J . Chem. SOC., Faraday Trans. I , 1982, 78, 2183-2194 Decomposition Reactions in the Flame Ionization Detector BY ANTHONY J. C. NICHOLSON CSIRO Division of Chemical Physics, P.O. Box 160, Clayton, Victoria, Australia 3 168 Receiued 8th July, 198 1 A computer simulation has been made of reactions postulated as occurring in the flame ionization detector, f.i.d. Upstream of the luminous zone, alkanes, alcohols and ethers decompose in an atmosphere of hydrogen to give methane. For each additive the calculated yield of methane equals the experimentally determined relative ionization yield. The response characteristics of the f.i.d. follow from this, provided that the additive decomposition and the ionization reaction occur in separate regions of the flame. For the very different response of an f.i.d.using carbon monoxide as fuel instead of hydrogen, the methane yield and the ionic yield are also equal. Although the flame ionization detector, f.i.d., has been used in gas chromatography' as a sensitive and versatile detector of organic molecules for over twenty years, the chemical reactions which give it its particular electrical response are still not established. The f.i.d., as normally operated, is a hydrogen diffusion flame burning in air. The most detailed and satisfactory attempt to fit the experimental observations on the f.i.d. into current theories of hydrocarbon decomposition and of ionization in flames was that of Blades., He proposed that the 'equal-per-carbon' response of the f.i.d. to hydrocarbon additives followed if all additives were converted to the same distribution of single-carbon hydrides prior to ionization or oxidation.Nicholson and Swingler3 suggested that this distribution was set by the equilibrium of H,, H, CH,, CH,, CH and C. The similar suggestion had been made by Hayhurst and Vince4 that a number of phenomena in hydrogen flames, including ionization, could be correlated by assuming a series of stripping reactions CH, +H+CH,-, +H, of which one was the rate-determining step for production of CH. Blades2 proposed that after the first reaction producing radicals the hydrocarbon chain was degraded by a series of H-atom cracking reactions such as C,H5+H+2CH,. The availability of a computer program5 for simulating the kinetics of a number of coupled chemical reactions provided an opportunity for testing these postulates in a quantitative manner.A mechanism for the degradation process in the f.i.d. was set up using rate constants from the literature. In spite of uncertainties in our knowledge of some of these rate constants, the calculation gives an insight into which reactions contribute to the overall decomposition and which reactions are too slow to be relevant. It also shows whether equilibria can be attained in the time available. A mechanism involving methane production that does not require the equilibrium hypothesis will be shown to fit the facts adequately. 21832184 THE FLAME IONIZATION DETECTOR The experimentally observed f.i.d. relationships that must be deducible from a mechanistic theory are as follows. (a) If & is the ion yield per additive molecule, then K is constant for additive concentrations from to of the concentration of the carrier gas.6 (b) If RK is the yield relative to methane taken as one, then Rx equals the carbon number for hydrocarbon additives (the ‘equal-per-carbon ’ response).Blades2 showed that this relationship only holds if the additive is carried up to the flame in a gas stream containing hydrogen; this observation has recently7 been questioned. The value for acetylene, R ( c ) For a carbon attached to a hetero-atom an ‘effective carbon number’ less than one can be determined.6 RX for a molecule then equals the sum of the actual and effective carbon numbers, the latter being roughly transferable from molecule to molecule. Quantitative values (C=O = 0, CH30 = 0.3 in dimethyl ether but zero for its homologues, COH = 0.75 and so on) must be reproduced by the mechanism.( d ) When carbon monoxide replaces hydrogen as fuel in an f.i.d., q varies as the additive concentration at low concentrations and is constant at high concentrations,s for a carbon monoxide flame in the high-concentration region is ca. 0.08 of that for a hydrogen flame (from a comparison of the figures of McWilliam8 and Sternberg et a1.6 for C,H16). FLAME MODEL = 2.6, is exceptional.2g A representative f.i.d. flame, as described in our earlier experirnent~,~ had a blue luminous zone in the shape of a blunt-ended cylinder ca. 1.2 mm in diameter and 4 mm long. From the measurements of Ohline et aL9 and of Mitchell et al.l0 on similar diffusion flames, a reasonable estimate of the temperature 0.1 mm inside the luminous zone is 1800 K.For a cold gas flow of 2 cm3 s-l the velocity of the gas as it passes through a 1 mm diameter x 4 mm long cylinder at 1800 K is ca. 1 m s-l. The time for the gas to travel 0.1 mm is then 100 ps; the program was arranged to calculate the products formed in this time from various concentrations of the chosen additive reacting at 1800 K. The assumption is made that additive decomposition occurs in a hydrogen atmosphere in a region separate from that in which the ionizing and combustion reactions involving oxygen take place. Calcote’l has shown that the maximum rate of ionization occurs downstream from the luminous zone where the temperature is also at a maximum. Mitchell et al.l0 give plots of [HCN] and [NO] as a function of position in a diffusion flame of methane burning in air. They show that [HCN] peaks at ca.2 mm upstream of the luminous zone, where the temperature is 1400 K, whereas [NO] is almost separated from [HCN] and peaks at the temperature maximum, 2000 K, ca. 1 mm downstream from the luminous zone. In other words, the reducing and oxidizing regions of such flames are physically separate. The calculations given below are for the reducing region, and the half-lives obtained for the additive decomposition are a further justification for neglecting reactions between an additive and 0 or 0,. COMPUTER PROGRAM The program, developed by Davis,5 uses Gear’s algorithm for integrating a group of ‘stiff’ coupled differential equations.It has an option in which a system is considered at constant pressure with variable temperature and volume which corres- ponds reasonably to a flame. Preliminary runs showed that the compositions after 150 ps differed little regardless of whether a temperature rising in steps of 50 ps at each of 1600,1700 and 1800 K or a fixed temperature of 1800 K was used, so the latterA. J. C . NICHOLSON 21 85 was chosen for convenience. The total gas concentration, [MI, was taken as 4 x 10l8 molecule ~ r n - ~ , of which 3 x 1018 molecule cm-3 was hydrogen. Concentrations of various additives from 10l6 molecule ~ m - ~ downwards were used. The program does not accept zeros so all free-radical and atom concentrations were set initially at of the additive concentration.All reasonable reactions of the additive were written into the program, which calculated the concentrations of all the atoms, radicals and molecules as a function of time. RATE CONSTANTS The reactions and rate constants used for alkane pyrolysis [reactions (1)-(27)] are given in table 1. In general the rate constants used were taken from the review of Jensen Y and Jones.12 This-is a self-consistent set chosen for their applicability to TABLE RATE CONSTANTS FOR THE PYROLYSIS OF PARAFFINS k = ATn exp (-E/RT), A/cm3 molecule-' s-l or s-l, T/K, E/kJ mol-1 - initiation (unimolecular) 1 M+H,+ H+H +M 2 3 CzH, CH, + CH, 4 6 M +CH, e CH,+ H + M C,H, + C,H5 + H C3H8 * C3H7 + H 5 C3Ha + CH3 +C,H5 7 C4HIo s CzH5 + CzH5 8 9 10 C,H,, s C,H, + CH, C4H,, e C,H, + H M +C,H, =C,H, +H +M initiation (bimolecular) 11 13 14 H + C,H5 + CH, + CH, H, + C,H, + CH, + CH, H, + C3H, + C,H, + CH, 12 H, + CZH, G= CzH3 + H 15 CzH4 + CzH4 + CzH5 + CH3 radical decomposition 16 M+C,H, +C,H,+H+M 17 M +C,H, -P C,H,+H + M 18 C3H7 + C,H, + H 19 C3H7 -+ CzH4 + CH, 20 C4H, + C3H6 + CH, 21 C4H, + C2H4 + C,H5 22 23 CzH, +CH3eCH4+C,H,.hydrogen abstraction H, + C H , g CH, + H 24 H, + CzH3 S C,H4 + H 25 28 29 H, + C2H5 + CzHe + H M +HCO+ CO+H + M H, +CO =$ HCO +H 26 27 H, + C,H, + C3H8 + H H, + C4H, + C,H,, + H forward _ _ _ _ _ log A n E log A __________ -5.10 5.70 16.04 16.04 16 16 15.30 15.30 15.30 -6.22 -9.3 - 1 1.28 -9.3 -9.3 -8.96 - 8.92 -9.47 13 13 13 13 - 10.52 - 24.04 -11.32 - 16.64 - 1 1 -11 -9.4 - 10.85 438 454 36 1 403 343 393 323 356 376 41 1 X O 269 234 222 268 126 126 173 139 137 120 62.4 34.6 24 38.6 32 32 66 400 - 29.52 -20.70 - 11.00 -9.3 - 9.3 -9.3 -9.3 - 9.3 - - 32 - -11.15 - - - - 33 - 32 -9.3 -9.3 - - -9.15 -24.52 - 10.52 - 15.30 - 10 - 10 - 32.7 -9.7 back n E - 1 0 -3 0 - 4.2 __ - - - 40 - - 62.4 4 60.3 29 2 29 32 32 7 21 - - - - - flame ref.12 12 12 13 13 14 13 13 15 16 12 __ - - 15 12 15 13 13 13 13 12 15 12 12 - - 12 12 a Ref. (12) gives log A,, = - 17.64 and log A-25 = - 16.30. The values used here are a t the upper limit of their error bounds. temperatures. More recently Baulch and Duxburyl' have recommended a slightly faster rate for reaction (3) but consider that its rate is falling off at ca. 1019 molecule ~ m - ~ as it moves towards second-order kinetics. This makes little difference to the rate of 71 FAR I2186 THE FLAME IONIZATION DETECTOR reaction (3) at 1800 K.On the other hand, reaction (17) is in its transition region at this total gas concentration, so the second-order rate constant given by Koike and Gardner15 is preferred. Uncertainties in knowledge of rate constants of hydrocarbons increase with increasing molecular weight; the values chosen here are rounded off values from the review by Benson and O’Neal.13 They are slightly smaller than those given in a more recent review.lS Provided that the rate constant for unimolecular decomposition is > 7 x lo4 s-l (99.9% decomposition in 100 ps) the results obtained below are not sensitive to the value of this rate constant. The back reactions between radicals given in table 1 are unimportant compared to the reactions of radicals with hydrogen.They only become significant when hydrogen is replaced by carbon monoxide, with which there is no equivalent metathetical reaction. Values of these rate constants were calculated from the equilibrium constants given by Jensen and Jones12 or from those in the JANAF Tables,19 or put equal to 5 x 1O-lo cm3 molecule-’ s-l. Decompositions involving the breaking of C-H bonds in hydrocarbons, reactions (4), (6) and (9), are rarely mentioned in the literature because they are usually overshadowed by reactions involving breaking of C-C bonds. If these rate constants are estimated by taking the same pre-exponential factor and correcting for the higher bond dissociation energy, D, then the C-H breaking rates are slower than the C-C breaking rates by exp[(D,,-D,,)/RT].This factor has a value of ca. 10 at 1800 K, and these reactions are in fact significant. In table 2 are given rate constants for a mechanism with methanol and dimethyl ether as additives. TABLE 2.-RATE CONSTANTS FOR THE PYROLYSIS OF CH30H AND CH30CH3 k = ATnexp( -E/RT), A/cm3 molecule-l s-’ or s-l, T / K , E/kJ mol-l forward back log A n E log A n E ref. initiation 30 31 32 33 34 35 36 37 CH, + CH,OCH, CH,OCH, + CH, 39 41 M + CH,OH =$ CH, + OH + M M +CH,OH s C H , O + H + M CH,OCH, + CH,O + CH, CH,OCH, + CH,OCH, + H M +CH,O g CH,O + H + M M +CH,OCH, g CH,O + CH, + M radical decomposition hydrogen abstraction OH + CH,OH g CH,O + H,O 38 OH + H,+ H,O+H 40 CH,O + H, + CH,OH + H CH, + CH,OH + CH,O + CH, CH,OCH, + H, + CH,OCH, + H - 5.3 -3 15 15 - 10 - 10.4 - 11.18 - 10.3 - 14.72 -9.3 -9.3 - 12.52 - 335 -32 - __ 418 -32 - - 318 -11.7 - - 418 -11.7 - - 121 -11.7 - - 60 -11.7 - - 8.4 -9.3 - 63 -9.3 - 1.3 15 - 14.08 1.3 - 48 -10.7 - 20 -9.3 - - 41 -11.2 - - - 0 20 0 0 21 0 - - 0 20 0 21 71 20 75 20 77.6 12 29 16 55.6 22 - .- . RESULTS ETHANE, PROPANE A N D BUTANE Fig. 1 shows the products formed as a function of time when lo9 molecule ~ r n - ~ of C,H6 is taken as the initial concentration of additive in the scheme of table 1. This figure shows that: (i) The major molecular product is CH,, and in 10 ps it has reached a limit set by the amount of C,H6 initially present. Little change in products occurs between 10 and 100 ps. (ii) For reactions (I) and (- 1) the equilibrium valueA. J.C. NICHOLSON I I I I I H2 18 t \c*&- C2H6 2187 0 20 40 60 80 100 time/ps FIG. 1 .-Plots of the logarithm of the concentrations of the products from log molecule cmW3 of C,H, as additive against time at 1800 K. Hydrogen flame. of [HI (1.25 x 1015 atoms cmW3 at 1800 K) has not been reached. Dissociation of H, is negligible. (iii) C,H, is present to ca. 0.05% of the CH,. Its amount is decreasing but so slowly as not to be visible on this scale. The amount is lo5 times greater than it would be if the equilibrium C,H, + 2H, 2CH, had been established. (iv) Reactions (22) and (-22) are almost in equilibrium: [CH,][H,]/[CH,][H] = 47, whereas the equilibrium value is 23. Other products have reached ' steady states' that are not equilibria.Fig. 2 gives similar plots for an initial ethane concentration of 1015 molecule ~ m - ~ . Once again CH, is the main product, although the relative proportions of the minor products have changed. [HI is higher than the equilibrium value, and equals [CH,] throughout. The equilibrium of reactions (22) and (-22) is established. [C,H,] at 100 ps is now 0.4% of [CH,]. Reactions (3) and ( - 3 ) are almost in equilibrium. If initial concentrations of propane or butane are substituted for ethane into the scheme of table 1, the general shape of the concentration against time curves remains the same and methane remains the dominant product. Product concentrations after 100 ps for equivalent concentrations of the first four alkanes are given in table 3. The postulate that every alkane produces the same distribution of minor products is wrong, but the quantities of minor products are so small relative to the quantities of methane that their effect on the yield of methane is negligible.Fig. 3 is a log-log plot of the concentrations of some of the products against the initial additive concentration. This particular plot is for C3H, additive, but the plots are the same for C,H, and C4H1, except for the absolute values of the ordinate. Over a dynamic range of at least lo7, [CH,] is linear with initial additive concentration, 71-22188 - 14 m . 2 12 .- Y E 4 d c 10 s 9 8 W 2 00 THE FLAME IONIZATION DETECTOR - \ \ C4HIO 0 20 40 60 80 100 timelps FIG. 2.-Plots of the logarithm of the concentrations of the products from I O l 5 molecule of C,H, as additive against time at 1800 K.Hydrogen flame. TABLE 3.-DECOMPOSITION PRODUCTS AFTER 100 / l S FOR PARAFFIN ADDITIVES initial concen- tration /lo14 molecule product after 100 p s concentration/molecule ~ r n - ~ additive ~ r n - ~ CH4 H CH, ‘ZH4 CZHZ C3H6 ‘ZH5 ~~ ~ ~ C4H10 1 3.7 x 1014 5.7 x 1014 1.6 x lo’* 1.5 x 1013 1.0 x 1012 4.4 x 109 3.1 x 107 CZH, 2 4.0 x 1014 4.2 x 1014 5.3 x 1011 3.8 x 1011 1.2 x 1011 6.2 x 108 2.7 x 105 CH4 4 4.0 x 1014 4.0 x 1012 1.9 x 1010 3.5 x 109 1.3 x 109 1.1 x 109 4.0 x 104 C3H8 1.3 3.8 x 1014 5.4 x 1014 4.0 x lo1* 6.8 x 10l2 3.2 x 10l2 1.6 x log 9.3 x los , CH4 H CH3 C2H2 C2H4 r 8 10 12 14 16 log,,(initial concentrati~n/cm-~) FIG. 3.-Log-log plot of concentration of products after 100 p s against initial concentration of additive, C,H,, at 1800 K.Hydrogen flame.A. J . C. NICHOLSON 2189 a, 0 4 E l 2 . ... x 22 - ... x t: a, m ... - .3 Y c 2 I I 1 I J 8 10 12 14 16 2 1 ' u log,,(initial concentrati~n/cm-~) FIG. 4.-CH, yield, molecule at 100 p s per initial additive molecule plotted against the logarithm of the initial concentration. The circles give the ion yield, ions per additive molecule relative to CH, = 1 . Hydrogen flame. i.e. its yield per additive molecule, YCH4, is constant, and as shown in fig. 4 this yield is almost equal to the carbon number of the additive. Methane is the only product for which this is true, and neither relationship holds for methyl radical concentration. Rather, over the range 1012-1014 molecule cm-, of additive the slope of the log [CH,] line is 2.The point (ca. lo1, molecule ern-,) where this curve steepens corresponds to the point where the [HI curve approaches the [CH,] straight line. At this point also the equilibrium of reactions (22) and (-22) is established, i.e. [CH,] = [CH,][H]/K,,[H,], and the square law follows since both [CH,] and [HI vary as the initial additive concentration. The only liberty taken in selecting rate constants was to put k,, at the upper limit of the range given in ref. (12). YCH, is sensitive to the ratio of the rates of reaction (25) to reaction (7) and, if this is reduced by a factor of 10, YCH, for C,Hl, is reduced to 3.2 although the yield from C,H, or C,H, is barely altered. CARBON MONOXIDE FLAME Reactions in a carbon monoxide flame are simulated by the reactions for alkanes plus reactions (28) and (29) with the initial H, replaced by CO.The distribution of products, fig. 5, is now quite different (cf. fig. 1). CH, is the major product, CH, a minor one and it is the equilibrium of reactions (3) and (- 3) which is reached. In fig. 6 a log-log plot of the products against initial additive concentration also demonstrates a reversal since [CH,] is linear with additive at low additive concentra- tions, whereas over the range 1011-1014 molecules of additive per cm3, [CH,] varies as the square of the initial additive concentration. Thus for both flames it is the methane yield which parallels the RT behaviour given in the introduction under (a), (b) and ( d ) . In the CO flame at low concentrations [H,] is also proportional to the square of the additive concentration, but once the absolute amount of hydrogen becomes appreciable the system moves towards that of a hydrogen flame, the [CH,] and [CH,] curves cross and the [CH,] tends towards being proportional to the first power of the additive concentration.This behaviour is also paralleled by Rq in CO flames,, although the changeover occurs at a lower concentration than in the model system. The ratio of the methane yields at initial [C,H,] = lo1, molecule cm-, in CO and H, is 0.025, whereas the corresponding ratio of ion yields is 0.08. Agreement is fair when one considers that experimental data on CO flames are scarce and that there2190 THE FLAME IONIZATION DETECTOR 0 time/ps FIG. 5.-Plots of the logarithm of the concentrations of products from los molecule ~ m - ~ of C,H, as additive against time at 1800 K.Carbon monoxide flame. FIG. 6.-L0g-l0g plot 4 ' IOJ 1; Id 1's I; ' log,,(initial concentrati~n/cm-~) of concentration of products after 100 ,us against initial concentration of additive, C,H,, at 1800 K. Carbon monoxide flame. are doubts8 about the purity of the carbon monoxide used in some of the published experiments. O X Y GEN-CONTAIN ING MOLECULES CH+O+HCO The reaction is so exothermic that there is enough energy available to ionize HCO. Conversely, the reactions reducing a carbonyl group to CH are so endothermic, and accordingly slow, that a carbonyl group gives zero ionic response in the f.i.d. It follows that anyA. J . C. NICHOLSON 2191 reaction of the O-containing molecule that produces a carbonyl group will lower the value of Rq if it is fast enough to compete with the reactions producing ions.If, as indicated above, CH, production is an essential step in the path to ion production, this possibility can be further tested by seeing whether YCH, is lower than the carbon number for dimethyl ether and methanol. Formaldehyde is the obvious carbonyl compound that could be formed from these additives. The question is whether its rate of formation is comparable with that of CH, formation. These decompositions have been modelled by adding the reactions of table 2 to reactions (1)-(4), (1 l), (1 3), (1 7), (22), (23) and (25) of table 1 ; the rate constant values of table 2 are more uncertain than those of table 1. DIMETHYL ETHER The yields of methane, YCH,, for CH,OCH, over a range of additive concentrations are shown in fig.4. The value k,, = 1 x exp (- 121 000/RT) gives YCH, = 1.3 in exact agreement with R for CH,0CH3.2 For all concentrations, the sum of YCH, and YHpco is two. The agreement is only of limited significance since YCH, is sensitive to the value of k,,; specifically, YCH, is reduced to 1 by increasing k,, by a factor of 40. However, this sensitivity means that a slight increase in the rate constants of the equivalent of k,, (e.g. CH,CH20+CH3 + H,CO for diethyl ether) would account for the zero Rq of the ether group in all ethers other than dimethyl ether.6 METHANOL For CH,OH decomposition, CH,O has been used to represent the two radicals CH,O and H2COH, since no quantitative data are available to determine any difference in their reactivities.The hydrogen abstraction reactions, (36), (- 39) and (41), are not fast enough to produce H2C0 via CH30 in appreciable quantities in the times considered here. Accordingly, reaction (31) is postulated with a rate making it competitive with reaction (30). The YCH, value calculated is then 0.93. This is higher than the RF value of 0.75 given by Sternberg et al.6 However, YCH, falls off at high additive concentrations more sharply than the yield from alkanes (fig. 4) and, since the measurements of Sternberg et al. were made over the range 1013-1015 molecule ern-,, the discrepancy is not large. As for CH,OCH,, the sum of YcH, and YHzC0 equals the carbon number of the additive. ETHYLENE A N D ACETYLENE In an earlier paper3 we assumed that ethylene and acetylene would be reduced to alkanes by the excess of hydrogen present sufficiently rapidly for degradation to methane to occur. The calculations of the present paper make this unlikely, but the values chosen for the rate constants are uncertain.The only measurements of gas-phase hydrogenation date from the 1930s, of Pease23 on C2H, and Taylor and van Hook2, on C2H,. For such chain reactions it is possible to calculate the rate constant for the initiation reaction from the overall rate of the reaction; SemenofF gives such a calculation for Pease’s measurements. Even if the assumed mechanism is correct, the initiation reaction contributes only a small fraction to the overall reaction and cannot be specified very accurately.Reactions (1 I)-( 15), grouped under the heading of ‘ Bimolecular Initiation’ have been assigned rate constants at the upper limit of what seems reasonable in an endeavour to find a reaction path from unsaturated to saturated that could be significant in 100 ps. All are inadequate and their omission, along with reaction (10) and all reactions involving C2H2 and C,H,, makes a negligible alteration to the calculations described above for the alkanes. The fastest reaction2192 THE FLAME IONIZATION DETECTOR removing C,H, is reaction (lo), and its half-life is 0.36 s at 1800 K and 2.3 x s at 2200 K, which is about the highest temperature reached in f.i.d. flames. On present knowledge of rate constants, C2H4 and C,H, must be regarded as exceptions which do not degrade to CH4 within 100 ps at 1800 K as do alkanes, alcohols and ethers.METHANE Methane, like C2H4 and C,H,, decomposes by breaking a C-H bond which is stronger than the C-C bonds in its homologues. The lifetime of CH4 for decomposition by reaction (2) is 4.3 x lo-, s at 1800 K and 3.9 x lo-, s at 2200 K. CH,, C,H, and C,H, are unique among the gases considered here in that their lifetimes are long with respect to the time-scale used in the calculations. A calculation of products after 100 p s is given in table 3. For the same final amount of CH, the amounts of minor products are markedly different for CH, itself than for its homologues decomposing to CH,. DISCUSSION The calculations given here were planned to give quantitative expression to the mechanism given in a previous paper,3 which was itself an extension of the mechanism given by Blades., Two assumptions made there3 are shown here to be wrong.The first is that equilibrium is reached in the H, f 2H reaction [reactions (1) and (- l)]. In 100 ps, [HI is above its equilibrium value at high additive concentrations and below it at low concentrations. The equilibrium CH3+H, *CH,+H is only attained at the high end of the additive concentration range. The calculations made here give concentrations reached after 100 ps, and whether these are equilibrium values or not is irrelevant. The second, made originally by Blades, that 'all additives give the same distribution of single carbon hydrides before ionization' is not quite correct. All additives give CH, as a major product, while the variable amounts of minor products make a negligible contribution to the decomposition.It has been shown that YcH, from C,H,, C,H, and C4H1" parallels the relative ionization yield in the usual f.i.d. (H, flame) in its major characteristics; the constancy with additive concentration, the wide range of this constancy and its equality to the carbon number. Considerable alterations to the rate constants given in table 1 would be needed to cause a departure from this parallelism. In the f.i.d. with CO as fuel, R F is linear with concentration over a small range and then becomes almost constant; this is also the behaviour of YCH,. Although the rate constants needed are less accurately known, the calculation also correlates YCH, from oxygen-containing compounds with the effective carbon numbers in the f.i.d.of the COH and CH30 groups. A complete explanation of the f.i.d. relationships then requires a mechanism yielding ions in amounts proportional to CH, concentration. Experimentally this was established by Peeters et a1.26 for a premixed flame with H, : 0, = 7 : 3 and it holds for CH, in the f.i.d. diffusion flame. A surprising result of the calculations given above is that while [CH,] is linear with additive concentration over the whole range over which ionization is linear, [CH,] is at first linear, then follows a square law and finally requires an intermediate exponent as the additive concentration rises (fig. 3). This suggests that the reaction path giving ionization (whether via CH or in any other way) proceeds from CH, without passing through CH,.A possible reaction by-passing CH, formation is CH, CH, + H, for which Chen et al.27 give a rate constant of 6.3 x 1014exp (-474000/RT) s-l, stating that this is probably a maximum value. We used a value of one quarter ofA. J. C . NICHOLSON 2193 this in a simuIation of the reactions of the single-carbon hydrides with H, H, and C and found that [CH,], [CHI and [C] were proportional to the initial [CH,] while [CH,] was proportional to the square of the initial [CH,]. No more is claimed for this calculation than that it provides a possible mechanism, since most of the rate constants used were estimated. This calculation also shows that the hydrogen-atom cracking mechanism postulated by Blades is relatively unimportant.For example, in the reaction plotted in fig. 2 the rate of removal of C,H, by atom cracking, reaction (1 I), is 1 O6 [C,H,] molecule ~ m - ~ s-l, whereas by hydrogen abstraction, reaction (25), it is 1.7 x lo7 [C,H,] molecule ~ m - ~ s-l. At lower additive concentration reaction (1 1) is even slower, 10, [C,H5] molecule ~ r n - ~ s-l for the reaction of fig. 1, whereas k25 [H,] does not change with additive concentration. The calculations show that it is reactions with the hydrogen molecule that are important and that alkane decomposition is complete before an appreciable amount of hydrogen has dissociated. This high concentration of hydrogen, in suppressing radical-radical reactions, is essential to give total conversion to the single-carbon compound, CH,.Recently Wagner et al.’ have examined the response of alkanes in the hydrogen atmosphere flame ionization detector, h.a.f.i.d, in which the additive is carried in an oxygen stream, in the inner of two concentric jets, to burn in an atmosphere of hydrogen. They observed an equal-per-carbon response, which seems to be in direct contradiction to measurements of Blades,, and state that ‘an explanation unifying the results reported in this study and in that of Blades remains unclear’. They also state that ‘if only one mechanism is to be proposed as the origin of the equal-per-carbon response it must operate in both hydrogen-rich and oxygen-rich environments ’. The mechanism involving CH, formation advanced here cannot operate in an oxygen-rich flame and a postulate that more than one mechanism exists seems more likely.A difference worth noting is that while the response and methane formation are linear over seven orders of additive concentration in the f.i.d., the h.a.f.i.d. response is only linear over two. Response- per-carbon is only meaningful in the linear region. Additive decomposition in oxygen atmosphere might resemble the carbon monoxide experiments modelled in this paper, where methane formation exhibits both linear and square-law variations with concentrations. Since the mechanism involving CH, production cannot be true for C,H, and C,H, an ad hoc postulate must be made that some other route to ionization is available for these two gases. Blades pointed out the anomalous position of C,H, referring to ‘ the embarrassing situation of basing the mechanism of ion formation on work on C,H,, the very compound for which there is evidence of unique behaviour’.He emphasized that C,H, catalyses H-atom recombination28 rather than being itself reduced, that ionizing reactions other than CH +O could not be ruled out with certainty, and that a C,H +O,* ionizing reaction might be peculiar to C,H,. Later, Blades29 compared ionization rate with CH emission rate and showed that their ratio was higher for C,H,, C,H, and CH, than for other alkanes. He argued that this required an additional ionization reaction to the generally accepted reaction via CH. The most detailed study of the chem-ionization of C,H, by 0 atoms, that of Vinckier et goes no further than saying that ‘an intermediate’ gives HCO+, either in a reaction with 0 or via CH which reacts with another 0 atom.These observations and the still unexplained R x value of 2.6 for acetylene are not inconsistent with the view that acetylene is a special case. It would seem that C,H, is also, although this has not been apparent since its R value of 1.98 fits into the normal pattern. The differences between CH, as additive and CH, produced from higher alkanes might be associated with the differences in minor products shown in table 3.2194 THE FLAME IONIZATION DETECTOR I am indebted to Dr T. McAllister who pointed out the program to me and taught me how to use it, and to Dr H. L. Davies for information about the program. I also thank the referees for helpful suggestions that have improved the presentation of this paper.I. G. McWilliam and R. A. Dewar, Nature (London), 1958, 181, 760 A. T. Blades, J. Chromatogr. Sci., 1973, 11, 251. A. J. C. Nicholson and D. L. Swingler, Combust. Flame, 1980, 39, 43. A. N. Hayhurst and I. M. Vine, Nature (London), 1977, 266, 524. H. L. Davies, CSIRO Division of Computing Research Program Simulation of Chemical Kinetics: H. L. Davies and M. Y. Smith, Proc. 8th Australian Computer Conference, Canberra, 1978 (Australian Computer Society, Canberra, 1978), p. 277. J. C. Sternberg, W. S. Gallaway and D. T. L. Jones, in Gas Chromatography, ed. N. Brenner, J. E. Callen and M. D. Weiss (Academic Press, New York, 1962), p. 231. I. G. McWilliam, Nature (London), 1970, 228, 356. R. W. Ohline, E. Thall and Ping Hwat Oey, Anal. Chem., 1969, 41, 302. ' J. H. Wagner, C. H. Lillie, M. D. Dupuis and J. J. Hill Jr, Anal. Chem., 1980, 52, 1614. lo R. E. Mitchell, A. F. Sarofim and R. Yu, Combust. Sci. Technol., 1980, 21, 157. * l H. F. Calcote, 8th Symp. (Int.) on Combustion (Williams and Wilkins, Baltimore, 1962), p. 184. l2 D. E. Jensen and G. A. Jones, Combust. Flame, 1978, 32, 1. l3 S. W. Benson and H. E. O'Neal, Kinetic Data on Gas Phase Unimolecular Reactions (NSRDS-NBS21, U.S. Department of Commerce, Washington, D.C., 1970). l4 D. G. Hughes, R. M. Marshall and J. H. Purnell, J. Chem. Soc., Faraday Trans. 1, 1974, 70, 594. l5 T. Koike and W. C. Gardiner Jr, J. Phys. Chem., 1980, 84, 2005. l6 P. Camilleri, R M. Marshall and J. H. Purnell, J. Chem. Soc., Faraday Trans. I, 1974, 70, 1434. l7 D. L. Baulch and J. Duxbury, Combust. Flame, 1980, 37, 313. l8 D. L. Allara and R. Shaw, J. Phys. Chem. Ref. Data, 1980, 9, 523. l9 D. R. Stull and H. Prophet, JANAF Thermochemical Tables (NSRDS-NBS37, U.S. Department of 2o C. R. Westbrook and F. L. Dryer, Combust. Sci. Technol., 1979, 20, 125. 21 P. D. Pacey, Can. J. Chem., 1974, 53, 2742. 22 D. Aronowitz, R. J. Santoro, F. D. Dryer and I. Glassman, 17fh Symp. (Int.) on Combustion (The Combustion Institute, Pittsburgh, 1979), p. 633. 23 R. N. Pease, J. Am. Chem. Soc., 1932, 54, 1877. 24 H. A. Taylor and A. van Hook, J. Phys. Chem., 1935, 39, 81 1. 25 N. N. Semenov, Some Problems of Chemical Kinetics and Reactivity, transl. J. E. S. Bradley 26 J. Peeters, C. Vinckier and A. van Tiggelen, Oxid. Combust. Rev., 1969, 4, 93. 27 C-J. Chen, M. H. Back and P. A. Back, Can. J. Chem., 1975, 53, 3580. 28 J. V. Michael and €3. Niki, J. Chem. Phys., 1967, 46, 4969. 29 A. T. Blades, Can. J. Chem., 1976, 54, 2919. 30 C. Vinckier, M. P. Gardner and K. D. Bayes, J. Phys. Chem., 1977, 81, 2137. Commerce, Washington D. C., 1971). (Pergamon, London, 1958), p. 218. (PAPER 1/1417)