Anal. Chem. 2002, 74, 4885-4888
Profiling pH Changes in the Electrospray Plume Shaolian Zhou,† Benjamin S. Prebyl, and Kelsey D. Cook*
Department of Chemistry, University of Tennessee, Knoxville, Tennessee 37996-1600
Laser-induced fluorescence spectrometry is used to profile pH changes as droplets evaporate in an electrospray plume by measuring emission spectra of 2(or 4)-[10(dimethylamino)-3-oxo-3H-benzo[c]xanthene-7-yl]-benzenedicarboxylic acid (carboxy SNARF-1), a pH-sensitive fluorescent dye. The observed pH changes depend on initial droplet pH and polarity. In some instances, small or negligible changes of pH are observed, consistent with expected buffering. The pH of initially acidic droplets decreases along the spray axis in both positive and negative ion modes, to an extent larger than expected from solvent evaporation. This phenomenon may be a manifestation of droplet cooling, droplet subdivision, or heterogeneous charge distribution within the spray plume or within individual ES droplets. The formation of ions in electrospray (ES) mass spectrometry1-3 and the efficiency of desorbing those ions4-6 can be critically sensitive to the chemical environment in the sprayed droplets. Thorough understanding of the ES mechanism therefore requires understanding of the chemical changes that take place in the droplets, including changes in pH, solvent composition, analyte concentration, and charge state.1,4,7-10 For example, production of abundant protonated ions from solutions in which their equilibrium concentration is low has been attributed to lowering of the droplet pH by solvent oxidation,10,11 simple proton enrichment due to solvent evaporation,4,12 or both. Localized enrichment of protons at the droplet surface may also contribute.2,3,7 Direct measurement of these changes is challenging because of the small size and complex dynamics of the droplets;4,10 few such measurements have been reported.1,8,10-13 * Corresponding author. Phone: (865) 974-8019. E-mail:
[email protected]. † Current address: Covance, Inc., Madison, WI 53704. (1) Zhou, S.; Cook, K. D. Anal. Chem. 2000, 72, 963-969. (2) Gatlin, C. L.; Turecˇek, F. Anal. Chem. 1994, 66, 712-718. (3) Zhou, S.; Cook, K. D. J. Am. Soc. Mass Spectrom. 2000, 11, 961-966. (4) Kebarle, P.; Tang, L. Anal. Chem. 1993, 65, 972A-986A. (5) Iribarne, J. V.; Thomson, B. A. J. Chem. Phys. 1976, 64, 2287-2294. (6) Thomson, B. A.; Iribarne, J. V. J. Chem. Phys. 1979, 71, 4451-4463. (7) Fenn, J. B. J. Am. Soc. Mass Spectrom. 1993, 4, 524-535. (8) Kiselev, P.; Rosell, J.; Fenn, J. B. Ind. Eng. Chem. Res. 1997, 36, 30813084. (9) Chillier, X. F. D.; Monnier, A.; Bill, H.; Guelacar, F. O.; Buchs, A.; McLuckey, S. A.; Van Berkel, G. J. Rapid Commun. Mass Spectrom. 1996, 10, 299304. (10) Zhou, S.; Edwards, A. G.; Cook, K. D.; Van Berkel, G. J. Anal. Chem. 1999, 71, 769-776. (11) Van Berkel, G. J.; Zhou, F.; Aronson, J. T. Int. J. Mass Spectrom. Ion Processes 1997, 162, 55-67. (12) Tang, K.; Smith, R. D. J. Am. Soc. Mass Spectrom. 2001, 12, 343-347. 10.1021/ac025960d CCC: $22.00 Published on Web 08/29/2002
© 2002 American Chemical Society
Our efforts in this area1,10 have involved use of laser-induced fluorescence spectrometry to monitor the pH and solvent composition of sprayed droplets. In previous studies, we limited the laser power at the spray plume, fearing photodamage to the probe dyes or other invasive effects of the laser beam. As a result, sensitivity has been inadequate to determine the pH more than 2 mm downstream from the ES emitter.10 In the present study, we raised the laser power an order of magnitude and doubled the dye concentration, enabling assessment of the pH at much later stages of the droplet evolution. Some of the resulting data coincide with intuitive expectations, whereas other observations appear to be reflective of the complexity of the chemistry of the spray. EXPERIMENTAL SECTION The fluorescent pH indicator 2(or 4)-[10-(dimethylamino)-3oxo-3H-benzo[c]xanthene-7-yl]-benzenedicarboxylic acid (carboxy SNARF-1, hereafter designated SNARF; Catalog No. C-1270, Molecular Probes, Eugene, OR; Figure 1) was used as received. The doubly (A2-) and singly (HA-; pKa 7.5) deprotonated species are the fluorescent forms of this indicator.14 The neutral form (H2A) is not appreciably soluble in water; its pKa has not been reported. All solutions were aqueous, using water which was passed through a Milli-Q system (Millipore, Bedford, MA) then boiled to remove residual CO2. Solutions of the dye (40 µM) were prepared by dilution of a 2 mM stock solution. Where noted, solution pH was adjusted using 1 mM KOH prior to final dilution. A “calibration curve” of the emission ratio (I638/I587, where I indicates the measured fluorescence signal at the subscripted wavelength) versus pH was obtained from a series of 20 µM dye solutions at varying pH contained in a 1-cm quartz cuvette. As in ref 10, these measurements employed solutions prepared with 0.05 M phosphate buffers. In addition, use of an AMANI-1000 mini pH sensor (Harvard Apparatus; South Natick, MA) allowed pH measurement on the small volumes necessitated by the expense of the dye, thereby enabling acquisition of calibration curves without buffer. These curves did not differ significantly from those obtained with buffer, suggesting that the spectroscopic effects of the phosphate buffer17 were negligible at these concentrations. Fluorescence temperature dependence was assessed using an Aminco-Bowman (Urbana, IL) Series 2 luminescence spectrometer (13) Rodriguez-Cruz, S. E.; Khoury, J. T.; Parks, J. H. J. Am. Soc. Mass Spectrom. 2001, 12, 716-725. (14) Timperman, A.; Tracht, S. E.; Sweedler, J. V. Anal. Chem. 1996, 68, 26932698. (15) Olumee, Z.; Callahan, J. H.; Vertes, A. Anal. Chem. 1999, 71, 4111-4113. (16) Olumee, Z.; Callahan, J. H.; Vertes, A. J. Phys. Chem. A 1998, 102, 91549160. (17) Whitaker, J. E.; Haugland, R. P.; Prendergast, F. G. Anal. Biochem. 1991, 194, 330-344.
Analytical Chemistry, Vol. 74, No. 19, October 1, 2002 4885
Figure 1. Structure of carboxy-SNARF 1. Table 1. Conditions, Slopes, and Standard Deviations of Slopes for Curves in Figure 3 initial pH
curvea
spray voltage (kV)
m ( sm (pH/mm)b
6.9 6.3 6.9 7.5 7.5 7.5
1 (9) 2 (b) 3 (\) 4 (2) 5 (1) 6 ([)
+4.0 -3.0 -3.3 +4.0 +3.5 -3.0
-0.100 ( 0.018 -0.121 ( 0.016 -0.081 ( 0.005 -0.002 ( 0.007 -0.010 ( 0.004 -0.030 ( 0.006
a Curve number and symbol used in Figure 3. b Linear least squares slope (m) and corresponding standard deviation (sm) calculated using the points at 1, 4, and 8 mm in the curves of Figure 3.
equipped with a 150 W ozone-free xenon lamp. The excitation wavelength was tuned to 514 nm with a band-pass of 2 nm for monitoring a 40 µM SNARF solution (pH 6.6) in a 1-cm quartz cuvette. The spectrometer was configured to acquire emission spectra from 580 to 640 nm in 6 s. Results are the average of triplicate analyses at each respective temperature. The temperature of the water flowing through the cell holder was regulated by a Haake (Berlin, Germany) model F 4391 circulating water bath. The temperature of the fluorescence cell and cell holder were measured with separate type K thermocouples and a Wavetek (San Diego, CA) DM23XT multimeter. The thermocouples were calibrated against an alcohol thermometer. The details of the experimental spray apparatus were described previously.1,10 Briefly, the ES source was comprised of stainless steel inner (160 µm i.d. × 310 µm o.d.) and outer (510 µm i.d. × 820 µm o.d.) capillaries arranged concentrically using a Swagelok (Solon, OH) stainless steel “tee” fitting and graphite ferrules. The inner capillary extended ∼0.5 mm beyond the outer one. The distance between the tip of the spray needle (inner capillary) and a 2.5-cm-diameter brass counter electrode was maintained at 14 mm. Prepurified-grade N2 nebulizing gas (National Welders; Charlotte, NC) was introduced through the annulus between the capillaries at a flow rate of 1.2 L/min. The sample was infused at a flow rate of 5 µL/min using a Harvard (South Natick, MA) model 11 syringe pump. The high voltage for electrospray was supplied by a Fluke (Seattle, WA) model 408B power supply. The ES voltage varied between -3.0 and +4.0 kV (see Table 1), which resulted in ES currents j50 nA.10 The ES assembly was mounted on a two-dimensional positioner (L. S. Starrett Co.; Athol, MA), allowing translation in the lateral (x) and axial (z) directions (the z axis coincides with the spray). Along the optical (y) axis, the system focusing optics were used to maximize fluorescence intensity, so that data collection was from the center of the plume 4886 Analytical Chemistry, Vol. 74, No. 19, October 1, 2002
Figure 2. Comparison of the fluorescence spectra obtained by spraying a 40 µM SNARF aqueous solution in the positive-ion mode and probing the plume at z )1 mm and 8 mm. The spectra were normalized to the isoemissive point at 607 nm. ES voltage: +3.6 kV. Arrows indicate the increase in the relative emission from the acidic form of the dye and the corresponding decrease in emission from the basic form at the downstream position.
(depth of field is ∼1.5 mm10). Positioning in all dimensions was estimated to be reproducible to within ∼0.1 mm. Fluorescence spectra of the spray were obtained in the coaxial (backscattering) mode using the macro stage of a Dilor XY Raman spectrometer (Instruments S. A. Inc.; Edison, NJ) equipped with an EG&G (Trenton, NJ) OMA 4 CCD detector cooled to -70 °C. The spectrometer was fitted with a 150 groove/mm grating; with the 100-µm slits used, this resulted in a band-pass of ∼2 nm. A Lexel 3500 (Fremont, CA) argon ion laser provided 514.5-nm excitation radiation. The laser power (50 mW at the source, 7 mW at the sample for spray experiments) was 10 times higher than used previously and provided the expected 10-fold signal enhancement. Even under this condition, no spectral changes were observed when ∼1 mL of a dye solution in a 1-cm quartz cell was exposed to the laser for 50 s; photodegradation was apparently not a problem under these conditions. In view of the short exposure times of individual droplets, photodamage was therefore assumed to be negligible in the spray. Spectra were acquired for 50 s over the wavelength range from 520 to 930 nm, then smoothed with a 5-point Savitsky-Golay smooth. Three such acquisitions were averaged in each case; error bars represent one standard deviation (σ) of these triplicate measurements. RESULTS AND DISCUSSION Figure 2 shows fluorescence spectra acquired from the plume at z ) 1 mm and 8 mm when electrospraying a 40 µM aqueous solution of SNARF in the positive ion mode. The initial pH of this solution (estimated from the I638/I587 emission ratio for a hanging drop) was adjusted to 6.90 using dilute KOH in an attempt to maintain the spray pH near the optimum working range of the dye (pKa ( 1 ∼7.5 ( 1). Within 1 mm of the spray tip, the pH had decreased to 6.36, in very good agreement with that predicted (6.29) based on the electrolytic generation of protons by solvent oxidation, as described previously.10 It is clearly evident from the data of Figure 2 that the pH of the droplets continues to decrease as they move down the plume;
Figure 3. Plots of the pH values estimated from the fluorescence spectra of SNARF in the spray plume versus the axial distance (z) from the emitter tip. ES voltage: curve 1 (9), + 4.0 kV; curve 2 (b), -3.0 kV; curve 3 (\), -3.3 kV; curve 4 (2), + 4.0 kV; curve 5 (1), +3.5 kV; and curve 6 ([), -3.0 kV. Left-most point in each plot reports the pH of the bulk solution prior to spraying. Error bars represent one standard deviation, based on triplicate measurements of each point.
arrows in the Figure highlight the increase in emission attributable to the protonated form of the dye at 8 mm (HA-, emitting at 587 nm) and the corresponding decrease for the deprotonated form (A2-, 638 nm). Comparison with emission from buffered solutions indicates a pH near 5.67 at 8 mm downstream; the proton concentration has increased 5-fold relative to that measured at 1 mm. The decrease is gradual but reasonably steady (curve 1 in Figure 3), consistent with the intuitive expectation that the pH of a solution that is even slightly acidic should drop as the volume decreases and the acid becomes more concentrated as a result of solvent evaporation. The relatively large error bars at the lowest pH reflect the extension beyond the “optimum” pH range of the dye, but leave no doubt of the decrease. Very similar results were obtained when spraying a solution with initial pH 6.30 (no added KOH) in the negative ion mode (curve 2 in Figure 3). After an electrolysis-induced increase to pH ∼6.77 at z ) 1 mm (in good agreement with the 6.65 value expected based on the electrolysis current), the pH falls to ∼5.92 at z ) 8 mm, representing a ∼7-fold increase in the proton concentration. The rough parallelism of these semilogarithmic curves (slope m ) -0.100 ( 0.018 for curve 1 and -0.121 ( 0.016 for curve 2; Table 1) might be expected for simple solvent evaporation at a rate insensitive to the initial droplet pH or polarity. However, the observed 5-7-fold increase in proton concentration is much larger than expected for this system. Kebarle4 predicted a ∼4-fold change in volume for a 1.5 µm (initial radius) methanol droplet in the time required to travel ∼8 mm (∼0.4 ms, based on Vertes’ velocity measurements15,16). Arguably, a purely aqueous droplet would be larger (due to higher surface tension) and should evaporate more slowly (as a result of lower vapor pressure); using equations from ref 4, we estimate that under conditions similar to Kebarle’s (sample flow rate, capillary diameter and voltage, gas flow, and temperature), aqueous droplets would be ∼2.4 µm in radius and would lose only ∼10% of their volume in 0.4 ms, vs 80-86% for 5-7-fold enrichment. Furthermore, the presence of 40 µM dye should serve to buffer the system; with this quantity
of a pKa 7.5 weak acid, even a 4-fold volume change beginning at pH 6.30 should cause the pH to decrease by less than 0.03, that is, by less than the uncertainty in the measured pH. It is clear from Figure 3 that the decrease exceeds even the relatively large error bars at low pH. Perhaps more surprising is the observation of a similar decrease in pH when electrospraying a pH 6.90 solution in the negative ion mode (curve 3 in Figure 3). Here, there is roughly a 4-fold increase in hydrogen ion concentration in moving from 1 to 8 mm, despite the fact that this solution becomes slightly basic following electrolysis (pH ∼7.20 at 1 mm) and would therefore be expected to become more basic (positive slope) if evaporation were the only process active. Comparison with the other two curves for which pH at 1 mm is ∼7.12 and 7.27 (curves 4 and 5 in Figure 3) confirms that factors in addition to simple solvent evaporation must be operative; the latter two curves are essentially horizontal from 1 to 8 mm (Table 1), as expected for the slow evaporation described above and for solutions for which the pH is very near pKa so that buffer capacity is optimized. The anomaly persists for the most basic solution (curve 6 in Figure 3); even here, there is a small decrease in pH as the droplets move away from the emitter (m ) -0.03; Table 1). This curve lies well within the range of pH for which SNARF is a useful pH indicator, so that error bars are relatively small and the small decrease appears to be statistically significant. The slopes for the curves of Figure 3 generally correlate with initial pH and ionic strength; slopes are smaller for those solutions to which KOH was added to raise the prespray pH. If our adaptation of Kebarle’s model4 for some reason underestimates the evaporation rate for curves 1-3, the shallow slopes of curves 4-6 may simply reflect a slower rate of evaporation due to a colligative effect of the KOH added to raise the initial pH. However, only on the order of 10-20 µM KOH (final formal concentration) was needed to reach the initial pH for the top three curves of Figure 3. The total electrolyte concentration may be enhanced some by electrolysis (mainly of the acidic solution in positive ion or of the basic solution in negative ion mode), but at the currents and flow rate employed, this addition will be small (on the order of micromolar). In any case, since the vapor pressure (and therefore, the rate of evaporation) is proportional to the solvent mole fraction (Raoult’s law), such low electrolyte concentrations should have only a small effect at these early stages of evaporation. The spectroscopic (as opposed to colligative) influence of ionic strength is also small; Whitaker et al. report very little change in the emission wavelengths, intensities, or ratios when 200 mM NaCl is added to aqueous SNARF solutions.17 Similarly, effects of self-quenching or dye aggregation should be negligible at these low concentrations and would in any event cause an apparent increase in pH; the emission yield of the protonated form of the dye is reportedly attenuated more strongly (decreasing by a factor of 6.7) than the deprotonated form (decreasing by a factor of only 2.3) when the concentration is increased from 10 µM to 1 mM.18 The fluorescence of SNARF is slightly temperature-sensitive;17-19 lowering the temperature of a pH 6.6 sample (in a cuvette) from 50 to 10 °C resulted in a shift of the relative emission intensities (18) Parker, J. W.; Laksin, O.; Yu, C.; Lau, M. L.; Klima, S.; Fisher, R.; Scott, I.; Atwater, B. W. Anal. Chem. 1993, 65, 2329-2334. (19) Szmacinski, H.; Lakowicz, J. R. Anal. Chem. 1993, 65, 1668-1674.
Analytical Chemistry, Vol. 74, No. 19, October 1, 2002
4887
equivalent to that caused by lowering the pH ∼0.3 units. If evaporative cooling overwhelms collisional warming of the droplets, this effect could enhance the apparent pH change of acidic droplets and offset that of basic droplets. Further study will be needed to assess fully any contribution from this mechanism; use of a more highly temperature-sensitive and less highly pH-sensitive fluorophore would be advantageous. One final, relatively complex potential contributor to the “anomalous” behavior of Figure 3 should be considered. It has been reported that the mutual Coulombic repulsion among excess protons in positive ion ES can lead to a 103- to 104-fold increase of proton concentrations within a layer 5-27 nm from the droplet surface.2,3,7 Enrichment of the hydrophobic dye near the acidic surface might enhance the abundance of HA- to an extent that increases as evaporation increases the surface-to-volume ratio or as the dye has more time to diffuse to the surface. (An organic anion with a typical 10-5 cm2/s diffusion coefficient would be expected to diffuse ∼1 µM in 0.4 ms.) This effect could accelerate the reduction in pH observed in curve 1. For the other two positive-ion experiments (curves 4 and 5), droplets are initially slightly basic, so that most electrolytically generated protons would combine with hydroxide, leaving potassium ions as the cation in excess at the surface. This, combined with better buffering (because the initial pH is closer to pKa) could mitigate any surface effect. The case is more complex for the three negative-ion experiments (curves 2, 3, and 6). For curves 2 and 3, the hydroxide concentration at z ) 1 mm is lower than the dye concentration; electrolytically generated hydroxide can react with HA- so that the excess negative charge will likely be HA- and A2-. The Coulombic force driving excess charge to the surface will be stronger for the dianion, but it may be opposed by a potential barrier resulting from the drive to solvate this ion. Net partitioning of HA- into a distinct surface layer may contribute to its overall enrichment.20,21 For curve 2, this effect would be enhanced by the drop in pH resulting from evaporation of an initially acidic droplet; this synergy (plus possible thermal effects; see above) may account for the fact that this curve has the steepest slope (-0.121; Table 1). For the more basic droplets of curve 6, hydroxide and dye concentrations are roughly equal, and the majority of the dye is fully deprotonated (pH > pKa). Because of the greater mobility of the smaller hydroxide ion (compared with the dye), the surface excess charge is likely to be OH-. Thus, the analyte ion may be (20) Enke, C. Anal. Chem. 1997, 69, 4885-4893. (21) Zhou, S.; Cook, K. D. J. Am. Soc. Mass Spectrom. 2001, 12, 206-214. (22) Gomez, A.; Tang, K. Phys. Fluids 1994, 6, 404-414.
4888
Analytical Chemistry, Vol. 74, No. 19, October 1, 2002
left behind in the droplet interior, wherein pH may fall slightly as hydroxide ions partition to the surface. If inhomogeneity contributes to the trends of Figure 3, droplet subdivision may enhance the effects. Dye at or near the droplet surface may be enriched in the small, highly charged offspring droplets that result from uneven droplet fission.4,22 For curve 1, the offspring droplets would be more acidic than the bulk, since the excess charge is primarily H+. As discussed above, for curves 2 and 3, the dye itself comprises the excess charge; removing HA- into offspring droplets may favor a shift toward more protonation in the large residual droplets. Effects would be mitigated in curves 4 and 5, where the excess charge is primarily sodium. Finally, stripping OH- into offspring from the basic droplets of curve 6 may lower the pH of the residual drops, contributing to the small negative slope. All of these effects may be convoluted with the differing spatial distributionsand resulting differential samplingsof large and small droplets in the plume.10,12 Directly probing inhomogeneity within or between droplets is difficult; the fluorescence method used here can measure only the average chemical changes within the dimension of the laser spot. Further study will be needed to better resolve the relative contributions from the various mechanisms discussed (and perhaps others). Nevertheless, these findings clearly indicate that there are spray-induced chemical changes within the plume and suggest a surprising mechanistic dependence on the initial pH. If kinetics are fast enough (on the ES sampling time scale), these changes may have a significant impact on the relation between ES mass spectra and sample solution chemistry. ACKNOWLEDGMENT We thank Gary J. Van Berkel (Oak Ridge National Laboratory) for the SNARF sample. Student assistantships from the University of Tennessee Science Alliance (B.P., S.Z.), the University of Tennessee Measurement and Control Engineering Center (a National Science Foundation Industry/University Cooperative Research Center; B.P., S.Z.), and the University of Tennessee Scholarly Activity and Research Incentive Fund (S.Z.) are gratefully acknowledged. Reviewers of the manuscript offered constructive suggestions that contributed to the discussion. NOTE ADDED AFTER ASAP POSTING. This article was inadvertently posted ASAP August 29, 2002, without citations for footnotes 20 and 21. The corrected version was posted on August 30, 2002. Received for review July 19, 2002. Accepted August 2, 2002. AC025960D
