding to the empirical formula be N. In principle, the N value can be found by structure generation using appropriate software. Wesay that a set of n structures (14n4N)538 selected after the ES operation is a solution of the problem. The solution is correct if it contains a proper structure, otherwise it is incorrect.A correct solution and proper structure can be found only by checking each structure from the answer file using prediction of its spectra and additional experiments. If the answer file is empty (n=0), the problem has no solution consistent with experimental data, information stored in the knowledge base and system options. A description of the RASTR expert system created to elucidate the molecular structure using IR and NMR spectra is available (see, e.g., Ref. 22). This system is capable not only of revealing possible structures, but also of generating, constructing and depicting spatial models of all stereoisomers corresponding to a given structural formula. The RASTR system was used as a prototype and basis for a user-friendly X-PERT system 29 which operates under MS WINDOWS.Let us consider the most important elements of this ES, which in one form or another are incorporated into several other systems. 1. The X-PERT system The X-PERT system was created to elucidate the structure of organic molecules containing up to 30 ± 35 skeletal atoms (C, N, O, P, S, B, Si, Hal) using IR, 1H and 13C NMR spectra. The knowledge base of the system comprises a set of libraries of molecular fragments and ranges of changes in their characteristic spectral features. The libraries are composed taking into account that the fragments belong to different classes of chemical com- pounds and respond specifically toward different types of spec- troscopic techniques. Libraries belonging to different categories play different roles. The first category comprises universal libraries used for both fragment selection and structure filtra- tion.Libraries belonging to the second category are adjusted to different types of spectroscopic techniques and are mainly used as filters for structure verification. Each fragment in every library is associated with its characteristic features in IR andNMRspectra, which are also stored in the knowledge base. There are four hierarchically organised universal libraries. 1. Library of principal functional groups (LPF), e.g., C=O, C:N, etc. The widest ranges of changes in characteristic spectral features are used for these groups irrespective of the nearest environment of a fragment.These libraries are mainly used for preliminary evaluation of a specimen to assign it to a particular class of chemical compounds and for recognition of the structure of non-additive compounds. 2. Library for structural-group analysis contains functional groups and information on the admissible nearest environment, 7CO C C 7O7C=O, 7CH27OH, etc. The , e.g., C library is used for SGA and structure filtration. 3. Library of aromatic fragments (AROM) contains 12 types of benzene ring substitution. 4. Library of phosphorus-containing fragments. There are the following specialised libraries: Alibrary of 13CNMRdata; three libraries of 1HNMRdata to reveal CH3, CH2 andCH groups with consideration of the nearest environment; a library containing fragments (mostly cyclic) characterised by strong specific absorption bands (e.g., four- and five-membered cyclic ketones, lactones, lactams, etc.); and a library of fragments containing an OH group.For each fragment, the system generates a `visit card' containing the following information: the fragment structure; ranges of changes in the characteristic features in IR (frequen- cies, halfwidths and intensities of absorption bands) and NMR spectra (chemical shifts, SSCC, signal multiplicities and corre- sponding numbers ofCorHatoms); structural descriptors used to set the requirements for the fragment environment (additivity conditions) and provide retention of its spectral features on going from one molecule to another.Descriptors characterise the frag- ME Elyashberg ment atoms with free valence bonds. The descriptor indicates the possibility of adding a hydrogen atom, hetero-atoms (they can be enumerated), aromatic cycles and multiple bonds to a given atom. The possibility (or impossibility) of the formation of a multiple bond and incorporation of the fragment into cyclic structures of specified size is also indicated. The descriptor can have three types of meanings: `possible' (P), `forbidden' (F) and `obligatory' (O). The program finds classes of topologically equivalent atoms with free bonds in the fragment structure and colours them using individual colours. Each coloured atom is associated with a column of corresponding colour in the table of descriptors.This makes it possible easily to form, interpret and correct the values of descriptors characterising equivalent atoms. The total number of fragments in the libraries is 450.Experimental data (the empirical formula or molecular weight and available spectra) can be input manually or imported using the OPUS and WIN-NMR programs. The system checks the data for consistency and prompts the chemist to double-check the reliability of the data in questionable situations. If the empirical formula is unknown or questionable, the program generates all possible molecular formulae corresponding to the specified molecular weight taking into account the constraints imposed on the qualitative and quantitative composition of the substance.In this case the most probable molecular formula is chosen using a recently developed method.156 { From this time on, the system operates in the interactive mode. This makes it possible to use the experience and intuition of the chemist and the capability of computer to perform fast search for necessary information and complex logic-combinatorial calcula- tions. In this case, the basic requirement for ES is also met: at any step in the solution the program can explain the reasons for one or another decision and the user can be prompted to activate a corresponding option. Prior to starting the operation, the program assesses complex- ity of the problem and displays messages concerning the method of its solution. If the problem can be solved automatically, the program generates all structural isomers and selects possible structures by passing them through a system of filters.The first stage of the standard method for solution is generation of fragment sets. First, possible fragments are selected with consideration for their empirical formulae, spectral intervals and the unsaturation degree. In this case, necessary libraries of fragments are chosen by the user. The fragments are selected taking into account that the reliability of finding a fragment is 100% if the experimental band falls into the middle of the characteristic interval. It (reliability) linearly decreases as the interval bounds are approached and decreases to zero on both sides of the bounds.A trapezoidal membership function 19 of the interval is chosen so that the decrease in the reliability near the bounds with increasing the length of the interval be more gradual. The fragments selected and their reliability indices are dis- played. The `Show Spectra' procedure makes it possible to represent experimental spectra and characteristic intervals for selected fragments in clear graphical form to allow the chemist to assess the reliability of the fragments and to edit the list by adding (if needed) his own fragments. Then the table with interpretation data of all available experimental spectra is displayed. The chemist can analyse logical relations between the fragments and correct the `reasoning' of the program using his experience and a priori information.The final set of consistent logical equations is used as input information for the `inference machine'. The result of the solution of the logical equations [function f(A), see Section II.1] is displayed as a set or several sets of fragments consistent with the spectra, empirical formula and the unsaturation degree of the compound analysed. The program automatically adds sets of { See also a recent study by M E Elyashberg, Yu Z Karasev and E R Martirosian Anal. Chim. Acta 388 353 (1999).Expert systems for structure elucidation of organic molecules by spectral methods microfragments such as CH3, CH2, etc., which were not included as constituents of the structures of selected fragments. The chemist can also correct the selected fragment sets or compose his own sets.The special-purpose `Multiple Fragments' procedure allows for the inclusion of a fragment appearing repeatedly in the structure, which reduces dimensionality of the problem. Structure generation using fragment sets is the next step of the program operation. In this stage, the possibility of imposing structural constraints listed in Section III (GOODLIST, BAD- LIST, constraints associated with the ring size, bond multiplic- ities, etc.) is provided. There is an additional option which enables or cancels the descriptor control in the course of structure generation. The descriptor control is useful to preclude genera- tion of a large number of wrong (bad) structures. Cancellation of the descriptor control can appear to be useful in the case of repeated solution of the problem if the structures found are questionable.The structure generation is carried out by the program based on the known algorithms.84, 85 The structure counters allow the chemist to monitor the process and assess the efficiency of the constraints imposed. Structure generation using fragment sets and free atoms can lead to such substructures whose presence contradicts the spectra and/or concepts of organic chemistry, therefore all generated structures are verified using the following six filters (see Section IV).1. Checking for consistency with characteristic spectral features. 2. Rough simulation of 13CNMRspectra (see Section IV.1.a). 3.Checking for consistency with principles of structural chemistry and stereochemistry (unlikely fragments, Bredt's rule). 4. Checking each isomer for its appropriateness to be used in interpreting experimental spectra. 5. Search for and exclusion of identical structures. 6. Prediction of chemical shifts in 13C NMR spectra to reveal the most probable structures. The system chooses necessary libraries on the basis of generalised `portrait' of the answer file to perform a two-stage check. In the first stage, the list of fragments that contradict the spectra is generated. In the second stage, only contradictory fragments are sifted using structural formulae to save time and reject those structures in which at least one of these fragments is found.Options used in the filtration stage allow the user to specify the parameters determining hardness of constraints imposed (i.e., the degree of fuzziness of the interval bounds). Structure verification using the first and second filters can be performed in both the standard and supervised modes. The first of them provides automated filtration (the so-called `blind' filtra- tion). In this case, only messages on the current status of program operation and the running number of `good' and `bad' structures are displayed. In the second mode, the filtration process is visualised and the user can control over it as well as get all information necessary to explain the decisions made by the system. Each structural formula to be checked is displayed.If no contradictory fragments are found in the structure (it should be kept in mind that only these are sifted), the structure is indicated as good. Otherwise contradictory fragments found in the structural formula are coloured and information on characteristic spectral features (frequencies, chemical shifts, parameters of bands or signals) not confirmed by the spectra is displayed in the form of a `visit card' of the fragment. Depending on the number and type of contradictory frag- ments found in the structure, the user decides to reject the structure as unlikely or accept it for further checking. Structure filtration using this user-friendly mode of the program operation is visualised, fully controllable and understandable. Checking the number of signals with different multiplicity in 13C NMR spectra in the supervised mode is accompanied by displaying the data on the differences between the number of predicted and observed signals. 539 After passing the data through the first five filters the system automatically uses the answer file as input data for the SpekEdit program which predicts the chemical shifts in 13C NMR spectra.Then the standard deviations (S) of calculated spectra from experimental ones are calculated and the structures are ranked in ascending order of S. The potentialities of the system will be assessed below; here we consider only one example to illustrate the operation of the program.29 Let an unknown compound have a molecular weight of 295 and possible composition C1 ±50 H0 ± 120 N0±30 O0±30; 197 differ- ent molecular formulae correspond to this set of parameters.Using IR and NMR spectral data, the program found the only molecular formula C17H13N1O4 and created the fragment sets (1) C=O, C6H5, 1,2-C6H4 and (2) C=O, C6H5, 1,3-C6H4. These sets were used to generate 6995 structures. The proper structure ON O O HO was elucidated using spectral and structural filtration. 2. The SpecSolv system The SpecSolv system { is a new module of the well known SpecInfo system 157 designed for storage and structural interpretation of NMR, IR and mass spectra. In particular, the SpecInfo knowl- edge base consists of over 200 000 assigned 13C NMR spectra which serve as a basis for the development of the program.The SpecSolv system is designed to elucidate the molecular structure using the 13C NMR spectrum only and does not require the knowledge of the molecular weight and molecular formula. It uses parameters of 13C NMR spectra [the chemical shifts, signal multiplicities and intensities (the latter as the number of carbon atoms)] as input data. The knowledge base of the system is a library of fragments and their subspectra. The library was created by automated `cutting off' fragments from the structures stored in the SpecInfo database and contains more than 400 000 fragments encoded by the three-sphere HOSE code 73 and more than 100 000 fragments encoded by the two-sphere HOSE code with respect to the central carbon atom.The library is carefully checked by experts. Information on a fragment includes the connectivity matrix, HOSE codes of all heavy atoms and the subspectrum. Parameters of 13C NMR spectrum (chemical shift, multiplicity, intensity and root-mean-square deviation R of the chemical shift) are stored for each carbon atom. The R values were calculated for the sets of equivalent substructures in the SpecInfo database. All fragments are completely defined including ring closures in the outer sphere. The first step of structure elucidation with the program is the fragment search. All subspectra in the library of fragments are compared with the experimental spectrum and the fragments whose subspectra match corresponding regions in the experimen- tal spectrum are selected.A typical case is selection of *500 fragments for a molecule containing*25 heavy atoms. Fragment search includes three stages: preliminary (pre- search), main and fine search. After the first stage the subspectra are coded as strings encoding the presence of signals with a certain multiplicity within discrete regions of the 13C NMR spectrum. At the second stage it is tested as to whether the number of signals in the regions selected after presearch has a matching number of lines in the experimental spectrum. The fine search includes checking for exact deviations of the chemical shifts in both spectra. { The first article contained the description of the SpecSolv system 30 appeared in 1996 and had an intriguing title `Fully Automated Structure Elucidation �A Spectroscopist's Dream Comes True'.540 Typically, the deviation is 2 to 5 ppm.The selected fragments are then ranked in descending order of match factor, which is defined as the mean deviation of the chemical shift of each atom from the signals in the experimental spectrum. Despite the fact that the program can operate without knowl- edge of the molecular formula, it is necessary to specify possible types of atoms (especially, hetero-atoms) since they cannot be immediately revealed by 13C NMR spectroscopy. It is essential that there is no need for a chemist to analyse all selected fragments, since bad fragments should be excluded at the stage of structure generation. However, if a priori information is available (most often, such is indeed the case), it can be expected that analysing the list will make it possible to exclude the bad fragments to save the working time.The main feature of the generation algorithm used in the SpecSolv system is that the fragments are linked by overlapping common atoms. The largest fragment whose subspectrum best matches a particular region of the experimental spectrum is taken as the first one and the program tries to add successively any other fragment to this fragment. If the new fragment passes the plausibility test, it is accepted and the enlarged fragment is considered as starting structure. Before adding the next fragment, the program pricts the 13CNMRspectrum of the enlarged fragment. This is achieved by simple combination of the subspectra, which provides a high performance of prediction. Intermediate structures for which the deviations of the chemical shifts exceed a prescribed value are excluded.In this fashion the final structure is being assembled at the same time with cross-validation of intermediate structures. The absence of such a verification tool in structure generators that admit assembly of structures from overlapping fragments (e.g., GENOA87) is a grave drawback. The assembly continues until all chemical shifts in the experimental spectrum are assigned and free bonds in the result- ing structure are absent. Usually, 10 generation cycles based on the first 10 best (the largest with minimum deviation of the chemical shifts) fragments are sufficient to elucidate the answer structure.This is possible since the selected fragments are ranked by match factor of their subspectra with the experimental spectrum. Most often, only one structure is found after exhaus- tive search in the entire fragment space, though several isomers of similar structure can also be obtained (in general, if no empirical formula is given, generation of nonisomeric structures is also possible). Will, Fachinger and Richert 30 reported recognition of rather large molecules, e.g., such as 1.O HO OH O OH HO O HO 1 However, it is obvious that structure generation is possible only if the fragment library contains large overlapping fragments of such molecules. Hence successful identification requires the presence of a rather large number of large molecules related to the molecule to be recognised in the system database.However, even if the database contains appropriate fragments, the absence of com- mon atoms will make generation of the proper structure impos- sible, which means a strong dependence of the result on the diversity and quality of the database. The program is also very `sensitive' to the solvent in which the experimental spectrum is recorded and, in addition (this is a grave drawback) does not recognise stereoisomeric forms. These factors can have a deter- mining effect on the results of comparison of predicted subspectra of intermediate substructures and the overall model spectrum with the experimental one.It is reported that the system elucidated the ME Elyashberg structure of 80% of all compounds analysed in the laboratory. The system was adapted for exploitation on a VAX 6610 minicomputer. As can be seen, at present `a spectroscopist's dream' does not come true with the aid of the SpecSolv system since 20% of all problems remain unsolved. If the system cannot solve the prob- lem, a corresponding message is generated and the program is terminated. Therefore new approaches are required which make it possible to resume the analysis in those cases where the program reaches a deadlock. A procedure for overcoming a deadlock has been implemented in the ACD/Structure Elucidator system.31 3. The ACD/Structure Elucidator system The ACD/Structure Elucidator system integrates the recognition strategy developed by the authors of the SpecSolv system 30 and conventional methodology used in the RASTR,22 X-PERT,29 CHEMICS37 and other systems.Two structure elucidation modes, a standard mode and the classical ES mode are permitted. The first mode is to a great extent based on the known general strategy.30 As in the SpecSolv system, the set of experimental data includes a 13C NMR spectrum, chemical shifts, signal multiplicities and intensities. If it is difficult to indicate the exact number of carbon atoms assigned to a signal, their minimum and maximum values are specified. To facilitate measurements of signal intensities, it is appropriate to record a 13C NMR spectrum with an admixture of a paramagnetic reagent, e.g., Cr(acac)2, which shortens and equalises the spin-lattice relaxation time T1 for all carbon atoms including quaternary ones.The experimental spectrum can be imported directly from the spectrometer or the user can input spectral parameters manually. Usually, only an NMR spectrum is required for structure elucidation; however, it is helpful to have IR and 1H NMR spectral data to facilitate the solution of the problem. The database of the system is a library consisting of more than 400 000 fragments and their 13C NMR subspectra that were derived from 135 000 structures with assigned 13CNMR spectra (the structural database). This ES has an appreciably smaller knowledge base than the SpecSolv system has.The operation of the program begins with the search for 13C NMRspectrum in the knowledge base. Once a spectrum is found, a corresponding structure is displayed and the program is completed. Otherwise the structure is reconstructed from frag- ments. At first, the program selects fragments whose subspectra do not contradict the experimental spectrum. The fragments found are sorted in descending order of the number of skeletal atoms and mean deviations of their subspectra from corresponding signals in the experimental spectrum. Since it is impossible to accept or reject the presence of, e.g., OH,NHand C:Ngroups or to differentiate between dC of the O=C7N group and aromatic ring signal (150 ± 160 ppm) using the 13C NMR spectrum, then, if the IR spectrum is available, filtration of selected substructures is performed using a library of functional groups having the most characteristic features in the IR spectra.Analogous filtration can be performed using fragments having clearly defined character- istic features in the 1H NMR spectra. After filtration the number of selected fragments is reduced (in some instances, it becomes halved), which favours placing good fragments at the beginning of the list. Then, beginning with the first fragment, the fragments are checked for the possibility of linking by common atoms to perform successive assembly of the structure. As in the SpecSolv system, linking the fragments is accompanied by prediction of the 13C NMR subspectrum of the enlarged fragment to verify the correctness of the assembly.The problem is considered solved if after checking of all possible combinations of the fragments the program generates one or several structures. The most promising fragments (i.e., the largest ones whose subspectra match well the experimental spectrum) are placed at the beginning of the list; therefore the answer structure is often generated within minutes after startingExpert systems for structure elucidation of organic molecules by spectral methods the operation of the program. In the case of good correspondence between the 13C NMR spectrum predicted for the answer structure and the experimental spectrum the process can be terminated before completion of exhaustive search for all combi- nations of the fragments.If the attempt to assemble the structure fails (the knowledge base contains no appropriate fragments having common atoms or good fragments were not selected), the program automatically switches on the classical ES mode to operate with DUS. As was mentioned in Section III, operation in this mode requires knowledge of the molecular formula or at least the molecular weight. If the molecular formula is unknown, the program first selects functional groups that would be present in the molecule analysed. To this end, a library containing a small number of the most characteristic functional groups and their spectral features is used. Then search for selected functional groups in the structure of previously selected fragments is performed.The elemental composition and unsaturation degrees of the verified functional groups, as well as the numbers of carbon and hydrogen atoms counted fromNMRspectral data are used by the program for automated determination of structural con- straints for the generator of empirical formulae. One or several possible empirical formulae established by this generator are passed to the structure generator for verification. This structure generator operates with the atoms and frag- ments having no common atoms. It is based on the known algorithms.84 ± 86 The fragment sets are composed as follows. The first N (reference) fragments in the primary list of selected and filtered fragments (usually, the chemist varies the N value from 10 to 30 on his own) are chosen.Obviously, these are the largest fragments and this set almost always includes those fragments that are constituents of the structural formula of analysed compound. Then, the fact is used that, according to the ranking algorithm, small fragments (most often functional groups with their nearest environment) are placed at the end of the primary list. The subspectrum of the first reference fragment is compared with the experimental spectrum to select the unmatched chemical shifts in the latter. Then, small fragments appropriate for interpreting these chemical shifts are chosen. Only those combi- nations of small fragments are added to the first reference fragment that provide the highest degree of interpretation of the experimental spectrum.If thereafter certain signals in the query 13C NMR spectrum remain uninterpreted, the program adds corresponding microfragments (CH3, CH2 , CH, C=O, etc.) to the set. Each of the sets obtained is checked for consistency with the current empirical formula of the molecule. The sets corre- sponding to the second, third, etc. reference fragment are generated analogously. If no small fragments appropriate for composing the sets were found, the latter are formed using microfragments so that the number of sets becomes equal N. The chemist can add his own fragments to the sets, compose the sets based on one of the reference fragments and use all types of constraints described in Section III.To facilitate this, the program displays a `generalised portrait' of the molecule under study, i.e., the distribution of typical functional groups over frequency of meeting them in the selected fragments. High frequency of meeting indicates that the group is, probably, a constituent of the molecule, whereas the absent groups are used to generate the BADLIST. Before starting the structure generation using the current fragment set the program assesses the working time (to this end, the percentage of the job done in a time set using system options, e.g., 20 s, is measured) and, if it appears to be longer than a reasonable time, this set is skipped during the first program run and the program begins testing the next fragment set. All structures resulting from the structure generation process are checked for consistency with the available spectra and principles of organic chemistry and stereochemistry.A special program developed to search for tautomeric structures excludes unstable 541 forms from the list, which often substantially reduces the size of the answer file. Then the structures automatically arrive at the input of the program for prediction of 13C NMR spectra (the ACD/CNMR Predictor program 118). When prediction is completed, the struc- tures are ranked in ascending order of mean deviations of the predicted spectra from the experimental spectrum. Our experience shows that calculations of the spectrum using a personal computer equipped with a Pentium CPU operating at 133 MHz take from 0.5 to 1 s, so prediction of the spectra of 1000 structures takes only 5 to 15 min depending on the number of carbon atoms.Testing of the system showed that in the majority of cases the proper structure is ranked 1st or 2nd from the beginning of the answer file or is at least placed among the first dozen of the structures. However, if the difference between the chemical shift deviations of the preferred structure and the succeeding structure lies within the accuracy of the calculation of the 13C NMR spectrum, further verification of the solution is required. To this end, automated prediction of 1H NMR spectra is performed using the ACD/HNMR Predictor program 137 (see Section IV.2). As a rule, only the spectra of the first dozen of structures are predicted, so the CPU working time is relatively short (*10 ± 60 s per structure depending on the complexity of the spin systems calculated).Experimental and calculated 1H NMR spectra are compared by a chemist who makes the 3 3 final choice of the preferred structure or makes a decision on the necessity of additional experiments. The calculated spectra, integral curve and dH values for all lines are displayed. If the predicted spectra of several structures appear to be equally close to the experimental spectrum, the proper choice may be made by using the DCH values for the dCH singlets DCH3=|dCH3 exp7dCH3 calc| and comparing the integrals for all multiplets. In most cases, at DCH350.2 ± 0.3 ppm the structure can be rejected.Resolving of questionable situations requires additional experiments, in partic- ular, using 2D NMR spectroscopy. Thus, the ACD/Structure Elucidator system differs from other systems in the possibility of using two interacting strategies for solving the problem, each of them being based on correspond- ing structure generator. If in both operating modes the program cannot automatically find an adequate solution, the possibility is provided of solving the problem in the step-by-step mode under the supervision of a chemist who has considerable opportunities to interfere in the `reasoning' of the program. Testing of the system on a test set of 220 problems showed that correct solution can be found in 90% of cases. 4.2D NMR spectroscopy in expert systems The systems created have progressively been developing. In addition to improving the algorithms and extending the knowl- edge bases, attempts are made to use 2DNMRspectroscopic data in ES, since methods employed in this area of NMR spectroscopy have received wide acceptance and, actually, became routine. For instance, instrumental tools for using 2D NMR spectro- scopy data were incorporated into the CHEMICS system to extend its capabilities.158, 159 Methods of processing the data obtained in 2D-INADEQUATE (C7C correlations through one bond) and 1H713C NMR COSY (C7H correlations through one bond)/1H71H NMR COSY (H7H correlations through three bonds) spectroscopic experiments were developed. Using the data of 2D NMR experiments, it is possible to reveal CHn groups (n=0 ± 3) bonded by chemical bonds and to correlate this additional information with the library fragments selected by the system, thus testing the possibility of the formation of a bond between the fragments.The data obtained are used for controlling the process of structure generation from the fragments. A new SESAMI system 161 was created as a result of improv- ing the CASE system.160 The library of 13C NMR spectra containing 5100 ACF with one-layer environment is used as the542 database in this system. First, SESAMI generates an exhaustive list of fragments, which then is made more `sparse' using a special- purpose subprogram. Another subroutine adds extra constraints imposed on the structures.The reduced list of fragments and the constraints are used as input data by the COCOA structure generator.79 Preliminary selection of the structures generated is performed on the basis of prediction of their 13C NMR spectra and an expert makes the final decision. Enhancement of potentialities of the SESAMI system is attained by incorporating the software 162 developed for process- ing the data of 2D NMR experiments. The system receives information on correlations between signals and builds frag- ments consistent with these data. The structures are generated using theCOCOAprogram. Further improvement of the system is performed to develop a module based on application ofANN and designed to elucidate the fragments in the case of simultaneous use of IR and 13C NMR spectra.In parallel with improving the known ES by incorporating tools for processing 2D NMR data, investigations on creation of computer systems designed to retrieve and efficiently use struc- tural information `hidden' in 2D NMR spectra for molecular structure generation began in the last decade. The demand for the development of such systems is caused by the necessity of establishing the structures of very large molecules of natural products containing up to 50 ± 60 skeletal atoms. Of interest is a recently developed approach.32, 163, 164 The authors started from a rather logical position that if once enormous efforts were made to isolate a natural product, carrying out a number of rather labour- consuming 2D NMR experiments seems to be entirely justified.Fragments designed immediately in the course of experiments are mostly used for structure elucidation. As a result, a CISOC-SES expert system was created whose operation requires the data of 1D (1H and 13C) and 2D DEPT, DQF-COSY, HMQC, HMBC, HETCOR, COLOC and 2D-INADEQUATE NMR experi- ments. In this approach, structural data obtained from 2D NMR experiments are mostly used (the structure is reconstructed as if by applying `the first principles'), therefore the authors called this approach `ab initio'. Obtaining information from 2DNMRexperiments requires a skilful choice of procedures (pulse sequences) and correct use of data. Hence the efficiency of such systems will be to a great extent determined by the success in implementing the man ± computer `symbiosis'.Since the systems based on fragment selection retrieve valuable structural information, one should expect that both approaches will eventually integrate. VI. Strategy of molecular structure elucidation Since the requirements for the ES architecture are, on the whole, already formulated, the development of an optimum strategy of the ES application is topical. This problem can hardly be solved O O N O O O O O OH N O O 3 (n=2, np=1) 2 (n=5, np=4) O Cl O NH O O Cl N OH O 8 (n=4, np=2) 7 (n=4, np=2) 6 (n=6, np=5) ME Elyashberg theoretically because it is impossible to imagine the whole versatility of situations that may arise in the course of structure elucidation.Efficient approaches can be found only after analysis of the solution protocols of a rather large number of problems. An attempt to develop the methodology of the molecular structure elucidation using ES was made.141 Particular emphasis was placed on the study of difficulties and contradictions that arise when solving the problems and on the development of procedures for overcoming them. The authors carried out a large-scale computational experiment using the X-PERT sys- tem.29 About 200 structural problems taken mostly from mono- graphs 165 ± 168 or spectral databases and analytical practice were solved. The molecules analysed contained from 10 to 23 skeletal atoms, which is characteristic of the compounds commonly used in industrial laboratories.Most of empirical formulae contained hetero-atoms (from 1 to 10). Over 80% of the total number of compounds belonged to the cyclic ones and about 45% of them contained benzene rings. As a rule, experimental data included IR, 1H and 13C NMR spectra, as well as the empirical formula or the molecular weight of the substance. To find possible ways of solving the problems, each of them was repeatedly solved using different methods and strategies. Particular emphasis was placed on elucidation of the reasons for difficulties and failures. To facilitate the logic- statistical analysis of the data obtained, the system automatically stored the solution protocols as text files incorporated into the active database.Plausible structures were elucidated on the basis of prediction of chemical shifts in the 13C NMR spectra using specially developed software.169 Computational experiments were carried out on an AT 486 personal computer operated at 66 MHz. Analysis of the protocols has shown that it is possible to find a correct solution in virtually all the cases considered using the interactive mode. The percentage of unambiguous solutions was 40% and the number n of candidate structures did not exceed 5 for 70% of the problems. The n value exceeded 10 for 15% of the problems. Prediction of 13C NMR spectra of ambiguously solved structures increased the total percentage of unambiguous and right answers from 40% to 75%, i.e., the proper structure was ranked 1st and placed at the beginning of the file.For instance, the answer structures of molecules 2 ± 9 were obtained with indication of n values, reduced size of the answer file after prediction of 13CNMRspectrum (n p) and the number of the proper structure in the ranked answer file n t. The n p value was defined as the number of structures for which the root-mean- square deviation s does not exceed 5.5 ppm (the accuracy of the program 169). About 60% of problems were solved automatically by using SGA or straightforward structure generation (SSG) from empiri- cal formulae. In those cases where the structure was unambigu- ously elucidated by means of SSG followed by structure filtration, exhaustive information on the molecular structure was obtained O O O Cl O N S NH S O O 4 (n=4, np=1) 5 (n=7, np=3) ON OH O O O 9 (n=4, np=1)Expert systems for structure elucidation of organic molecules by spectral methods using characteristic features not observed in the experimental spectra.Moreover, filtration has led to recognition of certain structures containing fragments that were absent in the database. In essence, the strategy of solving a problem is a set of methods for overcoming various difficulties. The authors 141 analysed typical difficulties encountered when using ES. 1. Examples of difficult situations Mixed-valence atoms. Depending on the valence state, mixed- valence chemical elements can form various functional groups (e.g., SH, S=O, SO2, PH, P=O).Each functional group generates its own family of isomers. Therefore, strictly speaking, structural problems should be repeatedly resolved until all valence states of atoms present in the empirical formula have been searched through. The problem is appreciably simplified if available information on the origin of a specimen or data of independent experiments make it possible to determine the valences of all elements in the empirical formula. Specific character of IR spectrum. It is known that IR spectrum is extremely sensitive to slight changes in the molecular structure and electron density. This makes it possible to reveal even minor changes in the molecular structure. However, frequen- cies or intensities can be interpreted only if the structural formula of the molecule is known.This is impossible until the molecular structure has been recognised. For this reason, rearrangements of functional groups in the isomers, which have little effect on the additivity of the molecule, can cause the characteristic frequencies and band intensities to go beyond the limits indicated in the fragment libraries. For instance, in some instances the bands at 3000, 1600 and 1500 cm71 characteristic of benzene derivatives appear to be inactive or too weak to be detected in the spectrum. However, if the NMR spectra of such a compound contain signals in the aromatic region, the X-PERT system assesses the situation and prompts the chemist to accept or reject the presence of an aromatic fragment (in other words, to decide between `yes' and `no', respectively). Another problem arises if a compound is dissolved in a solvent of the CCl4 or CHCl3 type.In this case diagnostically important aromatic bands in the region 700 ± 800 cm71 cannot be used for determination of the types of benzene substitution, which results in increasing the size of the answer file. The IR spectral pattern in the region of stretching vibrations of the OH group is dependent on many factors (the presence of neighbouring atoms and functional groups, solvent properties, etc.). Because of a very strong effect of the intramolecular O7H. . .X hydrogen bond, the parameters of some IR bands [ns(OH), ns(C=O), etc.] are shifted beyond the limits of character- istic intervals.Therefore, deciding between the presence and absence of the OH group often requires the interference of a chemist (the X-PERT system generates a corresponding message). It turned out that joint use of IR andNMRspectra often leads to the loss of right fragments whose characteristic features were not confirmed because of unusual IR spectral pattern. Therefore, the IR spectrum should sometimes be `switched off' during the first program session. Of course, it does not mean that the ES can exclude IR spectra from its `reasoning', since there is a rather large number of functional groups which contain no C and H atoms (e.g., NO2, SO, SO2, NO, ONO, P=O, P=S, etc.) and cannot be immediately detected by NMR technique without using the IR spectral bands for confirmation.The influence of structural peculiarities on the complexity of the problem.At first glance, it may appear that the larger the molecule, the larger the expected n value. Actually, the n value is a very complex function of different structural parameters. Therefore it is quite possible that a large molecule will be identified unambig- uously, whereas the answer file created when recognising a relatively small molecule will contain a large number of struc- tures. Let us consider the most important structural factors affecting the problem solution. If the symmetry of the structure of analysed compound is higher than C1, the solution of the problem becomes easier, since a large number of wrong structures are often rejected after compar- ing the predicted number of signals and their multiplicities in the 13C NMR spectrum with the experimental data.For instance, unique and proper answer structures were found for molecules 10 ± 17. It should be noted that structure 12 containing the O=C7N7C=O fragment was unambiguously recognised despite the fact of the absence of this fragment in the knowledge base. 10ON O 14 O 16 Symmetry of the structure also allows the chemist to use more efficiently the `Multiple Fragments' procedure 29 to generate sets containing admissible (from the viewpoint of the empirical formula) numbers of identical fragments, which makes possible generation of large structures. For instance, a set containing four ester groups and four ethyl groups was created for the empirical formula C14H22O8 and the unique structure 17 was recognised after the filtration procedure.As is known, the number of signals in a 13C NMR spectrum calculated as the number of groups of topologically equivalent carbon atoms does not always correspond to experimental data. For instance, only three signals are observed in the 13C NMR spectrum of molecule 18 instead of the six expected. In this case the proper structure was unambiguously elucidated using prediction of 13C NMR spectrum. Accidental degeneracy of signals in the NMR spectrum or splitting of signals of carbon atoms of gem-dimethyl, isopropyl and other groups near the stereocentres makes computer-assisted structure recognition more difficult and requires interference by a chemist.If there are several stereocentres in the molecule, the answer file often contains a larger number of structures than usually. This is explained by complexity of 1H NMR spectra (a small number or even the absence of the first-order signals, different dH values of geminal protons) and additional splittings of signals in the 13C NMR spectra. For instance, in the case of compound 19 having three stereocentres the program for dC prediction ranked the proper structure 3rd from the beginning of the answer file containing 31 structures in the structural similarity criterion. Large number of candidate structures in the answer file is also characteristic of hetero- and carbocyclic compounds, e.g., for 18 ± 25.This is associated with a large number of rearrangements 543 O N O N 11OH 13 12 N NH 15 O O O OO OO O O 17544 possible in heterocyclic structures with several substituents. Compound 20 can serve as an example. The set including all fragments of this structure, 3( N7CH3), 2( C=O), C=C , C=N7, was created manually. The program generated 181 structures which were subjected to filtration to select 40 structures containing five- and six-membered rings. O N N N N NH OH N N 20 (n=40, nt=14) 18 (n=7, nt=1) 19 (n=31, nt=3) O O N O N NH N O 23 (n=18, nt=1) 22 (n=28, nt=4) 21 (n=14, nt=1) O N S OH 25 (n=35, nt=1) OH O 24 (n=25, nt=1) The S value of 11.6 ppm was obtained for the predicted 13C NMR spectrum of the proper structure 20 (the maximum deviation was 25.6 ppm.).As a result, the proper structure was ranked 14th from the beginning of the list of structures in the structural similarity criterion. Thus, we get a `fuzzy' solution even if the fragment set used is a priori suitable for recognition of the proper structure; therefore, elucidation of a proper structure requires additional information (obtained, e.g., from 2D NMR experiments) and application of more reliable methods for spectra prediction. Analysis of the ways of overcoming difficult situations made it possible to propose methodology resulting in correct solutions of the problems. 2. Strategies of problem solution To obtain a correct solution of a problem using an ES, i.e., to find a proper structure in the answer file containing a minimum number of possible structures (ideally, one structure), it is necessary (but not sufficient) to provide the possibility of select- ing as many fragments constituting the analysed structure as possible.To this end, a large knowledge base should be incorpo- rated into the system and characteristic spectral ranges of the fragments should be rather wide lest the correct fragments be lost. When these conditions are met, it appears that the number of selected fragments is too large since many superfluous (wrong) fragments are also selected. This requires finding and excluding superfluous fragments. Editing the list of selected fragments is hardly possible without participation of a chemist. This procedure can be excluded only if all logically admissible fragment sets and generation of structural formulae are created automatically.As a result, the number of generated structures will be too large, structure generation and filtration will take a long time and the problem even may appear to be virtually unsolvable. In addition, the answer file can contain the structures that contradict a priori information. Hence, in any case checking the list of selected fragments by a chemist seems to be virtually unavoidable rather than desirable. ME Elyashberg Thus, meeting the requirement for the presence of the largest possible number of right fragments in the list of selected sub- structures results in increasing the total number of fragments and, as a consequence, in increasing the size of the answer file.In this case the solution will likely be right, but `fuzzy'. This should be considered as a pay for correctness. The selectivity of fragment selection is strongly dependent on the environment descriptors of the fragments. The stiffer the requirements for the fragment environment, the narrower the intervals of characteristic spectral features and, hence, the higher the SGA selectivity. At the same time, the narrowing of the intervals increases the probability of the loss of correct fragments because of the deviation of an experimental feature beyond the bounds of the characteristic interval. There is also a clear dependence between the size of the structural file and the rigour of definition of descriptors describ- ing the fragment environment.The point is that only those structures which meet all the requirements imposed by the environment descriptors are stored in the course of structure generation. This drastically reduces the number of structures to be checked later. On the one hand, there are problems which cannot be solved at all without descriptor control over structures during the structure generation stage (calculations take too long a time). On the other hand, there is a possibility of losing a proper structure in the case of too rigorous descriptor control. For instance, if the O=C7C=C fragment defined in the library as acyclic (whereas the proper structure contains the C=C bond in the ring) was selected at the SGA stage, all structures containing this fragment in the ring (including the proper structure) are rejected in the course of structure generation. This leads to an incorrect solution.Ideally, it is desired to have an ES characterised by high selectivity at SGA stage and, at the same time, by the capability to generate unambiguous and correct solutions. It is hardly probable to meet both requirements, therefore a compro- mise should be found. Since the basic requirement for the solution is that it should be correct, it is possible to put up with its ambiguity. Therefore, the optimum strategy of structure elucidation can be formulated as follows: the problem should be solved in the interactive mode, which is a sequence of iterations followed by increasing the hardness of imposed constraints.For instance, if dimensionality of the problem is not very large, the library of principal functional groups (LPF, see Section V.1) should be used to select the fragments in the first stage. This provides the possibility of using softest structural constraints to generate a maximum number of fragment environments. Recognition of a large molecule requires additional use of the SGA and AROM libraries. To control the hardness of imposed constraints at the stages of fragment selection and structure generation, the chemist can switch on (or off) the following options: the use of any type of spectra from the available experimental data; the use of signal multiplicities and SSCC in NMR spectra; the use of band intensities and halfwidths in IR spectra; the environment descrip- tor control over structures in the course of structure generation; the environment descriptor control over structures in the course of structure filtration; setting of various tolerances for the predicted number of signals with different multiplicity in the 13C NMR spectrum; the use of structural constraints at the stages of structure generation and filtration (the limiting size of the ring, the maximum bond multiplicity, satisfying Bredt's rule, GOODLIST, BADLIST, list of unlikely fragments, etc.).If no solution was obtained during the first program run which, as a rule, is carried out automatically, rational use of the above-mentioned switchable parameters in the course of iterative (trial-and-error) procedure usually makes it possible to establish the reasons for contradictions, thus facilitating the search for a correct solution.The search for contradictions is particularly efficient when using the controlled filtration mode (see Section V.1). Storing all intermediate results in the active database allows the user to return to the solution of the problem at any step.Expert systems for structure elucidation of organic molecules by spectral methods Fast operation of programs for prediction of 13C NMR spectra (0.5 ± 1 s per spectrum) makes possible effective employ- ment of the strategy outlined above.Thus, to attain a correct solution, the initial constraints may sometimes be slacked to such a degree that the answer file will contain thousands of structures. Prediction of the chemical shifts and the maximum likelihood ranking of the structures are carried out automatically; therefore CPU time taken can be considered justified and admissible. The choice of appropriate strategy of solving a certain problem depends on optimum combination of two basic modes of operation of the system, which can be called the `Archaeologist' mode and the `Sculptor' mode. The `Archaeologist' mode is used to perform SGA and structure generation (to `glue' fragments and atoms). Then the `Sculptor' mode should be switched on to carry out structure filtration and reject wrong structures (in this mode, the strategy is to cut off all superfluous structures).Analysis of a large number of solutions made it possible to divide all problems into the following categories: (1) problems solved automatically using SSG from molecular formulae followed by application of the `Sculptor' mode strategy; (2) problems solved automatically using strategies of the `Archaeologist' and `Sculptor' modes; (3) problems solved using elements of the interactive mode (a chemist edits the `reasoning' of the program using diversified information easily obtained from experimental data); and (4) problems which can be solved only in the interactive mode (manual creation of fragment sets, imposing various constraints due to the origin of a specimen, etc.).Rating a problem in a particular category depends on different factors (e.g., the size of the molecule, the presence of large fragments with characteristic spectral features, symmetry, etc.) and, first of all, on the number of skeletal atoms in the empirical formula. In many instances, SGA can be performed automati- cally; however, the appearance of at least one contradiction between the axioms of the SGA theory and experimental data makes the interference of an expert virtually inevitable. It should be kept in mind that even coincidence of predicted spectra with experimental ones does not guarantee against incorrect choice of the structure. Additional experiments can appear to be necessary, in particular, an X-ray study or structure confirmation by INADEQUATE 2D NMR technique.Despite carrying out addi- tional labour-consuming procedures, the expenses are incommen- surate with the cost of independent synthesis, isolation and purification of the product. Thus, AI programs must be considered as powerful tools of extending the intellectual potentialities of the investigator rather than those of a method designed to exclude a chemist from the structure elucidation process. This idea was clearly expressed by the authors of monograph:170 `Computer-based automated or interactive versions of similar approaches have also been devised for structural elucidation of complex natural products, such as the SESAMI (Systematic Elucidation of Structures by using Artificial Machine Intelligence), but there is no substitute for the hard work, experience and intuition of the chemist.' We cannot but agree with this.VII. Conclusion The first studies reported 30 years ago laid the foundations of an independent line of investigation, the development of ES for elucidation of the molecular structure of organic compounds by spectral methods. It is closely related to many branches of chemistry (organic, physical, analytical, quantum chemistry, stereochemistry, molecular spectroscopy, molecular mechanics), mathematics (theory of graphs, combinatorial analysis, mathe- matical logic, theory of fuzzy sets, mathematical statistics), cybernetics (AI, programming, information theory, pattern rec- ognition, ANN theory, computer graphics, mathematical linguis- tics), psychology (the man ± computer interface) and philosophy of science (the role of the complementarity principle in studying 545 the structure of matter).171 Therefore rapid progress in the development of ES that absorbs new achievements of interdisci- plinary sciences should be expected.Three major stages of the process of structure recognition, viz., the SGA, structure generation and prediction of the spectra of candidate structures, have been transformed into independently developing investigation lines. It has been possible to computerise the structure generation and prediction of NMR spectra to a greater extent than SGA. Several methods were developed to reveal molecular fragments; however, this stage remains the most vulnerable and dependent on the a priori information.Recently, intensive studies on the application of ANN to revealing the fragments have begun. Up to now, all attempts to develop a fast and reliable method for the prediction of IR spectra have failed. Solving this problem requires investigation of the potentialities of ANN (encouraging results have already been obtained) and development of new approaches. The known methods of calculation of IR spectra must be essentially improved to accelerate calculations. It has been shown 172 ± 174 that calculations of IR spectra can also be used for quantitative spectral analysis of organic compounds without any calibration. Therefore incorporation of databases (e.g., such as it was made by Gribov and Zinovyev 138) into ES will allow one to create systems for both qualitative and quantitative spectral analysis of organic substances.Taking into account the accumulated experience in solving many structural problems, the general strategy of elucidation of the molecular structure using ES was established. It consists of performing a number of successive iterations followed by a gradual increase in the hardness of constraints imposed. The use of this strategy makes it possible to obtain correct solutions. Currently, most of ES are capable of recognising small and medium size (up to 20 ± 25 skeletal atoms) molecules; however, this does not mean that ES have limited area of application. Millions of organic structures can correspond to a relatively simple molecular formula (e.g., about 48.5 mln of structures correspond to the empirical formula C9H10O3); since `only' *14 mln of compounds are known currently, it is easy to conclude that the area of application of ES is very wide.Since at present the structures generated by the use of the most sophisticated structure generators contain no more than 10 ± 12 molecular fragments, the ES capable of revealing large fragments are rather promising for recognition of large molecules. Rapid progress of ES in which 2D NMR spectroscopy data are used should also be expected. References 1. M E Elyashberg, L A Gribov Zh. Prikl. Spektrosk. 8 296 (1968) 2. M E Elyashberg, L A Moskovkina Zh. Prikl.Spektrosk. 8 996 (1968) 3. J Lederberg, G L Sutherland, B G Buchanan, E A Feigenbaum, A V Robertson,A M Duffield,C Djerassi J. Am. Chem. Soc. 91 2973 (1968) 4. S I Sasaki, H Abe, T Ouki, M Sakamoto, S Ochiai Anal. Chem. 40 2220 (1968) 5. D B Nelson, M E Munk, K B Gasli, D L Horald J. Org. Chem. 34 3800 (1969) Kvantovaya Mekhanika (Classical Theory of Structure of Molecules 6. V M Tatevskii Klassicheskaya Teoriya Stroeniya Molekul i and Quantum Mechanics) (Moscow: Khimiya, 1973) 7. M E Elyashberg, L A Gribov, V V Serov Molekulyarnyi Spektral'nyi Analiz i EVM (Molecular Spectral Analysis and Computers) (Moscow: Nauka, 1980) 8. R K Lindsay, B G Buchanan, E A Feigenbaum, J Lederberg Applications of Artiécial Intelligence for Organic Chemistry.The 9. N A B Gray Computer-Assisted Structure Elucidation (New York: DENDRAL Project (New York: McGraw-Hill, 1980) Wiley, 1986) 10. T H Pierce, B A Hohne Artiécial Intelligence Applications in Chemistry (Washington, DC: American Chemical Society, 1986)546 11. J Klassens, G Kateman Fresenius Z. Anal. Chem. 326 203 (1987) 12. N A B Gray Anal. Chim. Acta 210 9 (1988) 13. G W Small Anal. Chem. 59 535A (1987) 14. M E Elyashberg Matematicheskie Metody i EVMv Analiticheskoi Khimii (Mathematical Methods and Computers in Analytical Chemistry) (Moscow: Nauka, 1989) p. 132 15. M E Elyashberg, V V Serov, L A Gribov Talanta 34 21 (1987) 16. L A Gribov, M E Elyashberg, V V Serov J. Mol. Struct. 50 371 (1978) 17. A M Denisov Vvedenie v Teoriyu Obratnykh Zadach (Introduction to Theory of Inverse Problems) (Moscow: Moscow State University, 1994) 18.A N Tikhonov, V Ya Arsenin Metody Resheniya Nekorrektnykh Zadach (Methods of Solution of Incorrect Problems) (Moscow: Nauka, 1979) 19. V V Serov, L A Gribov,M E Elyashberg J. Mol. Struct. 129 183 (1985) 20. A Kaufman Theory of Fuzzy Set (New York: Academic Press, 1975) 21. H J Luinge Vibrat. Spectrosc. 1 3 (1990) 22. M E Elyashberg Zh. Analit. Khim. 47 966 (1992) a 23. W A Warr Anal. Chem. 65 1087A (1993) 24. J Zupan, J Gasteiger Neural Networks for Chemists (Weinheim: VCH, 1993) 25. E W Robb, M S Madison,M E Munk Mikrochim. Acta. II 505 (1991) 26. C Affolter, J T Clerc Chemometr. Intell. Lab. Syst. 21 121 (1993) 27. D Ricard, C Cachet, D Cabrol-Bass, T P Forrest J.Chem. Inf. Comput. Sci. 33 202 (1993) 28. M A Sharaf, D L Illman, B R Kowalski Chemometrics (New York: Wiley, 1986) 29. M E Elyashberg, E R Martirosian, Yu Z Karasev, H Thiele, H Somberg Anal. Chim. Acta 337 265 (1997) 30. M Will,W Fachinger, J R Richert J. Chem. Inf. Comput. Sci. 36 221 (1996) 31. ACD/Structure Elucidator (Advanced Chemistry Development Inc., www.acdlabs.com) 32. C Peng, S Yuan, C Zheng, Y Hui J. Chem. Inf. Comput. Sci. 34 805 (1994) 33. L A Gribov, M E Elyashberg J. Mol. Struct. 5 179 (1970) 34. V V Serov,M E Elyashberg, L A Gribov Dokl. Akad. Nauk SSSR 232 592 (1977) b 35. M E Elyashberg, L A Gribov, V N Koldashov, I V Pletnev Dokl. Akad. Nauk SSSR 268 112 (1983) b 36.M E Elyashberg, V V Serov, E R Martirosian, L A Zlatina, Yu Z Karasev, V N Koldashov, Yu Yu Yampolskiy J. Mol. Struct. 230 191 (1991) 37. K Funatsu, Y Susuta, S Sasaki Anal. Chim. Acta 201 55 (1989) 38. G N Andreev, O K Argirov, P N Penchev Anal. Chim. Acta 284 131 (1993) 39. G N Andreev, O K Argirov J. Mol. Struct. 347 439 (1995) 40. V V Serov,M E Elyashberg Zh. Strukt. Khim. 27 (2) 32 (1986) c 41. L J Bellamy The Infrared Spectra of Complex Molecules (Chichester: Wiley, 1975) 42. R R Yager (Ed.) Fuzzy Set and Possibility Theory (New York: Pergamon Press, 1982) 43. C L Chang, R C T Lee Symbolic Logic and Mechanical Theorem Proving (New York: Academic Press, 1973) 44. V V Serov,M E Elyashberg, V M Petrov Matematicheskie Metody i EVM v Analiticheskoi Khimii (Mathematical Methods and Computers in Analytical Chemistry) (Moscow: Nauka, 1989) p.150 45. H J Luinge, G J Kleywegt, H A van`t Klooster, J H van der Maas J. Chem. Inf. Comput. Sci. 27 95 (1987) 46. H J Luinge, J N van der Maas Anal. Chim. Acta 223 135 (1989) 47. H J Luinge Trends Anal. Chem. 9 66 (1990) 48. H Huixiao, X Xinquan J. Chem. Inf. Comput. Sci. 30 203 (1990) 49. C Cleva, C Cachet, D Cabrol-Bass, T P Forrest Anal. Chim. Acta 348 255 (1997) 50. C Klawun, C L Wilkins J. Chem. Inf. Comput. Sci. 36 69 (1996) 51. M E Munk,M S Madison, E W Robb J. Chem. Inf. Comput. Sci. 36 231 (1996) 52. M E Munk, M S Madison, E W Robb Mikrochim. Acta II 505 (1991) 53. N Sbirrazzuoli, C Cachet, D Cabrol-Bass, T P Forrest Neural Comput.Applic. 1 229 (1993) ME Elyashberg 54. J R M Smith, P Schoenmakers, A Stehmann, F Sijstermans, G Kateman Chemometr. Intell. Lab. Syst. 18 27 (1993) 55. C Klawun, C L J Wilkins Chem. Inf. Comput. Sci. 34 984 (1994) 56. E W Robb, M E Munk Mikrochim. Acta I 131 (1990) 57. R J Fessenden, L Gyorgi J. Chem. Soc., Perkin Trans. 2 1755 (1991) 58. O C van Est, P Schoenmakers, J R M Smits, W P M Nijssen Vib. Spectrosc. 4 263 (1993) 59. M Meyer, T Weigelt Anal. Chim. Acta 265 183 (1992) 60. M Meyer, H Hobert Anal. Chim. Acta 282 407 (1993) 61. U M Weigel, R J Herges J. Chem. Inf. Comput. Sci. 32 723 (1992) 62. B J Hare, J H Prestegard J. Biomol. NMR 4 35 (1994) 63. C Klawun, C L Wilkins J. Chem. Inf. Comput. Sci. 36 249 (1996) 64.I I Strokov, I V Gritsenko, K S Lebedev Izv. Sib. Otd. Akad. Nauk. Ser. Khim. Nauk (2) 78 (1987) 65. K S Lebedev, E A Otmakhova, I V Gritsenko Izv. Sib. Otd. Akad. Nauk. Ser. Khim. Nauk (5) 73 (1990) 66. K S Lebedev, O N Sharapova, I K Korobeinicheva, V A Kokhov Sib. Khim. Zh. (1) 50 (1993) 67. K S Lebedev Zh. Analit. Khim. 48 851 (1993) a 68. K S Lebedev, I V Gritsenko Zh. Strukt. Khim. 34 (3) 51 (1993) c 69. K S Lebedev, B G Derendyaev Khim. Interesakh Ustoichivogo Razvitiya 3 269 (1995) 70. K S Lebedev, D Cabrol-Bass J. Chem. Inf. Comput. Sci. 38 410 (1998) 71. K Varmuza, P N Penchev, H Scsibrany J. Chem. Inf. Comput. Sci. 38 420 (1998) 72. L Chen,W Robien J. Chem. Inf. Comput. Sci. 34 934 (1994) 73. W Bremser Anal. Chim.Acta 103 355 (1978) 74. M Carabedian, I Dagane, J E Dubois Anal. Chem. 60 2186 (1988) 75. V V Raznikov, V L Tal'roze Zh. Strukt. Khim. 11 357 (1970) c 76. Y Kudo, S Sasaki J. Chem. Inf. Comput. Sci. 16 43 (1976) 77. V V Serov,M E Elyashberg, L A Gribov Dokl. Akad. Nauk SSSR 224 109 (1975) b 78. C A Shelley, M E Munk Anal. Chim. Acta 133 507 (1981) 79. B D Christie, M E Munk J. Chem. Inf. Comput. Sci. 28 87 (1988) 80. A Kerber, R Laue, D Moser Anal. Chim. Acta 235 221 (1990) 81. C Benecke, R Grund, A Kerber, R Laue, T Wieland J. Chem. Inf. Comput. Sci. 72 403 (1995) 82. C Benecke, R Grund, R Hohberger, A Kerber, R Laue, T Wie- land Anal. Chim. Acta 314 141 (1995) 83. T Wieland, A Kerber, R Laue J. Chem. Inf. Comput. Sci. 36 413 (1996) 84.S G Molodtsov Commun. Math. Chem. (MATCH) 30 213 (1994) 85. S G Molodtsov Commun. Math. Chem. (MATCH) 30 203 (1994) 86. S G Molodtsov Commun. Math. Chem. (MATCH) 37 157 (1998) 87. R E Carhart, D H Smith, N A B Gray, J G Nourse, C Djerassi J. Org. Chem. 46 1709 (1981) 88. S G Molodtsov, Candidate Thesis in Chemical Sciences Novosi- birsk Institute of Organic Chemistry, Siberian Branch, Russian Academy of Sciences, Novosibirsk, 1997 89. M S Molchanova, Candidate Thesis in Chemical Sciences Institute of Organic Chemistry, Russian Academy of Sciences, Moscow, 1997 90. L M Masinter, N S Sridharan, J Lederberg, D H Smith J. Am. Chem. Soc. 96 7702 (1974) 91. M S Molchanova, V V Shcherbukhin, N S Zefirov J. Chem. Inf. Comput. Sci. 36 888 (1996) 92.Ch-L Hu, L Xu Anal. Chim. Acta 298 78 (1994) 93. J-M Nuzillard,W Naanaa, S Pimont J. Chem. Inf. Comput. Sci. 35 1068 (1995) 94. R Barone, F Barberis, M Chanou Commun. Math. Chem. (MATCH) 32 19 (1995) 95. I P Bangov J. Chem. Inf. Comput. Sci. 34 318 (1994) 96. V Kvasnicka, J Pospichal J. Math. Chem. 9 181 (1992) 97. V Kvasnicka, J Pospichal J. Chem. Inf. Comput. Sci. 36 516 (1996) 98. L A Zlatina, M E Elyashberg Zh. Strukt. Khim. 32(4) 92 (1991) c 99. L A Zlatina, M E Elyashberg Commun. Math. Chem. (MATCH) 27 191 (1992) 100. E V Gordeeva, A R Katrizky, V V Shcherbukhin, N S Zefirov J. Chem. Inf. Comput. Sci. 33 102 (1993) 101. J Gasteiger, C Rudolph, J Sadowski Tetrahedron. Comput. Methodol. 3 537 (1990) 102. J Sadowski, J Gasteiger Chem. Rev. 93 2567 (1993) 103. J Sadowski, J Gasteiger, G Klebe J. Chem. Inf. Comput. Sci. 34 1000 (1994)Expert systems for structure elucidation of organic molecules by spectral methods 104. C A Shelley, M E Munk Anal. Chem. 50 1522 (1978) 105. J T Clerc, H A Sommerauer Anal. Chim. Acta 95 33 (1977) 106. L Chen,W Pobien Anal. Chem. 65 12282 (1993) 107. K L Jensen, A S Barber, G W Small Anal. Chem. 63 1082 (1991) 108. D L Clouser, P C Jurs Anal. Chim. Acta 295 221 (1994) 109. P C Jurs, J W Ball, L S Anker, T L Friedman J. Chem. Inf. Comput. Sci. 32 272 (1992) 110. A Furst, E Pretsch Anal. Chim. Acta 229 17 (1990) 111. W Bremser Magn. Reson. Chem. 23 271 (1985) 112. L Chen,W Robien Chemometr. Intell. Lab. Syst. 19 217 (1993) 113. C W Crandell,N A B Gray,D H Smith J. Chem. Inf. Comput. Sci. 22 48 (1982) 114. H N Cheng, L J Kasehagen Anal. Chim. Acta 285 223 (1994) 115. C W von der Lieth, J Seil, I Kohler, H J Opferkuch Magn. Reson. Chem. 23 1048 (1985) 116. A Furst, E Pretsch,W Robien Anal. Chim. Acta 233 213 (1990) 117. R C Schweitzer, G W Small J. Chem. Inf. Comput. Sci. 37 249 (1997) 118. ACD/CNMR Predictor (Advanced Chemistry Development Inc., www.acdlabs.com) 119. V Kvasnicka J. Math. Chem. 6 63 (1991) 120. J P Doucet, A Panaye, E Feuilleaubois, P Ladd J. Chem. Inf. Comput. Sci. 33 320 (1993) 121. A Panaye, J P Doucet, B Fan, E Feuilleaubois, S R El Azzouzi Chemometr. Intell. Lab. Syst. 24 129 (1994) 122. V Kvasnicka, S Sklenak, J Pospichal J. Chem. Inf. Comput. Sci. 32 742 (1992) 123. L S Anker, P C Jurs Anal. Chem. 64 1157 (1992) 124. Y Miyashita, H Yoshida, O Yaegashi, T Kimura, H Nishiyama, S Sasaki J. Mol. Struct. (Theochem) 311 241 (1994) 125. B E Mitchell, P C Jurs J. Chem. Inf. Comput. Sci. 36 58 (1996) 126. D L Clouser, P C Jurs J. Chem. Inf. Comput. Sci. 36 168 (1996) 127. O Ivanciuc, J-P Rabine, D Cabrol-Bass, A Panaye, J P Doucet J. Chem. Inf. Comput. Sci. 36 644 (1996) 128. O Ivanciuc, J-P Rabine, D Cabrol-Bass, A Panaye, J P Doucet J. Chem. Inf. Comput. Sci. 37 587 (1996) 129. D M Grant, E G Paul J. Am. Chem. Soc. 86 2984 (1964) 130. E Breitmaier,W Voelter Carbon-13 NMR Spectroscopy (Weinheim: VCH, 1987) 131. E Pretsch, T Clerc, J Seibl,W Simon Tables of Spectral Data for Structure Determination of Organic Compounds (Berlin: Springer, 1989) 132. N A B Gray, J G Nourse, C W Crandell, D H Smith, C Djerassi Org. Magn. Res. 15 375 (1981) 133. K R Beebe, B R Kowalski Anal. Chem. 59 1007A (1987) 134. R B Shaller, E Pretch Anal. Chim. Acta 133 507 (1981) 135. R B Shaller, M E Munk, E Pretch J. Chem. Inf. Comput. Sci. 36 239(1996) 136. C A Shelley, M E Munk Anal. Chim. Acta 296 295 (1994) 137. ACD/HNMR (Advanced Chemistry Development Inc., www.acdlabs.com) 138. L A Gribov, K A Zinovyev J. Mol. Struct. 268 191 (1992) 139. L A Gribov, W J Orville-Thomas Theory and Methods of Calculation of Molecular Spectra (Chichester: Wiley, 1988) 140. M E Elyashberg, Yu Z Karasev, V A Dement'ev, L A Gribov Interpretirovannye Kolebatel'nye Spektry Uglevodorodovì Proizvodnykh Tsiklogeksana i Tsiklopentana (Interpreted Vibrational Spectra of Hydrocarbons �Derivatives of Cyclohex- ane and Cyclopentane) (Moscow: Nauka, 1988) 141. M E Elyashberg, E R Martirosian, Yu Z Karasev, H Thiele, H Somberg Anal. Chim. Acta 348 443 (1997) 142. J E Dubois, G Mathieu, P Peguet, A Panaye, J P Doucet J. Chem. Inf. Comput. Sci. 30 290 (1990) 143. Ch Affolter, J T Clerc Chemometr. Intell. Lab. Syst. 21 151 (1993) 144. U M Weigel, R J Herges Anal. Chim. Acta 331 63 (1996) 145. J H Schuur, P Selzer, J Gasteiger J. Chem. Inf. Comput. Sci. 36 334 (1996) 146. J Gasteiger, J Sadowski, J Schuur, P Selzer, L Steinhauer, V Steinhauer J. Chem. Inf. Comput. Sci. 36 1030 (1996) 147. J Gasteiger, J Schuur, P Selzer, L Steinhauer, V Steinhauer Fresenius J. Anal. Chem. 359 50 (1997) 148. K Baumann, J T Clerc Anal. Chim. Acta 348 327 (1997) 149. G W Adamson,M F Lynch, W G Town J. Chem. Soc., C 3702 (1970) 547 150. J T Clerc, A Tercovics Anal. Chim. Acta 235 93 (1990) 151. P Willett, J M Barnard, G M Downs J. Chem. Inf. Comput. Sci. 38 983 (1998) 152. N A B Gray Chemometr. Intell. Lab. Syst. 5 11 (1988) 153. B G Buchanan, E A Feigenbaum, J Lederberg Chemometr. Intell. Lab. Syst. 5 33 (1988) 154. N A B Gray Chemometr. Intell. Lab. Syst. 5 37 (1988) 155. M E Munk J. Chem. Inf. Comput. Sci. 38 997 (1998) 156. M E Elyashberg, Yu Z Karasev, E R Martirosian Chimia 52 324 (1998) 157. W Bremser Angew. Chem., Int. Ed. Engl. 27 247 (1988) 158. K Funatsu, Y Susuta, S Sasaki J. Chem. Inf. Comput. Sci. 29 6 (1989) 159. K Funatsu, S Sasaki J. Chem. Inf. Comput. Sci. 36 190 (1996) 160. M E Munk, B D Christie Anal. Chim. Acta 216 57 (1989) 161. M E Munk, V K Velu,M S Madison, E W Robb, M Christie, M Razinger, in Recent Advances in Chemical Information II (Ed. H Collier) (Cambridge: Royal Society of Chemistry, 1993) p. 247 162. M Christie, M E Munk J. Am. Chem. Soc. 113 3750 (1991) 163. C Peng, S Yuan, C Zheng, Y Hui, H Wu, K Ma, X Han J. Chem. Inf. Comput. Sci. 34 814 (1994) 164. C Peng, S Yuan, C Zheng, Z Shi, H Wu J. Chem. Inf. Comput. Sci. 35 539 (1995) 165. P L Fuchs, C A Bunnell Carbon-13 NMR Based Organic Spectral Problems (New York: Wiley, 1979) 166. E Pretsch, J Seibl, A Manz, J Simon Aufgabensammlung zur Strukturaufklarung Organischer Verbindungen mit Spektroskopi- schen Methoden (Berlin: Springer, 1985) 167. R B Bates,W A Beavers Carbon-13 NMR Spectral Problems (Clifton: Humana Press, 1981) 168. J T Clerc, E Pretsch, J Seibl Structural Analysis of Organic Compounds by Combined Application of Spectroscopic Methods (Budapest: Akademiai Kiado', 1981) 169. E Pretsch, A F Furst,M Badertscher, R B Shaller, M E Munk J. Chem. Inf. Comput. Sci. 32 291 (1992) 170. N F Atta-ur-Rahman,M I Choudhary Solving Problems with NMR Spectroscopy (New York: Academic Press, 1996) p. 391 171. L A Gribov J. Mol. Struct. 300 415 (1993) 172. L A Gribov,M E Elyashberg, Yu Z Karasev Anal. Chim. Acta 316 217 (1995) 173. M E Elyashberg, L A Gribov, Yu Z Karasev, E R Martirosian Anal. Chim. Acta 353 105 (1997) 174. L A Gribov, D I Sidelov, I V Maslov Zh. Analit. Khim. 53 706 (1998) a�J. Anal. Chem. (Engl. Transl.) b�Dokl. Chem. Technol., Dokl. Chem. (Engl. Transl.) c�J. Struct. Chem. (Engl.
ISSN:0036-021X
出版商:RSC
年代:1999
数据来源: RSC
|
2. |
Three-body ion–molecule processes |
|
Russian Chemical Reviews,
Volume 68,
Issue 7,
1999,
Page 549-561
Gennadii V. Karachevtsev,
Preview
|
|
摘要:
Russian Chemical Reviews 68 (7) 549 ± 561 (1999) Three-body ion± molecule processes GV Karachevtsev (deceased), P S Vinogradov Contents I. Introduction 549 II. Theoretical models for three-body reactions 550 III. Experimental methods for the study of three-body ion± molecule reactions 554 IV. Temperature dependences of association rate constants 556 V. Some results of studies 556 VI. Conclusion 559 Abstract. Theoretical models used in the studies of three-body ion ± molecule reactions are considered and the experimental methods for the study of these processes are described. The results of experimental studies are presented, including those for gas phase reactions at ultra-low temperatures and for the kinetic isotope effect induced by the symmetry of the species.The bibliography includes 130 references. I. Introduction Three-body ion± molecule processes of the type A++B+M products, (1) where A+ is an atomic or polyatomic ion (in this case, positively charged) andBandMare neutral species, constitute an important class of reactions which can affect substantially the composition and electrophysical properties of the low-temperature plasma. In gas-phase chemical kinetics, the term`three-body reactions' denotes reactions the rates of which depend on the concentrations of the components A+, B and M. The rate of a third-order reaction is represented as W=kt[A+][B][M], where kt is the rate constant independent of the component concentrations, these reactions may be regarded as a special case.Such a reaction order is characteristic, for example, of the association reaction at sufficiently low pressure. The dependence of the reaction rate on the concentrations of the components over a wide pressure range is more complex. This dependence is determined by a combination of several processes: the formation of transition states (with a definite dispersion of lifetimes), decomposition of these states through different chan- nels (into different quantumstates with account of the conserva- tion laws), as well as energy exchange between the species and its intramolecular conversion. It should be noted that the rates and GB Karachevtsev Moscow Institute of Physics and Technology, Institutskii per. 9, Dolgoprudnyi, Moscow region, 141700 Russian Federation.Fax: (095) 408-6869 P S Vinogradov Institute of Energetic Problems of Chemical Physics, Russian Academy of Sciences, Leninsky pr. 38/2, Moscow, 117829 Russian Federation. Fax: (095) 137-8258 Received 11 September 1998 Uspekhi Khimii 67 (7) 605 ± 618 (1998); translated by S S Veselyi #1999 Russian Academy of Sciences and Turpion Ltd UDC 539.186 branching ratios between the channels for some exothermic reactions which are formally represented as second-order reac- tionsA++B products, can depend on the buffer gas pressure. For some of the ion± molecule processes, this dependence can be noticeable even at a pressure of several Torr. The majority of studies regarding three-body ion± molecule processes are devoted to the investigation of association reactions.The pressure range inwhichquantitative measurements of the rate constants have been carried out { usually does not exceed 10 Torr. The most reliable values of the rate constants are obtained by the mass spectrometric method where reactions occur in a flowor ion drift in an electrical field at a buffer gas pressure of *1 Torr. Under these conditions, a third order reaction is observed for the majority of systems studiedinvolving a small number of atoms. As the number of atoms is increased (for some systems, starting from seven or eight), a transition mode is established, which corre- sponds to a change froma third to a second order reaction. The importance of the association processes is due to the fact that it is the composition of the positive ions that largely determines the loss rate of free electrons, i.e., one of the basic parameters characterising the properties of an ionised gas.The recombination coefficients of various types of positive ions with electrons can differ by several orders of magnitude. This can be seen fromtwo examples. In the inert gas plasma, three-body association results in the formation of molecular ions such as Heá2 , Neá2 , Ará2 , which recombine with electrons much more efficiently (in dissociative recombinationprocesses) thanthe corresponding atomic ions that have different neutralisation mechanisms. In the atmosphere of Earth, as the altitude decreases, three- body association affects essentially the kinetics of ionic processes starting fromtheD-layer (height range 75± 90 km). Rocket-based in situ measurements detected complex ions in the D-layer.1, 2 The formationof complex (cluster) positive ions results ina decrease in the concentration of free electrons, as the dissociative recombina- tion coefficients for these ions are about two orders greater than those for non-cluster primary ions.In the Earth stratosphere, where the high-energy intergalactic species are the major ionisa- {Or rather, effective rate constants, since such processes are not elemen- tary in the strict sense.550 tion source, the ion chemistry is completely determined by trans- formations of positive and negative cluster ions.3, 4 Three-body processes play an important role in high-pressure analytical mass spectrometry,5, 6 for example, where chemical ionisation and ionisation at atmospheric pressure are used, as well as in plasma chromatography.7, 8 Termolecular reactions can compete with some slow bimolecular reactions of species exchange 9±17 and with energy exchange processes.11, 18 ±20 In this review, we discuss the specific features of three-body ion± molecule reactions and the application of the developed theories to processes involving ions; the theories based on statistical approaches occupy a special place among them.The experimental methods for the study of termolecular reactions are described. A separate section deals with the analysis of results of studies of the processes occurring at very low temperatures. II.Theoretical models for three-body reactions Three-body reactions involve several steps. The Lindemann scheme of three-body association is rather simplified, however, it gives a qualitative idea of the multistage character of the process. The first step implies the formation of an excited complex in a bimolecular collision; the complex can then undergo a sponta- neous decay into the original species kf A++B AB+*, where kf is the rate constant for the formation of the collisional complex, kd AB+* A++B, where kd is the rate constant for the spontaneous decay of the collisional complex. In parallel with the decay, stabilisation of the excited complex can occur upon collision with the third speciesM ks AB+*+M AB++M, where ks is the rate constant for the collisional stabilisation of the excited complex. Assuming the concentration of AB+* to be quasi-steady, the rate of, e.g., the association process can be presented as Wa à dâABáä à ksâMäâABáä à kfks âAäâBäâMä dt kd áksâMä .The rate constant for a three-body reaction kt (which is of the third order and a dimensionality of cm6 s71) equals kt=kd áksâMä . (2) kfks At the low pressure limit, where [M]55 kd, ks kt does not depend on [M] kt=kfks . kd If the equality (2) does not hold, the association rate constant is usually represented as an apparent rate constant of a quasi- bimolecular second-order reaction (dimensionality cm3 s71) which depends on [M]: k kef=kd áksâMä .fksâMä The dependence on [M] disappears only at the high pressure limit; in this case, kef=k?=kf. GV Karachevtsev (deceased), P S Vinogradov Some authors supplement the scheme with the formally possible decomposition of the AB+* complex upon collisions withM k7s AB+*+M A++B+M. Note that the scheme does not take radiative stabilisation into account kr AB+* AB++hn. This process does not require stabilisation by a third species and can compete with the stabilisation only at very small concentrations ofM([M]4kr/ks). The Lindemann model is based on formal kinetics concepts. Nevertheless, the model shows the key role of the intermediate excited complex AB+*. It is the formation of this complex with an energy sufficient for its decay that allows one to consider three- body association using various approaches developed in the studies on unimolecular decomposition.21± 25 An important role in the transformations of the AB+* complex belongs to energy exchange, including both intramolec- ular energy conversion of the highly excited species AB+* (`reactive molecule' in terms of the unimolecular decomposition theory) and energy exchange with the buffer gas (collisional relaxation7activation). The contemporary theoretical approaches used for the description of the energy exchange and reactions of highly excited polyatomic molecules, as well as the results of optical experimental studies of these processes, are described in sufficient detail elsewhere.26 1.Association of atomic species In the simplest case, where A+ and B are atomic species, the rate constants can be calculated using the method of trajectories. It is also useful to consider other ways of estimating that give a qualitative idea of the magnitudes of the rate constants for this type of reactions. For ion± atom processes at low energies, the ion±neutral species interactions are determined by the long-range potential and kf and ks have a typical value of 1079 cm3 s71. The minimum lifetime of a collisional complex tc=1/kd can approximately be estimated as the time during which one species is located near another one, i.e., as the ratio of several characteristic sizes of the atoms to the velocity of their relative motion.Such an estimation yields a value of the order of kd^1013 s71 and an association rate constant of atomic species of the order of kt^10731 cm6 s71 (see Refs 21 and 27). Let us note an interesting approach to the estimation of the rate constant for ion-atom association based on the data on the mobility of ions and polarisability of atoms. The rate constant for the process can be calculated using the formula 28 (cm6 s71), (3) k=3.27610727 Z5=2ÖaMm=TmAÜ3=4 m0ÖmA ámBÜ where Z is the ion charge in elementary charge units; m is the reduced mass m à mA ámM , mAmM mA, mB and mM are the masses of species A, B andMin atomic mass units, respectively; T is the temperature in K; m0 is the reduced mobility of the ion in cm2 (V s)71 estimated using the Langevin formula (which is valid if the species A and M are different) m0=13.8 aBmAmB , mA ámB aB and aM are the polarisabilities of B andMin 10723 cm3.The experimentally measuredmobilities m0 are usedinthe calculations. This approach is based on the fact that the mobility is determinedThree-body ion ¡¾ molecule processes by the cross-section and kinematics of collisions, i.e., the averaged characteristics of energy transfer in collisions are used. Formula (3) provides a fairly good description of the exper- imental data for the association of atomic species and can be used for reliable estimates of the association rate constants at temper- atures from 70 K to 300 K. Similar calculated values of rate constants are obtained using an approach suggested elsewhere.29 A quantum-mechanical calculation with account of species tunnelling through a centrifugal barrier (i.e., with account of resonance quasi-bound states) yields a correction (less than 20% at ambient temperature) to the data of a classical calculation even for the reaction of light species,30 for example, He++He+M Hea2 +M.Considerable discrepancies arise at very low temperatures. Experimental results for a flow cooled to 1 K after gas expansion through a nozzle 31 ¡¾ 34 do not correlate with the temperature dependence which follows from formula (3). The reasons for this discrepancy are not known yet. Probably, one has to take the quantum effects in scattering processes into account in order to provide agreement between the experimental results and the calculation.35 The importance of these effects has been noted, e.g., in studies of the ion mobility in a gas at low temper- atures.36 ¡¾ 38 In addition, the dynamics of the collisional complex at very low gas temperatures can be affected by thermal infrared radiation of the relatively `hot' walls.39 This radiation affects the energy redistribution in the system, which affects the lifetime of the intermediate complex.The approach described above is of interest for the estimation of the association rate constants of electronically exited atomic ions. It is known that the mobility of such ions can differ from the mobility of ions in the ground state.40 The difference is due to partial electron density transfer from the neutral species to the ion in collision. The transfer is possible over rather long distances (10A).This results in a decrease in the effective polarisability of the neutral species and should affect the collision cross-section and the association rate constant. An analysis 41 showed that the available experimental results on the association rate constants involving polyatomic species are poorly described by the formula (3). Nevertheless, the following qualitative conclusions may be made from comparison of the experimental results and those of calculations presented in Fig. 1: the points above the straight line, which characterises the equality between the measured rate constants and those calculated using the formula (3), correspond to the formation of complexes with high bond energy (more than 1.5 eV); the points below the straight line correspond to the formation of weakly bound complexes.This correlation is related to the characteristic features of the energy exchange processes in polyatomic systems and may be explained 7ln kexp 76 74 72 70 68 66 64 71 70 69 68 67 7ln ktheor Figure 1. Correlation between the experimental rate constants for molecular ion association and those calculated using the formula (3).41 551 based on more rigorous theoretical approaches with the use of the statistical theory. Qualitatively, the divergence between the calculations and experimental results can be explained by the fact that the characteristic lifetime of a complex is longer than that determined by the kinematics of collisions of `point' species having no molecular structure.An increase in the lifetime of the complex in molecular systems virtually does not affect its mobility but results in an increase in the association rate constant. For weakly bound complexes, the overestimated calculated rate constants for the processes involving molecular species may be explained by a decrease in the efficiency of the complex stabilisation. Upon a collision with the third body, the latter takes away a rather small fraction of the excess internal energy of the complex (*kT). Therefore, even after collision, the energy in the internal degrees of freedom of the complex is sufficient for its decomposition. 2.Characteristic features of ion ¡¾ molecule interactions Close similarities exist between the termolecular reactions involv- ing only neutral species 42, 43 and ion ¡¾ molecule reactions. On the other hand, there are also differences caused by the specific character of interaction between ions and neutral species. They are characterised by longer-range interactions than colliding neutral species. In addition, there are also certain specific features in the interactions at short distances. The peculiarity of long-range interactions results in the cross- sections of ion ¡¾ molecule collisions, and hence the rate constants for collisions, being much greater than the corresponding values for neutral species.The formation of a collisional complex occurs in the first step of three-body processes, and this is one of the reasons why the overall rate for ion ¡¾ molecule interaction is greater than that for neutral ¡¾ neutral species. The typical rate constants for collisions between ions and molecules at room temperature are of the order of 1079 cm3 s71. Therefore, the rate constants for three-body reactions of ions may be expected to be *2 ¡¾ 3 orders greater than similar reactions of neutral species at the low pressure limit due to the greater values of kfks.The dependence of the collision rate constant on the transla- tional energy of species is determined mainly by the polarisation interaction (i.e., the interaction of an ion with an induced dipole) and the ion ¡¾ permanent dipole interaction if the neutral species has a permanent dipole moment md.If one assumes that the dipole is always directed along the radius-vector r connecting the colliding species regarded as point masses, the corresponding effective potential has the form U (4) ef a L2 2mr 2 ¢§ 2arq24 ¢§ mrd2q , where L is the orbital angular momentum, m is the reduced mass, a is the polarisability of the neutral species, and q is the ion charge. The first term, i.e., the centrifugal potential, considers the fact that the relative motion occurs in the central ion. If the molecule is non- polar (md=0, the Langevin model), the collision rate constant is determined by the expression 44 (5) rAAA ma .kL=2pq If the neutral species has a permanent dipole moment, the dynamics of the collision process becomes much more complex. Formula (4) describes the so-called locked dipole model, while the dipole rotation, which is the closer to the free rotation the greater is r, is not taken into account. One of the coordinates of the potential energy surface of species interacting over long distances, U(r,0), is the angle y between the direction of the dipole axis and the radius-vector552 U(r,0)=¢§aq2 q 2r 4 ¢§md r 2 cos y. The effect of the ion¡¾dipole interaction on the collision rate constants has been studied in sufficient detail 45 ¡¾47 and proce- dures have been developed for the calculation of these con- stants.45, 48 In spite of the fact that the ion¡¾ dipole interaction is long- ranged (*71/r2), its contribution to the overall collision cross- section is not great as comparedto the contributioninthe case of a locked dipole model.Because of the dipole rotation, the time phases of attraction (cos y>0) and repulsion (cos y<0) are averaged to some extent, and the total effect is comparable with the contribution from the polarisation interaction; however, dependence of the collision rate constant on temperature appears in this case. For some molecules, the polarisability a can be anisotropic, i.e., it can depend on the molecular orientation in an electrical field. However, taking into account the relationship (5) and the averaging effect described above, it may be noted that the contribution of this anisotropy is insignificant (usually <10%).If the molecule has a quadrupole moment Q, a is also anisotropic. The corresponding contribution to the interaction potential is determined by the equation QqO3 cos2 y¢§1U . r3 UQ(r,y)=12 Estimates show that the contribution to the collision rate constant due to ion¡¾ quadrupole interaction (!Q2/3/a1/2 ) is not very large in comparison with the Langevin constant (*10%). The contributions due to other types of interactions proportional to 71/rN (the exponent N depends on the interaction type) are also relatively small.45 To summarise the above considerations, it should be noted that long-range interactions of ions with molecules and atoms are mostly determined by electrostatic forces represented as a sumof terms with different exponents N.The main contribution to the collision rate constant is given by the almost isotropic polarisation interaction (! 71/r4). For polar molecules, the rotation of dipoles is not `frozen' at long distances, and the contribution to the collision rate constant due to the ion¡¾ dipole interaction can be comparable with the Langevin constant only for a sufficiently large dipole moment (>3 Debye units). The Langevin model provides rather precise estimates of actual collision rate constants at translational energies of species from0.01 to 1 eV. For a typical value of kL=1079 cm3 s71, the cross-section s(u)=kL/u=3.3610714 cm2 at the initial (r=?) velocity of the relative motion of species u=3610714 cms71.In this case, the maximum value of the aiming parameter b*(u) allowing a collision to occur is =10.5A, b*(u)=sOpuU 1=2 while the `orbiting' radius rc(u) defined from a condition where Uef(r,u) is a maximum(or the top of the `centrifugal hump'), is rc(u)= 2bOuU=7.4A. pAAA This value is muchlarger thanthe characteristic sizes of atoms and small molecules. A criterion for the collision to occur is that the initial translational energy exceeded Uef(rc), i.e., the image point passed above the `centrifugal hump' into the region of short distances. The value of rc may be regarded as an arbitrary border dividing the regions of interaction at short and long distances.The value of rc enables an approximate estimate of the position of the `critical' surface (transition state) in the theoretical analysis of the processes with the use of statistical and quasi-equilibrium approaches. The actual position of this surface should be GV Karachevtsev (deceased), P S Vinogradov determined by the more precise variational methods. The corre- sponding criteria are found, for example, fromthe condition qaWs¢§1OE¢§UOrU; rUa a 0, qr where Ws71(E7U(r),r) is the sum of the allowed states for all degrees of freedom, except for the reaction coordinate r, in the energy range E7U(r); E is the total energy. At short distances, the interaction between the species cannot be represented as a sum of electrostatic interactions.In this case, one has to use various quantum-chemical methods (ab initio and modifications of the semiempirical methods). Nowadays, it is possible to calculate the optimum geometry of stable ionic complexes, their dissociation energies, frequencies of normal vibrational modes and rotational constants with sufficient preci- sion using computer programmes. In turn, these data allowone to find equilibriumconstants and other kinetic characteristics. When comparing ion¡¾ molecule processes with those involv- ing neutral species, it should be noted that the bond energy of the complexes A+B, i.e., products of association of a molecular ion formed froma stable moleculeAand a stable molecule B, is often much higher than the bond energy of neutral van der Waals clusters AB. At short distances, charge delocalisation is typical; distribution of the effective charge between the species can be represented as A+(17d)7Bd+.2)a It is noteworthy that some stable complex ions, such as (O 2 , (CO)a2 and (NO)a2 , have a zigzag structure, i.e., the axes of the moleculesO2,COandNOare not collinear but are parallel toeach other and are coplanar. Let us also note that the multiplicity of the ground state of the (O2)a2 complex equals four, which means that this state is formed with parallel spins of the ion (doublet) and the O2 molecule (triplet). The reactions of organic molecular ions with the precursor neutral molecules (A++A) most commonly result in protonated ions AH+ as the main products.In the association products of such ions (protonated dimers AH+A), the speciesAappear non- equivalent, i.e., they have different structures and effective charges. The structure of the complex ion should be represented as A7H+7A, where one of the bonds is longer than the other by 40%¡¾ 60%. Thus, ion¡¾ molecule interactions are characterised by greater collision cross-sections than interactions of neutral species and wider `potential wells' in the interactions at short distances. However, despite these features, the fundamental ideas of the theories used for three-body ion¡¾ molecule processes and those involving neutral species are alike. Still, the purely technical details can differ essentially in the calculations for particular systems.3. Statistical methods for the calculation of the rate constants of three-body reactions The statistical approaches have already been used for theoretical calculations of chemical reaction rates for nearly 70 years. The best known among them are the activated complex theory, the RRKMmethod (the quasi-equilibriumtheory) and the statistical theory (phase space theory) taking the conservation of angular momentum into account. To date, these methods have been described in sufficient detail (see, e.g., Refs 21 ¡¾ 26). The wide application of statistical methods to ionic processes startedinthe early 1950s after the statistical theory of mass spectra had appeared; this theory was used for the calculation of the energy distribution of fragmentation products (decomposition pathways) in the ionisation of molecules by electrons with differ- ent energies.Later, they found use in ion¡¾ molecule reactions. The simplest formof the formula fromtransition state theory is exp , Q# Q kOTU a kBT h ¢§k Ea BT Three-body ion ¡¾ molecule processes where kB is the Boltzmann constant, h is the Planck constant, T is the temperature, Ea is the activation energy, Q# is the partition function for the transition state, and Q=QAQBQtr is the product of the partition functions for the reacting species A and B and the translational statistical sum of their relative motion. The micro- canonical equation of the RRKM theory has the form kOEU a W#OE ¢§ E0U , hr where W#(E7E0) is the number of states (vibrational and rota- tional) in the energy range (E7E0), E is the total energy, E0 is the dissociation energy and r is the total density of vibrational- rotational states of the reagents.The most reliable results are obtained in the calculations of the rate constants with the use of phase space statistical theory that takes conservation of the total angular momentum (J) into account. This approach was used for calculations of the rate constants of three-body ion ¡¾ molecule reactions (see, e.g., Refs 49 ¡¾ 51). An analogue of the Lindemann scheme of the association process is presented by the following scheme within the framework of the statistical model:52 kf(E7E0) A++B AB+*(E7E0 , J ), kd(E7E0 , J ) AB+*(E7E0 , J) A++B, ks AB++M.AB+*(E7E0 , J)+M The rate constants of the reactions are presented in a micro- canonical form, but a two-dimensional one, kOE ¢§ E0; JU a W#OE ¢§ E0; J U hr and then used in this form in the expression for the overall reaction rate. For example, the association rate constant at small pressures is represented as:52 k (6) dE FOE ¢§ E0; J U k t a kfbkc O? O JOE¢§E0U d a kcbaMa dJ , 0 0 where b is the stabilisation probability, kc is the rate constant for the corresponding collision, F(E ¡¾ E0,J) is the function of distri- bution over the states taking the conservation of the angular momentum of the AB+* complex into account, and J* corre- sponds to the greatest possible angular momentum for the current value of E.For the sake of clarity, the stabilisation rate constant in the formula (6) is presented not in the microcanonical, but in a simplified form ks=bkc, i.e., as the product of the stabilisation efficiency per collision of the AB+* complex with the speciesMby the rate constant for the corresponding collision. The relation- ships for the calculation of F(E7E0,J), kd(E7E0,J) and the effective constant are reported in Ref. 52. Formula (6) is a stricter analogue of the expression (4). The statistical approach was used to obtain the temperature dependence of the association rate constant at the low pressures limit. This constant was found to be mainly determined by the difference in the rotational degrees of freedom of the complex and the reagents (S) kt^T7S/2 .A correction (X) is introduced in order to take the contribution of vibrations into account, which depends on temperature and is estimated from the ratio of the statistical vibrational partition functions kt^T7(S/2+X) . 553 The effect of X becomes noticeable at rather high temper- atures. The use of the statistical theory requires detailed data on the activated complex and rather complicated calculations. The success of its application for particular systems often depends on how successfully the parameters of the complex are chosen and how meticulously the distribution functions are calculated. It should be noted that even rather simple approaches can sometimes be used successfully for the interpretation of the experimental results.For example, the simplified statistical approach has been used 53 for the reaction C3H++H2+He C3Ha3 +He . Statistical methods were used to find the boundaries of the area of experimental results processing where the dependences should conform to the behaviour at the low pressure limit for the ion ¡¾ induced dipole 54 and ion ¡¾ dipolar molecule association reactions.55 A stricter description of the three-body association process involving complex molecules is based on the solution of the two- dimensional master equation [such as equation (6)]. The corre- sponding approaches have been considered in detail;56 the methods of its solution with account of the specificity of the ion ¡¾ molecule interaction can be found in Refs 57 ¡¾ 59.The approach with the use of the master equation allows one to take into account the contribution of weak stabilising colli- sions, where the third bodyMaccepts a relatively small fraction of energy from AB+*. The remaining energy is sufficient for the decomposition of the complex, while complete stabilisation requires several collisions of this kind.56 ¡¾ 59 For unimolecular decomposition of the AB+* complex with the characteristics E7E0 and J, the two-dimensional master equation taking into account both the change in the internal state of the complex and its decomposition into the original species, has the following form at any pressure: aP2DOe; J?e1; J1UgOe1; J1U¢§ (7) OO kunig(e,J )=o ¢§P2DOe1; J1 ?e; J UgOe; J Uade1dJ1 ¢§ kdOe; J UgOe; J U, where kuni is the rate constant of the process, e=E7E0, o is the frequency of collisions between AB+* and the species M, P2D(e,J?e1,J1) are the probabilities of transitions between the states (e,J?e1,J1) induced by collisions, and g(e,J) and g(e1,J1) are the populations of the states with a given energy and angular momentum.The separation of variables method was used to simplify equation (7):57 P2D(e,J?e1,J1)=Pe(e,e1)PJ(J,J1), and present the functions in a form convenient for calculations. Let us note some features of the use of the statistical theory for ion ¡¾ molecule association. One of them concerns taking into account rotational freezing if the species B is a dipole. At long distances betweenA+and B, the species B is considered as a rotor, while at a distance shorter than some critical value rcr estimated from the relationship mdq cos y 4kBT , r2cr where md is the dipole moment, it becomes an oscillator.For example, rcr=13 A for the HCN molecule, and rcr=23 A for acetonitrile. The expressions for the partition functions of the oscillator and the rotor are different. There is an approach which takes the `hidden rotor' into account.57 Yet another specific feature is the rather large moment of inertia of the transition state (I*). Many of the methods for the calculation of the association rate constants from approximate solutions of the two-dimensional microcanonical master equa- tion, which are used for the calculation of the association rate554 constants of neutral species, can give wrong results when applied to ion± molecule processes.The approaches used for neutral systems are valid if the ratio of the moments of inertia of the transition state and the reagent is I*/I<6. This inequality usually holds for reactions of neutral species. If the inequality is not valid, the role of weak collisions increases. This factor can be taken into account using the reported procedures.57 III. Experimental methods for the study of three- body ion ±molecule reactions Mass spectrometric methods are the most popular and convenient ones for the study of three-body ion± molecule reactions. The ratio of the number of three-body collisions (Z3) to the number of two-body ones (Z2) is proportional to the concentration of the speciesMin the reactor Z3 Z2 à ktâMä . kf Therefore, one has to increase the pressure in the region where these processes occur to increase the yield of the products.At a pressure of 1078 Torr in the mass spectrometer, the effective time of formation of, e.g., a neon molecular ion in the reaction Ne++2Ne Neá2 +Ne is*103 years.60 The contribution of two-body ion± molecule reactions to the mass spectra obtained with mass spectrometers with conventional ion sources becomes appreciable at pressures of*1076 Torr and a characteristic reaction time of ca. several microseconds. The study of three-body processes requires muchhigher pressures, viz., at least 1072 Torr.This implies that the reactionzone andthe zone where ion separation according to their masses occurs should be separated, and in the second area the vacuum required for the analysis (no less than 1076 Torr) should be provided. In ordinary mass spectrometers, small ionisation chambers (with a characteristic size of 1 cm and even smaller) with rather narrow diaphragms (to minimise the gas flow into the high- vacuumzone) are used as a reactor. Primary ions are usually formed under the action of a fast- electron beam. Diffusion of primary ions to the ionisation chamber walls is accompanied by various reactions. In the vicinity of the diaphragm, the ions are carried away by the gas flow into the high-vacuum zone, separated according to their masses bythe mass analyser anddetected.The processes occurring in the ionisation chambers at higher pressures have been consid- ered elsewhere.7, 61, 62 The ratio between the current of secondary ions I2 formed in three-body reactions [see reaction (1)] to the current of primary ions I1 at lowdegree of conversionis determinedby the expression 2 II à ktâBäâMät , 1 where kt is the reaction rate constant (with the assumption of the low pressure limit) and t is the lifetime of ions in the ionisation chamber. If B=M (it is for such systems that ordinary mass spectrometric methods are most commonly used), then 2 à ktâBä2t . 1 IIThis formula is used to determine the rate constants of three- body reactions from the experimental results.Of greatest diffi- culty is the determination of t, as the residence time of ions in the ionisation chamber is determined by many factors: diffusion, drift in the field of the space charge of ions, `penetration' of the extracting field into the chamber, gas flow and geometric parameters of the chamber. The problem can be simplified by using the pulse method suggested by Tal'roze and Frankevich.63 These methods are GV Karachevtsev (deceased), P S Vinogradov mainly used for the study of ion± molecule reactions at elevated pressures in the ionisation chamber.64 ±67 Ionisation is carried out by a short pulse of electrons. The time dependences of the intensity of ion currents with different masses (over a sufficiently long time period) are studied with the use of a multichannel pulse analyser.64 ±66 Ionisation pulses are repeated with a certain frequency, and acquisition of data on the time dependence of the current for each type of ions is performed. This is necessary to obtain reliable results as regards the signal-to-noise ratio.Processing of the kinetic dependences of the ion mass spectra obtained in such a way enables one to find the reaction rate constants. However, the pulse method suffers a number of drawbacks, the main one of which is that a large excess of the buffer gas is not allowed. If B=M, the mechanismof stabilisation of the complex is much more complicated, as the substitution of the species B, i.e., BA+*+B B+AB+#. can contribute to certain extent. Instead of the actual reaction time, the kinetic dependences include its analogue which is averaged in a rather complicated way.The flow and drift mass spectrometry methods are the most reliable.The use of these methods for the study of the ion± molecule reaction kinetics has been reviewed rather thor- oughly (see Refs 68 and 69). The flow method (this is also called the flowing after glow method) makes it possible to study the reactions of ions thermo- lysedinthe excess of a buffer gas (He, Ar,N2, etc.) at temperatures from80Kto several hundred degrees (Kelvin scale) and pressures fromtenths of a Torr to several Torr.70± 73 In this method, the region where the primary ions are formed is spatially separated fromthe reaction zone, which begins where the neutral reagent is added to the flow of the buffer gas containing primary ions that had already been thermolysed.The reaction zone is ended by a nose cone with an orifice in the top, which withdraws a part of the flow for the ionic composition analysis. The time period during which the reactions are studied is determined by the flow rate and the place of injection of the neutral reagent and amounts to several milliseconds. A useful modification of this method is the injection of ions, which have previously been separated by mass with another mass spectrometer, into the flow [the so-called SIFT (selected ion flow tube) technique]. The main drawback of the SIFT technique is that the conditions of injectionfromthe lowpressure zone into the reactor do not allow increasing the pressure in the latter above 1 Torr.Thus, despite the main advantage that only one type of primary ions is considered, the procedure for the processing of the kinetic result is simplified and the reliability is enhanced, the possibility of monitoring the changes in the reaction order occurring at pressures above 1 Torr for many three-body reac- tions is lost. A similar principle for studying ion±molecule reactions underlies the drift methods. The motion of ions along the reactor is determined by the rate of the ion drift in an electrical field (ud 44uf) generated by a potential difference between the guard rings located along the drift tube rather than by the gas flow rate.The drift methods allow one to study the dependences of the reaction rates on the energy of the relative translational motion of the colliding partners 74 ±76 which can vary over a wide range (from the value close to that of ambient-temperature motion to several electron-volts in the centre-of-mass system) by varying the electrical field. At present, drift methods make use of the injection of mass-selected ions into the drift tube. This method is known as selected ion flow-drift tube (SIFDT) method. In the drift methods, there are nonequilibriumdistributions of species according to their energies. These can be characterised as the translational collision temperature of an ion and a molecule, the effective temperature of the internal degrees of freedomof an ion (the relevant formulas can be found in Refs 74 and 77) and theThree-body ion ± molecule processes 1010 kef /cm3 s71 123456789 10 54321 0.6 0.4 0.2 0 PHe /Torr Figure 2.Dependence of the effective rate constant for the process (8) on pressure at different values of E. E /meV: (1) 41; (2) 42; (3) 47; (4) and (5) 49; (6) 59; (7) 61; (8) 75; (9) 87; (10) 100. temperature of the neutral reactant, which is equal to the temper- ature of the reactor walls. If the translational energy (E) in collisions of rather heavy species (528 amu) is less than 0.1 eV and helium is used as the buffer gas, the effective temperature of the internal degrees of freedom of an ion does not differ much from the ambient temperature.This allows one to study correctly enough the dependences of the association rate constants on E in the range from 0.04 to 0.1 eV.78 Figure 2 can be regarded as an example, it presents the dependences of the effective rate constant for the reaction (8) Oá2 CH3CN+He Oá2 +CH3CN+He on the helium pressure, and Fig. 3 presents the dependence of the ratio of the rate constants for stabilisation in collisions of the complex with helium to the spontaneous reverse decay rate constant of the collisional complex at different E values. ks/kd /Torr 3210 0.09 E /eV 0.07 0.05 Figure 3. Dependence of the ratio of the rate constant for collisional stabilisation to the rate constant for the collisional complex decay on E for the process (8).The drift method allows one to study, over a wide energy range, the competition of the association channel with other channels, for example, endothermic charge transfer,79 or with reactions accompanied by species rearrangement.80 At sufficiently high energy of species relative translational motion in the centre-of-mass system, the SIFDT method is very helpful for the studies of the collision-induced dissociation of stable ionic complexes AB+, since the reaction zone contains no neutral reagent B and the reverse process, viz., association, does not occur there. The dissociation of a series of complex ions has been studied in detail.77 A procedure with a modified flow reactor has been developed for the measurement of the rate constants for ion ± molecule reactions at low temperatures in which the reactions were studied 555 1 ~~ 2 5 3 6 ~ ~ 4 ~ ~ Figure 4.A scheme of an experimental set-up for low-temperature studies of ion-molecule reactions:89 (1) gas-dynamic nozzle with a pulse valve for gas flooding; (2) focus of ionising laser irradiation; (3) a plate for the deflection of ions to the input of the mass spectrometer; (4) time-of-flight mass spectrometer; (5) micro- channel plates; (6) collector of secondary electrons. in a cooled supersonic flow.31, 81 ± 86 A strongly cooled flow is obtained by adiabatic gas expansion following its discharge through a nozzle or an orifice in a thin wall. A scheme of this type of experimental set-up is shown in Fig.4. The gas expands through a pulse nozzle. The ionising laser radiation is focused on the axis of the supersonic flow at a point located sufficiently close to the nozzle. The ions moving together with the supersonic flow react with molecules. When the ions reach the level of the repeller plate, the pulse electrical field directs them into the time-of-flight mass-spectrometer moved along the flow; the mass-spectrometer records the ion composition of the beam. The yield of secondary ions in reactions can be increased not only by increasing the gas pressure, but also by increasing the time of ion residence in the reactor under low pressure conditions. Various types of ionic traps are widely used for this purpose in mass spectrometry.87 ± 93 In the flow and drift methods, the order of characteristic reaction time is 1073 s, while in ionic traps this time can be many orders longer.In ion cyclotron resonance (ICR) spectrometers, the cell retention time for an ion can amount to several hours at a very high vacuum. However, kinetic measure- ments require higher pressures, but under these conditions the reaction time is usually less than a minute. In radio frequency traps, such as QISTOR, effective retention of ions requires the presence of a buffer gas (usually, helium) at a pressure of 1073 Torr. However, the ion translational energy in these traps is rather high (>1 eV), hence, they can be used only for studying the dissociation of complex ions that have been pre- viously injected into them.Yet more promising are multipole radio-frequency traps (of the `ion guide' type) 93 in which the energy of ions is not high. The ionic-trap and ICR methods, like the drift method, allow variation of the translational energy of the reacting ions. The ICR method is limited to small pressures. Therefore, it can be used for studying the association reactions of relatively complex species, with rather long complex lifetimes (of about a microsecond and more). Only the reagent can act as the stabilising species in this case. However, this method proved to be convenient for radiation stabilisation studies. This can be clearly demon- strated using the Lindemann scheme with allowance for radiative stabilisation with the rate constant kr à 1 , tr where tr is the radiation lifetime of the species.The effective rate constant of association (represented as a second-order reaction) has the form556 k (9) ef a k kfksaMa d akr aksaMa a kd akr aksaMa kfkr and has a nonzero value when [M] is extrapolated to zero. Experimental studies of the processes taking into account the competition between the collisional and radiative stabilisation are described in Refs 94 ¡¾ 100. A more perfect statistical model for data interpretation [in comparison with that used to derive formula (9)] is reported elsewhere.101, 102 The results of the rate constant measurements for three-body ion¡¾ molecule reactions using the methods described above are tabulated in reference books.103, 104 IV.Temperature dependences of association rate constants As noted above, the temperature dependence of the association reaction is a negative power function k^T¢§n. (10) At relatively low temperatures, where the energy difference between the vibrational levels is much greater than the thermal energy (hu44kBT) for simple molecules, the contribution of the relative vibrational partition functions to the rate constant can be neglected. Under nonequilibrium conditions where the effective temperatures of the internal degrees of freedom for ions (Ti), neutral molecules (Tm) and collisional complexes (Tc) differ (as in the case of ion drift), the temperature dependence of the associa- tion rate constant at the low pressure limit has the form 76 Tcrc=2 , k! T3=2 tr Tiri=2Tm rm=2 where Ttr is the temperature that characterises the energy of the relative motion of the colliding molecules; rc, ri and rm are the numbers of the rotational degrees of freedomof a complex, an ion and a neutral molecule, respectively.It should be noted that the electrical field can affect the translational temperature of ions and their internal energy but not the temperature of neutral species. The validity of this formula was confirmed experimentally in studies of ion drift in an electrostatic field. In the case of equilibrium temperature, the exponent n is determined by the relationship n=3¢§rc ari arm . (11) 2 It should be noted that both the statistical theory, which takes conservation of the angular momentum into account, and the simplified theory, which disregards this, give identical expressions for n.105 It should be noted that the statistical theory is applicable only to the reactions occurring with the formation of an intermediate complex with a sufficiently `deep potential well', which is the reason for its long lifetime necessary for the efficient energy exchange between the degrees of freedom.For example, the temperature dependence of the rate constant for the process (12) Na2 +N2+N2 Na4 +N2+96.5 kJ mol71 is rather well described by the statistical theory, n=2, whereas for a similar process (13) NO++N2+N2 NO+N2+N2+17.4 kJ mol71 n=4. For complexes with a small bond energy, a transition from statistical to dynamic control of the lifetime of the intermediate complex takes place.105, 106 In this case, the lifetime of the complex is givenby the effective time of collision, i.e., the time during which the species remain sufficiently close to each other.This value is estimated by numeric calculation of trajectories. The large exponent n for the formation of the intermediate complex GV Karachevtsev (deceased), P S Vinogradov (NO+N2)* is a consequence of dynamic restrictions for the total rotational energy of the colliding species (at fixed energy, collisions are possible only at small values of the total orbital momentum). As the molecules become more complex and/or the temper- ature increases, it is necessary to take the contribution of the vibrational partition functions of the colliding species and the intermediate complex into account.Relatively simple models of harmonic oscillators are used for the estimation of this contribu- tion. The corresponding correction of the parameter n in the formula (10) gives results which agree satisfactorily with the experiment.107, 108 The contribution of the aforementioned radiative processes to the stabilisation of complex molecules, viz., association products, can increase considerably. The competition between the channels for complicated hydrocarbon ion¡¾ molecule complexes was considered in detail;94, 109 it was shown that their ratio evaluated using the approach mentioned above agrees satisfactorily with the experiment.For very complicated systems, the assumption of the uniform energy distribution between the degrees of freedom of a collisional complex may no longer agree with the experimental results.110 V. Some results of studies Anew type of isotope effect related to the involvement of several potential energy surfaces with different symmetry types in one process has been discovered in three-body ion¡¾ molecule associ- ation processes. The simplest of the processes studied is the association in helium 111 He++2He Hea2 +He. In a mixture with approximately equal concentrations of 3He and 4He, the molecular ions with the isotope composition 3He4He+ are formed with a higher efficiency than the 3He3He+ and 4He4He+ ions.The relative concentration of the mixed molecular helium ions [3He4He+] exceeded approximately 2.5 times the statistical value, which would be observed if equal association rate constants independent of the isotope composi- tion were assumed. This isotope effect was rationalised as follows. Only for a half g , of collisions between the species of identical isotope composition is the interaction potential defined by the term 2Pau of the ion Hea2 , which corresponds to attraction and passes through a minimum; for the other half of the collisions, the potential interaction energy corresponds only to the repulsive term 2Pa and the species cannot approach for short distances where the formation of a bound state becomes possible (Fig.5 a). a U b U 1 12 2 r r Figure 5. Dependences of the potential energy on the internuclear 4 g , (2) 2Pau ; (b) the 3He++4He (1) and distance for the systemHe++He.111 (a) the 4He++4He system, (1) 2Pa He++3He (2) systems. If the isotopes are different, at sufficiently long distances between the species, along with the term of the original species there is a termwith a very small energy difference correspondingThree-body ion ¡À molecule processes to the charge transfer products, i.e., there is a splitting of terms for the 3He++4He and 4He++3He systems (Fig. 5 b). It is assumed that the binding state of the molecular ion becomes available in every collision, as the transition to the binding term occurs in a nonadiabatic transition in the area indicated in the Figure.The quantum-mechanical restrictions for symmetry (more exactly, the restrictions due to the identity of species) forbid a transition into the ground state from the 2P�¢g state for identical isotopes. Thus, the isotope effect depends not only on the masses of the species involved in the process, but also on this type of quantum-mechanical restrictions. Let us compare, for example, the kinetic isotope effect related to the mass difference with the isotope effect due to the symmetry (identity or nonidentity) of the species for the process 20Ne22Ne++He. 20Ne++22Ne+He This process has not been studied experimentally yet. The interest in it is due to the fact that the difference between the ionisation potentials of 20Ne and 22Ne, which determines the energy difference of terms, is much smaller than that for helium.In addition, the hyperfine structure for ions is absent (its possible effect has not been discussed) because the nuclei of these isotopes have no magnetic moments. The magnitude of the isotope effect due to the mass difference only is estimated as*3%. Assuming, as in the case of helium, that upon collision of homonuclear species the potential of interaction between them is determined with equal probability by the terms u and 2P�¢g and that for every collision of heteronuclear species 2P�¢ the potential corresponds only to the binding term, we find that the ratio of the association rate constant k20 ¡À 22, estimated taking into account the mass difference and the nonidentity of the nuclei, to the rate constant kst found ignoring these effects is k20¡¦22 �� 2:06.kst The even greater magnitude of the kinetic isotope effect for the 3He74He mixture is due to the fact that the symmetry restrictions manifest themselves not only in the complex formation step but also in its stabilisation step. Let us note that the isotopes formed in the three-body association can then participate in fast bimolecular switching reactions of the type 4 3 He 4He++3He, He 4He++4He 20Ne 20Ne++22Ne . 20Ne 22Ne++20Ne There are no symmetry restrictions for these reactions, and they can rapidly result in an equilibrium isotope distribution in the molecular ions. Therefore, the isotope effect described above is observable only at relatively small degrees of conversion deter- mined by the ratio between the association and switching rate constants. This type of kinetic isotope effect has been also found experimentally in three-body reactions involving molecular spe- cies.The theoretical explanation of the effect in this case is much more complicated than in the case of systems involving atomic species. The enrichment in the 17O and 18O isotopes upon the formation of the O�¢4 ions in the reaction O�¢4 + O2 + M was studied.112 It was found that the isotope effect strongly depends on the energy of the ionising electrons used for the formation of the primary O�¢2 ions. The relative enrichment of the dimers in the labelled molecules (in comparison with the statistical enrichment) was found to be *10 at Ei close to the O2 ionisation threshold (12.077 eV).However, the enrichment at an energy of the electrons above 40 eV decreases to a virtually statistical value. 4 from O�¢2 ions formed In addition, the yield of ionic dimers O�¢ by electrons at Ei close to the ionisation threshold is much lower than that from ions obtained at higher electron energies. All this is explained 112 by the above-mentioned symmetry restriction upon the formation of a bound electronic state ofO�¢4 in a collision ofO�¢2 557 with O2 (3Pg). The restriction disappears if O�¢2 (or O2) is a heteronuclear species. 2)�¢2 formation in the CO�¢2 +CO2+ The efficiency of the (CO M reaction also increases as the energy of ionising electrons is increased.113 The symmetry-induced kinetic isotope effect was also studied for this process with the use of 13C and 18O labels.If the heavy oxygen label 18O is used, an increase in the ionising electron energy from 25 to 100 eV results in a gradual decrease in the heavy isotope enrichment factor from 27 to 8. For 12Cand 18O, the heavy isotope enrichment factor is *5 and virtually does not depend on the energy of the ionising electrons. It is of note that the kinetic isotope effect is also manifested in the clusterisation of ions (CO2)�¢n where one of the CO2 molecules is replaced by another species with a closed electron shell. For example, the isotope effect was observed in the formation of cluster ions in reactions of ArCO�¢2 complexes (see Ref.114). It was shown that the CO2 molecules containing heavy isotopes 17O and 18O are incorporated into the ArCO�¢2 clusters 20 times more efficiently than a statistical combinatorial estimate would predict. The studies of the symmetry-induced kinetic isotope effect in three-body ion ¡À molecule reactions have been reviewed.115 Let us discuss the results of the experiments at ultralow temperatures in a cooled gas-dynamic flow. Figure 6 presents the temperature dependence of the rate constant for the reaction (14) Kr++2Kr Kr�¢2 +Kr. In a wide range from moderate (*300 K) to low temperatures (*0.1 K), the rate constant cannot be described by the relation- ship (10) with an index n which does not depend on T.It was found that n=1.8 at low temperatures, while at moderate temperatures n is much smaller. The average value hni over the entire temper- ature range studied is equal to 1.2. It should be recalled that the theoretical n value for the association of monatomic species should 1072 1073 1074 1075 T /K 100.0 10.0 1.0 0.1 Figure 6. Temperature dependence of the rate constant for reaction (14).116 Similar deviations from the theoretical dependence are also observed for other reactions involving monatomic species: Ar�¢2 +Ar, n=1.4, hni=1.0; ArNe++Ne, n=2.4, hni=0.8; KrAr++Ar, n=1.8, hni=1.2; Xe�¢2 +Xe, n=2.3, hni=1.1; XeKr++Kr, Ar++2Ar Ar++2Ne Kr++2Ar Xe++2Xe Xe++2Kr n=2.3, hni=1.3.The values of n are given for T ^ 0.1 K. For the above process (12) involving molecular nitrogen in the same wide temperature range, the expression (10) at n=2 describes well the experimental temperature dependence of the rate constant (Fig. 7). It would be interesting to study analogous temperature dependences for n at low T for the formation of weakly bound complexes, for example, for the reaction (13) for which n is anomalously high at moderate temperatures.558 1025 k /cm6 s71 1071 1072 1073 1074 1075 5 10 50 100 300 T /K Figure 7. Temperature dependence of the rate constant for reaction (12).116 The points refer to the data fromdifferent authors. The radio frequency ion trap method with cooled walls was used to measure the rate constants for clusterisation and fragmen- tation Hái +2H2 Háiá2+H2 at 10 K for odd i in normal hydrogen n-H2 with an equilibrium ratio of the ortho- and para-isomers and in hydrogen enriched in the para-component, p-H2.92 The data obtained are presented in Fig.8. The difference between the curves for n-H2 and p-H2 is due to the rotational energy of the molecule with total nuclear momentum J=1 in the ortho-modification of molecular hydro- gen. The bond energy of the H2 molecule with the Hái cluster changes fromseveral kcal mol71 at small i to a value comparable with the heat of evaporation of liquid hydrogen (0.2 kcal mol71). 1025 k /cm6 s71 2 1071 1 1072 1073 21 23 19 21 17 19 15 17 13 15 11 13 9 11 79 57 35 i iá2 Figure 8.Dependence of the rate constant for the formation of odd hydrogen ionic clusters in the reaction Hái + 2H2 on the relative size change i/(i+2) of an ionic cluster Hái at lowtemperatures.92 The points in the curve 1 were obtained in an experiment with an equilibrium concen- tration of the para- and ortho-modifications in H2, those in the curve 2, in hydrogen enriched with the para-component. This example demonstrates a transition from clusters with sufficiently strong chemical bonds to clusters with van der Waals bonds with an increase in the number of species. Experiments also give information on the cluster structure.For example, a decrease in the association rate constant with an increase in the cluster size at i=9 is due to the beginning of formation of the second shell around the Há5 skeleton. The rate decrease at i>15 is due to a decrease in the efficiency of stabilisation by the third body.41, 73 The effect of the particular features of the complex structure on the association rate constant is demonstrated in Fig. 9, which GV Karachevtsev (deceased), P S Vinogradov 1025 k3 /cm6 s71 1072 1072 1073 1074 1075 0 4 6 10 n Figure 9. Dependence of the rate constant for reaction (15) on the number of atoms in an ionic copper cluster.73 The points refer to the experiment; the curve gives the calculation in accordance with the RRKM theory under the assumption of a `soft' transition state.The dotted line connects the calculated points obtained for a `hard' transition state for n=1 and 2. shows the experimental dependence of the association rate constant for copper ionic clusters Cuán with COmolecules Cuán +CO+He Cuán CO+He (15) on the magnitude of n.73 On going fromthe Cu+ ion to its dimer Cuá2 , the rate constant increases by about an order of magnitude. However, the rate constant decreases somewhat on going from Cuá2 to Cuá3 . This rather unexpected result is related to the specific features of the trimer electronic structure, which are unknown.{ The passage from Cuá3 to Cuá4 again results in an increase in the reaction rate constant by an order of magnitude.Further increase in the cluster size results in a slower increase in the rate constant, which approaches a steady level at n=7. The agreement between the calculated and experimental points for the rate constants of the reactions of the Cun+ ions with n53 is good, but the experimental points for Cu+ and Cu2+ `fall out'. If another type of transition states (so-called hard ones) are assumed for the Cu+CO and Cuá2 CO complexes, the agree- ment for these points with the calculated data is improved considerably. The rates of three-body association processes usually decrease with an increase in the temperature and translational energy of ions. There is only a single publication 117 where the effective rate constant for the reaction C6Há6 +2C6H6 C12Há12+C6H6 is reported to be independent of the translational energy of the ions over a wide energy range.However, in this case it cannot be ruled out that the C12Há12 ions are products of associative ionisation of highly excited benzene molecules 118 C6H 6 +C6H6 C12Há12+e; the rate of this process is not affected by electrical fields that accelerate the ions.117 However, some data indicate that the efficiency of association in certain systems is increased with an increase in the translational energy of the colliding species. For example, it was found that collisions of accelerated fullerene ions with inert gas atoms produce long-lived excited complexes with a noble gas atom inside the molecular cage of the fullerene.119, 120 These are thresh- oldprocesses that occur only at sufficiently highcollision energies.{A semiempirical calculation by one of the authors suggests that for Cuá2 COthe almost T-shapedstructure of the (Cu7CO7Cu)+complex in which the carbon atomis arranged between the two copper atoms has the lowest energy.Three-body ion ± molecule processes The formation of long-lived excited complexes in bimolecular processes also occurs in collisions of ions of large biological molecules with small molecules like NH3 and CH4.118 Probably, association is possible in these systems, and its efficiency increases with an increase in the translational energy of ions. Of the three-body ion ± molecule processes, the association processes have been studied most thoroughly to date.However, three-body exchange reactions, which have been little studied thus far, are of considerable interest. There are reasons to believe that at higher gas pressures three-body reactions resulting in the same products as bimolecular ones C++D+M A++B+M can compete successfully with some slow exothermic bimolecular reactions of the type C+ + D . A++B The reactions SF3O++2HF+M, SFá5 +H2O+M (rate constant 3610727 cm6 s71) and SFá5 +H2O SF3 O++2HF, with a rate constant less than 10711 cm3 s71 can serve as examples.14, 121 The detailed mechanisms of these processes have not been sufficiently studied yet. Studies of the dependence of the reaction rate on pressure 122, 123 and temperature 124 revealed an abnormal behaviour of the [SF5 + H2O] complex.This complex can probably exist in different configurations with considerably different reactivities. VI. Conclusion Intense studies on ion ± molecule reactions in gases have already been continuing for almost half a century after V L Tal'roze discovered these reactions in mixtures of hydrocarbons in the early 1950's. About 15 years later, intense studies of three-body reactions began in parallel with bimolecular reactions; the contribution of the former becomes important upon increasing the pressure in the reaction region. At present, there is a rather extensive experimental data base on the rate constants for such reactions; however, its bulk is about 30 times smaller than that of a similar base for bimolecular ion ± molecule reactions.After the discovery of ion ± molecule reactions, it was consid- ered that all of them are fast if this is allowed by their thermo- chemistry. The main reason in favour of this was the assumption that thtivation barriers appearing in the rearrangement of species (characteristic of reactions between neutral species) are already inside the `potential well' arising due to strong electro- static ion-dipole or polarisation attraction. In addition, it was shown that the majority of ion ± molecule reactions result in an intermediate long-lived complex, and this mechanism is more typical than the `direct' mechanisms. It was thus concluded that exothermic ion ± molecule reac- tions are a specific type of process different from similar reactions between neutral species.However, many exceptions from this concept were found later. The presence of energy barriers and the effect of various types of internal excitation on the reaction rates were observed. In view of this, it was concluded that there is no fundamental difference between the mechanisms of ion ± molecule reactions and reactions of neutral species. In fact, the energy barrier exists in the direction of the reaction coordinate, while the potential well originating from electrostatic attraction between an ion and a molecule corresponds to another coordinate describing the distance between the species. The effect of the electrostatic attraction can be demonstrated in a simplified form as a decrease in potential energy relative to the level of dissociation into the original species. The difference between the total energy and the energy barrier level (which, in rare cases, can 559 even be negative) for ion ± molecule system is most commonly different from that for neutral species.However, this is most likely a quantitative parameter, and the theoretical approaches for the consideration of ion ± molecule reactions do not differ fundamen- tally. It is of note that the role of the ion7molecule attraction should not be overestimated. The electrostatic interactions cannot be extrapolated for the short distance range. When the reacting partners approach each other, considerable redistribution of the effective charge in a system of colliding species occurs, and the character of interaction alters.However, owing to redistribution of the effective charge, which can occur even at rather long distances between the partners, the properties of the reagents (geometry, bond ener- gies, vibration frequencies) also change. This charge redistribu- tion is more typical of ion ± molecule collisions than of collisions involving neutral species (except for the systems with a weakly bonded electron). Activation is usually non-typical of ion ± molecule association reactions. However, specific effects related to the presence of charge in the system manifest themselves in full measure. This corresponds to a deeper global minimum of the potential energy surface for the interaction between the reagents, i.e., an increase in the dissociation energy.The potential well becomes much wider, which results in an increase in the number of internal energy levels. As a result, the rate constants for ion ± molecule association appear to be much higher than those for similar systems of neutral species. Upon this is further superimposed the effect of larger collision cross-sections due to long-range interaction. Despite the similarity of the theoretical approaches for three- body processes involving only neutral and ion-neutral species, the calculation details differ in the practical calculations, where it is necessary to use certain approximations. For ion ± molecule association processes, anisotropic effects are essential (an exam- ple of their calculation within the framework of the statistical theory was described above).Calculations of potential energy surfaces (which commonly have local minima in addition to global ones in ion ± molecule systems) and the corresponding trajectory calculations are also important. Rather recently, a quasi-classical method has appeared, viz., SACM (statistical adiabatic channel model), for the calculation of ion capture by molecules.125 Rather accurate algorithms of trajectory calculations for these processes 126 ± 129 and three-body collisions 130 have also been developed. However, a strict theoret- ical consideration of low-temperature experiments requires the development of purely quantum models.This work was financially supported by the INTAS (Grant No. 93-1128 IXT) and Russian Foundation for Basic Research (Project No. 96-02-1858A). References 1. R A Goldberg, A C Aiken J. Geophys. Res. 76 8352 (1971) 2. R S Narcisi, A D Bailey, L E Wlodyka, C R Philbrick J. Atmos. Terr. 34 647 (1972) 3. G Beig, S Walters, G Brasseur J. Geophys. Res. 98 767 (1993) 4. G Beig, S Walters, G Brasseur J. Geophys. Res. 98 775 (1993) 5. A F Harrison Chemical Ionisation Mass Spectrometry (Boca Raton, CA: CRC Press, 1983 6. A A Polyakova, I A Revel'skii, M I Tokarev, L O Kogan, V L Tal'roze Mass-Spektral'nyi Analiz Smesei s Primeneniem Ionno-Molekulyarnykh Reaktsii (Mass-Spectrum Analysis of Blends with Use of Ion ± Molecule Reactions) (Ed.A A Polyakova) (Moscow: Khimiya, 1989) 7. TWCarr (Ed.) Plasma Chromatography (New York: Plenum, 1984) 8. C B Collins, F W Lee J. Chem. Phys. 68 1391 (1978) 9. C B Collins, F W Lee J. Chem. Phys. 71 184 (1979) 10. In V Vsesoyuz. Konf. po Fizike Nizkotemperaturnoi Plazmy (Tez. Dokl.) [The Fifth All-Union Conference on the Physics of Low- Temperature Plasma (Abstracts of Reports)] (Kiev: Institute of560 Nuclear Investigation, Academy of Sciences, Ukr. SSR, 1979) Part 2, p. 315 11. G V Karachevtsev, IM Mazurin,A Z Marutkin,V V Savkin, V L Tal'roze Khim. Fiz. 136 (1983) a 12. P P Wickramanayake Int. J. Mass Spectrom. Ion Processes 69 39 (1986) 13. J A Stone,W J Wagtenberg Int. J. Mass Spectrom. Ion Processes 94 269 (1989) 14. I A Kaltashev,G V Karachevtsev,A Z Marutkin Khim.Vys.Energ. 24 489 (1990) b 15. H Bohringer Chem. Phys. Lett. 122 185 (1985) 16. S Matsuoka,H Nakamura J. Chem. Phys. 89 5663 (1988) 17. R J Gordon J. Chem. Phys. 74 1676 (1981) 18. J Chesnoy J. Chem. Phys. 79 2793 (1983) 19. G V Karachevtsev,A Z Marutkin,V V Savkin,V L Tal'roze Khim. Fiz. 798 (1983) a 20. M J Pilling Int. J. Chem. Kinet. 21 267 (1989) 21. V N Kondrat'ev,E E Nikitin Khimicheskie Protsessy v Gazakh (Chemical Processes in Gases) (Moscow: Nauka, 1981) 22. J P Robinson,K AHolbrook Unimolecular Reactions (New York: Interscience, 1972) 23. W Forst Theory of Unimolecular Reactions (New York: Academic Press, 1973) 24. N M Kuznetsov Kinetika Monomolekulyarnykh Reaktsii (Kinetics of Unimolecular Reactions) (Moscow: Nauka, 1982) 25.E E Nikitin Teoriya Elementarnykh Atomno-Molekulyarnykh Prot- sessov v Gazakh (Theory of Elementary Atomic and Molecular Processes in Gases) (Moscow: Khimiya, 1970) 26. I S Zaslonko Usp. Khim. 66 537 (1997) [Russ. Chem. Rev. 66 483 (1997)] 27. B M Smirnov Iony i Vozbuzhdennye Atomy v Plazme (Ions and Excited Atoms in Plasma) (Moscow: Atomizdat, 1974) 28. B Chatterjee,R Johnsen J. Chem. Phys. 93 5681 (1990) 29. B M Smirnov Cluster Ions and Van der Waals Molecules (New York: Gordon and Breach, 1992) 30. J E Russell, J S Shyu J. Chem. Phys. 91 1015 (1989) 31. M Hawley,T L Mazely, L K Randeniya,R S Smith,X K Zeng, M A Smith Int. J. Mass Spectrom. Ion Processes 97 55 (1990) 32. M A Smith,M Hawley, in Advances of Gas Phase Ion Chemistry Vol.1 (San Francisco: JAI Press, 1992) p. 167 33. M Hawley,M A Smith J. Chem. Phys. 96 326 (1992) 34. M A Smith, in Unimolecular and Bimolecular Reaction Dynamics (Eds CYNg, T Baer, I Powis) (New York: Willey, 1994) p. 183 35. G V Karachevtsev Khim. Vys. Energ. 24 195 (1990) b 36. J Sanderson,H Tanuma,N Kobayashi,Y Kaneko J. Phys. B, At. Mol. Opt. Phys. 26 L465 (1993) 37. N Saito,T M Kojima,N Kobayashi,Y Kaneko J. Chem. Phys. 100 5726 (1994) 38. J Sanderson,H Tanuma,N Kobayashi,Y Kaneko J. Phys. B., At. Mol. Opt. Phys. 27 L433 (1994) 39. M A Haney, J L Franklin J. Phys. Chem. 73 4328 (1969) 40. D R C Dougherty, Ming Xu, in Proceedings of the 45th ASMS Conference on Mass Spectrometry and Allied Topics, PalmSprings, CA, 1997 p.246 41. G V Karachevtsev,DM Manakov Khim. Fiz. 12 510 (1993) a 42. G A Kapralova,AM Chaikin Khim. Fiz. 10 807 (1991) a 43. G A Kapralova,T V Suchkova,A M Chaikin Khim. Fiz. 14 (5) 55 (1995) a 44. P M Langevin Ann. Chim. Phys. 5 245 (1905) 45. T Su,M T Bowers Classical Ion ± Molecule Collision Theory in Gas Phase Ion Chemistry (Ed.MTBowers) (NewYork: Academic Press, 1978) Vol. 1, p. 83 46. L Bass,T Su,M T Bowers Int. J. Mass Spectrom. Ion. Phys. 28 389 (1978) 47. MXu,T Solouki,A G Marshall,R C Dougherty, in Proceedings of the 45th ASMS Conference on Mass Spectrometry and Allied Topics, PalmSprings, CA, 1997 p. 258 48. T Su,W J Chesnavich J. Chem. Phys. 76 5183 (1982) 49. W J Chesnavich,M T Bowers J.Chem. Phys. 66 2306 (1977) 50. L MBass,W J Chesnavich,M T Bowers J. Am. Chem. Soc. 101 5493 (1979) 51. L M Bass, P R Kemper,V G Anicich,M T Bowers J. Am. Chem. Soc. 103 5283 (1981) GV Karachevtsev (deceased), P S Vinogradov 52. L M Bass,K R Jennings Int. J. Mass Spectrom. Ion Processes 58 307 (1984) 53. S A Maluendes,A D McLean,K Yamashita, E Herbst J. Chem. Phys. 99 2812 (1993) 54. D R Bates J. Chem. Phys. 84 6233 (1986) 55. D R Bates J. Chem. Phys. 89 192 (1988) 56. J Troe Z. Phys. Chem. 154 73 (1988) 57. S C Smith,M J McEwan,R G Gilbert J. Chem. Phys. 90 1630 (1989) 58. S C Smith,M J McEwan,R G Gilbert J. Chem. Phys. 90 4265 (1989) 59. R G Gilbert,S C Smith Theory of Unimolecular and Recombination Reactions (Oxford: Blachwell, 1990) 60.J D C Jones,D G Lister,D P Wareing,N D Twiddy J. Phys. B., At. Mol. Phys. 13 3247 (1980) 61. M W Siegel Science 260 95 (1993) 62. I A Kaltashev,G V Karachevtsev,A Z Marutkin Khim. Vys. Energ. 25 305 (1991) b 63. V L Tal'roze,E L Frankevich Zh. Fiz. Khim. 34 2709 (1960) c 64. L Operti,M Splendore,G A Valio,AM Franklin, J F Todd Int. J. Mass Spectrom. Ion. Processes 136 (1) 25 (1994) 65. P Kebarle J. Am. Soc. Mass Spectrom. 3 1 (1992) 66. U A Arifov, S L Pozharov, I G Chernov,Z A Mukhamediev Khim. Vys. Energ. 7 394 (1973) b 67. A P Babaev, I A Kaltashev,G V Karachevtsev,A Z Marutkin Prib. Tekh. Eksp. 200 (1990) 68. D Smith,N G Adams, in Gas Phase Ion Chemistry (Ed.MT Bowers) (New York: Academic Press, 1979) Vol. 1, p.2 69. S T Graul,R R Squires Mass Spectrom. Rev. 7 263 (1988) 70. E E Ferguson,E C Fehsenfeld,A L Schmeltekopf Adv. At. Mol. Phys. 5 1 (1969) 71. P S Vinogradov,O V Dmitriev Khim. Vys. Energ. 24 483 (1990) b 72. I Bohringer,F Arnold Int. J. Mass Spectrom. Ion Phys. 49 61 (1983) 73. R E Leuchtuer,A C Harms,AW Castleman Jr J. Chem. Phys. 92 6527 (1990) 74. L A Viehland, in Swarms of Ions and Electrons in Gases (EdsWLindinger, TDMark, F Howorka) (Wien: Springer, 1984) 75. W Lindinger Int. J. Mass Spectrom. Ion Processes 80 115 (1987) 76. N G Adams,D Smith Int. J. Mass Spectrom. Ion Processes 81 273 (1987) 77. J Glosõ k,V Skalsky',C Praxmarer,D Smith,W Freysinger,W Lin- dinger J. Chem. Phys. 101 3792 (1994) 78. P S Vinogradov,D M Borisenko,A Hansel, J Taucher, W Lindinger,R Flannery, in Proceedings of the 45th ASMS Con- ference on Mass Spectrometry and Allied Topics, PalmSprings, CA, 1997 p.266 79. P S Vinogradov,D M Borisenko,A Hansel, J Taucher, W Lindinger,R Flannery In Proceedings of the 44th ASMS Conference on Mass Spectrometry and Allied Topics, Portland, OR, 1996 p. 238 80. V ZakourÏ il, J GlosõÂk,V Skalsky',W Lindinger J. Phys.Chem. 99 15890 (1995). 81. B R Rowe, J-B Marquete Int. J. Mass Spectrom. Ion Processes 80 239 (1987) 82. B R Rowe, J B Marquett,C Rebrion J. Chem. Soc., Faraday Trans. 85 1631 (1989) 83. B R Rowe,A Canosa,V LePage Int. J. Mass Spectrom. Ion Processes 149/150 573 (1995) 84. T L Mezely,M A Smith J. Chem. Phys. 89 2048 (1988) 85. D Gerlich,T Rox Z.Phys.,D13 259 (1989) 86. M A Smith, in Unimolecular and Bimolecular Reaction Dynamics (Eds CYNg, T Baer, I Powis) (New York: Willey, 1994) 87. V L Talrose,G V Karachevtsev Adv. Mass Spectrom. 3 211 (1966) 88. M H Prior,R Marrus,C R Vane Phys. Rev. A, Gen. Phys., Ser. 3 28 141 (1983) 89. D Shen,H O Pritchard, J Todd Mol. Phys. 80 1135 (1994) 90. M V Gorshkov, S Ghan,A G Marshall Rapid Commun. Mass Spectrom. 6 166 (1992) 91. J P Honovich,G V Karachevtsev, E N Nikolaev Rapid Commun. Mass Spectrom. 6 429 (1992) 92. W Paul,B Lucke, S Schlemmer,D Gerlich Int. J. Mass Spectrom., Ion Processes 149/150 373 (1995) 93. D Gerlich Adv. Chem. Phys. 82 1 (1992) 94. R C Dunbar Int. J. Mass Spectrom. Ion Processes 100 423 (1990)561 Three-body ion ± molecule processes 95. D Gerlich, S Hornig Chem. Rev. 92 1509 (1992) 96. G V Karachevtsev Khim. Fiz. 13 67 (1994) a 97. A D Sen,W T Huntress Jr , V G Anicich J. Chem. Phys. 94 5462 (1991) 98. J Fisher, T McMahon Int. J. Mass Spectrom. Ion Processes 100 423 (1990) 99. A D Sen,W T Huntress Jr, V G Anicich, M J McEwan, A D Denison J. Chem. Phys. 94 5462 (1991) 100. S C Smith, P W Wilson, P Sudkeaw, R G A R Maclagan, M J McEwan, V G Anicich, W T Huntress Jr J. Chem. Phys. 98 1944 (1993) 101. V G Anicich, D Sen,W T Huntress Jr , M J McEwan J. Chem. Phys. 94 4189 (1991) 102. V Anicich, A D Sen,M J McEwan, S C Smith J. Chem. Phys. 100 5696 (1994) 103. L I Virin, R V Dzhagatspanyants, G V Karachevtsev, V K Potapov, V L Tal'roze Ionno-Molekulyarnye Reaktsii v Gaz- akh (Ion ± Molecule Reactions in Gases) (Moscow: Nauka, 1979) 104. Y Ikezoe, S Matsuoka,M Takebe, A Viggiano Gas Phase Ion ± Molecule Reactions Rate Constants Through 1986 (Tokyo: Ion Reaction Research Group of the Mass Spectroscopy Society of Japan, 1987) 105. R Patric, D M Golden J. Chem. Phys. 82 75 (1985) 106. L F Phillips J. Chem. Phys. 92 6523 (1990) 107. A A Viggiano J. Chem. Phys. 84 244 (1986) 108. R A Morris, A A Viggiano J. Phys. Chem. 98 3740 (1994) 109. R C Dunbar, J D Faulk Chem. Phys. Lett. 214 5 (1993) 110. E W Schlag,R D Levin Chem. Phys. Lett. 163 523 (1989) 111. G I Gellene J. Phys. Chem. 97 34 (1993) 112. K S Griffith, G I Gellene J. Chem. Phys. 96 4403 (1992) 113. R K Yoo, G I Gellene J. Chem. Phys. 102 3227 (1995) 114. R K Yoo, G I Gellene J. Chem. Phys. 105 177 (1996) 115. G I Gellene, in Advances in Gas Phase Ion Chemistry (Eds N G Adams, LMBabcock) (Greenwich, CT: JAI, 1996) Vol. 2, p. 4 116. L K Randenia, X K Zeng, R S Smith, M A Smith J. Phys. Chem. 93 8031 (1989) 117. V M Matyuk, I N Pobezhimova, V K Potapov Khim. Vys. Energ. 28 204 (1994) b 118. P S Vinogradov, D M Borisenko, O V Dmitriev, I N Veretennikov, in Contributions. Symposium on Atomic, Clusters 119. H J Cooper, P J Derrick, H Donald, B Jenkins, E Uggerud and Surface Physics (SASP) (Innsbruck, Austria: Institut fuÈ r Ionenphysik UniversitaÈ t 1994) p. 144 J. Phys. Chem. 97 5443 (1993) 120. D E Giblin, M L Gross, M Saunders, H Jimenez-Vazquez, R J Cross J. Am. Chem. Soc. 119 9883 (1997) 121. X Cheng, C Fenselau J. Am. Chem. Soc. 115 10327 (1993) 122. P S Vinogradov, O V Dmitriev, G V Karachevtsev, D M Borisenko, I N Veretennikov Khim. Vys. Energ. 28 92 (1994) b 123. P S Vinogradov, D M Borisenko, O V Dmitriev, G V Karachevtsev, I N Veretennikov, in Contributions. Sympo- sium on Atomic, Clusters and Surface Physics (SASP) (Innsbruck, Austria: Institut fuÈ r Ionenphysik UniversitaÈ t 1994) p. 233 124. G V Karachevtsev, A Z Marutkin, V V Savkin, V L Tal'roze Khim. Fiz. 3 695 (1984) 125. J Troe J. Chem. Phys. 105 6249 (1996) 126. A I Maergoz, E E Nikitin, J Troe, V G Ushakov J. Chem. Phys. 127. A I Maergoz, E E Nikitin, J Troe, V G Ushakov J. Chem. Phys. 128. A I Maergoz, E E Nikitin, J Troe, V G Ushakov J. Chem. Phys. 129. R T Pack, R B Walker, B K Kendrick J. Chem. Phys. 109 6701 105 6263 (1996) 105 6270 (1996) 105 6277 (1996) (1998) 130. A I Maergoz, E E Nikitin, J Troe, V G Ushakov J. Chem. Phys. 108 5265 (1998) a�Russ. J. Chem. Phys. (Engl. Transl.) b�High Energy Chem. (Engl. Transl.) c�Russ. J. Phys. Chem. (Engl. Tr
ISSN:0036-021X
出版商:RSC
年代:1999
数据来源: RSC
|
3. |
High-resolution solid-state NMR spectroscopy in studies of conversions of hydrocarbons and alcohols on zeolites |
|
Russian Chemical Reviews,
Volume 68,
Issue 7,
1999,
Page 563-580
Alexander G. Stepanov,
Preview
|
|
摘要:
Russian Chemical Reviews 68 (7) 563 ± 580 (1999) High-resolution solid-state NMR spectroscopy in studies of conversions of hydrocarbons and alcohols on zeolites { A G Stepanov Contents I. Introduction II. Alkene oligomerisation III. Dehydration of alcohols IV. Reactions of alkenes and alcohols with carbon monoxide and acetonitrile V. Analysis of hydrocarbon products and intermediates formed from alkene oligomers VI. Activation of alkanes at low temperatures. Isotope (H/D) exchange between alkanes and acidic OH groups of zeolite VII. Activation of alkanes at low temperatures. Carbonylation VIII. The nature of intermediates in conversions of alkanes, alkenes and alcohols IX. Conclusion Abstract. The review surveys advances in high-resolution solid- state 1H and 13C NMR spectroscopy applied to studies of conversions of hydrocarbons and alcohols on zeolite catalysts of an acidic nature.The potential of NMR spectroscopy in studies of mechanisms of chemical reactions and analysis of compounds formed in situ is considered. The bibliography includes 134 references. I. Introduction NMR spectroscopy has been used for studying solid heteroge- neous catalysts and surface chemical reactions over several decades.1 However, impressive progress in this field has been achieved only with the advent of special techniques for recording NMR spectra and with the development of superconducting magnets producing strong magnetic fields (7 ± 18 T). Thus the application of the magic-angle-spinning (MAS) technique in which the sample is rotated rapidly about an axis subtending a magic angle (57 8480800) 2±4 with respect to the external magnetic field (2 ± 35 kHz) leads to substantial narrowing of NMR signals for nuclei of both solid catalysts (1H, 17O, 27Al, 29Si and 51V) and organic molecules adsorbed on the catalyst surface (1H, 13C, 15N and 31P).The implementation of a technique for recording NMR spectra with the use of cross-polarisation simultaneously with high-power proton decoupling 5 ±7 enhances the sensitivity of NMR spectroscopy a hundred times and makes possible observa- tions of signals of molecules adsorbed on the catalyst surface. Presently, high-resolution solid-state NMR spectroscopy is widely used for both characterising structures of heterogeneous catalysts 1, 8±13 and studying conversions of organic molecules on their surfaces.13 ± 17 In many cases, NMR spectroscopy appears to be the unique physical method which provides data on reaction mechanisms and enables one to perform reliable analysis of A G Stepanov Boreskov Institute of Catalysis, Siberian Branch of the Russian Academy of Sciences, prosp. Akademika Lavrent'eva 5, 630090 Novosibirsk, Russian Federation.Fax (7-383) 234 30 56. Tel. (7-383) 239 73 50. E-mail a.g.stepanov@catalysis.nsk.su Received 23 October 1998 Uspekhi Khimii 67 (7) 619 ± 637 (1998); translated by T Safonova #1999 Russian Academy of Sciences and Turpion Ltd UDC 543.422.25 : 542.97 563 564 567 570 572 575 577 578 578 compounds formed in situ, to determine their compositions and structures and to make conclusions about chemical processes that occur on the catalyst surface.Hence, one can go from the hypothetical notion of reaction mechanisms to the understanding of their true mechanisms. NMRspectroscopy is particularly helpful in studying catalytic conversions at low (as regards traditional organic catalysis) temperatures, for example, in studying low-temperature alkene conversions on zeolites. This field of potentially important low- temperature chemical conversions remains poorly studied due to problems associated with analysis of undesorbable compounds. Recently, studies of conversions of hydrocarbons and alcohols on zeolites by NMR spectroscopy have attracted growing interest in connection with the solution of the following practical problems: �a search for ways of enhancing the selectivity of the yields of C4±C5 alkenes in the course of catalytic cracking based on zeolite catalysts (HY, USY and HZSM-5); � synthesis of a motor fuel from methanol on HZSM-5 zeolite.18 Discussion about the role of various intermediates, viz., alkoxides and carbenium ions, in catalytic hydrocarbon conver- sions on solid acids (zeolites and amorphous aluminosilicates), which has lasted for the last decade, gave additional impetus to the application of NMR spectroscopy to the studies of hydrocarbon conversions.The present review surveys the most interesting results of studies of conversions of various hydrocarbons on solid zeolite catalysts of acidic nature which have been performed in the leading world's laboratories with the use of NMR spectroscopy.The first results of studies of hydrocarbon (alkene) conversions are compared with the results obtained by various research groups within the last five years. The possibilities of other methods (primarily, of IR spectroscopy) in studies of hydrocarbon con- versions are considered. In the present review, the discussion is restricted to the results of chemical studies { obtained by NMR spectroscopy. { Dedicated to the memory of Academician K I Zamaraev. { Experimental details and discussion of various possibilities of the use of high-resolution solid-state NMR spectroscopy for characterisation of catalysts and processes that occur on their surfaces have been reported in the literature.10 ± 13, 16, 19, 20564 II.Alkene oligomerisation 1. Adsorbed products of low-temperature alkene oligomerisation Until recently, the nature of hydrocarbon products which were formed upon low-temperature oligomerisation of alkenes (ethyl- ene, propene or isobutene) on acidic forms of zeolites (for example, on the HZSM-5 zeolite), remained unclear and was the subject of discussion.21 ± 38 It was reasonable to suggest that alkene oligomerisation in the presence of an acidic catalyst should afford a long-chain alkene. However, signals at 1660 ± 1670 [n(C=C)] and 3020 ± 3090 cm71 [n(=C7H)] expected for the double bonds 39 were not observed in the IR spectra of adsorbed oligomeric products.21 ± 24, 27 ± 29, 31, 33 Signals with the chemical shifts at d 110 ± 140 typical of alkene double bonds 40 were absent in the 13C NMR spectra as well.24 ± 26, 30, 32, 34 ± 38 The only identified signals were those of saturated hydrocarbon fragments of oligomeric products.21 ± 38 The absence of signals of alkene fragments in the IR and NMR spectra has been explained within the frameworks of two alternative hypotheses.According to the first hypothesis, adsorbed oligomers are hydrocarbon fragments covalently bonded to oxygen atoms of the zeolite lattice, i.e., the hypothesis postulates that oligomeric alkoxides (alkyl silyl ethers), which do not contain double bonds, are formed.32 ± 34, 38 According to the second hypothesis, adsorbed oligomers are oligomeric aliphatic carbenium ions,27, 28, 37 which do not contain double bonds as well.Unfortunately, both hypotheses are self-contradictory. Actually, if oligomers were alkoxides, the 13C NMR spectra should have signals for the carbon atoms of the C±O± Si fragment at d 70 ± 90.40, 41 However, in a series of stud- ies,24 ± 26, 30, 32, 34 ± 38 no mention of these signals was presented, while in other studies 15, 35, 38, 42 ± 48 these signals were observed only in the case of small hydrocarbon fragments (Me, Pri, But and Bui) bound to the zeolite framework. The same is true for carbenium ions containing long-chain hydrocarbon fragments for which signals at d 300 ± 330 (a carbon atom that carries a positive charge) 49 ± 51 in the 13C NMR spectra or bands at 1260 ± 1295 cm71 [n(+C± C) stretching vibration] 52, 53 in the IR spectra would be expected.However, neither the IR spectra nor the NMR spectra had these signals. a. Alkoxide nature of the adsorbed oligomers The 13C CP/MAS NMR spectrum } of products of ethylene oligomerisation on the HZSM-5 zeolite at 296 K is shown in Fig. 1.54 On the whole, this spectrum is identical with that obtained previously by van den Berg and co-workers ethylene oligomers.30 The signals at d 14, 22 and 33 were assigned to the terminal CH3 groups, to the adjacent CH2 groups and to internal CH2 groups of linear hydrocarbon fragments of oligomers, respectively. The spectrum has no signals with noticeable intensities at d 110 ± 140 (for alkene fragments) 40 or at d 300 ± 330 (for carbenium ions).49 ± 51 However, substantial enhancement of the spectrum made it possible to detect a weak broad signal at d 89, which indicates that adsorbed oligomers contain carbon atoms bound to oxygen atoms of the zeolite framework, i.e., is indicative of the presence of C±O± Si fragments.41 ± 45 An analogous signal for C±O± Si fragments at d 86, while substantially more intense, was observed for isobutene oligomers formed upon dehydration } The 13C CP/MAS NMR spectrum was recorded with the use of the magic-angle-spinning (MAS) technique at the spinning frequency of 2500 ± 5000 Hz using cross-polarisation (CP).A G Stepanov 89 * (32) * * 50 0 750 d /ppm 300 250 200 150 100 Figure 1.The 13C CP/MASNMRspectrumof oligomerisation products of ethylene (40%-13CH2=CH2) on HZSM-5 zeolite at 296 K. Hereinafter, peaks due to the rotation of the sample are marked with asterisks. of tert-butyl alcohol on HZSM-5 zeolite.35, 38 Therefore, evidence for the alkoxide nature of the adsorbed oligomers was obtained.} b. Carbenium-ion properties of the adsorbed oligomers An interesting experiment, which gave a deeper insight into the nature of the adsorbed oligomers, was described in Ref. 37. Studies of a possible ethylene oligomerisation product, viz., oct- 1-ene containing the selectively 13C-labelled terminal =CH2 group, adsorbed on the HZSM-5 zeolite demonstrated that the terminal =13CH2 group of oct-1-ene was converted into the 13C- labelled terminal methyl group immediately after adsorption (Fig.2). Then the intensity of the signal for the carbon atom of the terminal 13CH3 group at d 14.3 decreased and the intensities of the signals of internal CH2 groups of oct-1-ene at d 30.0 increased with time. This is indicative of the transfer of the selective 13C label from the methyl group of oct-1-ene to its CH2 groups with the characteristic half-reaction time t1/2 *4 h (at 290 K). The redistribution of the carbon label in the adsorbed oct-1-ene was completed in t > 20 h. It is known that the redistribution of b a e c d 14.3 33.0 14.3 33.0 * * * * 100 100 100 0 0 100 0 0 0 d /ppm 100 Figure 2. Change in the 13C CP/MAS NMR spectrum of oct-1-ene (containing the 13C-labelled terminal =CH2 group) adsorbed on HZSM-5 zeolite (T=290 K) with time.37 Time /h: 0.17 (a), 2.83 (b), 6.17 (c), 12.83 (d), 53 (e).} An explanation of the reasons for the low intensity of the signal for the carbon atom of the COSi fragment in adsorbed ethylene oligomers is offered below, where the interpretation of the spectra is discussed.High-resolution solid-state NMR spectroscopy in studies of conversions of hydrocarbons and alcohols on zeolites carbon labels is typical only of carbenium ions existing as stable species in solutions of superacids.49 ± 51, 55 ± 57 Based on the afore- said, it was concluded that carbenium ions were formed upon adsorption of oct-1-ene on the HZSM-5 zeolite. The changes observed in the NMR spectrum are attributable to the following reactions (Scheme 1): Scheme 1 + H H313C CH(CH2)5CH3+ H213C 7 O CH(CH2)5CH3 , O Si Al Si Al + CH2 H313C CH(CH2)5CH3 H313C +CH(CH2)4CH3 13CH3 H313C + HC CH(CH2)3CH3 C(CH2)4CH3 + H3C CH3 + CH3CH213CH2CH(CH2)3CH3 .In the adsorption of oct-1-ene on zeolite, the proton of the acidic OH group of the zeolite is transferred to the oct-1-ene molecule. In this case, the terminal=CH2 group is converted into the terminal CH3 group simultaneously with the formation of a carbenium ion. In the cation formed, the 13C label is redistributed through the formation of protonated cyclopropane intermedi- ates,49, 50, 55 ± 57 which leads to a decrease in the intensity of the signal for theCH3 group and to an increase in the intensities of the signals for the internal CH2 groups with time (Fig.2). In spite of the label redistribution typical only of carbenium ions, signals at d 300 ± 330, which are indicative of the presence of carbenium ions in the system, were not observed in the spectrum. It was thus concluded that either the concentration of the cations formed was too low or these cations were short-lived transient species. Later, the conclusion that carbenium ions exist in small amounts among adsorbed oligomers was confirmed by the chemical reaction at room temperature involving the adsorbed oligomers, which is characteristic of carbenium ions. It is known that carbenium ions readily react with carbon monoxide in solutions of strong acids 58, 59 to form carboxylic acids.Consequently, carbenium ions formed among oligomeric prod- ucts would be expected to react with CO and water to form carboxylic acids. Actually, simultaneous adsorption of CO and water on a zeolite specimen containing ethylene oligomers resulted in conversion of oligomers into a mixture of the corresponding oligomeric carboxylic acids (Scheme 2), which is evidenced by the signal of carboxyl groups at d 189 in the 13C CP/MAS NMR spectrum (Fig. 3).60, 61 Scheme 2 + CO+H2O CH3CH(CH2)nCH3 CH3CH(CH2)nCH3 7 O O Si Al Si Al CH3CH(CH2)nCH3+ H C O O HO Al Si The formation of carboxylic acids and the redistribution of the selective 13C label unambiguously suggest that carbenium ions were present among the adsorbed oligomers. In addition, after completion of the reaction of oligomers with CO andH2O, signals for oligomeric alkoxides at d 89 were no longer observed in the 13C NMR spectrum (cf.Figs 1 and 3), which is indicative of the conversion of alkoxides into acids. These results altogether allowed the conclusion that alkoxides existed in the equilibrium with oligomeric carbenium ions. Though the concentrations of the 565 COOH 189 * * * ** 200 0 100 300 d /ppm Figure 3. The 13C CP/MAS NMR spectrum of products formed after additional adsorption of 13CO + H2O on zeolite containing ethylene oligomers. latter are too low to be detected by 13C CP/MAS NMR spectro- scopy, the reactivities of the oligomers are determined by the reactivities of the equilibrium carbenium ions.Taking into account that conversion of C2±C4 alkenes in a flow reactor at T <450 K afforded C6±C18 alkenes,62, 63 i.e., oligomeric products can be desorbed as alkenes under particular conditions, { it was concluded that alkenes were present in small amounts among the adsorbed oligomeric products. However, the concentrations of alkenes were so low that these compounds, like carbenium ions, were `invisible' to NMR spectroscopy. Hence, the NMR spectral data allow the conclusion that oligomeric products of low-temperature alkene conversions, which are adsorbed on acidic forms of zeolites, exist as three interconvertible forms, viz., as alkoxides, carbenium ions and alkenes (Scheme 3).Scheme 3 H2C CH(CH2)nCH3 CH(CH2)nCH3 CH3 H O Si Al O Si Al + CH3 CH(CH2)nCH3 O7 Si Al Oligomeric alkoxides present the major adsorbed form. 2. Compositions and structures of products of alkene oligomerisation on HZSM-5 zeolite In the first studies devoted to oligomerisation of alkenes on HZSM-5 zeolite, qualitative conclusions about the formation of linear oligomers 25, 26, 30, 34 which contain approximately 16 car- bon atoms from alkenes (ethylene, propene or isobutene) were made based on the analysis of the positions of the signals for CH3 and CH2 groups in the 13C MAS NMR spectra.30 However, signals for the CH2 groups of linear fragments and signals for the CH3 groups of branched fragments of oligomers are observed in the same region of the 13C NMR spectrum (at d 20 ± 30).40 Consequently, one cannot infer the structures and lengths of hydrocarbon chains of oligomeric products exclusively from the chemical shifts of signals in the 13C NMR spectra.It was suggested that 2D J-resolved 13C MAS NMR spectro- scopy followed by GC-mass spectrometric analysis and tradi- tional high-resolution 13C NMR spectroscopy (in solution) of oligomers which have been extracted from zeolite, be used for establishing the compositions and structures of oligomeric prod- ucts formed from ethylene on HZSM-5 zeolite.54 { Signals of double bonds are also absent in the NMR spectrum of the adsorbed oligomer.566 a.2D J-resolved 13C MAS NMR spectroscopic data Two-dimensional J-resolved MAS NMR spectroscopy offers advantages over 1D NMR spectroscopy because not only does it provide data on the chemical shifts of signals for carbon atoms of oligomers but it also allows the determination of the multiplicity of the signals that occurs due to spin-spin coupling between the 13C nuclei and the adjacent protons [the spin-spin coupling constants J(13C±1H)].64, 65 In the 2D NMR spectrum, the assign- ment of the signal with a particular F2-dimension chemical shift to the CHn group is made based on the calculation of the number of peaks of this signal on the F1 axis.36, 38, 64 ± 66 As a result, the ambiguity of the assignment of the signals for carbon atoms is excluded and the true structure of oligomers in the adsorbed form can be determined.The contour plot of the 2D J-resolved 13C MAS NMR spectrum of ethylene oligomers on the HZSM-5 zeolite and the corresponding 1DNMRspectrum are shown in Fig. 4 b. It can be seen that the signals at d 30 and d 34 exist as triplets with the constant J(13C±1H)=140 16 Hz, which suggests that these signals belong to theCH2 groups of oligomers. The signal at d 22 is a singlet characteristic of the quaternary carbon atom. However, it is observed in the region typical of either the terminal CH3 groups of branched oligomers (in this case, the signal should exists as a quartet on the F1 axis 66) or the CH2 group (a triplet on the F1 axis 66) adjacent to the terminal CH3 group of the linear fragment of the hydrocarbon chain rather than being located in the region of chemical shifts of the quaternary carbon atom of alkanes (at d 30 ± 40 40).The existence of a singlet rather than a doublet at d 22 suggests that this signal is in fact a triplet for which only the most intense central component is distinct in the spectrum against a background of the noise level.35, 67 Hence, it was concluded that the signal at d 22 belongs to the CH2 groups of oligomers. The existence of four peaks on the F1 axis for the signal at d 14 indicates that the latter belongs to the terminal CH3 groups of the oligomer. To summarise, analysis of the multiplicities of signals of the adsorbed oligomers allowed the conclusion that signals in the ~~ 14.3 20 d /ppm 24 15 13 22 Figure 4.Spin-echo 1D 13C MAS NMR spectrum (a) and the contour plot of the 2D J-resolved 13C MAS NMR spectrum (b) of oligomerisation products of ethylene on HZSM-5 zeolite at 296 K. The observed value of the scalar spin-spin coupling constant J(13C±1H) on the F1 axis is half as large as the real value J(13C±1H) due to the use of a special technique for recording the 2D spectrum.65 ~~ a ~~~~ 34.229.9 22.4 10 30 40 0 710 b F1 7100 0100 Hz F2 d /ppm 35 33 30 28 ~~~~ A G Stepanov region from d 22 to d 34 belong to the CH2 groups, i.e., the oligomers formed are linear hydrocarbon fragments. b. High-resolution solution 13C NMR spectroscopic data To confirm the conclusions about the structures of oligomers formed from ethylene, a sample of HZSM-5 zeolite with the adsorbed oligomers was dissolved in 10% NaOH61 and the liberated hydrocarbon products were extracted with diethyl ether, dissolved in benzene-d6 and analysed by high-resolution 13C NMR spectroscopy. The major characteristics (chemical shifts, the ratio of the intensities of the signals and the multi- plicities of the major, most intense signals) of the spectrum of the products desorbed from the zeolite (Fig.5) are identical with those of the adsorbed oligomers (see Fig. 1). The most intense signals at d 22 and 33 ± 34 exist as triplets with J(13C±1H)=125 Hz and the signals at d 10 ± 14 exist as quartets with J(13C±1H)=125 Hz.CH2 (32) CH3 CH2 64 68 72 76 40 60 0 20 80 d /ppm Figure 5. The high-resolution 13C NMR spectrum (in benzene-d6) of ethylene oligomers extracted from HZSM-5 zeolite after dissolution of the zeolite in NaOH solution. In the spectra of the extracted products, the intensities of the signals at d 60 ± 80 are as small as those of the adsorbed oligomers. The fact that the ratio of the signals of the adsorbed products is identical with that of the extracted products as well as the presence of signals for the carbon atoms of the CHx ±OH fragment of alcohols indicate that the oligomers are extracted from the zeolite as alcohols and the structures of the extracted alcohols are analogous to those of the adsorbed alkoxide oligomers. The estimation of the ratio of the signals at d 14, 22 and 33 (Fig.5) demonstrated that an average linear oligomeric chain consists of seven carbon atoms [ ± (CH2)6CH3] bound to an oxygen atom of the zeolite framework. From the spectrum shown in Fig. 5 it follows that not only linear but also branched oligomers are formed on the HZSM-5 zeolite, which is evidenced by the signals at d 35 ± 50 belonging to secondary and tertiary carbon atoms.30, 68 However, linear oligomers predominate apparently due to the specific structure and the size of the channels in the HZSM-5 zeolite.69 The fact that the 13CNMRspectrum (see Fig. 5) has more than 20 signals from alcohols at d 60 ± 80 indicates that a mixture of no less than 20 oligomeric alkoxides containing predominantly linear fragments is formed upon adsorption of ethylene on the HZSM-5 zeolite.For a mixture of adsorbed oligomers, only one broad low- intensity signal at d 89 (see Fig. 1) is observed, which is indicative of the alkoxide nature of the adsorbed oligomers.54 Previ- ously,30, 32, 34, 70, 71 attempts to identify this signal have failed for several reasons. First, the adsorbed oligomers exist as a mixture of oligomeric alkoxides (>50, according to GC-mass spectrometric analysis 53). Second, the signals for oligomeric alkoxides are broadHigh-resolution solid-state NMR spectroscopy in studies of conversions of hydrocarbons and alcohols on zeolites but their intensities are low. Third, chemical shifts of the signals vary within the range of >15 ppm.In some cases, signals at d 87 ± 89 observed for various alkenes (Pri and But) 35, 45, 48, 72 were assigned to small hydrocarbon fragments bound to the oxygen atoms of the zeolite framework. However, the fact that the chemical shifts of the signals for alkoxides observed upon propene oligomerisation on HY zeo- lite 45, 73 or on HZSM-11 zeolite 48 are identical with those observed upon isobutene oligomerisation on HZSM-5 zeolite 35 allows the conclusion that the signals at d 87 ± 89 (see Refs 35, 45 and 48) should be assigned to oligomeric alkoxides rather than to short hydrocarbon fragments. c. Data of GC-mass spectrometric analysis The compositions of ethylene oligomers formed were also deter- mined by GC-mass spectrometric analysis.To do this, the adsorbed oligomers were first converted into a mixture of the corresponding carboxylic acids according to Scheme 2. Then, carboxylic acids were extracted from the zeolite by dissolving the latter in 10% alkali. Subsequent treatment of the mixture of acids with diazomethane 61 made it possible to convert them into the corresponding methyl carboxylates, which were readily separated and analysed by GC-mass spectrometry. It was found that the mixture extracted was comprised of more than 50 different oligomeric products (C5±C14). III. Dehydration of alcohols 1. Methyl alcohol NMR studies of alcohol conversions on zeolites attracted interest immediately after `Mobil' had discovered conversions of meth- anol into a motor fuel on HZSM-5 zeolite.74, 75 The first NMR spectral study devoted to conversions of methanol 76 had already demonstrated the possibility of the application of 13C MASNMR spectroscopy in the analysis of intermediates and reaction prod- ucts.Evidence was obtained for the conversion of methanol first into dimethyl ether and then into a mixture of hydrocarbons via intermediate carbenium ions rather than carbenes, as had been suggested previously by Chang and Silvestri.18 Anderson and Klinowski 78, 79 noted the difference in the composition of the products which were present in the gas phase and in zeolite pores. Thus in no case were C10 aromatic products observed among the desorbed products, but these compounds were found in situ in zeolite pores.It was thus concluded that products were formed within zeolite pores under conditions of the equilibrium reaction, the composition of the desorbed products being determined by the pore size of the zeolite. Hence, the effect of the internal specific size and the pore structure on the composition of the products desorbed was experimentally demonstrated. In the study devoted to methanol conversions,79 it was noted that ethylene was formed as an important intermediate. It was also reported that the reaction performed under conditions similar to those used in a flow reactor (small contact time) yielded inter- mediate C4 alkenes and cyclopentenyl cations (CPC).80 Below is given the scheme of conversions of methanol into a mixture of hydrocarbons on the HZSM-5 zeolite.H2C CH2 CH3OCH3 CH3OH alkanes ethylene oligomers aromatic compounds polyenes (stabilised as CPC) 2. tert-Butyl alcohol Conversions of aliphatic C3±C4 alcohols on acidic zeolites were studied with the aim of obtaining more detailed information on the mechanism of dehydration. The molecular sizes of these alcohols are comparable with the pore sizes of zeolites. 567 Dehydration of C3±C4 alcohols starts at substantially lower temperature (300 ± 400 K) than that of methanol (550 ± 700 K). This enables one to obtain more detailed information on the nature of intermediates based on the data of NMR spectroscopy and, consequently, to draw conclusions about the mechanism of these reactions.In addition, it becomes possible to reveal the effect of the pore size of a zeolite on the course of dehydration of isomeric alcohols and to analyse subsequent conversions of the products formed upon primary dehydration. In the present review, the results of studies of dehydration of isomeric butyl alcohols are considered in detail. Dehydration of tert-butyl alcohol on HZSM-5 zeolite occurs at room temperature.81 NMR spectral studies of this reaction, which is not complicated by secondary processes of cracking of the products formed, are of considerable interest for the understand- ing of the mechanisms of dehydration and subsequent oligomer- isation of isobutene formed upon dehydration.35, 38, 46, 70 It was found 46 that the dehydration of ButOH containing the 13C-labelled tertiary carbon atom on HZSM-5 zeolite (Si :Al = 36) was accompanied by complete redistribution of the labels in the oligomeric products.It was thus concluded that carbenium- ion intermediates were involved in dehydration of the alcohol and that tert-butyl silyl ether (TBSE) was formed as a stable inter- mediate, i.e., the tert-butyl fragment was covalently bound to the oxygen atom of the zeolite framework. The authors 46 believed that the formation of TBSE is evidenced by the signal at d 77 corresponding to the carbon atom bound to the oxygen atom of the zeolite { as well as by the unusually large chemical shift of the carbon atom bound to the oxygen atom of the zeolite framework compared to that of the carbon atom in the initial ButOH, which is generally observed in neat alcohol or in neutral solutions.More recent studies 33, 35 demonstrated that the conclusion about the formation of TBSE as a stable intermediate 46, 70 was erroneous. A special NMR spectral study 35 showed that dehy- dration of ButOHon an HZSM-5 zeolite with Si :Al=24 at room temperature proceeded slowly and was completed in>15 h (Figs 6 and 7). With an HZSM-5 zeolite in which Si :Al = 36, i.e., the zeolite in which the concentration of reactive acidic OH groups is 1.5 times lower, the dehydration time could only increase. Hence, it was concluded that dehydration of alcohols cannot be com- pleted in 1.5 h (the reaction time in the study 46) and, consequently, the signal at d 77 can be assigned only to the initial unconsumed alcohol.35 The difference between the chemical shift of the carbon atom of the C±Ofragment in the adsorbed alcohol (at d 82 35, 70 or at d 77 46) and that of this carbon atom in the initial solution (at d 68.97 40) should be attributed to specific interactions between the alcohol and the centres of its adsorption (Brùnsted acidic OH groups) as well as to a possible change in the geometry of the alcohol upon its adsorption in zeolite channels.82 IR spectral studies 33 demonstrated that the rates of oligomer- isation and isomerisation of isobutene liberated upon dehydration are substantially higher than that of its dehydration. Hence, the products observed in theNMRspectrum are isobutene oligomers.The identity of the spectra of isobutene oligomers prepared from ButOH containing selectively 13C-labelled CH3 or C±OH groups suggests the complete redistribution of the selective 13C labels (cf. Figs 6 and 7) and confirms the conclusions made by Aronson et al.46 that carbenium-ion intermediates are involved in dehydration and oligomerisation of ButOH. The formation of the tert-butyl carbenium ion (TBCI) and isobutene as intermediates, which exist in equilibrium with the initial alcohol,35 is evidenced by the experimental data on the incorporation of the deuterium label from D2O into the methyl group of tert-butyl alcohol which had previously been adsorbed on zeolite (Fig. 8, Scheme 4). { It was suggested that dehydration of the alcohol on HZSM-5 zeolite with the Si :Al ratio of 36 was completed in 1.5 h.568 81.6 * *** 1 0 * * * * ** 1 ** 2 0 * * ** ** ** 2 * * 3 0 * * ** * 4 0 * * * ** 86 5 0 * * * *6 0 * * * * **** * 7 0 * * * ** 34567 0 100 80 40 d /ppm Figure 6.Changes in the 13C CP/MAS NMR spectra of tert-butyl alcohol (containing the selectively 13C-labelled C±OH group) with time at 296 K (1 ± 5), 373 K (6) and 448 K (7) after adsorption on HZSM-5 zeolite (Si :Al=24). Spectra 1 0 ± 7 0 at d 60 ± 110 are given with a fourfold amplification. Time /h: (1) 0.33; (2) 2.83; (3) 5.5; (4) 8.0; (5) 15.5; (6) 1; (7) 1. The signal at d 82 belongs to the 13C±OH group of the unconsumed initial alcohol.33.5 24.9 14.2 29.8 * 1 0 * 14.2 * 1 86 * 2 0 * * ** 2 * 3 0 * * * ** 3 * * * * * * * 4 0 ** ** 5 0 * * * * 6 0 * ** 456 80 40 0 100 d /ppm Figure 7. Changes in the 13C CP/MAS NMR spectra of tert-butyl alcohol (containing the selectively 13C-labelled CH3 group) with time at 296 K (1 ± 4), 373 K (5) and 448 K (6) after adsorption on HZSM-5 zeolite (Si :Al = 24). Spectra 1 0 ± 6 0 at d 60 ± 110 are given with an eightfold amplification. Time /h: (1) 0.33; (2) 2.83; (3) 5.5; (4) 15; (5) 1; (6) 1. The signal at d 29.8 belongs to the 13CH3 group of the initial alcohol. ab0 kHz 50 100 Figure 8. The static 2H NMR spectra observed after adsorption of D2O on a HZSM-5 zeolite specimen with the previously adsorbed ButOH recorded within 5 min (a) and 1 h (b) after adsorption ofD2O(T=296 K).(a) the shape of the spectral line corresponds to deuterated water; (b) the shape of the spectral line corresponds to superposition of three signals, viz., a signal for the (CD3)3C fragment of alcohol with the observed quadrupole splittingQ3&10 kHz and signals for theCD3 andCD2 groups (weak signals) of oligomers with Q2&38 kHz and Q1&123 kHz, respectively. H CH3 7H2O O H3C C CH3+Al Si OHH3C C CH2+Al H3C H3C C CH2+Al H3C CH3 DHO + 7 H3C C CH2D O Si Al It was also demonstrated that if zeolite containing a smaller number of acidic OH groups was used, i.e., in the case of slower dehydration of the alcohol, the selective 13C label from the C±O group of the alcohol was transferred to its CH3 group (Fig.9).38 This redistribution of the 13C label confirms that TBCI is formed as an intermediate, which exists in equilibrium with the initial alcohol, in the course of dehydration. An analogous redistribution of the carbon labels was also observed in the case of carbenium ions in superacid solutions (Scheme 5).49, 50 In spite of the above-considered experimental evidence for the formation of TBCI, the signal at d 330 characteristic of the carbenium ion was not observed.51 It was thus concluded that the cation formed in zeolite is either transient or present in an amount insufficient for detection by 13C NMR spectroscopy. The signal at d 86, which was observed in the 13C NMR spectrum in the course of dehydration of ButOH, was originally assigned to stable tert-butyl silyl ether formed in the reaction (see Figs 6 and 7).35 However, it was found that this signal appeared simultaneously with signals of oligomers and persisted after completion of dehydration.This signal disappeared only at a temperature higher than 373 K at which cracking of oligomers A G Stepanov 750 7100 Scheme 4 CH3 + 7 H2O H3C C CH3 O Si Al H D2O O Si 7DHO DO Si D CH3 O H3C C CH2D+Al Si OHHigh-resolution solid-state NMR spectroscopy in studies of conversions of hydrocarbons and alcohols on zeolites H CH3 7H2O 13 O H3C C CH3+ Si Al OH 13 CH3 H3C CH CH2 +CH3 H2O + H313C 7 C CH3 O Al Si * 100 150 d /ppm Figure 9.The 13C NMR spectrum of tert-butyl alcohol (containing the selectively 13C-labelled C±OH group) recorded within 4 h after its adsorption on HZSM-5 zeolite (Si :Al = 44). The signal at d 82 belongs to the C±OH group of the unconsumed alcohol. The narrow signal at d 29.7 against the background of the signals of oligomers at d 14 ± 40 occurs due to the incorporation of the 13C label from the C±OH group into the CH3 group of the alcohol. accompanied by their conversions into a mixture of alkanes and cyclopentenyl cations occurred 38 (see below). Hence, it was concluded that the signal at d 86 should be assigned to C±O groups of oligomers covalently bound to the oxygen atoms of the zeolite framework.It was thus suggested that dehydration of ButOHon the HZSM-5 zeolite proceeds according to Scheme 6 to form intermediate TBCI, which reacts with isobutene formed from TBCI to give oligomeric products. These products in the adsorbed state are oligomeric alkoxides which exist in equilibrium with oligomeric carbenium ions. It is to these alkoxides that the signal at d 86 belongs. H CH3 7H2O O H2O H3C C CH3+Al Si OH H H3C C O CH2+ Si Al H3C CH3 H3C C CH CHCH3 H3C H3C O Si Al isobutene oligomer Scheme 5 CH3 13 +7 H2O H3C C CH3 O Si Al CH3 +CH CH3 H213C H CH3 O H313C C CH3+ Si Al OH 29.8 81.6 14.3 * * * 50 0 Scheme 6 CH3 +7 H3C C CH3 O Si AlTBCI TBCI H3C + C +H2O CH O7 H3C H3C CH3 CHCH3 Al Si 569 3. Isobutyl alcohol IR spectral studies of the kinetics of dehydration of isobutyl alcohol on an HZSM-5 zeolite (Si :Al=20) 83 demonstrated that dehydration of the alcohol at 373 K was completed in 30 min.An adsorbed intermediate retained the structure of the hydrocarbon skeleton of the initial alcohol. It was thus concluded that diisobutyl ether was formed as an intermediate. However, analysis of the 13C NMR spectra demonstrated that the inter- mediate was in fact isobutyl silyl ether (IBSE).47 The chemical shifts of the carbon atoms of the CH3 and CH groups of the alkoxide intermediate (d 19.2 and d 30.5, respectively) coincide with those of the carbon atoms of the analogous groups in the initial alcohol, while the chemical shift of the carbon atom of the CH2 group (d 73.0, Fig.10 c) is close to that of the CH2 group in the initial alcohol (d 74.9, Fig. 10 a) but differs substantially (Dd *10 ppm) 47 from that of the carbon atom of the CH2 group in diisobutyl ether (d 84.6).47 Prolonged storage of BuiOH adsorbed on the zeolite at 373 K or 398K resulted in an increase in the intensities of the signals at d 19.2 and d 30.5 (Fig. 10 c,d) due to redistribution of the selective 13C label in IBSE, which suggests the formation of the carbenium ion as an intermediate. In this case, the intensity of the signal at d 19.2 increased due to the incorporation of the selective carbon label from the CH2 group to the CH3 group of IBSE 47 and the intensity of the signal at d 30.5 increased due to the predominant formation of TBCI and its stabilisation as TBSE whose methyl group resonates at d 30.5.This signal is close to that of the methyl group of tert-butyl alcohol (d 31 35). An increase in the intensity of the signal at d 30.5 was originally attributed to the incorporation of the selective 13C label into the CH3 group of TBSE 47 or into the CH group of IBSE 84 via the linear sec-butyl cation, which is thermodynamically more stable in the zeolite.85 However, more recent studies of the reaction of isobutyl alcohol with carbon monoxide on the zeolite demonstrated that trimethylacetic acid was selectively formed in this reaction 61 (see below), which is unambiguously indicative of the predominant formation of TBCI in the reaction of IBSE with carbon monoxide.Hence, it is reasonable to relate an increase in the intensities of the signals at d 19.2 and d 30.5 (Fig. 10 c,d) to the equilibria on the zeolite shown in Scheme 7. Scheme 7 CH3 H3C CH3 H3C H CH 7H2O 13 O CH 13CH2 CH2 H2O +Al Si O OH Al Si CH3 H3C CH3 H3C +C CH 13 13 + CH2 CH3 +CH2 H3C CH 13CH3 H313C CH3 CH3 CH C CH3 H313C CH2 O O Si Si AlTBSE AlIBSE570 Ô 73.0 * b H3C13 CH3 CH CH2 O Si Al * a H3C13 CH3 CH CH2 OH * 100 d /ppm Figure 10. Changes in the 13C NMR spectrum of isobutyl alcohol (containing the selectively 13C-labelled CH2 group) adsorbed on HZSM-5 zeolite (Si :Al=24) upon successive heating of the specimen at different temperatures. Temperature /K (duration of heating, min): (a) 296 (without heating); (b) 343 (70)+373 (30); (c) 398 (40); (d) 413 (60); (e) 428 (60); (f ) 448 (60).Further increase in the temperature (Fig. 10 d ± f) caused subsequent conversions of IBSE and TBSE into a mixture of alkanes (Fig. 10 f ) and CPC (see Scheme 12 below) via butene oligomers.38 IV. Reactions of alkenes and alcohols with carbon monoxide and acetonitrile The experimental evidence for the involvement of carbenium ions in conversions of alkenes and alcohols on zeolites obtained in NMR studies was further confirmed by a series of chemical reactions which occurred on zeolites under mild condi- tions.49, 50, 86 1. The Koch reaction Selective conversions of isomeric butyl alcohols in the presence of carbon monoxide and of alkenes (ethylene, isobutene and oct-1- ene) in the presence of CO and water on the HZSM-5 zeolite to form carboxylic acids (the Koch reaction 59, 87), which were studied with the use of 13C NMR spectroscopy,60, 61 were found to occur at lower temperatures (20 ± 100 8C) compared to those (300 ± 500 8C) normally used in catalytic reactions on zeolites.Actually, the use of labelled reagents (alcohols, alkenes and CO) for the analysis of products of their reactions on the HZSM-5 zeolite provided conclusive evidence of selective conversions of alcohols and alkenes into carboxylic acids. Thus the following changes in the 13C NMR spectra (Fig. 11) were observed in the reaction of ButOH with CO at room temperature: when ButOH containing the 13C-labelled tertiary carbon atom was used, a signal at d 40.7 belonging to the tertiary carbon atom of trimethyl- acetic acid (TMAA) was observed (Fig.11 a); when ButOH containing the 13C-labelled CH3 group was used, a signal at d 27.8 belonging to the methyl group of the acid was observed (Fig. 11 b); and when labelled carbon monoxide was used, a signal H313C CH3 CH CH2 O Si AlIBSE 30.5 19.2 CH3 H313C C CH3 O Si Al TBSE * 50 0 fed 100 d /ppm at d 186 ± 194 belonging to the carboxyl group was observed (Fig. 11 c). The formation of TMAA from ButOH and CO on a zeolite catalyst at room temperature agrees well with the general concepts of the mechanism of this reaction in solutions of acids and superacids.58, 59 In both cases, the carbenium ion is the key intermediate (Scheme 8).CH3 H 81.4 29.7 7H2O H3C C OH+ O Si Al CH3 CH3 + CO 7 H3C C CH3 O Si Al CH3 40.7 186;194 27.8 C H3C CCH3 OH Hereinafter, the chemical shifts for the carbon atoms are given above the corresponding atoms. The special role of carbon monoxide in this reaction is to capture TBCI formed upon elimination of the water molecule from the alcohol and to hinder its rapid oligomerisation 35, 38, 46 by converting it into the acylium cation, which reacts with water to form acid. Trimethylacetic acid is also selectively formed from BuiOH and isobutene, which suggests the presence of TBCI.Ethylene is selectively converted into 2-ethyl-2-methylbutyric acid, which is indicative of rapid oligomerisation of ethylene to form the C6 12.2 A G Stepanov 25.4 7 Si 33.5 30.5 50 22.9 14.7 0 Scheme 8 H H3C C CH2+ O Al Si H3C CH3 H2O C+ O H3C C O CH3 Al H O+ O Si AlHigh-resolution solid-state NMR spectroscopy in studies of conversions of hydrocarbons and alcohols on zeolites a 40.7 81.4 * * * * b 27.8 29.7 c 194 * * * 186* * * * * * * * * * * d 184 * * * ** * ** * 100 0 200 d /ppm Figure 11. The 13C CP/MAS NMR spectra of the product formed upon adsorption of ButOH and CO on HZSM-5 zeolite at 296 K.ButOH containing the 13C-labelled tertiary carbon atom (a); ButOH containing the 13C-labelled methyl group (b) and labelled CO (c). The spectrum (d) is the spectrum (c) recorded without CP. carbenium ion followed by termination of the growing oligomeric chain by carbon monoxide. Subsequent NMR spectral studies demonstrated also that benzene can undergo carbonylation to form benzoic acid on HY zeolite 88 and sulfatised zirconia.89 2. Friedel ± Crafts-like acylation of alkenes In the absence of water adsorbed on zeolite, the alkyl cation which is formed upon adsorption of alkenes on zeolite reacts with CO to give the acylium cation. The latter reacts with another alkene molecule to form unsaturated ketone. This represents the Friedel- Crafts-like acylation of alkenes,90 ± 92 the acylating agent being formed in situ from the alkene and CO on the zeolite.Actually, the 13CNMRspectrum of the product formed in the case of simultaneous adsorption of isobutene and 13CO on the HZSM-5 zeolite has two intense signals for the 13C-labelled carbonyl group 93 at d 225 and d 250 (Fig. 12 a,b). The signal at d 225 belongs to unsaturated ketones formed in the reaction and the signal at d 250 belongs to stable cyclic five-membered carboxo- nium ions (Scheme 9).94 The 13C CP/MAS NMR spectra of the products formed after the adsorption of 13C-labelled ethylene and CO on the HZSM-5 zeolite are shown in Fig. 12 c,d. If a zeolite specimen was kept in air, i.e., was exposed to atmospheric moisture, the signal for the carboxonium ion at d 250 disappeared, the signal for unsaturated ketones was shifted downfield (to d 217) and a signal belonging to carboxylic acids appeared at d 186 (Fig.12 e). It was thus concluded that carboxo- nium ions were decomposed in the presence of moisture and the a 225 * 250 185 * * * * * b ** (5) * * * * * * c 250225d (20) e 186 217 * * 300 200 d /ppm Figure 12. The 13C CP/MAS NMR spectra of the products formed after adsorption of alkenes and CO on HZSM-5 zeolite in the absence of water at 296 K. (a) (CH3)2C=CH2+13CO; (b) (CH3)2C=CH2+CO; (c) CH2=CH2+ 13CO; (d ) 13CH2=CH2 + CO; (e) (CH3)2C=CH2 + 13CO + H2O. Curves (1 ± 5) without amplification; (2 0) with a fivefold amplification; (4 0) with a twenty fold amplification. H+ + (CH3)2C CH3 (CH3)2C CH2 (CH3)2C CH3 (CH3)2C CH2 + CO acylium cation H3C + H3C O C(CH3)3 equilibrium shown in Scheme 9 was shifted towards acylium cations, which reacted with water to form carboxylic acids.R2C CH3 H2O + R2CCOOH C CH3 O 571 * * 1 * * 123 * 2 0 2 * * 3 * * 112 * * 4 0* * 4 * * 5 * * 0 100 Scheme 9 CO H3C +O 250C(CH3)3 H3C 225 O C(CH3)3572 Hydroxy ketones, which are the products of the reaction of unsaturated ketones or carboxonium ions with water, were not detected in the reaction under consideration. 3. The Ritter reaction Since the discovery of the reactions of alcohols and alkenes with nitriles (the Ritter reaction) in 1948,95 ± 98 it was believed that the formation of N-alkylamides occurs via an intermediate N-alkylni- trilium cation CH3C:N+R.108.0 69.6 59.5 + H+ NR H3CC ROH+CH3CN 7H2O However, the experimental spectral supporting evidence for the formation of N-alkylnitrilium cations was long absent, which has cast doubt on the existence of a cationic intermediate.98 It was suggested that sulfate this reaction afforded a CH3C(OSO3H)=NR (the reaction was carried out in sulfuric acid) as an intermediate.95, 98 It was demonstrated 99, 100 by 13C and 15N MAS NMR spectroscopy that the Ritter reaction can be performed on zeolite catalysts. These studies provided the first experimental proof for the formation of N-alkylnitrilium cations as stable intermediates in the reactions of alkenes (or alcohols) with carbonitriles.The 13C NMR spectra of the intermediate N-tert-butylaceto- nitrilium cation and of the final N-tert-butylacetamide, which were formed in the reaction of ButOH with CH3CN on the a 59.5 69.6 (CH3)3COH** 28.4 * * b 59.5 27.5 * * Ô 177.1*ÊH3CN * * * * * *** * d 177.1 * * * ** ** ** * * * ***50 100 200 150 59.5 28.4 27.5 24.8 0 750 d, uÈ .o . Figure 13. The 13C NMR spectra of the products formed upon simulta- neous adsorption of tert-butyl alcohol and acetonitrile on HZSM-5 zeolite at 296 K; (a) and (b) simultaneous adsorption of ButOH containing the labelled tertiary carbon atom and CH3CN; (c) and (d) adsorption of labelled ButOH (10% enrichment with the 13C isotope) and CH3CN (80% enrichment with the 13C isotope); (a) and (c) the spectra were recorded within 4 h after the adsorption; (b) and (d ) the spectra were recorded within 4 days after the adsorption.O H3CCNHR H2O 7H+ CH3 69.6 H3C C N C CH3 CH3 CH3 O 27.5 59.5C CH H3C C NH 3 CH3 CH3 108.0 69.6 H3C C N C CH3 CH3 * CH3 O 24.8 27.5 59.5C CH H3C C NH 3 177.1 CH3 A G Stepanov HZSM-5 zeolite, are shown in Fig. 13. It was found that N-alkylnitrilium cations are long-lived stable species in the absence of moisture. Hence, the reaction mechanism proposed for the Ritter reaction 95 was additionally confirmed by the detection of stable N-alkylnitrilium cations on zeolites using NMR spectroscopy.The behaviour of the acetonitrile molecule in the Ritter reaction is analogous to that of the CO molecule in the Koch reaction, i.e., acetonitrile acts as a trap for the alkylcarbenium ion thus stabilising the unstable alkyl cation (Scheme 10) and preventing secondary reactions involving highly reactive alkyl cations (oligomerisation and cracking). Scheme 10 H CH3 CH3 7H2O CH3CN + O H3C C OH+ H2O 7 H3C C CH3 O Al Si CH3 Al Si O H CH3 CH3 + C O NH C CH3+ H3CC C CH3 H3C 7 NO Al Si CH3 CH3 Al Si The identification of stable N-alkylnitrilium cations con- firmed the suggestions made previously that alkylcarbenium cations can be stabilised on zeolite through their interactions with acetonitrile molecules 101 ± 103 and disproved the sugges- tion 104 that strong Lewis centres to which acetonitrile molecules are coordinated are formed on the zeolite.V. Analysis of hydrocarbon products and intermediates formed from alkene oligomers Adsorbed hydrocarbon products were analysed in situ by studying dehydration of ButOH on HZSM-5 zeolite in a wide temperature range (296 ± 637 K) as an example using 13C CP/MAS NMR spectroscopy.38 It was found that heating (448 K) of a zeolite specimen containing adsorbed alcohol resulted in the appearance of two groups of signals at ca. d 10 ± 40 and d 150 ± 250 in the 13C NMR spectrum (Fig. 14). The signals at d 10 ± 40 correspond to a mixture of acyclic alkanes (Fig.14 a) and those at d 100 ± 450 correspond to stable cyclopentenyl cations (Fig. 14 d ). The simultaneous formation of alkanes and CPC from oligomers of isobutene, which is formed in the first stage of dehydration of tert- butyl alcohol, occurs as a result of cracking of oligomers (for example, of oct-1-ene; Scheme 11). The diene formed upon cracking of the alkene undergoes further dimerisation and cyclisation to form alkyl-substituted cyclopentadienes, which are stabilised within channels of the HZSM-5 zeolite as CPC. Scheme 11 CH CH2 H3C (CH2)5 H3C CH CH3+ H2C CH CH CH2 , H3C 2H2C CH CH CH2 H3C CH CH CH CH CH CH CH3 CH3 CH3 H+ CH3 H3C CH3 H3C + CH3 CH3 CH3 H3C CH3 H3C + +High-resolution solid-state NMR spectroscopy in studies of conversions of hydrocarbons and alcohols on zeolites a * * 33.61 b ** c ** 50 40 30 d /ppm d * * * 254 * (16) 400 300 d /ppm Figure 14.The 13C MAS NMR spectra of the conversion products of ButOH on HZSM-5 zeolite at 448 K at d 10 ± 60 (a ± c) and at d 100 ± 450 (d). (a) experimental data; (b) a model spectrum obtained by superposition of 25 signals of the mixture of 11 hydrocarbons; (c) the spectrum of the cyclopentenyl cation (a 16-fold amplification of the intensities of the signals). The simultaneous formation of acyclic alkanes and CPC on the HZSM-5 zeolite indicates that the mechanisms of conversions of alkenes on zeolites and in sulfuric acid are similar.Actually, as early as 1936, Ipatieff and Pines 105, 106 demon- strated that a mixture of alkanes and cycloalkenes was formed when isobutene was passed through concentrated sulfuric acid at 0 8C. Cycloalkenes, as was demonstrated by Deno and co-workers in the early 1960s using NMR spectroscopy,107 exist in 96% H2SO4 as stable alkyl-substituted CPC. The use of 2D J-resolved 13CMASNMRspectroscopy for the analysis of the multiplicities of signals at d 10 ± 40 corresponding to hydrocarbon products (Fig. 15) allowed the determination of the number of hydrogen atoms attached to the carbon atoms in the CHn groups (n=0 ± 3). Based on the analysis of the 13C chemical shifts, the relative intensities of the signals (see Fig. 14) and the multiplicities (see Fig.15), it was found that a mixture of hydrocarbons was formed 30.67. 25.5 20 23.00 20.20 154 100 20014.82 12.31 159 10 0 710 * ** * * * * * ** ** 20.38 16.96 573 a F1 7100 0 10.0 15.0 20.0 25.0 30.0 35.0 100 Hz F2 d /ppm b25.50 * 20 30 26.7 Figure 16. The 13C CP/MAS NMR spectra of the conversion products of ButOH on HZSM-5 zeolite at 573 K [spectrum (b) is given with an eightfold amplification of the intensities of the signals at d 40 ± 240]. 14.70 12.28 0 10 40 50 710 d /ppm Figure 15. Contour plot of the 2D J-resolved 13CMASNMRspectrum of the products of ButOH conversion on HZSM-5 zeolite at 448 K (a). The 1D spin-echo 13CNMRspectrum (b) corresponding to the spectrum of the standard one-pulse excitation (see Fig.14 a). The observed value of the spin-spin coupling constant J(13C±1H) on the F1 axis is half as large as the true value of J(13C±1H) due to the use of a special technique for recording the 2D NMR spectrum.35, 61, 63 at 448Kand the relative amount of hydrocarbons that occurred in the adsorbed state (Table 1) was determined. An increase in the reaction temperature to 573 K led to the redistribution of the intensities of signals for alkanes at d 10 ± 40 (Fig. 16). In this case, propane and butanes were prevailing a 128.8 * * * * * * 0 100 d /ppm b ** ** * * ** * * * ** * * * * * * 100 36.37 33.65 30.68 28.38 23.34 22.98 200 d /ppm574 Table 1. Assignment of the signals in the 13C NMR spectrum and the relative amounts of alkanes, isobutene oligomers and aromatic hydrocarbons formed upon conversion of ButOH on HZSM-5 zeolite at 373 ± 673 K.Hydrocarbon d /ppm a 13CH37 713CH27 13CH3CHCH3 713CH 7.5 (5.7) 17.0 (15.4) 14.7 (13.1) 18.0 (15.9) 27.1 (24.9) 25.5 (25.0) 25.5 (24.3) Ethane Propane n-Butane Isobutane n-Pentane 14.7 (13.7) 24.5; 36.6 (22.6; 34.6) 33.6 (31.7) 30.7 (29.7) 23.0 (21.9) 11.8 (11.4) Isopentane Neopentane n-Hexane 14.7 (13.7) 28.4 (27.6) 23.5 (22.4) 14.7 (14.0) 2-Methylpentane 24.5; 33.7 (22.8; 31.9) 20.2; 40.9 (20.5; 41.6) 34.9 (33.9) 20.2 (19.1) 8.5 (8.5) 43.4 30.7 isobutene oligomers 36.6 (36.5) 22.3; 23.4; 30.7; 34.2; 36.6; 40.9 o- and p-Xylenes 2,3-Dimethylbutane 2,2-Dimethylbutane Alkanes (C7 and above)+ 10.6 11.8 13.9 20.7 (19.6; 20.9) 22.5 (32.3) Toluene+m-xylene Note.The assignment was made relative to the total amount of hydrocarbons whose CHn signals are observed at d 10 ± 40. a The chemical shifts of CHn groups of the adsorbed hydrocarbons were taken from the spectra shown in Figs 14 ± 17 (the accuracy of the determination was 0.5 ppm). The chemical shifts of the same hydrocarbons in a solution were taken from the literature 40 and are given in parentheses. b The values were determined only from the intensities of the CH3 groups without considering the intensities of the carbon atoms of the benzene ring of toluene or xylenes. At 373 and 296 K, alkanes were also formed from isobutene oligomers.38 However, at these temperatures, isobutene oligomers covalently bound to the oxygen atoms of the zeolite framework, viz., alkoxides, were obtained as the major conversion products of ButOH on HZSM-5 zeolite, which is evidenced by the presence of a broad signal for the C±O groups at d 89 in the spectrum of oligomeric products (Fig.18). alkanes (Table 1), signals for CPC at d 150 and d 250 disappeared and a signal at d 128, which belongs to a mixture of monocyclic and fused aromatic hydrocarbons with the latter predominating, appeared. The changes in the 13CNMRspectrum are indicative of cracking of alkanes formed at 448 K. When the temperature was further increased to 673 K, C2±C4 alkanes (with propane predominant) and aromatic hydrocarbons, viz., toluene and xylene, were obtained as the major reaction products (Fig.17). The sequence of conversions (according to the 13C MAS 17.03 NMR spectral data) of alkene oligomers adsorbed on acidic zeolite in the temperature range of 296 ± 637 K is shown in Scheme 12.38 25.58 7.52 129.13 * * * 0 50 100 150 d /ppm d /ppm Figure 18. The 13C CP/MAS NMR spectrum of the conversion products of ButOH on HZSM-5 zeolite at 373 K. Figure 17. The 13C MAS NMR spectrum of the conversion products of ButOH on HZSM-5 zeolite at 673 K. 40.86 A G Stepanov Relative amount (mol %) at different temperatures (K) 13C 673 573 448 3737 7 7 8.7 59.2 5.3 5.3 7 29.8 18.5 19.6 3.7 1.0 <1.7 17.2 17.2 <1 76 7 7 31.8 (31.4) 33.43 11.7 7 7 7 7 17.9 0.7 6.9 5 77 30.11 24.88 22.84 14.26 4.1 7 7 7 30.7 (30.2) 3.4 2.1 27.8 7 7 7 7 7 7 7782 13.4 b 7 7 16.7 b 7 7 7 8.1 b 89 * * * * (3) * ** * 0 50 100 150High-resolution solid-state NMR spectroscopy in studies of conversions of hydrocarbons and alcohols on zeolites Scheme 12 CH3CH CH(CH2)nCH3 HO Al Si + CH3CH2CH(CH2)nCH3 7 CH3CH2CH(CH2)nCH3 O O Al Si Al Si 296 ± 473KCH3 CH3 alkanes C4±C7+ H3C + 473 ± 573K CH3 CH3 CH3 + alkanes C3±C5+ 673K CH3 CH3 + + alkanes C2±C4 CH3 VI.Activation of alkanes at low temperatures. Isotope (H/D) exchange between alkanes and acidic OH groups of zeolite The results of the studies 108 ± 112 demonstrate that NMR spectro- scopy is a particularly informative and unique method for investigating conversions of lower alkanes (propane and isobu- tane) on acidic zeolites at low (as regards traditional conversions of alkanes on zeolites) temperature (350 ± 550 K).It was found 108 that the selective 13C label of theCH2 group was incorporated into theCH3 group of propane adsorbed on Ga/HZSM-5 zeolite at 573 K. According to the results of the studies,50, 55 ± 57 the redistrib- ution of the selective carbon label indicates that the isopropylium ion which is responsible for this redistribution is formed as an intermediate on the zeolite, although Derouane and co-work- ers 108 believed that the label redistribution observed in the case of Ga/HZSM-5 zeolite proceeds with the participation of a transi- tion complex of propane with the Ga atom,108 while in the case of pure HZSM-5 zeolite,109 the redistribution involves the C-ethyl- methanium (carbonium) ion.113 13CH2 H+ H2C CH2 Ga/H-ZSM-5 d+ d7H H O7 +Ga O27 Z CH313CH2CH3 + 13CH3 H-ZSM-5 H3C CH3 13CH3CH2CH3 The observed redistribution of the 13C label in propane indicates that propane is activated and can undergo subsequent chemical conversions (cracking and alkylation 114).Yet another example of the use of NMR spectroscopy for studying activation of alkanes at low temperature is investigation 575 of hydrogen isotope exchange 110, 111 and activation of alkanes in the presence of CO and H2O.112 The H/D isotope exchange between alkanes and acidic OH groups of zeolites proceeds at rather low temperatures (20 ± 200 8C) at which the alkane does not undergo isomerisation and cracking.115 ± 121 The isotope exchange involving stages of cleavage and formation of C±H bonds in alkane and O±H bonds in zeolite is indicative of activation of alkane.Hence, it is not surprising that a large number of experimental 115 ± 124 and theoretical 121 ± 124 studies were devoted to the elucidation of the mechanism of the hydrogen exchange. Recently, 1H MAS NMR spectroscopy was used for investigating the kinetics and mechanism of theH/Dexchange of alkanes.110, 111 1. Propane In studies of the kinetics of the H/D isotope exchange between deuterated propane and acidic groups of HZSM-5 zeolite (150 ± 250 8C), the simultaneous incorporation of protium from the acidic HO groups into the deuterated methyl and methylene groups of propane was observed.It was found that the intensities of the signals for the CH3 and CH2 groups of propane at d 1.0 and d 1.45, respectively, in the 1HMASNMRspectrum increased with time (Fig. 19). Analysis of the kinetics of the incorporation of protium into the methyl and methylene groups of propane 0 1 d /ppm 2 Figure 19. Changes in the intensities of the signals for the methyl (d 1.0) and methylene (d 1.45) groups with time at 519 K in the 1H MAS NMR spectrum of propane C3D8 adsorbed on HZSM-5 zeolite.The first spectrum (at the bottom) and the last spectrum were recorded within 3 min and 5 h after the adsorption. The intervals at which the spectra were recorded were 5 min. revealed a noticeable difference in the rates of hydrogen exchange for these groups (Fig. 20). The corresponding rates, which were I (arbitrary unit) 1 15 10 2 50 200 100 t /min Figure 20. Kinetic curve of the transfer of the proton from acidic OH groups of zeolite to the methyl (curve 1) and methylene (curve 2) groups of propane C3D8 at 519 K.576 calculated by the kinetic equation It = I?[17exp(7Rt)] [It and I? are the integral intensities of the signals for the CHn groups (n = 2 or 3) in the 1H MAS NMR spectrum at times t and t=? (equilibrium)] are R(CH3) = 0.03816 min71 and R(CH2) = 0.0436 min71.The time t = 0 corresponds to the beginning of the experiment when the temperature was increased from room temperature to 519 K. The activation energies of hydrogen exchange for the methyl and methylene groups were determined (ECH3 = 118 2 kJ mol71 and ECH2 = 117 4 kJ mol71) from the temperature dependence of the reaction rate of hydrogen exchange (from the Arrhenius parameters; Fig. 21).110 R /min71 1 1 2 0.1 0.01 0.001 0.0001 0.0021 0.0019 1/T /K71 0.0017 Figure 21. Dependence of the rate of the H/D exchange for the methyl (1) and methylene (2) groups of propane on HZSM-5 zeolite in terms of the Arrhenius parameters. Taking into account the absence of selective or predominant enrichment of the methyl groups with protium, the fact that the H/ D exchange is not inhibited by carbon monoxide and the fact that the experimental activation energies are close to those theoret- ically calculated for methane and ethane,122 ± 124 it was concluded that H/D exchange with the acidic OH protons occurs independ- ently for each group (without the formation of a single carbenium- ion intermediate or a transition state) via an intermediate pentavalent transition state in which the hydrogen atoms that are exchanged lie halfway between the carbon atom of the methyl (methylene) group and the oxygen atoms of the zeolite.H H CH3 CH2 C H D O O Si Si Al The data obtained do not also contradict the formation of the carbonium ion as a two-electron three-centre intermediate state involved in the hydrogen exchange in the propane ± superacid DF± SbF5 system.125 The difference in the rate of exchange for the methyl and methylene groups was attributed to the difference in the s-basicity of the C±H bonds of the primary (CH3) and secondary (CH2) groups of the alkane.126 2.Isobutane In the case of the hydrogen exchange in isobutane, the situation is different (Fig. 22). In this case, only methyl groups are subjected to H/D exchange,118 ± 120, 127 which is manifested in an increase in the intensity of the signal at d 1.0 (Fig. 22 b).111 The expected increase in the intensity of the signal for the CH group at d 1.85 (Fig. 22 c) was not observed. Measurements of the rates of hydrogen exchange in the temperature range of 80 ± 150 8C allowed the determination of the activation energy of the exchange (502 kJ mol71) which is half as large as that of A G Stepanov a Si7OH 2.0 H Si7O7Al CH3 1.0 4.9 * * b CH3 1.0 Si7OH 2.0 * c CH3 1.0 CH 1.85 * 0 5 10 75 d /ppm Figure 22.The 13C MASNMRspectra of isobutane adsorbed on HZSM- 5 zeolite at 295 K. (a) the spectrum of isobutane-d10 before the temperature was increased to 425Kor of isobutane-d10 in the presence of CO after storage of the sample at 475 K for 10 h; (b) the spectrum of isobutane-d10 after storage of the sample at 425 K for 70 min; (c) the spectrum of nondeuterated isobutane. propane. It was also found that carbon monoxide completely inhibits the regiospecific hydrogen exchange (Fig.22 a). The mechanism of the regiospecific hydrogen exchange for isobutane was explained within the framework of the catalytic cycle 118, 120 shown in Scheme 13. Scheme 13 AH CH3 CH2 CH3 ? + H H3C C C C CH3 H3C A7 CH3 H3C CH3 CH2D AD H H3C C CH2D + CH3 C CH3 A7 CH3 H3C H H3C CCH3 The regiospecificity of the H/D exchange is achieved due to formation of TBCI and isobutene as intermediates, which exist in equilibrium with each other. Carbon monoxide, which is present in the system, reacts with TBCI to give the oxocarbenium ion. The latter cannot undergo reversible deprotonation to form the alkene, which results in complete inhibition of the hydrogen exchange. CH2 C CH3 CH3 H3C ? + A7 H C H3C C H3C CH3 CO CH3 +AH CH3 CH3+C O H3C CCH3High-resolution solid-state NMR spectroscopy in studies of conversions of hydrocarbons and alcohols on zeolites VII.Activation of alkanes at low temperatures. Carbonylation In addition to the hydrogen exchange in alkanes at 20 ± 200 8C, NMR spectral studies allowed observation of carbonylation at 100 ± 200 8C,112 which is also indicative of the possibility of activation of alkanes on acidic zeolites at low temperatures. Based on the assumption that alkanes are formed from carbenium ions and alkenes 118, 119, 127 as intermediates upon hydrogen exchange at low temperature and taking into account that alkenes (carbenium ions) can react with carbon monoxide and water on zeolites 61 to form carboxylic acids, it was suggested that carboxylic acids can be formed from alkanes at low temper- ature 112 at which hydrogen exchange is observed but alkanes do not undergo cracking.1. Propane It was found that heating of a zeolite specimen containing adsorbed propane, CO and water at 373 ± 473 K led to changes in the NMR spectra compared to those of the initial reagents. In the case of simultaneous adsorption of 13C-labelled propane, carbon monoxide and water, the 13C CP/MAS NMR spectrum has a signal at d 191 for the carboxy group of isobutyric acid formed along with signals at d 17.0 and d 18.2 belonging to the initial propane (Fig. 23 a). The 1H MAS NMR spectrum of the same specimen has a signal for methane at d 0.25, signals for the initial propane at d 1.0 and d 1.45 and a signal for water adsorbed on acidic OH groups of zeolite at d 6.2 (Fig.23 c). In the case of adsorption of labelled propane-2-13C, the 13C MAS NMR spectrum has a signal for ethane-1-13C at d 7.3, a signal for the carbon atom of the 13CH group of isobutyric acid at d 36.4 and signals for unconsumed propane-2-13C and CO at d 18.2 and d 185, respectively (Fig. 23 b). These changes agree well with 191 (610) 185 (625) 150 d /ppm 6.2 * 10 5 d /ppm Figure 23. NMR spectra of propane, carbon monoxide and water adsorbed on HZSM-5 zeolite at 473 K. The samples were heated at 473 K for 3 h. (a) the 13C CP/MAS NMR spectrum, simultaneous adsorption of propane, 13CO and water; (b) the 13C MAS NMR spectrum, simulta- neous adsorption of propane-2-13C, CO and water; (c) the 13CMASNMR spectrum, simultaneous adsorption of propane, 13CO and water.a 17.0 18.2 b 18.27.3 * 36.4 * 0 50 100 c1.0 0.25 (610) 1.45 2 6 10 72 * 0 710 715 75 activation of propane-2-13C on zeolite and protolytic cleavage of the C±C bond of propane to form 13CH4, 13CH3CH3 and (CH3)213CHCOOH according to Scheme 14. Actually, in the course of the reaction of 13C-labelled propane with CO and H2O, the selectively 13C-labelled CH2 group of propane was converted into the 13CH group of isobutyric acid (d 36.4). Methane, which is observed only in the 1H MAS NMR spec- trum, is formed in the stage (1) due to the protolytic attack of the acidic group of zeolite on the C±C bond of propane.The ethyl cation formed in the same stage reacts with another molecule of labelled propane to give ethane and the isopropyl cation [stage (2)]. The latter reacts with CO and water to form isobutyric acid [stage (3)]. H-ZSM-5 H3C13CH2CH3 373 ± 473K + H3C13CH2+CH4 , 7 O Al Si + CH313CH2+CH313CH2CH3 + CH313CHCH3+CO+H2O 2. Isobutane The reaction of isobutane with carbon monoxide and water on zeolite proceeds analogously.112 After heating of a zeolite speci- men containing adsorbed isobutane, 13CO and water at 423 K for 3 h, the 13C CP/MASNMRspectrum has four signals, viz., signals forTMAAat d 193 and d 27.7, a signal for the initial isobutane at d 25.5 and a signal for the intermediate Si ± C(OOH) ±Al at d 172 (Fig.24). The 1H MAS NMR spectrum of the same specimen has signals for the initial isobutane at d 1.0 and d 1.8, a signal forH2O/ Si ±OH± Al at d 6.6 and signals for the reaction products, viz., H2 and TMAA, at d 4.1 and d 1.45, respectively (Scheme 15). No conversion products were detected in a special NMR spectral study of conversions of propane and isobutane on HZSM-5 zeolite in the absence of CO and water in the same temperature range. However, cracking products of these alkanes, 172 193 * d /ppm 200 * 10 d /ppm Figure 24. NMR spectra of conversion products of isobutane, 13CO and water adsorbed on HZSM-5 zeolite. (a) the 13C CP/MAS NMR spectrum; (b) the 13C MAS NMR spectrum.577 Scheme 14 Al + (1) CH37O H3C 13CH2 H Si+ (2) CH313CH3+CH313CHCH3 , (3) (CH3)213CHCOOH +H+. a 25.5 * 27.7 (68) * 0 50 100 150 b1.0 1.45 6.6 4.1 (615) 1.8 8 4 0 * 0 5 75 710 715578 Scheme 15 H Al H3C H 373 ± 423 K + O 7O (CH3)3CH+ H3C C H Al Si Si H3C H CH3 + CO, H2O (CH3)3CCOOH+H2+ O 7 H3C C CH3+H2 O Al Si Al Si such as hydrogen (in the case of isobutane), ethane and methane (in the case of propane), were detected by gas-chromatographic analysis. The amounts of these compounds were so small that they could not be detected by NMR spectroscopy. It was concluded that in the presence ofH2Oand CO(a trap of carbenium ions), the cracking equilibrium of alkanes was shifted to the corresponding acids formed from CO and carbenium ions.In this case, the reaction product, viz., a carboxylic acid, was accumulated in amounts detectable by NMR spectroscopy. The low degrees of conversion of propane and isobutane into the corresponding acids at 373 ± 423 K (1% ± 2%) are attributable to the presence of insignificant amounts of superacidic centres in the zeolite (no more than1%± 2%of the total amount of acidic protonic centres) on which cracking occurs at these low temperatures. VIII. The nature of intermediates in conversions of alkanes, alkenes and alcohols The nature of intermediates in the reactions of hydrocarbons and alcohols on heterogeneous catalysts of the acidic nature has long been the subject of discussion.According to the classical concepts of conversions of organic molecules in the presence of acids, alkyl- type carbenium ions represent reactive intermediates in these reactions. However, the absence of direct experimental evidence for the formation of alkyl-type carbenium ions on acidic centres of zeolites has cast doubt on the possibility of their formation as stable species. The first results of quantum-chemical calculations 128, 129 demonstrated that carbenium ions are not located in a local minimum on the potential energy surface and these ions can be formed on zeolites only as an unstable transition state. More recent studies 130 ± 132 demonstrated that alkoxides are key inter- mediates, while carbenium ions are formed upon thermal excita- tion of alkoxides where the C±O bond is elongated.Theoretical calculations 133 also indicated that acid catalysis on zeolites occurs with the involvement of alkoxide intermediates and without the participation of carbenium ions. Some NMR spectral data (for example, those obtained in studies of oligomerisation products of propene 134) were interpreted without regard for the participation of carbenium ions in the reaction. Recently, considerable progress has been achieved in the identification of various intermediates (carbenium ions and alkoxides) by NMR spectroscopy and in the understanding of the role of these intermediates in conversions on solid acids. The reliable evidence was obtained for the formation of alkoxide intermediates in the case of lower hydrocarbons (MeO,42 ± 44 PriO134 and BuiO47) and oligomeric alkoxides 54 and their role as intermediates was revealed.The methoxide (A) and isobutoxide intermediates (C) are stable species at 293 ± 373 K, while the isopropoxide intermediate (B) oligomerises even at 200 K.134 Apparently, this fact indicates that isopropoxide rather than the isopropylium ion is a reactive intermediate in the oligomerisa- tion of propene.134 However, these conclusions were drawn without considering the characteristic time of the 13C-carbon redistribution in carbenium ions, which should be approximately A G Stepanov CH3 H3C CH CH3 H3C CH CH3 CH2 O O O Si Al Si Al Si Al C B A 39 days at 230 K.134 Taking into account that oligomerisation proceeds substantially faster (is completed in several hours), the formation of the carbenium ion, which rapidly undergoes oligo- merisation, from isopropoxide must not be ruled out.The redistribution of the selective 13C label in oct-1-ene at room temperature, which is indicative of the formation of carbenium ions, as well as the alkoxide nature of adsorbed alkene oligomers 54 allow conclusions to be drawn about the existence of the equilibrium between the alkoxide and carbenium-ion inter- mediates, the latter being formed in small amounts undetectable by NMR spectroscopy. The formation of carbenium ions on zeolites is additionally confirmed by the fact that chemical reactions typical of carbenium ions proceed under mild conditions (room temperature).It appeared that unstable carbenium ions undetectable by NMR spectroscopy can be stabilised by introducing these ions into the reaction with CO and CH3CN to obtain cyclic carboxonium 93 and alkylnitrilium 99, 100 cations, respectively, which can be detected by NMR spectroscopy. The hydrogen H/D exchange between D2O and the methyl groups of ButOH35, 81 and the transfer of the 13C label from the tertiary carbon atom to the methyl group of ButOH38 are also indicative of the formation of the carbenium ion upon dehydration of ButOH on zeolite. Recent quantum-chemical calculations 85 of the stability of alkylcarbe- nium ions adsorbed on zeolite taking into account the stabilising effect of the zeolite lattice (it was believed that 16 oxygen atoms rather than two atoms, as was assumed in all the previous studies,128 ± 133 are involved in stabilisation of the intermediate formed) revealed a local minimum for carbenium ions.It was found that the stability of alkyl carbenium ions is 10 ë 20 kJ mol71 lower than that of the corresponding p-complexes of alkenes. Taking into account the experimental data obtained by NMR spectroscopy, it may be concluded that alkylcarbenium ions exist in zeolites as real structures (although they are very unstable and highly reactive) along with more stable alkoxide intermediates. Alkyl-substituted cyclopentenyl cations belong to another type of carbocationic intermediates that exist in zeolites.45 Their formation from alkene oligomers on zeolites simultaneously with alkanes indicates that conversions of alkenes on zeolites and in a solution of concentrated sulfuric acid occur analogously.38 These cations are intermediates, which form aromatic hydrocarbons at higher temperature.IX. Conclusion Recently, considerable progress has been made in the under- standing of the mechanisms of hydrocarbon conversions on the surface of zeolites of acidic nature. This progress was achieved primarily due to the successful use of high-resolution solid-state NMRspectroscopy for the in situ analysis of intermediates and the final products of hydrocarbon conversions. In some cases where the expected intermediate was not observed due to its instability or low concentration (alkyl-type carbenium ions), this was identified from the characteristic reactions, viz., from the redistribution of selective carbon labels or from the reactions with cation traps (CO or CH3CN).Based on the NMR spectral data, possible pathways of conversions of alkenes and alcohols on zeolites at low temper- ature were suggested. Carbonylation of alkenes and alcohols (the Koch reaction), Friedel ± Crafts-like acylation of alkenes where an acylating agent is formed directly from the alkene and CO onHigh-resolution solid-state NMR spectroscopy in studies of conversions of hydrocarbons and alcohols on zeolites zeolite and the reactions of alkenes with acetonitrile to form N- alkylamides (the Ritter reaction) were studied.Interestingly, the N-alkylnitrilium cation as a stable intermediate in the Ritter reaction was first detected and characterised by NMR spectro- scopy on zeolite rather than in solution. Hence, presently not only does solid-stateNMRspectroscopy confirm or refute the parallels between the mechanisms of conversions of hydrocarbons in solutions and on the surface of solid acidic catalysts, but it also enables one to obtain original data on the mechanisms of chemical conversions. The Friedel ± Crafts-like reaction of alkenes with CO adsorbed on zeolite to form unsaturated ketones and carbon- ylation of alkanes (propane and isobutane) to form carboxylic acids are examples. The use of NMR spectroscopy for the analysis of products formed in situ on zeolites provides a deeper insight into the effect of the internal structure of zeolite pores on the composition of the products, which was clearly exemplified by conversions of methanol on HZSM-5 zeolite.NMR spectroscopy is more informative than IR and UV spectroscopy in the analysis of the structures of adsorbed hydrocarbons. As a result, NMR spectral studies made it possible to identify stable alkyl-substituted cyclo- pentenyl cations, which were formed in zeolites as intermediates in the conversions of alkenes into aromatic compounds, and to establish their structures. In situ studies of the kinetics of the hydrogen exchange of alkanes by NMR spectroscopy allowed conclusions about the nature of intermediates of this exchange to be drawn.It would be expected that the experimental data on hydro- carbon conversions obtained by NMR spectroscopy and conclu- sions about the mechanisms of these conversions will be rationally used in the development of new industrially important processes of hydrocarbon conversions, such as alkylation or isomerisation of alkenes and alkanes, and the data on the nature of intermedi- ates will provide the basis for quantum-chemical calculations. References 1. C A Fyfe Solid State NMR for Chemists (Guelf, OR: CFC Press, 1983) 2. E R Andrew Prog. Nucl. Magn. Reson. Spectrosc. 18 1 (1971) 3. M M Maricq, J S Waugh J. Chem. Phys. 70 3300 (1979) 4. E R Andrew Trans. R. Soc. London, A 299 505 (1981) 5. S R Hartmann, E L Hahn Phys Rev.128 2042 (1962) 6. A Pines,M G Gibby, J S Waugh J. Chem. Phys. 59 569 (1973) 7. D E Demco, J Tegenfeldt, J S Waugh Phys. Rev. B, Solid State, Ser. 3 11 4133 (1975) 8. J M Thomas, J Klinowski Adv. Catal. 33 199 (1985) 9. C A Fyfe, Y Feng, H Grondey, G T Kokotailo, H Gies Chem. Rev. 91 1525 (1991) 10. V M Mastikhin, I L Mudrakovsky, A V Nosov Prog. Nucl. Magn. Reson. Spectrosc. 23 259 (1991) 11. O B Lapina, V M Mastikhin, A A Shubin, V N Krasilnikov, K I Zamaraev Prog. Nucl. Magn. Reson. Spectrosc. 24 457 (1992) 12. V M Mastikhin,O B Lapina, I L Mudrakovskii Yadernyi Magnitnyi Rezonans v Geterogennom Katalize (Nuclear Magnetic Resonance in Heterogeneous Catalysis) (Novosibirsk: Nauka, 1992) 13. A Bell, A Pines (Eds) NMR Techniques in Catalysis (New York: Marcel Dekker, 1994) 14.O V Krylov, V A Matyshak Usp. Khim. 63 585 (1994) [Russ. Chem. Rev. 63 559 (1994)] 15. A G Stepanov Catal. Today 24 341 (1995) 16. J F Haw, J B Nicholas, T Xu, L W Beck,D B Ferguson Acc. Chem. Res. 29 259 (1996) 17. I I Ivanova Ross. Khim. Zh. 42 67 (1998) a 18. C D Chang, A J Silvestri J. Catal. 47 249 (1977) 19. T A Carpenter, J Klinowski, D T B Tennakoon, C J Smith, D C Edwards J. Magn. Reson. 68 561 (1986) 20. L W Beck, J L White, J F Haw J. Magn. Reson. 99 182 (1992) 21. P E Eberly Jr J. Phys. Chem. 71 1717 (1967) 22. J Nova'kova', L Kubelkova', Z DolejsÏ ek, P JtruÊ Collect. Czech. Chem. Commun. 44 3341 (1979) 579 23. L Kubelkova , J Nova'kova', Z DolejsÏ ek, P JtruÊ Collect.Czech. Chem. Commun. 45 3101 (1980) 24. V Bolis, J C Vedrine, J P van den Berg, J P Wolthuizen, E G Derouane J. Chem. Soc., Faraday Trans. 1 76 1606 (1980) 25. E G Derouane, J-P Gilson, J B Nagy J. Mol. Catal. 10 331(1981) 26. E G Derouane, J-P Gilson, J B Nagy Zeolites 2 42 (1982) 27. J Datka Zeolites 1 113 (1981) 28. J Datka, in Catalysis on Zeolites (Eds D Kallo , KhM Minachev) (Budapest: Akade miai Kiado , 1988) p. 467 29. J Haber, J Komorek-Htodzik, T Romotowski Zeolites 2 179 (1982) 30. J P van den Berg, J P Wolthuizen, A D H Clague, G R Hays, R Huis, J H C van Hooff J. Catal. 80 130 (1983) 31. A K Ghosh, R A Kydd J. Catal. 100 185 (1986) 32. J-P Lange, A Gutsze, J Allgeier, H G Karge Appl. Catal. 45 345 (1988) 33.C Williams,M A Makarova, L V Malysheva, E A Paukshtis, E P Talsi, J M Thomas, K I Zamaraev J. Catal. 127 377 (1991) 34. K P Datema, A K Nowak, J van Braam Houckgeest, A F H Wielers Catal. Lett. 11 267 (1991) 35. A G Stepanov, K I Zamaraev, J M Thomas Catal. Lett. 13 407 (1992) 36. A G Stepanov, V N Zudin, K I Zamaraev Solid State NMR 2 89 (1993) 37. A G Stepanov, M V Luzgin, V N Romannikov, K I Zamaraev Catal. Lett. 24 271 (1994) 38. A G Stepanov, V N Sidelnikov, K I Zamaraev Chem. Eur. J. 2 157 (1996) 39. L M Sverdlov, M A Kovner, E P Krainer Kolebatel'nye Spektry Mnogoatomnykh Molekul (Vibrational Spectra of Polyatomic Molecules) (Moscow: Nauka, 1972) 40. E Breitmaier,W Voelter 13C NMR Spectroscopy, Methods and Applications in Organic Chemistry (Weinheim: Verlag Chemie, 1978) 41.V G Malkin, V V Chesnokov, E A Paukshtis, G M Zhidomirov J. Am. Chem. Soc. 112 666 (1990) 42. V BosacÏ ek J. Phys. Chem. 97 10732 (1993) 43. D K Murray, J-H W Chang, J F Haw J. Am. Chem. Soc. 115 4732 44. D K Murray, T Howard, P W Goguen, T R Krawietz, J F Haw 45. J F Haw, B R Richardson, I S Oshiro, N D Lazo, J A Speed 46. M T Aronson, R J Gorte,W E Farneth,D White J. Am. Chem. Soc. 47. A G Stepanov, V N Romannikov,K I Zamaraev Catal. Lett. 13 395 48. I I Ivanova, A Corma, Z Gabelica, in Proceedings of the 11th (1993) J. Am. Chem. Soc. 116 6354 (1994) J. Am. Chem. Soc. 111 2052 (1989) 111 840 (1989) (1992) International Zeolite Conference (Abstracts of Reports), Seoul, 1996 RP93 49.G A Olah Angew. Chem., Int. Ed. Engl. 12 173 (1973) 50. G A Olah, A M White J. Am. Chem. Soc. 91 5801 (1969) 51. G A Olah, D J Donovan J. Am. Chem. Soc. 99 5026 (1977) 52. G A Olah, E B Baker, J C Evans, W S Tolgyesi, J S McIntyre, I J Bastein J. Am. Chem. Soc. 86 1360 (1964) 53. G A Olah, J R DeMember, A Commeyras, J L Bribes J. Am. Chem. Soc. 93 459 (1971) 54. A G Stepanov, M V Luzgin, V N Romannikov, V N Sidelnikov, E A Paukshtis J. Catal. 178 466 (1998) 55. M Saunders, E L Hagen J. Am. Chem. Soc. 90 2436 (1968) 56. M Saunders, P Vogel, E L Hagen, J Rosenfeld Acc. Chem. Res. 6 53 (1973) 57. M Saunders,A P Hewett, O Kronja Croat. Chem. Acta 65 673 (1992) 58. H Hogeveen Adv. Phys. Chem. 10 29 (1973) 59. H Bahrmann, in New Syntheses with Carbon Monoxide (Ed. J Falbe) (Berlin: Springer, 1980) p.372 60. A G Stepanov, M V Luzgin, V N Romannikov, K I Zamaraev J. Am. Chem. Soc. 117 3615 (1995) 61. A G Stepanov, M V Luzgin, V N Romannikov, V N Sidelnikov, K I Zamaraev J. Catal. 164 411 (1996) 62. W E Garwood ACS Symp. Ser. 218 383 (1983) 63. S A Tabak, F J Krambeck, W E Garwood AIChE J. 32 1526 (1986) 64. R R Ernst, G Bodenhausen, A Wokaun Principles of Nuclear Magnetic Resonance in One and Two Dimensions (Oxford: Oxford University Press, 1987)580 65. A E Derome Modern NMR Techniques for Chemistry Research (Oxford: Pergamon Press, 1987) p. 259 66. R K Harris Nuclear Magnetic Resonance Spectroscopy. A Physicochemical View (London: Pitman, 1983) p. 107 67.M W Anderson, J Klinowski Chem. Phys. Lett. 172 275 (1990) 68. P A Couperus, A D H Clague, J P C Mvan Dongen Org. Magn. Res. 8 426 (1976) 69. G T Kokotailo, S L Lowton, D H Olson, W M Meier Nature (London) 272 437 (1978) 70. N D Lazo, B R Richardson, P D Schettler, J L White, E J Munson, J F Haw J. Phys. Chem. 95 9420 (1991) 71. F G Oliver, E J Munson, J F Haw J. Phys. Chem. 96 8106 (1992) 72. I I Ivanova, D Brunel, J B Nagy, E G Derouane J. Mol. Catal., A Chem. 96 243 (1995) 73. J L White, N D Lazo, B R Richardson, J F Haw J. Catal. 125 260 (1990) 74. S L Meisel, J P McCullough, C H Lechthaler, P B Weisz CHEMTECH 6 86 (1976) 75. C D Chang Catal. Rev. Sci. Eng. 1 25 (1983) 76. E G Derouane, J B Nagy, P Dejaifve, J H C van Hoof, B P Spekman, J C Vedrine, C Naccache J.Catal. 53 40 (1978) 77. M W Anderson, J Klinowski Nature (London) 339 200 (1989) 78. M W Anderson, J Klinowski J. Am. Chem. Soc. 112 10 (1990) 79. E J Munson, A A Kheir, N D Lazo, J F Haw J. Phys. Chem. 96 7740 (1992) 80. P W Goguen, T Xu, D H Barich, T W Skloss, W Song, Z Wang, J B Nicholas, J F Haw J. Am. Chem. Soc. 120 2650 (1998) 81. M T Aronson, R J Gorte, W E Farneth J. Catal. 98 434 (1986) 82. A G Stepanov, A G Maryasov, V N Romannikov, K I Zamaraev Magn. Reson. Chem. 32 16 (1994) 83. C Williams,M A Makarova, L V Malysheva, E A Paukshtis, K I Zamaraev, J M Thomas J. Chem. Soc., Faraday Trans. 86 3473 (1990) 84. A G Stepanov, K I Zamaraev Catal. Lett. 19 153 (1993) 85. K Teraishi J. Mol.Catal., A Chem. 132 73 (1998) 86. G A Olah, S K Prakash, J Sommer Superacids (New York: Wiley, 1985) 87. H Koch Brennstoff-Chemie 36 321 (1955) 88. T H Clingenpeel, A I Biaglow J. Am. Chem. Soc. 119 5077 (1997) 89. T H Clingenpeel, T E Wessel, A I Biaglow J. Am. Chem. Soc. 119 5469 (1997) 90. G A Olah Friedel ë Crafts and Related Reactions Vol. 1 (New York: Wiley, 1963) 91. J K Groves Chem. Soc. Rev. 1 73 (1972) 92. G A Olah Friedel ë Crafts Chemistry (New York: Wiley, 1973) 93. M V Luzgin, V N Romannikov, A G Stepanov, K I Zamaraev J. Am. Chem. Soc. 118 10 890 (1996) 94. O V Lyubinskaya, V A Smit, A S Shashkov, V A Chertkov, M I Kanishchev, V F Kucherov Izv. Akad. Nauk SSSR, Ser. Khim. 2 397 (1978) b 95. J J Ritter, P P Minieri J. Am. Chem. Soc. 70 4045 (1948) 96. S Patai (Ed.) Chemistry of Cyano Group (New York: Interscience, 1970) 97. E N Zil'berman Reaktsii Nitrilov (Reactions of Nitriles) (Moscow: Khimiya, 1972) 98. I D Gridnev, N A Gridneva Usp. Khim. 64 1091 (1995) [Russ. Chem. Rev. 64 1021 (1995)] 99. M V Luzgin, A G Stepanov Mendeleev Commun. 238 (1996) 100. A G Stepanov, M V Luzgin Chem. Eur. J. 3 47 (1997) 101. D S Bystrov Zeolites 12 328 (1992) 102. D S Bystrov, A A Tsyganenko, H FoÈ rster, in Proceedings of the 10th International Congress oÕ Catalysis (Eds L Guczi, F Solymosi, P Te te nyi) (Budapest: Akade miai Kiado', 1993) p. 268 103. S Jolly, J Saussey, J C Lavalley Catal. Lett. 24 141 (1994) 104. A S Medin, V Yu Borovkov, V B Kazansky, A G Pelmentschikov, G M Zhidomirov Zeolites 10 668 (1990) 105. V N Ipatieff, H Pines J. Org. Chem. 1 464 (1936) 106. H Pines The Chemistry of Catalytic Hydrocarbon Conversions (New York: Academic Press, 1981) 107. N C Deno, D B Boyd, J D Hodge, C U Pittman Jr , J O Turner J. Am. Chem. Soc. 86 1745 (1964) 108. E G Derouane, S B Abdul-Hamid, I I Ivanova, N Blom, P E Hojlund-Nielsen J. Mol. Catal. 86 371 (1994) A G Stepanov 109. I I Ivanova, E B Pomakhina, A I Rebrov, E G Derouane Top. Catal. 6 49 (1998) 110. A G Stepanov, H Ernst, D Freude Catal. Lett. 54 1 (1998) 111. J Sommer, D Habermacher, R Jost, A Sassi, A G Stepanov, M V Luzgin, D Freude, H Ernst, J Martens J. Catal. 181 265 (1999) 112. M V Luzgin, A G Stepanov, J Sommer, in The II International Memorial G K Boreskov Conference Catalysis on the Eve of XXI Century. Science and Engineering (Abstracts of Reports), Novosibirsk, 1997 Pt. 2, p. 250 113. G A Olah J. Am. Chem. Soc. 94 808 (1972) 114. I I Ivanova, N Blom, E G Derouane J. Mol. Catal. 109 157 (1996) 115. A Ozaki Isotopic Studies of Heterogeneous Catalysis (Tokyo: Kodansha, 1977) 116. C J A Mota, R L Martins J. Chem. Soc., Chem. Commun. 171 (1991) 117. C J A Mota, L Nogueira, W B Kover J. Am. Chem. Soc. 114 1121 (1992) 118. J Sommer, M Hachoumy, F Garin, D Barthomeuf, J Vedrine J. Am. Chem. Soc. 117 1135 (1995) 119. J Engelhardt, W K Hall J. Catal. 151 1 (1995) 120. J Sommer, D Habermacher, M Hachoumy, R Jost, A Reynaud Appl. Catal. 146 193 (1996) 121. T F Narbeshuber, M Stockenhuber, A Brait, K Seshan, J A Lercher J. Catal. 160 183 (1996) 122. G J Kramer, R A van Santen, C A Emeis, A K Nowak Nature (London) 363 529 (1993) 123. G J Kramer, R A van Santen J. Am. Chem. Soc. 117 1766 (1995) 124. S R Blaszkowski,M A C Nascimento, R A van Santen J. Phys. Chem. 100 3463 (1996) 125. J Sommer, J Bukala Acc. Chem. Res. 26 370 (1993) 126. G A Olah, Y Halpern, J Shen, Y Y Mo J. Am. Chem. Soc. 93 1251 (1971) 127. S G Hindin, G A Mills, A G Oblad J. Am. Chem. Soc. 73 278 (1951) 128. K I Zamaraev, G M Zhidomirov, in Proceedings of the 5th International Symposium on Relations between Homogeneous and Heterogeneous Catalysis (Eds Yu I Ermakov,V A Likholobov) (Utrecht: VNU Science, 1986) p. 23 129. A G Pel'menshchikov, N U Zhanpeisov, E A Paukshtis, L V Malysheva, G M Zhidomirov, K I Zamaraev Dokl. Akad. Nauk SSSR 293 915 (1987) c 130. V B Kazansky, I N Senchenya J. Catal. 119 108 (1989) 131. V B Kazansky Acc. Chem. Res. 24 379 (1991) 132. P Viruela Martin, C M Zicovich-Wilson, A Corma J. Phys. Chem. 97 13713 (1993) 133. M V Frash, V B Kazansky, A M Rigby, R A van Santen J. Phys. Chem. 101 5346 (1997) 134. J F Haw, J B Nicholas, T Xu, D Barich Top. Catal. 6 141 (1998) a�Mendeleev Chem. J. (Engl. Transl.) b�Russ. Chem. Bull. (Engl. Transl.) c�Dokl. Chem. Technol., Dokl.Chem. (Engl. Tr
ISSN:0036-021X
出版商:RSC
年代:1999
数据来源: RSC
|
4. |
N-Functionally substituted imines of polychlorinated (brominated) aldehydes and ketones |
|
Russian Chemical Reviews,
Volume 68,
Issue 7,
1999,
Page 581-604
Galina G. Levkovskaya,
Preview
|
|
摘要:
Russian Chemical Reviews 68 (7) 581 ± 604 (1999) N-Functionally substituted imines of polychlorinated (brominated) aldehydes and ketones G G Levkovskaya, T I Drozdova, I B Rozentsveig, A N Mirskova Contents I. Introduction II. Synthesis of polychlorinated (brominated) aldimines and ketimines N-substituted by carboxylic, carbamic, sulfonic and phosphoric acid residues III. Structure of imines of polyhalogenated aldehydes and ketones: syn ± anti isomerism; enamide ± acylimine tautomerism IV. Reactivity of N-acyl-, N-sulfonyl- and N-phosphorylimines of polyhalogenated aldehydes and ketones V. Conclusion Abstract. Published data on the synthesis, structure and chemical transformations of N-acyl-, N-alkoxy(aryloxy)carbonyl-, N-car- bamoyl-, N-sulfonyl- and N-phosphoryl-imines of chlorine-(and bromine)-containing ketones and aldehydes are surveyed and described systematically.The influence of polyhaloalkyl and functional substituents at the azomethine bond on their proper- ties is considered. The bibliography includes 198 references. I. Introduction The chemistry of imines presents interest for both theoretical and applied organic chemistry.1 ±21 On the one hand, azomethines are convenient models for the investigation of stereochemistry of organic compounds, tautomerism and isomerism phenomena, molecular rearrangements and mechanisms of radical, electro- philic and nucleophilic addition at the double bond and cyclo- addition.1 ±5 On the other hand, imines are used as the starting compounds for organic synthesis,6 for the synthesis of biologically active7 ±10 and natural products,5 monomers 1 and various nitro- genous heterocycles.1 ± 5, 11 In addition, compounds containing an azomethine group are known to play an important role in living organisms.1, 5 N-Functionally substituted polyhalogenated aldimines and ketimines present special interest due to their enhanced reactivity imparted by electron-withdrawing polyhaloalkyl substituents and by acyl, alkoxy(or aryloxy)carbonyl, -carbamoyl, sulfonyl or phosphoryl group at the N atom.The chemistry of polyhalogenated N-acyl aldimines and ketimines started to develop in the last two or three decades, because it was not until the early 1960s that preparative methods for their synthesis appeared. Quite a large number of review papers devoted to these interesting compounds have been pub- lished.The synthesis and properties of N-fluoro-, N-acyl-, N-sulfonyl- and N-phosphoryl-imines of polyfluorinated ketones G G Levkovskaya, T I Drozdova, I B Rozentsveig, A N Mirskova Irkutsk Institute of Chemistry, Siberian Branch of the Russian Academy of Sciences, ul. Favorskogo 1, 664033 Irkutsk, Russian Federation. Fax (7-395) 239 60 46. Tel. (7-395) 246 64 32. E-mail: admin@irioch.irk.ru Received 24 July 1998 Uspekhi Khimii 68 (7) 638 ± 662 (1999); translated by Z P Bobkova #1999 Russian Academy of Sciences and Turpion Ltd UDC 547.288 : 547.415.3 : 547.416 581 581 588 589 601 and aldehydes have been considered in the literature fairly comprehensively;12 ± 17 therefore, they will not be discussed here.However, known reviews 18 ± 25 dealing with the corresponding polychloro(bromo)-containing derivatives do not cover adequately the chemistry of N-acyl-, N-carboxy-, N-carbamoyl-, N-sulfonyl- and N-phosphoryl-imines of polyhalogenated alde- hydes and ketones. Therefore, it appears pertinent to generalise and describe systematically the data concerning the methods of synthesis and chemical transformations of N-acyl-, N-alkoxy(aryloxy)car- bonyl-, N-carbamoyl-, N-sulfonyl- and N-phosphoryl-imines of polychlorinated (brominated) aldehydes and ketones R1N=CXCHalnR2, where R1=Alk(Ar)CO, Alk(Ar)OCO, Alk2NCO, NH2CO, Alk(Ar)NHCO, Alk(Ar)SO2, Alk2NSO2, (AlkO)2P(O), (AlkO)2P(S), Ar2P(O), (ArO)2P(O); X=H, Ar, COOR3, CCl3, OAlk(Ar), CN, NR4R5, Cl, etc.; n=2, 3; Hal=Cl, Br; R2=Alk, ArAlk, etc.II. Synthesis of polychlorinated (brominated) aldimines and ketimines N-substituted by carboxylic, carbamic, sulfonic and phosphoric acid residues The known methods for the synthesis of acid N-poly- chloro(bromo)alkylideneamides can be divided into several groups differing in the approach to the formation of the azome- thine bond. 1. Preparation of N-acylated ketimines and aldimines from carboxamides and polyhalogenated aldehydes and ketones. 2. Syntheses based on amide derivatives�N-sulfinylsulfona- mides and N-sulfonyl isocyanates�and carbonyl compounds. 3. Successive transformation of a,b-polyhaloalkyl isocyanates into amides and azomethines upon reactions with nucleophiles.4. Creation of an azomethine bond by a free radical reaction of N,N-dihaloamides with 1,2-polyhaloethenes and alkynes. 5. Photochemical halogenation of N-alkyl(or aryl)amides. Other methods, which have not found wide application, are also known; they are considered at the end of this Section. 1. Preparation of polychlorinated (or brominated) aldimines and ketimines based on amides and carbonyl compounds Condensation of carbonyl compounds with amines is a classical method used to introduce an azomethine bond into a molecule. This is a two-stage process involving the intermediate formation of an amino alcohol.582 7O H + C N R C O+RNH2 H OH R C N C NR+H2O H However, in the general case, this reaction cannot be used to prepare imines of polyhalogenated compounds. The presence of electron-withdrawing groups in the molecule enhances the stabil- ity of the intermediate amino alcohol.Therefore, the rate of the synthesis of polyhalogenated aldimines and ketimines is limited by the second stage, that is, dehydration of hemiaminals. Thus it is known that condensation of carboxamides with aliphatic and aromatic aldehydes gives N,N-alkylidene- or N,N-arylidene-bis- amides,18 whereas polyhalogenated aldehydes and ketones react with amines and amides to give stable 1-hydroxyalkylamines.18, 26 In the case of aliphatic and aromatic amines and chlorinated aldehydes and ketones, the problem of transformation of amino alcohols into imines was solved by using TiCl4.27 For the synthesis of acyl-, alkoxy(or aryloxy)carbonyl- and -carbamoyl-, sulfonyl- and phosphoryl-imines of polyhalocar- bonyl compounds, methods based both on dehydration and on other chemical transformations of 1-hydroxypolyhaloalkyla- mides have been devised.a. Dehydration of N-(1-hydroxypolyhaloalkyl)amides In the earliest studies 28 ± 30 devoted to N-trichloroethylidene- amides, it was found that treatment of N-(1-hydroxy-2,2,2-tri- chloroethyl)acetamide, N-(1-hydroxy-2,2,2-trichloroethyl)chloro- (bromo)benzamides and N-(1-methoxy-2,2,2-trichloroethyl)bro- mobenzamides with dehydrating agents gives bis(1-acylamino- 2,2,2-trichloroethyl) ethers instead of the corresponding chloral imines.An example of fairly easy thermal dehydration giving rise to the target imines has been reported. Chloroacetamide reacts with chloral and bromal 31 yielding N-(2,2,2-trichloroethylidene)- and N-(2,2,2-tribromoethylidene)chloroacetamides 1a,b. 100 8C ClCH2CONH2+ CX3CHO ClCH2CON CHCX3 1a,b X=Cl (a), Br (b). Subsequently, synthetic expedients have been developed which permit dehydration of some chloral hemiaminals. N-(2,2,2-Trichloroethylidene)-4-toluenesulfonamide 2a has been prepared by dehydration of the corresponding N-(1-hydr- oxyalkyl) derivative induced by P2O5 at 200 8C in benzene or xylene.32 P2O5 4-MeC6H4SO2N CHCCl3 4-MeC6H4SO2NHCH(OH)CCl3 2a (71%) In our opinion, the sulfonamide derivative is esterified in this reaction by phosphoric anhydride and then phosphoric acid splits off.However, phosphorylimines cannot be prepared by dehydra- tion of N-(1-hydroxy-2,2,2-trichloroethyl) dialkyl phosphorami- dates.33 An attempt to dehydrate N-(1-hydroxy-2,2-di- chloroalkyl)amides was also unsuccessful.34 Apparently, the limited scope of the method for the synthesis of imines based on dehydration of hemiaminals of polyhalogen- ated aldehydes and ketones is due to the fact that these com- pounds are unstable in the presence of dehydrating agents and decompose to give the initial components or condensation products (esters) under the reaction conditions (see, for exam Refs 28 ± 30). G G Levkovskaya, T I Drozdova, I B Rozentsveig, A N Mirskova b.Dehydrochlorination of chloro(bromo)polyhaloalkylamides Dehydrochlorination of N-(1-haloalkyl)amides, which are pre- pared by chlorination of the corresponding hydroxy derivatives by SOCl2 or PCl5, is a fairly general method for the synthesis of acyl-, alkoxy- and -carbamoylimines of polyhalogenated carbonyl compounds as well as their sulfonyl and phosphoryl analogues. This synthetic route was first used 35 to prepare N-(2,2,2- trichloroethylidene)benzamide 3a and N-(2,2,2-trichloroethyl- idene)-N-benzyloxycarbonylamine 3b. PCl5 Et3N Cl3CCH(Cl)NHR Cl3CCH(OH)NHR Cl3CCH NR 3a,b R=PhCO (a), PhCH2OCO (b). The same method, except that thionyl chloride has been used as the chlorinating agent, has been employed to prepare trichlor- omethyl- and tribromomethyl-ethylidene-N-(alkoxycar- bonyl)amines,36 including various acyl- and alkoxy(aryloxy)- carbonylimines of chloral 37 ± 41 and bromal,42 acylimines and alkoxycarbonylimines of bromodichloroacetaldehyde,43, 44 and N-acetyl(benzoyl)imines and N-methoxycarbonylimines of 1,1,2- trichloro-2-phenylpropionaldehyde and 1,1,2-trichlorobutyralde- hyde,45 their yields ranging from 52% to 92%.Several N-(2,2,3- trichloropropylidene)amides and N-(2,2,3-trichloropropyli- dene)ethoxycarbonylamine have been prepared in the same way (yields 52%± 72%).46 SOCl2 RHalCH(OH)NHCOR RHalC(O)H+ RCONH2 Et3N NCOR RHalCH RHalCH(Cl)NHCOR RHal= CCl3, CBr3, CBrCl2, PhCHClCCl2, CH2ClCCl2, MeCHClCCl2, AlkCCl2; R=Alk, Ar, PhCH2, OEt, OMe.Synthesis of N-(2,2-dichloroalkylidene)acetamides by con- densation of dichlorinated aldehydes with acetamide followed by chlorination and then dehydrochlorination of the resulting com- pounds has been reported.34 In the same study, 34 dichloroethyli- denebenzamide was prepared in a low yield and an attempt to prepare 2,2-dibromoaldimines failed, probably due to the insta- bility of the intermediate hemiaminals formed from 2,2-dihaloal- dehydes and amides. The condensation of carboxamides with methyl trichloropyr- uvate gave the corresponding N-acylhemiaminals, which were then converted into methyl 2-acylimino- and 2-alkoxycarbonyli- mino-3,3,3-trichloropropionates 4a ± d.47 SOCl2 RCONHC(OH)COOMe RCONH2+Cl3CC(O)COOMe CCl3 Et3N RCONHC(Cl)COOMe RCON C(CCl3)COOMe 4a ± d CCl3 R=Me (a), Et (b), OMe (c), OEt (d).N-(2,2,2-Trichloroethylidene)-N0,N0-dimethylaminosulfon- amide has been synthesised by a traditional multistage procedure based on the condensation of chloral with N,N-dimethylamino- sulfonamide followed by chlorination by PCl5. Dehydrochlorina- tion of the intermediate compound was carried out on heating.48 PCl5 Cl3CCHO+ H2NSO2NMe2 Cl3CCH(OH)NHSO2NMe2 D Cl3CCHClNHSO2NMe2 Cl3CCH NSO2NMe2 5 Preparation of the phosphoryl analogues of N-acylimines of halogenated aldehydes was first reported in 1965.49 It was shown that the products of acidolysis of trichlorophosphazo com-N-Functionally substituted imines of polychlorinated (brominated) aldehydes and ketones pounds � N-polychloroalkylaminophosphoryl dichlorides con- taining chlorine atoms at the a-position with respect to nitro- gen � readily abstract an HCl molecule on treatment with triethylamine to give N-(2-chloro-2-methylpropylideneamino)- and N-(2,2-dichloro-3,3-dimethylbutylideneamino)phosphoryl dichlorides.49 Et3N RCHClNHPOCl2 RCH NPOCl2 6a,b R=Me2CCl (a), ButCCl2 (b). Several dialkyl N-(2,2,2-trichloroethylidene) phosphoramid- ates have been prepared in 50%± 76% yields from chloral and dialkyl phosphoramidates using chlorination and dehydrochlori- nation of intermediate products.33, 50 SOCl2 Cl3CCH(OH)NHP(O)(OR)2 Cl3CC(O)H + H2NP(O)(OR)2 Et3N Cl3CCH(Cl)NHP(O)(OR)2 Cl3CCH NP(O)(OR)2 7a ± e R=Me (a), Et (b), Prn (c), Pri (d), Bui (e). Dimethyl N-(2,2,2-trichloroethylidene) phosphoramido- thioate has been synthesised in a similar way.51 It has been shown 52, 53 that some tetrachloroalkanesulfon- amides can be dehydrochlorinated thermally without using dehydrochlorinating agents.For instance, N-trichloroethylidene- methane- and -propane-sulfonamides were synthesised in 73% and 93% yields on distillation of the corresponding tetrachloro- ethylalkanesulfonamides. Thus, this method is fairly versatile and permits the prepara- tion of various N-functionally substituted imines of polychlori- nated (or brominated) aldehydes and ketones. The drawbacks of this method include the large number of stages, the use of corrosive reagents and also the facts that the intermediate hemi- aminals are relatively unstable and haloaldehydes with different halogen atoms are inaccessible.c. Condensation of N-sulfinylsulfonamides and N-sulfonyl isocyanates with haloaldehydes Chloral N-4-tolylsulfonylimine 2a has been prepared in 91% yield by the reaction of chloral with unstable N-sulfinyl-4-toluenesul- fonamide in the presence of AlCl3;32, 54, 55 the starting amide was synthesised by the reaction of 4-toluene sulfonamide with thionyl chloride in the presence of AlCl3.56 N-(Trichloroethyl- idene)benzenesulfonamide 2b (yield 40%) and other chloral N-arylsulfonylimines (e.g., 2c,d) were prepared in a similar way.57 4-RC6H4SO2N S O+Cl3CCHO 7SO2 4-RC6H4SO2N CHCCl3 2a ± d R=Me (a), H (b), Cl (c), MeO (d).However, an attempt to prepare (N-2,2-dichloroethyli- dene)arenesulfonamides from N-sulfinyl-4-toluenesulfonamide and dichloroacetaldehyde 32 resulted in the isolation of 1,1-bis- (4-tolylsulfonylamino)-2,2-dichloroethane. 4-MeC6H4SO2N S O+Cl2CHCHO (4-MeC6H4SO2NH)2CHCHCl2 An attempt to synthesise 1-chloroimines by condensation of N-sulfinylarenesulfonamides with acid chlorides was also unsuc- cessful because the condensation products proved to be unstable under the reaction conditions.58 Thus, the applicability of this method is quite limited. For example, it is impossible to prepare N-trichloroethylidene dialkyl phosphoramidates by the condensation of the corresponding sulfinylamides with chloral.33 583 A general method for the synthesis of chloral sulfonylimines consists in the interaction of chloral with sulfonyl isocyanates in the presence of amines and quaternary ammonium bases.59 C16H33CH(NMe2)COMe Cl3CCHO+RSO2NCO Cl3CCH NSO2R 2e,g R=Cl (e), F (g), Alk(C1±C20), ArAlk(C7±C12), Ar(C6±C10). The method for the synthesis of acyl-, X-carbonyl- and phosphoryl-imines of polyhalogenated aldehydes and ketones based on condensation of N-derivatives of amides with carbonyl compounds has not been further developed. However, a large number of sulfonylimines of aromatic and aliphatic aldehydes and ketones 60 and N-tert-butyl- and N-fluoroalkyl-imines of alkyl, aryl, polyfluoro and chloro aldehydes and ketones have been prepared from sulfinylimines.61 2.Preparation of imines of polychlorinated aldehydes and ketones from isocyanates Methods for the synthesis of halo-containing imines based on 1-R- 1,2,2,2-tetrahaloethyl isocyanates and perchlorinated isocyanates have found wide application.Thus the reaction of 1-aryl-1,2,2,2-tetrachloroethyl isocya- nates with alcohols, phenols, benzenethiols and amines taken in equimolar amounts affords N-(1-aryl-1,2,2,2-tetrachloro- ethyl)amides. These compounds are dehydrogenated on heating or on treatment with bases. N-R-Oxycarbonyl-, arylthiocarbonyl- and carbamoyl-imines of trichloromethyl phenyl ketone, N-R- oxycarbonylimines of trichloromethyl 4-methylphenyl ketone and phenoxycarbonylimines of trichloromethyl aryl ketones were prepared in this way.62HX Cl3CCAr(Cl)NHCOX Cl3CCAr(Cl)NCO 7HCl Cl3CC(Ar) NCOX Ar=Ph: X=MeO, PhO, 4-MeC6H4O, 3,5-Me2C6H3O, 4-BrC6H4, O(CH2CH2)2N, Ph2N, PhNH, 3-CF3-4-NO2C6H3NH; Ar=4-MeC6H4: X=MeO, PhO; X=PhO: Ar=4-ClC6H4, 4-Cl3CC6H4, 4-FC6H4.N-Alkyl(aryl)-N0-(perchloroethylidene)ureas Later,63, 64 8a ± d were synthesised by the reaction of perchloroethyl iso- cyanate with primary amines at720 8C. RNH2 RNH2 [Cl3CCl2CNHC(O)NHR] Cl3CCl2CN C O Cl3CClC NC(O)NHR 8a ± d (20% ± 70%) R=Pri(a), But (b), Ph (c), 2,4,6-Me3C6H2 (d). Tetrachloroethyl isocyanate was prepared by treatment of N-(1-hydroxy-2,2,2-trichloroethyl)alkoxycarbonylamide with PCl5 3d pentachloroethyl isocyanate was obtained by its photochemical chlorination.63 The readily available ylide Ph3P=CHCN easily adds to 1,2,2,2-tetrachloroethyl isocyanate giving rise to a new biphilic reagent, which contains the clearly electrophilic group CClNHCO7, apart from the nucleophilic centre of the ylide fragment. This new reagent was used to create biphilic phospho- nium ylide 9.65 Cl3CCHClN C O+NCCH PPh3 Et3N Cl3CCH Cl3CCHClNHC(O)C PPh3 NC(O)C PPh3 CN CN 9 The reactions of 1,2,2,2-tetrachloroethyl isocyanate with orthoesters of carboxylic acids and with benzaldehyde aminals584 result in the formation of chloral N-alkoxycarbonylimines 10a,b and N,N0-substituted dialkoxyureas 11a,b.66 Cl COOR1 a Cl3CCH N Cl3CCHClNCO 7R1Cl,7R2COOR1 C(R2)OR1 R1O Cl3CCH NCOOR1 10a,b Cl CONR32b Cl3CCH N Cl3CCHClNCO 7PhCH=N+R32 Cl7 CHPh R32 N 2 Cl3CCH NCONR3 11a,b (a) (R1O)3CR2; (b) PhCH(NR32 )2; R1=Me (10a), Et (10b); R2=H, Ph, 4-ClC6H4; R32 =(CH2)2O(CH2)2 (11a); (CH2)5 (11b).It was found 66 that acetals do not enter into similar reactions with tetrachloroethyl isocyanate. 1,2,2,2-Tetrachloroethyl isocyanate reacts with (dialkyl- amino)alkoxymethanes and bis(dialkylamino)ethoxymethanes giving rise to the chloral N-alkoxycarbonylimines 10a,b and the N,N0-substituted ureas 11c,d.67 The reactions with dimethylfor- mamide acetals afford a mixture of chloral imines and 1-alkoxy- 2,2,2-trichloroethyl isocyanates. Et2NCH2OR1 Me2NCH(OR2)2 Cl3CCH NCOOR1 10a,b 10a,b +Cl3CCH(OR2)NCO Cl3CCHClNCO (R32 N)2CHOEt 2 Cl3CCH NCONR3 11c,d R3=Me (11c), Et (11d).Apart from tetrachloroethyl isocyanate, the products result- ing from replacement of one chlorine by an aryloxy or alkoxy group are also employed in the synthesis of imines. Thus the reaction of 1,2,2,2-tetrachloroethyl isocyanate and 1-aryloxy- 2,2,2-trichloroethyl isocyanate with trimethylsilylamines gives the N,N0-substituted ureas 11a,d.68Cl3CCH N CONR2 Cl3CCH(X)NCO +Me3SiNR2 7Me3SiX X SiMe3 Cl3CCH NCONR2 11a,d X=Cl, PhO, 4-ClC6H4O; R=Et, R2=(CH2)2O(CH2)2. The compound 11a was obtained in the reaction of diethyl(trimethylsilyl)amine with 1-alkoxytrichloroethyl isocya- nates; the first stage of the reaction yields silylated ureas, which decompose on heating to give the imine 11d and the corresponding alkoxysilanes.69 NCONEt2 Cl3CCH Cl3CCHNCO+Me3SiNEt2 7Me3SiOCH2X SiMe3 XCH2O OCH2X Cl3CCH NCONEt2 X=Cl2CH, H(CF2)2, H(CF2)4, COOMe.This method was used to synthesise a large number of acyl-, alkoxy(aryloxy)(thio)carbonyl- and N0-substituted carbamoyl- imines of chloral and trichloromethyl aryl ketones. G G Levkovskaya, T I Drozdova, I B Rozentsveig, A N Mirskova In conclusion, we would like to note that, despite the fact that these imines are produced from polychloroethyl isocyanates in preparative yields, isocyanates themselves are synthesised from polyhalogenated amides, nitriles, etc., using PCl5, COCl2 and other similar reagents, which restricts the scope of application of this method. 3. Reactions of N,N-dihaloamides with 1,2-polyhaloethenes as a new pathway to azomethines The reactions of 1,2-polyhaloethenes with sulfonic, carboxylic, alkoxycarboxylic and phosphoric acid N,N-dichloroamides have been vigorously studied for about 20 years.These reactions were found to give azomethines rather than saturated adducts. They provide the basis for one-stage procedures for the preparation of polyhaloaldehyde imines and their derivatives. Studies dealing with the synthesis of acyl-, ethoxycarbonyl- and sulfonyl-imines of chloral and other haloaldehydes, N-(2-bromo-2,2-dichloroethyl- idene)benzenesulfonamide, N-(2,2-dibromo-2-chloroethylidene)- benzenesulfonamide and N-(2,2-dichloroethylidene)arenesulfon- amides from N,N-dichloro- and N,N-dibromo-amides and 1,2-polyhaloethenes (trichloro-, tribromo- and dichloro-ethenes) published before 1989 have been surveyed in several reviews.23 ± 25 Here we consider the studies that have not been covered in these reviews.For instance, when N,N-dichloronitrobenzenesulfonamide, aminobenzenesulfonamide andN,N-dichloroamides of perfluoro- alkanesulfonic acids are made to react with trichloroethene (TCE), the chloral sulfonylimines 2e ± h are formed;70, 71 the reaction of N,N-dichlorourethanes with TCE affords chloral N-alkoxycarbonylimines 10a ± c in preparative yields.72, 73 CCl2 RSO2NCl2+2CHCl 7Cl2 Cl3CCH NSO2R+Cl2CHCCl3 2a ± c, e ± h R=4-MeOC6H4 (a), Ph (b), 4-ClC6H4 (c), 3-NO2C6H4 (e), 4-NO2C6H4 (f), 4-MeOCONHC6H4 (g), AlkF (h). Cl2NCOOR +2CHCl CCl2 7Cl2 Cl3CCH NCOOR+Cl2CHCCl3 10a ± c R=Me (a), Et (b), Bu (c).The reactions of TCE with N,N-dichloroarenesulfonamides in the presence of Lewis acids (SnCl4, AlCl3) also afford the sulfonylimines 2a ± c, which are formed, however, in lower yields (up to 47%); N-(1-arylsulfonylamino-2,2,2-trichloroethyl)arene- sulfonamides are formed as side products.74 It has been reported previously 23 that TCE reacts with N,N- dibromobenzenesulfamide orN,N-dibromourethane giving rise to the corresponding imines of 2-bromo-2,2-dichloroacetaldehyde BrCl2CCH=NSO2Ph and BrCl2CCH=NCOOMe. The reaction of TCE with N-bromourethane, which is known to disproportionate to give amide and dibromoamide, results in the formation of 2-bromo-2,2-dichloro-1,1-di(ethoxycarbonyl- amino)ethane.75 The reactions of N,N-dichlorourethane and N,N-dichloro- benzenesulfonamide with tribromoethene (TBE) follow a similar route.N,N-Dichlorourethane is converted into ethyl N-(2,2- dibromo-2-chloroethylidene)carbamate 12.76 The reaction is accompanied by evolution of chlorine, bromine and 1,1,2- tribromo-1,2-dichloroethane. EtOCONCl2+CHBr CBr2 7Cl2,7Br2 EtOCON CHCClBr2+CHClBrCClBr2 12 It has been reported in studies cited in a review 23 that N,N- dichloroarenesulfonamides react with 1,2-dichloroethene (DCE) to give N-(2,2-dichloroethylidene)arenesulfonamides 13a ±c; the formation of these products was proved by physico-chemical methods and by chemical transformations. In another study,77 they were isolated and characterised for the first time.The relativeN-Functionally substituted imines of polychlorinated (brominated) aldehydes and ketones yields of the compounds 13, 14 and 2 depend on the reaction conditions.76, 77 RNCl2+CHCl CHCl 2a ± c 14a ± c RN CHCHCl2+ (RNH)2CHCHCl2+ RN CHCCl3 13a ± c R=PhSO2 (a), 4-MeC6H4SO2 (b), 4-ClC6H4SO2 (c). The reaction ofN,N-dichlorourethane with DCE was found 76 to afford highly reactive ethyl N-(2,2-dichloroethylidene)carb- amate 15. However, like the reaction of N,N-dichloroarenesulfo- namides with DCE, this process yields the imine 10b and the product of addition of dichlorourethane to the C=Nbond of this imine, namely, 2,2,2-trichloro-1,1-bis(ethoxycarbonylamino)- ethane 16, together with the target imine 15.This is due to the fact that the process is accompanied by the chlorination of DCE giving rise to trichloroethene, tetrachloroethane, tetrachloro- ethene and urethane. EtOOCNCl2+CHCl CHCl EtOOCN CHCHCl2+ 15 +(EtOOCNH)2CHCHCl3 +EtOOCN CHCCl3 10b 16 N,N-Dichloroarenesulfonamides have been found to react readily with 1-bromo-1,2-dichloroethene yielding N-(2-bromo- 2,2-dichloroethylidene)arenesulfonamides 17a,b.78 ArSO2N CHCCl2Br ArSO2NCl2+CHCl CClBr 17a,b Ar=Ph (a), 4-ClC6H4 (b). Systematic studies dealing with the influence of the natures of the polyhaloethene and dihaloamide used on the structure of the imine RN=CHCClnBr37n made it possible to refine the scheme proposed previously23 for this reaction. It was demonstrated that the radical adducts formed either from N,N-dibromoamides and polychloroethenes or from N,N-dichloroamides and TBE do not undergo 1,2-halotropic rearrangement because these reactions afford neither polychloroaldehyde imines nor bromal imines.23, 75 ± 78 The method for the synthesis of polyhaloaldehyde imines based on the reaction of amidyl radicals with polyhaloethenes permits polychlorinated (or polybrominated) azomethines includ- ing those with different halogen atoms to be prepared from available initial compounds and to be used subsequently without isolation for synthetic purposes.However, the process of the preparation of halo-containing acyl- and phosphoryl-imines requires optimisation. The applicability of this method to the preparation of carbamoylimines and fluorinated imines contain- ing residues of various types of acids also has to be studied.4. Synthesis of N-arylsylfonylimines of polyhaloaldehydes by the reaction of N,N-dihaloamides with acetylene derivatives The method for the synthesis of imines from acid dichloroamides and acetylene derivatives was developed in the last decade. It has been shown 79, 80 that the reaction of N,N-dichloroar- enesulfonamides with phenylacetylene occurs by a free radical mechanism and results in the formation of N-(2,2-dichloro-2- phenylethylidene)arenesulfonamides 18a,b. A detailed study of the process 80 demonstrated that N,N-dichlorobenzenesulfona- mide reacts with phenylacetylene giving rise to a mixture of compounds, the imine 18a being the major product (yield up to 75%); adduct 19a, resulting from the addition of benzenesulfon- amide to the C=N bond of 18a, is also formed in a noticeable amount.585 ArSO2N CHCCl2Ph 18a,b ArSO2NCl2+PhC CH (ArSO2NH)2CHCCl2Ph 19a [ArSO2NClCH CClPh] ArSO2NHCH CClPh Ar=Ph (a), 4-ClC6H4 (b). The reaction of phenylacetylene with N,N-dibromoarenesul- fonamides gives 2,2-dibromo-2-phenyl-1,1-di(phenylsulfonyl- amino)ethane and benzenesulfonamide.80 The reactions of N,N-dichlorourethanes and N,N-dichloro- amides with phenylacetylene have also been studied.81, 82 How- ever, the results obtained are contradictory and require more precise determination of the structures and transformations of the reaction products. The reaction of N,N-dichloroarenesulfonamides with prop- argyl alcohol and 3-chloroprop-1-yne 83, 84 gave the products of the subsequent addition of arenesulfonamides to the imines formed, namely, 3-chloro- and 3-hydroxy-2,2-dichloro-1,1- di(phenylsulfonylamino)propanes or 1,1-di(4-chlorophenylsulfo- nylamino)propanes.(ArSO2NH)2CHCCl2CH2X ArSO2NCl2+CH CCH2X X=Cl, OH; Ar=Ph, 4-ClC6H4. 5. Chlorination and bromination of amide derivatives Among the methods for the synthesis of polyhaloaldehyde imines, mention should be made of chlorination of N-alkyl(aryl)amides; however, this method has already been considered in several reviews.17 ± 21 We would only like to emphasise that chlorination ofN-acylamides has been shown 85 to give products resulting from chlorination of both the aldehyde and the amide fragments.Photochemical high-temperature chlorination of isopropyl- amides of benzoic and 4-chlorobenzoic acids affords the corre- sponding acylimines of hexachloroacetone. The transformation of N-(1-phenyl-2,2,2-trichloroethyl)- amides under conditions of photochemical chlorination 86 and the transformation of N-(1-diethoxyphosphoryl-2,2,2-trichloro- ethyl)benzamide on treatment with the chlorine complex of pyridine 87 yield N-(1-phenyltrichloroethylidene)trichloro- acet(or benz)amides 19a,b 86 and N-(1-diethoxyphosphoryltri- chloroethylidene)benzamide 19c (yield 68%).87 It was shown 87 that the latter product is more convenient to prepare by the chlorination of the silyl derivative CCl2=C[P(O)(OEt)2]. .N(SiMe3)COPh (yield 84%).Cl2, hn CCl3CCl(X)NHCOR CCl3CH(X)NHCOR D 7HCl CCl3C(X) NCOR 19a ± c X=Ph, R = CCl3 (a, 58%), Ph (b, 62%); X=(EtO)2P(O), R=Ph (c). Yet another study deserves special attention88 in which it was reported that N-acetyl-(1-tert-butoxycarbonyl-2,2-dibromo)- propan- and -butanimines 20a,b are formed upon successive bromination of the corresponding enamides withN-bromosuccin- imide (NBS). NBS MeC(O)NHC CHR MeC(O)NBrC CHR COOBut COOBut NBS MeC(O)NBrC CBrR MeC(O)NHC CBrR COOBut COOBut586 MeC(O)N CCBr2R COOBut 20a,b R=Me (a), Et (b). Finally, some polyhalogenated ketimines have been prepared by halogenation of compounds with C=N bonds. Thus N-perchlorovinylcarboximidoyl chlorides were used to synthesise N-substituted imino esters with a perchlorovinyl group.89 When these compounds are subjected to chlorination, the perchlorovinyl group is converted into a perchloroethyl group.Methyl N-perchloroethylbenzimidates eliminate chloromethane on heat- ing to give N-(tetrachloroethylidene)benzamides 21a ± c in high yields. Bromination of esters of N-perchlorovinylimidic acids under mild conditions is accompanied by elimination of MeBr and gives rise to N-(2-bromotrichloroethylidene)benzamides 22a ± c. MeONa Cl2C CClN C(OMe)R CClN CClR Cl2C Cl2 C(OMe)R Cl3CÊ l2CN Cl3CClC 7MeCl NC(O)R 21a ± c Br2 C(OMe)R] [Cl2(Br)CCClBrN Cl2(Br)CClC 7MeCl NC(O)R 22a ± c R=4-XC6H4; X = H (a), Cl (b), NO2 (c).Chlorination has been employed to prepare imines of cyclic halogen-containing ketones.90 Thus chlorination of both N,N0- bis(arylsulfonyl)-1,4-benzoquinonediimines 23 and N-arylsul- fonyl-substituted p-phenylenediamines 24 in dimethylformamide results in the isolation of the same products, 2,3,5,6-tetrachloro- N,N0-bis(arylsulfonyl)-1,4-phenylenediamines 25. Further chlori- nation affords tetrachlorobis(arylsulfonyl)benzoquinone- diimines 26, which, in turn, add a chlorine molecule at a double bond of the quinone ring to give the final products, 1,4-bis(arylsulfonylimino)-2,3,5,5,6,6-hexachlorocyclohex-2-enes 27a ± c.90 Cl Cl2 NHR RHN NHR RHN Cl2 24 Cl NR RN 23 Cl Cl Cl Cl Cl2 NR RN NHR RHN Cl Cl Cl Cl 25 26 Cl Cl Cl Cl NR RN Cl Cl 27a ± c R=PhSO2 (a), 4-MeC6H4SO2 (b), 4-ClC6H4SO2 (c).Similarly, 4-arylsulfonylimino-2,3,5,5,6,6-hexachlorocyclo- hex-2-en-1-ones 28a ± c 91 and 4-arylsulfonylimino-2,3,5,5,6,6- hexachloro-1-oxo-1,2,3,4-tetrahydronaphthalenes 29a ± c 92 were obtained by chlorination ofN-arylsulfonyl-1,4-aminophenols and N-arylsulfonyl-1,4-aminonaphthols. G G Levkovskaya, T I Drozdova, I B Rozentsveig, A N Mirskova Cl Cl Cl Cl Cl Cl Cl Cl O ArSO2N O ArSO2N Cl 29a ± c Cl 28a ± c Ar=Ph (a), 4-MeC6H4 (b), 4-ClC6H4 (c). 6. The use of organophosphorus compounds and PCl5 for the synthesis of polyhalogenated imines Several methods for the synthesis of N-functionally substituted polyhalogenated imines based on organophosphorus compounds and PCl5 are known.Dialkyl N-(1-amino-2,2-dichloroethylidene) phosphorami- dates were prepared by the reaction of trichloroacetylamidines with trialkyl phosphites.93 Cl3CC(NR1R2) NH+P(OR3)3 [Cl2C C(NR1R2)NHP(O)(OR3)2] NP(O)(OR3)2 Cl2CHC(NR1R2) 30a ± e R1=H,R2=Me: R3=Me (a), Et (b), Prn (c), Pri (d); R1=H,R2=But, R3=Et (e). Diethyl N-(1-dimethylamino-2,2,2-trichloroethylidene) phos- phoramidate 31 was synthesised from N-chloro-N0,N0-dimethyl- trichloroacetylamidines and triethyl phosphite.93 Cl3CC(NMe2) NCl+P(OEt)3 Cl3CC(NMe2) NP(O)(OEt)2 31 Trialkyl phosphites were shown 94 to react with polychloroni- trosoethanes under mild conditions (7208C) giving rise to the corresponding dialkyl N-chloroalkylidene phosphoramidates 32a ± h.720 8C (R2O)2P(O)N CClR1 O NCCl2R1 + (R2O)3P 7RCl 32a ± h R1=CH2Cl: R2=Prn (a), Bun (b), Bui (c), C5H11 (d), ClCH2CH2 (e); R1=CHCl2: R2=Prn (f), Bun (g), C5H11 (h). The reaction of N-trihaloacetyl diphenylphosphinic amide with PCl5 affords a mixture of N-(tetrachloroethylidene) (33a) and N-(trifluoro-1-chloroethylidene) (33b) diphenylphosphinic amides (70%) and phosphazo compounds 34 (30%).95 The rearrangement, which occurs presumably via a four-centre tran- sition state, is facilitated by the presence of strong electron- withdrawing groups. PCl5 X3CC(Cl) NP(O)Ph2 X3CC(O)NHP(O)Ph2 33a,b Cl X3CC(O)N P(Cl)Ph2 X3C C N 34a,b O PPh2 X=Cl (a), F (b). However, trichloroacetamide reacts with excess PCl5 to give N-dichlorophosphoryltetrachloroethanimine 35.96 This product reacts with sodium phenoxide giving rise to the products of substitution of phenoxyl residues for the halogen atoms � N-diphenoxyphosphoryl-2,2,2-trichloro-1-phenoxyethanimine 36.97N-Functionally substituted imines of polychlorinated (brominated) aldehydes and ketones PhONa PCl5 Cl3CCONH2 NPOCl2 Cl3CClC35 Cl3CC(OPh) NPO(OPh)2 36 Imines of type 35 have also been used to prepare chlorinated and fluorinated imides of phosphorothioic acids 37a ± d.98 P2S5 RClC NP(O)Cl2 RClC NP(S)Cl2+ P2O5 37a ± d R=CCl3 (a), CH2ClCCl2 (b), MeCCl2 (c), EtCCl2 (d).When the imines 37a,c are made react with alcohols and amines, the chlorine atoms at phosphorus and carbon are substituted to give compounds 38a ± e and 39a ± f.98 R2R3NH NP(S)(NR2R3)2 R4OH, Et3N R1CCl NP(S)Cl2 37a,c R1C(OR4) R1C(NR2R3) 38a ± e NP(S)(OR4)2 39a ± f R1=Cl3C: NR2R3=PhNH (38a), 4-ClC6H4NH (38b), N(CH2CH2)2O (38c); R1=Cl2CMe: NR2R3=4-ClC6H4NH (38d), N(CH2CH2)2O (38e); R1=Cl3C: R4=Me (39a), Et (39b), Pr (39c), Bu (39d), Ph (39e); R1=Cl2CMe, R4=Et (39f). Several N-phosphorylimines have been synthesised using trichlorophosphorazoperchloroethane 40,99 which reacts with alcohols to yield imides of various structures depending on the reaction temperature and the reactant ratio. When an equimolar amount of an alcohol is employed, one chlorine atom is replaced giving rise to unstable alkyl N-(perchloroethyl)dichlorophos- phorimidates, which decompose on heating with evolution of an alkyl chloride to give N-dichlorophosphoryltetrachloroethan- imine 35.D Cl3CCl2CN PCl2(OR) 7RCl Cl3CCl2CN PCl3+ ROH 40Cl3CClC NPOCl2 35 R=Me, Et, Pr, Bu, C8H17. When a threefold excess of an alcohol is used, this reaction results in the formation of a mixture of dialkyl phosphoramidates 41a ± h and the corresponding alkoxy chloro derivatives 42a ± h in *60 : 40 ratio. 3ROH Cl3CClC NPO(OR)2+ Cl3CC NPO(OR)Cl 40 73HCl 42a ± h 41a ± h OR R=Me (a), Et (b), Prn (c), Bun (d), Bui (e), C5H11 (f), iso-C5H11 (g), C8H17 (h). When the alcohol : imine 40 ratio is 4 : 1, dialkyl N-(1-alkoxy- 2,2,2-trichloroethylidene) phosphoramidates 43 are produced in 65%± 85% yields. These compounds can also be obtained by treatment of the compounds 35, 41 or 42 with an alcohol.3ROH Cl3CClCN POCl2 35 ROH Cl3CC NPO(OR)2 Cl3CClC NPO(OR)2 41 OR 43 ROH Cl3CC NP(O)Cl(OR) OR 42 587 The imine 41a was converted into dimethyl N-trichloroacetyl phosphoramidate 44 using a procedure proposed by Kirsanov et al.96 The product 44 can exist in solution as the imine tautomer 45.100H2O Cl3CCNHP(O)(OMe)2 Cl3CC NP(O)(OMe)2 41a 44 O OH Cl3CCN P(OMe)2 OH 45 The compound 6a can be prepared by treatment of N-(1,2- dichloro-2-methylpropylidene)aminophosphorus trichloride with sulfur dioxide.49 SO2 Me2CClCH NP(O)Cl2 6a Me2CClCHClN PCl3 46 Tetrachloroethylideneaminodioxaphospholane 48 was pre- pared by the reaction of 2-amino-2-trichloromethyl-5,6-dichloro- benzo-1,3-dioxolane 47 with PCl5.101 Presumably, a phosphazo compound is formed initially and then it rearranges into the dioxaphospholane via a seven-membered intermediate.Cl Cl O O PCl5 NH2 N PCl3 CCl3 72HCl CCl3 O O Cl Cl 47 O Cl Cl CCl3 O PCl3 N PN O Cl Cl Cl O CCl3 48 7. Other methods of synthesis It is worth noting that chlorovinylamides, structural isomers of chloroethylideneimines, can be used in some cases for the syn- thesis of imines. Thus bis(dialkoxyphosphoryl)trichlorovinyl- amines react with two equivalents of a dialkylamine to give dialkyl N-(1-dialkylamino-2,2-dichloroethylidene) phosphorami- dates 49a,b.102 Cl3CClC NPO(OEt)2+P(OEt)3 R2NH Cl2CHC(NR2) NPO(OEt)2 Cl2C CClN[PO(OEt)2]2 49a,b R=Me (a), Et (b).It has been mentioned above that imines with a similar structure have also been prepared from trichloroacetamidines.93 (2,2-Dichloro-1-cyanovinyl)benzenesulfonamide 50 reacts with amines, in the opinion of the researchers cited,103 in the imine form with replacement of the cyano group by an amino group; N-(1-dialkylamino-2,2-dichloroethylidene)arenesulfon- amides 51a ± e were isolated in good yields. HNR1R2 [Cl2CHC(CN) NSO2Ph] Cl2C C(CN)NHSO2Ph 50 Cl2CHC(NR1R2) NSO2Ph 51a ± e NR1R2=PhCH2NH (a), PhNH (b), 4-MeC6H4NH (c), Me2N (d), O(CH2CH2)2N (e). Aquite promising method for the preparation of imines is that based on ethyl N-[trichloro-1-alkoxy(aryloxy)ethyl]carbamates and chlorotrimethylsilane.When ethyl N-(1-aryloxy-2,2,2-tri- chloroethyl)carbamates are silylated with chlorotrimethylsilane in the presence of triethylamine, the intermediate product decom- poses to give N-(ethoxycarbonyl)chloral imine 10b.69588 Me3SiCl, Et3N Cl3CCH(OR)NHCOOEt D Cl3CCH NCOOEt 7ROSiMe3 Cl3CCH NCOOEt 10b RO SiMe3 R=Ph, 2,4-Br2C6H3. It has been shown that 1-phenyl-2,2,2-trichloroethylidene- ureas 52 can be prepared from phenyl N-(1-phenyl-2,2,2-trichlo- roethylidene)carbamate 53.62 Under rigorous conditions, the phenoxy group is replaced by the amine residue.62 HNR2, D Cl3CC(Ph) NCONR2 52 Cl3CC(Ph) NCOOPh 53 The reaction of N,N-dichlorobenzenesulfonamide with dichlorovinyl ketones followed by dechlorination of the resulting N-(1-acyl-2,2,2-trichloroethyl)benzenesulfonamides on treatment with alkali resulted in the formation of N-(1-acyl-2,2-dichloro- ethylidene)benzenesulfonamides 54a ± c. They can exist as two tautomers.104 OH7 PhSO2NCl2+RCOCH CCl2 PhSO2NHCH(COR)CCl3 PhSO2NHC(COR) CCl2 PhSO2N C(COR)CHCl2 54a ± c R=Me (a), Pr (b), CH2Cl (c).III. Structure of imines of polyhalogenated aldehydes and ketones: syn ± anti isomerism; enamide ± acylimine tautomerism 1. Structure of N-trichloroethylidenearenesulfonamides according to the data of 35Cl NQR and NMR spectroscopy. syn ± anti Isomerism of polyhaloaldehyde imines Despite the considerable attention devoted to the chemistry of carboxylic, sulfonic and phosphoric acid N-polyhaloalkylidene- amides, their structures certainly have not been adequately studied.Most often, these compounds have been characterised by IR and NMR data; the results of NQR studies of trichloro- ethylidenearenesulfonamides can be found only in two publica- tions.105, 106 The 35Cl NQR spectra, as well as 1H and 13C NMR spectra,70 indicate that chloral arylsulfonylimines 2a ± c occur as one of the two possible isomers. According to quantum-chemical calcula- tions, anti-isomer A is more favourable than syn-isomer B; this is confirmed by experimental and theoretical 35Cl NQR data.105 ArSO2 ArSO2 H CCl3 N C N C H CCl3 B A Ar=4-MeC6H4, Ph, 4-ClC6H4. Br2ClCCH NSO2Ph 55 Br2ClCCH NCOOEt 12 It has been shown by NMR spectroscopy that the dibromo- chloroacetaldehyde imines 12 and 55 exist as mixtures of syn- and anti-isomers in 3 : 4 107 and 1 : 3 76 ratios, respectively. Indeed, the 1H NMR spectrum of the imine 55 exhibits, in addition to the multiplet for the phenyl protons, two singlets at 8.30 and 8.40 ppm due to the protons at the double bond, the ratio of their intensities being 3 : 4.The 13C NMR spectrum contains two signals for the carbon atoms of theCH=Ngroup (167.9 and 167.6 ppm) and the CClBr2 group (135.7 and 135.6 ppm).107 More evidence support- ing the existence of syn- and anti-isomers of imines is provided by the fact that addition of nucleophiles at the C=N bond affords two enantiomers.107 G G Levkovskaya, T I Drozdova, I B Rozentsveig, A N Mirskova It should be noted that syn- and anti-isomers cannot be detected in the 1H NMR spectra of chloral sulfonylimines 2a ± c or ethoxycarbonylimines 10b,108, 73 dichloroacetaldehyde sulfo- nylimines 13a ± c or ethoxycarbonylimines 15,76, 77 or bromodi- chloroacetaldehyde sulfonylimines 17a,b or alkoxycarbonyl- imines.23, 78, 109 The correlations between the experimental 35Cl NQR fre- quencies and the C± Cl bond lengths proposed in Ref.110 were used to calculate the bond lengths in the imines 2a ± c, which were estimated as ranging between 175.8 and 176.3 pm.105 These values are typical of C± Cl bond lengths in compounds with a trichloro- methyl group. It has been noted that the three chlorine atoms in trichloro- ethylidene-benzenesulfonamide and -4-toluenesulfonamide 2a,b are nonequivalent in the 35ClNQRspectra, whereas the 35ClNQR spectrum of the corresponding 4-chlorophenylsulfonylimine 2c exhibits a single signal for the trichloromethyl group,105, 106 which is difficult to explain based on its molecular structure.Apparently, the electron environment of all the chlorine atoms in this imine becomes equivalent due to intermolecular contacts. Unfortunately, no data of 35Cl NQR spectra for imines that are known to exist as both syn- and anti-isomers can be found in the literature. 2. Enamide ± acylimine tautomerism in dichloroethylideneimines An interesting problem of theoretical chemistry is the question of the existence of enamide ± acylimine tautomerism in haloalkylide- neimines N-substituted by acid residues.R1CXHal CH NR2 R1CHal CH NXR2 X=H, Hal. It has been concluded in reviews 20, 21 that, among N-acyl- imines of chloro(bromo)(fluoro)aldehydes and ketones, stable structures with the imine bond can be found for ald- and ketimines containing no b-hydrogen atoms in the haloalkyl groups. However, now it is known that arylsulfonylimines of dichloro- acetaldehyde ArSO2N=CHCHCl2 13a ± c, Ar=Ph (a), 4-MeC6H4 (b), 4-ClC6H4 (c), which do have a hydrogen atom in the b-position with respect to the imine group, exist in the stable imine form, as indicated by the presence of characteristic absorp- tion bands of the C=N bond in the IR and NMR spectra (the signal for the azomethine proton at d 8.35 ± 8.44 ppm and for the carbon atom of the CH=N group at d 166 ppm).77 Similarly, the dichloroethylidenecarbamate 15 (the signal of the azomethine proton at d 8.06 ppm 76) and 1-alkylamino-substituted dialkyl 2,2-dichloroethylidene phosphoramidates also exist as stable imine forms.93, 102 According to the data of IR and 1H NMR spectra, N-dichlorovinylamides and N-dichlorovinylthioamides exist as stable enamide forms.111, 112 Nevertheless, they react with nucleo- philes to give stable products of addition at the azomethine bond.112 ± 114 Some functional substituents, for example CN,103, 115, 116 CONH2,117 POPh2,118 PO(OEt)2 119, 120 and P+Ph3Cl7,121 present in a dichloroalkyl group stabilise the enamide form Cl2C=C(R1)NHCOR2, where R2=Alk, Ar, OMe, OPh.These amides react with primary and secondary amines in the enamide tautomeric form to give the products of replacement of chlorine atoms.An exception is provided by N-(1-carbamoyl-2,2-dichloro- vinyl)benzamide 56, which reacts with methyl- and ethylamine giving rise to stable addition products 57.117 Cl2CHC(CONH2)NHCOPh Cl2C C(CONH2)NHCOPh +RNH2 56 NHR 57 R=Me, Et.N-Functionally substituted imines of polychlorinated (brominated) aldehydes and ketones It has already been mentioned (see Section II.7) that N-(2,2- dichloro-1-cyanovinyl)benzenesulfonamide 50 reacts with amines with substitution of the cyano group to afford stable 1-amino-2,2- dichloroethylideneamides 51a ± e.103 It has also been found that N-(1-acyl-2,2-dichloroethyl- idene)benzenesulfonamides 54a ± c can exist as two tautomers;104 the amide form predominates in the solid state, whereas in solutions the equilibrium shifts towards the imine form.The facts that the equilibrium is established over a relatively long period and that the tautomers can be observed separately indicate that the energy barrier separating these forms is higher than 107 kJ mol71.104 It should be emphasised that bromination of the NH group in N-(1-tert-butoxycarbonyl-2-methyl- and 2-ethylvinyl)acetamides yields unstable compounds which undergo a 1,3-halotropic rearrangement with migration of the halogen atom giving rise to the imines 20a,b (see Section II.4). Similarly, N-chloro-N- (2-chloro-2-phenylvinyl)arenesulfonamides and N-bromo-N-(2- bromo-2-phenylvinyl)benzenesulfonamide, formed in the reac- tion of N,N-dichloro(bromo)arenesulfonamides with phenylace- tylene (see Section II.4), are converted into the corresponding N-(2,2-dihalo-2-phenylethylidene)arenesulfonamides of the type 18 as a result of 1,3-halotropic rearrangement.79, 80 Thus, the imine form of dichloroacetaldimines is stable in the case of arylsulfonylimines and alkoxycarbonylimines and for N-dichlorovinyl phosphoric amides if the a-position of the vinyl group has electron-donating substituents.N-Dichlorovinyl-sub- stituted dialkyl phosphoramidates, carboxamides and thiocar- boxamides containing no substituents at the a-position of the vinyl group are stable in the enamide form; however, they react with nucleophiles in the imine form.The presence of electron- withdrawing substituents at the 1-position of the vinyl group of dichlorovinylamides stabilises the enamide tautomer. The intro- duction of a halogen atom in the NH group facilitates tautomer- isation of enamides to give halogen-containing imines. IV. Reactivity of N-acyl-, N-sulfonyl- and N-phosphorylimines of polyhalogenated aldehydes and ketones The reactivity of N-functionally substituted polyhalogenated aldimines and ketimines is due to the presence of the highly electrophilic C=N bond. Electron-withdrawing substituents at this bond increase its reactivity towards nucleophiles and electron- rich species such as dienes, alkenes, alkynes or arenes. 1. Reactions of polychlorinated(brominated) aldimines and ketimines with nucleophiles Acyl-, alkoxy(aryloxy)(amino)carbonyl-, sulfonyl- and phos- phoryl-imines of polyhalogenated aldehydes and ketones have been studied in reactions with O-, N-, S- and P-nucleophiles and bifunctional nucleophiles.To elucidate the relationship between the structure of imines and their reactivity towards nucleophiles, the kinetics have been measured for the addition of ethanethiol and 2,4,6-trichloroaniline to chloral imines with various acyl residues, Cl3CCH=NCOR, where R=Ph (3a), PhCH2O (3b), Me (3c), Et (3d), Pri (3e), But (3f), 4-MeC6H4 (3g), 4-MeOC6H4 (3h), 4-ClC6H4 (3i), 4-O2NC6H4 (3j), 3-MeC6H4 (3k), 3-ClC6H4 (3l), and to the imines 10a,b, 2a,b and 7b.41, 122 It was found that the reactions obey second-order kinetics, the reactivity of imines sharply decreasing in the series 3>10>7. Cl3CCH=NR R 7b (EtO)2PO 4.7 10b EtOCO 14.6 3d EtCO 77.0 3a PhCH2CO k /litre mol71 h71 226.0 The rate constants for the addition of ethanethiol diminish by factors of 20 to 52 on passing from the acylimines 3 and 10 to the phosphorylimine 7.Evidently, this is due to the decrease in the 589 electronegativity of the acid residues, resulting in a higher electron density on the carbon atom of the azomethine group. The substantial decrease in the reactivity observed on passing from N-ethoxycarbonylchloral imine 10b to N-propionylchloral imine 3d is due to the decrease in the conjugation of the azomethine group with the carbonyl group involving the alkoxy residue.The considerable influence of conjugation is also supported by the fact that the reactivity of an imino group attached to a phosphonyl residue is even lower because in this case, conjugation is insignif- icant if at all present.122 Study of the reactivity of the chloral aroylimines 3i ±m towards 2,4,6-trichloroaniline in benzene, dioxane and acetoni- trile demonstrated that the rate of addition of the amine in acetonitrile depends on the substituents in the benzene nucleus.41 This reaction is accelerated by electron-withdrawing substituents and retarded by electron-donating ones. Comparison of the reactivity of chloral imines towards 2,4,6- trichloroaniline 41 demonstrated that the influence of the substitu- ent at the nitrogen atom on the capacity of the C=N bond for the addition reactions decreases in the sequence COAr>COAlk>SO2Ar and also in the sequence COAlk>PO(OAlk)2>COOAlk. Although arylsulfonamide groups possess high electronega- tivity, they have a lesser influence on the reactivity of the C=N bond towards trichloroaniline.The addition of the amine to phosphoryl- and sulfonylimines is characterised by greater neg- ative values of activation entropy than the addition to acylimines. This was interpreted 41 by assuming that the reactions of the amine with the C=N±C=O, C=N±P=O and C=N± SO2 systems occur via different cyclic transition states. In the case of acyl- imines, a six-membered transition complex is formed, whereas for phosphoryl- and sulfonylimines a four-membered transition state is involved.This assumption is confirmed by the known fact that acylamines enter into 1,4-cycloaddition reactions, whereas aryl- sulfonylimines tend to undergo 1,2-cycloaddition. a. Reactions of imines with O-nucleophiles The reactions of imines with oxygen-containing nucleophiles have been studied most extensively; it was shown that acyl-, alkoxy(aryloxy)carbonyl-, carbamoyl-, sulfonyl- and phos- phoryl-imines of polyhaloaldehydes add water and alcohols with heat evolution to give stable hydroxy- 58 and alkoxy-substituted 59 haloalkylamides. RN CHCXHal2+H2O RNHCH(OH)CXHal2 58 Ref. Hal X R 37 35 65 36 36, 72 72 35 68 Cl Cl Cl Cl Cl Cl Cl Cl Cl Cl Cl Cl Cl Cl Cl Cl MeCO PhCO Ph3P=C(CN)CO MeOCO EtOCO BuOCO PhOCO Et2NCO O(CH2CH2)2N Cl Cl 6832, 74, 123, 124 74, 123, 124 74, 123, 124 70 70 33, 50 33 33 33 33 51 43 43 Cl Cl Cl Cl Cl Cl Cl Cl Cl Cl Cl Cl Cl Cl Cl Cl Cl Cl Cl Cl Cl Cl Cl Cl Br Br 4-MeC6H4SO2 PhSO2 4-ClC6H4SO2 4-NO2C6H4SO2 3-NO2C6H4SO2 (EtO)2P(O) (MeO)2P(O) (PrnO)2P(O) (PriO)2P(O) (BuiO)2P(O) (MeO)2P(S) MeCO EtCO590 R X MeOCO EtOCO PhSO2 4-ClC6H4SO2 PhSO2 MeOCO EtOCO 4-MeC6H4SO2 PhSO2 4-ClC6H4SO2 Br Br Br Br Cl Br Br HHH R1N CHCXHal2+R2OH R2 R1 EtOCO Me2NCO 4-MeC6H4SO2 PhSO2 4-ClC6H4SO2 (MeO)2P(S) Me Me Me Me Me Me Ph3P=C(CN)CO Me Me Me Me Me Me Me Me Et Et Et Et Et Et Et Et Et Et Et Et Et Bui MeCO EtCO MeOCO EtOCO EtOCO MeCO MeCO PhCO PhOCO EtOCO Et2NCO 4-MeC6H4SO2 PhSO2 4-ClC6H4SO2 (EtO)2P(O) (MeO)2P(S) PhSO2 4-ClC6H4SO2 PhSO2 PhSO2 (BuiO)2P(O) PhSO2 CH2 O Bun CH2=CHCH2 CH2=CHCH2 CH2=CHCH2 PhSO2 PhSO2 4-MeC6H4SO2 4-ClC6H4SO2 Allyl alcohol adds to chloral arylsulfonylimines under mild conditions without heating.129 The addition of alcohols and ammonia at the C=N bond in trichloromethyl phenyl ketimines is accompanied by simultaneous replacement of the fragmentXby an alcoholic residue or an amino group to give 1-[alkoxy(amino)carbonylamino]-2,2,2-trichloro- ethylbenzenes 60.62 HY Cl3CC(Ph) NCOX Cl3CCY(Ph)NHCOY 60 X=OAlk, SAr, NAlk2; Y=MeO, EtO, NH2.Reactions of acyl-, sulfonyl- and dialkoxyphosphoryl-tri- chloroethylideneamines with phenols and 1-naphthol have been studied. It was shown that phenol 33, 36, 37, 50, 123, 124, 130 and p-chlorophenol 33, 39, 130 add to almost all types of imines giving rise to aryloxy derivatives 61. Ref. Hal 43, 44 109 78 125, 78 107 42 36 77, 126 77, 126 77, 126 Cl Cl Cl Cl Br Br Br Cl Cl Cl R1NHCH(OR2)CXHal2 59 Ref. Hal X 36 68 32, 72, 124 8, 124 8, 124 51 65 43 43 44 109 36 34 34 35 35 36 68 32, 124 124 74 33 51 78, 125 788, 77 107 33 127 Cl Cl Cl Cl Cl Cl Cl Cl Cl Cl Cl Br Cl Cl Cl Cl Cl Cl Cl Cl Cl Cl Cl Cl Cl Cl Cl Cl Cl Cl Cl Cl Cl Cl Cl Cl Br Br Br Br Br Prn Bun Cl Cl Cl Cl Cl Cl Cl Cl Cl Br Br HBr Cl Cl 128 129 129 129 Cl Cl Cl Cl Ph Cl Cl Cl G G Levkovskaya, T I Drozdova, I B Rozentsveig, A N Mirskova RN CHCXCl2+ ArOH Ar=Ph, 4-ClC6H4; X = Cl, CH2Cl; R =AlkCO, AlkOCO, (AlkO)2PO, ArSO2. 1-Naphthol readily adds to phosphorylimines.33, 50 However, these imines give no addition products with 2,6-dimethylphenol, 2,6-dinitrophenol or 2,4,6-trinitrophenol even on heating.33 This is caused not only by steric reasons but also by the low nucleophilicity of nitrophenols.Chloral arylsulfonylimines, which also react with phenol, do not react with chloro-substituted phenols.123 2-Chloroalkanols vigorously react with chloral imines and dichloroacetaldehyde sulfonylimines yielding N-[(2,2-dichloro-2- X-1-(2-chloroalkoxy)ethyl]amides 62, which cyclise on treatment with an alkali giving rise to 1-arylsulfonyl(ethoxycarbonyl)-2- trichloromethyl- and -dichloromethyloxazolidines.77, 131 R1N CHCXCl2 R1=PhSO2 (a), 4-ClC6H4SO2 (b), 4-MeC6H4SO2 (c), EtOCO (d), R2=H, X=Cl; R1=PhSO2, R2=CH2OMe, X=Cl (e); R1=PhSO2, R2=CH2Cl, X=Cl (f); R1=PhSO2, R2=H, X=H(g). The reactions of imines with carboxylic acids have been studied extensively. Although carboxylic acids are weak nucleo- philes, they add at the C=N bond of imines (sometimes on heating) to afford 1-acyloxy-2-polychloroethylalkylamides 64.R1N CHCXCl2+ R2COOH R1 MeCO PhSO2 (EtO)2PO (BuiO)2PO PhSO2 MeCO PhSO2 (MeO)2PO (PrnO)2PO (PriO)2PO (BuiO)2PO MeCO MeCO EtCO PrnCO MeCO PrnCO PhSO2 The imine 2b reacts with halocarboxylic acids to give addition products � sulfonamides 65a,b. An attempt to dehydrohalogen- ate these products in order to prepare heterocyclic compounds 66a,b failed.131 PhSO2N CHCCl3 2b RNHCHCXCl2 OAr 61 NaOH ClCH2CH(R2)OH R1NHCHCXCl2 R1 OCH(R2)CH2Cl 62a ± g N R2 CXCl2 O 63a ± g R1NHCHCXCl2 OC(O)R2 64 X R2 Ref. Me Me Me Me CH2Cl Me Ph Ph Ph Ph Ph Ph Ph Ph Ph 37 124 33 33 128 34 124 33 33 33 33 34 130 130 130 1-naphthyl 130 1-naphthyl 130 127 Cl Cl Cl Cl Ph Bun Cl Cl Cl Cl Cl Prn CH2Cl CH2Cl CH2Cl CH2Cl CH2Cl Cl 2-furyl X(CH2)nCOOH PhSO2NHCHCCl3 65a,b OCO(CH2)nXN-Functionally substituted imines of polychlorinated (brominated) aldehydes and ketones CHCCl3 PhSO2N (H2C)n O O 66a,b X=Cl, n=1 (a); X=Br, n=2 (b).The sulfonylimine 2b vigorously reacts with oximes of aliphatic, cycloaliphatic and aromatic ketones being thus con- verted into the addition products 67a ± d.8, 11, 124 PhSO2N CHCCl3+R1R2C NOH PhSO2NHCHCCl3 2b ONCR1R2 67a ± d R1=Me: R2=Ph (a), But (b), Me (c); R1±R2=(CH2)5 (d). An attempt was made to carry out addition of hydroxylamines to imines.Cyclic hydroxylamines, namely, 4-hydroxy-2H-1,4- benzoxazin-3(4H)-ones, were found to add to the acylimines 1 and 3. Compounds 68 were patented as antimicrobial prepara- tions.132 R3 O R2 R1C(O)N CHCCl3+ 1, 3 O NOH R3 O R2 O N 68 OCH(CCl3)NHCOR1 R1=Alk, Ar, ArO; R2, R3= H, Hal, AlkO, Alk, Ar, CN, NO2. b. Reactions of imines with NH-nucleophiles Only two examples of addition of ammonia to the imines in question have been reported. The reaction of the benzoylimine 19b with ammonia has resulted in the isolation of N-(1-amino-2,2,2- trichloro-1-phenylethyl)benzamide 69.86 PhCON C(Ph)CCl3+ NH3 PhC(O)NHC(Ph)CCl3 19b 69 NH2 It was shown 65 that biphilic phosphonium ylide 9 reacts with ammonia to give the product of addition to the C=N bond, Ph3P=C(CN)COCH(NH2)CCl3, in 95% yield, the cyano group and the C=P bond remaining intact.It was also reported that the reaction of ammonia with trichloromethyl phenyl ketimine affords 1-amino-1-phenyl-2,2,2-trichloroethylurea.62 The reactions of N-functionally substituted imines of poly- halogenated aldehydes and ketones with amines, amides and imines have been studied in detail. Acyl- and alkoxy(aryloxy)carbonyl-imines of polyhaloalde- hydes add primary and secondary amines to give addition products 70 in good yields.34 ± 37, 41, 43 ± 45, 65 HNR2R3 R1N CHCXCl2 R1NHCHCXCl2 70 NR2R3 R1= AlkCO, AlkOCO, ArCO, ArOCO, Ph3P=C(CN)CO; R2, R3=H, Alk, Ar; R2±R3=(CH2)2O(CH2)2, (CH2)4; ; X=Cl, MeCHCl, PhCHCl, Br, Pr, Bu.NR2R3= N NH Trichloroethylideneureas show a similar behaviour in the reactions with amines. Thus the amides 11a ± c add diethylamine, morpholine or piperidine at 20 8C to give diethylamide, morpho- lide and piperidide, respectively, of 1-diethylamino- (71a), 1-morpholino- (71b) and 1-piperidino-2,2,2-trichloroethylcarba- mic acid 71c.66, 68 591 HNR22Cl3CCH(NR22 )NHCOR1 71a ± c Cl3CCH NCOR1 11a ± c R1=N , N O ,Me2N; R2=Et (a), (CH2)2O(CH2)2 (b), (CH2)5 (c). Phenoxy- and methoxycarbonylimines of trichloromethyl aryl ketones also add amines at the C=N bond yielding compounds 72a ± c.62 HY Cl3CC(Ar) NCOX Cl3CCY(Ar)NHCOX 72a ± c X=PhO, Ar=4-Cl3CC6H4, Y=PhCH2NH (a); X=MeO, Ar=Ph, Y=BuNH (b), PhCH2NH (c).However, it was found that the reaction between phenyl phenyltrichloroethylidenecarbamate 53 and secondary amines, for example, morpholine or diphenylamine, can occur without participation of the C=N bond. Under rigorous conditions, the phenoxy group is replaced by an amine residue to give the imines 52 62 (see Section II.7). The reactions of dialkyl N-trichloroethylidene phosphorami- dates 7 with amines have also been studied in detail. It was shown 41, 133 that these imines and their sulfur-containing ana- logues 51 readily react with primary aromatic amines to give stable addition products, O,O-dialkyl N-(1-amino-2,2,2-trichloeroethyl) phosphoramidates and -phosphoroamidothioates. R1NHCH(NR2R3)CCl3 R1N CHCCl3+ R2R3NH 7 R1=PO(OAlk)2 (Alk=Me, Et, Prn, Pri, Bui); PS(OMe)2; R2= H: R3= Ph, PhCH2, 4-ClC6H4, 4-MeC6H4, 2,4,5-Cl3C6H2; R2= R3= Et; R2±R3=(CH2)5.The imines 7 add aniline, 4-chloroaniline and 4-methylaniline at room temperature; the reaction with 2,4,6-trichloroaniline occurs at *80 8C and that with the least nucleophilic 2,4,6- trinitroaniline does not occur even at 100 8C.133 Whereas acyl-, phosphoryl-, alkoxycarbonyl- and carbamoyl- imines ofrichloro- and dichloroacetaldehydes, trichloropropion- aldehyde and other aldehydes react to give the products of addition to the azomethine bond, the trichloroethylidenearene- sulfonamides 2 can either give the products of addition to the C=N bond or undergo haloform decomposition. Aniline,32, 124 trichloroaniline,41 morpholine,124 N-ethylani- line,136 N-methylaniline 136 and primary amines (methylamine, tert-butylamine, allylamine 129) yield products of nucleophilic addition.Diphenylamine does not react with chloral arylsulfonyl- imine.73 Meanwhile, highly basic dialkylamines induce haloform decomposition; this affords the products of substitution of the CCl3 N-arylsulfonyl-N0,N0-dialkylformamidines group, 73,73, 134 ± 136 in high yields. ArSO2N CHNR2 73 ArSO2N CHCCl3+ NHR2 2 Ar=Ph, 4-MeC6H4, 4-ClC6H4; R=Me, Et, Pr, Bu, C6H13; R±R=(CH2)2O(CH2)2, (CH2)5. In the case of sulfonylimines, this reaction pathway is explained 73 by the fact that the ArSO2 group, unlike acyl or alkoxycarbonyl groups, cannot participate in conjugation and, hence, it cannot ensure stabilisation of the transition state.Therefore, CHCl3 is ejected from the intermediate to give the compound 73. The reaction scheme proposed in the literature 136 includes the addition of amine to the double carbon ± nitrogen bond and the reaction of the intermediate with a second amine molecule to give ionic species 75. Fragmentation of this product yields formamide and a trichloromethyl anion.592 Cl3CCH NSO2Ar+R2NH 2 Cl3CCH(NR2)NSO2Ar Imines add readily to the azomethine bond of the trichloro- ethylideneamides 1 and 7 giving rise to azomethines.37, 136 Cl3CCH NX+HN CR1R2 1, 7 X=MeCO, R1=R2=Ph; X=PO(OEt)2: R1=R2=Ph; R1=Ph, R2=4-MeC6H4; R1=Me, R2=OEt. When 2,2-dichloropentanal N-acetylimine 76 34 or the phenyl- sulfonylimine 2b 124 are made to react with hydrazines, the addition products, N1-aryl-N2-[1-amido-2-(trichloroethyl)- and -(2,2-dichloropentyl)]hydrazines 77a ± c, are produced in good yields.124 R1Cl2CCH NX+H2NNHR2 X=PhSO2, R1=Cl, R2=Ph (77a); X=MeCO, R1=Pr, R2=Ph (77b), 2,4-(NO2)2C6H3 (77c).N-2,2,2-Trichloro-, bromodichloro-, dibromochloro- and 2,2- dichloro-ethylideneamides and 2-R-2,2-dichloroethylidene- amides of carboxylic, carbamic and sulfonic acids and of dialkyl phosphates add acid amides and substituted ureas on moderate heating giving rise to polyhaloalkanediylbisamides 78 and 79 in good yields. Cl3CCH NR1+H2NR2 R1 MeCO MeCO MeCO MeCO MeCO MeCO MeCO MeCO MeCO MeOCO MeOCO PhCO (EtO)2PO (EtO)2PO (EtO)2PO (EtO)2PO (EtO)2PO (EtO)2PO (EtO)2PO (EtO)2PO (PrnO)2PO (PriO)2PO (PriO)2PO EtOCO PhSO2 PhSO2 PhSO2 PhSO2 PhSO2 PhSO2 Cl3CCH(NR2)NHSO2Ar 74 7 R2NCH NSO2Ar+CCl3+R2NH2 + 75 R2NH2 Cl3CCH(N CR1R2)NHX R1Cl2CCHNHX 77a ± c 2b, 76 Cl3CCH(NHR2)NHR1 R2 MeCO PhCO PhSO2 (EtO)2PO (PhO)2PO (MeO)2PS (PriO)2PO (PriO)2PS (MeO)(MeS)PO MeOCO (EtO)2PO PhCO (EtO)2PO MeCO Cl3CCO ButCO PhCO 4-MeC6H4CO H2NCO Me2NCO MeCO MeCO H2NCO EtOCO MeCO PhCO H2NCO NC(O)(CH2)3 PhNHCO PhSO2 R2NH 7 NHNHR2 78 Ref.137 37 37 37 37 138 138 138 138 39 39 137 133 133 133 133 133 133 133 133 133 133 133 139 140 1408 124 124 74, 123 G G Levkovskaya, T I Drozdova, I B Rozentsveig, A N Mirskova R1 R2 + PhSO2 4-MeC6H4SO2 4-ClC6H4SO2 4-MeC6H4SO2 4-MeC6H4SO2 4-ClC6H4SO2 R1Hal2CCH NR2+ H2NR3 R2 Hal R1 HBr Cl MeCHCl MeCHCl Br Br HHH EtOCO EtOCO EtOCO MeOCO PhCO PhSO2 4-ClC6H4SO2 PhSO2 4-MeC6H4SO2 4-ClC6H4SO2 MeCO MeCO MeCO MeCO EtCO EtCO EtCO ClCH2CO PhCO Cl Cl Br Cl Cl Cl Cl Cl Cl Cl Cl Cl Cl Cl Cl Cl Cl Cl Cl CH2Cl CH2Cl CH2Cl CH2Cl CH2Cl CH2Cl CH2Cl CH2Cl CH2Cl This reaction can be performed for amides of aliphatic and aromatic carboxylic acids unsubstituted at the nitrogen atom as well as phosphoric and phosphorothioic amides and thioamides; N-substituted amides do not react with acetylimines even on refluxing for several hours.34, 138 Detailed investigation of the reactions of the imines 7 with amides has revealed several characteristic features.133 The capacity of amides for the addition sharply changes on variation of the electronegativity in the acyl residue.Comparison of the reactivities of acetamide, trichloroacetamide and the amide of pivalic acid demonstrates that the steric factor is insignificant. When the reaction mixture is heated to *90 8C, pivalamide and acetamide add quantitatively, whereas trichloroacetamide reacts only at 167 8C.133 In a study of the addition of amides and thioamides of four- coordinate phosphorus to trichloroethylideneacetamide 3c, it has been found that amides of phosphorus thioacids add more readily than the oxygen analogues to give the corresponding substituted acetamides 79a ± d.138 N-Acylated amides and thioamides of phosphorus-containing acids do not enter into this reaction due to steric shielding and low nucleophilicity of the amide nitrogen atom.N-Thioacylamides of phosphorus acids, in which the sulfur atom of the C=S group possesses high nucleophilicity, undergo exothermic addition giving rise to carboximidothioates 80a ± g.138 Addition at a sulfur atom also occurs in the reaction of the aldimine 3c with bis(diphenylthiophosphinyl)amine; this affords phosphinimidothioate 81.138 Cl3CCH NCOMe 3c R1R2P(X)NH2 R1R2P(X)NHCH(CCl3)NHC(O)Me R1R2P(X)NHC(S)R3 R1R2P(X)N Ph2P(S)NHP(S)Ph2 Ph2P(S)N PPh2SCH(CCl3)NHC(O)Me Ref.123 74 74 R1Hal2CCH(NHR3)NHR2 79 Ref. R3 76 109 76 45 45 125, 78 78 77, 126 77, 126 77, 126 130 130 130 130 130 130 130 130 130 EtOCO EtOCO EtOCO MeOCO MeOCO PhSO2 4-ClC6H4SO2 PhSO2 4-MeC6H4SO2 4-ClC6H4SO2 MeCO EtCO PrnCO Me2NCO EtCO PrnCO Me2NCO ClCH2CO PhCO 79a ± d C(R3)SCH(CCl3)NHC(O)Me 80a ± g 81N-Functionally substituted imines of polychlorinated (brominated) aldehydes and ketones R3 Compound R2 R1 X PriO PriO OR=PhCH2CO (a), EtCO (b), EtOCO (c), (EtO)2PO (d), (MeO)2PS (e). PriO PriO S 79a 79b 79c 79d 80a 80b 80c 80d 80e 89f 80g MeO MeS Et Pri Pri Pri Et Pri Pri MeO MeO Et Pri Pri Pri Et Pri Pri Ph Ph 4-MeOC6H4 3-NO2C6H4 Ph Ph MeO SOSSSSOOS The reactivity of amides and thioamides of phosphorus acids towards chloral acetylimine increases in the sequence R1R2P(X)NHC(O)R3 < R1R2P(O)NH2 < R1R2P(S)NH2 <R1R2P(S)NHP(S)R32 <R1R2P(X)NHC(S)R3.Promising complex-forming reagents and biologically active compounds 82 have been prepared by the reaction of diaza-18- crown-6 with two equivalents of the imine 3c. 141 O O NH HN Cl3CCH NAc+ 3c O O O O NHAc AcHN N CH CH N CCl3 Cl3C O O 82 These compounds are smectic liquid crystals with fan tex- ture.141 c. Reactions with S-nucleophiles Much attention has been devoted to the reactivity of polyhalo- aldehyde imines N-substituted by functional groups towards sulfur-containing nucleophiles (hydrogen sulfide, thiols, benzene- thiols and thioamides), which add to imines very easily.The N-acetaldimine 3c and dialkyl phosphoramidates 7d,e react with hydrogen sulfide at 20 8C (ratio 2 : 1) to give bis(1-ami- do-2,2,2-trichloroethyl) sulfides 83a ± c.33, 37 H2S Cl3CCH(NHR)SCH(NHR)CCl3 Cl3CCH NR 83a ± c 3c, 7d,e R=MeCO (3c, 83a), (PriO)2PO (7d, 83b), (BuiO)2PO (7e, 83c). Unlike the reaction of imines with water, which yields stable hydroxy derivatives, in this case, the corresponding product with a thiol group has not been isolated. Evidently, this difference is related to the higher reactivity of N-(2,2,2-trichloro-1-sulfanyl- ethyl)amides, formed in the first stage, towards the addition to the C=N bond of the initial imine compared to that of the hydroxy analogue.As has already been noted, the rate of the reaction of ethanethiol with chloral imines depends on the nature of the substituent at the nitrogen atom; the reaction gives N-(2,2,2- trichloro-1-ethylthioethyl)amides 84.122 Later it has been shown that O,O-dimethyl N-(2,2,2-trichloroethylidene) phosphoramido- thioate 85 also reacts with ethanethiol being thus converted into the compound 84e.51 593 Cl3CCH(SEt)NHR 84a ± e Cl3CCH NR+EtSH 85 Chloral arylsulfonylimines 2 add butanethiol and prop-2-ene- 1-thiol.142, 143 The reactions of the phosphorus-substituted imines of chloral 7 with thiols and diphenylphosphinothioic acid have been studied extensively.33, 50 These reactions occur on mixing the reactants without heating, the yields of N-(2,2,2-trichloro-1-R-thio- ethyl)amides 86 being quantitative.HSR2 Cl3CCH NP(O)(OR1)2 Cl3CCH(SR2)NHP(O)(OR1)2 86 7a ± e R1=Me, Et, Prn, Pri, Bui; R2=4-ClC6H4, 4-NO2C6H4, Et, Ph, Ph2P(S). The acylimine 3c,37 the imine 9, functionally substituted in the acyl fragment,65 methoxycarbonylimine 10a 39 and 2,2-dichloro- alkylideneacetamides 76 34 also react with ethanethiol,34 4-nitro- benzenethiol,39 benzenethiol 34, 39 and 4-chlorobenzenethiol,65 with heat evolution. Imines 87 have been introduced in situ in the reaction with 4-chlorobenzenethiol; this gave N-[polychloroalkyl-1-(4-chloro- phenylthio)]amides 88.45 HSR2 XCl2CCH(SR2)NHR1 XCl2CCH NR1 87 88 X=MeCHCl, PhCHCl; R1=MeOCO, PhCO; R2=4-ClC6H4.d. Reactions of imines with bifunctional O,O-, O,N-, S,N- and N,N- nucleophiles The reactions of trichloropropionaldehyde acetylimine with 4-aminophenol and para-phenylenediamine have been studied.130 These reactions were found to involve the amino group of the aminophenol or both amino groups of phenylenedi- amine and to give 4-(1-acetamido-2,2,3-trichloro- propyl)aminophenol 89 or bis(N-1-acetamido-2,2,3- trichloropropyl)-1,4-phenylenediamine 90. ClCH2CCl2CH NAc H2NC6H4OH-4 ClCH2CCl2CHNHCOMe HNC6H4OH-4 89 NHAc AcHN H2NC6H4NH2-4 NHCH CHNH CCl2CH2Cl ClCH2Cl2C 90 The sulfonylimine 2b reacts with monoethanolamine at 720 8C giving compound 91 in a quantitative yield.144 H2NCH2CH2OH Cl3CCHNHSO2Ph Cl3CCH NSO2Ph 2b OCH2CH2NH2 91 Treatment of the sulfonylimines 2 with triethanolamine or dimethylethanolamine induces only haloform decomposition of azomethines.73 However, the ethoxycarbonylimine 10b reacts with N,N-dimethylethanolamine giving rise to a stable addition product 92.73594 Me2NCH2CH2OH Cl3CCHNHC(O)OEt Cl3CCH NC(O)OEt 10b OCH2CH2NMe2 92 (66%) Only decomposition products � sulfamides, chloroform and aminothiol hydrochloride � have been isolated upon the exo- thermic reaction of the imines 2 with aminoethanethiol.144 The reaction of ethylene glycol with the imines 3 has given both mono- (93) and bis-O-amidoalkylation products, [1,2-bis(1- arylsulfonamido-2,2,2-trichloroethoxy]ethanes 94.144 The addition of 2-hydroxyethanethiol to the sulfonylimines 2 at an equimolar ratio of the reactants occurs as an exothermic reaction involving the SH group and affords N-[2,2,2-trichloro-1- (2-hydroxyethylthio)ethyl]arenesulfonamides 93b.144 Successive treatment of these products with thionyl chloride and alkali results in the formation of the oxazolidines 63 and thiazolidines 95.131, 144 HXCH2CH2OH Cl3CCHNHSO2Ar Cl3CCH NSO2Ar 2a ± c XCH2CH2OH 93a,b CCl3 CCl3 Cl3CCH NSO2Ar ArSO2NHCHXCH2CH2OCHNHSO2Ar 94 SO2Ar Cl3C SOCl2 NaOH Cl3CCHNHSO2Ar N X XCH2CH2Cl 63, 95 Ar=Ph, 4-MeC6H4, 4-ClC6H4; X=O(63, 93a, 94), S (93b, 95). The sulfonylimines 2 react with glycolic and sulfanylacetic acids under mild conditions to give nucleophilic addition prod- ucts, O- or S-(2,2,2-trichloro-1-arylsulfonylamino)glycolic 96a and -thioglycolic 96b acids, respectively.124, 145 When these prod- ucts are treated with thionyl chloride and then with triethylamine, intramolecular cyclisation occurs giving rise to 3-arylsulfonyl-2- trichloromethyl-1,3-oxazolidin-4-ones 97a and -1,3-thiazolidin-4- ones 97b.HXCH2COOH SOCl2 Cl3CCHNHSO2Ar Cl3CCH NSO2Ar 2 XCH2COOH 96a,b SO2Ar Cl3C Et3N N Cl3CCHNHSO2Ar X XCH2COCl O 97a,b Ar=Ph, 4-ClC6H4, 4-MeC6H4; X=O(a), S (b). 2. Reactions of phosphorus compounds with polyhaloaldehyde imines The sulfonylimines 2a,b readily react with triethyl phosphite to give diethyl N-(2,2-dichlorovinyl)-N-(arylsulfonyl) phosphorami- dates 98a,b in high yields.71, 146 (EtO)3P Cl2C CHN(SO2Ar)P(O)(OEt)2+ EtCl Cl3CCH NSO2Ar 98a,b 2a,b Ar=4-MeC6H4 (a), Ph (b).Dialkyl N-trichloroethylidene phosphoramidates react with trialkyl phosphites giving rise to bis(phosphorylated) compounds 99.147 G G Levkovskaya, T I Drozdova, I B Rozentsveig, A N Mirskova (R2O)3P Cl3CCH NP(O)(OR1)2 Cl2C CHN[P(O)(OR2)2][P(O)(OR1)2] 99 R1=Et, R2=Et, Pr. The reactions of dialkyl N-tetrachloroethylidene phosphor- amidates with trimethyl and triethyl phosphites occur in a similar way and result in the formation of trichlorovinylamides Cl2C=CClN[P(O)(OR1)2] (R1=R2 = Me, Et; R1 = Me, R2 = Et).102 The reaction of benzoylchloral imine 3a with triethyl phos- phite follows a different pathway yielding a complex mixture of products, in which diethyl 1-benzamido-2,2-dichlorovinyl- phosphonate 100 and diethyl 1-benzamido-2,2,2-trichloroethyl- phosphonate 101 have been detected.147 Trimethyl phosphite reacts with the imine 3a giving only dimethyl 1-benzamido-2,2-dichlorovinylphosphonate 102; the yield of this product is not given in the study cited.148 (EtO)3P Cl2C C(NHCOPh)P(O)(OEt)2 100 (EtO)3P Cl3CCH(NHCOPh)P(O)(OEt)2 Cl3CCH NCOPh 3a 101a (MeO)3P PhCONHCP(O)(OMe)2 102a CCl2 Detailed study of the reaction of the imine 3c with trimethyl phosphite 148 showed that, depending on the reaction temper- ature, different products can be obtained.Cl3CCH NCOMe 3c CNHCOMe Cl3CCHNHCOMe+Cl2C P(O)(OMe)2 P(O)(OMe)2 101b 102b CCl3 Cl Cl (MeO)3P N MeCONH Me O (MeO)2(O)P 103 Me N O Cl2C P MeO OMe OMe 104 At 713 8C, a mixture of dimethyl a-(N-acetyl)amino-2,2,2- trichloroethylphosphonate 101b and dimethyl a-(N-acetyl)amino- 2,2-dichlorovinylphosphonate 102b is formed. The same reaction carried out without cooling (40 ± 45 8C) affords 1,3-oxazine 103, and at740 8C, 1,2,4-oxaphosphazole 104 is obtained.149 Organophosphorus acids and thioacids react with imines to give addition products.36, 37,150, 151 N-Acetyl-,37 N-ethoxycar- bonyl- 36 and perchloroalkylsulfonyl-trichloroacetaldimines 71 readily add various organophosphorus reagents. In the case of dialkyl phosphites, dialkyl 1-acetyl-, 1-ethoxycarbonyl- and perchloroalkylsulfonyl-amino-2,2,2-trichloroethylphosphonates 105 are formed.Cl3CCH(NHR1)P(O)(OR2)2 Cl3CCH NR1 + (R2O)2P(O)H 105 R1=EtOCO, MeCO, AlkFSO2; R2=Alk.N-Functionally substituted imines of polychlorinated (brominated) aldehydes and ketones The P-substituted imine 7b also adds diethyl phosphite being thus converted into diethyl N-(1-diethoxyphosphoryl-2,2,2-tri- chloroethyl) phosphoramidate.50 The reaction of diphenylphosphinic acid 37 with the acetyl- imine 3c results in the formation of the O-addition product, 1-acetylaminotrichloroethyl diphenylphosphinate 106a; in the case of phosphorothioic and -dithioic acids, S-addition products, namely, 1-acetylamino-2,2,2-trichloroethyl diphenylphosphino- dithioates and -thioates 106b,c are produced.37, 150, 151 Cl3CCHNHCOMe Cl3CCH NCOMe +R2P(X)YH YP(X)R2 106a ± c R=Ph,X=Y=O(a); R=Ph, MeO, EtO, PrnO, PriO,X=Y=S(b); R=EtO, PrnO, PriO, X=S, Y=O (c).Phosphorus pentachloride is widely used in the reactions with derivatives of alkyl- and alkoxycarbonylimines of polyhaloalde- hydes in order to prepare isocyanates.62, 63 However, heating of N-perchloroethylideneureas 8 63, 64 with PCl5 yields highly reac- tive halo-substituted diazadienes 107a,b in 41% ±76% yields. PCl5 Cl3CClC NCCl NR+POCl3+HCl Cl3CClC NC(O)NHR 8 107a,b R=Pri(a), Ph (b). N-Mesityltrichloroethylideneurea reacts with phosphorus pentachloride to give a mixture of two isomers, carbodiimide and diazadiene.63 Cl3CCl2CN C NMes PCl5 Cl3CClC NC(O)NHMes Cl3CClC NCCl NMes 7POCl3, 7HCl An unusual reaction of N-tert-butyl-N0-perchloroethylide- neurea with PCl5 was discovered; in this case, phosphorus pentachloride acts simultaneously as a chlorinating and a phos- phorylating agent.64 Cl3CCl2CN CClN PCl3 Cl3CClC NC(O)NHBut+PCl5 When N-dichlorophosphorylimines of polychloroaldehydes are treated with phosphorus pentachloride, (dichlorophosphor- ylthio)polyhaloethylideneimines 37a ± d are formed (see Section II.6).99 3.Reactions of acylimines with organometallic compounds and metals Reactions of polyhaloaldehyde imines with organometallic com- pounds have been less studied. A review has been published 19 dealing mainly with reactions with organomagnesium and -lithium compounds.In the general case, alkoxycarbonylimines of chloral 10a,b are converted into nucleophilic addition products 108 upon reactions with non-branched Grignard reagents.39, 152 Cl3CCH(R2)NH Cl3CCH NCOR1+ R2MgX 108 10a,b R1=MeO, EtO; R2=Me, Et, Pr, Bu, Ph, PhCH2, CH2=CHCH2. Branched organometallic compounds reduce the imine 10b to the corresponding amine, which adds to the initial compound to give the aminal. This product cyclises giving rise to azetidine 109.152 595 RMgX Cl3CCH NCOOEt Cl3CCH2NHCOOEt Cl3CCH NCOOEt 10b COOEt Cl3C N Cl3CCH2NCOOEt Cl Cl3CCHNHCOOEt NHCOOEt Cl 109 R=Pri, Bui. Acylimines of halogenated aldehydes, like other halogenated imines, react with lithium aluminium hydride yielding substituted aziridines.19, 153, 154 In particular, N-acetyl-R-2,2-dichloroacetal- dimines are converted into 1,2-substituted aziridines 110, in which the acyl group has been reduced to an ethyl group.Et COMe LiAlH4 LiAlH4 N N RCl2CCH NCOMe Cl R R 110 R=Pr, Bu, C5H11. The behaviour of trichloroethylideneacetamide 3c in reactions with zinc and electrophilic reagents has been studied.155 The interaction of the imine 3c with zinc affords a zinc-containing intermediate, which is a polydentate reagent with three nucleo- philic centres � the carbon, nitrogen and oxygen atoms. The reaction of this intermediate with a-chlorinated ethers or acyl halides gives N-addition products, namely, substituted amides and imides.The reaction with chlorotrimethylsilane affords the O-attack product, N-(2,2-dichlorovinyl)-N-(1-trimethylsilyloxy- ethylidene)amine. Zn [Cl2C CH N C(Me) O]+Cl7 Cl3CCH NCOMe 3c ClCH2OBu Cl2C CHN(COMe)CH2OBu ClCOR Cl2C CHN(COMe)COR ClSiMe3 Cl2C CHN CMeOSiMe3 R=Me, Pri, Bun, But, Ph. 4. Reactions of imines with inorganic acids The reactions of imines with hydrogen chloride have been studied in the greatest detail; it has been shown that polyhaloalkylidene- amides of dialkylphosphinic, carboxylic and carbamic acids are converted into chloroalkyl derivatives.33, 35, 37, 39, 46, 50 On treatment of the sulfonylimines 2a,b with hydrogen cyanide in the presence of triethylamine, 1-(arylsulfonylamino)- 2,2-dichloroacrylonitriles 111a,b are readily formed.156 7 [Cl3CCH(CN)NSO2Ar Cl3CCH NSO2Ar+HCN 2a,b 7 Et3N Cl3CC(CN)NHSO2Ar] 7HCl Cl2C C(CN)NHSO2Ar 111a,b Ar=4-MeC6H4 (a), Ph (b).Similarly, when the benzoylimine 3a is treated in situ with hydrogen cyanide, N-(2,2-dichloro-1-cyanovinyl)benzamide is produced.38 On the other hand, the methoxycarbonylimine of trichloro- methyl phenyl ketone adds hydrogen cyanide to give the amide Ph(CCl3)C(CN)NHCOOMe.62596 The reaction of chloral imines with hydrazoic acid has been carried out; this gave 1-amino-2,2,2-trichloroethyl azides 112.40 Cl3CCH NR+HN3 Cl3CCH(N3)NHR 112 R=PO(OEt)2, COEt, COPr, COCH2Ph, COOEt. 5. Cycloaddition reactions involving alkoxycarbonyl-, acyl- and sulfonyl-imines of polyhaloaldehydes The capability of acylimines, including halogenated ones, of exhibiting the properties of both dienophiles and dienes has been discussed in reviews.157, 158 Sulfonylimines can act as dieno- philes.158 It should be noted that these reviews 157, 158 are far from embracing all publications devoted to this topic.The high reactivity of chloral sulfonylimine 2a towards dienes (2,3-dimethylbutadiene and cyclopentadiene) was first reported in 1964.159 Reactions of chloral sulfonylimines with a large number of dienes (piperylene,160 butadiene,161 cyclopentadiene,57, 158 ± 164 dimethylbutadiene 161, 162, 165 and isoprene 160, 161) have been studied much later. The cycloaddition products, 1-arylsulfonyl- and 1-alkoxycarbonyl-2-trichloromethyl-1,2,3,6-tetrahydropyri- dines 113 and N-arylsulfonyl- and N-alkoxycarbonyl-2-azabicy- clo[2.2.1]hept-5-enes 114 are formed in 50%± 94% yields.N-Ethoxycarbonyl-2-azanorbornenes have been synthesised and some of their reactions, for example, dehydrochlorination and bromination, have been studied.161, 162 N-Ethoxycarbonyl-2-azabicyclo[2.2.2]oct-5-ene 115 has been prepared 163 by the reaction of the imine 10b with cyclohexadiene. Chloral 4-tolylsulfonylimine 2a does not enter into the cyclo- addition reaction with cyclohexadiene either on heating or under acid catalysis.163 R4 R3 R2 R3 NR1 R4 CHCCl3 R2 113 NR1 R1N CHCCl3 114 CCl3 NCOOEt R1=COOEt 115 CCl3 R1=MeOCO, EtOCO, MeCO, 4-XC6H4SO2 (X=H, Me, Cl, MeO); R2, R3, R4=H, Me.Cycloaddition of halogenated azomethines to non-symmet- rical dienes occurs regiospecifically giving only one of the two possible isomeric heterocycles.160 Study of the 1H NMR spectra of 4-tolylsulfonyl-, alkoxycar- bonyl-, and acetyl-substituted 2-azanorbornenes and bicyclooc- tenes, prepared by the reactions of chloral imines 2a, 3c, and 10a,b with cyclopentadiene and cyclohexadiene, showed that these bicyclic systems are mixtures of two isomers with exo- and endo- arrangement of the trichloromethyl group, the isomer ratio being dependent on the reaction temperature and on whether or not a Lewis acid is present in the system.161 ± 164 Acylimines of polyhaloaldehydes are able to participate in reactions with dienes not only as dienophiles but as hetero- dienes.162 G G Levkovskaya, T I Drozdova, I B Rozentsveig, A N Mirskova CCl3 CCl3 R2 N N R2 + Me O Me O R1 CHCCl3 R2 R1 117a ± c 116a ± c N + CCl3 O R1 3c NCOMe R2 R1 R1=R2=Me (a); R1=Me, R2=H(b); R1=OMe, R2=H (c).Thus the acetylimine 3c reacts with 2,3-dimethyl-, 2-methyl- and 2-methoxybuta-1,3-dienes mainly as a diene to give mixtures of 5,6-dihydro-4H-1,3-oxazine derivatives 116a ± c and 117a ± c. When the acetylimine reacts with 2,3-dimethylbuta-1,3-diene, it acts as a dienophile; a compound of the type 113 was isolated.165 The methoxycarbonylimine 10a reacts with 1-alkoxydienes as a dienophile, the final reaction product, 1-methoxycarbonyl-2- trichloromethyl-1,2-dihydropyridine 119, being formed in 84% yield upon elimination of an alcohol molecule with simultaneous migration of the double bond in the intermediate 118.166CCl3 CCl3 CHCCl3 + N N NCOOMe COOMe 119 10a COOMe 118 OR OR R=Me, Et.Chloral acylimines introduced in reactions with vinyl ethyl ether act as heterodienes, which leads to 2-alkyl-6-ethoxy-4- trichloromethyl-5,6-dihydro-4H-1,3-oxazines 120a ± d.167R N O Cl3C RC(O)N CHCCl3+H2C CHOEt OEt 120a ± d R=Me (a), Et (b), Pr (c), Ph (d). Sulfonylimines are capable of forming [2+2]-cycloadducts. Thus the imine 2f reacts with trimethylsilyl ketene or ketene at room temperature to afford 1-propylsulfonyl-4-trichloromethyl- azetidin-2-ones 121a,b.168, 169 R O Cl3CCH NSO2Pr +RCH C O NSO2Pr 2f Cl3C 121a,b R=H(a), SiMe3 (b).The reaction of the sulfonylimine 2f with trimethylsilyl ketene diethylacetal gives 3-(N-trimethylsilylpropylsulfonylamino)-1,1- diethoxy-4,4,4-trichlorobut-1-ene 122 instead of a cyclic product. The compound 122 undergoes desilylation and hydrolysis yielding ethyl 3-(propylsulfonylamino)-4,4,4-trichlorobutanoate 123.53 The researchers 53 suggested that the reactions with trimethyl- silyl ketene acetals occur via bipolar intermediate A, which is stabilised upon migration of the trimethylsilyl group. An alter- native pathway in which the silyl group migrates in [2+2]- cycloadduct 124 formed from the intermediate A is also possible.N-Functionally substituted imines of polychlorinated (brominated) aldehydes and ketones 7 Cl3CCHNSO2Pr Me3SiCH C(OEt)2+ Cl3CCH NSO2Pr 2f + H2O Cl3CCHN(SiMe3)SO2Pr Me3SiCHC(OEt)2 A Cl3CCHNHSO2Pr CH2COOEt 123 CH C(OEt)2 122 OEt Cl3CCH CHC(OEt) NSO2Pr Cl3C NSO2Pr 124 O-Silylated analogues of the compound 122, 1 : 1 adducts 125, have been isolated upon reactions of trimethylsilyl ketene dialkyl- acetal with the acetylimine 3c.No desilylation was observed in this case.169, 170 It is noteworthy that the reaction of the imine 3c with ketene acetals occurs without heating and gives [2+4]-cycloadducts, 6,6-dialkoxy-2-methyl-4-trichloromethyl-5,6-dihydro-4H-1,3-ox- azines 126.Under hydrolysis conditions, they are converted into alkyl 3-(methylcarbonylamino)-4,4,4-trichlorobutanoates, for example, compound 127.170 Cl3CCH NCOMe + R1CH C(OR2)2 3c *Me3Si MeC NCH(CCl3)CH C(OR2)2 R1=Me3Si 125 OSiMe3 CCl3 Cl3CCHCH2COOEt H2O N C±O cyclisation OR2 R2=Et R1=H Me OR2 NHCOMe 127 O 126 R1=H,Me3Si; R2=Me, Et, Pri.No cyclic products have been isolated after the reactions of the imines 10a,b with ketene dietl- and dimethylacetals. However, the formation of these products was proved by spectroscopy and by isolation of esters 128, resulting from their transformations, in quantitative yields.171 Cl3CCH NCOOR1+H2C C(OR2)2 10a,b CCl3 CH2COOR2 H2O N Cl3CCH OR2 NHCOOR1 O R1O OR2 128 R1, R2=Me, Et. The reaction of chloral acetylimine 3c with O-acetyl-O-ethyl- trimethylsilyl ketene acetal gives a 1 : 1 product 130, which is an ester of substituted butyric acid containing a trimethylsilyl group at the a-position. Presumably, this product results from migration of the acetyl group in the intermediate 129.172 Cl3CCH NCOMe + Me3SiCH COAc 3c OEt CCl3 Me3Si *COMe N Cl3CCHCH(SiMe3)COOEt 7 + N C(Me)OAc EtO Me O 130 OAc 129 597 Meanwhile, 2-trimethylsilyloxy-3-trimethylsilylacrylonitrile reacts with the imine 3c only in the presence of a catalyst, trimethylsilyl trifluoromethanesulfonate. The reaction involves the activated cyano group and results in [4+2]-cycloadduct, 4H- 1,3,5-oxadiazine.173 Me3SiOSO2CF3 3c+Me3SiCH C(CN)OSiMe3 CCl3 N N Me3SiCH O Me OSiMe3 The reactions of imines with several acetylene derivatives have been studied.The chloral imine 3c reacts with silicon- and germanium-substituted alkoxyacetylenes at room temperature without a catalyst to give a mixture of [4+2]-cycloadducts � 6-alkoxy-2-methyl-4-trichloromethyl-5-trialkylsilyl(trialkylgerm- yl)-4H-1,3-oxazines 131 � and N- and O-element-substituted acyclic compounds isomeric to them.172 COR2 Cl3CCH NCOMe+R13 MC 3cCCl3 MR13N Cl3CCHC COR2 + Cl3CCHC COR2 + OR2 Me N C(Me)OMR13 N(MR13 )COMe O 131 R1, R2=Me, Et; M=Si, Ge. It is noteworthy that the reactions of ethoxyacetylene and methoxymethylacetylene with the imine 3c occur over a period of 20 h at room temperature and give [4+2]-cycloadducts of type 131, 6-alkoxy-2-methyl-4-trichloromethyl-1,3-4H-oxazines.52 The reactions of trialkylsilyl(germyl)ethoxyacetylenes with sulfonylimines follow the [2+2]-cycloaddition route.The result- ing cycloadducts, 1-alkylsulfonyl-2-alkoxy-3-trialkylsilyl(germ- yl)-4-trichloromethyl-2-azetines, undergo ring opening under thermolysis conditions. The addition of water to the product of recyclisation of the triethylgermyl derivative of azetidine involves the C=C bond and gives compound 132, the azomethine group remaining intact.52 OR3 R2 D Cl3CCH NSO2R1+R2C COR3 NSO2R1 Cl3C R1=Pr, R2=Et3Ge, R3=Me Cl3CCH C C NSO2R1 H2O R2 OR3 Cl3CCH(OH)CHC NSO2Pr 132 Et3Ge OMe R1=Me, Pr; R2=Me3Si, Et3Si, Me3Ge, Et3Ge; R3=Me, Et. Arylsulfonyl-, acetyl- and alkoxycarbonyl-imines of chloral enter into cycloaddition reactions with azomethines and aza- dienes.Thus cycloaddition of diphenyl isocyanatophosphite to the imine 3c yields cycloadducts, the dimerisation of which during the reaction leads to tricyclodecanes 133.174 When the imine 3c reacts with dimethyl alkynylphosphonites, unstable cycloadducts are formed, which are hydrolysed to give linear derivatives 134.598(PhO)2PNCO Cl3CCH NAc (3c) (MeO)2PC CR O Cl3C PhO OPh O N OPh N P P NR RN N N P OPh Ac CCl3 O PhO OPh CCl3 133 Me O R C RC CP(O)CHNHAc P N 134 MeO CCl3 OMe C OMe CCl3 R= Me, Ph. The imine 19c containing a P(O)(OEt)2 group reacts with acyclic phosphines in a similar way, giving rise to unstable [4+1]- cycloaddition products. The reaction of this imine with P-diethylaminobenzo-1,3-dioxaphospholane gives a stable spiro- phosphorane.87 Cl3C PO(OEt)2 R3P N PR3 O Ph NCOPh Cl3CC O PNEt2 (EtO)2P(O) Ph O O O P N OCl3C PO(OEt)2 Ethyl 2-R-3-aryl-5-trichloromethyl-1,2,4-oxadiazolidine-4- carboxylates 135 have been prepared by the reaction of azo- methine N-oxides with the imine 10b.175 COOEt Cl3C N ArCH N(O)R+Cl3CCH NCOOEt O Ar 10b NR135 R=Ph; Ar=Ph, 4-NO2C6H4, 4-MeOC6H4. It was shown that the arylsulfonylimines 2a ± c react with benzal- and 4-methoxybenzal-azines by a two-stage cross-[2+3]- cycloaddition mechanism giving bicyclic triazole derivatives 136 but do not react with azines derived from other aromatic aldehydes.176, 177 Arylsulfonylimines give no cycloaddition prod- ucts when react with aliphatic ald- and ketazines; only the products of C-arenesulfonamidoethylation of azines were iso- lated.177 Ar 0 CCl3 N NSO2Ar ArSO2N ArSO2N CHCCl3+(Ar 0CH N)2 N 2a ± c Ar 0 Cl3C 136 Ar=Ph, 4-MeC6H4, 4-ClC6H4; Ar 0=Ph, 4-MeOC6H4.Dimethoxycarbene acts as the dienophile in the reactions with the heterodiene system of the benzoylimine of chloral 3a; 5,5-dimethoxy-2-phenyl-4-trichloromethyl-4,5-dihydro-1,3-oxa- zolidine 137 thus formed is dehydrochlorinated under the reaction conditions.178 G G Levkovskaya, T I Drozdova, I B Rozentsveig, A N Mirskova MeO MeO O MeO Ph C: +Cl3CCH NC(O)Ph 3a MeO N Cl3C 137 MeO MeO O Ph N Cl2C 6. C-Amidoalkylation with polyhaloaldehyde imines C-Amidoalkylation reactions had been unknown before 1970; it had only been reported 179 that chloral acylimines do not react with anisole.The first example of involvement of imines of polyhaloalde- hydes in C-amidoalkylation is represented by the reaction of chloral N-fluorosulfonylimine with isobutene, which gives the product of alkylation of the methyl group � 4-fluorosulfonyl- amino-5,5,5-trichloro-2-methylpent-1-ene 138.59 FSO2NHCHCCl3 FSO2N CHCCl3+ Me2C CH2 2 CH2C(Me) CH 138 Later it has been shown that it is a general type of reaction. Arylsulfonylimines of chloral, dichloroacetaldehyde and dichlo- rophenylacetaldehyde can also amidoalkylate aromatic and heteroatromatic compounds. Trichloro- 2,180 dichloro- 13 77 and phenyldichloro-ethyl- idenearenesulfonamides 18 181 react with electron-enriched are- nes, namely, with anisole and thioanisole in the presence of boron trifluoride etherate and with dimethylaniline without a catalyst; the substituent enters the para-position with respect to the electron-donating group.The sulfonylimines 2 C-amidoalkylate benzene, toluene, chlorobenzene, naphthalene and chlorothio- phene only in the presence of oleum.182 ArSO2NH Y Y ArSO2N CHCXCl2+ Cl2XC 2, 13, 18 Ar=Ph, 4-ClC6H4, 4-MeC6H4; X=H, Cl, Ph; Y=H, Me, OMe, SMe, NMe2, Cl. Pyrrole,183 1-methylpyrrole,183 thiophene,77, 182, 184 furan and sylvan [a-methylfuran, Ed.] 77,127 have been C-arylsulfamidoalky- lated by the imines 2 and 13, the substituent entering the 2-position.4,5-Tetramethylene-, 3H- and 3-methyl-4,5-pentam- ethylene-pyrroles react in a similar way.183, 185 2-Chlorothiophene reacts only in the presence of oleum.182 CXCl2 ArSO2N CHCXCl2+ R R Z Z NHSO2Ar Ar=Ph, 4-ClC6H4, 4-MeC6H4; X=H, Cl; R=H, Cl, Me; Z=S, O, NH, NMe. 2-Acetyl- and 2-ethoxyfuran do not react with the imines 2, and the reactions of 2-hydroxymethylfuran- and 2-furan-carbox- ylic acids involve the hydroxy or the carboxy group and afford addition products of types 59 and 64, respectively, (see Section IV.1.a).127 The sulfonylimines 2a ± c react with 1,8-bis(dimethyl- amino)naphthalene without heating or catalysts giving rise to 4-[1-(arylsulfonylamino)-2,2,2-trichloroethyl]-1,8-bis(dimethyl- amino)naphthalenes 139.186599 N-Functionally substituted imines of polychlorinated (brominated) aldehydes and ketones R1 R1 ArSO2NHCHCCl3 O N O R2 R2 Cl3C Cl3C ArSO2N CHCCl3+ 2a ± c 146 NHCOR3 NHCOR3 Me2N NMe2 Me2N NMe2 139a ± c Ar= 4-MeC6H4 (a), Ph (b), 4-ClC6H4 (c).R1±R2=CH2SCH2, (CH2)3, (CH2)4, OC(Me)2O; R3=Me, Ph, PhCH2O, PhCH=CH. MeO NHCOR3 NHCOR3 Cl3C Cl3C The sulfonylimines 2 react equally easily with indole and its substituted derivatives being thus converted in good yields into compounds 140, resulting from C-amidoalkylation at the 3-position.187 O N ArSO2NHCHCCl3 147 R3=Ph, PhCH2O. R2 R2 NR1 NR1 ArSO2N CHCCl3+ 2 140 Ar=Ph, 4-ClC6H4; R1, R2=H, Me. Reactions of various types of monohydroxypyrimidines with N-(benzoyl)trichloroacetaldimine 3a and other acylimines 3, generated in situ from the corresponding tetrachloroethylamides have been studied.190, 191 It was found that 2-hydroxypyrimidines and 4-hydroxypyrimidines containing no substituents at the 2-position form the products of amidoalkylation of the NH group � 1-(1-acylamino-2,2,2-trichloroethyl)-2- (148) and -4-hydroxypyrimidines 149, respectively.190 The sulfonylimine 2b C-amidoalkylates azines and hydra- zones of aldehydes and ketones containing a-methyle-methine groups; this leads to 3-arylsulfonylamino-4,4,4-tri- chlorobutyl ketazines 141a,b, aldazines 141c,d and hydrazone 142.177 R3 R1 NCH(CCl3)NHCOR NH (CHR1R2CR3 N)2 PhSO2NHCH C C N +RCON CHCCl3 O O N N 148 CCl3 2 R2 141a ± d PhSO2N CHCCl3 2b MeC(Ph) NNMe2 R=Ph, H, But, .O PhSO2NHCHCH2C NNMe2 O O Ph NCH(CCl3)NHCOR2 NH+R2CON CHCCl3 N N 149 R1 R1 CCl3 142 R1=R2=H: R3=Me (a), Ph (b); R1=R2=Me, R3=H(c); R1=R3=H, R2=Et (d). R1=H: R2=Ph, H, But, , 4-ClC6H4; O R1=Me: R2=Ph, H, But, . O The use of polyhaloaldehyde imines for C-amidoalkylation of heterocyclic compounds extends the capacity for the preparation of new derivatives. Thus chloral imines of carboxylic and carbamic acids have been used in C-amidoalkylation of pyrazo- lones. This resulted in the synthesis of 2-aryl(alkyl)-4-[1-ace- tyl(benzoyl)aminoethyl-2,2,2-trichloro]-3-alkyl(phenyl)-2-pyr-az- olones 143 in high yields.188 R3OCNHCHCCl3 O O In the case of 2-alkyl-4-hydroxypyrimidines, the reaction involves the oxygen atom and gives O-(1-acylamino-2,2,2-tri- chloroethyl)-4-hydroxypyrimidines 150a ± d; in the opinion of the authors cited,190 this is due to the steric shielding of both nucleophilic centres at the nitrogen atoms.NHCOR3 NR2 NR2 +Cl3CCH NCOR3 R1 R1 N N OCHCCl3 O 143 R3CON CHCCl3 R1=R3=H, R2=Me N NH R1=Me, Et, Bui, Ph; R2=Ph, 2,4,6-Cl3C6H2, Me; R3=Me, Ph, OEt. Et3N, D R1 N R2 R1 N R2 150a ± d O NH The reactions of N-acyl-N-(2,2,2-trichloroethylidene)amines (which were prepared in situ from the corresponding 1,2,2,2- tetrachloroethylamides) with enamines 144 and 145 occur with high diastereo- and enantioselectivity yielding amidoalkylation products, which are hydrolysed to give chiral 2,2,2-trichloroethyl- amides 146, 147.189 N CH2CHCCl3 R1 151a NHCHO N O R2 144 R3CON CHCCl3 MeO N R1= H, R2= Me: R3= H (a), Ph (b); R1= Me, R2= Pri: R3= Ph (c), 4-ClC6H4 (d).Long refluxing of the imine 3a with 4-hydroxy-2-methylpyr- imidine induces C-amidoalkylation involving the methyl group of the pyrimidine, which affords 2-(2-benzamido-3,3,3-trichloro- propyl)-4-hydroxypyrimidine 151b.190 The product 151a with a 145600 similar structure was also isolated when the pyrimidine 150a was heated in the presence of triethylamine.190 O O PhCON CHCCl3 NH NH N N Me CH2CHCCl3 151b NHCOPh C-Amidoalkylation of 4-hydroxy-6-methylpyrimidine does not proceed even under rigorous conditions.190 When 4-methoxy-6-phenacylpyrimidine 152 reacts with N-(1,2,2,2-tetrachloroethyl)amide of 4-chlorobenzoic acid in the presence of Et3N, monoamidoalkylation product 153 is formed.When 152 reacts with the imine 3j, the activated methylene group in the side chain and the ring nitrogen atom may be amidoalkyl- ated simultaneously giving rise to compound 154.191 OMe OMe N N N CHCOPh CH2COPh NH 152 OMe a N CHCOPh N Cl3CCHNHCOC6H4Cl-4 153 OMe b N CCl3 N C(COPh)CHNHCOC6H4Cl-4 Cl3CCHNHCOC6H4Cl-4 154 (a) Cl3CCHClNHCOC6H4Cl-4, Et3N, D; (b) Cl3CCH=NCOC6H4Cl-4 (3g). C-Amidoalkylation of the closest analogues of pyrimidine 152 � 4-dimethylamino-, 4-methyl-, 4-benzoylmethyl-6-(ben- zoyl)methyl-pyrimidines � has been accomplished by introduc- ing these compounds into reactions with N-(1,2,2,2-tetra- chloroethyl)amide of 4-chlorobenzoic acid in acetonitrile in the presence of triethylamine.191 The alkylation involves the benzoyl- methylene group and affords compounds of the type 153 � 4-dimethylamino-, 4-methyl-, 4-benzoylmethyl-6-[3,3,3-trichlo- ro-2-(4-chlorobenzoylamino)-1-benzoylpropyl]pyrimidines.191 Under similar conditions, amidoalkylation of some 5-R- uracils by tetrachloroethylamides of benzoic acids involves both nitrogen atoms and affords 1,3-bis[2,2,2-trichloro-1-(4-chloro- benzoylamino)ethyl]-5-R-pyrimidine-2,4-diones 155.192 O R NEt3 NH +Cl3CCHNHCC6H4Cl-4 MeCN O O Cl NH O O CCl3 R NCHNHCC6H4Cl-4 O N Cl3CCHNHCC6H4Cl-4 O 155 R=H, Ph, OMe.G G Levkovskaya, T I Drozdova, I B Rozentsveig, A N Mirskova The reaction of 2-thiouracil with tetrachloroethylamides in the presence of triethylamine follows a more complex pathway, which includes not only amidoalkylation at the nitrogen and sulfur atoms but also dehydrochlorination; this gives 2-[(1-ami- do-2,2-dichlorovinyl)thio]-4-(1-amido-2,2,2-trichloroethoxy)pyri- midines 156.193 OH O OCH(CCl3)NHCOR 2Cl3CCHClNHCOR N NH N 2NEt3 SH N S SCNHCOR N NH 156 CCl2 R=But, Ph, C6H4Cl-4, OMe.7. Other reactions of polyhaloaldehyde imines Reduction of imines presents interest as a method for the synthesis of amines. Only a single example of diastereoselective reduction of an azomethine group in cyclic dichlorocamphorsulfonylimine 157 to camphorsultam 158 (yield 93%± 97%) was reported.194 In the case of the corresponding bromo-derivative, debromination is the predominant reaction route.Cl Cl NaBH4, MeOH Cl Cl N NH O2S O2S 157 158 It was found 195 that when benzylsulfonamide reacts with chloral in the presence of trifluoromethanesulfonic anhydride or another similar reagent, the chloral benzylsulfonylimine formed at the first stage undergoes intra- and intermolecular heterocycli- sation under the reaction conditions to afford 3,4-dihydro-1H- 1,3-benzothiazine 2,2-dioxide 159 (yield 3%± 58%) and 1,3,5- tribenzylsulfonylhexahydrotriazine 160 (yield 17%). (F3CSO2)2O PhCH2SO2 PhCH2SO2NH2+Cl3CCHO N CHCCl3 CCl3 PhCH2O2S SO2CH2Ph N N SO2 + NH N CCl3 Cl3C 159 CCl3 160 SO2CH2Ph 8. Reactivity of arylsulfonyliminopolychlorocyclohexenes The reactions of bis(arylsulfonylimino)-2,3,5,5,6,6-hexachloro- cyclohex-2-enes 27b,c with reducing agents, organophosphorus and organosilicon compounds, alcohols, primary aromatic amines, secondary cyclic amines 196, 197 and with hydrazoic acid 198 have been studied.When the compounds 27b,c react with Na2S2O4, zinc in acetic acid or with dialkyl phosphites, elimination of two chlorine atoms from the 5- and 6-positions occurs to give the benzoid structure 161.196 The reactions with methanol give products 162 resulting from addition to one azomethine bond. It was shown that 1,4- bis(arylsulfonylimino)hexachlorocyclohex-2-enes 27b,c do not react with any other alcohols.196 The reactions of the bis(sulfonylimines) 27b,c with aromatic amines containing no electron-withdrawing groups in the aro- matic ring also involve the addition to the C=N bond and give rise to compounds 163.196N-Functionally substituted imines of polychlorinated (brominated) aldehydes and ketones Cl Cl [H] NHR RHN Cl Cl 161 Cl Cl Cl Cl RHN MeOH NR MeO Cl Cl 162Cl Cl Cl Cl Cl Cl Cl Cl RHN XC6H4NH2 NR RN NR XC6H4NH Cl Cl Cl 163 Cl 27b,c Cl Cl Cl Cl NH Z RN NR Cl Z N 164Cl Cl OPri 1.Me3SiOP(OPri)2 2. H2O N(R)PO RHN OPri Cl Cl 165 R=4-MeC6H4SO2 (b), 4-ClC6H4SO2 (c); X=H, 4-Me, 3-Me; Z=O, CH2. The reactions with morpholine and piperidine occur as nucleophilic substitution of the chlorine atom at the sp2-hybri- dised carbon atom of the bisimine and give products 164.196 Two processes�elimination of two chlorine atoms and 1,6-addition to the system of conjugated double bonds�occur simultaneously in the reactions of 27b,c with trimethylsilyl diisopropyl phosphite.The addition products are subsequently hydrolysed with elimina- tion of the trimethylsilyl group, which leads to compounds 165.196 When the bis(sulfonylimines) 27b,c react with pyridine, this gives the products of substitution of the vinylic halogen atom, N-[3,6-di(arylsulfonylimino)-2,4,4,5,5-pentachlorocyclohex-1-en- yl]pyridinium chloride 166; on exposure to UV radiation or under electron impact, these compounds eliminate two chlorine atoms to give quinone structures, N-[3,6-di(arylsulfonylimino)-2,4,5-tri- chlorocyclohex-1,4-dienyl]pyridinium chlorides 167.196 Cl Cl Cl Cl N hn 27b,c NSO2Ar ArSO2N + Cl7 N Cl166 Cl Cl NSO2Ar ArSO2N Cl7 N +Cl167 Ar=4-MeC6H4 (b), 4-ClC6H4 (c).601 The reaction of 4-arylsulfonyliminohexachlorocyclohex-2-en- 1-ones 28b,d with pyridine results in the isolation of arenesulf- amides and compounds 168, formed upon the substitution of a pyridinium ion for the chlorine atom in the resulting 1,4-dioxo- 2,3,5,5,6,6-hexachlorocyclohex-2-ene.197 It is of interest that 4-arylsulfonylimino-2,2,3,3-tetrachloro-1- oxo-1,2,4,4-tetrahydronaphthalenes 29 do not react with pyri- dine. Cl Cl Cl Cl Cl Cl O O ArSO2N 7ArSO2NH2 Cl7 Cl N +Cl Cl 28b,d 168 Ar=4-ClC6H4 (b), 4-MeC6H4 (d).The reaction of the imines 28a ± c with hydrazoic acid occurs as nucleophilic replacement of the chlorine atom at the sp2- hybridised carbon atom located at the 2-position, accompanied by competing elimination of chlorine with restoration of the benzoid structure.198 Cl Cl Cl Cl Cl Cl Cl Cl HN3 O O ArSO2N ArSO2N Cl Cl Cl N3 28a ± c Ar=Ph (a), 4-MeC6H4 (b), 4-ClC6H4 (c). V. Conclusion Analysis of the published data indicates that N-functionally substituted halogenated imines are highly stable and can be prepared by reactions of N,N-dihaloamides with 1,2-polyhalo- ethenes, transformation of an isocyanate bond and by other non- trivial methods, which require further development.Electronegative substituents at theC=Nbond account for the high reactivity of N-functionally substituted halogenated imines towards nucleophiles. A specific feature of these reactions is that nucleophiles attack the sp2-hybridised carbon atom of the C=N bond. The enhanced electrophilicity of the C=N bond in polyhaloaldimines permits the use of acyl- and sulfonyl-imines for C-amidoalkylation of aromatic and heterocyclic compounds. Search for conditions under which acyl- and phosphonyl-imines of polyhalogenated aldehydes and ketones could be involved in similar reactions appears to be topical task. Chloral acyl- and alkoxycarbonyl-imines can enter into cyclo- addition reactions as heterodienes but can also act as dienophiles.Unfortunately, imines of the largest class, polyhaloalkylphospho- nylimines, have not been studied in cycloaddition reactions. Halogen-containing imines are of interest regarding investi- gation of the enamide ± imine tautomerism and syn ± anti isomer- ism of azomethines and determination of the energy parameters of tautomerisation and isomerisation. The synthetic potential of imines is now far from being completely utilised. The data on the reactivity of halogenated imines can serve as the basis for targeted design of required structures with important biogenic groups: acyl-, phosphonyl- and sulfonyl-amide groups and polyhalomethyl groups. It can be expected that more extensive studies along this line would permit the researchers to perform intra- and intermolecular heterocycli- sation of imines, reactions involving haloalkyl groups, etc.602 References 2.S Patai (Ed.) The Chemistry of the Carbon ± Nitrogen Double Bond 1. D Barton,WD Ollis Comprehensive Organic Chemistry Vol. 3 (Oxford: Pergamon Press, 1979) (New York: Interscience, 1970) 3. S Patai (Ed.) The Chemistry of Double Bonded Functional Groups (New York: Interscience, 1977) 4. R W Layer Chem. Rev. 63 489 (1963) 5. W N Speckamp, H Hiemstra Tetrahedron 41 4367(1985) 6. O N Bel'kova, G G Levkovskaya, A N Mirskova Fiz.-Tekhn. Probl. Razrab. Polezn. Iskopaemykh (1) 93 (1997) 7. USSR P. 1 432 826; Byull. Izobret. (28) 412 (1997) 8. A N Mirskova, T I Drozdova, G G Levkovskaya, I T Gogoberidze, Yu D Ochirov, V N Zarubina, I F Zhovtyi, M G Voronkov, in Fiziologicheski Aktivnye Veshchestva (Physiologically Active Substances) (Kiev: Naukova Dumka, 1989) p.84 9. I B Rozentsveig, G G Levkovskaya, A G Stupina, A N Mirskova, in International Conference on Natural Products and Physiologically Active Substances (ICNPAS-98) (Abstracts of Reports) (Novosibirsk: Novosibirck Institute of Organic Chemistry, 1998) p. 152 10. T I Drozdova, A N Mirskova, A N Nikitin, Yu D Ochirov, in Virusnye Rikketsioznye i Bakterial'nye Infektsii, Perenosimye Kleshchami (Tez. Dokl. Mezhd. Konf.) [Virus Rickettsiosal and Bacterial Infections Carried by Mites (Abstracts of Reports of the International Conference)] (Irkutsk: Institute of Epidemiology and Microbiology, Siberian Branch, Russian Academy of Medical Sciences, 1996) p.153 11. B S Drach, V S Brovarets, O B Smolii Sintez Azotsoderzhashchikh Geterotsiklicheskikh Soedinenii na Osnove Amidoalkiliruyushchikh Agentov (Synthesis of Nitrogen-Containing Heterocyclic Compounds Based on Amidoalkylating Agents) (Kiev: Naukova Dumka, 1992) 12. I L Knunyants, V R Polishchuk Usp. Khim. 45 1139 (1976) [Russ. Chem. Rev. 45 573 (1976)] 13. A V Fokin, A T Uzun, V P Stolyarov Usp. Khim. 46 1995 (1977) [Russ. Chem. Rev. 46 1057 (1977)] 14. V L Dyatkin, R N Makarov, I L Knunyants Tetrahedron 27 51 (1971) 15. N P Gambaryan Usp. Khim. 45 1251 (1976) [Russ. Chem. Rev. 45 630 (1976)] 16. A V Fokin, A F Kolomiets, N V Vasil'ev Usp. Khim. 53 398 (1984) [Russ. Chem. Rev.53 238 (1984)] 17. A C Cope (Ed.) Organic Reactions Vol. 14 (New York: Wiley, 1965) 18. N De Kimpe, N Schamp Org. Prep. Proceed. Int. 11 115 (1979) 19. N De Kimpe, R Verhe, L De Buyck, N Schamp Org. Prep. Proceed. Int. 12 49 (1980) 20. I Malassa, D Matthies Chem.-Ztg. 111 (6) 181 (1987) 21. I Malassa, D Matthies Chem.-Ztg. 111 (9) 253 (1987) 22. M G Voronkov, A N Mirskova Zh. Vsesoyuz. Khim. O-va im. 23. A N Mirskova, T I Drozdova, G G Levkovskaya, M G Voronkov 24. R V Kaberdin, V I Potkin Usp. Khim. 63 679 (1994) [Russ. Chem. D I Mendeleeva 30 294 (1985) a Usp. Khim. 58 417 (1989) [Russ. Chem. Rev. 58 250 (1989)] Rev. 63 691 (1994)] 25. N N Labeish, A A Petrov Usp. Khim. 58 1844 (1989) [Russ. Chem. Rev. 58 1048 (1989)] 30. N W Hirwe, B V Patil Proc.Indian Acad. Sci. 13A 273 (1941); Chem. 26. G Giesemann, I Ugi Synthesis 788 (1983) 27. N De Kimpe, R Verhe, L De Buyck, L Moems, N Schamp Synthesis 43 (1982) 28. F Feist Chem. Ber. 45 945 (1912) 29. N W Hirwe, K N Rana J. Univ. Bombay 7 (3) 174 (1938); Chem. Abstr. 33 3778 (1939) Abstr. 35 6250 (1941) 31. A Thomas, T George Agra Univ. J. Research 9 (1) 11 (1960); Chem. Abstr. 55 24 517 (1961) 32. R Albrecht, G Kresze, B Mlakar Chem. Ber. 97 483 (1964) 33. B S Drach, A D Sinitsa, A V Kirsanov Zh. Obshch. Khim. 39 1480 (1969) b 34. N de Kimpe, R Verhe, L de Buyck, W Dejonghe, N Schamp Bull. Soc. Chim. Belg. 85 763 (1976) G G Levkovskaya, T I Drozdova, I B Rozentsveig, A N Mirskova 35. F Weygand,W Steglich, I Lengyel, F Fraunberger, A Maierhofer, W Oettmeier Chem.Ber. 99 1944 (1966) 36. H Ulrich, B Tusker, A A R Sayigh J. Org. Chem. 33 2887 (1968) 37. B S Drach, A D Sinitsa, A V Kirsanov Zh. Obshch. Khim. 39 2192 (1969) b 38. B S Drach, E P Sviridov, T Ya Lavrenyuk Zh. Org. Khim. 10 1271 (1974) c 39. B S Drach, A D Sinitsa, A V Kirsanov Zh. Org. Khim. 5 2181 (1969) c 40. B S Drach, A D Sinitsa, A V Kirsanov Zh. Obshch. Khim. 40 1933 (1970) b 41. B S Drach, T P Popovich, A A Kisilenko, O M Polumbrik Zh. Org. Khim. 15 31 (1979) c 42. B S Drach, E P Sviridov Zh. Org. Khim. 9 1074 (1973) c 43. B S Drach, E P Sviridov Zh. Obshch. Khim. 44 348 (1974) b 44. B S Drach, E P Sviridov Zh. Org. Khim. 9 680 (1973) c 45. B S Drach, G N Mis'kevich Zh. Org.Khim. 11 316 (1975) c 46. H Zinner,W E Siems, G Erfurt J. Prakt. Chem. 316 63 (1974) 47. D Matthies, I Malassa Synth. Commun. 177 (1985) 48. B S Drach, A P Martynyuk, G N Mis'kevich, O P Lobanov Zh. Org. Khim. 13 1404 (1977) c 49. I N Zhmurova, B S Drach, A V Kirsanov Zh. Obshch. Khim. 35 1018 (1965) b 50. B S Drach, A D Sinitsa Zh. Obshch. Khim. 38 2778 (1968) b 51. H Teichman, M Schnell J. Prakt. Chem. 329 871 (1987) 52. G S Zaitseva, O P Novikova, L I Livantsova, V S Petrosyan, Yu I Baukov Zh. Obshch. Khim. 61 1389 (1991) b 53. G S Zaitseva, L I Livantsova, O P Novikova Zh. Obshch. Khim. 64 1750 (1994) b 54. G Kresze, R Albrecht Angew. Chem. 74 781 (1962) 55. G Kresze, R Albrecht Angew. Chem., Int. Ed. Engl. 595 (1962) 56. G Kresze, A Maschke, R Albrecht, K Bederke, H P Patzschke, H Smalla, A Trede Angew.Chem. 74 135 (1962) 57. F Kasper, S Dathe J. Prakt. Chem. 327 1041 (1985) 58. G Kresze,W Wucherpfennig Chem. Ber. 101 365 (1968) 59. BRD P. 2 645 280; Chem. Abstr. 89 5908 (1978) 60. G Kresze,W Wucherpfennig Angew. Chem. 79 109 (1967) 61. L N Markovskii, A V Solov'ev, N P Kolesnik, Yu G Shermolovich Zh. Org. Khim. 31 656 (1995) c 62. Ya G Bal'on, V A Smirnov Zh. Org. Khim. 16 738 (1980) c 63. V I Gorbatenko, Yu I Matveev, M N Gertsyuk, L I Samarai Zh. Org. Khim. 20 2543 (1984) c 64. Yu I Matveev, V I Gorbatenko Zh. Obshch. Khim. 62 324 (1992) b 65. O B Smolii, S Ya Panchishin, L V Budnik, E A Romanenko, B S Drach Zh. Obshch. Khim. 67 391 (1997) b 66. A D Sinitsa, N A Parkhomenko, S V Bonadyk Zh.Org. Khim. 12 974 (1976) c 67. A D Sinitsa, S V Bonadyk, L N Markovskii Zh. Org. Khim. 13 721 (1977) c 68. A D Sinitsa, N A Parkhomenko Zh. Obshch. Khim. 47 609 (1977) b 69. A D Sinitsa, N A Parkhomenko, V S Krishtal', L N Markovskii Zh. Obshch. Khim. 49 130 (1979) b 70. I B Rozentsveig, D A Matveev, in Molodezhnaya Nauchnaya Shkola po Organicheskoi Khimii Pamyati Akademika I Ya Postovskogo (Tez. Dokl.), Ekaterinburg, 1998 [Young Research School on Organic Chemistry Dedicated to the Memory of Academician I Ya Postovskii (Abstracts of Reports), Ekaterinburg, 1998] p. 154 71. B Xu, S Zhu Heteroat. Chem. 8 (4) 309 (1997); Chem. Abstr. 127 190 795 (1997) 72. Ya G Bal'on, V E Paranyuk Zh. Org. Khim. 19 1346 (1983) c 73.A N Mirskova, G G Levkovskaya, A A Bryuzgin, T I Drozdova, I D Kalikhman,M G Voronkov Zh. Org. Khim. 26 140 (1990) c 74. A N Mirskova, T I Drozdova, G G Levkovskaya, O B Bannikova, I D Kalikhman,M G Voronkov Zh. Org. Khim. 22 763 (1986) c 75. T I Drozdova, A N Mirskova, G G Levkovskaya Zh. Org. Khim. 26 1002 (1990) c 76. I T Evstaf'eva, A N Mirskova, G G Levkovskaya, O B Bannikova Zh. Org. Khim. 26 998 (1990) c 77. T I Drozdova, A N Mirskova, G G Levkovskaya, I D Kalikhman, M G Voronkov Zh. Org. Khim. 23 1685 (1987) c 78. T I Drozdova, G G Levkovskaya, A N Mirskova Zh. Org. Khim. 27 2281 (1991) c603 N-Functionally substituted imines of polychlorinated (brominated) aldehydes and ketones 121. O P Lobanov, A P Martynyuk, B S Drach Zh.Obshch. Khim. 50 2248 (1980) b 122. B S Drach, A D Sinitsa, A V Kirsanov Zh. Obshch. Khim. 40 934 (1970) b 123. USSR P. 803 361; Byull. Izobret. (36) 259 (1987) 124. A N Mirskova, T I Drozdova, G G Levkovskaya, O B Bannikova, I D Kalikhman,M G Voronkov Zh. Org. Khim. 18 1407 (1982) c 125. T I Drozdova, A N Mirskova, G G Levkovskaya, I D Kalikhman,M G Voronkov Zh. Org. Khim. 24 1240 (1988) c 126. A N Mirskova, T I Drozdova, G G Levkovskaya, I D Kalikhman,M G Voronkov Zh. Org. Khim. 23 1248 (1987) c 127. G G Levkovskaya, I T Evstaf'eva, A N Mirskova, S N Zhuravlev, V G Kul'nevich Zh. Org. Khim. 23 1991 (1987) c 128. T I Drozdova, A N Mirskova Zh. Org. Khim. 33 1591 (1997) c 129. I T Evstaf'eva, G G Levkovskaya, O B Kozyreva, A N Mirskova Zh.Org. Khim. 33 487 (1997) c 130. N Zinner,W E Siems, G Erfurt J. Prakt. Chem. 316 491 (1974) 131. A N Mirskova, I T Gogoberidze, G G Levkovskaya, I D Kalikhman,M G Voronkov Zh. Org. Khim. 20 1502 (1984) c 132. GDR P. 270 707; Chem. Abstr. 112 98 548 (1990) 133. B S Drach, A D Sinitsa, A V Kirsanov Zh. Obshch. Khim. 39 1940 (1969) b 134. USSR P. 1 336 494; Byull. Izobret. (3) 237 (1990) 135. V L Krasnov, G I Vasyanina, I V Bodrikov Zh. Org. Khim. 23 441 (1987) c 136. V L Krasnov, G I Vasyanina, I V Bodrikov Zh. Org. Khim. 27 1552 (1991) c 137. A N Mirskova, G G Levkovskaya, I T Gogoberidze, I D Kalikhman,M G Voronkov Zh. Org. Khim. 21 269 (1985) c 138. N G Zabirov, R A Cherkasov Zh. Obshch. Khim. 60 1251 (1990) b 139.A N Mirskova, G G Levkovskaya, I T Gogoberidze, T I Drozdova, I D Kalikhman,M G Voronkov Zh. Org. Khim. 19 1110 (1983) c 140. A N Mirskova, T I Drozdova, G G Levkovskaya, O B Bannikova, I D Kalikhman,M G Voronkov Zh. Org. Khim. 17 1108 (1981) c 141. N G Zabirov, N I Galyautdinov, V A Shcherbakova, R A Cherkasov Zh. Obshch. Khim. 60 1247 (1990) b 142. I T Evstaf'eva, G G Levkovskaya, A N Mirskova, in 19-ya Vsesoyuz. Konf. po Khimii i Tekhnologii Organicheskikh Soedinenii Sery. (Tez. Dokl.), Kazan' 1995 [The 19th All-Union Conference on Chemistry and Technology of Organic Compounds of Sulfur (Abstracts of Reports), Kazan, 1995] Part 1, p. 69 143. USSR P. 908 012; Byull. Izobret. (7) 292 (1997) 144. A A Bryuzgin,G G Levkovskaya,A N Mirskova, I D Kalikhman Zh.Org. Khim. 26 1296 (1990) c 145. I T Evstaf'eva, G G Levkovskaya, O B Bannikova, A N Mirskova Zh. Org. Khim. 29 794 (1993) c 146. B S Drach, O P Lobanov Zh. Obshch. Khim. 45 472 (1975) b 147. A D Sinitsa, B S Drach Zh. Obshch. Khim. 43 211 (1973) b 148. VanK Burger, J Fehn, J Albanbauer, J Friedl Angew. Chem. 84 258 (1972) 149. B A Arbuzov, N A Polezhaeva, V S Vinogradova Izv. Akad. Nauk Ser. Khim. 1112 (1973) d 150. M G Zimin, N G Zabirov, A N Pudovik Zh. Obshch. Khim. 49 2189 (1979) b 151. M G Zimin, N G Zabirov, R A Cherkasov, A N Pudovik Zh. Obshch. Khim. 50 1458 (1980) d 152. C Kashima, Y Aoki, Y Omoto J. Chem. Soc., Perkin Trans. 1 2511 (1975) 153. L Duhamel, J-Y Yalnot Tetrahedron Lett. 3167 (1974) 79. N N Labeish, Yu I Porfir'eva, A A Petrov Zh.Org. Khim. 21 659 (1985) c 80. T I Drozdova, G G Levkovskaya, A N Mirskova Zh. Org. Khim. 28 1236 (1992) c 81. Ya G Bal'on, R N Moskaleva Zh. Org. Khim. 14 147 (1978) c 82. Ya G Bal'on, R N Moskaleva Zh. Org. Khim. 19 2456 (1983) c 83. T I Drozdova, O B Kozyreva, G G Levkovskaya, A N Mirskova Zh. Org. Khim. 30 381 (1994) c 84. T I Drozdova, O B Kozyreva, A N Mirskova Zh. Org. Khim. 33 276 (1997) c 85. K Grohe, E Degener, H Holtschmidt, H Heitzer Liebigs Ann. Chem. 730 133 (1969) 86. Ya G Bal'on, V A Smirnov Zh. Org. Khim. 26 2377 (1990) c 87. N V Kolotilo, A A Sinitsa, P P Onys'ko Izv. Akad. Nauk, Ser. Khim. 2101 (1998) d 88. C-G Shin, Y Sato, J Yoshimura Bull. Chem. Soc. Jpn. 49 1909 (1976) 89. B S Drach, V A Kovalev, T Ya Lavrenyuk Zh.Org. Khim. 11 1913 (1975) c 90. A P Avdeenko, N V Velichko, A A Tolmachev, V V Pirozhenko, E A Romanenko Zh. Org. Khim. 30 136 (1994) c 91. A P Avdeenko,N V Velichko, E A Romanenko, V V Pirozhenko, V I Shurpach Zh. Org. Khim. 27 2350 (1991) c 92. A P Avdeenko, N V Velichko, E A Romanenko, V V Pirozhenko Zh. Org. Khim. 27 1747 (1991) c 93. A D Sinitsa, V S Krishtal', V I Kal'chenko, L N Markovskii Zh. Obshch. Khim. 50 2409 (1980) 94. I V Martynov, A N Ivanov, T A Epishina, V B Sokolov Izv. Akad. Nauk, Ser. Khim. 1086 (1987) d 95. N V Kolotilo, P P Onys'ko, A A Sinitsa Zh. Obshch. Khim. 67 160 (1997) b 96. A V Kirsanov, G I Derkach Zh. Obshch. Khim. 26 2009 (1956) b 97. A V Kirsanov, G I Derkach Zh.Obshch. Khim. 26 2631 (1956) b 98. V Ya Semenii, G F Solodushchenko, V P Kukhar', V A Bondar' Ukr. Khim. Zh. 53 397 (1987) 99. V I Shevchenko, A A Koval' Zh. Obshch. Khim. 37 1111 (1967) b 100. V M Amirkhanov, V A Trush Zh. Obshch. Khim. 65 1120 (1995) b 101. V I Boiko, L I Samarai, V V Pirozhenko Zh. Obshch. Khim. 65 1054 (1995) b 102. A D Sinitsa, V S Krishtal', V I Kal'chenko Zh. Obshch. Khim. 50 1288 (1980) b 103. B S Drach, G N Mis'kevich Zh. Org. Khim. 13 1398 (1977) c 104. G G Levkovskaya, A N Mirskova, T I Drozdova, I D Kalikhman, O B Bannikova, M G Voronkov Zh. Org. Khim. 21 620 (1985) c 105. G V Dolgushin, G G Levkovskaya, I B Rozentsveig, P A Nikitin, A N Mirskova Zh. Obshch. Khim. 66 2031 (1996) b 106.G V Dolgushin, G G Levkovskaya, I B Rozentsveig, I T Evstaf'eva, A N Mirskova Zh. Obshch. Khim. 67 598 (1997) b 107. A N Mirskova, G G Levkovskaya, T I Drozdova, O B Bannikova, I D Kalikhman,M G Voronkov Zh. Org. Khim. 18 1632 (1982) c 108. A N Mirskova, G G Levkovskaya, T I Drozdova, I D Kalikhman,M G Voronkov Zh. Org. Khim. 18 452 (1982) c 109. A N Mirskova, I T Gogoberidze, G G Levkovskaya, M G Voronkov Zh. Org. Khim. 20 2235 (1984) c 110. S Brummer, A Weiss Ber. Bunsenges. Phys. Chem. 94 497 (1990) 111. B A Shainyan, A N Mirskova Usp. Khim. 48 201 (1979) [Russ. Chem. Rev. 48 107 (1979)] 112. B S Drach, V A Kovalev Zh. Org. Khim. 13 1597 (1977) c 113. A N Mirskova, E F Zorina, A S Atavin Zh. Org. Khim. 8 1150 c (1972) 114.T K Vinogradova, V V Turov, B S Drach Zh. Org. Khim. 26 1302 (1990) c 115. B S Drach, E P Sviridov, A A Kisilenko, A V Kirsanov Zh. Org. Khim. 9 1818 (1973) c 116. V V Dovlatyan, E N Ambartsumyan Arm. Khim. Zh. 22 135 (1969) 117. B S Drach, G N Mis'kevich Zh. Org. Khim. 10 2315 (1974) c 118. B S Drach, E P Sviridov, A V Kirsanov Zh. Obshch. Khim. 45 12 (1975) b 119. B S Drach, E P Sviridov Zh. Obshch. Khim. 43 1648 (1973) b 120. B S Drach, E P Sviridov, Ya P Shaturskii Zh. Obshch. Khim. 44 1712 (1974) b 154. N De Kimpe, R Verhe, N Schamp Bull. Soc. Chim. Belg. 84 701 (1975) 155. V V Shchepin, D I Efremov Zh. Org. Khim. 29 2156 (1993) c 156. B S Drach, G N Mis'kevich Zh. Org. Khim. 12 465 (1976) c 157. S M Weinreb, P M Scola Chem. Rev. 89 1525 (1989) 158. J S Sandhu, B Sain Heterocycles 26 777 (1987) 159. G Kresze, R Albrecht Chem. Ber. 97 490 (1964) 160. A N Mirskova, I T Gogoberidze, G G Levkovskaya, I D Kalikhman, O B Bannikova, A S Kisin,M G Voronkov Zh. Org. Khim. 19 1744 (1983) c 161. T Imagava, K Sisido, M Kawanisi Bull. Chem. Soc. Jpn. 46 2922 (1973)G G Levkovskaya, T I Drozdova, I B Rozentsveig, A N Mirskova 604 162. M S Raasch J. Org. Chem. 40 161 (1975) 194. F A Davis, P Zhou, B-C Chen Phosphorus Sulfur Silicon Relat. Elem. 115 85 (1996) 195. O O Orazi, R A Corral, R Bravo J. Heterocyclic. Chem. 23 1701 (1986) 196. A P Avdeenko, A L Yusina Zh. Org. Khim. 31 753 (1995) c 197. A P Avdeenko, A L Yusina Zh. Org. Khim. 31 458 (1995) c 198. A P Avdeenko, Yu V Menafova, S A Zhukova Zh. Org. Khim. 34 237 (1998) c a�Mendeleev Chem. J. (Engl. Transl.) b�Russ. J. Gen. Chem. (Engl. Transl.) c�Russ. J. Org. Chem. (Engl. Transl.) d�Russ. Chem. Bull. (Engl. Transl.) e�Chem. Heterocycl. Compd. (Engl. Transl.) 163. G R Krow, C Pyun, R Rodebaugh, J Marakowski Tetrahedron 30 2977 (1974) 164. G Krow, R Rodebaugh, J Marakowski, K C Ramey Tetrahedron Lett. 1899 (1973) 165. Yu A Arbuzov, E I Klimova, N D Antonova, Yu V Tomilov Zh. Org. Khim. 10 1164 (1974) c 166. T N Maksimova, V B Mochalin, B V Unkovskii Khim. Geterotsikl. Soedin. 273 (1980) e 167. A D Sinitsa, B S Drach, A A Kisilenko Zh. Org. Khim. 9 685 (1973) c 168. O P Novikova, L I Livantsova,G S Zaitseva Zh. Obshch. Khim. 59 2630 (1989) b 169. O P Novikova, L I Livantsova, E A Zubkov, G S Zaitseva, Yu I Baukov, in Sintez i Reaktsionnaya Sposobnost' Organicheskikh Soedinenii Sery (Tez. Dokl. 17-i Vsesoyuz. Konf.), Tbilisi, 1989 [Synthesis and Reactivity of Organic Compounds of Sulfur (Abstracts of Reports of the 17th All-Union Conference), Tbilisi, 1989] p. 214 170. L I Livantsova, O P Novikova, G S Zaitseva, Yu I Baukov Zh. Obshch. Khim. 59 2293 (1989) b 171. T Akiyama, N Urasato, T Imagawa,M Kawanisi Bull. Chem. Soc. Jpn. 49 1105 (1976) 172. G S Zaitseva,O P Novikova, L I Livantsova Zh. Obshch. Khim. 65 701 (1995) b 173. G S Zaitseva, L I Livantsova Zh. Obshch. Khim. 65 804 (1995) b 174. I V Konovalova, Yu G Trishin, L A Burnaeva, E K Khusnutdinova, V N Chistokletov, A N Pudovik Zh. Obshch. Khim. 58 1292 (1988) b 175. R Consonni, P Dalla Croce, R Ferraccioli, C La Rosa J. Chem. Res. (S) (1) 32 (1992); Ref. Zh. Khim. 17 Zh 308 (1992) 176. A N Mirskova, G G Levkovskaya, A A Bryuzgin, I D Kalikhman,M G Voronkov Zh. Org. Khim. 22 2173 (1986) c 177. A N Mirskova, G G Levkovskaya, A A Bryuzgin, I D Kalikhman,M G Voronkov Zh. Org. Khim. 25 1695 (1989) c 178. R W Hoffmann, K Steinbach, W Lilienblum Chem. Ber. 109 1759 (1976) 179. H E Zaugg Synthesis 49 (1970) 180. I T Gogoberidze, G G Levkovskaya, A N Mirskova, I D Kalikhman, O B Bannikova, M G Voronkov Zh. Org. Khim. 21 633 (1985) c 181. T I Drozdova, A N Mirskova Zh. Org. Khim. 34 948 (1998) c 182. I B Rozentsveig, Candidate Thesis in Chemical Sciences, Irkutsk Institute of Chemistry, Irkutsk, 1999 183. A N Mirskova, T I Drozdova, G G Levkovskaya, B F Kukharev, I D Kalikhman,M G Voronkov Zh. Org. Khim. 25 1312 (1989) c 184. I T Gogoberidze, G G Levkovskaya, A N Mirskova, M G Voronkov Zh. Org. Khim. 20 1100 (1984) c 185. A N Mirskova, T I Drozdova, in V Vsesoyuz. Konf. po Khimii Azotsoderzhashchikh Geterotsiklicheskikh Soedinenii (Tez. Dokl.), Chernogolovka, 1991 [The Vth All-Union Conference on the Chemistry of Nitrogen-Containing Heterocyclic Compounds (Abstracts of Reports), Chernogolovka, 1991] p. 255 186. I B Rozentsveig, G G Levkovskaya, A N Mirskova, O B Kozyreva Zh. Org. Khim. 33 623 (1997) c 187. G G Levkovskaya, E V Krivonos, I B Rozentsveig, A N Mirskova, in Simpozium po Khimii i Primeneniyu Fosfor-, Sera- i Kremniiorganicheskikh Soedinenii `Peterburgskie Vstrechi 98' (Sb. Nauch. Tr.), S.-Peterburg, 1998 (Proceedings of Symposium on Chemistry and Application of Organophosphorus, Organosolfur and Organosilicon Compounds `Petersburg Meetings 98', St-Petersburg,1998) p. 126 188. D Sicker,W BoÈ hlmann, D Bendler, G Mann Synthesis 493 (1987) 189. W Miltz,W Steglich Synthesis 750 (1990) 190. L P Prikazchikova, L I Rybchenko, S V Klyuchko, V V Pirozhenko, B S Drach Khim. Geterotsikl. Soedin. 1424 (1994) e 191. B M Khutova, S V Klyuchko, E A Romanenko, L P Prikazchikova, V M Cherkasov Ukr. Khim. Zh. 56 396 (1990) 192. B M Khutova, S V Klyuchko, L P Prikazchikova Khim. Geterotsikl. Soedin. 512 (1991) e 193. S V Klyuchko, B M Khutova, A B Rozhenko, E A Romanenko, S I Vdovenko, L I Rybchenko, L P Prikazchikova Khim. Geterotsikl. Soedin. 9
ISSN:0036-021X
出版商:RSC
年代:1999
数据来源: RSC
|
5. |
Combustion of char-forming polymeric systems |
|
Russian Chemical Reviews,
Volume 68,
Issue 7,
1999,
Page 605-614
Aleksei V. Antonov,
Preview
|
|
摘要:
Russian Chemical Reviews 68 (7) 605 ± 614 (1999) Combustion of char-forming polymeric systems A V Antonov, I S Reshetnikov, N A Khalturinskij Contents I. Introduction II. Intumescent systems III. Types of intumescent systems IV. Chemical reactions occurring in the intumescence V. Heat-insulating characteristics of foamed cokes formed upon combustion and thermal degradation of polymeric systems containing intumescent fire retardants VI. Conclusion Abstract. Chemical reactions occurring in the combustion and decomposition of char-forming polymeric systems and the effects of the nature of the polymer and additives on the reaction pathways are discussed. The thermal characteristics of the chars formed upon combustion are considered. The bibliography includes 86 references.I. Introduction The lowering of ignitability and flammability of polymers and the creation of fire retarding materials is an urgent problem which demands immediate solution. Regulation concerning the prohib- ition or limited use of flammable polymeric materials in industrial and civil engineering, construction of means of transportation (aircraft, automobiles, railway cars, ships), electrical and elec- tronic engineering and production of special-purpose materials have been passed in many countries. This circumstance has given a strong impetus to the studies aimed at elaboration and practical application of fire retarding polymeric materials. In the coming years, many commodities prepared on the basis of polymeric materials will be manufac- tured only in a fireproof form.The production of fire retarding polymeric materials is following three main directions, viz., synthesis of fire retarding polymeric materials, chemical and physical modification of polymers and the use of fire retardants. The first trend, which holds especially great promise, consists in the synthesis of polymers of two types, viz., polymers yielding incombustible gases upon decomposition and highly thermostable rigid-chain polymers. The first group of compounds includes fluorine- or nitrogen-containing polymers (e.g., fluoroplastics, A V Antonov, I S Reshetnikov N S Enikolopov Institute of Synthetic Polymeric Materials, Russian Academy of Sciences, 117393 Moscow, ul. Profsoyuznaya 70, Russian Federation.Fax (7-095) 420 22 29. Tel. (7-095) 332 58 37. E-mail: ant@lx.ispm.ac.ru (A V Antonov). Tel. (7-095) 332 58 39. E-mail: ris@lx.ispm.ac.ru (I S Reshetnikov) N A Khalturinskij N N Semenov Institute of Chemical Physics, Russian Academy of Sciences, 117977 Moscow, ul. Kosygina 4, Russian Federation. Fax (7-095) 938 21 56. Tel. (7-095) 938 21 56. E-mail: khalt@chph.ras.ru Received 4 August 1998 Uspekhi Khimii 68 (7) 663 ± 673 (1999); translated by R L Birnova #1999 Russian Academy of Sciences and Turpion Ltd UDC 541.64 : 546.14 605 606 606 609 610 613 nitroso and triazine elastomers) 1 and organoelement polymers (e.g., polyphosphazenes, polysiloxanes and polycarboranes).2 ±4 The second group comprises polymers containing heterocyclic fragments, such as polysulfones (arylox, bakelite, radel, udel, etc.), polyphenylene sulfides, liquid-crystalline polyesters (vectra, econol, etc.), polyamides (phenilon, kevlar, polymers containing mesogenic groups), polyimides (kapton, epilex, ultem, torlon, etc.) and polyheterocycles, such as polypyrazoles, polyoxazoles, poly- benzoimidazoles, polybenzooxazoles, polyquinoxalines and poly- imidazobenzophenanthrolines (the BBB polymer).5 Rigid-chain polymers combine low flammability and high thermostability with other valuable properties, e.g., high chemical and radiation stability, high durability and dielectric characteristics.Studies 6 ±8 of effects of the chemical structure and conditions of preparation of polymers on their flammability and thermal characteristics have demonstrated a good correlation between the yields of non-volatile residues formed upon decomposition of the polymers and their flammability (oxygen index, OI).The yields of carbonised products depend on the chemical structures of poly- mers.6, 7 The smaller the number of flexible `hinge' groups in the polymer molecule the higher the char yield. It is of note that the flammability of poly(ester imides) depends on the method used for their synthesis.8 It was shown also that the thermal stability of a polymer is determined by the nature of its end groups and the presence of admixtures. Unfortunately, the high cost of organoelement and high- performance polymers restricts the fields of application of materials based on them.Today, the production of low-cost polymers which employs less power-consuming synthetic proce- dures and less expensive monomers seems to have especially great potential. Chemical and physical modification of polymers is the second trend in the production of fire retarding polymeric materials. In order to decrease the flammability of carbochain polymers, carbochain monomers are copolymerised with halogen- or phos- phorus-containing monomers. In addition, chemical modification of synthetic polymeric products can be achieved by treating them with various reagents: this can be effected, e.g., by bromination, chlorination, phosphorylation 9±12 and polymer-analogous trans- formations aimed at incorporating halogen- and phosphorus- containing groups into polymers and copolymers.13 ± 15 This approach was used to obtain brominated polystyrene 16 and polyphenylene oxides,17 copolymers of vinyl acetate, ethylene and derivatives of halogen-containing acids 18 and phosphorus- containing epoxy resins (PER).19606 Physical modification of polymers is based on the treatment of the polymer surface with various types of energy (plasma, thermal and laser beams, UV or IR irradiation, etc.).Using this approach, one can obtain fire retarding polymeric materials based on various polymers.20 Unfortunately, modification very often induces significant changes in the properties of the polymers and cannot be used to decrease the flammability of industrial polymers.The third approach to the preparation of fire retarding polymeric materials, which employs fire retardants, is the most popular and efficient way. Polybrominated biphenyls and diphenyl oxides are the most widespread compounds endowed with the ability to decrease the flammability of industrial plastic materials. However, there is evidence that combustion and thermal degradation of polymeric materials based on them is accompanied by the formation of highly toxic products.21 ± 23 Thus, the need arose for fire retarding materials devoid of low-molecular-weight polybrominated com- pounds. High-molecular-weight bromine-containing fire retard- ants meet this demand. It was established 16, 17, 23 that thermal degradation of brominated polyphenylene oxides and polystyrene was not accompanied by the formation of highly toxic products either in the presence or in the absence of plastic materials.Fire retarding styrene-based plastics have been developed which contain high-molecular-weight fire retardants, their properties are comparable with those of styrene-based materials containing low-molecular-weight fire retardants.24 II. Intumescent systems The problem of formation of toxic compounds in the combustion of polymeric materials is completely eliminated if intumescent systems (IS) devoid of polybrominated diphenyls or diphenyl oxides are used. An IS represents a mixture of compounds which form foamed cokes on the polymer surface upon heating and thus prevent combustion.The fire-resistance due to foamed coke layers is associated with thermal insulation and barrier properties in mass transfer.25, 26 The superficial foamed coke layer is formed as a result of complex physical and chemical processes, each contributing to the decrease in the polymer flammability. The factors affecting the flammability of IS-containing polymers can be classified as follows: 1. Reversal of the route of the polymer thermal degradation towards the formation of foamed cokes and incombustible volatile products. 2. A change in the heat balance of combustion due to the thermal effects of chemical reactions occurring on intumescence and dilution of fuels with incombustible gaseous products. 3. Thermal insulation of the polymer surface from reverse heat flows.4. Prevention of fuel diffusion into the combustion zone. 5. Prevention of oxygen diffusion from the environment to the polymer surface. The first two factors exert their influence on combustion through chemical reactions between individual components of the polymeric composition, while the rest are closely connected with the physical structure of the foamed coke. III. Types of intumescent systems 1. The use of intumescent systems for decreasing the flammability of carbochain polymers The following compounds can be used as IS components:27, 28 inorganic acids (in free form or as derivatives which generate acids on heating), polyhydroxy derivatives, nitrogen and halogen derivatives. This classification is arbitrary, since any component may contain several functions. The presence of a nitrogen- or a halogen-containing compound is not always necessary.It is A V Antonov, I S Reshetnikov, N A Khalturinskij believed that acids favour carbonisation of polyhydroxy deriva- tives (carbon source) due to dehydration, whereas nitrogen- or halogen-derivatives (foaming agents) facilitate intumescence of the system due to intense degradation associated with evolution of large quantities of gaseous compounds. IS that are to be used in polymeric materials should meet the following requirements :29 1. IS must be resistant to temperatures above 200 8C, which are often used in polymer processing. 2. Thermal degradation of polymers must not interfere with the foaming reaction.3. IS should produce a protective foamed coke layer over the whole polymer surface despite the dilution of the additive with the polymer. 4. IS should not induce any significant changes in the physical and mechanical properties of the polymer or interact with other additives (pigments, stabilisers, etc.). 5. The mineral acid formed upon heating of the corresponding derivative used as IS should be high-boiling and should not be a strong oxidant. Phosphoric acid derivatives (amides, esters, imides) meet these requirements most completely. Of these, high-molecular-weight ammonium polyphosphate (APP), 7[P(ONH4)(O)O]n7, where n=20 ± 3000, is the most popular fire retardant.30 The IS containing APP used for decreasing the flammability of carbochain polymers are listed in Table 1.Several other compounds tested as fire retardants for various polymers are listed in Table 2. These data suggest that efficient reduction of flammability demands a high degree of polymer filling, which inevitably increases its cost, deteriorates its physical and mechanical proper- ties and impedes its processing. The mechanisms of action of IS systems are different and depend on the structure of the polymer and the additives. The polymer can take part in the formation of the carbonised layer on the surface of the burning material, but the foamed coke layer can be also formed in the absence of the polymer. 2. The use of intumescent fire retardants for the preparation of polymeric materials based on polycondensation polymers Polycarbonates are widely employed in electrical engineering, instrument making and automobile construction. These com- pounds are endowed with valuable and unique properties, such as higher thermal stability and lower flammability compared with polyalkenes and styrene-based plastic materials.Many branches of industry have a great demand for fire retarding polymeric materials based on polycarbonates (PC). The flammability of PC can be lowered if alkali metal salts of organic acids are used as additives. The preparation of materials with OI above 35% demands only small quantities of the additive (43%). The mechanism of action of these additives has been studied.48, 49 It is known that thermal degradation of polycarbon- ates occurs through two main routes.The first route consists in isomerisation followed by the formation of aryloxybenzoic acids 1. Further thermal conversions of the thermal degradation products yield ethers, xanthones and cross-linked structures: D PC COOH Me Me O C C Me Me 1 cross-linking. 7CO2,7H2O The second route involves intramolecular transfer reactions, resulting in cyclic oligomers of type 2. The reaction of product 2 with water and subsequent elimination of CO2 yield phenols:Combustion of char-forming polymeric systems Table 1. APP-based intumescent systems. Polymer Polypropylene Shock-resistant polystyrene Polystyrene Polyacrylonitrile Table 2. Fire retardants as additives to polyalkenes. Polymer Polypropylene Additive APP content in the polymer (%) Mass ratio APP : additive 25 ± 30 2 : 1 a condensation product of substituted ureas with formaldehyde 30 ethylenebis[tris(2-cyanoethyl)]phosphonium 2 : 1 bromide N NCH2 20 3 : 1 n O 20 3 : 1 NC(O)O N n NC(O)O N O 20 3 : 1 n 20 3 : 1 N NC(O)O(CH2)4O n Me N NC(O)OCH2CO 20 3 : 1 Me n 24 7 30 30 30 a condensation product of aromatic diisocyanates with pentaerythritol and melamine pentaerythritol phosphate 7 : 3 pentaerythritol phosphate, melamine 6 : 3 : 1 pentaerythritol, guanidinium carbonate 6 : 3 : 1 ethylenebis[tris(2-cyanoethyl)]phosphonium 1 : 1 30 bromide 24 7 a condensation product of aromatic diisocyanates with pentaerythritol and melamine hexabromocyclododecane 20 1 : 1 IS a IS content in the polymer (%) 30 27 a condensation product of P2O5 with pentaerythritol and melamine O 30 P(O)O Xá2O O O(O)PO 30 O CH2O P(O)O7X+ 2 30 O P O O a mixture of piperazine pyrophosphate and melamine phosphate 20 CH2OH 30 condensation products of triglyceryl isocyanurate with ortho-(poly)phosphoric acid O C(CH2OH)4, O P O O O CH2O P(O)O7X+ 30 O P O O 2 N N NH2 H2N N N N O 30 CH2OH, N N O P O O NH2 607 Ref. OI 30 ± 33 34 ± 35 34 30 35 26.5 35 25 36 26.5 36 25.5 36 23.5 32 37 37 26 37 38 37 36 34 30 38 37 39 28 ± 30 Ref.OI 40 32 41 28.5 41 33 42 36 ± 38 43, 44 24 ± 25 37 33 37 32.5 37 34.5608 Table 2 (continued).Polymer Polypropylene A copolymer of ethylene with propylene A copolymer of ethylene with vinyl acetate Polystyrene Polyethylene N a Designation: X+=H2N D PC It has been found that the high efficacy of alkali metal salts of organic acids is determined by the fact that they accelerate thermal degradation of PC. This results in the increase in the amount of the char formed and evolution of CO2 as one of the polymer decomposition products at lower temperatures. In addition, decomposition of PC in the presence of organic acid salts is IS a X+HPO¡4 , C(CH2OH)4 O O P(O)Me, X Me(O)P O O 27 O(O)PCMe(OH)P(O)O C(CH2OH)4, Xá2OH OH C(CH2OH)4, HO(O)PCMe(OH)P(O)O7 X+ OH OH HO(O)PCMe(Cl)P(O)O7 X+ Cl OH (HO)2PCH2XCH2XCHCH2P(O)(OH)2 O OH a mixture of piperazine pyrophosphate and melamine phosphate O Me X+ HP2O¡7 , P(O)CH2 N Me O 3 O CH2O P(O)O7X+ 2 O P O OO OP(O)Me, X Me(O)P O O HO(O)PCMe(Cl)P(O)O7 X+ Cl OH (HO)2PCH2XCH2XCHCH2P(O)(OH)2 O OH O O P(O)Me, X Me(O)P O O C(CH2OH)4, HO(O)PCMe(OH)P(O)O7 X+ OH OH HO(O)PCMe(Cl)P(O)O7 X+ OH Cl (HO)2PCH2XCH2XCHCH2P(O)(OH)2 O OH NH2 N + N NH3 O CH3 H2O phenols.OCO C 7CO2 n CH3 2 A V Antonov, I S Reshetnikov, N A Khalturinskij IS content in the polymer (%) 30 30 30 30 30 30 30 35 30 30 30 30 30 30 30 30 accompanied by the Fries rearrangement (migration of acyl groups of phenol esters to the o-position of the benzene ring), eventually resulting in the formation of cross-linked structures of the type 3: D PC Me CMe cross-linking.Ref. OI 37 32.5 45 24 45 31 45 30 45 21 45 28 46 28 ± 30 47 27 41 25 ± 26 45 25 45 23 45 22 45 23 45 30 45 22 45 25 O Me C CMe HO OH 3Combustion of char-forming polymeric systems Thermal decomposition of polycarbonates occurring along this route leads to the formation of a solid foamed coke layer on the polymer surface, which prevents further spreading of the flame. It was found 49 that foamed coke is made up of the polymer degradation products, i.e., PC takes an active part in its forma- tion, whereas the additive catalyses this process but is not directly involved in it.It is known that the main drawback of PC is a drastic decrease in their impact strength at low temperatures.50 Polymeric materi- als based on PC and an acrylonitrile ± butadiene ± styrene copoly- mer (ABS) and polyethylene terephthalate (PETP) are free from such defect. Some aromatic sulfonates as well as phosphorus- and bromine-containing compounds were tested as fire retardants for PETP ± PC51 and ABS± PC mixtures.49, 52 It was found that aromatic sulfonates become less efficient after addition of ABS to PC.49 The decrease in the fire retarding efficacy with the change in the composition of the polymeric matrix is due to the lowering (by 50 8C) of the temperature at which decomposition of PC containing ABS begins.As a result, the fire retardant does not affect the composition of PC pyrolysis products and the antipyr- ogenous effect of aromatic sulfonates on the PC ± ABS mixture is not manifested. Thus, IS prepared from aromatic sulfonates exert rather specific selective action and even minor alterations in the compo- sition of the polymeric mixture may significantly decrease their efficiency. Phosphorus-containing fire retardants can be recommended for the preparation of polymeric high-resistant materials based on a PC± ABS mixture. Although these compounds are less efficient than the bromine-containing ones, the materials based on them possess better mechanical properties.39 The bromine- and phos- phorus-containing fire retardant is especially efficient in PC ± PETP and PC ± ABS mixtures.51, 52 The combustion of systems prepared from polyurethane and ammonium polyphosphate 53 or polyamide and ammonium poly- phosphate 54 yields foamed cokes, predominantly due to the reaction of the polymer with the additive resulting in the formation of cross-linked products.Recently it was shown 55 that certain melamine salts are efficient as fire retardants for polycapramide, because these compounds also favour the forma- tion of cross-linked products in the reaction of the additive with the polymer. The reactions resulting in the formation of foamed cokes have been considered. Phosphoric acid is known to be an efficient fire retardant for cellulose.56 Phosphoric acid reacts with the polymer, which alters the mechanism of thermal degradation of cellulose towards char formation; therefore, this system can be related to IS.57 By virtue of their stability, adhesiveness, dielectric properties, thermal stability and technological efficiency, epoxy resins (ER) and phenol ± formaldehyde resins (PFR) are widely employed as binding agents in preparation of reinforced plastic materials in many branches of technics. Fire retardants affecting the mecha- nism of their thermal degradation are the most efficient for these polymers.Thus the efficiency of some transition metal salts (e.g., cobalt acetate) is due to their catalytic effect on dehydration and dehydrogenation of ER and PFR.58, 59 This significantly increases the char yield, while the yield of volatile flammable products formed in thermal degradation of ER decreases in the presence of organophosphorus compounds 19 and certain phosphines.60 The effects of the composition of ER on the mechanism of thermal degradation and the composition of the pyrolysis prod- ucts responsible for the flammability of ER have been discussed.61 It was shown that enhanced char formation is the determining factor in this process.Chemical reactions occurring in the presence of phosphoric acid used as IS for carbamide ± formaldehyde resin (CFR) have been discussed in detail.62 It was shown that in the course of thermal conversions phosphoric acid reacts with the amide groups of CFR to give N-phosphorylated products, which lose water, 609 ammonia and carbon dioxide and eventually turn into foamed cokes O +P(O)(OH)3 7H2O O CH2NHCNHCH2OCH2 n O HO CH2NCNHCH2OCH2 7H2O,7CO2 P(O)(OH)2 n O foamed cokes.HNPNH 7H2O,7NH3 OH n A viscous melt of poly(amidophosphates) favours the forma- tion of a swollen carbonised layer and suppresses thermal oxidation of foamed cokes, which enhances the fire-proof effect. Phosphorus-containing compounds, like phosphoric acid, react with functional groups of the polymer to yield foamed cokes which protect the polymer surface from the heat flow. It is the formation of foamed cokes that is responsible for the significant lowering of the flammability of the polymers to which a fire retardant has been added.IV. Chemical reactions occurring in the intumescence Chemical reactions occurring in the intumescence are most extensively studied for IS consisting of ammonium polyphos- phate and pentaerythritol. The sequence of chemical and physical processes has been established: decomposition of ammonium polyphosphate at 215 8C, esterification of the alcoholic groups of pentaerythritol with liberation of H2O, which results in the increase in the bulk of IS, solidification of the swollen carbon ± phosphorus residue at 360 8C.30 Studies using thermogravimetric analysis (TGA), thermovo- lumetric analysis and 31P NMR spectroscopy have shown 63 ± 69 that pentaerythritol is phosphorylated with APP at 210 8C with the scission of the main polyphosphate chain of APP and the formation of the 7[P(O)OCH2]7 phosphate groups.This reac- tion does not yield volatile products. O +C(CH2OH)4 O P H4NO n O O P O OH+(HOCH2)3CCH2 O P H4NO n H4NO n This is followed by the formation of cyclic esters of polyphos- phoric acid with elimination of ammonia and water. CH2OH O O (HOCH2)2C O P CH2O P 7H2O,7NH3 H4NO H4NO n O HOCH2 O P O P(O) HOCH2 O H4NO n Repetition of these reactions gives the structures of type 4.610 O O O O (O)P P(O) (O)P P(O) O O O O O O 4 The fact that the APP ± pentaerythritol pyrolysis products are paramagnetic has led the authors to suppose that their formation occurs via an intermediate carbene 5.69 O P OO OH (HOCH2)4C 7H2O (HOCH2)3CCH 5 O O (HOCH2)3CCH2OP O An alternative hypothesis suggests 63 that esterification of APP is preceded by the formation of a carbenium ion.The formation of cyclic structures as intermediates in thermal con- versions of APP ± pentaerythritol mixtures has made it possible to use bisphosphate of pentaerythritol as a model compound for investigation of a intumescence mechanism.63, 64 It was shown that condensation with elimination of water and formation of struc- tures of type 4 is the first step in the degradation of this bisphosphate. O O O O P P n 7(n71) H2O OH O O HO O O O O P P O O H O HO n An analysis of 31P and 13C NMR spectra made it possible to propose the following mechanism for char formation.Its first step is the formation of the carbenium ion 6 owing to protonation and cleavage of the C7O bond. This ion can be rearranged into a stable tertiary carbocation 7; deprotonation of the latter leads to the formation of the double bond. Subsequent pyrolysis of the ester 8 can occur with the elimination of phosphoric acid resulting in diene derivatives 9 which enter into the Diels ± Alder reaction to give polycyclic compounds of the type 10. O OH O O O O O H+ P P O+PP HO OH HO O O OH O O OH O O OP P + O OH HO CH2 O 6 O O + P (CH2)2OP(OH)2 HO O O 7 O O P (CH2)2OP(OH)2 7H3PO4 HO O 8 O O O 9 P CH CH2 HO O 9 O O OH P O O O P HO O A V Antonov, I S Reshetnikov, N A Khalturinskij Repetition of these reactions yields conjugated aromatic structures.70 The reactions can be accompanied by condensation of phosphoric acids and polymerisation; therefore, the reaction products possess irregular structures.Intumescence occurs within a narrow temperature range (300 ± 350 8C) and results in the formation of a bulky carbonaceous residue containing admix- tures of phosphoric acid. The lack of a sufficient body of evidence does not allow any valid conclusion on the possibility of the interaction of polymers with IS. It was found 63 ± 68 that polypropylene (PP) has virtually no effect on the mechanisms of intumescence. In turn, IS do not influence the composition of products formed upon polymer decomposition. Experiments with TGA in air have shown 38, 71 that the contribution of the polymer to char formation is insignificant.Presumably phosphoric acid reacts with alcohols formed in the oxidation of PP with atmospheric oxygen. Enhanced paramagnetic properties of the residue obtained upon pyrolysis of the polymeric composition in comparison with pure IS are attributed to the reaction of free radicals formed upon thermal degradation of the polymer with IS (APP + penta- erythritol).69 (HOCH2)3CCHR (HOCH2)3CCH+R 5 This assumption was supported by the finding that the mass of the char residue diminishes in the presence of free radicals. An analysis of the chemical composition of the chars formed upon thermal degradation of PP, polyethylene (PE) and polystyr- ene (PS) with IS (APP+pentaerythritol) has shown 39 that the chars formed at any temperature and with any polymer or without it are made up of identical chemical structures, which represent a combination of fragments containing quaternary carbon atoms with irregular aromatic and phosphorus ± oxygen units. However, there is conclusive evidence 72 that the chemical structure of a comonomer and its content in the polymer strongly affect the interaction of IS based on APP and pentaerythritol with ethylene copolymers.It was found that butyl acrylate ± ethylene and maleic anhydride ± ethylene copolymers markedly increase the efficiency of IS due to their interaction in the condensed phase. It was also shown that aluminosilicate zeolites also significantly increase the efficiency of IS.72 Reactions of ammonium polyphosphate with polyols (pen- taerythritol, starch, polyvinyl alcohol, glycerol) have been studied in the search for the optimum ratio of reaction components to be used for the filling of polymers in order to decrease their flammability.45 Irrespective of the polyol/APP ratio, all the systems examined displayed exothermic peaks of various intensities in the DPA curves.Presumably, this was due to the reaction of the polyol with APP with elimination of ammonia and water in this temperature range. The magnitude of the thermal effect was estimated on the assumption that the maximum thermal effect corresponds to the most complete reaction. It was established 45 that all the systems under study were characterised by identical mass loss and had identical masses of char residues, but differed in thermal effects.The optimum APP/polyol ratio for all the systems tested was 7 : 3 (mass %) as can be judged from the identical mass loss in the reaction. However, no strict correlation between the char yield, mass loss upon esterification and maximum thermal effect was found for systems containing methylphosphonic acid and pen- taerythritol.45 V. Heat-insulating characteristics of foamed cokes formed upon combustion and thermal degradation of polymeric systems containing intumescent fire retardants The efficiency of IS is determined primarily by heat-insulating characteristics of foamed cokes formed in the combustion ofCombustion of char-forming polymeric systems OI (%) 55 1 45 35 2 25 15 10 5 0 PER content (w/w) Figure 1.The OI values for specimens burnt under ordinary conditions (1) and for specimens from which `char caps' were removed during combustion (2).19 polymers. The role of `char caps' in the lowering of flammability has been established 19 by comparing OI of specimens burnt by routine methods with those where the foamed coke layer formed in the course of combustion was continuously removed (Fig. 1). The considerable difference in the characteristic curves and much greater values of OI in the former case (Fig. 1, curve 1) suggest that the presence of foamed cokes is the main factor decreasing flammability. Measurements of hydrodynamic resistances have shown 73 that foamed cokes cannot prevent the diffusion of degradation products to the combustion zone.It was con- cluded 73 that heat-insulating properties of `char caps' play the crucial role in flammability reduction. These results are in good agreement with the finding 74 that the foamed coke layer is permeable by low-molecular-weight liquids. However, the results of microstructural analysis of foamed cokes formed in the combustion of IS-containing PP compositions by scanning electron microscopy led the authors to conclusion 62 that diffu- sion of degradation products through the foamed coke layer plays an important role in combustion. The compositions with low OI displayed multiple fissures and breaks in the carbon structure, whereas foamed cokes formed from the compositions with high OI had closed pores.Evidently, heat-insulating characteristics (HIC) of foamed cokes are responsible for the high efficiency of IS in PP. Model studies of forced combustion, 36, 75 where a polymer sample with a thermocouple inserted at a definite distance from its surface was flame-burnt, have demonstrated that the foamed coke layer formed significantly lowers the polymer temperature. Under these conditions, the rate of the polymer temperature changes, i.e., HIC, of the foamed coke correlated with the OI. No such correlation was found for PP charged with various fillers (10% ± 30%) in addition to IS.30 The HIC of foamed cokes formed from different compositions and polymeric matrices but containing the same amounts of IS have been investigated.76 A great variety of compositions allowed the authors to establish a relationship between the flammability of polymer material and thermal characteristics of foamed cokes in a broad range of OI values.Thus it was found that a satisfactory correlation between OI and HIC was observed only for relatively low (<25) OI values. In addition, it was shown that the flammability of IS-containing polymeric systems is determined not only by the HIC of `char caps' but also by the rate of foamed coke layer formation. Most authors believe that HIC of foamed cokes play the major role in flammability reduction, although no satisfactory correla- tion between these parameters has been found so far.Various factors determining the HIC of foamed cokes formed upon combustion have been investigated.62, 77 A unit used in these studies allows the registration of the light flux and the response in the sample's rear (Fig. 2a). The sample to be tested, which represents a disc (*1 mm thick) 10 mm in diameter (Fig. 2b) is mounted on a thin metal stage. The sample surface is irradiated with the light generated by a carbon dioxide laser. Part of the radiant flux is diverted to an ILD-24M irradiation meter by means of a semitransmitting mirror made of a chalcogenide glass. A a 2 1 3 7 b 8 3 Figure 2. A schematic representation of a unit (a) and a stage (b) for studying heat-insulating characteristics of various materials; (1) carbon dioxide laser; (2) sample; (3) a thermocouple; (4) chalcogenide glass; (5) radiation meter; (6) interface; (7) computer; (8) the radiant flux; (9) asbestos support; (10) base. chromel-copel thermocouple (0.08 mm thick) is placed on the sample's rear.Having passed through the mirror, the radiant flux reaches the sample and warms it up. The signal of the thermo- couple response is passed to a computer. The experiments were carried out with both continuous and pulsating radiant fluxes (in the latter case, the ratio of the pulse length to the pause was equal to two) and ignition of the sample did not occur. Figure 3 shows the temperature vs time curves for the combustion of PS, PP and PE containing IS prepared from APP and pentaerythritol.39 Complete burn-out of polymers containing no fire retardants normally takes 15 s at temperatures 4800 K and does not depend on their chemical nature.Comparison of experimental curves points to significant differences in the proper- ties of foamed cokes at the moment of their formation. The shape of these curves is of complex type; each portion of the curve corresponds to a definite state of the system. A drastic increase in the temperature in the initial moment corresponds to the heating of the composition prior to foamed coke layer formation. Subsequent decline (curves 2 and 3) or a bend (curve 1) corre- spond to the formation and growth of the foamed coke layer on the polymer surface. The plateau corresponds to the intumescence of the sample in bulk and is associated with heat absorption upon intense degradation.A slow increase in temperature up to the steady-state level in the terminal steps indicates that the intumes- cence process is completed and the whole mass of the sample has been converted into foamed coke. The shape of the curve for the composition prepared from PE (curve 1) suggests that the 1 T /K 800 700 600 500 400 300 30 0 60 Figure 3. Heat-insulating characteristics of foamed cokes prepared from IS-containing polymeric compositions based on APP and pentaerythritol; (1) PE; (2) PP; (3) PS. 611 4 5 6 2 9 10 23t /s612 formation of foamed coke occurs at an extremely slow rate and it has low HIC. A small bend on the curve for the compositions prepared from PS (curve 3) points to a relatively high rate of foamed coke formation, which is followed by a decrease in the rate of warm-up for a short period.The shapes of the curves for the compositions prepared from PP and PS (curves 2 and 3) indicate that the formation of, and accumulation of, the foamed coke occur much faster than in the case of the compositions based on PE. These foamed cokes have enhanced HIC, but their shapes and the positions of their extrema point to differences in the kinetics of their intumescence. The sharp maximum on the curve for the composition based on PP (curve 2) suggests that the activation energy of the intumescence reaction is low, while the rate of intumescence and the HIC of foamed cokes formed even at insignificant heat fluxes are high.The maximum on the curve for the composition derived from PS (curve 3) points to the higher rate of intumescence and formation of foamed coke at higher temper- ature in comparison with the composition prepared from PP. In the final step, the rates at which the temperature changes occur and their limiting values for the compositions prepared from PP and PS coincide (curves 2 and 3), which testifies to identical properties of the `char caps' formed. The differences in the HIC of `char caps' at the moment of their formation and the similarity of their HIC in the case of IS-containing PS and PP compositions can be attributed to the differences in their internal structure, which is schematically presented in Fig.4. In both cases, `char caps' represent thick walls with large cavities inside. The walls of the `caps' have porous structures, which differ in pore size and in wall thickness. The foamed coke formed from the composition based on PP has smaller pores and thinner walls; the air fraction (v/v) is greater (Fig. 4a). a b Figure 4. A schematic representation of foamed cokes formed under the action of a heat flux; (a) a composition prepared from PP; (b) a composition prepared from PS.39 Investigations into the effects of phosphorus-containing fire retardants based on methylphosphonic acid on the flammability of PE, PP and PS have demonstrated 47 that in the case of PE and PP the fire retardants prepared from the adduct of melamine ± formaldehyde resin and methylphosphonic acid (MFMA), from the adduct of methylphosphonic acid and melamine (DPM-3) combined with pentaerythritol and from methylphosphonic diamide (MPD) are more efficient than halogen-containing fire retardants currently used.The temperature vs time curves for the combustion of PP and PE containing MFMA as the fire retardant are shown in Fig. 5. Comparison of these curves suggests thatMFMAstrongly affects the properties of foamed cokes at the moment of their formation. In the case of a PP composition, foamed cokes are formed at a faster rate and have enhanced HIC. Therefore, MFMA can be regarded as a typical IS, although its efficiency and the HIC of the foamed cokes formed are inferior to those of compositions based on APP.Figure 6 shows the temperature vs time curves for the combustion of PP, PS and PE containing a 4 : 1 DPM-3 ± pentaerythritol mixture.47 Comparison of these curves points to a strong effect of the polymeric matrix on the properties of foamed cokes at the moment of their formation. In the case of PS compositions, the foamed cokes are formed at slow rates and have low HIC. The foamed cokes formed at a fast rate in the A V Antonov, I S Reshetnikov, N A Khalturinskij T /K 600 21 500 400 300 60 0 30 t /s Figure 5. Heat-insulating characteristics of fire retardant-containing compositions prepared from the adduct of melamine ± formaldehyde resin with MFMA; (1) PE+MFMA; (2) PP+MFMA.47 T /K 3 400 350 12 300 90 t /s 0 30 60 Figure 6.Heat-insulating characteristics of fire retardant-containing polymeric compositions prepared from the adduct of methylphosphonic acid with DPM-3;47 (1) PS+DPM-3; (2) PE+DPM-3; (3) PP+ DPM-3. combustion of PP and PE compositions have high HIC. The HIC of foamed cokes show a good correlation with the efficiency of the fire resistant agent. For PP, the efficiency of DPM-3 is close to that of APP, whereas for PE, DPM-3 is more efficient than APP. The temperature vs time curves for the combustion of PP, PS and PE containingMPDare shown in Fig. 7.78 As can be seen, the presence of a polymer significantly lowers the rate of foamed cokes formation. It is of note that the chars formed in the decomposition of compounds prepared from PE and PS have very low HIC, whereas the chars formed in the combustion of compositions derived from PP have high HIC. T /K 600 1 500 3 400 2 300 t /s 0 20 40 Figure 7.Heat-insulating characteristics of fire retardant-containing compositions prepared from MPD; (1) PS+MPD; (2) PE+MPD; (3) PP+MPD. The plateau (Fig. 7) corresponding to the intumescence of the bulk of the sample is very low. It may thus be assumed that the amount of the char formed is insufficient for long-term protection of the polymer. Thus, the mechanism of action of MPD as the fire resistant agent is based primarily on the reversal of the destructive processes occurring in PE, which results in smaller quantities of volatile flammable products.Although MPD favours char for- mation in polymers, it is not the main factor responsible for the flammability reduction, since the char formed cannot protect the polymer in the course of combustion.Combustion of char-forming polymeric systems MPD has a limited use because fire retarding polymeric compositions containing this fire retardant are difficult to proc- ess: phase separation occurs in moulding and casting. One of the clues to the solution of this problem is microencapsulation of the fire retardant prior to its introduction into the system. Besides, the choice of an appropriate shell may not only facilitate compati- bility of the fire retardant with the polymeric matrix, but also increase its efficiency.The effects of various types of shells on HIC and flammability of polymeric materials prepared from PP and PE have been studied.77 The following polymeric shells were employed for microencapsulation: high-density PE, polyvinyltriethoxysilane and an aromatic polyamide based on terephthalic acid and m- phenylenediamine. It was shown that the chemical nature of the polymeric shell strongly affects the flammability and HIC of polymeric compositions based on PE and PP. The temperature vs time curves for the combustion of PP containing non-encapsulated MPD and MPD encapsulated in PE and polyvinyltriethoxysilane are shown in Fig. 8. As can be seen, the shell strongly affects the physical structure of the char at the moment of its formation.As was noted above, the formation of char in the combustion of PP containing MPD occurs at a very slow rate (Fig. 8, curve 2). The polyethylene film does not influence the combustible and protective properties of the chars formed upon combustion (Fig. 8, curve 3). Char formation accompanying the decomposition of MPD-containing composi- tions enclosed in polyvinyltriethoxysilane occurs much faster and the chars formed possess enhanced HIC (Fig. 8, curve 1). T /K 1 700 600 2 3 500 400 4 3000 20 40 t /s Figure 8. Heat-insulating characteristics of fire retardant-containing compositions prepared from MPD encapsulated in polymeric films:77 (1) PP; (2) PP+MPD; (3) PP+MPD encapsulated in PE; (4) PP+MPD encapsulated in polyvinyltriethoxysilane.Thus, the efficiency of IS (OI of the polymeric compositions) correlates well with the HIC of foamed cokes formed upon combustion. The higher the HIC of the chars formed the greater the efficiency of IS in these compositions. VI. Conclusion The major effect of fire retardants used for decreasing the flammability of polymers consists predominantly in increasing the char yields in the combustion and thermooxidative degrada- tion of polymeric materials. The additives favouring char for- mation upon the interaction of polymeric components on heating are successfully employed for the preparation of fire retarding polymeric materials. The studies 79 ± 84 published in 1996 ± 1998 have demonstrated that metals and metal-containing materials are best suited for this purpose.This trend of research holds especially great promise, for some metal-containing additives are able to catalyse chemical cross-linking reactions in polymers containing functional groups.85, 86 Their addition not only affects the yields of chars and the kinetics of their formation but also changes their physical structure. 613 References 1. H J Fabris, J G Sommer Rubb. Chem. Technol. 50 523 (1977) 2. K A Andrianov Polimery s Neorganicheskimi Glavnymi Tsepyami Molekul (Polymers with Inorganic Molecular Backbones) (Moscow: Nauka, 1962) 3. F Stone,W Grahem (Ed.) Inorganic Polymers (New York: Wiley, 1963) 4. G S Gol'din, S G Fedorov, S F Kravtsova Antipireny i Ograni- chenno Goryuchie Materialy na Osnove Oligo- i Polifosfazenov (Fire Retardants and Fire Retarding Materials Based on Oligo- and Polyphosphazenes) (Moscow: NIITEKhIM, 1988) 5.K-U BuÈ ller Spezialplaste (Berlin: Akademie-Verlag, 1978) 6. A L Rusanov, B B Serkov, E G Bulycheva, T N Kolosova, T V Lekae, I I Ponomarev, N G Matvelashvili Makromol. Chem., Makromol. Symp. 74 189 (1993) 7. D W van Krevelen Polymer 16 615 (1975) 8. A V Antonov, S V Lavrov, A A Kuznetsov, R M Gitina, B V Kotov Vysokomol. Soedin., Ser. A 36 20 (1994) a 9. I C McNeill,M Coskun Polym. Degrad. Stab. 23 175 (1989) 10. S L Malhotra, P Lessard, L P Blanchard J. Macromol. Sci., Chem. A15 1577 (1981) 11. L A Mango III J. Polym. Sci., Polym. Chem. Ed. 15 513 (1977) 12. L Costa, C Camino, A Chiotis, G Clouet, J A Brossas, M Bert, A Guyot Polym.Degrad. Stab. 6 177 (1984) 13. B A Zhubanov, S A Nazarova, Z G Karzhaubaeva, K M Gibov Vysokomol. Soedin., Ser. B 18 150 (1976) a 14. Z Janovic, K Saric, O Vogl J. Macromol. Sci., Chem. 22 85 (1988) 15. C P R Nair, G Clouet, Y Guilbert Polym. Degrad. Stab. 26 305 (1989) 16. A V Antonov, R M Gitina, E D Rogozhkina, A L Izyumnikov, I L Zhuravleva, Yu S Bogachev, S N Novikov Vysokomol. Soedin., Ser. A 32 860 (1990) a 17. R M Gitina, L A Oksent'evich, A A Kuznetsov, L I Danilina, A L Izyumnikov, E D Rogozhkina, V V Kopylov, S N Novikov, A N Pravednikov Vysokomol. Soedin., Ser. A 26 1060 (1984) a 18. MYu Belov, Candidate Thesis in Chemical Sciences, L Ya Karpov Research Physical Chemistry Institute, Moscow, 1994 19.A V Kochubei, Candidate Thesis in Chemical Sciences, Institute of Synthetic Polymer Materials, Russian Academy of Sciences, Moscow, 1990 20. E M Nechvoda, A G Gal'chenko, S Z Rogovina, E V Prut, I M Bel'govskii, N A Khalturinskij Khim. Fiz. 6 696 (1987) b 21. A Survey Chemosphere 19 2023 (1989) 22. J Troitzsch Makromol. Chem., Makromol. Symp. 74 125 (1993) 23. R Luijk, J Pureveen, J Commandeur, J Boon Makromol. Chem., Makromol. Symp. 74 235 (1993) 24. A V Antonov, R M Gitina, S N Novikov Vysokomol. Soedin., Ser. A 32 1895 (1990) 25. G Bertelli, P Busi, G Camino, L Costa, R Locatelli Polym. Degrad. Stab. 18 307 (1987) 26. K M Gibov, Doctoral Thesis in Chemical Sciences, Kazan Chemical Technology University, Kazan, 1987 27.H Vandersall J. Fire Flammability 2 (4) 97 (1971) 28. M Kay, A F Price, I Lavery J. Fire Retard. Chem. 6 69 (1979) 29. G Camino, L Costa, G Martinasso Polym. Degrad. Stab. 23 359 (1989) 30. G Bertelli, E Marchetti, L Costa, E Casorati, R Locatelli Polym. Degrad. Stab. 25 277 (1989) 31. BRD P. 2 723 877; Chem. Abstr. 88 62 971 (1978) 32. BRD P. 2 839 710; Chem. Abstr. 90 188 015 (1979) 33. G Camino, L Costa,M Luda Makromol. Chem., Makromol. Symp. 74 71 (1993) 34. A Granzow, C Savides J. Appl. Polym. Sci. 25 2195 (1980) 35. A Ballistreri, G Moutaudo, C I C Puglisi, E Scamporrino, D Vitalini J. Appl. Polym. Sci. 27 3369 (1982) 36. G Moutaudo, E Scamporrino, G Puglisi, D Vitalini J. Appl. Polym. Sci. 30 1449 (1985) 37.E V Gnedin, Candidate Thesis in Chemical Sciences, L Ya Karpov Research Physical Chemistry Institute, Moscow, 1991 38. O Ciccetti, A Pagliari, G Camino, in Proceedings of the 3rd Meeting on Fire Retardant Polymers (Abstracts of Reports), Torino, 1989 p. 178614 39. A Ballistreri, G Moutaudo, G Puglisi, E Scamporrino, D Vitalini J. Appl. Polym. Sci. 28 1743 (1983) 40. D G Brady, C U Moberly, J R Norell, H C Walters J. Fire Retard. Chem. 4 150 (1977) 41. Y Halpern, D M Mott, R H Niswander Ind. Eng. Chem. Prod. Res. Dev. 23 233 (1984) 42. US P. 4 253 972; Chem. Abstr. 94 176 192 (1981) 43. F Severini, R Gallo, G Audisio, M Pegoraro Polym. Degrad. Stab. 17 71 (1987) 44. A De Chirico, G Audisio, F Provasoli, M Armanini, R Franzese Makromol.Chem., Makromol. Symp. 74 343 (1993) 45. I Reshetnikov, A Antonov, T Rudakova, G Aleksjuk, N Khalturinskij Polym. Degrad. Stab. 54 137 (1996) 46. US P. 4 257 931; Chem. Abstr. 94 209 805 (1981) 47. E D Weil, W Zhu, N Patel, S M Mukhopadhyay Polym. Degrad. Stab. 54 125 (1996) 48. A Ballistreri, G Moutaudo, E Scamporrino, C Puglisi, D Vitalini, S Cucinella J. Polym. Sci., Part A, Polym. Chem. 26 2113 (1988) 49. A V Antonov, S N Novikov Vysokomol. Soedin., Ser. A 35 1442 (1993) a 50. P-R MuÈ ller Plastverbeiter 41 32 (1990) 51. J Green J. Fire Sci. 12 257 (1994) 52. J Green Polym. Degrad. Stab. 54 189 (1996) 53. S F Egorov, V N Kuz'min, I A Grishin Plast. Massy (3) 42 (1986) 54. S V Levchik, G Camino, L Costa, G F Levchik Fire Mater.19 1 (1995) 55. S V Levchik, G F Levchik, A I Balabanovich, G Camino, L Costa Polym. Degrad. Stab. 54 217 (1996) 56. C F Cullis,MMHirshler (Eds) The Combustion of Organic Polymers (Oxford: Clarendon Press, 1980) 57. B Kandola, A R Horrocks Polym. Degrad. Stab. 54 289 (1996) 58. T V Popova, Candidate Thesis in Chemical Sciences, L Ya Karpov Research Physical Chemistry Institute, Moscow, 1988 59. B Ya Kolesnikov, Doctoral Thesis in Chemical Sciences, S M Kirov Kazakh State University, Alma-Ata, 1987 60. S V Levchik, G Camino, L Costa,M P Luda Polym. Degrad. Stab. 54 317 (1996) 61. N Rose, M Le Bras, S Bourbigot, R Delobel, B Costes Polym. Degrad. Stab. 54 355 (1996) 62. B T Sarsembinova, Candidate Thesis in Chemical Sciences, S M Kirov Kazakh State University, Alma-Ata, 1990 63.G Camino, L Costa, L Trossarelli Polym. Degrad. Stab. 7 394 (1984) 64. G Camino, L Costa, L Trossarelli, in The 6th Conference `Italian Science of Macromolecules' (Abstracts of Reports), Atti, 1983 p. 341 65. G Camino, L Costa, L Trossarelli Polym. Degrad. Stab. 12 213 (1985) 66. G Camino, L Costa, L Trossarelli, F Costanzi, G Landoni Polym. Degrad. Stab. 8 13 (1984) 67. G Camino, L Costa, G Martinasso Polym. Degrad. Stab. 27 285 (1990) 68. G Camino, L Costa,G Martinasso,R Cobetto Polym. Degrad. Stab. 28 17 (1990) 69. J Rychly, L Matisova-Rychla,M Vavrekova J. Fire Retard. Chem. 8 82 (1981) 70. MLe Bras, S Bourbigot, C Siat, R Delobel, in Fire Retardancy of Polymers. The Use of Intumescence. (EdsMLe Bras, G Camino, S Bourbigot, R Delobel) (Cambridge: Royal Society of Chemistry, 1998) p. 266 71. R Delobel, N Ouassou,M Le Bras, J-M Leroy Polym. Degrad. Stab. 23 349 (1989) 72. S Bourbigot,M Le Bras, R Delobel, P Breant, J-M Tremillon Polym. Degrad. Stab. 54 275 (1996) 73. A V Kochubei, N A Khalturinskij, Al Al Berlin, N I Rakhmangulova Vysokomol. Soedin., Ser. B 31 659 (1989) a 74. K M Gibov, B A Zhubanov, L N Shapovalova Vysokomol. Soedin., Ser. B 26 108 (1984) a 75. G Moutaudo, G Puglisi, in Proceedings of the 3rd Meeting on Fire Retardant Polymers (Abstracts of Reports), Torino, 1989 p. 36 76. E Gnedin, S Novikov, N Khalturinskij Makromol. Chem., Makromol. Symp. 74 329 (1993) 77. A Antonov, I Reshetnikov, E Potapova, T Rudakova, N Zubkova,M Tuganova, N Khalturinskij, in The 12th Symposium of the Israeli Section of the Combustion Institute (Abstracts of Reports), Haifa, 1996 p. 6 A V Antonov, I S Reshetnikov, N A Khalturinskij 78. N S Zubkova,M A Tyuganova, N G Butylkina, N A Khalturinskij, I S Reshetnikov, E V Potapova, M S Vilesova, L I Voronkova, M S Bosenko Plast. Massy (5) 35 (1996) 79. M Le Bras, G Camino, S Bourbigot, R Delobel, in Fire Retardancy of Polymers. The Use of Intumescence (EdsM Le Bras, G Camino, S Bourbigot, R Delobel) (Cambridge: Royal Society of Chemistry, 1998) p. 343 80. S Bourbigot,M Le Bras, in Fire Retardancy of Polymers. The Use of Intumescence (EdsMLe Bras, G Camino, S Bourbigot, R Delobel) (Cambridge: Royal Society of Chemistry, 1998) p. 222 81. G E Zaikov, S M Lomakin Polym. Degrad. Stab. 54 223 (1996) 82. J Wang, in Fire Retardancy of Polymers. The Use of Intumescence (EdsMLe Bras, G Camino, S Bourbigot, R Delobel) (Cambridge: Royal Society of Chemistry, 1998) p. 159 83. E D Weil, W Zhu, H-K Kim, N Patel, L Rossi di Montelera, in Fire Retardancy of Polymers. The Use of Intumescence (EdsMLe Bras, G Camino, S Bourbigot, R Delobel) (Cambridge: Royal Society of Chemistry, 1998) p. 35 84. M Lewin, in Fire Retardancy of Polymers. The Use of Intumescence (EdsMLe Bras, G Camino, S Bourbigot, R Delobel) (Cambridge: Royal Society of Chemistry, 1998) p. 3 85. C A Wilkie, in Recent Advances in Fire Retardancy of Polymer Materials Vol. 9 (Ed.MLewin) (Norfolk: BCC, 1998) 86. J Li, C Wilkie Polym. Degrad. Stab. 57 293 (1997) a�Polym. Sci. (Engl. Transl.) b�Russ. J. Chem. Phys. (Engl. Tran
ISSN:0036-021X
出版商:RSC
年代:1999
数据来源: RSC
|
|