International Journal of Mass Spectrometry and Ion Processes, 89 (1989) 265-285 Elsevier Science Publishers B.V., Amsterdam - Printed in The Netherlands

265

CHARGE SEPARATION MASS SPECTROMETRY 1. DOURLY-IONIZED PERFLUORO COMF’OUNDS

J.H.D. ELAND, L.A. COLES and H. BOUNTRA

Physical Chemistry Laboratory, South Parks Road, Oxford OXI 3&Z (Gt. Britain)

(First received 10 October 1988; in final form 5 December 1988)

ABSTRACT

Perfluoro compounds have been used as model compounds to investigate the charge separation mass spectra of polyatomic molecules using a triple coincidence photoionization technique. The two-dimensional m, - m2 spectra contain detailed information on the dou- bly-charged ion breakdown pathways, and show the importance of both primary and secondary loss of neutral fragments. The major characteristics of the reactions are rational- ized by a simple model, which can also be used to predict the characteristics of charge separation in other classes of compound and following triple or higher ionization.

INTRODUCTION

Doubly-charged ions are generally rather minor, though extremely inter- esting, contributors to the mass spectra of small to medium sized molecules. Their low abundance is due partly to the lower probability of double ionization rather than single ionization, but mainly to the propensity of small doubly-charged species to dissociate by charge separation into the singly-charged ion pairs [l] of high initial kinetic energy, whose origin was thus explained correctly in the early days of mass spectrometry [2,3]. The major process of fast cation-pair formation following photoionization of small molecules has been studied recently using the double coincidence technique PIPICO [4,5] (photoion-photoion coincidence), in which corre- lated ion pairs are accumulated as a function of the difference of their flight times to a single detector in a time-of-flight mass spectrometer. The PIPICO technique is excellent for measuring energy released in the fast charge separation reactions of diatomics and other very small molecules, but because different pairs of product ions may have similar time-of-flight differences (proportional to rn\” - m:12), and the peaks are greatly broad- ened by the kinetic energy release, the technique is unsuitable for large molecules. In conventional mass spectrometers of high resolution only slow

0168-1176/89/$03.50 0 1989 Elsevier Science Publishers B.V.

266

charge separation reactions can be seen as metastable peaks [6], or collision- induced charge exchange can be used to study neutral fragment loss leading to “stable” doubly-charged fragments [7].

We have recently introduced a triple coincidence technique, photoelec- tron-photoion-photoion coincidence (PEPIPICO) [8] which provides more detailed information on the dynamics of charge separation in small doubly- charged ions [9]. It also offers much better mass separation than PIPICO, giving the full resolution of the TOF spectrometer used, and so opens the way to study large molecular ions. Because the resolution of our current TOF spectrometer is only about 100 even for thermal ions, however, it is still not possible to resolve individual mass peaks from large hydrogenic mole- cules. We have therefore chosen perfluoro derivatives as model compounds for this first investigation of fast, unimolecular charge separation reactions following double ionization of polyatomic molecules.

The conventional electron-impact mass spectra of many perfluoro com- pounds have been examined [lO,ll] and are available in standard compila- tions, though no comprehensive survey seems to be available. Some of them show evidence of “isolated state behaviour” rather than complete energy pooling [12,13]. The question of “statistical” vs. “isolated state” behaviour probably needs special consideration for doubly-charged species because coulombic repulsion in addition to internal excitation can only lead to faster dissociation, which may make the energy randomization incomplete.

EXPERIMENTAL

In our technique [8] molecules in an effusive jet are ionized either by wavelength-selected vacuum ultraviolet photons (hv 12-50 ev), or by high energy electrons, admitted continuously to the ionization region. The elec- tron impact technique, which closely resembles the pioneering experiment of McCulloh et al. [14], will be the subject of a future publication [15]. The occurrence of an ionization event is signalled by the detection of a low-en- ergy electron at the channel electron multiplier (Fig. 1) with a few nano- seconds’ delay. The electron signal serves as the start pulse for a multihit time-to-digital converter, which registers the arrival times of up to eight ions detected within some 20 pus of the electron. Ion arrivals unrelated to the electron signal build up to a flat false-coincidence background in the TOF mass spectra, while correlated ion arrivals accumulate as peaks. Events in which one, two or three ions are registered after a single start are recorded separately.

The light source used in the present work was a low pressure magnetically confined discharge in helium, from which the 30.4 nm resonance line of He(I1) or other lines were selected by a grazing incidence toroidal grating

267

TDC Amplifier and

p\ 4

CF discriminator stop

-2ooom start

Fig. 1. Scheme of the PEPIPICO apparatus, not to scale. The drift region is 16-cm long, while the source and acceleration regions are l-cm and OS-cm long, respectively. Ionization is at the centre of the source region. The TDC (time to digital converter) is an eight-hit unit (Le Croy 4208).

monochromator (Jobin-Yvon LHT 30). The TOF mass spectrometer (Fig. 1) is a dual field Wiley and Maclaren design [16], with a resolution for thermal ions of about 100. For ions of high initial kinetic energy its munerica!l resolution (half height definition) R is a simple function of the source electric field E and the initial kinetic energy U

R = 1.4( E/Z+*

where E is in V cm-’ and U in eV. For most of the ions formed by charge separation from the fluorocarbons studied here the initial energy is about 1 eV. At the electric fields of 200-600 V cm-’ employed, resolution for these ions is thus only 35 at best. This is sufficient to resolve fluorine loss from the heaviest ion studied (m/z 615 from heptacosafluorotributylamine), though it does not, by itself, permit unambiguous distinction of fluorine and carbon numbers over the full mass range. In practice the centres of peaks, and hence masses, can be determined with an accuracy several times better than the resolution, and no ambiguity of interpretation remains.

The fluorocarbons used in this work were commercial samples, of ade- quate purity, but not necessarily isomerically pure.

RESULTS

Nature of the spectra obtained

The coincidence experiments produce two sets of data for each compound at each wavelength used. First there is the mass spectrum, dominated by

268

Cycle-C,F, 304 nm

20 40 60 80 100 120 140 m,lamu

Fig. 2. View of the tw~~ension~ PEPIPICO spectrum of pe~uor~yclobut~e at 30.4 nm on a linear mass scale. The largest peak represents the pair of ions of mass 31 and 100.

normal single ionization, and directly comparable with conventional elec- tron-impact spectra. It also includes “stable” doubly-charged ions, though these are not prominent for the fluorocarbons, except perfluoronaphthalene. Secondly, the triple coincidence spectrum is a two-dimensional spectrum, representing the intensity of correlated pairs of ions as a function of the masses (or times) of both ions. These spectra can be drawn on the printed page as two-dimensional representations of three-dimensional figures, as exemplified in Fig. 2. Although such figures are elegant and informative, they are difficult to interpret in qu~titative detail, and for clarity we use spectra projected from two to one dimension; detailed analysis requires listings of in~~dual ion pairs and their characteristics extracted by com- puter from the raw data.

The simplest one-dimensional spectrum projected from the two-dimen- sional data is the mass spectrum of (singly-charged) ions from charge separation of doubly ionized precursors. It is evident (Fig. 3b) that in such a spectrum the parent ion is absent, ions are concentrated at low mass, and the resolution is much reduced, particularly at high mass by the large kinetic energy releases. Nevertheless, most of the same characteristic ions as seen in the normal mass spectrum are present.

Next, spectra of apparent precursors, which we call “pair-sum spectra”, can be derived by adding together the masses of the two ions observed in each correlated pair. These spectra show a notable reappearance of high masses, often including the parent mass, and concentration of intensity into relatively few ions (Fig. 3~). Pair-sum spectra are very easily obtained from the raw data and have many valuable characteristics; nevertheless they are not the spectra of true precursors, as will be explained below.

269

Cycle-C,F, 304nm

120 160 21 jo masslamu

Fig. 3. One dimensional spectra from perfluorccyclobutane at 30.4 run: (a) photoionization mass spectrum of normal singly-charged ions; (b) spectrum of singly-charged ions seen as constituents of pairs; (c) spectrum of apparent doubly-charged precursors, or “pair-sum spectrum”, obtained by adding together masses of correlated ion pairs.

Thirdly, a useful form of spectrum, which cannot be put on a common scale with the others, is one of intensity as a function of (t, + t2), corre- sponding to (ml ‘I2 + WZ\,~) (Fig. 4). The advantage of this type of spectrum

Cycle-C,F, 30.4 nm

Fig. 4. Spectrum of (t, + tz) for perfluorocyclobutane at 30.4 run.

270

comes from the fact that where ion pairs are formed as the sole products of a doubly-charged precursor decay (two-product reaction)

f -+,+ 1 +d

the initial momenta of the two ions must be equal and opposite. The deviation of an individual ion from the nominal time-of-flight corresponding to its mass is directly proportional to the component of its initial momentum along the spectrometer axis. Therefore in the sum (t, + f2) the effect of initial kinetic energy release is cancelled out, and sharp peaks can be seen in this spectrum. There is partial cancellation even where three fragments, two charged and one neutral, are formed as in the common secondary fragmen- tations. As a result peaks are often separated in the (t, + t2) spectrum even when overlapped in other spectral projections, and this fact can be used as an aid to detailed analysis. Because the data are stored in digital form, (t, + t2) spectra can be drawn for specific groups of reactions, such as those including ions of one particular mass, further aiding the process of separat- ing and identifying distinct ion pairs.

Peak shapes and secondary ion dissociations

Two-dimensional spectra showing intensity as a function of t, and t, (or m, and m2) contain all the information available from the PEPIPICO experiment. However, much of the most interesting detail is not apparent unless the figures are drawn out on a very large scale. Each individual peak in the spectrum has a shape, which reflects both the initial kinetic energies of the two ions, and also the correlation between their initial momenta [9]. Most peaks are found (Fig. 5) to be more or less long and thin, with their long dimensions at various angles to the axes. A long, thin peak at an angle 8 to the t, axis means that the initial momenta pi of the ions rnc and ml observed are related by p2= -pl tan8

Several peaks, including all those that represent two-body dissociation of parent ions have B = 45 O, meaning that the ions’ momenta are equal and opposite, as required by momentum conservation.

Although more complex interpretations are possible, and have been detailed elsewhere [9], the normal meaning of angles other than 45” is that one or both of rnc and ml is formed by secondary decay from a heavier singly-charged primary fragment. The overall course of pair formation may be considered as a sequence of steps

P 2+

+m$+ +nO 0)

2+ m0 +m,+ +mt (2)

m,++m,++n, (3)

271

Cycle-CLF, 304nm m,lamu

3LL----J9 2 3 4

w

Fig. 5. Part of the PEPIPICO spectrum of perfluorocyclobutane at 30.4 nm viewed from above as an array of t, vs. t2. As a result of contrast enhancement for presentation, intense peaks are broader than weaker ones, but lengths and slopes are shown correctly.

The kinetic energy release in charge separation (step 2) is much larger than in neutral loss (1 or 3), and if the secondary fragmentation of the primary daughter ions (mc) occurs after separation to 1 nm or more from the other primary ion (i.e., after about 5 X lo-l3 s), the slope of the experimental peak for coincidences between m: and ml is simply tan 19 = - ml/m3 because the momenta p1 and p2 are equal and opposite while the daughter ion ml continues with the same velocity as rnc giving

P3=(m3/ml)P1* Other similar mass ratios apply if the other primary ion, or both primary

ions, undergo secondary fragmentation. Conversely, therefore, from mea- sured slopes we can deduce the mass ratios, and so identify the fragmenta- tion sequence. Once any neutral masses have been identified, the length of the line in the t, vs. t, spectrum also gives the magnitude of the kinetic energy release in the charge separation step. A computer can be used to examine each peak, to determine its slope and length, as well as the masses of the ions involved, and so allow a mechanism to be inferred for formation of the observed ion pairs. Part of such a Table, for charge separations of CF2C1Br2+ is shown as an example in Table 1. Kinetic energies are not given in this case because they are made unreliable by the presence of unresolved isotope contributions.

We find from the slopes of the t, vs. t, peaks that a large proportion of final ion pairs from charge separation actually arise from secondary reac- tions. It is evident therefore that the pair sum spectra which are derived simply by adding together the masses of ions detected in pairs are apparent,

272

TABLE 1

Observed ion pairs from CF,CIBr*+

Intensity Observed Observed Secondary reaction Calculated

1365 CF+ 963 CF,+ 881 CFCl+ 648 Br’ 404 CF+ 271 cl+ 251 cl+ 208 cl+

55 Cl+ ? cc1 +

ions

Br+

slope

-1.7 fO.l CF,+ --) CF+ +F Br+ -1.5 fO.l CF,Cl+ + CF,+ + Cl Br+ - 0.95 f 0.05 CF,Cl + - 0.98 f 0.05 cl+ -3.5 f0.4 CFBr+ --) CFf + Br CF; -0.4 f0.05 CF2Br+ + CF; + Br Br+ -0.7 fO.l CF2Br+ + Br+ + CF, CF,Br+ -0.8 f0.2 CFBr + -0.8 f0.2 [CF2Brf -) CFBr” +F Brf (not fully resolved)

slope

- 1.61 - 1.71 -1.0 - 1.0 -3.6 - 0.38 - 0.61 -1.0 - 0.851

rather than true precursor spectra. When secondary reactions have been taken into account by analyzing the slopes of the t, vs. t, peaks, “true” precursor spectra can be derived by adding together the masses of primary pairs; these show the masses of the doubly-charged intermediates from which charge separation actually occurs, and they exhibit very strong concentration on high masses. Unfortunately definitive analysis of this kind is not always possible, because there are sometimes ambiguities in the interpretation of the t, vs. t, slopes. Nevertheless, the raw slopes show directly whether ion pairs are primary or secondary, and identify various possible primary precursors, even if a definite assignment cannot be made, It has already been demonstrated that apparent kinetic energy releases from PIPICO peak widths may be falsified by secondary reactions [17], but that they can be derived correctly using the present technique [9]_

The richness of PEPIPICO data raises a severe problem of presentation. Even the raw two-dimensional spectra of correlated ion pairs can be shown in many different ways, and the substantial, though still partial information from peak slopes is almost impossible to represent graphically. Because peak heights in raw time-of-flight spectra depend upon widths from kinetic energy release and mass effects, they are not a suitable quantitative repre- sentation. For the majority of compounds considered here we therefore present formalized spectra, with heights proportional to measured areas. The two spectra shown for each compound are the photoionization mass spec- trum of singly-charged ions, and the pair-sum spectrum of (apparent) doubly-charged precursors. The identities of the correlated ion pairs which contribute to each pair-sum peak are shown as a formalized array of m, vs. (m, + m2), with intensities crudely encoded by spot size.

273

CF;+F

Fig. 6. Breakdown scheme for doubly-charged CF, 2+ formed by photoionization at 25.6 nm. The percentages give amounts of ions or pairs observed in the mass spectrum and PEPIPICO

spectra.

. Carbon tetrafluoride The doubly-ionized CF, molecule exemplifies much of the characteristic

behaviour found in larger species, but also exhibits some idiosyncrasies. Its second ionization potential is so high that 30.4 nm (40.8 eV) ionization results in very limited ion pair production: the breakdown pathways derived from spectra taken at 25.6 run (48.4 eV) are presented in Fig. 6.

Of CF$+ ions initially formed none at all survive to be detected after 2 p.s. The most intense decay pathway after 25.6 nm double ionization is loss of a neutral fluorine atom; of the CF:+ ions so produced a small percentage remain stable, while the majority decay further. The second most abundant decay route is direct formation of CF: + F+ pairs from the transitory CF:+, amounting to 33% of the total, while formation of CF:+ with ejection of two fluorine atoms (whether combined or not) is the third and least abundant pathway. There is some production of CF+ + F+ pairs (7% of total) which arises in part from the decay of CFi+ (though the majority of these ions are stable), while part must come from rapid secondary decay of CFC ions (from CFi+) within the coulomb field of their F+ ion partners. This is an example of an ambiguous peak slope, which cannot be interpreted uniquely.

CClBrF, The doubly-charged ion dissociation channels of this bromine- and chlo-

rine-substituted compound, listed in Table 1, resemble those of tetrafluoro- methane, with greater richness from the variety of atoms. Two substantial differences immediately stand out, however. First, there is no F+ ion as an ion pair partner in any product channel from double ionization, whereas both Cl+ and Br+ ions appear. This is evidently explained by the very high heat of formation of F+. Secondly, the majority of the initial CClBrFt+ ions decay with loss of a neutral fragment either immediately or in a secondary process, only 18% of observed pairs having masses which add up to the parent mass.

214

CF&lBr 304nm

~

0

Fig. 7. Formalized spectra of CF,ClBr at 30.4 nm: (a) photoionization mass spectrum; (b) pair-sum spectrum of apparent doubly-charged precursors, with the q vs. (ml + mz) array below to allow identification of the correlated ion pairs contributing to each pair-sum peak. The dot size roughly encodes intensity.

Both these characteristics are probably related to the lower ionization energies of almost all species other than an F atom. Thus in charge separation reactions ionic products lower in energy than F+ are always favoured, and for equal initial excitation energy in the doubly-charged parent these products will then have more internal excitation energy, leading to secondary reactions.

It is interesting that the energized species CF,Br+ is found to decay with about equal probability to CFC + Br and to CF, + Br+. By an extension of Stevenson’s rule, this would indicate near equality of the ionization poten- tials of CF, radical and the bromine atom, in agreement with the known values (11.64 eV and 11.81 eV, respectively).

For comparison with the other species, formalized spectra of CF,ClBr are presented in Fig. 7.

Perfluoropropyl iodide Both the photoionization mass spectrum (Fig. 8a) and the ion pair mass

spectrum of C,F,I (not shown) are free from F+ ion, and have CF,+ (m/z

275

C ,F,I 304nm

151

m,lamu

101

m,+m,lomu

Fig. 8. Formalized spectra of C,F,I at 30.4 nm, as Fig. 7. An estimate of the significance of the unresolved process (Table 1) has been incorporated.

69) as base peak. The pair sum spectrum (Fig. 8b) shows a substantial shift to higher mass numbers compared with the normal mass spectrum and contains peaks at the parent mass (p), (p - F), (p - F2), (p - CF,) and ( p - CF,). The parent peak in the pair-sum spectrum, representing pairs formed with no neutral fragment formation, has less than 1% of the total intensity, however. The vast majority of charge separation reactions (93%) involve either simultaneous or prior loss of a neutral particle, which may be F, F2, CF, or I in order of decreasing abundance. Most of the primary singly-charged ions ( - 85%) undergo secondary dissociation, irrespective of whether neutral loss has or has not already occurred in (or before) the charge separation step, so formation of many observed pairs involves ejection of two separate neutral fragments.

Large open-chain saturated perfuoro compounds As molecular size increases mass spectra of open chain perfluorocom-

pounds become ever more strongly dominated by CFT, and the pair

276

12c m,lamu

9(

60

3c

0

n-&F,, 30.4nm

I ! .,!.I! . ...,‘.. ,, ,.I,.

(b)

Fig. 9. Formalized spectra of perfluoro-n-pentane, as Fig. 7. Other isomers may also have been present in small abundance in the sample used.

reactions are dominated by CF,+ + anything+. The fraction of pair forma- tion reactions not involving loss of neutrals drops to zero, and almost all primary singly-charged fragments undergo secondary decay. The exceptional reactions not involving secondary decay, and leading to the highest mass peaks in the pair sum spectrum, are those in which charge separation leads to two large fragment ions, rather than one large and one small one. The pair sum spectra show a greater emphasis on high mass ions than the sample mass spectra, the average mass of intense peaks being higher by a factor of about two. Thus the mass spectrum of perfluoropentane (Fig. 9a) has its centre of gravity around mass 80, while the pair sum spectrum (Fig. 9b) is centred around about 180. For perfluorohexane where the spectra (not shown) are closely similar to those of perfluoropentane, the corresponding masses are about 90 and 200, while for perfluorotributylamine (“heptacosa”, Fig. 10) they are about 130 and 230. Because of the increasing tendency to secondary breakdown and consequent domination of the mass spectra of the heavier molecules by low-mass ions, these numbers do not increase in direct proportion to the molecular weight, but slightly less rapidly.

277

Fig. 10. Focalized spectra of heptacos~uor~t~bu~l~e ~p~fluorot~butyl~ne) at 30.4 nm, as Fig. 7.

CycbC,F, 30.4nm:True precursors

Fig. 11. Formalized spectra of cycle-C,F, at 30.4 nm showing spectra and primary corresponding to “true” precursors, after allowance for secondary reactions.

ion pairs

278

Saturated ring compounds The raw mass spectrum and pair data for perfluorocyclobutane have been

shown in Figs. 2-5; analysis of the peak slopes allows us to determine a complete breakdown scheme for the doubly charged ions of this molecule, from which formal spectra of “ true” precursors, shown in Fig. 11, are derived. By ~rnp~son with the spectra of raw pairs (Fig. 2) and apparent precursors (Fig. 3~) this shows much more concentration of intensity into a few doubly-charged precursors of high mass. The parent molecule mass is more strongly represented even in the apparent precursor spectrum than in corresponding spectra of open chain molecules of similar size (Figs. 8 and 9), presumably because of the extra stability conferred by the ring structure. Even so, 90% of the doubly charged parent ions split up into three, rather than two fragments at this wavelength, and about 85% of all primary fragments undergo secondary dissociation. When ionization is brought about at the longer wavelength of 35.6 rrm (34.8 ev), however, there is very much less fragmentation, and the parent mass becomes the base peak of the pair-sum spectrum.

C,oF,s 304nm 1

200 m,iamu

150

Fig. 12. Formalized spectra of perfluorodecalin at 30.4 mn from an isomeridy mixed sample, as Fig. 7.

279

The pair-sum spectrum and pair identities for perfluorodecalin charge separation (Fig. 12b) at 30.4 nm are extremely rich, and concentrate at an average precursor mass much higher than the average mass in the photoioni- zation mass spectrum. Nevertheless, the highest apparent precursor mass (393) is less than the highest mass in the singly-charged ion spectrum, where the parent ion (462) is seen weakly. If even shorter wavelength light (25.6 nm, 48.4 eV) is used this lack of high-mass apparent precursors is strongly enhanced; the reason, to be discussed below, is evidently the tendency towards neutral fragment emission at high excitation energies.

Aromatic perfuoro compounds The stability of both doubly and singly-charged ions from perfluoroben-

zene and perfluoronaphthalene dominates the mass spectra and charge separation spectra, Figs. 13 and 14. In hexafluorobenzene 36% and in decafluoronaphthalene 35% of initial pair separations involve no concom- itant neutral loss, and of primary fragments only 63% from C,F, and a smaller percentage from C,,F, undergo secondary fragmentation. From

80 m,lamu

6t

Fig. 13. Formalized spectra of perfluorobenzene ionized at 30.4 in. The mass spectrum

may include some GFi+, which cannot be distinguished from C,F: at mass 93.

C,F, 30.4nm

(al

m,+m,famu

(4

280

C,,F, 304nm

150

m,lamu

100

I .

m,+m,lomu

Fig. 14. Formalized spectra of perfluoronaphthalene at 30.4 nm. The peak at mass 136 in the mass spectrum is probably dominated by C,,,Fi+.

perfluoronaphthalene, doubly-charged parent ions and some doubly-charged fragments are detected as stable species. The charge-separation reactions in both compounds are dominated by CF+ + anything+, with a very few pair separations not of this type for C6Ft’. The pair-sum spectra have parent ions as base peak in both cases. The neutral fragments ejected before or during the initial charge separation steps are mainly F’ atoms, F, or CF moieties.

DISCUSSION

Perfluoro compoundr

The perfluoro compounds are well known as marker molecules giving mass spectral peaks at widely separated intervals in electron impact instru- ments, though seldom including the parent mass [ll]. The 30.4~nm photo- ionization mass spectra reported here recapitulate this behaviour showing weak or absent parent ions from saturated compounds and charge sep-

281

aration characteristics which vary systematically according to the degree of “ unsaturation”, from the open chain compounds through the saturated cyclic compounds derived from cyclobutane and decalin to the aromatic naphthalene and benzene derivatives. The two-dimensional charge-sep- aration spectra change along this sequence of unsaturation, being dominated by CF,” + anything+ in the open chain compounds and by CF+ + anything+ in the aromatics, while the saturated ring compounds show richer patterns involving wider choice including both light ion partners. When the observed ion pairs are combined in the spectra of apparent precursors there is a strong shift to higher masses as compared with the normal mass spectra. The average mass of the apparent doubly-charged precursors becomes a smaller fraction of the full molecular mass as the size of the molecules increases; however, for molecules of comparable size the fraction is smallest for open chain compounds, intermediate for saturated rings and largest for aromatic compounds. Doubly-charged parent ions show up as apparent precursors only from aromatics and very small saturated compounds; there is some tendency for the highest mass precursor to be ( p - 2F)‘+, in contrast to the ion (p - F)+ in the normal mass spectrum. These characteristics are, how- ever, dependent on the wavelength of light used for ionization, as there is more fragmentation at shorter wavelengths, and less at longer wavelengths. At the fixed wavelength mainly used here the excess energy available for transfer to the molecular doubly-charged ion, and to its fragmentation products, increases as the ionization potential decreases, and therefore it increases with increasing molecular size in each series.

The identities of the low mass species observed both as singly-charged product ions and as neutral fragments from dication decay can be under- stood to some extent in terms of their heats of formation. The heats of formation of the neutral species F, F2, CF, CF,, CF, and CF, of 79, 0, 255, - 182, -470 and -933 kJ mol-’ respectively [18] favour the heavier fragments, though not dramatically. In fact in primary fragmentation of doubly-charged species the most common neutral ejecta are F, 2F/F,, CF, and CF, while secondary dissociations of singly-charged primary fragments most often eject F, CF and CF,. The most abundant singly-charged ions, whether as constituents of correlated pairs or not are the characteristic species CF+ (31), CF;t (69), C,Fl (100) and C,F,+ (131), which must be of special thermodynamic or kinetic stability. The heats of formation [18] of CF+ (1149 kJ mol-l), CFj’ (421 kJ mol-‘) and C,FT (321 kJ mol-‘) are certainly significantly less than that of F+ (1758 kJ molll), but are hardly enough by themselves to explain the exceptional abundances of these ions.

The overwhelming majority of observed ion pairs from all the perfluoro compounds are accompanied by neutral fragments, lost either in the initial stage of formation of a doubly-charged precursor, or in secondary fragmen-

282

tation of one or both of the singly-charged ions from the first charge separation reaction. We cannot distinguish between simultaneous break- down of a doubly-charged parent into two ions plus a neutral fragment, and sequential breakdown by loss of a neutral followed by charge separation of the doubly-charged ion that remains [9]. Because of this ambiguity there is always some uncertainty over the masses of “true” precursors; if decay is by simultaneous breakdown the true precursor is always the doubly-charged parent. Despite this, it is absolutely clear that in the vast majority of cases no doubly charged ions of the same mass as the charge-separation pre- cursors are found in the mass spectrum. This means that the precursors whose existence we infer are intermediates with lifetimes shorter than a few hundred ns at most.

General application and theory

The perfluoro compounds are notorious for having weak or undetectable parent ion peaks, and for possibly showing non-statistical behaviour in conventional mass spectrometry. Nevertheless, we believe that some char- acteristics of their charge separation spectra will also apply to a wide range of compounds. The most noticeable of these characteristics is the impor- tance of neutral fragment ejection. Even among the very small compounds hitherto studied by PIPICO, the prevalence of three-body dissociation (charge separations involving a neutral fragment) has been remarked upon [5]. Its origin can be understood qualitatively in terms of the large amount of excitation energy deposited in ions during double ionization.

The minimum energy transfer required to doubly ionize a closed-shell neutral molecule is given by the rule of thumb [19] as approximately 2.81, where I is the first ionization potential in eV. The energy of the products, taken as an ion pair, can be estimated as D + 2 I’ where D is the energy of the bond(s) to be broken, and I’ is the ionization potential of neutral radical fragment, assumed to be the same for all. If we assume, in accor- dance with the Wannier cross-section law for double ionization [20] that the minimum energy transfer in double ionization is also the most probable energy transfer, then the probable, or at least the minimum excess internal energy of a cation pair can then be expressed approximately as

E* = 2.81- (D + 21’) - KE

where KE is the kinetic energy release due to coulomb repulsion. Its magnitude decreases as the species becomes larger, and amounts to no more than 2 or 3 eV for species larger than triatomic. For fluorocarbons, I can reasonably be estimated as 12 eV, I’ as 10 eV, D as 4 eV, and KE as 2 eV. The excess internal energy to be disposed of then amounts to 7.6 eV, which

283

is more than enough, if held within one fragment, to break another bond. If a neutral fragment is not ejected at once, a statistical redistribution of this amount of excess energy between two ions will usually result in secondary dissociation, unless the two ions are of nearly equal size and so can take half the excess energy each. If higher excited states of the doubly charged ion are populated, rather than just the ground state, the higher internal energy content is certain to entail more extensive bond breaking.

This simple excess energy model explains the main features of perfluoro compound charge separation. The less disruptive fragmentation of saturated ring compounds follows from the increased number of bonds to be broken in forming a cation pair, while the relative stability of the aromatic com- pounds is explained both by the number of bonds to be broken and the low ionization potentials of the parent compounds. The difference between fluorocarbons and the related molecules containing heavy atoms, Br or I, is a consequence of the low ionization potentials of both parent and of one fragment. In accordance with the model, the only cation pairs formed without simultaneous or later loss of a neutral are those in which both ions are complex molecular species, able to contain considerable internal energy. Primary pairs with one light or structureless partner always undergo subse- quent neutral loss from the heavier partner.

The model may also be extended to other classes of compounds, and to triple or higher ionization. The energy required for triple ionization, for instance, may be estimated as 5.61 [20], so the excess energy to be shared between three cations is

E* = 5.61- (20+ 31’) - KE

If the kinetic energy release is estimated at 10 eV, this gives 19.2 eV excess energy among three ions, predicting universal neutral ejection accompanying charge separation. This is indeed observed, as will be reported elsewhere [15]. For double ionization of hydrogenic molecules there is a wide range of ionization potentials according to the degree of unsaturation. For the aromatic hydrocarbons, for instance, which have also been studied exten- sively by the present technique [21], the first ionization potentials can be estimated as 8 + 1 eV, while those of the fragments are likely to be at least as high. Since two or more bonds must be broken to form a pair the excess energy is estimated as - 3.6 +. 5 eV, which is effectively zero. This is clearly in accordance with the observation that the aromatic hydrocarbon dipositive ions are stable in their low-lying electronic states, and seldom suffer sec- ondary fragmentation after primary charge separation. It follows that charge separation occurs only from excited states of the doubly-charged ions of these compounds.

284

The general model proposed above, and also our explanation of the identities of prominent ions and neutral species involved in fragmentation, are based upon assumptions of competitive reactions with choice of pathway decided mainly on energetic grounds. Such assumptions, which seem to work, are the foundations of a statistical model; our observations suggest that the doubly-charged ion reactions of perfluoro compounds may largely be described by such a model.

CONCLUSIONS

Triple coincidence spectroscopy of charge separation reactions allows us, for each dissociation:

(1) to observe two products of each dissociation directly; (2) to deduce the existence of secondary reactions occurring as fast

majority processes within lo-’ s. Slow reactions, with characteristic lifetimes around 10V7 s, which give rise

to “metastable” signatures in coincidence spectra [8] as well as normal mass spectra, are always a small minority. By comparison with conventional mass spectrometry the PEPIPICO technique offers much more detailed informa- tion on major dissociative pathways of both singly-charged and doubly- charged ions.

The spectra produced by the new technique may be a useful tool for structure determination or identification, because pair-sum spectra, and even more the “true precursor” spectra, concentrate intensity into higher mass and more “structurally significant” peaks than normal mass spectra. Fur- thermore, for each peak in such a precursor spectrum the identities of one or more pairs of fragments into which the precursor breaks up are known, which must be even more structurally informative. The mass spectrum is always measured concurrently with the ion pair spectrum, so that the new information would complement and extend conventional mass spectral interpretation, rather than replacing it. The recent development of high resolution time-of-flight mass spectrometers [22], in which triple coincidence measurements can be made concurrently with the measurement of the mass spectrum, makes this a practical possibility.

A major observation made in applying this technique to the perfluoro compounds is the prevalence of neutral fragment ejection both in secondary decays of primary fragments, and in primary doubly-charged ion dissocia- tions. One immediate consequence is that one-dimensional PIPICO peak widths cannot be relied upon for correct kinetic energy releases in any three-body dissociations of large doubly-charged ions, even where they can be satisfactorily mass resolved. According to our model, neutral fragment ejection occurs with great frequency because double ionization by vertical

285

transition from the neutral molecule produces doubly-charged parent ions with energies greatly in excess of the ion-pair dissociation limits. Even though it is thus clearly a “hard” ionization method, the detection of both members of the cation pair, with measurement of the correlations of their initial motions, gives the whole technique characteristics of “soft” ioni- zation. The pair-sum spectra, directly derived from the raw data, and especially the “ true precursor spectra”, derived with more difficulty, contain higher mass ions, and more structural information than normal mass spec- tra. The technique is as yet at an early stage of development, and further extensions and advances are to be expected.

REFERENCES

9 10 11 12 13 14 15 16 17 18

19 20

21 22

B. Brehm and G. de F&es, Adv. Mass Spectrom., 8 (1979) 138. P. Kush, A. Hustrulid and J.T. Tate, Phys. Rev., 52 (1937) 843. R. Fuchs and R. Taubert, Z. Naturforsch., Teil A, 19 (1964) 494. G. Dujardin, S. Leach, 0. Dutuit, P.M. Guyon and M. Richard-Viard, Chem. Phys., 88 (1984) 339. D.M. Curtis and J.H.D. Eland, Int. J: Mass Spectrom. Ion Processes, 63 (1985) 241. W. Higgins and K.J. Jennings, Chem. Commun., (1965) 99. J.H. Beynon, R.M. Caprioli and W.E. Baitinger, Org. Mass Spectrom., 3 (1970) 455. J.H.D. Eland, F.S. Wort and R.N. Royds, J. Electron Spectrosc. Relat. Phenom., 41 (1986) 297. J.H.D. Eland, Mol. Phys., 61 (1987) 725. F.L. Mohler, E.G. Bloom, J.H. Lengel and C.E. Wise, J. Am. Chem. Sot., 71 (1949) 337. H.M. Bibby and G. Carter, Trans. Faraday Sot., 59 (1963) 2455. C. Lifshitz and F.A. Long, J. Phys. Chem., 69 (1965) 3746. LG. Simm, C.J. Danby and J.H.D. Eland, Int. J. Mass Spectrom. Ion Phys., 14 (1974) 285. K.E. McCulloh, T.E. Sharp and H.M. Rosenstock, J. Chem. Phys., 42 (1965) 3501. D.A. Hagan and J.H.D. Eland, to be published. W.C. Wiley and I.H. Maclaren, Rev. Sci. Instrum., 26 (1955) 1150. G. Dujardin, L. Hellner, D. Winkoun and M.J. Besnard, Chem. Phys., 105 (1986) 291. M.W. Chase, Jr., C.A. Davies, J.R. Downey, Jr., D.J. Frurip, R.A. McDonald and A.N. Syverud, J. Phys. Chem. Ref. Data, 14 Suppl. 1 (1985). B.P. Tsai and J.H.D. Eland, Int. J. Mass Spectrom. Ion Phys., 36 (1980) 143. R.G. Kingston, M. Guilhaus, A.G. Brenton and J.H. Beynon, Org. Mass Spectrom., 20 (1985) 406. S. Leach, S.D. Price and J.H.D. Eland, to be published. R. Grix, R. Kutscher, G. Li, U. Grtiner and H. Wollnik, Rapid. Comm. Mass Spectrom., 2 (1988) 83.