Characterization and Improvement of Signal Drift ... - ACS Publications

CC.C. EEE. III.... JJJ............. LL. MM. NN. OO.. loading. To add this web app to the home screen open the browser option menu and tap on Add to ho...
0 downloads 0 Views 2MB Size
Anal. Chem. 2010, 82, 6480–6486

Characterization and Improvement of Signal Drift Associated with Electron Ionization Quadrupole Mass Spectrometry Ward D’Autry, Kris Wolfs, Sitaramaraju Yarramraju, Ann Van Schepdael, Jos Hoogmartens, and Erwin Adams* Laboratory for Pharmaceutical Analysis, Catholic University of Leuven, Herestraat 49, O&N2, PB 923, B-3000 Leuven, Belgium Quadrupole mass spectrometry with electron ionization (EI-QMS) is a very popular detection technique in combination with gas chromatography. It is deployed for the analysis of volatile and semivolatile analytes in many industry domains. Although a very important factor for quantitative analysis, little is known about the stability of ion source performance. Only a few papers and patents report possible signal instabilities due to sample adsorption, degradation, or insulating deposits on the hot stainless steel surface of the ion source. In this study, a conventional stainless steel ion source was used to investigate possible signal drifts. It was observed that the EI-QMS instrument indeed suffered from continuous signal instability. It was found that the key parts which are responsible for the signal instabilities are those that regulate the ion beam toward the mass analyzer: the repeller, exit plate, and focusing lenses. The voltage of the repeller was found to have a major influence on the signal stability. The surface of the repeller, exit plate, and focusing lenses was modified by applying a gold coating. It was demonstrated that the signal stability of the MS dramatically improved when using the gold-coated parts. The contribution of each part to the stability improvement was quantitatively determined and compared with the standard stainless steel source performance. It was assumed that the signal drift observed with the stainless steel EI source originated from charge buildup on the surfaces. This hypothesis was supported by software simulations. Electron ionization quadrupole mass spectrometry (EI-QMS) is an additional separation and detection technique which can be easily coupled to gas chromatography (GC). The ion source of the mass spectrometer is responsible for ionizing the injected analytes, which are further analyzed through the mass analyzer. It is one of the key components of the mass spectrometer as its design and operational parameters are critical for efficient, sensitive, and stable operation of the instrument. The EI sources used nowadays still have many similarities with the first-generation * To whom correspondence should be addressed. E-mail: Erwin.Adams@ pharm.kuleuven.be. Fax: +32 16 323 448.

6480

Analytical Chemistry, Vol. 82, No. 15, August 1, 2010

Figure 1. (A) Schematic drawing of the EI source used in this study: (1) filament, (2) anode, (3) ion volume, (4) repeller, (5) exit plate, (6) ion source body, (7) source magnets, (8) focusing lenses. (B) Software simulation showing the equipotential lines of the electric fields inside the ion source. The model was built using real dimensions and voltages of the ion source parts. The voltage on the repeller was set at 1 V, the anode at 70 V, lens 1 at 7 V, and lens 2 at 12 V. The ion source body was kept at the ground potential.

EI sources designed more than 60 years ago by Nier et al.1,2 A schematic representation of the EI source which was used in this study is shown in Figure 1A. Ionization is achieved through the interaction of an analyte with an energetic electron beam, which results in the loss of an electron from the analyte molecule and the production of a radical cation: M + e- f M•+ + 2eThis radical cation may further undergo several fragmentations into fragment ions. Although much less studied, negative ions are also formed through electron ionization. The mechanisms of negative ion formation include ion pair formation and electron attachment.3-6 The electrons for the ionization are produced by thermionic emission from a heated metal filament. Indeed, the filament of the EI source is heated to a temperature where electrons exceed (1) Nier, A. O. Rev. Sci. Instrum. 1940, 11, 212–216. (2) Nier, A. O. Rev. Sci. Instrum. 1947, 18, 398–411. (3) Reese, R. M.; Dibeler, V. H.; Mohler, F. L. J. Res. Natl. Bur. Stand. 1956, 57, 367–369. (4) Tsuda, S.; Yokohata, A.; Kawai, M. Bull. Chem. Soc. Jpn. 1970, 43, 1649– 1656. (5) McAllister, T. J. Chem. Soc., Chem. Commun. 1972, 4, 245. (6) Rapp, D.; Briglia, D. D. J. Chem. Phys. 1965, 43, 1480–1489. 10.1021/ac100780s  2010 American Chemical Society Published on Web 07/13/2010

the work function barrier of the filament material.7 The produced electrons are accelerated toward the anode through a narrow slit. The interaction with the sample molecules takes place in the ion volume, which is the open space between the filament and the anode. The positive ions produced during the ionization process are pushed away by the repeller positioned perpendicular to the electron trajectory. The repelled ions are extracted by an exit plate and directed toward the analyzer by a series of lenses. A simulation of the equipotential lines in an EI source, based on the dimensions and parameters of our EI-QMS system, operating under standard and ideal conditions, is shown in Figure 1B. The electric fields in the ion volume (except the borders near the anode) are fairly weak compared to the electric fields at the lens systems. Ion extraction results from the potential difference between the repeller and exit plate. Subsequently, the focusing lenses control the ions’ kinetic energy for optimal quadrupole separation.8 Few papers deal with improvements in ionization and transmission efficiency by modifying the concept of Nier. The use of a shorter ionization path was investigated by Kuhnke et al. in 1994.9 The use of a cylindrical repeller shape instead of a flat shape for improved ion beam focusing was reported by Park and Ahn.10 A new medium for EI, defined as a supermolecular beam interface, was introduced by the group of Amirav. Advantages include reduction of background noise and less ion source degradation.11 For quantitative analysis, the number of produced ions must be proportional to the amount of injected analyte molecules. Hence, stable instrument performance is an important factor for quantitative work. However, several events can contribute to signal intensity variation. Some of them, such as filament emission variability, electron energy variation, and repeller voltage, are listed by Millard.12 The hot stainless steel surface can lead to sample adsorption, degradation, and adverse ion-surface interactions. This could affect the accuracy of a quantitative measurement.13 Other undesirable effects of analyte-surface interactions are the presence of unexpected ions, peak tailing, loss of sensitivity, nonlinearity, and general erratic performance. Moreover, nonvolatile materials can form insulating deposits on the stainless steel surface which distort the electrical fields in the ion source.14 Dodds et al. reported nonlinear behavior of their calibration curves obtained with both EI-QMS and EI ion trap MS, whereas the flame ionization detector (FID) exhibited linear responses for the analysis of fatty acid methyl esters.15 To overcome these problems, efforts have been made to modify or change the ion source material. To our knowledge, modifications are only reported in the patent literature.13,14,16,17 All cited patents have in common that stainless steel sources were coated or completely replaced (7) Chao, B. Y. K.; White, F. A. Int. J. Mass Spectrom. Ion Phys. 1973, 12, 423–432. (8) Grill, V.; Shen, J.; Evans, C.; Cooks, G. R. Rev. Sci. Instrum. 2001, 72, 3149–3179. (9) Kuhnke, K.; Kern, K.; David, R.; Comsa, G. Rev. Sci. Instrum. 1994, 65, 3458–3465. (10) Park, C. J.; Ahn, J. R. Rev. Sci. Instrum. 2005, 76, 044101. (11) Fialkov, A. B.; Steiner, U.; Lehotay, S. J.; Amirav, A. Int. J. Mass Spectrom. 2007, 260, 31–48. (12) Millard, B. J. Quantitative Mass Spectrometry; Wiley: New York, 1977. (13) Kroska, J. D. United States Patent 6,878,932, 2005. (14) Perkins, P. D. United States Patent 7,304,299, 2007. (15) Dodds, E. D.; McCoy, M. R.; Rea, L. D.; Kennish, J. M. Lipids 2005, 40, 419–428. (16) Taylor, D. M.; Amy, J. W.; Stafford, G. C., Jr. United States Patent 5,055,678, 1991.

with another metal or alloy to improve the “inertness” of the source. The described coatings include noble metals and carbon nanotubes, while specialized alloys such as incoloy, inconel, and hastelloy are proposed as entirely new massive ion source materials. As a result, cleaner mass spectra and improved signal intensities are reported. However, no signal drift improvements are claimed. Moreover, as a consequence of patented work, essential research details and quantitative information about occurring processes are lacking. In this work, a common stainless steel ion source was used to investigate signal drifts with EI-QMS. All parts of the ion source which could have a role in signal drift were thoroughly studied, and which played the most important role were investigated. The mechanisms involved were explored and linked to possible negative charge retention on the stainless steel surface. In an attempt to improve the signal stability, the surfaces of the ion source parts which were found to be responsible were electroplated with gold. For each modified part, the gain in signal stability was determined in a quantitative way. EXPERIMENTAL SECTION Reagents and Materials. The purity of all reference chemicals used in this study was more than 99.9% by GC. Methanol, ethanol, and acetone were obtained from Fisher Chemicals (Loughborough, England). Methyl isobutyl ketone (MIBK) was obtained from Acros Organics (Geel, Belgium). Ultrapure water was produced in the laboratory with a Milli-Q system from Millipore (Molsheim, France). The 22 mL headspace vials (HS) and the aluminum crimp caps were obtained from Filter Service (Eupen, Belgium). The gold coating procedures for the repeller, exit plate, and focusing lenses were carried out at Britech Private Ltd. (Hyderabad, India). Standard Mixture. For all experiments, an aqueous standard solution containing methanol and ethanol at a concentration of 10 µL/mL and acetone and MIBK at a concentration of 1 µL/mL was prepared. Sample Preparation. In this study 1.0 mL of the standard solution was pipetted into 22 mL headspace vials which were immediately sealed with aluminum crimp caps with PTFE septa. Instrumentation. All instruments and software of the hyphenated HS-GC/MS system were from Perkin-Elmer (Waltham, MA). The headspace autosampler used in this study was a Turbomatrix HS 40 XL. The static headspace mode was used. The vials were thermostated for 30 min at 60 °C. The pressurization time and injection time were 3.0 and 0.04 min, respectively. The injection pressure was 100 kPa. The temperature of the transfer line was maintained at 80 °C. The GC instrument was an Autosystem XL. The analytical column used was an AT-Aquawax column, 30 m × 0.53 mm ×0.5 µm (Alltech, Deerfield, IL). The bonded polyethylene glycol gives resistance toward aqueous injections. The carrier gas used was helium 5.6 at a flow rate of 4.0 mL/min and a split ratio of 1:5 at the injector. The end of the column was connected to the MS instrument through a split connection. A Y-connector was used for the split, where one outlet was connected to the MS instrument using an undeactivated fused silica transfer line with an internal diameter of 100 µm, whereas the other outlet was connected to (17) Joyce, T. H.; Lu, J. Q. United States Patent 7,075,067, 2006.

Analytical Chemistry, Vol. 82, No. 15, August 1, 2010

6481

the FID using an undeactivated fused silica transfer line with an internal diameter of 530 µm. This led to a flow of 1 mL/min to the MS instrument and 3 mL/min to the FID. The temperatures of the injector and FID were maintained at 140 and 250 °C, respectively. An isothermal temperature program of 50 °C for 10 min was applied. The mass spectrometer coupled to the GC instrument was a Turbomass EI-QMS instrument. The mass spectrometer was equipped with a standard stainless steel EI source. The temperatures of the interface and ion source were 180 and 200 °C, respectively. The chromatograms were recorded in total ion current (TIC) mode with an m/z range of 16-150. All data were acquired and integrated by Turbomass software. The reference compound used for tuning and calibration was perfluorotributylamine (PFTBA). When a cleaned ion source was installed, a software tuning procedure was invoked. This autotune procedure ramps the settings for the tuning parameters until they are optimized to give the best intensity, resolution, and peak shape. Immediately after autotune, the MS instrument was calibrated. Finally, the repeller voltage was manually adjusted to the examined values without adjustment of the other tuning parameters, allowing exclusive monitoring of the influence of the repeller voltage on the signal intensity. Software simulations were carried out with Maxwell SV software (Ansoft, Pittsburgh, PA). RESULTS AND DISCUSSION The retention times of acetone, methanol, ethanol, and MIBK were 2.2, 2.7, 3.1, and 4.0 min, respectively. All four compounds were well separated, and their chromatographic performance remained stable throughout the whole investigation period. All experimental results reported and discussed in this section were obtained with static headspace sampling. As this is a repeatable injection method, this allowed easier data interpretation. Signal Stability Using a Standard Ion Source. A series of 10 vials containing 1.0 mL of the standard solution was injected consecutively using the instrumental parameters mentioned in the Experimental Section. The cycle time was 15 min, making the total analysis time of one sequence 175 min. Figure 2 shows an overlay of chromatograms and the plot of the analyte peak areas obtained with EI-QMS against the injection number. As an example, the FID signal of ethanol was added to the plot. For all other compounds, the FID signal pattern was found to be similar. The peak areas obtained with EI-QMS decreased by more than 50% after 10 injections, while the FID produced repeatable signals. In terms of relative standard deviation (RSD) values, these ranged from 28% for MIBK to 38% for acetone (n ) 10). The results in Figure 2 are shown as a graph so that the progress of the signal decrease can be observed. It is clear that it was not a random phenomenon, but a continuous loss of signal. Moreover, signal drift patterns and total signal loss were slightly dissimilar for the different compounds. Probably many GC/MS users are not aware of the signal drift as the use of an internal standard is the established quantification method, especially with direct injection autosampling. The advantage of an internal standard is that errors are “canceled out” as the magnitude of systematic error is similar for the analyte and internal standard. Thus, in an ideal situation instrumental drift 6482

Analytical Chemistry, Vol. 82, No. 15, August 1, 2010

Figure 2. (A) TIC chromatogram overlay of 10 consecutive injections of standard solution. (B) Analytes’ residual peak areas with respect to the first peak area against the injection number together with the FID signal of ethanol.

affects the analyte and internal standard equally.18 However, with our experiments it was observed that the extent of signal loss and the drift pattern were not the same for different compounds, even for homologue compounds such as methanol and ethanol. As a consequence, independent of the injection method, the use of any other internal standards than isotopically labeled compounds is only “almost right” and thus not good enough for quantitative measurements where high levels of accuracy are needed. All TIC chromatograms obtained from the 10 injections were compared to investigate the possibility of sample molecule adsorption and degradation on the EI source surfaces. The mass spectra at any particular scan point in all TIC chromatograms remained similar. In none of the TIC chromatograms were secondary ions or unexpected ions observed. The same sequence of 10 injections was repeated in quadruplicate with a strict intersequence time gap of 8 h. A fifth sequence was recorded after the ion source was cleaned by polishing. The obtained signal changes are shown in Figure 3. It was observed that for each sequence the residual peak areasswhich are defined throughout the work as the ratio percentage of the final run peak areas over those of the first runswere only about 50%. The signal loss of one sequence was partially recovered the next day (for about 70%), but the signal continued to decrease with the next sequences of repeated injections. The same experiment was repeated with longer time intervals of 16, 32, 72, and 168 h. It was observed that an intersequence time of at least 32 h was necessary to recover a maximum amount (95%) of lost signal from the previous sequence. Signal recovery of more than 95% was never achieved, even with time gaps of up to 1 week. (18) Busch, K. L. Spectroscopy 2007, 22, 14–19.

Table 1. Residual Peak Area (%) after 10 Injections of the Standard Solution Obtained with Increasing Repeller Voltagesa acetone

methanol

ethanol

MIBK

voltage (V)

SS

gold

SS

gold

SS

gold

SS

gold

1 2.5 5

34 77 108

97 107 103

39 81 110

97 109 111

43 83 108

96 108 108

46 78 106

95 103 99

a The results obtained with the stainless steel (SS) ion source are compared with those of the modified source (gold).

Figure 3. Residual peak area of ethanol with respect to the first injection over 5 sequences of 10 injections (A-B, C-D, E-F, G-H, I-J) of the standard solution. The time gap between two sequences was 8 h. Before the last sequence, the ion source was cleaned by abrasive polishing.

From the signal progress shown in Figure 3, two phenomena could be distinguished: short-term signal instability and long-term signal instability. Short-term signal instability was the drastic signal loss within a sequence. These are routes A-B, C-D, E-F, and G-H in the figure. Such short-term signal losses could be partially recovered by using longer time gaps between the sequences. The long-term signal instability was the slow decrease of signal within repeated sequences over longer periods, which is route A-C-E-G in the figure. Such long-term signal decrease never improved unless the ion source was cleaned. This is represented by point I in the figure. The higher peak area at point I compared to that at point A is explained by the fact that the first sequence was not recorded just after an ion source cleaning procedure but after some previous sequences. Instrument Components Responsible for Signal Drift. Several instrument parts were first ruled out as possible causes of signal drift. Although the headspace sampler was not suspected due to the repeatable FID signals, a sequence of 10 injections of the standard solution with direct injection autosampling was carried out. As could be expected, signal drift still occurred with large, random variations in absolute signal intensity. A series of water-free injections was also tried, with no significant improvement. A possible failure of the photomultiplier was ruled out by the fact that the noise levels remained similar while the peak intensities were dropping. Also different multiplier voltages did not have any influence on the stability of the signal. The quality of the mass spectra of the different peaks remained unaffected over time, indicating that the quadrupoles as well as the ionization process were functioning well. These results indicate that the ion source should be responsible for the observed signal drift. The first ion source parameters that were investigated were the filament material (rhenium instead of tungsten) and ion source temperature (from 150 to 250 °C), both without any effect on the signal stability. Further experiments were performed by varying the MS tune parameters. It was observed that the repeller voltage has a large influence on the signal stability. New sequences of 10 repeated injections were performed at repeller voltages of 1, 2.5, and 5 V. Residual peak areas for each compound at different repeller voltages are summarized in Table 1 (stainless steel part). As can be seen, signal drift improved with increasing repeller voltages. With a repeller voltage of 5 V, the RSD (n ) 10) of the peak areas was found to be less than 5% for all investigated analytes, which is within acceptable limits for HS-GC. Also the

signal intensity increased with higher repeller voltages. Therefore, elevated repeller voltages resulted in less signal decrease or even signal increase. However, when an automatic tuning is done with the software, the repeller voltage is never set to values higher than 1.2 V. Hence, when the repeller voltage was manually adjusted to 5 V for these experiments, other tuning parameters such as lens and photomultiplier voltages were no longer optimized. This affects the resolution as the lens voltages are responsible for focusing the ion beam and controlling the ions’ kinetic energy within the optimal range for quadrupole m/z separation.8 Consequently, working at high repeller voltages should be avoided as the mass spectrometer cannot be automatically tuned for these repeller voltages. Another drawback of working with high repeller voltages was an increased cleaning frequency. Dirt spots on the repeller and exit plate were much more intense when the repeller voltage was set at 5 V, compared to dirt spots seen with a similar instrument work load but with a repeller voltage of 1 V. Interlaboratory Investigation. The standard mixture was also analyzed on a similar HS-GC/EI-QMS system of the same brand in a different laboratory by a different analyst. The parameter settings of the HS-GC instrument were identical to our settings. The MS instrument was autotuned and calibrated. The results obtained were similar to those obtained in our laboratory (Figure S-1, Supporting Information). Also the data sets of three other laboratories, working with EI-QMS equipment of various brands, showed a continuous signal decrease for their analytes (Table S-1, Figure S-2, Supporting Information). Hence, the problem seems to be universal for such instrumentation. Surface Modification. It was assumed that the observed signal instability was linked to the stainless steel surface. Therefore, a gold coating of 5 µm was electroplated over a protective nickel layer on the existing surface of our repeller, exit plate, and focusing lenses. Again, sequences of 10 repeated injections of the standard mixture were analyzed. All the experiments were carried out with a fixed repeller voltage of 1 V. Four different setup combinations (A-B-C-D) of original parts and gold-coated parts were investigated. The obtained results are shown in Figure 4. For each compound, the residual peak area after 10 injections is reported. Situation A represents an unmodified ion source. The residual peak area for each compound can be found in Table 1 (stainless steel, 1 V). In situation B, the stainless steel focusing lenses were replaced by gold-coated ones. The stainless steel repeller and exit plate remained in place. It was observed that the residual peak areas increased by 10-20% compared to those Analytical Chemistry, Vol. 82, No. 15, August 1, 2010

6483

Figure 4. Residual peak areas of the 4 compounds after 10 consecutive injections of standard solution with 4 different setups: (A) original ion source, (B) gold-coated focusing lenses (stainless steel repeller and exit plate), (C) gold-coated repeller and exit plate (stainless steel focusing lenses), (D) gold-coated repeller, exit plate, and focusing lenses.

in situation A. Situation C represents the use of a gold-coated repeller and exit plate, but stainless steel focusing lenses. The residual peak area is higher than 80% for all compounds, which is already a drastic improvement compared to that of the original setup. Finally, in situation D, the repeller, exit plate, and focusing lenses were all replaced with gold-coated parts. With this setup, the signal remained stable with all residual peak areas equal to or above 95%. For all compounds, RSD values were lower than 2% (n ) 10). The sequences of situation D were repeated at different repeller voltages as was done with all the original parts. The results are summarized in Table 1. As can be seen, with the gold-coated parts the signal rather increased at 2.5 and 5 V. However, compared to the results obtained with the original source (situation A), the signal stability is clearly better at 1 and 2.5 V. After a work load of 8 weeks, the gold-coated repeller and exit plate were visually inspected. No coloration or dirt spots could be observed. Hence, the maintenance frequency could be drastically decreased compared to that of stainless steel parts. Moreover, the cleaning procedure itself is less time-consuming for goldcoated surfaces compared to stainless steel surfaces. Sonication in dilute ammonia for 1 min, followed by a citric acid solution for 5 min, and afterward rinsing with methanol does the job perfectly, while for stainless steel polishing with abrasives is inevitable. Consideration about the Cause of Signal Drift Observed with Stainless Steel EI Source Parts. The obtained results clearly indicate that there should be a link between the observed signal drift and the stainless steel surface of the repeller, exit plate, and focusing lenses in an operating EI source. However, it is likely to be a series of events that contribute to signal drift. Therefore, a step-by-step approach to potential mechanisms is presented based on software simulation and literature findings. First, the three key functions of the ion source were considered. Ionization and subsequent fragmentation must occur in a stable and repeatable manner. No aberrations were found in the analyte mass spectra, except the decreasing ion abundances. Moreover, no unexpected ions were found in all scans of a TIC chromatogram. Hence, the ionization and fragmentation appeared to function properly. Second, electric fields generated between the repeller and exit plate must push positive ions toward the mass analyzer. The equipotential lines of the electric fields in an ion source are 6484

Analytical Chemistry, Vol. 82, No. 15, August 1, 2010

depicted in Figure 1B. Perturbations in these electric fields will directly affect the shape and intensity of the ion beam toward the mass analyzer. Finally, the electric fields created by the focusing lenses should focus the ion beam and give the ions the optimum kinetic energy for quadrupole-based mass separation.8 When the electric fields are influenced at the lens system, the kinetic energy of the ions is no longer optimized and resolution will be affected. This could be observed when the repeller voltage was set at 5 V, without optimization of the lens voltages. However, within a sequence at a fixed repeller voltage, no resolution decrease was noticed. Hence, the attention was focused on the possibility of altered electric fields in the ion volume due to changing surface properties of stainless steel when molecules were injected inside the ion source. From all the charged species entering or created in the ion source, the cations are repelled toward the mass analyzer, the neutrals are removed by the vacuum, and the anions are attracted toward the positively charged repeller, where they should be discharged rapidly. Although much less abundant than positive ions, also negative ions are formed through the electron ionization process.3,4 With the repeller voltages used in this work, the formed negative ions will reach the repeller at energies around 1-5 eV. According to a review by Grill et al., these energies are in the lower part of the hyperthermal energy range.8 In this energy regime, molecules can be deposited onto surfaces by physisorption, chemisorption, or soft landing.19-21 Of particular interest is the possible charge retention of deposited ions on semiconducting or insulating surfaces. For example, in the work of Miller et al., it was demonstrated that intact polyatomic ions were trapped onto a self-assembled monolayer surface for many days.22 As a consequence, a potential difference will develop over the charged layer, which would interfere with electric fields for ion beam focusing.23 Next it was decided to consider similar events in a stainless steel EI source. The inert and corrosion-resistant nature of stainless steel is due to the presence of an invisible, nanoscale (±5 nm) layer of chromium oxide, which was characterized with Auger electron spectroscopy and X-ray photoelectron spectroscopy.24,25 This layer is commonly referred to as the oxide or passive layer on stainless steel. The Cr2O3 surface predominance has also been verified in a high-temperature vacuum.26,27 The thin chromium oxide film has poorer electrical conductivity than the underlying base metal.28,29 With a band gap in a range of 2.7-2.9 eV, Cr2O3 is positioned between semiconductors and insulators. A relative (19) Cooks, R. G.; Ast, T.; Mabud, A. Int. J. Mass Spectrom. Ion Processes 1990, 100, 209–265. (20) Gologan, B.; Green, J. R.; Alvarez, J.; Laskin, J.; Cooks, R. G. Phys. Chem. Chem. Phys. 2005, 7, 1490–1500. (21) Silbey, R. J.; Alberty, R. A. In Physical Chemistry, 3rd ed.; John Wiley & Sons, Inc.: Hoboken, NJ, 2001; pp 865-870. (22) Miller, S. A.; Luo, H.; Pachuta, S. J.; Cooks, R. G. Science 1997, 275, 1447. (23) Laskin, J.; Wang, P.; Hadjar, O. Phys. Chem. Chem. Phys. 2008, 10, 1079– 1090. (24) Wilde, M.; Beauport, I.; Stuhl, F.; Al-Shamery, K.; Freund, H.-J. Phys. Rev. B: Condens. Matter 1999, 59, 13401–13412. (25) Cole, C. R.; Outlaw, R. A.; Champion, R. L.; Holloway, B. C.; Kelly, M. A. Appl. Surf. Sci. 2007, 253, 3789–3798. (26) Park, R. L.; Houston, J. E.; Schreiner, D. G. J. Vac. Sci. Technol. 1972, 9, 1023. (27) Rezaie-Serej, S.; Outlaw, R. A. J. Vac. Sci. Technol., A 1994, 12, 2814.

Figure 5. (Left) Frontal view of a freshly polished, clean repeller. (Right) Frontal view of the repeller after an operational period of about 8 weeks. The intensely colored zone on top was closest to the filament and thus was exposed to excessive heat.

permittivity, ε, of 12 for vacuum-deposited films is reported.30 The temperature and partial oxygen pressure are two main factors affecting the thickness and morphology of the oxide layer.30-32 In normal operating conditions, the source is heated to temperatures around 200 °C. Under these conditions, the initial layer will grow, creating the typical interference colors of the surface. From this color pattern, a minimum thickness of about 100 nm is assumed.33 However, the proximity of the filament which reaches a temperature of around 1500 °C can lead to hot spots on the repeller. This was confirmed by the existence of more intense colored zones on the repeller surface closest to the filament. At these spots, the oxide layer thickness should be much higher than 100 nm.28 A picture is shown in Figure 5. An oxide layer is not suspected to directly influence the electric fields inside the ion source. However, the insulating properties of this thin film will inhibit charge transfer toward the ground potential. In the case of negative ions, the persisting negative charge on the repeller surface may distort the electric fields. In what follows, a study was performed simulating the Cr2O3 thin film and the buildup of a negatively charged layer on that Cr2O3 film of the repeller. Compared to the simulation model of Figure 1B, an insulating layer of 200 nm thickness was placed on top of the repeller surface. As expected, it was found that an insulation layer does not affect the electric fields inside the ion source. In the next step, an additional negatively charged layer was placed over the insulation film. For this purpose, only the negative oxygen ions originally coming from the air in the vial are taken into account as they are the most abundant with headspace injections. First, the amount of negative oxygen ions which are formed with every injection was estimated. Assuming an injection volume of 1 mL and 18% O2 presence in the air and taking into account the 1:5 injector (28) Li, Z. L.; Zheng, H. Y.; Teh, K. M.; Liu, Y. C.; Lim, G. C.; Seng, H. L.; Yakovlev, N. L. Appl. Surf. Sci. 2009, 256, 1582–1588. (29) Ferreira, M. G. S.; Da Cunha Belo, M.; Hakiki, N. E.; Goodlet, G.; Montemor, M. F.; Simo ˜es, A. M. P. J. Braz. Chem. Soc. 2002, 13, 433–440. (30) Goodlet, G.; Faty, S.; Cardoso, S.; Freitas, P. P.; Simo ˜es, A. M. P.; Ferreira, M. G. S.; Da Cunha Belo, M. Corros. Sci. 2004, 46, 1479–1499. (31) Cho, B.; Choi, E.; Chung, S. Appl. Phys. A: Mater. Sci. Process. 1999, 69, 625–630. (32) Zielin´ski, W.; Kurzydłowski, K. J. Scr. Mater. 2000, 43, 33–37. (33) Va´zquez-Santoyo, L. D.; Pe´rez-Bueno, J. J.; Manzano-Ramı´rez, A.; GonzalezHerna´ndez, J.; Pe´rez-Robles, J. F.; Vorobiev, Y. V. Inorg. Mater. 2005, 41, 955–960.

Figure 6. Simulation of the equipotential lines of the electric fields inside the ion volume: (A) repeller voltage of 1 V without a negative charge on the surface, (B) repeller voltage of 1 V with a negative charge, (C, D) similar to (A) and (B) but with a repeller voltage of 5 V.

split and 1:3 Y-split, about 2 × 1017 oxygen molecules enter the ion source with every injection. Subsequently, a value of 1:300 was taken as the ionization efficiency.34 Next, it was estimated that the ratio of O+ of O- was about 1000. An exact figure could not be found in the literature, but the value was estimated on the basis of negative ion formation of alcohols with electron impact.4 With the above-mentioned estimations, 6.7 × 1010 negative oxygen ions are formed, leading to a total charge of about 1 × 10-8 C per injection. Similar to the ideal situation in Figure 1B, the electric fields were simulated when a negative charge of 1 × 10-8 C was applied to the repeller surface. This charge implied that all negative oxygen ions were adsorbed to the surface and retained their charge. In reality this is an overestimation as part of the ions will be scattered or discharged, but these processes are very difficult to predict and are beyond the scope of this study. On the other hand, other negative ions resulting from analyte or solvent (H2O) ionization are not taken into account. Moreover, the most relevant outcome of the simulation is the relative effect, rather than true values. Therefore, it was decided to fix the charge estimation at 1 × 10-8 C per injection. The outcomes of the simulation model with the charge buildup are shown in Figures 6 and 7. In Figure 6, the equipotential lines inside the ion volume are shown in four different situations. In situations A and B, the repeller voltage was set at 1 V, while it was 5 V for situations C and D. For both repeller voltages, the first simulation is without negative charge on the repeller surface (A and C). The other simulations (B and D) are done with a negative charge of 1 × 10-7 C on the repeller surface, which is virtually present after a sequence of 10 injections assuming all the negative charges are retained. It can be seen that, with a repeller voltage of 1 V, charging of the repeller dramatically decreased the electric field strengths. With a repeller voltage of 5 V, however, electric fields are less affected by the same charge. To have a (semi)quantitative idea about the influence of repeller charging on electric field disturbance, the potential differences inside the ion volume were plotted against the position on the x-axial direction. This was done for four different repeller voltages ranging from 1 to 5 V. These plots are shown in Figure 7. In plot A, the ideal situation similar to Figure 1B was simulated at repeller voltages of 1, 2, 2.5, and 5 V. In plot B, the fixed negative charge was applied on the repeller surface with identical voltages. It can be seen that, for a repeller voltage of 1 V, the tension is completely (34) Busch, K. L. Spectroscopy 2006, 21, 14–19.

Analytical Chemistry, Vol. 82, No. 15, August 1, 2010

6485

Figure 7. 2D plots of the potential differences in the ion volume against the position in the x-axial direction. The x-axis in the plots exactly corresponds to the line a-b in the ion source drawing on top of the figure. Plot A is the “ideal situation” without charge retention on the repeller surface, simulated for four different repeller voltages. Plot B is the same as plot A but with a negative charge of 1 × 10-7 C applied on the repeller surface, which is the virtual charge after 10 injections. Plot C represents the progressive influence of every injection on the electric fields inside the ion source for a sequence of 10 injections at a repeller voltage of 1 V, assuming complete charge retention on the repeller surface. Finally, plot D is similar to plot C but at a repeller voltage of 5 V.

countered by the charged layer. Moreover, the voltage only turns positive at a distance of about 3 mm from the repeller surface, only because of the influence of the anode and lens voltages. The plot also revealed that the shape of the curve changes, the potential difference reaching a minimum value in the middle of the ion volume. As a consequence, the direction vectors of the electric fields were changed. With higher repeller voltages, the potential differences in the ion volume also decreased, but the shape of the curve and the electric field direction vectors were less or not affected. In plots C and D, the progressive charge buildup on a repeller with every injection is shown for repeller voltages of 1 and 5 V, respectively. These outcomes clearly suggest that a progressive negative charge buildup on the repeller surface dramatically affects the electric fields inside the ion source at low repeller voltages. With higher repeller voltages, the electric field strengths also decrease but probably without a significant influence on the ion beam. In reality, progressively less amount of charge will be retained due to Coulomb repulsion, surface saturation, etc. until a balance is reached between adsorption and desorption.21 This may explain the nonlinear curve of the signal drifts observed at the standard repeller voltage of 1 V.

6486

Analytical Chemistry, Vol. 82, No. 15, August 1, 2010

CONCLUSIONS With this study it was confirmed that EI-QMS coupled to HSGC suffers from signal instability. Hence, accurate quantitative measurements may not be possible with instrumentation having a stainless steel EI source. It was found that the key parts responsible for the signal drift were those regulating the ion beam trajectory and ion beam focusing. Moreover, the mechanism involved probably deals with altered electric fields due to the surface characteristics of stainless steel in a stressful environment such as an EI source. Where the reasons for material change and related improvements are rather vague in the cited patents, the contribution of each part of the ion source was determined quantitatively in this study. With the typical repeller voltage setting of 1 V, the modified EI source produced a stable signal with improved resolution compared to the stainless steel EI source. Although the signal stability was already radically improved by modifying the repeller and exit plate, complete stability was only achieved by an additional modification of the focusing lenses. Using the modified EI source, instrument repeatability remarkably improved as relative standard deviation values for repetitive analyses were as low as 2%. From the results obtained it can be concluded that a simple and relatively cheap intervention such as applying a gold coating on three ion source parts radically improved the quantitative performance of the EI-QMS instrument, whereas the maintenance frequency and cleaning work were decreased. As a last part of the work, simulations revealed that negative charge retention on the repeller surface may cause the observed signal drifts. Although negative ion formation is thought to be negligible in EI source operation, it may cause serious electric field distortions assuming that (part of the) negative ions are not discharged immediately on the repeller surface. This assumption was made on the basis of similar findings of charge retention on insulating surfaces in other research fields. ACKNOWLEDGMENT We greatly acknowledge the Laboratory for Food Analysis, Ghent, the Laboratory for Toxicology, Ghent, the Laboratory for Toxicology, Antwerp, and Perkin-Elmer Co. for supplying helpful data for this work. We also thank Britech Private Ltd. for the gold platings. Financial support for this project was given by a research grant of the Research FoundationsFlanders (FWO). SUPPORTING INFORMATION AVAILABLE Additional information as noted in text. This material is available free of charge via the Internet at http://pubs.acs.org. Received for review March 26, 2010. Accepted June 15, 2010. AC100780S