246 GENERAL DISCUSSION GENERAL DISCUSSION prof. H. S. Gutowsky (University of Illinois) said : The Chairman has raised the question of the relative merits of electromagnets versus permanent magnets for magnetic resonance experiments. This is a difficult question to answer because the range in quality and characteristics of magnets of both types exceeds any intrinsic differences. But there are some intrinsic differences which, all other features being equal, recommend electromagnets for some applications and per- manent magnets for others. The characteristics of a magnet include: (i) maximum field obtainable, (ii) adjustable range in field, (iii) stability, (iv) field homogeneity, (v) gap dimensions, (Vi) cost of magnet and accessories, (vii) reliability, ease, and cost of operation, (Viii) commercial availability.The nature of available permanent magnet materials limits the fields economically obtainable in gaps of the dimensions required to about 7,000 to 8,000 gauss. Electromagnets do not suffer as great a limitation. Moreover, an electromagnet can provide a continuously adjustable field from zero to its rated maximum value, while the field of a permanent magnet can be adjusted only over a small range, say up to + 50 gauss from the usual value, by using direct current coils at the gap or about the poles.' However, in terms of stability a permanent magnet has a definite advantage. Electromagnets can be electrcnically regulated to a truly remarkable extent, to fluctuations no greater than 1 part in 107 per min and this will no doubt be improved.But considerable elcctronics are required in the power supply, at a cost com- parable to the cost of the magnet itself. Permanent magnets have a reversible change in magnetization with temperature, -0.02 %/deg. for Alnico V. But simple thermoregulation combines with the large heat capacity of the permanent magnets to give easily field stability an order of magnitude or more better than that cited for electromagnets. As to field homogeneity, this is determined more by the pole cap material and its shape and mechanical alignment than by the source of the magnetic field. So there is little here in favour of either type of magnet. The field contours for an electromagnet do show hysteresis effects, being different on the demagnetization compared to the magnetizing cycle, and also are different at high fields where the pole cap material becomes saturated.On the other hand, if a permanent mamet is not kept at constant temperature, small changes occur in the field homogeneity, presumably becaure of changes in domain structures with temperature. The gap dimensions of a permanent magnet are somewhat more critical than for electro- magnets, if the permanent magnet material is to be utilized efficient1y.l If stability and homogeneity are important, then the cost of a permanent magnet, for fields up to 6000 gauss, will generally be less by a factor of 1/2 to 2/3 1 Gutowsky, Meyer and McClure, Rev. Sci. Instr., 1953, 24, 644.GENERAL DISCUSSION 247 than a comparable electromagnet. And this is mainly because of the current supply and control.A permanent magnet is undoubtedly simpler and more reliable in operation than even the best electromagnet. You do not have to turn a permanent magnet on or off. But some applications such as electronic para- magnetic resonance require that the field be adjustable over a wide range, from zero to as high as 6,000 or 10,000 gauss depending on whether you are working at 3 cm or 1 cm wavelengths. And in nuclear magnetic resonance, sensitivities are increased by working at the higher fields of large electromagnets. And it is sometimes desirable to have a wide enough range of fields available to check the field dependence of a phenomenon. In America electromagnets probably are at present more prcvalent than permanent magnets. This is partially because electromagnets of guaranteed stability and homogeneity are available commercially.Excellent permanent magnets can also be contracted for but gap design is still your own problem and the homogeneity is not specifiable. In summary, it is usually a qucstion of personal prcjudice or availablility which determines whether a permanent or electromagnet is used for nuclear magnetic resonance. Insofar as I know good versions of both types of magnets have been used successfully in practically every variety of nuclear magnctic reson- ance experiment. Dr. N . Sheppard (Cambridge University) (communicated) : Prof. Gutowsky discussed the high resolution n.m.r. spectrum of dimethyl formamide and deduced from thc nonequivalence of the two methyl groups that a barrier of at lcast 10 kcal per molecule restricted rotation about the C-N bond.Would he mind outlining how this calculation was madc? The more general question of whether or not n.m.r. methods can distinguish between the different stable configurations (rotational isomers) corresponding to minima in the potential energy curve associated with less highly restricted internal rotation is also of interest. For molecules such as 1 : 2-dichloroethane certain of the barriers may be as low as 3-4 kcal. Whereas the rotational isomers can be separately distinguished by infra-red and Raman spectroscopy, is it probable that thc rate of transition from one minimum to another will be too rapid for observation of other than average effects by the n.m.r. method? Prof. H. S . Gutowsky (University of Illinois) (communicated) : The high resolution proton magnetic resonance spectrum of liquid dimethyl formamide, (CH&NCHO, was observed in our laboratory 1 to consist of a close doublet and a third line, with relative intensities of 3 : 3 : 1.The doublet was at a &value of - 0.24 which is in the range of -- 0.20 to - 0.30 found2 for methyl groups attached to nitrogen. And the + 0.34 for the weaker line agrees in position with other compounds2 having structures of the sort N-CHO and 0-CHO. A doublet structure of the CH3 group resonance might in principle arise either from a chemical shift due to nonequivalence of the two CH3 groups or from an indirect spin-spin coupling 3 with the proton of the CHO group. However, the latter possibility is eliminated by the absence of any corresponding splitting of the resonance line for the proton in the CHO group.So it is concluded that the two CH3 groups are nonequivalent. The nonequivalence of the methyl groups is explained most readily by assuming a barrier to internal reorientations about the N - C bond, because of a double bonded contribution to the structure : CH3 H 0- 1 Meyer, unpublished results. 2 Meyer, Saika and Gutowsky, J. Amer. Chem. SOC., 1953,75,4567. 3 Gutowsky, McCall and Slichter, J. Chem. Physics, 1953, 21, 279.248 GENERAL DISCUSSION A lower bond can be estimated for the barrier to internal rotation because re- orientations would average out the electronic differences of the two CH3 groups and give a single line if the reorientations were fast enough.The general theory 1 for the averaging effects of dynamic processes shows that the doublet separation of about 5 c/s would not be observed if the reorientation frequency Y was itself faster than 5 CIS. The reorientations are undoubtedly thermally activated and the reorientation frequency is given by a rate equation of the form : Y = vo exp (- V~/RT). vo is a frequency factor the order of kT1h or 6.25 x 1012 at room temperature and the activation energy is the potential barrier VO, which is related directly to the contribution of the double bond to the N-C bond energy. If we rcquire that v be less than 5 c/s, then Vo must be greater than 16 kcal. We are obtaining a more definite value for VO by observing the temperature dependence of the doublet separation.At higher temperatures the reorientation frequency increases enough to coalesce the doublet; so we can determine the temperature at which v is actually 5 c/s and this will give us VO. This general method is limited in its applications by the separation of the chemically shifted and multiplet resonances which are averaged by the dynamic processes in question, These separations range from a few cycles to the order of several kc. If the dynamic processes have normal frequency factors, the cor- responding range of barrier heights or activation energies is between 10 and 20 kcal mole-1. Prof. H. S. Gutowsky (University of Illinois) said: The results of Ford and Richards on the broad proton resonance in solid diketene provide an interesting confirmation of the molecular structure in the solid phase.However, there is some infra-red evidence 2 for the presence of a second form in thc liquid phase with increasing concentrations at higher temperatures. The versatility of nuclear magnetic resonance is illustrated by the fact that we could apply high resolution techniques in our laboratory 3 to the question of the structure of diketene in the liquid phase. At various times the following five structures have been proposed for diketene : 0 - G C H I II I I H2C--C=O C H 2 a H 2 I I I I o r - 0 O&O The number and relative intensities of the chemically shifted components in the proton magnetic resonance absorption spectra for these structural models are given simply by the number of non-equivalent structural sites and the fraction of protons in each type of site.Thus, structure (I) would have two resonance components with relative intensities of 3 : 1 ; (Ir) 2: 1 : 1 ; (111) 4 ; (IV) 3 : 1 ; and (V) 2 : 2. At room temperature the proton resonance of the liquid was found to consist of two components, with equal intensities, and with chemical shift &values 4 1 Gutowsky and Saika, J. Chem. Physics, 1953, 21, 1688. 2 Miller and Koch, J. Amer. Chem. Soc., 1948,70, 1890. 3 Bader, Gutowsky, Williams and Yankwich, unpublished results. 4 Meyer, Saika and Gutowsky, J. Amer. Chem. SOC., 1953,75,4567.GENERAL DISCUSSION 249 referred to H20 of - 0.07 and - 0.16. This proves that the structure of the liquid is (V), the 3-buteno-/%lactone as found in the solid. Moreover, the position of the - 0.07 line is close to that found 1 for vinyl groups while the - 0.16 line is not far removed from CH2 groups in cyclic compounds, confirming structure 03.At elevated temperatures, the spectrum became more complex indicating the formation of one or more new species. However, most of these effects were irreversible. In any event, it appears that no more than 5 to 10 % of the sample at room temperature could have been other than structure 0. And this estimate is based on a conservative value for the sensitivity of our measurements. Prof. W. N. Lipcomb (University of Minnesota) said: Confirmation of the structure for diketene is indeed pleasing, and some comments on the values of the interatomic distances may be of interest. Cox, Cruickshank and Smith have found that if the effect of torsional oscillations of benzene molecules in the crystal is ignored, an apparent shortening of the bonded C-C distance by about 0.0158, occws.An estimate of this same effect in diketene indicates that OW published valucs of bonded distances in the four-membered ring are about 042A shorter than the correct values, and that the correct external C=C and C=O distances are probably about 0.015 8, shorter than we reported. If these correc- tions are made the bonded distances in our X-ray diffraction study agrec to within an average of -i. 0.02A with the electron diffraction results for dimensions of this model.2 Finally, a small similar effect occurs in our X-ray diffraction study of pentaborane. While these changes are probably significant they do lie within the limits of errors assigned by us.Dr. Peter Gray (Cambridge University) said: Dr. Drain reports a value 1.042 f 0.01 8, for the N-H distance in N&F (in NH4C1 the value 3 is 1,038 f 0.004 A) and suggests that the change in dimensions of the ammonium ion produced by the strong hydrogen bonding in ammonium fluoride is small. On the other hand the change in the N-H distance in hydrazinium fluoride in which there is also strong hydrogen bonding is appreciable ; 4 the N-H distance is 1.075 f0.02 A. In hydrazinium fluoride this is accompanied by an appreciable4 shift in the 3000 cm-1 N-H frequency. A similar effect is observed in ammonium fluoride where, as a result of hydrogen bonding, the N&+ vibration frequency normally in the neighbourhood of 1400cm-1 has the value 1484cm-1 in NH4F, markedly different from its 1397 cm-1 in NI&Cl. Such a shift would again be expected to be associated with an appreciable lengthening of the N-H bond.Is it possible that the value reported here for the N-H bond length in NH4F is somcwhat low ? Prof. D. P. Hornig (Math. Inst., Oxford University) said : It is surprising that the N-H distance obtained in NI&F should be so close to that in N&Cl since the N-H stretching frequencies differ considerably, occurring at about 2850 cm-1 in NH4F and 3100 cm-1 in N€&CI. In hydrazine fluoride, according to Deeley and Richards, the N-H distance is 1.075A and the mean N-H stretching frequency is 2670 cm-1. These considcrations suggest a value for rNH of 1-05-1.06 A.The considerable increase in thc bending frequency of the NH4+ ion from 1400 cm-1 in NH4Cl to 1494 cm-1 in N&F also indicates the presence of a moderately strong hydrogen bond in the latter. The infra-red spectrum yields a torsional frequency of 523 cm-1 (1.7 x 1013 sec-1) for the NH4+ ion in the lattice, moderately close to the value 1.4 X 1013 estimated by Dr. Drain. The change should not greatly affect his calculations. Dr. Mansel Davies (Aberystwyth) said: The urea molecule is of particular interest as it is possibly one of the simplest compounds showing the planar form 1 Meyer, Saika and Gutowsky, J. Amer. Chem. SOC., 1953,75,4567. 2 Bauer, Bregman and Wrightson, J. Amer. Chem. Soc., 1955, to be published. 3 Gutowsky, Pake and Bersohn, J. Chern. Physics, 1954,22, 643.4 Deeley and Richards, Trans. Faraday Soc., 1954, 50,560. 5 Bovey, J. Chem. Physics, 1950, 18, 684.250 GENERAL DISCUSSION of the amide nitrogen valencies found by X-ray methods to occur in a variety of peptide and protein-like structures. At Aberystwyth we have been interested in the bonding within the simplc amide group and the changes which occur in it. Thus formamide, the simplest amidc, is certainly not planar in the vapour state 1 -the hydrogens of the NH2 group are symmetrically disposed about the OCN plane-but it may possibly become so in the liquid or solid state. The planar structure of the urea molecule in the crystal presumably involves the sp2 hybridized valency state of the nitrogen atom : this means that what might have been regarded as the lonepair electrons of the nitrogen are in pz orbitals and delocalized n-bonding takes place between the 0, C and N atoms-a picture which accounts qualitatively for the high order of the C-N bonds when compared with singIe bonds, and the somewhat reduced bonding in the C 4 linkage.Mr. L. H. Hopkins, who has been studying urea in the infra-red, has considered the possible implications of this valency state of the nitrogen upon the stretching v(N-H) frequency and he has also made some experimental observations to which I should like to refer. Compared with ammonia r(N-H) == 1-014& v(N-H) = 3372 cm-1 if we accept the bond length of Andrew and Hyndman, the stretching frequency in urea is remarkably high r(N-H) = 1.046 A, v(N-H) = 3396 cm-1; the increased N-H bond length would have been expected to decrease the frequency by about 300 cm-1.However, by analogy with the case of carbon, the hybridiza- tion change p3 + sp2 for nitrogen might shorten r(N-H) by about 4 %. Taking this and the influence of ionic tcrms into account, one then finds that, in the absence of hydrogen bonding the v(N-H) frequency in urea could be in the neighbourhood of 3650cm-1. Compared with the quoted frequency (which is the arithmetic mean of thc symmetric and asymmetric modes), this would be indicative of appreci- able hydrogen bonding. But this suggestion is by no means satisfactory either in itself or in the light of at least two other observations. The X-ray data show that the relevant 0-N distances in the solid are 2-99A and 3.04 A : thesc correspond to weak hydrogen bridges and, moreover, suggest that the two N-H bonds of the NH2 groups in crystalline urea are not strictly equivalent.Again, as Mr. Hopkins has found, there are no very pronounced changes in the urea spectrum on going from the solid to the vapour-at least, not in the region of the carbonyl frequency. Thus it is very difficult, if not impossible, to reconcile the N-H bond length deduced from the proton magnetic resonance spectrum with other observations on the urea structure. Prof. H. S. Gutowsky (University of Illinois) said : In their determination of the structure of the urea molecule, by observing the proton magnetic resonance in a single crystal, Andrew and Hyndman assumed the lattice to be "rigid" at room temperature.This was supported by the similarity of results at liquid air and room temperature, so observations made at the latter temperature were used in the analysis. However, Kromhout and Moulton,2 at the University of Illinois, have observed a transition in the proton line width in urea crystal powder, centred at about 50" C. The width changes from 14 to 7 gauss, as would be expected for rotation of the NH2 groups about the C-N bond. The transition to the broad line may not be complete at 25" C. In fact, the second moment measured at 0" C is 20.8 & 0.6 gauss2, from which Kromhout and Moulton calculate an N-H bond distance of 1.010 f 0-007 A, if the planar rnolccule with an H-N-H bond angle of 120" is assumed. 1 Evans, J. Chem. Physics, 1954, 22, 1228. 2 Kromhout and Moulton, J.Chem. Physics, in press.GENERAL DISCUSSION 25 1 This is somewhat less than the 1.046 f 0.01 A found by Andrew and Hyndman, as would be the case if a slight amount of motional narrowing were indeed present at room temperature. Such narrowing would not affect in any way the con- clusions as to the planarity of the molecule. However, the N-H distance and the bond angles may require some modifications. I believe that resolution of the small discrepancies would be useful in establishing the dependence of the N-H bond distance on hydrogen bonding, which is an important problem. Prof. E. R. Andrew (University College of North Wales) (comniiinicnted) : Kromhout and Moulton (private communication) have recently reviscd their N-H bondlength of 1-010&0-007 A quoted by Prof.Gutowsky to 1*036:1:0*009 A, which, within the combined limits of error is not in disagreement with our value of 1.046 f 0.01 A. Nevertheless, it is possible that our second momcnt values may be slightly less than the rigid lattice values, though our measurements were made at a rather lower temperature (18-20°C) than the tempcraturc suggested by Prof. Gutowsky. A correction to the observed second moment values which might help to explain the discrepancy mentioned by Dr. Davies is concerned with the vibration of the atoms within the molecule. Deeley and Richards 1 have treated the case Of atoms vibrating along the line joining them, which causes only the length of the inter- nuclear vector to vary. If this correction is applied the bondlength is further increased.However, in our casc the vibration of the protons also causes the directions of the internuclear vectors and their angles with respect to the applied field to vary, leading to a more complicated corrcction which might have opposite sign. We are extending thc work on monocrystalline urea to lower and higher tem- peratures to find out more definitely than can be done using polycrystalline material the nature of the motion responsible for the narrowing of the spectrum at higher temperatures. Dr. J. A. S. Smith (Leedr University) said: I think that the question of the mechanism of the transitions ip PTFE is still an open one. I should like to discuss briefly some of the evidence which seemed to us relevant to the problem. In the first place, the first nuclear resonance transition decrcases the second moment by about 6 gauss2, which seems too large to be accounted for by a change in the so-called amorphous regions of the polymer alone, for these constitute only about 30 % of the bulk of the material.This point is brought out by inaking a rough calculation of the percentage of the polymer which has undergone the transition at 270°K at which temperature the second moment is 5.5 gauss2. If we assume that the second moment of a specimen of the polymer in which the helices are effectively stationary is 11.4 gauss2 and one in which the helices are undergoing hindered rotation is 2.3 gauss2, and assume thesc values to be the Same for both amorphous and crystalline regions, this percentage comes to 67.This figure agrees better with a model in which two-thirds of the chains are rotating and the remainder are stationary. In the second place, the experimental evidence shows that this is a roughly quantitative explanation, for as already mentioned the total width of the line falls by about 5 gauss between 200" and 250" K. The line-shape at these inter- mediate temperatures is apparently not a simple superposition of two curves, one of which is due to a rigid lattice and the other to a lattice affccted by molecular rotation, but both to some extent must be affected by thc molecular motion. The comparison of the results for PTFE with those for polythene may throw Some further light on the nature of the transitions. It must be remembered, however, that as many as four in a hundred carbon atoms in the polythene chain may be branch mcthyl groups, so that extra complications may occur because of the presence of CH3 or CH groups in the polymer.1 Deeley and Richards, Trans. Furuduy Soc., 1954, 50, 560.252 QENERAL DISCUSSION Prof. G. E. Pake (Washington University, St. Louis) said : As noted in Dr. Smith's excellent paper, Dr. C. W. Wilson 1 and 1 investigated the nuclear magnetic resonance of polytetrafluoroethylene (PTFE) in 1952. Since the published report of this work was extremely brief, it should be of interest to compare here the Washington University data with those of Dr. Smith. The Washington University data agree very well with the curves presented in fig. 3 and 4 of Dr. Smith's paper. In addition, Dr. Wilson measured the spin- lattice relaxation time TI and, as mentioned by Dr.Smith, interpreted both line width and T1 studies to indicate that there are two distinct F19 environments in PTFE, each giving rise to a TI and a T;! (inverse line-width parameter). Thus the resonances which appear to have structure or an inflection peak were inter- preted as resulting from a superposition of two simple bell-shaped absorption curves which have, in certain temperature ranges, different widths. The strongest support for this interpretation comes from the TI measurements. At a temperature of 170" K, for example, no resolvable structure has appeared in the resonance, and the data indicate that both kinds of F19 nuclei are sufficiently " frozen in " to give the full rigid lattice second moment.However, the resonance intensity alters in a complex way when one makes a saturation measurement2 to determine T I , as indicated by the experimental curve shown in fig. 1. We Siqoal qenerator output voltaqe FIG. 1.-Saturation curves for the P 9 resonance in PTFE showing a normal type of curve (135" K) and the complex type (170" K). Measurements were made at 30 Mc/sec. suggest that the 170" K curve illustrates the saturation first of those F19 nuclei in an environment which contributes about 28 % of the low-power resonance intensity, followed at higher power by saturation of the resonance from the re- maining 72 % of the nuclei. The occurrence o f a normal curve at 135" K is explained when one plots the TI values against temperature, for one finds that the T1 of the 28 % region increases rapidly with decreasing temperature, and at 135" K, power sufficiently low to avoid saturation gives totally inadequate signal-to-noise ratios for its signal observation. Other complex saturation curves support these general conclusions, although the saturation measurements require more care in interpretation to find T1 values where the line shape is complex, one region having developed a narrower resonance.1 Wilson, Ph.D. Thesis (Washington University, St. Louis, Mo., U.S.A., 1952) ; 2 see Bloembergen, Purcell and Pound, Physic. Rev., 1948, 73, 679 for a discussion J. Polymer Sci., 1953, 10, 503. of the saturation technique for measuring TI.GENERAL DISCUSSION 253 We were able to saturate selectively either the broad line or the narrow line at temperatures where the one was still at the rigid lattice width (T2 = 0.9 X 10-5 sec) and the other had progressed to considerable narrowness (T2 = 5 X 10-5 sec).This progressive narrowing of the one line, it should be pointed out, renders the inflection peak ratio of the derivative ineffectual as a means of fixhg the percentage of F19 nuclei in each environment, since the area under the absorption curve is proportional to, for a given line-shape function, the product of the square of the width and the derivative maximum. The derivative maxima of fig. 1 are an indication of intensity at 170" K only because both lines are found to have the same width (that corresponding to a rigid lattice). The decomposition of the complex line shapes into a derivative of a broad curve and that of a narrow curve, which were then subsequently integrated as described in the published account of this work, is admittedly a somewhat arbitrary procedure.However, numerous curves at different temperatures gave results gratifyingly consistent with each other and with the 72-28 distribution obtained from fig. 1. If these procedures all measure the same quantity, we can quote 72 rt 5 % as including all experimental determinations of this quantity. On the basis of these data, we suggest that 72 % of the F19 nuclei contribute to a resonance with distinct TI and T2 against temperature curves, and the remaining F19 nuclei possess different TI and T2 against temperature curves. The further suggestion that these different fluorine environments are the crystalline and amorphous regions of the polymer is, we think, plausible, but additional work is certainly necessary to substantiate this hypothesis.Dr. N. Sheppard (Cambridge University) said : I am particularly interested in the application of nuclear magnetic resonance (n.m.r.) to the study of the structure of complex molecules. It appears to me that, even when allowance is made for the fact that the normal nuclei of carbon and oxygen atoms are not accessible by this method, n.m.r. may well be the technique of next usefulness for the study of organic molecules after the universally applied methods of infra-red and ultra-violet spectroscopy. Dr. Shoolery has given us some very good examples of this type of work. He has studied a number of cyclobutane derivatives (his compounds A, B and C) in which it is probable that there are CF2 and CH2 groups adjacent to each other.The n.m.r. spectrum shows that the two fluorine nuclei are not in chemically equivalent positions in the molecule and the spectrum due to one of these shows extra triplet fine-structure. It is suggested in the paper that this fine structure is due to the stronger interaction of this fluorine nucleus with the adjacent protons cause of non-planarity of the cyclobutane carbon skeleton, the CF2 and CH2 groups assume relative positions with respect to each other between the extremes of (1) F- - ~ ~ - - - eclipsed and staggered configurations. Looking along the connecting C--C bond below. 1s it likely that the fluorine atom one would then observe a structure as (1) which lies between the two hydrogen atoms is the one that is more strongly coupled with the protons? It may be that there are difficulties in this type of inter- pretation of the spectra which I have not appreciated. In case this is so I would like to generalize my question and to ask if, in a more favourable case such as a substituted cyclohexane molecule, it might be possible to tell from the fine structure due to interaction with neigh- bouring protons whether a F atom (or other suitable nucleus) is in an axial or of the CH2 group.It is possible that, be- / / @\\ / I H /' I I (2) F254 GENERAL DISCUSSION an equatorial position on the ring? If from high resolution n.m.r. measurements one is likely to be able to deduce in this type of manner not only the chemical nature of adjacent groups, but also their spatial relationships, the method is going to be a very powerful one for the study of molecular structure.Dr. J. N. Shoolery (Varian Associates, California) said : The cyclobutane carbon skeleton can indeed assume a non-planar configuration and the adjacent CF2 and CH2 groups very probably do assume some position other than the sym- metrical eclipsed configuration. One of the fluorine nuclei is then coupled more strongly to the protons. Since the fluorine atom (1) pictured by Dr. Sheppard lies between the hydrogens, and since experimentally the coupling appears to be about equally strong to the two hydrogens, there is a temptation to pick fluorine (1) as the more strongly coupled. However, the details of the coupling mechanism are not fully understood and such an interpretation is certainly open to question. In the general case, I would say that with sufficient experience we may eventually be able to deduce spacial relationships, at least in favourable cases, as well as the chemical nature of adjacent groups. Mr. R. P. Bell (Oxford University) said : It is of great interest that Ogg finds it necessary to invoke rhe tunnel effect in order to explain the apparent identity of the protons in AlB3H12. Calculations made about 20 years ago 1 indicated that this effect might be important in chemical reactions involving protons, and recent experimental work (Bell, Fendley and Hulett, to be published shortly) on proton and deuteron transfer in solution provides experimental support for this view. However, I am not sure that the case is proved for AIB3H12. The total frequency of interchange will be of the form zP(.i) exp (- Ej/kT), where p(.i) is the probability of tunnelling associated with the energy level ei. This will be temperature independent only if most of the transitions take place from the lowest level : this is certainly not the case in our own theoretical or experi- mental work, where the most important levels at ordinary temperatures appear to be those only a few kilocalories below the top of the barrier. This is probably also true for the problem treated by Ogg ; on the other hand, the observed spectrum could be independent of temperature even if the tunnelling rate is temperature- dependent, provided that the latter rate is great enough over the temperature range investigated. It would be of interest to know the actual rates of non- classical transfer estimated by Ogg in his problem. i 1 Bell, Proc. Roy. Soc. A , 1933, 139,466 ; 1935, 148, 241 ; 1936, 154,414.