|Home | About | Journals | Submit | Contact Us | Français|
The specific rates of solvolysis of chloromethyl phenyl sulfide [(phenylthio)methyl chloride] and its p-chloro-derivative have been determined at 0.0 °C in a wide range of hydroxylic solvents, including several containing a fluroalcohol. Treatment in terms of a two-term Grunwald-Winstein equation, incorporating terms based on solvent ionizing power (YCl) and solvent nucleophilicity (NT) suggest a mechanism similar to that for the solvolyses of tert-butyl chloride, involving in the rate-determining step a nucleophilic solvation of the incipient carbocation in an ionization process. A previous suggestion, that a third-term governed by the aromatic ring parameter (I) is required, is shown both for the new and for the previously studied related substrates to be an artifact, resulting from an appreciable degree of multicollinearity between I values and a linear combination of NT and YCl values.
It is well-established that alkyl chlorides where derivatization involves the introduction of a substituent atom such as oxygen, sulfur, or nitrogen on the α-carbon give, on ionization, resonance stabilized carbocations (equation 1).
The formation of the carbocations, with an appreciable amount of delocalized charge, is relatively rapid [1–3]. When X = O, they are often used to protect hydroxyl groups and the resulting acetals can be easily deprotonated by a mild treatment with acid .
The effect of the adjacent lone pairs of electrons can be dramatic. For example, the solvolysis of chloromethyl ethyl ether (CH3CH2OCH2Cl) in 36% dioxane proceeds at a rate of about 109 times that of butyl chloride  and, the ethanolysis of chloromethyl methyl ether proceeds about 1013 times as fast as that of propyl chloride  and of at least 1014 times as fast as that of methyl chloride . It is of interest that, in bimolecular reaction with ethoxide ion, the second-order rate coefficient is also considerable larger (by a factor of about 105) than for the methyl chloride .
The corresponding aryl α-chloroalkyl ethers are easily prepared and they also undergo unimolecular nucleophilic substitution in water and alcohols . The replacement of an alkyl group by an aryl group leads to an appreciable reduction in the solvolysis reaction rate. For example, in 98% dioxane (2% water) at 25.0 °C, chloromethyl methyl ether reacts 1090 times as fast as chloromethyl phenyl ether . Many of the more recent studies have involved cyclic substrates from the carbohydrate field [8,11]. The acetals involved in these reactions are among the most important types of biological functional groups .
There have also been extensive studies of the corresponding thioethers (X = S). Chloromethyl phenyl sulfide (1) is readily available and it is important  as a reactive electrophile, as an α-methylenating agent, as an acyl carbanion equivalent, as a source of phenylthiocarbene and the phenylthiomethyl carbanion, and as a precursor for halomethyl phenyl sulfoxides.
Indeed, the phenylthiomethyl group is so commonly encountered that it is frequently abbreviated to PTM. The compound 1, or a ring substituted derivative, is of use as a protecting agent for alcohols or phenols , as an alternative to the formation of the more common methoxythiomethyl (MTM) alkyl [15–17] or aryl  ethers. PTMCl was found to be an excellent reagent towards phenols but not, in general, towards alcohols (where considerable amounts of side products are formed). The solvolyses of 1 can be pictured as in Scheme 1.
When the substrate is acyclic, it has been found that, in a comparison of the rates of SN1 solvolyses of an α-chloroether with those of the corresponding α-chlorothioether, the ether reacts faster than the thioether. It was suggested that this was due to the larger sulfur atom not conjugating as well with the carbocation center . In terms of stabilities, there is both experimental  and theoretical  evidence that the adjacent sulfur stabilizes a carbocation better than an adjacent oxygen. On these grounds, one might expect a faster reaction to the more stable sulfur-containing carbocation.
However, it was found for solvolyses in aqueous dioxane of both the (dichlorodimethyl) ether and sulfate and the chloromethyl ethyl ether and sulfide, that the ether reacted faster than the sulfide .
Also it has been found  that when the X atom (O or S) is within a six-membered glucopyranosyl or thioglucopyranosyl ring, the fluorides hydrolyze with the thio-derivative reacting the faster. This represents a reversal of ordering observed for the previously studied  acyclic equivalents. It was proposed  that an enhanced anomeric effect is smaller for the S-C-F system than for the O-C-F system, leading to a smaller ground state stabilization for the S-C-F system. The mechanism of reactions of substrates containing these ring substituents has been reviewed .
It was shown that for hydrolysis in 5 M water in dioxane at 25 °C, the phenylthiomethyl chloride (chloromethyl phenyl sulfide, 1) reacted about 200 times slower than chloromethyl methyl sulfide . The effect of the introduction of meta- and para- substituents into the aromatic ring of 1 has been studied  in 50% dioxane at 35 °C. Use of the Hammett σ values (the original “normal” values) gave a better fit for the solvolyses of 1 than the σ+ values, which would be expected to apply if considerable positive charge was being distributed into the aromatic ring . The Hammett σ value was −2.6. This is an appreciable value for a reaction governed only by inductive and field effects, with significant resonance effects ruled out by the applicability of the σ values. The lack of resonance effects was attributed to the reluctance of divalent sulfur to expand its valence shell to ten electrons . It is found that, with fairly powerful nucleophiles in solvents of relatively weak ionizing power, bimolecular nucleophilic substitution (SN2 reaction) can be observed. A study has been made  of the reactions of para-substituted phenyl chloromethyl sulfides with aniline in dimethylformamide. An Hammett plot, incorporating σ values, led to a σ value of −0.84, indicating a modest accelerating effect of electron-supplying substituents, consistent with an SN2 reaction with bond-breaking running modestly ahead of bond-making. Similarly, it has been found that exchange reactions of chloride ion with chloromethyl aryl sulfide in acetonitrile follow second-order kinetics  and involve an initial complexation .
Another consideration is the extent to which replacement of a hydrogen on the α-carbon carrying the halogen by some other atom or group influences the rate and mechanism of the unimolecular solvolysis reaction of an α-chloroether or α-chlorothioether.
The influence of electron-withdrawing groups was probed using a successive introduction of additional chlorines into the β-position for the solvolyses of phenyl α-chloroethyl ether. The rates of reaction fell off appreciably with each additional chlorine introduced, with the group on the α-carbon going from CH3, to CH2Cl, to CHCl2, to CCl3. The CCl3CHClOC6H5 reacted 500,000 times slower than CH3CHClOC6H5 .
In a study with thioethers , it was found that the solvolyses of CH3SCH2Cl in 1 M water in dioxane at 25 °C increased in rate by a factor of 500 on the introduction of a methyl group into the chloromethyl, to give CH3SCH(CH3)Cl. If, in turn, the introduced methyl group was replaced by a phenyl group, it was found that, for reaction with 0.2 M water in dioxane at 25 °C, there was a further 140 fold increase in rate. These quite large increases were taken as indicating that no saturation level as regards electron supply had been approached.
There have also been studies of RSCH2Cl compounds which have an electron-withdrawing benzoyl or acetyl group replacing an α-hydrogen. This would be expected to lead to a reduction in the rates of SN1 reactions. If sufficiently powerful, it is possible that this effect could lead to the observation of contributions from a solvent-assisted SN2 process.
In Scheme 2 are given the structures of the six substrates given consideration in this report.
Initially, 2-chloro-2-(methylthio)acetophenone (2), and 2-chloro-2-(phenylthio)acetophenone (3) were studied . In a later paper, the solvolyses of 2-chloro-2-(4-methoxyphenylthio)acetophenone (4) and 2-chloro-2-(methylthio)acetone (5) were added to the study . These substrates are of additional interest in that the carbocation formed in an SN1 reaction formally has the positive charge on a carbon which is adjacent (α) to the carbonyl group. Carbocations of this type are normally difficult to form [33–35], and it is of interest to see what extent their formation is favored by having an alkylthio or arylthio substituent also at the reaction center.
The studies by Ryu et al [31,32] included a study of the effect of solvent variation on the specific rates of solvolysis (first-order rate coefficient for solvolysis) of compounds 2 – 5 (structures shown in Scheme 2). These specific rates were then analyzed using simple and extended forms of the Grunwald-Winstein equation (2). In this original  one-term equation, k represents the specific rates of solvolysis in a given solvent of ionizing power Y, ko represents the specific rate in the standard solvent (80% ethanol, Y = 0), m represents the sensitivity of the specific rates of solvolysis to changes in solvent ionizing power, and c
represents a constant (residual) term. In order to take into account the nucleophilic assistance from the solvent present in SN2 reaction (plus E2, when elimination reaction accompanies substitution) a second term, expressed as lN, was added to equation 2, to give equation 3 [37,38]. In equation 3, l represents the sensitivity of the specific rates of solvolysis towards changes in solvent nucleophilicity (N). This scale is also based on 80% ethanol being the standard solvent (both N and Y are set at zero).
It was observed early that reactions forming carbocations with aromatic substituents on the α-carbon deviated from the behavior described by equation 1 and, when several series of binary solvents of varied compositions were included in a study, dispersion into a series of linear plots was observed [39, 40]. In order to account for these aromatic ring effects, a third term, represented as hI, has been added to equation 3, to give equation 4. In equation 4, h
represents the sensitivity of the solvolysis to changes in the solvent property designated as the “aromatic ring parameter”, I . The development and uses of the extended forms of the Grunwald-Winstein equation have recently been reviewed . The scales used for solvent ionizing power (YX for a leaving group X) are now usually based on the solvolyses of the 1-adamantyl or 2-adamantyl derivative [43,44], as opposed to the original tert-butyl chloride scale  and scales of solvent nucleophilicity (NT) are usually based on studies of the solvolyses of the S-methyldibenzothiophenium ion [38,45].
Ryu et al applied equations 2 and 3 to the solvolyses of compounds 2 through 5 and, for the two compounds with an aromatic ring at the sulfur (3 and 4), they also applied the three-term equation 4. We have repeated the analyses reported using equation 4, with inclusion of all of the data points available. The validity of including the hI term for these systems is given additional consideration. We have also extended the studies to the solvolyses of 1 (compound 3 without the benzoyl substituent at the α-carbon) and to chloromethyl p-chlorophenyl sulfide [(p-chlorophenylthio)methyl chloride, 6].
The correlation analyses have been carried out using equations 2, 3, and 4. Particular attention will be paid to the sensitivities towards solvent nucleophilicity changes and towards changes in the aromatic ring parameter value, considered as a measure of solvent effects upon charge delocalization into aromatic rings.
The chloromethyl phenyl sulfide (1, Sigma-Aldrich, 97%) and chloromethyl 4-chlorophenyl sulfide (6, Sigma-Aldrich, 97%) were used as received. Solvents were purified and the kinetic runs carried out as previously described . A substrate concentration of approximately 0.007 M was used. Solubility considerations limited the range of aqueous-organic solvents which could be employed, especially for 6. For example, specific rate values could be observed in 40% ethanol and 60% methanol for 1 but this was not possible for 6. The specific rates and associated standard deviations, as presented in Table 1 for each of the two substrates, are obtained by averaging all of the values obtained from, at least, duplicate runs.
Multiple regression analyses were carried out using the Excel 2010 package from the Microsoft Corporation.
The specific rates of solvolysis of chloromethyl phenyl sulfide 1 and chloromethyl p-chlorophenyl sulfide 6 were determined at 0 °C in ethanol-water (100–40% ethanol), methanol-water (100–60% methanol), aetone-water (80–60% acetone), TFE-water (97–70% TFE), HFIP-water (97–70% HFIP) and across the full range of TFE-ethanol mixtures. The values obtained are listed in Table 1. In a footnote to the table, values are presented at 25–55 °C for the solvolyses of 6 in 100% ethanol, together with the calculated enthalpy and entropy of activation. The solvent nucleophilicity (NT) values  and solvent ionizing power (YCl) values [44,46] used in the correlation analyses are also listed.
The solvolyses of 1 and 6 proceed at a convenient rate over a wide variety of hydroxylic solvents at 0.0 °C. The introduction of a para-chlorine atom into 1, to give 6, results in a reduction in rate by a factor of four to six (Table 1). Calculation of a series of Hammett σ values based on the two data points for each solvent leads to surprisingly constant values over a wide composition range for the binary solvent mixtures. In aqueous ethanol (100–60% EtOH) and aqueous methanol (100–80% MeOH), eight determinations lead to a value of −2.9 ± 0.2 and four determinations in aqueous TFE (97–80% TFE) and 90% HFIP lead to a value of −3.7 ± 0.3. The average value of −2.9 in aqueous methanol and ethanol is in good agreement with the value of −2.6 for solvolyses of the parent and seven meta- and para- substituted derivatives in 50% dioxane . While serious consideration of Hammett correlations with only two data points is, as a general rule, not recommended, the observation of constant σ values over several EtOH-H2O and MeOH-H2O compositions and over four TFE-H2O and HFIP-H2O compositions does suggest that these values have relevance.
The major aspects of the present study are to apply the various forms of the Grunwald-Winstein equation (equations 2–4) to the kinetic data reported for the solvolyses of 1 and 6 in Table 1 and to revisit the Grunwald-Winstein correlations previously reported for compounds 2 – 5 [31,32] and, in particular, to give further consideration to the significance of the analyses including the hI term (use of equation 4).
The analyses suffer when solvolyses in TFE-ethanol solvent mixtures are included. This is often the case (discussed elsewhere ), due to the data points for these solvent systems lying below the best fit correlation line. Removal of these solvents considerably improves the correlations, as determined by comparison of simple or multiple correlation coefficients, F-test values, and the reduced probabilities that some of the terms making a small contribution are not statistically significant as regards their contribution to the overall correlation.
The analyses using equation 3 (two-term) lead for 16 solvolyses of substrate 1 to an l value of 0.27 ± 0.08 and an m value of 0.76 ± 0.06 and for substrate 6 to an l value of 0.22 ± 0.08 and an m value of 0.67 ± 0.06 for 17 solvents. These sensitivities correspond to l/m ratios of 0.36 and 0.33, respectively. The l/m ratios are similar to those obtained for solvolyses of tert-butyl chloride (l/m of 0.44) [46,48]. Accordingly the original Y scale [36,44] should give reasonably good one-term correlations for the solvolyses of 1 and 6, equivalent to using the solvolyses of tert-butyl chloride as a similarity model for the solvolyses [46,48,49]. It might be mentioned that the term “similarity model” in this context is often misunderstood as meaning similarity in structure. While the structures are often similar this is not a requirement. The requirement for a good similarity model is that the ratios of the sensitivities from a two-term treatment, in this case the l/m ratios are similar for the system under consideration and for the similarity model employed.
When the 16 data points for 1 and the 17 data points for 6 are plotted as log (k/ko) against Y values, values are obtained for the solvolyses of 1 of 0.97 ± 0.08 for the slope, 0.08 ± 0.09 for the intercept, 0.958 for the correlation coefficient, and 157 for the F-test value. The corresponding values for 6 are 0.92 ± 0.08, 0.15 ± 0.10, 0.947, and 131. Both correlations are acceptable and the slightly better correlation for the solvolyses of 1 is consistent with the l/m value being closer to that observed for the tert-butyl chloride similarity model. The plots against Y values are shown in Figures 1 and and2.2. When the log (k/ko) values are plotted against YCl values, reasonably good plots are obtained for both substrates but the correlations are improved when solvent nucleophilicity is also included in the correlation (equation 3). The values in Table 2 show a considerable improvement in the correlation coefficient accompanied by a small decrease in the F-test value. The l values (sensitivities to changes in solvent nucleophilicity) have a 0.005 for solvolyses of 1 and 0.015 for solvolyses of 6 probability that the lN term is not statistically significant in the overall correlation. Ideally, these probability values should be somewhat lower, with a value below 0.001 being considered desirable, but this is frequently the situation for low values for l, with the standard errors tending to be fairly constant, irrespective of the magnitude of the l value . The two-term correlations are presented in Figures 3 and and44.
It is of interest to give further consideration to the extent of further improvement (if any) on applying the three-term equation 4 to the data. The values obtained from these correlations are also presented in Table 2. For 1, there is a considerably improved correlation (F-test value increasing from 195 to 415) but the situation is less clear cut for 6, with a modest increase in the correlation coefficient and a modest decrease in the F-test value. Care must be taken because of the previously demonstrated  relationship between the I scale and a linear combination of NT and YCl values. We will return to this aspect after the presentation of new analyses (more data points) for the solvolyses of substrates 2 through 5. It was pointed out recently  that in consideration of multicollinearity one must consider not only collinearity between pairs of scales but also between any one scale and any linear combination of other applicable scales.
The previous studies of compounds 2–5 were analyzed in terms of the application of equations 2 and 3 and, for the two substrates 3 and 4 containing an arylthio substituent, also in terms of equation 4. The substrates 2 and 5 containing no aromatic rings were not originally analyzed in terms of equation 4 [31,32]. However, in view of Takeuchi’s finding  of a correlation against a combination of Y and I values for the solvolyses of some highly sterically hindered tertiary alkyl derivatives, we felt that application of equation 4 to these other two substrates (2 and 5) would also be worthwhile. As pointed out earlier, the experimental  and theoretical  evidence against a resonance interaction of an adjacent positively charged sulfur with an attached aromatic system suggests that the hI term may not be relevant in the correlation analyses of the solvolyses of any of the 1 through 6 substrates.
The correlation analyses of compounds 2 through 5 mimic those for 1 and 6 of Table 2, and they are reported in Table 3. In contrast to the earlier [31,32] correlations, the maximum number of data points were included in all of the correlations. Only for the solvolyses of 3 was there any indication that the analyses would be improved by omission of the specific rates measured in 100 and 90% ethanol and 100% methanol (the solvents most favorable for an SN2 mechanism to operate). Even here, the improvement in the correlations on omission of these three data points is, at best, marginal (Table 3).
For the correlations reported in Table 3, the m values obtained from the simple one-term treatment (equation 2) are just slightly lower than for the solvolyses of compounds 1 and 6. When the two-term equation (equation 3) is used, the l values for 3 and 4 of 0.31 and 0.33 are only slightly greater than for 1 and 6 but for substrates 2 and 5 considerably larger l values are obtained, reaching a value of 0.88 for 5 and at a value of 0.58 for 2. The m values, however, at 0.59–0.74 are very similar to those of 1 and 6 (0.67–0.76). The larger l values, sensitivities to changes in solvent nucleophilicity, are in accord with the need for additional nucleophilic solvation (or increased nucleophilic participation in a loose SN2 transition state) in the presence of the powerful electron-withdrawing acetyl or benzoyl substituent at the reaction center [33–35]. The study of 3 was at 50.0 °C  and in 80% ethanol a solvolysis specific rate of 4.00 × 10−4 sec−1 was reported. From specific rates at 0.0° and 25.0 °C (Table 1), a value of 24.4 × 10 −4 sec−1 can be estimated at 50.0 °C for the solvolyses of 1 in 80% ethanol at 50.0 °C: the introduction of an α-benzoyl group into 1 to give 3 results in a rate reduction by a factor of 6.1. This is a relatively small influence, consistent with favorable changes in the transition state structure being the rationale for the accompanying change in the l/m ratio from 0.36 to 0.56 (Tables 2 and and33).
When equation 4 (three-term) is applied to the solvolyses of substrates 1 and 6 negative values for h, with large standard error, are obtained when all of the solvents are considered. The values become positive, however, when the data points for solvolyses in TFE-ethanol are excluded from the correlations (Table 2). When solvolyses of substrates 2 through 5 are considered (Table 3), positive h values are obtained in three cases (0.36 to 0.99) and a negative value (-0.42) when 5 is the substrate.
In a recent review in this journal  of the correlations of highly hindered tertiary alkyl derivatives, it was demonstrated that the l and m values of the three-term equation were always lower than those of the two-term equation when h was negative and always higher when h was positive, as one would expect if the multicollinearity as expressed in equation (5) was operative. An inspection of Tables 2 and and33 shows that this behavior is also observed for
all nine of the analyses of the six solvolyses under consideration.
Such behavior strongly supports the conclusion previously reached  that the observation of an appreciable hI term, even for systems not containing any aromatic rings or conjugated π electron systems, is an artifact resulting from the presence of the multicollinearity as expressed in equation 5. It was also shown that, as desired, I correlates very poorly with Y values and poorly with NT values (presented for tert-butyl chloride solvolysis previously  and in Table 4), and only when correlated against both NT and YCl is there an important multicollinearity problem.
For each of the entries in Tables 2 and and3,3, the equation 5 has been applied to the set of solvents included in the correlations. It is shown in Table 4 that the multicollinearity as expressed in the equation is robust and the x, y, and c values vary only slightly with changes in the solvents included. Indeed, the average values are essentially identical to those recorded in the table for tert-butyl chloride solvolyses and these values (x = 0.29; y = 0.14; c =0.03) have been used in our analyses. The I within the hI term of equation 4 has been replaced by the expression given in equation 5, and incorporating the x, y, and c values for the 43 solvents used in the tert-butyl chloride solvolyses. In this way, with a combination of terms now based on NT and YCl, we convert from an equation 4 representation to an equation 3 representation. In Table 5, it is shown that the values obtained in this way, except for increased standard errors, are essentially identical to those obtained directly from an application of equation 3. This exactly parallels the observations previously  on application to the highly hindered tertiary alkyl derivatives.
The responses of the rates of solvolysis of the (arylthio)methyl chlorides 1 and 6 to changes in the composition of a hydroxylic solvent are quantified by the application of the extended Grunwald-Winstein equation (3). This results in the observation of a high sensitivity to changes in solvent ionizing power (m value) coupled with a modest sensitivity to changes in solvent nucleophilicity (l value). We favor an SN1 mechanism with a modest nucleophilic solvation of the developing carbocation, similar to the mechanism we proposed for tert-butyl chloride solvolysis . An alternative description  involving a loose SN2 transition state, very similar in detail, could also lead to the observed sensitivities. Indeed, for 1, SN2 reactions with strong nucleophiles in relatively inert solvent are well documented [28,29].
Previous analyses of the specific rates of solvolyses of compounds 2 through 5 [31,32], with an acyl substituent at the reaction center have been repeated with an increased number of data points, especially for the analyses applying the three-term equation. The sensitivities obtained in these analyses are quite similar to those reported earlier . One feature of the earlier report was a claim that there was a relevant contribution from the hI term (in equation 4) for compounds 3 and 4, which are derivatives of an (arylthio)methyl chloride. Subsequent to this claim being put forward, we have shown  that great care must be taken in the application of equation 4, due to a multicollinearity relationship between I values and a combination of the NT and YCl values for groupings of solvents. Indeed, it is now clear that the apparent utility of equation 4, not only for substrates 3 and 4 but also for 2 and 5, is an artifact due to the presence of this multicollinearity. The lack of a need to incorporate an hI term is not surprising because the positive charge adjacent to the aromatic ring in (arylthio)methyl chloride would lie on a sulfur atom, which has been shown to be reluctant to expand its valence shell to above eight electrons [25,27] as would be required to remove electron density by resonance from the directly attached aromatic ring.
In ethanol at 50.0 °C, the introduction of a benzoyl group into 1 to give 3 leads to a six fold reduction in the specific rates of solvolysis. This is a reasonable value considering that an acetyl group in the meta position has a Hammett σ value of +0.38 and the Taft σI value for the acetyl group is +0.28 .
This research was supported, in part, by the donors of the American Chemical Society Petroleum Research Fund (postdoctoral fellowship for BCP) and, in part, made possible by the Delaware INBRE program, supported by grants from the National Center for Research Resources - NCRR (5P20RR016472-12) and the National Institute of General Medical Sciences - NIGMS (8 P20 GM103446-12) from the National Institutes of Health (NIH); a National Science Foundation (NSF) EPSCoR grant EPS-0814251; and a NSF ARI-R2 grant 0960503.