Laser desorption mass spectrometry of nonvolatiles under shock wave

---

Anal. Chem. 1985, 57, 895-899

095

Laser Desorption Mass Spectrometry of Nonvolatlles under Shock Wave Conditions Buko Lindner and Ulrich Seydel* Forschungsinstitut Borstel, Parkallee 1-40, 0-2061 Borstel, Federal Republic of Germany

For transmission-type laser desorption mass spectrometers, as the commercially available laser microprobe mass analyzer LAMMA 500, Y is shown that a sultable combinatlon of sample thlckness and laser power denslty allows the reproduclbie selection of a defined desorptlon/lonlzatlon mode. The Irra0 at high laser power dendlance of thick samples ( ~ 2 pm) slties (=lb" W cm-'), which does not lead to a perforation, offers extremely soft lonlzation conditions. For these condltions a model is descrlbed for the desorption mechanlsm of molecules from organlc sollds based on a nonthermal, shock wave drlven process leadlng to the release of mainly Intact molecules from the soild sample surface.

Very soon after the laser had become available, the possibility of focusing its radiation to very small spot sizes and of achieving high power densities in short time intervalsyielding rapid heating of small sample volumes-called forth the interest of mass spectroscopists to use it as an desorption/ionization source. In the past 20 years, and with increasing tendency in the last 5 years, a large number of examples for the applicability of the laser desorption mass spectrometry LDMS-especially for the analysis of organic compounds-have been published (for review see ref 1 and 2). Mainly the construction of a modified laser focusing system by Hillenkamp et al. (3)brought considerable progress to the development of a laser microprobe mass analyzer, which became commercially available as LAMMA 500 (Leybold-Heraeus, Koln, FRG). Very soon the wide scope of this system-ranging from the sensitive tracing of elements and even organic molecules in biological specimens to the pure mass spectrometric application for the analysis of complex biomolecules-could be demonstrated in a large number of excellent papers (for review see ref 4 and 5). Considering the relatively long history of LDMS, it seems striking that there exists no unified model which would explain satisfactorily the physical processes leading to the desorption of intact molecules and fragments-or ions thereof-and with that of the structure of the mass spectra obtained from these particles. However, considering the diversity of instrumentation and experimental design meanwhile used in laser desorption mass spectrometry ranging in laser wavelengths from 249 nm to 10.6 pm, in laser power density from 20 to 10l2W cm-2,and in laser spot diameter a t the sample surface from less than 1 pm to 4 mm (for review see Hillenkamp (6),the described situation becomes understandable, especially when taking furthermore into account the different laser-sample-detector geometries (transmission- and reflectance-type instruments). Even with the same instrument, the variation of just one parameter like the irradiance or the sample thickness may require different models for the description of the steps involved in ion formation, as will be discussed below. A question of major concern is whether the desorption process requires high temperatures at the sample surface or not. In other words, does the mechanism proceed thermally, that means here under direct influence of the temperature in the desorption region,

or may other nonthermal mechanisms also lead to desorption. But keeping in mind the above outlined situation concerning the diversity of instrumentation and sample handling, it is not to be expected that there will by the one model describing all experimental conditions. Kistemaker (7-lo), Rollgen (11, 12), Cotter (13-17), and Heresch (18) favor a thermal mode for their experimental conditions: sample on bulk substrates and irradiation with moderately focused laser beams at moderate laser power densities at the sample side. However, as is pointed out in some cases (11,18,19),thermal processes cannot explain sufficiently all observed phenomena. Hillenkamp distinguishes between two different processes occurring with different sample geometries (20). For bulk solids, examined in a way described above, he assumes a thermal mechanism. For thin films analyzed in a transmission-type instrument he suggests nonequilibrium processes. This might be backed by the measured energy spreads of the produced ions of up to 50 eV (21). Comparable values (up to 25 eV) have also been reported by Hardin and Vestal (22). Hercules proposed several processes occurring concurrently a t different distances from the laser spot (5). We suggested a nonthermal shock-wave driven desorption process (23,24) to explain our findings that for a transmission-type instrument best results with respect to simplicity and reproducibility of spectra from complex organic compounds could be obtained at highest available laser power density ( = l o l l W cm-2) and relatively thick samples (=20 pm) not being perforated by the laser beam. In this paper we try to corroborate these first findings experimentally. For this purpose we used various oligosaccharide samples to get information on the thermal stress exerted on the desorbed molecules. Oligosaccharides are known to be thermally labile-stachyose more than raffinose and sucrose-and have been used repeatedly for similar applications (7-12, 18, 25). We found direct evidence for the existence of at least two different desorption modes, thermal and nonthermal, depending on laser power density and sample thickness, which may even coexist under particular conditions. The successful application of the described soft desorption/ionization conditions is demonstrated for some oligo/ polysaccharides containing up to 12 sugar units.

EXPERIMENTAL SECTION Laser Desorption Mass Spectrometer. The instrument used was the transmission-type laser microprobe mass analyzer LAMMA 500 (Leybold-Heraeus, Koln, FRG). The frequency quadrupled light of a Q-switched Nd-YAG laser (wavelength, 265 nm; pulse duration, 10 ns) is focused onto the front surface of the solid sample with a spot diameter down to approximately 1 Frn at maximal power densities of 10'l W cm-2 (adjustable over 3 orders of magnitude by neutral filters and laser amplifier setting). The ions produced on the opposite side (rear surface) of the freely suspended sample are registered by a TOF mass analyzer. The signals from the ion detector (SEM) were digitized by a Biomation 8100 transient recorder at a sampling time of 50 ns and transferred to a HP 1000 computer. A schematic of the experimental setup is given in Figure 1. Samples. Sucrose, raffinose, stachyose, and y-cyclodextrin were purchased from Sigma Chemicals (Munchen, FRG), the

0003-2700/85/0357-0895$01.50/00 1985 American Chemical Society

896

ANALYTICAL CHEMISTRY, VOL. 57, NO. 4, APRIL 1985

enlargement of A qlJUrtZ COVQf g/OSS Sample support grid

ion extraction laser irradiation

Incident Illumination

I

Meter

I

I

Computer Figure 1. Schematic of the LAMMA 500 instrument.

highly purified octa- and dodecasaccharides isolated from bacteria (Shigella flexneri) (26) were a generous gift from A. Lindberg (University Hospital, Huddinge, Sweden). Samples were prepared from aqueous solutions containing the sugar and NaI or KI at a molar ratio of approximately 5:l. Different amounts of these solutions were brought onto Formvar-coated copper grids (mesh width 100 wm) and dried to give thin layers homogeneous in thickness and consistency over the distance of several meshes. This way layers of different thickness could be achieved. The so prepared samples were mounted inside the TOF analyzer with the supporting grid facing the laser (see Figure 1, inlay). Measurements. From sucrose, raffinose, and stachyose, layers of thicknesses d, = 1pm and d z = 20 wm were mass analyzed at the laser irradiances of p1 = los W cm-2 and p z = 10l1W cm-2 (32X objective, diameter of the spot of direct laser-sample interaction =3 wm). To account for the limited reproducibility of the fragmentation patterns (mainly caused by variations in laser performance and focusing) as well as of total ion intensities of the spectra produced by each single laser pulse, 25 single spectra were registered and averaged for each combination of sample thickness and laser power density. With the sampling time of 50 ns the quasi-molecular peaks could only be registered if a suitable delay time was chosen. For this reason the spectra do not comprise the low mass region, which in each case showed mainly high cation signals.

RESULTS AND DISCUSSION Figure 2 gives the averaged positive ion LD mass spectra of stachyose obtained under different experimental conditions. Clearly, under the aspect of getting low-level fragment intensity, i.e., the "softest" desorption/ionization, the most favorable conditions are realized in spectrum 2a obtained from a thick sample layer at highest laser power density and without perforating the sample. The spectrum comprises only the abundant quasi-molecular peak [M + Na]' a t m / z 689 and one low-intensity fragment peak ClsH32016NaCat m / z 527, originating from unimolecular decomposition by cleavage of

l

'

300

400

600 r n l z

300

400

600 m l z 800

, 300

400

I

,

'

I

800

7--1

600 m / z

600

Figure 2. Positive ion LD mass spectra of a mixture of stachyose and

NaI (molar ratio ca. 5:l)-each spectrum representing the average of 25 single spectra-obtained for different sample thicknesses d and laser power densities p : (a) d = 20 pm, p = 10" W cm-', no perforation of sample; (b) d = 1 pm, p 10' W cm-', perforation of sample; (c) d = 1 Mum, p = 10'' w cmP, perforation of sample. a glycosidic bond and hydrogen transfer. It should be pointed out that similar spectra were observed under these conditions also when the laser beam was focused onto the supporting Cu grid, resulting in an indirect energy input.

ANALYTICAL CHEMISTRY, VOL. 57, NO. 4, APRIL 1985 250 mV

-

iM t N a l * 521

897

1

a

i

'

0--'

mV-

I

2 00

300

b

365 203

Flgure 3. Positive ion LD mass spectra of a mixture of raffinose and NaI (molar ratio ca. 5: 1)-each spectrum representing the average of 25 single spectra-obtained for different sample thicknesses d and laser power densities p : (a) d = 20 pm, p = IO" W om-', no perforation of sample; (b) d = 1 pm, p = I O 8 W cm-', perforation of sample.

200

mV

500

mlz

I

I

Spectrum 2b demonstrates the other extreme-low laser power density and a thin sample layer-showing a high level m/z of fragmentation with peaks at m / z 527 (C18H32016Na+), 365 (C12H22011Na+), and m / z 203 (C6H1206Na+), which again stem from glycosidic bond ruptures. The quasi-molecular peak, in this case, is reduced nearly to zero intensity, and additional fragment peaks can be observed beside intensive NaI-cluster peaks. Spectrum 2c (Figure 2c) resembles the pattern of spectrum 2b in respect to the intensive fragments at m/z 203,365, and 527 and the NaI cluster peaks. However, the intensity of the quasi-molecular peak at m / z 689 is higher. This spectrum seems to represent an overlap of the physical processes responsible for the patterns shown in Figure 2a,b. The application of low laser power density to thick sample layers did not lead to a detectable ion yield. Only in some cases weak cation signals were registered. A similar influence of the experimental conditions on the fragmentation patterns as described for stachyose were observed also for raffinose and sucrose (Figure 3, Figure4). Beside the fragment ion peak a t m / z 365 both spectra of raffinose (Figure 3) exhibit additional intense mass peaks at mlz 379 which could be shown to originate from ion formation by alkali attachment. (The admixture of CsI instead of NaI led to a corresponding shift of these peaks by 110 mass units). Presently we cannot offer a verified explanation of the underlying ion building processes. This is particularly difficult because impurities of the raffinose could be definitely excluded (applying extensive chemical analytical techniques), and for the other sugars respective peaks-if a t all-were observed only at negligible intensities. The obviously little influence of the experimental conditions on the patterns of the sucrose spectra (Figure 4)-even for a thin sample excited by low laser irradiance (Figure 4b) an intense quasi-molecular peak is observed-may be explained by the comparably high thermal stability. Figure 5 shows the average of 45 positive ion LD mass spectra of y-cyclodextrin obtained under the same experimental conditions as those in Figures 2a, 3a, and 4a. This spectrum does not comprise any fragment ion peaks. If spectra of thermally labile substances like the investigated sugar do not exhibit considerable fragmentation, a direct influence of temperature can be excluded. This holds for the spectra depicted in Figures 2a, 3a, and 4a which were obtained from thick sample layers irfadiated at the highest laser power density without producing perforations. The desorption/

A I

I

300

500

mlz

365

-

1

I

200

300

I

l

l

500

mlz

Figure 4. Positive ion LO mass spectra of a mixture of sucrose and NaI (mdar ratio ca. 5:l)-each spectrum representing the average of 25 single spectra-obtained for different sample thicknesses d and laser power densities, p : (a) d = 20 pm, p = IO" W cm-', no perforation of sample; (b) d == 1 pm, p = 10' W cm-', perforation of sample; (c) d = I pm, p = 10'~ w cm-', perforation of sample.

,oo - ,,hO,

~

voj* -~-______----

[ M . NO:' 1320

rnV

0,

,

- L - l - - _-u u - A . L L . r y ~ ~ r _ d 7----7---77

200

400

600

800

1000

cT

I

miz

1400

Flgure 5. Positive ion LD mass spectrum of a mixture of y-cyciodextrin and NaI (molar ratio ca. 5:1), representing the average of 45 single spectra (sample thickness d = 20 pm, laser power density p = 10" w cm-', no perforation of sample).

ionization must proceed at a locus spatially separated from the locus of direct laser-sample interaction. Since there is experimental and theoretical evidence that the impact of a laser beam as used in our experiments on a solid sample leads to a temperature rise of up to some thousand degrees at the surface within the laser pulse duration (8, 27-30), yielding heating rates of approximately 10l1K s-l for a laser pulse width of 10 ns. A direct influence of this high temperature would inevitably lead to drastic fragmentation. In other words, the desorption process as depicted in Figures 2a, 3a, and 4a must be nonthermal. We have therefore postulated a process triggered by a laser-driven shock wave (23, 24). The high heating rate of =lo" K s-l within the thin surface layer of direct laser-sample interaction well exceeds the limit of approximately lo9 K s-l above which an explosive vaporization (''phase explosion") should occur (31), which may be accompanied by plasma generation. This thermal ablation produces a shock wave which traverses the solid under energy dissi-

898

ANALYTICAL CHEMISTRY, VOL. 57, NO. 4, APRIL 1985

600

800

1000

12W

16W

rniz

ZOW

Figure 6. Positive ion LD mass spectrum of a mixture of approximately equal molar amounts of stachyose M,, an octasaccharide M, and a dodecasaccharide M, (both isolated from the bacterial species Shigella flexneri (30)and of K I (sample thickness d = 20 pm, laser power density p = 10'' W no perforation of sample).

pation. The principles of the thermodynamics governing this vaporization are described in detail for metals in ref 32. I t must be emphasized, however, that quantitative calculations for the organic substances under investigation can hardly be done a t present because of the lack of thermophysical data. At the rear surface the shock wave leads to the desorption of intact molecules (and possibly fragments) and of alkali ions via vibrational disturbance of the binding potentials. The latter stem from impurities or from alkali salts added to enhance the quasi-molecular ion yield. In experiments with relatively thick metal foils and organic layers (d i= 30 pm) which were covered on the back side with alkali salts, it could be shown that the shock wave mode also leads to the desorption of alkali ions from a solid, not perforated sample. Gas-phase reactions between the neutral molecules and the codesorbed alkali ions should lead to the formation of these quasi-molecular ions (18, 33-35). In contrast to the spectra produced by shock-wave-driven non-thermal desorption, the patterns of the spectra in Figures 2b, 3b, and 4b, which were obtained from thin samples with the laser beam perforating the layer at low irradiance, show considerable fragmentation. A possible explanation for this fragknentation is the thermal stress which is exerted on the molecules being desorbed at the locus of direct laser-sample interaction. However, other processes like fast collisional dissociation, which might put additional stress on the molecules (36), cannot be fully excluded. The spectra described so far represent the extremes in the experimental conditions: thick sample layers, high laser power density and thin sample layers low laser power density, respectively. A combination of these parameters (thin sample layer, high laser power density) obviously leads to patterns comprising features of both extremes (Figure 2c). From this it may be concluded that under such conditions a combination of the above described desorption modes may occur with varying contributions to the overall ion yield. It seems reasonable to assume that the two processes occur within the same time scale but are laterally separated. The shock-wave-drivendesorption mode from organic solids is favorable, especially for molecular weight determination. This is clearly demonstrated with y-cyclodextrin in Figure 5, which does not show any fragmentation in contrast to the results of Cotter who observed a sequence of fragmentation peaks (37)-besides the quasi-molecular peak-under his bulk analysis conditions. Such a high degree of fragmentation was not obtained in our experiments even under the conditions of high thermal stress. Another example for this application is given in Figure 6. Here, estimations of the molecular weight, based on biochemical data, of a dodecasaccharide isolated from a bacterial species (S.flexneri) were to be confirmed. For internal mass scale calibration the unknown compound was mixed with approximately equal molar amounts of stachyose and an octasaccharide of S. jlexneri of already known structure and with KI. The spectrum shows a very simple pattern with intense quasi-molecular ion peaks of the three sugars at m/z 705,1340, and 1982 and two peaks each 104 mass units lower than the

respective quasi-molecular ions also showing alkali attachment. At the moment it cannot be decided whether these peaks are due to laser-induced fragmentation or, more likely, originate from the chemical pretreatment. After statements of Denoyer et al. ( 4 ) Heinen (19),Cotter (37),and Simons (38),soft desorption/ionization should be achieved for organic solids from thin samples under as low a laser irradiance as possible for observation of ions. The above examples, however, may serve as an experimental proof that under the controlled conditions of high laser power density and thick homogeneous sample layers in a transition geometry a very soft desorption/ionization is realized.

ACKNOWLEDGMENT We thank F. W. Rollgen, University of Bonn, FRG, for many stimulating and fruitful discussions. We also gratefully acknowledge the skillful technical assistance of H. Luthje. Registry No. NaI, 7681-82-5; KI, 7681-11-0; stachyose, 47055-3; raffinose, 512-69-6; sucrose, 57-50-1; y-cyclodextrin, 17465-86-0.

LITERATURE CITED (1) Conzemius, R. J.; Capelien, J. M. I n t . J . Mass Spectrom. Ion Phys. 1980, 3 4 , 197. (2) Conzemius, R. J.; Simons, D. S.;Byrd, G. D. "Microbeam Analysis"; Gooley, R., Ed.; San Francisco Press: San Francisco, CA, 1983; p 301. (3) Hillenkamp, F.; Kaufman, R.; Nitsche, R.; Unsold, E. Appl. Phys. 1975, 8 , 381. (4) Denoyer, E.; Van Grieken, R.; Adams, F.; Natusch, D. F. S. Anal. Chem. 1982, 5 4 , 26. (5) Hercules, D. M.; Day, R . J.; Balasanumgam, K.; Dang, P. A,; Li, C. P. Anal. Chem 1982, 5 4 , 280. (6) Hillenkamp, F. "Ion Formation from Organic Solids"; Benninghoven, A., Ed.; Springer-Verlag: Berlin, Heidelberg, New York, Tokyo, 1983; Springer Series In Chemical Physics, Vol. 25, p 190. (7) Kistemaker, P. G.; van der Peyl, G. J. Q.; Haverkamp, J. "Soft Ionization Biological Mass Spectrometry"; Morris, H. R., Ed.; Heyden & Son: London, Philadelphia, Rheine, 1981; p 120. (8) Van der Peyl, G. J. 0.;Haverkamp, J.; Kistemaker, P. G. I n t . J . Mass Specfrom. Ion Phys. 1982, 4 2 , 125. (9) Van der Peyl, G. J. Q.; Isa, K. Haverkamp, J.; Kistemaker, P. G. I n t . J . Mass Spectrom. Ion Phys. 1983, 47, 11. (10) Van der Peyl, G. J. Q.; van der Zonde, W. J.; Bederski, K.; Boerboom, A. J. H.; Kistemaker, P. G. I n t . J . Mass Spectrom. Ion Phys. 1983, 4 7 , 7. (11) Stoli, R.; Rollgen, F. W. Org. Mass Spectrom. 1979, 14, 642. (12) Stoll, R.; Rollgen, F. W. Org. Mass Spectrom. 1981, 16, 72. (13) Cotter, R . J. Anal. Chem. 1980, 5 2 , 1767. (14) Cotter, R. J. Anal. Chem. 1981, 5 3 , 719. (15) Van Breemen, R. 8.; Snow, M.; Cotter, R. J. Inf. J . Mass Spectrom. Ion Phys 1983, 49 35. (16) Cotter, R. J; Tabet, J.-C. I n t . J . Mass Spectrom. Ion Phys. 1983, 5 3 , 151. (17) Tabet, J.-C.; Cotter, R. J. Anal. Chem. 1984, 5 6 , 1662. (18) Heresch, F. "Ion Formation from Organic Solids"; Benninghoven, A,, Ed.; Springer-Verlag: Berlin, Heidelberg, New York, Tokyo, 1983; Spinger Series in Chemical Physics, Vol. 25, p 217. (19) Heinen, H. J. I n t . J . Mass Specfrom. Ion Phys. 1981, 3 8 , 309. (20) Hillenkamp, F. Int. J . mass Spectrom. Ion Phys. 1983, 4 5 , 305. (21) Nitsche, R.; Kaufmann, R.; Hillenkamp, F.; Unsold, E; Vogt, H.; Wechsung, R. 1st'. J . Chem. 1978, 17, 181. (22) Hardin, E. D.; Vestal, M. L. Anal. Chem. 1981, 5 3 , 1492. (23) Seydel, U.; Lindner, B. "Ion Formation from Organic Solids"; Benninghoven, A., Ed.; Springer-Verlag: Berlin, Heidelberg, New York. Tokyo, 1983; Springer Series in Chemical Physics, Vol. 25, p 240. (24) Lindner, B.; Seydel, U. "Secondary Ion Mass Spectrometry"; Benninghoven, A., et al., Eds.; Spinger-Verlag: Berlin, Heidelberg, New York, Tokyo, 1984; Springer Series in Chemical Physics, Vol. 36, p 370. 125) . . Heinen. H. J.: Meier. S.:Voat, H.: Wechsuna, R. Adv. Mass SDectrom. 1980, 8 , 342. (26) Carlin, N. I.A.; Lindbera, A. A,; Bock, K.; Bundle, D. R. Eur. J . Biochem. 1984, 139, 189. (27) Ready, J. F. "Effects of High Power Irradiation"; Academic Press: New York, 1971. (28) Burns, A. R. SOC. Photo-Opt. Instrum. Eng. Proc. 1983, 380, 224. (29) Magili, J.; Bloem, J.; Ohse, R. W. J . Chem. Phys. 1982, 76, 622. (30) Cotter, R. J.; Tabet, J.-C. A m . Biotechnol. Lab. 1984, March, 10. (31) Martynyuk, M. M. Sov. Phys.-Tech. Phys. (Engl. Trans/.)1976, 2 1 , 430. (32) Seydel, U.; Fucke, F.; Wadle, H. "Bestimmung thermophysikalischer Daten flussiger hochschmelzender Metalle mit schnellen Pulsaufheizexperimenten"; Verlag P. Mannhold: Dusseldorf, 1980. (33) Van der Peyl, G. J. Q.; Isa, K.; Haverkamp. J.; Kistemaker. P. G. Nucl. Insfrum. Methods 1982, 198, 125. (34) Stoll, R.; Rollgen, F. W. 2.Naturforscl],. 1982, 37A, 9. (35) Cotter, R. J.; Snow, M.; Colvin, M. Ion Formation from Organic Solids"; Benninghoven, A., Ed.; Springer-Verlag: Berlin, Heidelberg,

.

.

I

Anal. Chem. 1985, 57,899-903 New York, Tokyo, 1983;Springer Series in Chemical Physics, Vol. 25, p 206. (36) . . Rolloen. F. W.. Drivate communication and Proceedinas of the 2nd L A i M A Workshop, D-2061 Borstel, FRG, 1983,p 57. (37) Cotter, R. J. Anal. Chem. 1984, 56, 485. (38) Simons, D. S. “Ion Formation from Organic Solids”; Benninghoven,

-

899

A., Ed.; Springer-Verlag: Berlin, Heidelberg, New York, Tokyo, 1983; Springer Series in Chemical Physics, Vol. 25, p 158.

RECEIVED for review September 17,1984. Accepted December 20, 1984.

Reliability Ranking and Scaling Improvements to the Probability Based Matching System for Unknown Mass Spectra Barbara L. Atwater, Douglas B. Stauffer, and Fred W. McLafferty* Chemistry Department, Cornell University, Ithaca, New York 14853

David W. Peterson Scientific Instrument Division, Hewlett-Packard, 1501 California Avenue, Palo Alto, California 94304

Statistical evaluations of the effects of flve matching parameters on the probablllty of retrieving a correct answer wlth the probability based matching (PBM) system have been made. Comblnlng the resulting values found In matching an unknown spectrum makes It posslbie to rank retrleved reference spectra according to the predicted match rellabllity. Thls ranklng substantlally improves the performance of PBM, and the reliability value is especially helpful In avoldlng the assumption that the best matching spectrum represents the correct compound when Its spectrum Is actually not in the reference flle. Quadratic scaling of the abundance values of the unknown compensates for spectral dlfferences caused by Instrumental varlations, a critlcal problem In matching reference spectra. Other improvements include a more effectlve “flagglng” technlque to remove spurious reference peaks. Extensive appllcatlons wlth a commercial GCIMS system have demonstrated the increased effectiveness made possible by these PBM modifications.

Thousands of gas chromatograph/mass spectrometers (GCIMS) are now used daily worldwide (1). A major application is the identification of unknown compounds, which in many laboratories results in the production of hundreds of unknown mass spectra per day, making obvious the need for computerized identification systems (2-16). For samples representing complex mixtures, incomplete GC separation is unavoidable (17,18); for the resulting spectra which represent more than one component reverse searching (only requiring the peaks of the reference to be in the unknown) improves retrieval performance (4-7). By far the most widely used retrieval algorithm of this type appears to be probability based matching (PBM) ( 4 , 7). Although other search systems (3, 10-12) are valuable, when evaluated under various conditions (7,10,12) none appears clearly superior to PBM. In the last decade PBM has been used extensively by individual implementation (161, through computer networks (Cornell Computer Services, Uris Hall, Ithaca, NY 14853) (19), and on a commercial GC/MS systems, resulting in a variety of helpful criticisms. A major problem to many users, which appears to be common to all retrieval algorithms, is that the match ranking

factors such as the “similarity index” (3)or “confidence ( K ) value” (4,7) give only a qualitative indication of the probability that the retrieved compound represents a correct answer. For example, although a K value of 150 implies a higher match confidence than a value of 100, it does not directly indicate whether the probability of a correct identification is 50% or 95%. This deficiency can cause a particularly serious problem when the unknown compound is not represented in the reference file, as almost always the algorithm retrieves a “best”, even though a poor, match. We show here that the effect of a variety of matching indicators can be evaluated statistically, so that a combination of these can serve as a quantitative measure of the predicted reliability of the match. A second serious problem for mass spectral matching systems utilizing a comprehensive reference file is the variation in peak abundances caused by mass discrimination and change in sample concentration during the spectrum scan. As suggested independently by Dromey (8), various methods of tilting and scaling the unknown spectrum to compensate for such spectral differences are investigated here. The PBM algorithm has also been modified to improve the “peak flagging” which discards anomalous peaks in the reference spectrum.

EXPERIMENTAL SECTION Computers used include a DEC PDP-11/45 containing 56 kilobyte memory and 64 megabyte random-access disk storage, an IBM 370/168 multiuser system, and the H/P-1000 computer of the H/P 5985 GC/MS system. The data base was the expanded Registry of Mass Spectral Data (ElectronicData Div., Wiley, 605 Third Ave., New York, NY 10158) containing 41 429 different spectra of 32 403 different compounds, from which 2091 isotopically labeled spectra were excluded. From those compounds in the file represented by more than one spectrum (measured under other experimental conditions) 900 were selected at random, with the restriction that all spectra of the compound must have a quality index (QI)2 0.5 (20). For each of these compounds the spectrum of highest QI value was used to make up the list of unknown spectra, which were excluded from the data base in testing. Every odd-numbered spectrum of this list was used to make up a second “odd” list of 450. The performances of these two lists and of the PBM program versions were evaluated by using recall/reliability plots (21,22),which show the proportion of correct answers which are retrieved as a function of the proportion of retrieved answers which are correct. For the “odd” list poor PBM retrievals were used to correct obvious errors and delete

0003-2700/85/0357-0899$01.50/00 1985 American Chemical Society