Thermodynamics and Molecular Orbital Calculations for the Parabens

The intent of this study was to follow two approaches of scientific methodology in the investigation of a chemically related series of pharmaceutical preservatives, the parabens (esters of para-hydroxybenzoic acid). One approach involved the experimental determination of the solubilities of these compounds and also included their partition coefficient determination. The other approach, strictly theoretical in nature, was manifest by the utilization of molecular orbital theory in the determination of molecular parameters and their possible correlation with physico-chemical parameters and/or drug action. The solubilities of a series of solutes related to each other by their increasing size in alkyl chain in an ester moiety are determined and compared. These compounds include benzoic acid, para-hydroxybenzoic acid and the methyl, ethyl, propyl, and butyl esters of parahydroxybenzoic acid. The effect of temperature upon the solubilities of these compounds in a series of n-alkanols allows for the calculation of thermodynamic parameters for these systems. Since heat of fusion values are available in the literature, then this leads to the determination of other theoretically important thermodynamic quantities which have been presented. This coupled with the experimentally evaluated thermodynamic elements, allows for some insight into the solubility mechanism. Previously enunciated empirical theory of dielectric requirements may be placed on a more quantitative basis. The rather extensive experimental support of the multiplicity of solubility peak maxima for a single solute can be interpreted on a thermodynamic basis. The partition coefficient, an important physicalchemical property of biologically active drugs, was determined from a biphasic system of n-octanol-water at 25°C for each of the solutes. Correlation of this work with the previous work of Hansch and with rank order of activity of these compounds is given. With respect to the determination of molecular parameters, appropiate computer programs were utilized to obtain the desired results for the f ollowing: aniline, para-aminobenzoic acid, phe~o l, benzoic acid, parahydroxybenzoic acid, methyl and ethyl esters of parahydroxybenzoic acid. It should be pointed out that the computer elements utilize the Iterated Extented Ruckel Theory (IEHT) approach, which contains assumptions of various types. For any series of molecules these assumptions are self consistent, however, so that comparisons can easily be made. The approach taken in this portion of the study was to characterize these systems in terms of atomic charge distributions, effect of hydroxy (-OH) and ester groups (both sterically and electronically), preferred molecular conformations and distinguishing features of the molecular orbitals. A hypothesis as to the possible site of action leading to biological activity of these compounds has been proposed.

quantities which have been presented.
This coupled with the experimentally evaluated thermodynamic elements, allows for some insight into the solubility mechanism. Previously enunciated empirical theory of dielectric requirements may be placed on a more quantitative basis.
The rather extensive experimental support of the multiplicity of solubility peak maxima for a single solute can be interpreted on a thermodynamic basis.
The partition coefficient, an important physicalchemical property of biologically active drugs, was determined from a biphasic system of n-octanol-water at 25°C for each of the solutes. Correlation of this work with the previous work of Hansch and with rank order of activity of these compounds is given.
With respect to the determination of molecular parameters, appropiate computer programs were utilized to obtain the desired results for the f oll owing: aniline, para-aminobenzoic acid, phe~o l, benzoic acid, parahydroxybenzoic acid, methyl and ethyl esters of parahydroxybenzoic acid.
It should be pointed out that the computer elements utilize the Iterated Extented Ruckel Theory (IEHT) approach, which contains assumptions of various types. For any series of molecules these assumptions are self consistent, however, so that comparisons can easily be made.
The approach taken in this portion of the study was to characterize these systems in terms of atomic charge distributions, effect of hydroxy (-OH) and ester groups (both sterically and electronically), preferred molecular conformations and distinguishing features of the molecular orbitals. A hypothesis as to the possible site of action leading to biological activity of these compounds has been proposed.

ACKNOWLEDGEMENT
The author wishes to express his appreciation to all those persons with whom he has come in contact at the University of Rhode Island and whose help he has gratefully received during his course of study there.
In particular, thanks are due to: Professor Anthony N. Paruta, under whose super-    Of the various descriptions of the ideal preservative for pharmaceuticals (1)(2)(3)(4), the list of characteristics given by Gershenfeld and Perlstein (3) are pertinent: It must be effective against all types of microorganisms that cause decomposition.
It must be soluble at the concentration used. It must not be toxic internally or externally depending upon where it is to be used at the concentration in which it is employed.
It must be compatible: i.e., it must not alter the character of the preparation as to give rise to odor, color, taste, etc.; and it should be practically neutral so that it will not affect the pH of the preparation.
The cost of the preservative should not increase the price of the preparation to any marked extent.
The inhibiting effect must be a lasting one; therefore, non-volatile substances are to be preferred.
The preservative agents which most nearly fulfill the above characteristics are esters of parahydroxybenzoic acid.
The larger the alkyl group in the ester, the greater the inhibiting action (11,21).
The para-hydroxybenzoic acid esters are two to three times as effective as benzoic acid in inhibiting bacterial growth (22).
The suitability of these compounds for the preservation of food stuff~ pharmaceuticals, leather, textiles, ink, etc. has been reviewed in the literature (23-28).

3
Contrary to these findings, early work of Boeseken and Waterman (29) showed that molds utilized para-hydroxybenzoic acid as a nutrient, and more recently, Davis (30) reported that para-hydroxybenzoic acid is indeed a bacterial vitamin, a possible intermediate in the synthesis of various aromatic compounds from shikimic acid. Nevertheless, Davis proposed using para-hydroxybenzoic acid as a model for chemotherapeutic analogs selectively toxic to microorganisms.
To date, however, the "true" mechanism of their preservative action has not been elucidated.
Microbiological proof of their microbial and therefore their preservative action has received acceptance despite the paucity of information regarding their microbial pharmacology or biochemistry.
The solubility phenomenon of a solute in a given solvent has been of interest to generations of scientists.
Pharmacists have been concerned with non-electrolyte solubility in particular.
The basis for much of the investigation in the field is the work of Hildebrand; indeed the applications of Hildebrand's theories and its use has been discussed recently by Mauger ( 32). Much of it is indispensible to the work at hand and will be dealt with in subsequent chapters. Chertkoff and Martin (33) initiated studies on the solubility of benzoic acid in mixed solvents. Restino and Martin (34) continued the investigation using esters of para-hydroxybenzoic acid, in a series of n- where Kx is the ionization constant of a benzoic acid derivative substituted by a group X, and KH is the ioniz- where PC' is the phenol coefficient for the drug activity. Table 7 gives the data necessary for the generation of regression analysis data and comparison with the data reported by Hansch (30). The regression analysis was performed on the reported data with the following result: log PC' = -0.1653(log P) 2 + l.7743(log P) -2.1897 The exclusion of the data for phenol itself gave log PC' = -0.1028(log P) 2 + l.3794(log P) -1.5915 Hence it can already be seen that regression coefficients may readily change by the inclusion or exclusion of per- The following five equations were tested for exactness of fit with the experimental data with or without the inclusion of Hansch data for missing members in the series log PC' = a p + b (8) log PC' = a log p + b (9) log PC' = a P2 + b p + c log PC' = a log (P)2 + b log p + c (11) l og PC' = a( log P)2 + b log p + c (12) where PC' is the phenol coefficient and P the partition coefficient. The results follow:    Tables 3 to 8 .
In Figure  The parent acid and the n-alkyl esters were approximately parallel to one another.             or (4) volume fraction. Mole fraction notation has been adopted due to its acceptance in the scientific community, and x 1 is solvent and x 2 solute.
Analytical expressions for free energy can be given as follows: Equating Eq. (1) and (2) a plot of The application of basic thermodynamics to the solubility phenomenon requires the following definition of the solution process: H Solid Solute _!'("I}-. Liquid State --r~J"'8olute in Solution (7) As the solid solute proceeds through step one to the liquid state, the enthalpy change is the heat of fusion,

Hf.
If the solution is "ideal" the enthalpy of the second step is zero. When a solid solute passes to a liquid state and then into solution, its behavior is said to be "non-ideal" if the enthalpy of the second step is not zero. Under conditions where both steps one and two are operative, the total enthalpy of the process is commonly expressed as the heat of solution, Hs , where Hs = Hf + X.
In this solubility pathway, X reflects the enthalpy involved in the transfer of the solute from the liquid to the solution phase. As the X term approaches zero, the heat of solution and the heat of fusion approach equality, and the system approaches ideal behavior.
When Hf = Hs then the pathway is ideal, the solubility process is athermal, and X, which is identified as a thermodynamic mix ing function (Hm=heat of mixing), is zero. Therefore Sm = enthalpy of mixing = -R ln x 2 •  (9) (10) (11) (12) If, however, we are dealing with a"non-ideal" solution process and therefore the equation becomes -log x 2 (non-ideal) AHs = 2.3 RT (14) which is in the form y = mx + b and when plotted yields   --Compa r i son of the •ideal a n d• e xp e rimental log mole fra ction solubility for the solut es noted above vs . reciprocal temperatu r e.
(Kelvin) 4 5 In order to describe the conditions which contribute to a "non-ideal" state it is necessary to ascertain the activity of the solute. Activity, a, of a component in a particular solution is the ratio of the thermodynamic escaping tendency in a system to its escaping tendency in some arbitrarily chosen "standard state".
Non-ideal solubility is thus associated with the activity coefficient as follows: and, therefore, the relationship between "ideal" and "non-ideal" solubility becomes and when solubility is "ideal", (°=l and a 1 =X 1 and equation 23 simplifies to (24) In the same manner the solute yields similar equations AF 2 = (F2 -F;) = -RT ln a 2 = -RT ln x 2 + RT ln y-; (25) and in the "ideal" case, · ~=l and a2=X2 to give (26) The overall process can therefore be expressed as follows where AF = the overall energy change for the system (the molal free energy change) Therefore, when the solubility is "non-ideal" the following expression is readily obtained AF (non-ideal) = -RT(X 1 ln x 1 + x 1 ln ~ + x 2 ln x 2 + X 2 ln r;) .

49
The energy which is present over and above that which is considered to be ideal is considered to be an "excess" function.
Once again the contributions of each component in solution must be recognized. Hence F~ = -RT ln ~ and F~ = -RT ln r-; where F~, 2 = the partial excess molal free energy and the overall contribution is seen as (30) where FE = the overall excess free energy change for the system (Molal excess free energy change).
In summary, then, solubility can be expressed as a function of the energy change in the following manner AF(non-ideal) = 4F(ideal) + FE(excess); (31) therefore, as the excess energy approaches zero, the non-ideal solubility approaches the ideal function. The activity coefficients for the solutes in a 50 particular solvent with respect to temperature were calculated by Eq. 22 and are presented in Table 11. The activity coefficients for the solvent in the above systems were similarly calculated and are presented in Table 12 . were derived from Eq. 8 in which X represents Hm and the ref ore (32) and (33) The results of the above thermodynamic determinations indicate several factors which soon become apparent.
During the course of the experimental work, the solutions were observed to become cold to the touch. In fact some even formed frost on the sides of their con-            In order for this to occur, heat must be absorbed from the external environment and therefore 6Hs must be positive for such a phenomenon. The data substantiate this observation.        zoic acid, para-hydroxybenzoic acid, methyl and ethyl paraben as seen in the tables can be associated with the salvation process; therefore, it may be said that energy can be either liberated or consumed with this secondary process.
The twin peak array observed in the solubility spectrum shown in Fig. 6 is seen throughout the thermodynamic picture of the investigational compounds. This phenomenon was first evidenced by the ~Hs and ~Ss values derived by graphical means in Tables 1 3 to 1

It
can be seen that any alteration in 4Hs was also accompanied by a corresponding change in 6Ss such that the free energy remains consistent for any given system. Thus, total inspection of the thermodynamic f t .

A -a
-E E unc ionsAHs, Ss, AF ,6F, F, and F truly reveals a trend for a multi-peak solubility array. Therefore, it can be said that a solute which is soluble in several or many solvents may approach ideal solubility in one or several of these solvents and thus exhibit one or several maxima or peaks in its solubility scan. The dielectric requirement of one peak is now felt to be related to the thermodynamics of the total system; hence, 81 a multi-peak system should signify that ideal solubility is being approximated in that particular solvent-solute system.

INTRODUCTION
During the past ten years, the mechanics of interaction between drugs and receptor sites have been intensively scrutinized in the hope of describing these interactions, their nature, rate, and steric dependence in quantitative terms by an estimation of chemical and physical parameters occurring in a viable system.  (4,5), Dewar (6), Streitweisser (7), Roberts (8), Purcell (9), and Martin (10 (1) the tendency to form insoluble polymers in aqueous media; (2) the capacity to bind phospholipid hydrophobically; (3) the capacity to interact with other membrane proteins to form water-soluble complexes; the capacity to bind small molecules; (5) the tendency of the polymeric form to disintegrate when exposed to certain chemical agents, e.g., dodecyl sulfate, 66% acetic acid, and dilute alkali. A distinction is often made between agents whose injurious effects on cell equilibria are reversible and those whose effects are irreversible. Reversible effects produce stasis or inhibition without immediate lethal action, while irreversible effects cause fairly prompt death of the cell.
There is no sharp distinction between bacteriostatic and bacteriocidal effects: the difference is quantitative rather than qualitative.
Nevertheless, the terms "bacteriostatic" and "bacterio- If the enzyme is essential, growth will cease.
However, some enzymes are not essential; that is, suitable alternate substrates may be supplied or some other mechanism may be available to form the required product. The normal reaction is depicted in B; where the enzyme and substrate interact to form a product or products and the enzyme is released for further activity. At A, an inhibitor that is chemically related to the normal substrate reacts with the enzyme, thus blocking the enzyme from reacting with the usual substrate. At C, a different inhibitor reacts with the normal substrate and forms a compound that cannot react with the enzyme. '° --J 98 Studies were initiated to elucidate the mechanism of action of phenol-type disinfectants (7)(8)(9). It is the molecular structure of these acids which would influence the ease with which they penetrate the cell, the part of the cell they attack, or the chemical nature of their attack (10).
The possible mode of action of benzoic acid has been postulated to be membrane action and competition with co-enzymes (11).  (3); for aniline from the work of Hulme (4); for phenol from the work of Giller-Pandraud ( 5 ) and Scheringer (6); for para-hydroxybenzoic acid from the work of Manjlovic (7); and for para-aminobenzoic acid from the work by Alleaume (8)  VSIP's were used to generate values for coulomb integrals, which are charge depe n dent and which, in turn, wer e used to obtain resonance integrals. However, since a given set of input VSIP's leads to MO's which, from a Mulliken population analysis, produce a charge distribution different from that assumed in the original evalua-1 0 1 tion of VSIP values, these input VSIP's had to be adjusted for the new charge distribution; and the calculation had to be repeated until the calculated charge distribution coincided within a preset limit to that used to evaluate VSIP's.
The VSIP's were obtained from the work of Cusacks (10), and Basch, Viste and Gray (11,12) and are listed in Table 27.  A recent question regarding MO-calculated conformations has been raised as to whether drug molecules treated in such a manner can be used to map complementary features of a receptor. One working hypothesis, which has been used in MO studies, is that these molecules engage the receptor in their preferred conforma- tions.
An alternative hypothesis is that the receptor is capable of a significant perturbing influence on a drug molecule within its vicinity, and that a nonpreferred conformation engages the receptor. This is not in violation of the concept that optimum stereochemistry maximizes efficacy, since the perturbed conformation would arise from an initial presentation of the drug molecule in its preferred conformation.
In this study, Iterated Extended Ruckel Theory The HOMO and LEMO basis set orb i tal coefficients for p a ra-Hydroxybenzoic a c id and co n former a re given in Ta ble 32 ; for Methyl p a ra-Hydr oxyb e nzo a te and conformers in Table 33 ; and for Ethyl para-Hydroxybenzoate and conformers in Tables 34 and 35 .
Resonance f orms , which can lead to molecular s t abiliza tion , can be shown to exist in the various com- • From the above it can be seen that para-hydroxy and para-aminobenzoic acid, as well as their esters, form more stable resonance structures than do benzoic acid, phenol or aniline. Hence, we should see that molecular conformers which tend to retain these resonance structures would be more stable than conformers which restrict resonance. However, the relationship of MO theor~ X-ray crystallography, solution and biological conformers must be kept in mind as the analysis of the data progresses.
The analysis which follows will use a charge/ HOMO/LEMO/configurational energy approach to study effects of various substituents and conformers, both by trends and by discontinuities. Conformational energy: Computer program did not converge.        LEMO: This is a pi-type orbital which is partly bonding and partly non-bonding; the energy is -8.60 eV.
Adding a carboxyl group planar to the ring has the effect of changing the character of the HOMO and its energy. The carboxyl group draws charge from the hydroxyl group.
Comparing benzoic acid (planar) (Figure 20) with para-hydroxybenzoic acid ( Figure 23) the carboxyl group again draws charge from the hydroxy group.
The HOMO and LEMO are similar to those of paraaminobenzoic acid.
As seen with para-aminobenzoic acid ( Figure 19) The perpendicular-under-the-ring configuration is 4 eV more stable than the perpendicular form.
The effect of the ethyl group as seen in Ethyl-parahydroxybenzoa te with the methyl ester and the parent acid: Ethyl-para-hydroxybenzoate in the hooked-planar position ( Figure 29) versus the linear-and-planar configuration ( Figure 32).
There is no significant charge redistribution. The planar-and-linear conformation is 4 eV more stable than the perpendicular-and-under-the-ring -and-down position.
The analysis above can be summarized by discussing the compounds as they relate to each other by a common physical parameter.

Energies
The most stable conformations are exhibited in those compounds which have a planar-and-linear configuration; while the next most stable conformation in the ester series is the perpendicular dangling chain (under-the ring-and-down).
A change in energy is seen in the Total Energy and the LEMO of the compounds but not in their HOMO's. This occurs in the methyl ester, parahydroxybenzoic acid and benzoic acid itself.

151
A conclusion which can be drawn, then, is that the molecule will tend to resonance stabilize if given a chance. This is evident throughout the whole series and is primarily reflected in the decreased energy of the LEMO.
It can be seen that the amino and hydroxyl groups in the para-position act in the same way and are therefore comparable.
There is no significant evidence that intramolecular hydrogen bonding exists or plays an important role in molecular conf igurational stabilization.

Charge Distribution
It is evident that, when a para-substituent is form, namely that of the planar, linear form. Moreover, a biologically active substance which exerts a specific action for its class usually contains a moiety or moieties common to all members of the drug class, which is responsible for its action. Hence, if one were to inspect the structure of the parabens, it would soon become apparent that if the stable forms were perturbed into some common · configuration, then they would be as depicted by Figure 25 for the methyl ester and Figure 32 for the ethyl ester.
There is a great similarity in their HOMO and LEMO energies as well.
The planar moieties, which are most stable, have an ELEMO of approxi- Para-hydroxybenzoic acid would not be as lipid-soluble as the methyl ester and may not readily be at such a site. Phenol and benzoic acid, while they may engage in the same mechanism of action do not have the same functional groups to act in precisely the same manner as the later members of the series.
The activity of these compounds should increase as the lipid portion of the molecule, the ester chain, becomes larger, since it may act as an anchor to retain the molecule at the site for a longer period of time.
If sufficient sites are filled, function may be interrupted, causing the observed biological effect. In fact, biological activity does indeed increase as the ester chain increases. The only drawback to the use of these compounds is their aqueous solubility, which diminishes rapidly as the ester chain length increases. It has been found that the most effective preservative is benzyl parahydroxybenzoate; however, its aqueous solubility is so low that effective concentrations cannot be obtained. Oleagenous solutions of benzyl para-hydroxybenzoate have been used effectively as a scabicide against lice and mites.
Pharmaceutical systems, however, are for the most part aqueous systems and, therefore, the problem of solubility in water is extremely important.
Enzyme systems within a biological environment are capable of splitting esters by the use of esterases, such that the parent acid as well as the ester form may be present. Since the parent acid is similar in character to the para-aminobenzoic acid, it is also safe to presume that either and/or both compounds would compete for reaction systems in which one or the other is involved.
Therefore, an additional mechanism to be assumed is inhibition of folic acid formation. Since biological systems are capable of converting compounds from inactive forms to biologically active ones, then it is possible to justify the conversion of para-hydroxybenzoic acid to the biologically active para-aminobenzoic acid.
Thus, when Boesken and Waterman (7) demonstrated that molds were capable of utilizing para-hydroxybenzoic acid, they may have in fact been witnessing a secondary shunt within that system to produce the biological material needed. This possibility was also shown by Davis (8) when he reported that para-hydroxybenzoic acid was a bacterial "vitamin". Since time is a factor for the development of a secondary shunt, however, we should discount this factor as the primary mechanism for preservative action.
Recently there has emerged an interest in the union of MO-calculated indices with thermodynamic properties, e.g., parameters reflecting partitioning between the lipid and aqueous phases in the body. Hansch (9) pioneered the technique, and the union of the two methods has been reported by several investigators (10).
This approach can be fruitful only if the deficiencies in MO methods are minimized and if the methods are wisely employed. The greatest danger lies in the oversimplification of the relationship between the two methods.
Molecular orbital theory can advance our understanding of the chemical events associated with molecules of biological significance; however, prudence should be observed in the application of this research technique. One must bear in mind that a biological process does not occur in a conservative state, nor does it occur in a reaction vessel; rather, it occurs in a complex medium which can influence and alter events to a degree beyond the reach of presently available predictive methods.
The greatest danger in the use of MO theory to study drug phenomena is the temptation to carry a correlation too far into the realm of mechanism prediction or receptor characterization. Therefore, limits must be placed on conclusions from MO studies so that the relationship between the method and the predicted result is always apparent. Nevertheless, the future opportunities for MO theory as a research technique are such that many scientists should be encouraged to familiarize themselves with this method.

CONCLUSION
The solubility of a related series of solutes namely benzoic acid, para-hydroxybenzoic acid and the methyl, ethyl, propyl and butyl esters of para-hydroxybenzoic acid, in a series of n-alkanols over a small temperature range, illustrated the usual polarity effects. A twin peak solubility array was evidenced for the above compounds in ethanol or propanol and hexanol.
Thermodynamic elements for the solution process in the above solutes were determined and tabulated, utilizing temperature as an experimental parameter. Both the direction and the magnitude of these elements were in keeping with expectations, in fact, the ideal values were seen to be a balance between two factors, entropy and enthalpy, which either oppose or support each other relative to their own magnitudes. The twin peak array observed in the solubility spectrum is seen throughout the thermodynamic picture of the investigational compounds with the following generalizations to be made about the solution process: (1) Any solute in solution whose actual free energy approaches the ideal free energy will approach the ideal state; (2) For any particular solution there may be many combinations of values for the heat of solution and the entropy of solution, depending upon the solvent employed, in which the actual free energy will approximate the ideal free energy. Therefore, it can be said that a solute which is soluble in a few or many solvents may approach ideal solubility in one or several of these solvents and thus exhibit one or several maxima or peaks in its solu-bility scan. The dielectric requirement of one peak is now felt to be related to the thermodynamics of the total system; hence, a multi-peak system should signify that ideal solubility is being approximated in that particular solvent-solute system.
Partitioning data were correlated with past work, notably that of Hansch.
The results indicate that partition coefficients are a useful tool and they should only be used with other substantial experimental techniques before definitive conclusions may be drawn. studied. This analysis can be summarized by a common physical parameter as it relates to the compounds studied.

Energies
The most stable conformations are exhibited in those compounds which have a planar-and-linear conf iguration; while the next most stable conformation in the ester series is the perpendicular dangling chain (under-the-ring-and-down). A change in the energy is seen in the Total Energy and the LEMO of the compounds but not in their HOMO's. A conclusion which can be drawn is that the molecules will tend to resonance stabilize if given a chance. This is evident throughout the whole series and is primarily reflected in the decreased energy of the LEMO. It was seen that the amino and hydroxyl groups in the para-position act as electron donors and are therefore comparable. There is no significant evidence that intramolecular hydrogen bonding exists or plays an important role in molecular configurational stabilization.

Charge Distribution
When a para-substituent is not present there is no tendency for resonance stabilization and therefore there is no real charge redistribution. The resonance stabilization as seen in the energies is not reflected in the charge trends on the carbonyl oxygen nor on the ester group as a whole. There does not seem to be a real depository for the charge removed from the para-hydroxy group, and it appears to be evenly delocalized over the entire molecule.
HOMO and LEMO Character planar and non-planar HOMO: In all cases the orbital is lone pair in character on the carbonyl oxygen.
LEMO: Planar compounds-the orbital is a delocalized pi-type, mainly non-bonding in character.
Non-planar-the charge is localized on the carboxyl group, and the orbital is of a non-bonding carbonyl pi-type.
This analysis leads to the hypothesis of a receptor site for the proposed biologically active form (Perpendicular-and-under-the-ring-and-down conformation). The receptor site should consist of a lipid area or areas in which the alkyl groups of varying lengths for the ester chain are solubilized, sites of positive or negative charge resulting from ionic or ion-dipole interac t ions with protein functional groups, and relatively uncharged areas composed of hydrocarbon fragments. Evidence is presented for the substantiation of this hypothesis as well as an alternative mechanism which would be time dependent. c D 1.