Electrospray Ionization Efficiency Scale of Organic Compounds

Mar 10, 2010 - widely differing sensitivities of ESI-MS to different ana- lytes. ...... Estonia. We are indebted to Dr Frank Eckert from Cosmologic. G...
0 downloads 0 Views 389KB Size
Download PDF
Anal. Chem. 2010, 82, 2865–2872

Electrospray Ionization Efficiency Scale of Organic Compounds Merit Oss, Anneli Kruve, Koit Herodes, and Ivo Leito* University of Tartu, Institute of Chemistry, 14a Ravila street, 50411 Tartu, Estonia Ionization efficiency (IE) of different compounds in electrospray ionization (ESI) source differs widely, leading to widely differing sensitivities of ESI-MS to different analytes. An approach for quantifying ESI efficiencies (as logIE values) and setting up a self-consistent quantitative experimental ESI efficiency scale of organic compounds under predefined ionization conditions (ionization by monoprotonation) has been developed recently. Using this approach a logIE scale containing 62 compounds of different chemical nature and ranging for 6 orders of magnitude has been established. The scale is based on over 400 relative IE (∆logIE) measurements between more than 250 different pairs of compounds. To evaluate which molecular parameters contribute the most to the IE of a compound linear regression analysis logIE values and different molecular parameters were carried out. The two most influential parameters in predicting the IE in ESI source are the pKa and the molecular volume of the compound. This scale and the whole approach can be a tool for practicing liquid chromatographists and mass spectrometrists. It can be used in any mass-spectrometry laboratory and we encourage practitioners to characterize their analytes with the logIE values so that a broad knowledge base on electrospray ionization efficiencies of compounds would eventually develop. Electrospray ionization (ESI) is an extensively used ionization method in mass spectrometry (MS).1 Its relative simplicity, ease of coupling with liquid chromatograph (LC), and usability for a large number of analytes have made it the primary ion source for today’s LC-MS instruments. It is generally accepted that the ESI process proceeds via the ion evaporation mechanism in the case of low molecular weight compounds (see refs 2-4). Efficiency of ESI for generating ions is strongly dependent on spray conditions and analytes. Ionization of some analytes may be highly efficient (even 100% efficiency has been discussed in case of nano-ESI5), but other analytes are not ionizable at all. In most cases only a fraction of the analyte molecules (or ions) in the liquid phase that is sprayed into the ESI source are eventually converted to gas-phase ions (via * To whom correspondence should be addressed. Phone: +372 5 184 176. E-mail: [email protected]. (1) Kostiainen, R.; Kauppila, T. J. J. Chromatogr., A 2009, 1216, 685–699. (2) Enke, C. G. Anal. Chem. 1997, 69, 4885–4893. (3) Kebarle, P. J. Mass. Spectrom. 2000, 35, 804–817. (4) Chalcraft, K. R.; Lee, R.; Mills, C.; Britz-McKibbin, P. Anal. Chem. 2009, 81, 2506–2515. (5) Smith, R. D.; Shen, Y. F.; Tang, K. Q. Acc. Chem. Res. 2004, 37, 269–278. 10.1021/ac902856t  2010 American Chemical Society Published on Web 03/10/2010

protonation, adduct formation, deprotonation, etc.). The term ionization efficiency (IE) has been coined to denote the extent to which analyte molecules in liquid phase are converted to gasphase ions and eventually detected in detector. Pursuant to that, by IE we mean the efficiency of generating gas-phase ions from analyte molecules or ions in the ESI source plus efficiency of detecting them.6 In consideration of the wide usage of ESI, development in knowledge of predicting the ionization efficiency of a particular compound would be very useful. In qualitative terms it is known that among the molecular properties that affect ESI ionization efficiency are, for example, pKa value, hydrophobicity, surface activity, etc.1,7 Also, numerous quantitative studies on IE of molecules in the ESI source have been carried out.1,4,6,8-10 Several of them have set as an aim to investigate the dependence of the response in ESI-MS from the molecular parameters of the analytes. Besides fundamental interest and general usefulness in analytical method development highly practical applications for quantitative ESI IE data have been envisaged, such as calibration without chemical standards.4 The summary of the studies is presented in Table 1. The following observations can be made from Table 1: (1) A limited number of analytes, often structurally similar, are included in each study. The investigated compounds are often polyfunctional, which hinders relating ionization efficiency to molecular structure in a straightforward way. (2) In most of the studies the set of compounds investigated had a limited range of ionization efficiencies. (3) Ionization efficiency in the ESI source and its relation to molecular structure are dependent on the used mobile phase.1 For example, hydrophobicity of molecule (and thus the resulting ion) may be a very strong factor in the case of waterrich mobile phases, but can be of low importance in the case of purely organic mobile phases (e.g., refs 11 and 8). (4) Different (and often restricted) concentration ranges of analytes have been used. (5) The measurement methodologies and the measurands vary considerably. In some cases absolute MS responses are recorded; in other cases relative MS responses of two simultaneously infused compounds are recorded. (6) Leito, I.; Herodes, K.; Huopolainen, M.; Virro, K.; Ku ¨ nnapas, A.; Kruve, A.; Tanner, R. Rapid Commun. Mass Spectrom. 2008, 22, 379–384. (7) Cech, N. B.; Enke, C. G. Mass. Spec. Rev. 2001, 20, 362–387. (8) Ehrmann, B. M.; Henriksen, T.; Cech, N. B. J. Am. Soc. Mass Spectrom. 2008, 19, 719–728. (9) Kamel, A. M.; Brown, P. R.; Munson, B. Anal. Chem. 1999, 71, 5481– 5492. (10) Cech, N. B.; Enke, C. G. Anal. Chem. 2000, 72, 2717–2723. (11) Tang, L.; Kebarle, P. Anal. Chem. 1993, 65, 3654–3668. (12) Bo ¨kman, C. F.; Bylund, D.; Markides, K. E.; Sjo ¨berg, R. J. Am. Soc. Mass Spectrom. 2006, 17, 318–324.

Analytical Chemistry, Vol. 82, No. 7, April 1, 2010

2865

Table 1. Overview of the Ionization Efficiency Studies Reported in the Literature (Ionization by Mono-Protonation) compounds

concentration range 10-3-10-7 M

6 tripeptides

-3

-6

7 esters, 3 aromatic 10 -10 M amines 19 N-bases with different 10-4 M structures 46 compounds, most of them 10-2-10-3 M amino acids and N-bases 16 cations (metal cations, protonated alkaloids)

10-2-10-8 M

tetraalkylammonium salts

10-4 M

dependence founda

solvent, flow rate MeOH or H2O 0.5% CH3COOH, 5 µL/min MeCN:0.1% aq. HCOOH, 8.3 µL/min MeOH or 0.5% AcOH in MeOH, 10 µL/min MeOH: 0.1% aq AcOH, 10 µL/min

NPSA, surface activity hydro-phobicity, chelating ability basicity in the solvent MV, logP, (absolute ion mobility)

MeOH, 20 µL/min

surface activity, ion evaporation rate constant different methanol/water ion mobility relative to mixtures, 15 µL/min surface activity

dependence not found

basicity in the solvent gas-phase basicity

quantitative information log response, direct ESI-MS response logIE values

direct esi-ms response (absolute ion relative response mobility), effective factors charge ion intensity (I) the surface partitioning coefficient (KA/KE)

approx. spanb ref 2.5

10

10 000

6

300

8

70

4

10

11

35

12

a NPSA, non-polar surface area; MV, molecular volume. b The span of ionization efficiencies (MS responses at the same concentration level) of the compounds studied, expressed as the ratio of the highest and lowest MS response under identical conditions.

Observations 1, 3, 4, and 5 are most probably the reasons why the main conclusions of the papers differ markedly. Observations 1 and 2 do not allow drawing truly far-reaching conclusions on the relations of molecular structure and ionization efficiency (under given experimental conditions). Obviously it would be beneficial if some “standardization” could be introduced into ESI ionization efficiency studies. Aiming to bring more methodological uniformity into ESI ionization efficiency studies and to make the results better comparable, we recently introduced a concept of ionization efficiency scale based on measurements of relative ionization efficiencies (RIE).6 The approach is based on the model of Enke.2 The classical model of Tang and Kebarle11,13 leads to identical results. The recently published more comprehensive model by Du and White14 allows to take explicitly into account also the effect of the background electrolyte but under our experimental conditions again eventually leads to the same mathematical expressions. The IE of a compound is dependent on its structure and ionization state as well as on the solvent and numerous parameters of the ESI ion source and the mass spectrometer.1,6,15,16 Several of these parameters are difficult to control in a reproducible way and therefore the absolute ionization efficiency IE(Bi) ) Ri/Ci is very difficult to measure for each compound. We showed that in contrast to the absolute ionization efficiencies of compounds their relative ionization efficiencies can be measured quite easily.6 Furthermore, by combining the RIE values of pairs of compounds it was possible to establish an ionization efficiency scale. The RIE of a compound B1 relative to B2 is

RIE(B1 /B2) )

K′1 IE(B1) R1C2 ) ) IE(B2) K′2 R2C1

(1)

where IE are the individual ionization efficiencies of the analytes, R are responses of ions B1H+ and B2H+ in the mass spectra, C is the concentration of the neutrals B1 and B2, and K′ ) K · R, where R are the protonation ratios of the compounds B1 and B2: R1 ) [B1H+]/C1 and R2 ) [B2H+]/C2 and K are the partition (13) (14) (15) (16)

Tang, L.; Kebarle, P. Anal. Chem. 1991, 63, 2709–2715. Du, L.; White, R. L. J. Mass. Spectrom. 2009, 44, 222–229. Taylor, P. J. Clin. Biochem. 2005, 38, 328–334. Constantopoulos, T. L.; Jackson, G. S.; Enke, C. G. Anal. Chim. Acta 2000, 406, 37–52.

2866

Analytical Chemistry, Vol. 82, No. 7, April 1, 2010

coefficient of the ions B1H+ and B2H+. The values of RIE can be determined by electrospraying solutions containing two analytes at known concentrations and measuring the intensities of their ions in mass spectrum. The currently available IE scale (expressed as logRIE values)6 spans for 4 orders of magnitude but contains a limited number of compounds thus not permitting to make generalizations or predictions. The largest community of users of the results of ESI IE studies are analysts developing and using LC-ESI-MS-based analysis methods. Thus the mobile phase and experiment parameters were chosen as close as possible to the real-life LC-ESI-MS practice (see ref 6 in Table 1). The goal of this paper was 2-fold. First, we report the most comprehensive scale of ESI ionization efficiency available to date. The scale ranges for 6 orders of magnitude and contains 62 compounds from the families of esters, pyridines, amidines, guanidines, phosphazenes, aromatic, aliphatic, and heterocyclic amines, tetraalkylammonium salts, amides, and some others. Second, on the basis of these data we attempt to give some insight into the dependence of ionization efficiency on the properties of the compounds of widely differing chemical nature and ionization efficiency, under experimental conditions that are similar to ones used in mainstream LC-ESI-MS analysis. We briefly discuss the ionization efficiency trends in compound families and between families. Most of the included compounds are monofunctional, allowing to interpret the variations in ionization efficiency based on the structure differences. We restrict our approach only with positive ESI: compounds that ionize in ESI via protonation and compounds that are ions already in solution. EXPERIMENTAL SECTION Equipment and Materials. The measurements were carried out using an Agilent XCT ion trap mass spectrometer. The MS and ESI parameters were not changed or optimized but the factory defaults were used: nebulizer gas pressure 15 psi, drying gas flow rate 7 L/min, drying gas temperature 300 °C. Only the MS parameter target mass6 (TM) was modified in order to assess the discrimination during ion transport. All RIE measurements were carried out using three different TM values: M+1 of the first compound, M+1 of the second compound and m/z ratio 500. The logRIE value was found as average of the values obtained with

the three target masses or average of two target masses, if one of the results was significantly different. Origin of the studied compounds: diphenyl phthalate (Riedelde Hae¨n, Assay 99.9% by GC), dimethyl phthalate (Merck, assay g99% by GC), dimethyl glutarate (Merck, assay g99% by GC), dimethyl succinate (Merck, assay g98% by GC), dimethyl malonate (Merck, assay g99% by GC), methyl benzoate (Fluka, assay g99.5% by GC),ethyl benzoate (Aldrich, 98%), ethylamine hydrochloride (Aldrich, 98%), diethylamine (Fisher, reagent grade), trimethylamine hydrochloride (Aldrich, 98%), tripropylamine (Aldrich, 99+%), tributylamine (Sigma-Aldrich, puriss. plus; g99.5% (by GC)), triphenylamine (Reakhim, pure, recrystallized once from ethanol (96%)), tetrapropylammonium chloride (Fluka Analytical, purum; g98.0% (AT)), tetramethylammonium chloride (Fluka Analytical, puriss. p.a., for ion pair chromatography), methiocarb (Dr. Ehrenstorfer analytical standard), methomyl (Dr. Ehrenstorfer analytical standard), aldicarb (Dr. Ehrenstorfer analytical standard), 1-ethyl-3-methylimidazolium trifluoromethyl sulfonate (Merck, high purity), 1-hexyl-3-methylimidazolium bis(trifluoro sulfonyl)imide (Merck, high purity), 2,6-dimethypyridine (Aldrich, 98%), 2-chloropyridine (Aldrich, 99%), acridine (Fluka, Assay g97%(HPLC)), benzoic acid (Reakhim, pure), 2-cyanophenol (Aldrich, 99%), diphenylguanidine (Reakhim, pure), benzophenone (Reakhim, pure), benzamide (Reakhim, pure) tetraethylammonium perchlorate (Fluka, puriss.), tetrabutylammonium perchlorate (Fluka, g99.0%), tetrahexylammonium benzoate (Kodak), guanidine hydrochloride (Reakhim, pure for analysis), sulfanilamide (Carlo Erba, for microanalysis). The following compounds are the same as those used in refs 17-20: phenyl benzoate,17 piperidine,20 N-methylpiperidine,20 aniline,19 2-nitroaniline,19 3-nitroaniline,19 4-nitroaniline,19 4-fluoro-3-nitroaniline,19 4-chloro-2nitroaniline,19 2,4-dinitroaniline,18 N,N-dimethylaniline,19 1-naphtylamine,19 diphenylamine,19 benzylamine,19 triethylamine,19 pyrrolidine,19 N,N,N,N-tetramethylguanidine,19 phenyl- N,N,N,Ntertamethylguanidine,19 DBU,19 pyridine,19 2-methoxypyridine,19 2,6-dimethoxypyridine,19 3-chloropyridine,19 ((CH3)2N)3-P ) N-C6H519,(C4H8N)3-P)N-C6H3-2,5-Cl219,(C4H8N)3-P)N-C6H44-CF319, (C4H8N)3-P ) N-(C4H8N)2-P ) N-C6H4-2-Cl19, 2-methylpyridine,19 2,4,6-trimethylpyridine.19 Measurement Procedure. used in this study is same as in reference 6. The solvent composition acetonitrile (Chempure, assay g99.5 by GC)/0.1% aqueous formic acid (Riedel-de Hae¨n, 98-100% puriss. p.a.) in volume ratio 80:20 was used. For every RIE measurement solutions of two compounds in the solvent were made and these were infused in either of the two ways. The first way was using two syringe pumps connected with a “T”-piece with around 1 mm3 of dead volume (for mixing the solutions). The concentration ratio of the compounds can be easily varied by varying the ratio of infusion rates of the two pumps. The second way was premixing the solutions and infusing them using a single syringe pump. Ca 90% of the measurements were made using the first way. Both ways gave identical results. Concen(17) Nummert, V.; Travnikova, O.; Vahur, S.; Leito, I.; Piirsalu, M.; Ma¨emets, V.; Koppel, I.; Koppel, I. A. J. Phys. Org. Chem. 2006, 19, 654–663. (18) Rodima, T.; Ma¨emets, V.; Koppel, I. A. J. Chem. Soc. Perkin Trans. 1 2000, 2637–2644. (19) Kaljurand, I.; Ku ¨ tt, A.; Soova¨li, L.; Rodima, T.; Ma¨emets, V.; Leito, I.; Koppel, I. A. J. Org. Chem. 2005, 70, 1019–1028. (20) Ro ˜o ˜m, E.-I.; Ku ¨tt, A.; Kaljurand, I.; Koppel, I.; Leito, I.; Koppel, I. A.; Mishima, M.; Goto, K.; Miyahara, Y. Chem.sEur. J. 2007, 13, 7631–7643.

trations of compounds in the sprayed solutions were in the range of n · 10-8 to n · 10-3 mol/L depending on the two compounds and their ratio in the mixture. With all compounds the concentration in the spray was varied by at least a factor of 4 (leading to the variation of the concentration ratio of the two compounds by a factor of 16). The obtained values of logRIE were not markedly influenced by changing the concentrations. The RIE measurements were carried out at the overall solution flow rate of 8.3 µL/min (0.5 mL/h). The mass spectra were registered over a time period of ca. 100 s (ca. 250 spectra) and were averaged. The RIE values were found according to eq 1 and expressed as logRIE values. As it is clarified in ref 6 to verify the relative measurement results, every value of logRIE(B1, B2) is verified by involving at least one additional compound B3 and by determining the values of logRIE(B3, B1) and logRIE(B3, B2) (cross-validation). The IE dependence in the solvent flow rate was also studied. Some examples: the measurements of dimethyl glutarate versus dimethyl succinate at flow rates 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, and 1.2 mL/h gave logRIE values 0.74, 0.78, 0.71, 0.79, 0.70, 0.70, 0.72, and 0.68 respectively; the logRIE values of dimethyl malonate versus methyl benzoate found at flow rates 0.2, 0.3, 0.5, 0.7, and 1.2 mL/h were 1.58, 1.48, 1.61, 1.62, and 1.67 respectively; the logRIE values of dimethyl glutarate versus methyl benzoate at same flow rates were 2.84, 2.83, 2.68, 2.78, and 2.63 respectively. The standard deviation of these results was smaller than the standard deviation of results of 37 measurements (dimethyl succinate vs dimethyl glutarate) at flow rate 0.5 mL/h, thus conclude that solvent flow rate does not markedly influence the results. Computational Methods. Gas-Phase Basicities. (GB) (Gibbs’ free energy change on protonation) were calculated for compounds lacking experimental data. GB values refer to the following: ∆Gb

B + H+ f BH+

(2)

GB is defined as follows: GB ) -∆Gb

(3)

The quantum-chemical computations were carried out using Gaussian 200321 and Spartan 200822 series of programs. Density functional theory (DFT) calculations were performed using the B3LYP hybrid functional with the 6-311+G** basis set. Full geometry optimizations and vibrational analyses were performed (21) Frisch, M. J.; Trucks, G. W.; Schlegel, H. B. Scuseria, G. E.; Robb, M. A.; Cheeseman, J. R.; Montgomery, J. A., Jr.; Vreven, T.; Kudin, K. N.; Burant, J. C.; Millam, J. M.; Iyengar, S. S.; Tomasi, J.; Barone, V.; Mennucci, B.; Cossi, M.; Scalmani, G.; Rega, N.; Petersson, G. A.; Nakatsuji, H.; Hada, M.; Ehara, M.; Toyota, K.; Fukuda, R.; Hasegawa, J.; Ishida, M.; Nakajima, T.; Honda, Y.; Kitao, O.; Nakai, H.; Klene, M.; Li, X.; Knox, J. E.; Hratchian, H. P.; Cross, J. B.; Bakken, V.; Adamo, C.; Jaramillo, J.; Gomperts, R.; Stratmann, R. E.; Yazyev, O.; Austin, A. J.; Cammi, R.; Pomelli, C.; Ochterski, J. W.; Ayala, P. Y.; Morokuma, K.; Voth, G. A.; Salvador, P.; Dannenberg, J. J.; Zakrzewski, V. G.; Dapprich, S.; Daniels, A. D.; Strain, M. C.; Farkas, O.; Malick, D. K.; Rabuck, A. D.; Raghavachari, K.; Foresman, J. B.; Ortiz, J. V.; Cui, Q.; Baboul, A. G.; Clifford, S.; Cioslowski, J.; Stefanov, B. B.; Liu, G.; Liashenko, A.; Piskorz, P.; Komaromi, I.; Martin, R. L.; Fox, D. J.; Keith, T.; Al-Laham, M. A.; Peng, C. Y.; Nanayakkara, A.; Challacombe, M.; Gill, P. M. W.; Johnson, B.; Chen, W.; Wong, M. W.; Gonzalez, C.; Pople, J. A. Gaussian, Inc.: Wallingford, CT, 2004. (22) Spartan ‘08; Wavefunction, Inc.: Irvine, CA.

Analytical Chemistry, Vol. 82, No. 7, April 1, 2010

2867

with all molecules and ions. This computational level has been demonstrated to describe with reasonable accuracy the gas-phase basicities23 of a wide variety of relatively simple molecules. All stationary points were found to be true (Nimag ) 0). Unscaled B3LYP/6-311+G** frequencies were used to calculate the GB values taking into account the zero point frequencies, finite temperature 0-298 K correction, the pressure-volume work term, and the entropy term as appropriate. Molecular Properties. Like molecular area (A), molecular volume (MV), polar surface area (PSA), and dipole moment (δ) were computed using software Spartan 200822 at the same basis set as used in GB calculations. Distribution Coefficients. As a hydrophobicity characteristic of the protonated compounds, the foral distribution coefficient between the used solvent and hexane (logPsolvent/hexane) was used. No experimental values of such distribution coefficients are available, thus theoretical calculations, based on the COSMORS approach24 were used. Full DFT geometry optimization of the ions was done with the Turbomole software package25 using B-P density functional26,27 with TZVP quality basis set using the RI approximation.28 The COSMO-RS calculations were done using the COSMOtherm software.29 The solvent phase used was mixture of water with acetonitrile (with mole fraction 0.58) and the second phase was pure hexane. Volume quotient value VQ ) V1/V2 ) 0.286655 was used in logPsolvent/hexane calculations. pKa Values. in water and in acetonitrile for compounds lacking experimental data were calculated using the same COSMORS approach.24 Geometry optimization procedure was the same as with distribution coefficient calculations. Data Analysis. In order to find out which molecular properties of the compounds influence the ionization efficiency a number of different molecular properties were tested in a linear model. All of the parameters, including logIE, were scaled. For the molecular volume of the cation it was found that a logarithmic relation should be used in the model and therefore the molecular volumes were first logarithmed and thereafter scaled. All the statistical analyses were perfomed at confidence level 95%. RESULTS AND DISCUSSION LogIE Scale. Altogether 407 logRIE measurements were carried out with 62 compounds in more than 250 different compound pairs. Although the process of interest was limited to ionization via monoprotonation a number of side-processes were observed, (23) Koppel, I. A.; Schwesinger, R; Breuer, T.; Burk, P.; Herodes, K.; Koppel, I.; Leito, I.; Mishima, M. J. Phys. Chem. A 2001, 105, 9575–9586. (24) Klamt, A. COSMO-RS: From Quantum Chemistry to Fluid Phase Thermodynamics and Drug DesignI; Elsevier Science Ltd: Amsterdam, The Netherlands, 2005. (25) Ahlrichs, R.; Ba¨r, M.; Baron, H.-P.; Bauernschmitt, R.; Bo ¨cker, S.; Ehrig, M.; Eichkorn, K.; Elliott, S.; Furche, F.; Haase, F.; Ha¨ser, M.; Horn, H.; Hattig, C.; Huber, C.; Huniar, U.; Kattannek, M.; Ko ¨hn, M.; Ko ¨lmel, C.; ¨ hm, H.; Scha¨fer, A.; Schneider, U.; Kollwitz, M.; May, K.; Ochsenfeld, C.; O Treutler, O.; von Arnim, M.; Weigend, F.; Weis, P.; Weiss, H. Turbomole Version 5.8, 2005. (26) Becke, A. D. Phys. Rev. A 1988, 38, 3098–3100. (27) Perdew, J. P. Phys. Rev. B 1986, 33, 8822–8824. ¨ hm, H.; Ha¨ser, M.; Ahlrichs, R. Chem. Phys. (28) Eichkorn, K.; Treutler, O.; O Lett. 1995, 242, 652–660. (29) Eckert, F.; Klamt, A. COSMOtherm, Version C2.1, Revision 01.06; COSMOlogic GmbH&CoKG: Leverkusen, Germany, 2006; http://www.cosmologic. de.

2868

Analytical Chemistry, Vol. 82, No. 7, April 1, 2010

depending on the compound family. With all esters in addition to the [B+H]+ ion also the ions [B+Na]+ and [B+H-ROH]+ (where ROH is the respective alcohol) were observed. In addition to the [B+H]+ ion N,N-dimethylaniline gives the ion [B-CH4+H]+, benzylamine gives the ion [B-NH3+H]+, phenyl-N,N,N,N-tetramethylguanidine and tetramethylguanidine gives the ion [B-N(CH3)2H+H]+, sulfanilamide gives the ions [B-NH3+H]+ and [B+Na]+ and DBU gives the ion [B+H3O]+ in mass spectrum. In mass spectra of the studied pesticides, there are different fragment ions in addition to the [B+H]+ ion and the ion [B+Na]+. The dNsOs, sOsC(dO)s, sC(dO)sNHs and sOsC(dO)s bonds tend to break in the protonated methomyl, methiocarb and aldicarb. The rest of the neutral compounds formed only [B+H]+ ions in the ESI source. From tetraalkylammonium salts tetraalkylammonium ions are formed without any fragmentation or adducts. MS Signals due to fragment ions and Na+-adduct ions were more intense in the mass spectra of esters than the signals of [B+H]+ ions. For all other compounds the MS signals of the additional ions were less intense than that of [B+H]+ in mass spectra. We restrict our approach only with compounds that ionize via protonation (or that are ions already in solution). Our assumption is that Na-adduct ions are generated at the same time as H-adduct ions.6 Therefore [B+Na]+ ions were not considered in calculations of logRIE values. On the other hand, fragment ions that can be assumed to form in the gas phase from the [B+H]+ ions initially formed from the solution droplets were considered as daughter ions of the [B+H]+ ions and were counted into the MS responses. Thus the MS responses were found as follows: R ) I([B + H]+) + I([B - x + H]+)

(4)

where x stands for a particular fragment that is eliminated during fragmentation. If multiple fragments are formed, then the intensities of all of them were added to the intensity of the [B+H]+ ion. From the obtained logRIE values a scale presented in Table 2 was compiled using the same approach as in ref 6. The logRIE values were anchored to the logIE value of methyl benzoate arbitrarily set to zero, chosen as the reference compound for the scale. The logIE values (i.e., ionization efficiencies vs methyl benzoate) for the other compounds were assigned using the leastsquares approach. The obtained values are the ionization efficiencies (excluding methyl benzoate) for all the compounds in the scale. These logIE values are obtained by minimizing the sum (SS) of squared differences between the differences of the assigned logRIE(Bi, Bj) values and the directly measured logIE(Bi) and logIE(Bj) values, which is minimized by varying congruous logIE values: nm

SS )

∑ {log RIE (B , B ) - [log IE(B ) k

i

j

k)1

i

(5)

2

log IE(Bj)]} f min where nm is the number of measurements and logRIEk(Bi, Bj) is the result of k-th measurement that has been conducted between compounds Bi and Bj. The minimization procedure

Table 2. Studied Compounds, Their Obtained LogIE Values, Scale Containing 116 Measurement Results, pKa Value in Water (pKa H2O), pKa Value in Acetonitrile (pKa MeCN), Gas Phase Basicity (GB) (kcal/mol), Molecular Mass of Ion (M) (g/mol), Molecular Area of Ion (A) (Å2), Polar Surface Area of Ion (PSA) (Å2), Molecular Volume of Ion (MV) (Å3), Dipole Moment of Ion (δ) (debye) and Log of Distribution Coefficient of Solvent/Hexane for Ionic Form of the Compound (logPs/h). Values in Italic Are Computational

a ref 19. b ref 20. c ref 31. d ref 32. e ref 33. f ref 34. g ref 35. h ref 36. i logIE values vs methyl benzoate obtained via least-squares minimization procedure (eq 5).

Analytical Chemistry, Vol. 82, No. 7, April 1, 2010

2869

has been systematically used in compilation of acidity and basicity scales and is described in refs 19 and 30. Consistency of the Scale is expressed by the consistency standard deviation:

s)



SS nm - nc

(6)

where SS is defined via eq 5, nm ) 407 is the overall number of measurements and nc ) 61 is the number of assigned logIE values. With our data s ) 0.30 logRIE units. Examining the scale reveals that 143 of the logRIE measurements disagree with the assigned logIE values by 0.2 units or more. By eliminating the results of these 143 measurements the consistency standard deviation of the scale containing the remaining 264 measurements would decrease to 0.1 logRIE units. The difference is smaller than the consistency standard deviation of the scale itself in all but two (triphenylamine and ethyl-methylimidazolium cation) cases. Therefore we keep all the compounds and all the measurements in the scale, in spite of the higher consistency standard deviation. The very good agreement of the logRIE values obtained with and without exclusion of the deviating measurements is an indication of the robustness of the approach. The following factors can be ruled out from the point of view of their possible contribution to the disagreements between different logRIE measurements: (1) The effect of large difference in molecular masses of two measured compounds. For some compound pairs with very different molecular masses we obtained logRIE values that agree with the assigned logIE differences. For example, measurement of (C4H8N)3-P ) N-C6H3-2,5-Cl2 (M ) 400.14 g/mol) versus diphenylguanidine (M ) 211.27 g/mol) resulted in logRIE that differed from the assigned logIE difference by -0.002 units and measurement of sulfanilamide (M ) 172.21 g/mol) versus diethylamine (M ) 73.14 g/mol) resulted in logRIE that differed from the assigned logIE difference by 0.04 units. At the same time measurements with compound pairs with similar molecular masses sometimes resulted in logRIE values disagreeing with the differences of the assigned values: Aniline (M ) 93.13 g/mol) versus pyridine (M ) 79.10 g/mol) gave logRIE value that disagrees by 0.4 units with the difference of the assigned value. (2) Difference in concentrations of two measured compounds. In addition to the example in the experimental section, also dimethyl phthalate vs dimethyl glutarate is a good example: concentration of these compounds varies correspondingly 6.81 ppm and 13.71, 0.61 and 6.81 ppm, 23.19 and 7.89 ppm, 23.19 and 10.28 ppm, 11.19 and 6.53 ppm, 9.88 and 9.38 ppm; logRIE values are, correspondingly, 0.50, 0.50, 0.52, 0.56, 0.44, 0.53, and these results agree well with the assigned logRIE values. Another (30) Ku ¨ tt, A.; Leito, I.; Kaljurand, I.; Soova¨li, L.; Vlasov, V. M.; Yagupolskii, L. M.; Koppel, I. A. J. Org. Chem. 2006, 71, 2829–2838. (31) Kaljurand, I.; Koppel, I. A.; Ku ¨ tt, A.; Ro ˜o ˜m, E.-I.; Rodima, T.; Koppel, I.; Mishima, M.; Leito, I. J. Phys. Chem. A. 2007, 111, 1245–1250. (32) Perrin, D. D. Dissociation Constants of Organic Bases in Aqueous Solution; Butterworths: London, England, 1965. (33) NIST Chemistry WebBook. http://webbook.nist.gov/chemistry/ (accessed March 2, 2010). (34) Bascombe, K. N.; Bell, R. P. J. Chem. Soc. 1959, 1096–1105. (35) Izutsu, K. Acid-Base Dissociation Constants in Dipolar Aprotic SolVents; IUPAC Chemical Data Series No. 35; Blackwell Scientific: Oxford, 1990.

2870

Analytical Chemistry, Vol. 82, No. 7, April 1, 2010

example could be diphenylamine vs diphenyl phthalate, with concentrations 3.01 ppm and 2.28 ppm the logRIE value is 0.12 units and with concentrations 0.61 ppm and 0.61 ppm the logRIE value is 0.12 units as well. (3) Dependence of the ion’s MS signal from concentration and the slope of this dependence. A main assumption of our approach6 is that when spraying a solution of two compounds, they do not influence each other’s ionization significantly, or that the influence is mutual (the compounds suppress or enhance each other’s ionization the approximately same way). This means that the compound’s response should be independent of the concentration in the used concentration range. In reality we observed that this was not always so. There was a thought that if the correlation of one compound’s response from concentration has a positive and another compound’s response from concentration has a negative slope then one compound has a suppression effect and another one the effect of enhancement. So in case like that it can be thought that results do not agree with assigned values, but it is not so. There are compound pairs with similar slopes that provide unsuitable results and there are compound pairs that have contrary slopes of mentioned correlation and still provide well adequate results. Correlation of logIE Values to Molecular Properties. Based on the general knowledge of the ESI mechanism and on the earlier works (see Table 1) the following properties of the studied compounds were considered in relation to ionization efficiency: (1) Basicity, i.e. the ability to be protonated and become a cation. Basicity was characterized by the pKa values in water and in acetonitrile as well as by the gasphase basicity GA. (2) Polarity, that is, the distribution of charge within the molecule. This was characterized by the dipole moment of the cation and by the polar surface area of the cation. (3) Distribution coefficient of the protonated forms between a polar and a nonpolar medium. This parameter is expected to characterize the affinity of the protonated forms toward the drop surface. (4) Molecular size. Generally, the larger the molecule the better stabilized is its protonated form in the gas phase. Molecular size was characterized by molecular weight, molecular surface area and molecular volume of the protonated compound. Logarithmic values (e.g., pKa instead of Ka, logP instead of P) are often used because it reduces the division of values into groups due to the differences in these values. Very few of the studied compounds have experimental data of all the mentioned properties. Therefore a large part of the data used in multilinear regression analysis is computational. In particular, logPsolvent/hexane was used instead of logPoct/w. Although there are no experimental logPsolvent/hexane values available, this parameter is in our case a better choice than logPoct/w: (1) Hexane is a solvent of significantly lower polarity, so that this value better describes hydrophobicity. (2) Even though experimental logPoct/w data for a number of compounds in our scale are available,37 these are often of limited use, because of different (or undefined) pH values that were used. This leads to poorly defined meaning (in terms of actual species in solution) of many of the experimental values. At the same time COSMO-RS computation allows to examine directly the protonated form of the compound. And once the values are computational, it does not make difference for the COSMORS, which of the two to compute. (36) Hall, H. K., Jr. J. Am. Chem. Soc. 1957, 79, 5439–5441. (37) Hansch, C.; Leo, A.; Hoekman, D. Exploring QSAR Hydrophobic, Electronic, and Steric Constants; American Chemical Society: Washington, DC, 1995.

Figure 1. Correlation between estimated (model 8) and experimental logIE values. Compound numbers are the same as in Table 2. The line represents the ideal correlation between logIEestimated and logIEexp with slope 1 and intercept 0.

Regression Model. Scaled parameters were used in the regression model: Xis )

Xi - mean(X) s(X)

(7)

where mean(X) and s(X) are the mean and standard deviation, respectively, of the parameter X over all compounds studied. In the initial model all the above-mentioned nine molecular properties were included. In the model with all compounds included only two parametersspKa value in water and logarithm of cation’s molecular volumeswere found to be statistically significant. The final model for scaled and centered parameters can be written as (the coefficients are given together with standard errors): log IEs ) (0.3941 ( 0.0752)pKas + (0.6599 ( 0.0752)log MVs (8) The scaled parameters were found according to eq 7. The mean(pKa) and mean(logMV) were 6.001 and 5.097, respectively. The s(pKa) and s(logMV) were 10.881 and 0.436), respectively. The logIE can be calculated from logIEs as follows: log IE ) mean(log IE) + log IEs · s(log IE)

(9)

where mean(logIE) is 3.200 and s(logIE) is 1.403. The squared correlation coefficient for this model is 0.675. The standard error of this model is 0.8135 logIE units. The residuals values range from -2.319 to 0.988 logIE units. The agreement between the measured and predicted logIE values is presented in Figure 1. As can be seen from Figure 1 the model for predicting logIE values works well at qualitative level. Predicting logIE values is problematic with compounds that tend to ionize poorly in ESI. Validation of the Model. In order to assess the predictive power of the model (eq 8) the data set was divided into two parts. 42 compounds were randomly selected for the “training set”, which

Figure 2. Correlation between estimated (model 9) and experimentally gained logIE values for test set compounds. Compound numbers are the same as in Table 2. The line represents the ideal correlation between logIEestimated and logIEexp with slope 1 and intercept 0.

was used to evaluate the coefficients for the prediction model. The average measured logIE of the training set was 3.121 and the standard deviation of the logIE was 1.456. The parameters were scaled and centered according to eq 7) (mean(pKa), mean(logMV), s(pKa), and s(logMV) values were 6.186, 5.089, 11.361, and 0.397, respectively). The model obtained for scaled and centered parameters was similar to eq 8: log IEs ) (0.4109 ( 0.0934)pKas + (0.6359 ( 0.0934)log MVs (10) with the squared correlation coefficient of 0.67 and the standard residual error 0.8572 logIE units. This model was used to predict the logIE values for the set of 20 compound left out from the training set (see Figure 2). The pKa values and logarithms of molecular volume for these compounds were scaled previously according to the mean and standard deviation of the training set compounds. The scaled logIE values were predicted and calculated back into nominal logIE values by using the mean and standard deviation of the logIE of the training set. While comparing the predicted and measured logIE values a good correlation was observed. The differences of measured and predicted logIEs ranged from -1.57 to 1.11 logRIE-units. It was studied if the regression model inside one compound group differs from the regression model for all compounds. We chose aminessboth aromatic and aliphatic, altogether 21 compoundssfor this study, as being the most abundant group of compounds in the present study. The regression analysis was carried out and following eq was found: log IEs ) (0.5480 ( 0.1050) + (0.8699 ( 0.1255)log MVs + s (1.5721 ( 0.3097)GBs - (0.6877 ( 0.1811)log P s/h

(11)

It can be seen, that for the amine group also gas phase basicity and solvent/hexane distribution coefficient become statistically Analytical Chemistry, Vol. 82, No. 7, April 1, 2010

2871

significant. At the same time pKa is statistically insignificant. The residual standard error of this model is 0.46 logIE units and the errors range from -1.33 to 1.09 logIE units. It is possible that for other compound groups still different regression models can be found but unfortunately in the current data set none of the groups have enough compounds to carry out a statistically meaningful analysis. Discussion on the Relation of logIE and Molecular Properties. The range of compounds in this study is much more diverse than in any previous study. The variety of molecular masses is in range of 46 g/mol up to 551 g/mol, hence other size characterizing properties of compounds like molecular volume and area also diverse. There are compounds with strong basic properties (guanidines) as well as compounds with weak basic properties (monoesters), compounds that are hydrophilic (phosphazenes) as well as hydrophobic (diesters), compounds with low polarity (trialkyl amines) as well as high polarity (nitroanilines) studied in this research. The results provide an implication that in the case of watercontaining mobile phases besides the basicity of the analytes their molecular size is very important in determining the ionization efficiency. Several authors6,10 have found that hydrophobicity (usually expressed as octanol/water partition coefficient) and/or nonpolar surface area are important characteristics in determining ESI IE. Like Cech and Enke10 have shown with some peptides we also have concluded that increasing the extent of nonpolar character in a compound leads to enhanced affinity for the surface phase and higher ESI response. Impressive examples in this respect are the tetraalkylammonium salts and trialkylamines studied in this work that are among the most efficient ionizers in ESI source. In the current work hydrophobicity (logPsolvent/hexane) was found not to be statistically significant. This is most likely due to the nonnegligible correlation between logPsolvent/hexane and logMV (R2 ) 0.38), meaning that logMV in part describes also hydrophobicity. In most works (including this work) the mobile phase contains a significant amount of water. Markedly different results have been obtained by other authors in pure organic solvents: Ehrmann, Henriksen and Cech8 have shown the value of logP is insignificant in describing ESI IE if methanol is used as solvent. The logIE values of most compound families have a rather wide distribution, so that in most cases it is impossible to compare families from the point of view of logIE values. Nevertheless, there are certain compound families that can be compared to others as wholes. It can be seen that phosphazenes have higher ionization efficiencies than all the other measured compound families. It can be explained by the fact that the studied phosphazenes are more basic than most of the other studied compounds (in water, acetonitrile, and also in gas phase), they also have larger molecular volumes. The same applies to most tetraalkylammonium ions. Esters, especially monoesters, on the other hand, have comparatively low ionization efficiencies. This can be explained by their low basicity and small molecule size. Other compound families measured have different ionization efficiencies depending on each compound’s individual set of parameters (for example the ionization efficiency for ethylamine and tributylamine differs over 5000 times and for 2,6-dimethoxy pyridine and 2,6demethyl pyridine the difference is about 200 times). However, amines do not have logRIE values below 1 or above 5 units.

2872

Analytical Chemistry, Vol. 82, No. 7, April 1, 2010

The clear trend within compound families is that increasing the number of alkyl chains (e.g., pyridine, 2-methylpyridine, 2,6-dimethylpyridine, 2,4,6-trimethylpyridine) or lengthening of alkyl chain (e.g., trimethylamine, triethylamine, tripropylamine, tributylamine) invariably increases the ionization efficiency of the compounds. Both these effects lead to increase of the molecule size (also lipophilicity) and also to a mild increase of basicity. Similar trend is observed in the series of five guanidines: the larger the molecule the higher the ionization efficiency, even though introducing phenyl groups into the guanidine moiety decreases somewhat the basicity. The trend changes when enlarging the molecules leads to a significant decrease of basicity. For example, in the seriessaniline, diphenylamine, triphenylaminesdiphenylamine has the highest ionization efficiency, beating triphenylamine by 0.5 units (and aniline by more than a unit). Apparently the reason is the significant decrease of basicity on going from diphenylamine to triphenylamine. CONCLUSIONS We have verified the earlier developed approach that allows setting up under predefined ionization conditions a self-consistent quantitative experimental scale of electrospray ionization efficiencies of organic compounds (ionization via protonation was only monitored). The range of compounds in this study is much more diverse than in any study published up to date. The scale of logRIE values containing 62 different compounds and spanning for 6 logIE unitssa million-fold difference in ionization efficienciesshas been established. The higher the ionization efficiency of a given compound, the better it is detectable by ESI-MS (ionization via monoprotonation). Compounds that have low ionization efficiency by monoprotonation may still be successfully detected using ESI-MS, if they readily give ions by some other ionization mechanism (e.g., ionization via forming of sodium adducts or forming anions). Initial correlations between logIE values of the compounds and their molecular parameters have been carried out using a linear regression model. The parameters that are most influential in predicting the ionization in ESI source are the pKa value of the compound in water and the logarithm of molecular volume of the compound. This scale and the whole approach can be a tool for practicing liquid chromatographists and mass spectrometrists. It can be used in any mass-spectrometry laboratory and we encourage practitioners to characterize their analytes with the logIE values so that a broad knowledge base on electrospray ionization efficiencies of compounds would start to form. ACKNOWLEDGMENT This work was supported by the grant No 7127 from the Estonian Science foundation and by the target financing project No SF0180061s08 from the Ministry of Education and Science of Estonia. We are indebted to Dr Frank Eckert from Cosmologic GmbH for valuable suggestions on COSMO RS computations and to mr Georgi Slavin for writing the software used to create the image of the logIE ladder presented in Table 2. Received for review December 16, 2009. Accepted February 18, 2010. AC902856T