Efficient interpretation of thermal desorption data for complex surface processes is often complicated further by species desorbing from heating elements, support materials, and sample holder parts. Multivariate curve resolution (MCR) can be utilized as an unbiased method to assign specific temperature-dependent profiles for evolution of different species from the target surface itself as opposed to traces evolving from the surroundings. Analysis of thermal desorption data for iodoethane, where relatively low exposures are needed to form a complete monolayer on a clean Si(100)-2 × 1 surface in vacuum, provides convenient benchmarks for a comparison with the chemistry of chloroethane on the same surface. In the latter set of measurements, very high exposures are required to form the same type of species as for iodoethane, and the detection and analysis process is complicated by both the desorption from the apparatus and by the presence of impurities, which are essentially undetectable during experiments with iodoethane because of low exposures required to form a monolayer. Thus, MCR can be used to distinguish desorption from the sample and from the apparatus without the need to perform complicated and multiple additional desorption experiments.

Temperature programmed desorption (TPD) has been used for decades for analyzing reaction mechanisms in surface science. This technique provides information about species desorbing from the target surface during thermal treatment. Most of the time, a linear temperature ramp is maintained to simplify kinetic analysis following the thermal desorption process. The desorbing species are ionized, and the resulting mass-to-charge (m/z) ratios are recorded by mass spectrometer with the corresponding intensities plotted against the sample temperature. Kinetic and thermodynamic parameters can be extracted from such experiments using a very straightforward Redhead method1 or sometimes more complex approaches.2,3 The key component of this analysis is the ability to assign the cracking patterns produced by a mass spectrometer to specific molecular species and then to construct the mechanism of surface processes that would yield such species and desorb them into a gas phase. When a reaction requires relatively low exposures and the mechanism of the surface process is simple, such as in desorption without chemical transformation4 or if a single reaction channel is available,5 the interpretation of TPD data is often straightforward. However, when extremely high exposures of a reactant are required or if the surface reaction mechanism is complicated and yields a large number of different products, the role of experimental setup design becomes very important. In order to minimize the effect of surroundings (sample holder, sample heater, mounting materials, etc.), temperature-programmed desorption experiments can be performed in ultrahigh vacuum with a mass-spectrometer encased in a differentially pumped shield, with only a small aperture positioned very closely to the middle of a solid sample under investigation. The assumption then is that only the species desorbing from the center of this sample are registered by the mass spectrometer during TPD. Although this approach does minimize the effect of the manipulator parts that do not directly face the mass spectrometer, eliminating the possible desorption from mounting materials, such as, for example, clips that may be used to attach the sample to the sample holder or heater, is much more difficult.

This paper will utilize a mathematical procedure based on multivariate curve resolution (MCR) and basic knowledge of surface processes to distinguish desorption from the target sample as opposed to the desorption from manipulator parts or mounting materials. MCR is a mathematical method used to analyze spectral data in the form of a matrix. The earliest work on this method was first reported in the early 1970s by Lawton and Sylvestre.6,7 More recently, this approach has been applied to analyzing spectral data in step-scan FT-IR and 2D-NMR.8,9 Complex temperature-programmed desorption data has also been analyzed by MCR, and previously reported work from our group utilized this method, both to distinguish desorption spectra of multiple products, and to analyze a product without a readily available mass spectrum in the literature.10 In MCR, a spectrum of different m/z traces as a function of temperature is converted into two matrices: one representing the cracking patterns of desorbing compounds (or components, where each component can be comprised of several compounds) and the other containing information about the evolution of these components as a function of temperature, as detailed in Sec. II.

In this work, we present the use of MCR for analysis of the thermal reaction of chloroethane with a clean Si(100)-2 × 1 surface, as compared with the similar reaction of iodoethane with the same surface. In both cases, halogen-terminated silicon surface with a well-defined coverage of the halogen is one of the reaction products. Based on previous studies,11 the adsorption and temperature-dependent transformations of haloethanes on silicon are summarized in Fig. 1.

Fig. 1.

Schematic representation of adsorption and thermally induced reaction of haloethanes on a clean Si(100) surface. The halogen (iodine or chlorine) is denoted as X.

Fig. 1.

Schematic representation of adsorption and thermally induced reaction of haloethanes on a clean Si(100) surface. The halogen (iodine or chlorine) is denoted as X.

Close modal

Thermal annealing of this surface releases ethylene as a result of β-hydrogen elimination and then H2 as a result of recombination of surface hydrogen. Generally speaking, the reaction of chloroethane would be expected to follow the same overall mechanism; however, as will be shown below, a much higher exposure is required to saturate a monolayer and thus the interpretation of the corresponding TPD traces is substantially more complicated by the desorption from mounting materials, heater, and sample holder. MCR is used to distinguish these processes without additional experiments by analyzing the desorption profiles and applying basic chemistry knowledge to this analysis.

For temperature-programmed desorption studies, a p-type double side polished Si(100) crystal sample (10 mm in diameter and 0.5 mm thickness, Wafer World) was mounted on a standard button heater (HeatWave Labs Inc.) made of molybdenum with a standard tantalum cup (Model 101028). This design leaves approximately 9 mm diameter aperture for the sample but also exposes the edges of the tantalum sample holder cup on the same side as the sample itself. This assembly was attached to a copper block of the sample manipulator and placed into an ultra-high-vacuum chamber with base pressure below 1 × 10−9 Torr. This chamber is equipped with a differentially pumped mass spectrometer (Hiden Analytical), with the detection range from 0 to 510 amu, for temperature-programmed desorption studies, an Auger electron spectrometer for confirming the cleanliness of the silicon sample, and a low-energy electron diffraction setup for confirming surface order following the preparation procedure. The sample was prepared by ion sputtering with argon (99.9999%, Matheson) and 20-min annealing to 1000 K. This procedure has been shown previously to result in a clean and well-ordered Si(100)-2 × 1 surface.11 Iodoethane (99%, Sigma-Aldrich) used in the experiment was purified by three freeze-pump-thaw cycles before being introduced into the chamber via a leak valve. The chloroethane-d5 gas (99%, Sigma-Aldrich) and ethylene (Matheson, research purity) were used without additional purification and introduced via a leak valve. All exposures are reported in Langmuirs (1 L = 10−6 Torr s) without additional corrections to the ion gauge sensitivity.

The exposure used for chloroethane-d5 (5000 L) corresponds to a saturated monolayer when dosed onto a clean Si(100)-2 × 1 surface at room temperature, according to the recorded experimental exposure profile. For iodoethane, exposure of 10 L was used to ensure submonolayer coverage, based on a separate experimental coverage profile. The TPD following this exposure produces all the key components still easily detectable in the setup described above. All TPD experiments with chloroethane were performed at a constant heating rate of 2 K/s following 25 simultaneously collected traces (m/z ratios) over the range of m/z = 2 to m/z = 170. All doses were performed at room temperature, and thermal desorption traces were collected between 300 and 1080 K. All experiments were performed with 70 eV electron ionization energy. For iodoethane, m/z in the range of m/z = 2 to m/z = 409 were followed.

The multivariate analysis presented here was performed using the pls_toolbox 7.8 software (Eigenvector Research, Inc., Manson, WA) for matlab (Mathworks, Inc., Natick, MA). The complete step-by-step procedure for the analysis is described in a free downloadable manual available at http://www.udel.edu/chem/teplyakov/MCRGuide.pdf. Briefly, the scheme is summarized in Fig. 2.

Fig. 2.

Schematic representation of MCR analysis as applied to a set of temperature-programmed desorption data. The data representing a collection of desorption traces for selected m/z ratios are mathematically decoupled into the spectral matrix that provides the mathematically identified mass spectra (that can include one or more compounds desorbing from a surface) and contribution matrix that describes the temperature-dependent desorption behavior of the components identified in the spectral matrix.

Fig. 2.

Schematic representation of MCR analysis as applied to a set of temperature-programmed desorption data. The data representing a collection of desorption traces for selected m/z ratios are mathematically decoupled into the spectral matrix that provides the mathematically identified mass spectra (that can include one or more compounds desorbing from a surface) and contribution matrix that describes the temperature-dependent desorption behavior of the components identified in the spectral matrix.

Close modal

The first set of TPD experiments is used as a benchmark and describes a well-understood reaction of iodoethane on Si(100)-2 × 1. The reaction has been investigated previously11 and the mechanism involves dissociation of a C–I bond, followed by the attachment of ethyl group and iodine to surface silicon atoms. The exact arrangement of these entities on the Si–Si dimers of the 2 × 1 reconstruction of the Si(100) surface may be a subject of further studies;12 however, it has been established that a relatively low exposure of iodoethane should be sufficient to saturate a monolayer and that following the initial dissociation, surface ethyl groups produced by this process upon heating yield ethylene and surface-bound hydrogen, which in turn recombines into H2 and desorbs around 800 K. In the setup described above, 10 L exposure is not sufficient to form a complete monolayer; however, it is sufficient to produce detectable amounts of ethylene and hydrogen in TPD spectra. During the dosing, the pressure of iodoethane was kept at 2 × 10−7 Torr and the selected resulting TPD traces are shown in Fig. 3. The relatively low exposure should help to minimize the impurities dosed into the chamber together with the target compound and the possibility of this compound being adsorbed onto parts of the manipulator and sample holder. The obtained TPD data are presented in Fig. 3. The characteristic peaks corresponding to ethylene desorption in a reaction-limited β-hydrogen elimination process are found in TPD spectra at approximately 700 K, in line with the previously reported temperatures,11,13 and a clearly detectable m/z = 2 peak indicative of hydrogen formation is recorded at 800 K.

Fig. 3.

Temperature programmed desorption following exposure of 10 L of iodoethane onto a clean Si(100) surface. Twenty-two masses (m/z = 2, 15, 16, 25, 26, 27, 28, 29, 30, 41, 42, 43, 44, 127, 128, 139, 140, 141, 155, 156, 254, and 409) are monitored simultaneously. All plots are zero-baseline corrected and smoothed.

Fig. 3.

Temperature programmed desorption following exposure of 10 L of iodoethane onto a clean Si(100) surface. Twenty-two masses (m/z = 2, 15, 16, 25, 26, 27, 28, 29, 30, 41, 42, 43, 44, 127, 128, 139, 140, 141, 155, 156, 254, and 409) are monitored simultaneously. All plots are zero-baseline corrected and smoothed.

Close modal

These data were used as a benchmark to test the applicability of the MCR analysis to the iodoethane/Si(100)-2 × 1 system. After applying MCR analysis to the spectral data with 171 temperature points and 22 m/z traces, two components were found to be sufficient to carry all the information from thermal desorption data. This information is summarized in Fig. 4. It identifies two components and includes 99.08% of the variance. The residual from this analysis, which carries only 0.92% of information, is not shown, as it corresponds to random fluctuations, which can be attributed to noise. In component 1, m/z = 2 is the most intense fragment with the desorption temperature of 800 K. Since this component clearly represents desorption of molecular hydrogen following surface recombination, all the m/z signatures above 2 (with very low intensity) can be eliminated from further description of this component. In other words, since the cracking fragment of molecular hydrogen should only have m/z = 1 and 2, with hard constraints applied, all the m/z signatures above 2 (with very low intensity) can be completely removed from further consideration of this product of reaction. Component 2 carries the thermal desorption information on C2H4. The most reliable part of this component, the ratio of intensities for m/z = 26 and 27, agrees with the data provided on the National Institute of Standards and Technology (NIST) standard database.14 The absolute intensity (before correcting for background) of m/z = 28 in our experiments is noticeably higher than that expected for ethylene. This issue is common for most thermal desorption studies, since background CO contributes to this mass-to-charge ratio. This is also why the m/z = 28 is often avoided in interpretation of surface reactions by thermal desorption. However, in this scenario, the shape of the thermal desorption peak is nearly identical for m/z = 26, 27, and 28. Thus, it can be concluded that in the original TPD result, the peaks for m/z = 26, 27, and 28 are all related to ethylene thermal desorption. Thus, component 2 can be identified as ethylene, as shown by the histogram spectrum in Fig. 4 and also compared with the mass spectrum of ethylene dosed into the same apparatus at 1 × 10−6 Torr, as shown at the bottom of the spectral matrix in Fig. 4. It should be emphasized that the absolute intensity of the “components” does not necessarily correspond to their absolute concentrations and should be calibrated separately. Thus, in summary, from the system of iodoethane adsorption on clean Si(100) surface, two components (hydrogen and ethylene) are found to indicate desorption at their expected temperatures.

Fig. 4.

(Color online) Multivariate curve resolution analysis of thermal desorption data for iodoethane adsorption on Si(100). Spectral matrix (a) with two components cracking patterns and contribution matrix (b) with two components thermal desorption spectra are displayed. Component 1, which is confirmed to be hydrogen, uses a hard constraint to remove all other low intensity m/z traces. Component 2 corresponds to ethylene desorption from the surface. This analysis is compared to the pure ethylene mass spectrum collected in the same chamber.

Fig. 4.

(Color online) Multivariate curve resolution analysis of thermal desorption data for iodoethane adsorption on Si(100). Spectral matrix (a) with two components cracking patterns and contribution matrix (b) with two components thermal desorption spectra are displayed. Component 1, which is confirmed to be hydrogen, uses a hard constraint to remove all other low intensity m/z traces. Component 2 corresponds to ethylene desorption from the surface. This analysis is compared to the pure ethylene mass spectrum collected in the same chamber.

Close modal

In a separate set of exposure profile measurements for chloroethane, it was found that much larger exposures are needed to form a saturated monolayer of this compound on Si(100)-2 × 1 compared to those for iodoethane. These high exposures may cause additional obstacles in interpreting TPD data. Thus, a deuterated compound was used in this case to avoid overlap with the components of the background gases. As an example, 5000 L of chloroethane-d5 was dosed into a vacuum chamber to react with the clean Si(100)-2 × 1 surface and to form a saturated monolayer. This dose was achieved by introducing 5 × 10−6 Torr of chloroethane-d5 into the vacuum chamber for 1000 s. At such a large dose, it is possible that pumping gas present in the background of the gas line could affect the adsorption process. The constituents of this background could interact with a clean silicon surface in competition with the chloroethane, but they could also adsorb on parts of the manipulator, sample holder, or sample mounting assembly and then interfere with the TPD traces collected during desorption measurements. Figure 5 shows the original TPD data. Baselines for all 25 collected spectra were adjusted to zero by linear functions. From Fig. 5, m/z = 2 and 4 desorption peaks at 800 K correspond to hydrogen and deuterium thermal desorption from silicon. Since hydrogen is not produced in a β-hydrogen elimination of ethyl-d5 formed on a silicon surface following C-Cl dissociation of chloroethane-d5, m/z = 2 can either represent the hydrogen generated by impurities or a cracking pattern of the desorbing D2. Additional contribution to these desorption traces coincides with the positions of the corresponding peaks from m/z = 26, 28, 30, and 32 traces, the characteristic m/z traces of ethylene-d4. However, one of these peaks, the one corresponding to m/z = 28 peak, may also have contributions from background gases. With such a high dose required, molecular adsorption on the target surface and on the surroundings is likely to occur. The molecular desorption was not found to contribute to the desorption process for the small exposures of iodoethane described above; however, for chloroethane-d5, it may become a rather prominent signature. In addition, CO from the background can affect the interpretation of the desorption data.

Fig. 5.

Temperature programmed desorption following exposure of 5000 L of chloroethane-d5 onto a clean Si(100) surface. Twenty-five different traces (m/z = 2, 4, 20, 26, 28, 30, 32, 34, 35, 37, 39, 44, 46, 48, 50, 51, 53, 69, 70, 71, 72, 133, 135, and 170) were monitored simultaneously. All plots were zero-baseline corrected and smoothed.

Fig. 5.

Temperature programmed desorption following exposure of 5000 L of chloroethane-d5 onto a clean Si(100) surface. Twenty-five different traces (m/z = 2, 4, 20, 26, 28, 30, 32, 34, 35, 37, 39, 44, 46, 48, 50, 51, 53, 69, 70, 71, 72, 133, 135, and 170) were monitored simultaneously. All plots were zero-baseline corrected and smoothed.

Close modal

For example, one possible interpretation of trace for m/z = 28, which is observed to have a very intense peak in this set of studies, could be that it is caused by CO desorption from sample mounting materials. CO adsorption had been studied in detail on Ta, Mo, W, and Cu,15–18 which are the main metallic components present in the mounting assembly for the silicon crystal used in this set of studies. Depending on the sample mounting approach, some of this contribution could be ruled out by repeating the TPD experiments with a sample not directly facing the shield aperture of a differentially pumped mass spectrometer. However, in the design used here and in the majority of designs used for silicon mounting, a part of the mounting assembly (specifically, metallic clamps, clips, or cup-shaped holders) susceptible to contributing to the TPD spectra is only exposed to the mass spectrometer at the same time as the target surface itself. To alleviate all these problems, MCR may become a reliable and unbiased method to identify which of these fragments correspond to the reaction with a target surface and which may be the result of the high dose interacting with the sample itself as opposed to the mounting materials and sample holder.

MCR was applied to the TPD spectra collected following exposure of a clean Si(100)-2 × 1 surface to 5000 L of chloroethane-d5. The spectra presented in Fig. 5 were used as data carrying 157 temperature points and 25 traces monitored in the range from 300 to 1000 K. The result obtained following MCR application with the component information (spectral matrix) and thermal desorption information (contribution matrix) is shown in Fig. 6.

Fig. 6.

Spectral matrix (a) and contribution matrix (b) following MCR analysis of chloroethane-d5 reaction with a clean Si(100) surface. The percentage in each component represents the distribution of variance with respect to data average. Residuals in the analysis presented in the top parts correspond to 0.38%. Component 1 corresponds to ethylene-d4, component 2 is D2, component 3 is CO desorption.

Fig. 6.

Spectral matrix (a) and contribution matrix (b) following MCR analysis of chloroethane-d5 reaction with a clean Si(100) surface. The percentage in each component represents the distribution of variance with respect to data average. Residuals in the analysis presented in the top parts correspond to 0.38%. Component 1 corresponds to ethylene-d4, component 2 is D2, component 3 is CO desorption.

Close modal

In this MCR analysis, the contribution matrix is plotted as 157 temperature points versus three components and the spectral matrix is plotted as three components versus 25 m/z traces. Components 1 and 2 represent ethylene-d4 (based on a comparison of the obtained cracking pattern of ethylene-d4 with the data in the NIST database14) and D2, respectively. Component 3 corresponds predominantly to the m/z = 28. The attempt to use component 4 in this fitting procedure resulted in a mass spectrum of a residual C2D5Cl followed as a function of sample temperature; however, this last component does not exhibit any desorption peaks and thus carries only information about the background change in the chamber as the sample is being heated. Since a 4-component fit includes this background variation, a 3-component MCR fitting procedure results in a complete description of the desorption of ethylene-d4, D2, and CO, as shown in detail in Fig. 6. The behavior of ethylene-d4 and D2 is identical to that for desorption of ethylene and H2 following iodoethane exposure to the same silicon surface. An important matter of interest, however, is the correct assignment of the m/z = 28 evolution. It is clearly the dominant signature in the mass spectrum of component 3 in Fig. 6; however, it is also expected to be present in the spectrum of deuterated ethylene. Constraining the observed mass spectrum of component 3 to represent ethylene-d4 could help this assignment unambiguously. However, in either case, a very substantial contribution of m/z = 28 clearly corresponds to a different component. Thus, component 3 would only correspond to CO evolution as a function of sample surface temperature. Now the question is: what is the origin of this desorption feature?

The desorption peak with m/z = 28 could only originate from five possible sources: (1) chloroethane-d5 desorption following its reaction with the silicon surface; (2) chloroethane-d5 desorption from the sample holder/heater/mounting on the silicon sample; (3) desorption of CO following pumping gas (predominantly CO) adsorption on the silicon surface; (4) desorption of CO following pumping gas (predominantly CO) adsorption on the sample holder/heater/mounting of the silicon sample; and (5) chloroethane-d5 and pumping gas (CO) desorption from the chamber walls.

In order to differentiate between these possibilities, TPD spectra were collected following the 5000 L exposure of chloroethane-d5 onto a clean Si(100)-2 × 1 surface with the sample facing away from the aperture of the mass spectrometer shield. However, in that case, both the signature of D2 evolution and that of the ethylene-d4 desorption decreased in intensity. Thus, both components seem to represent desorption from the face of the silicon sample and not from the chamber walls or parts of the sample holder that do not face the mass spectrometer. In other words, the observed desorption should either originate from the silicon sample itself or from the sample mounting elements that are very close to the sample and are facing the mass spectrometer at the same time as the sample. The possibility of ethylene desorption from tantalum could also be excluded, because in the previous studies, it was onfirmed that ethylene formed on tantalum would be dehydrogenated completely by 600 K.19 Thus, the questionable component of the m/z = 28 fragment corresponds to CO and does not represent interaction of chloroethane-d5 with any part of the vacuum chamber or sample manipulator or holder.

Further characteristics of this desorption product can be obtained by following the exposure-dependent evolution of components 2 and 3 from Fig. 6, as summarized in Fig. 7. Clearly, these two components have very different exposure profiles (thus supporting two different compounds as opposed to different behaviors of the same compound). If component 2 reaches a maximum around 2000 L exposure of chloroethane-d5, component 3 continues to increase linearly in signal intensity up to 5000 L, the highest exposure studied. This behavior of component 3 would also be consistent with the behavior of a low-concentration impurity, whose adsorption on a surface increases linearly as a function of exposure of the main component but does not effectively compete with this main reagent, most likely for kinetic reasons. Thus, if the behavior of component 3 reflects desorption of CO, what could be the source of this product? If this desorption followed the adsorption of pumping gas (which is predominantly CO) on a clean Si(100) surface, CO would be expected to desorb at a much lower temperature, 415 K instead of 700 K, based on the previous studies of CO interaction with this surface.20 Therefore, the only source for this intense mass 28 desorption peak is from CO adsorption on the sample holder or the button heater. The sample holder used in this study is made of tantalum and molybdenum and the button heater filament is made of tungsten. The previous studies of CO adsorption on clean Mo(100)15 recorded desorption temperatures of 840 K, 940 K, and 1200 K. CO adsorption on a clean W(111) surface17 showed CO desorption at 874 K. However, the behavior resulting in CO desorption at temperatures close to those observed in the present study has been reported for Ta(110).16,21 Thus, it is resonable to infer that the evolution of m/z = 28 described by component 3 reflects the reaction and desorption of CO on the tantalum cup sample holder rather than the interaction of chloroethane with a clean Si(100)-2 × 1 surface. Of course, the small amount of the pumping gas present in the dosing lines made it impossible to observed the effect of this impurity when iodoethane was investigated, as only a small exposure was required to obtain the high coverage. However, in the case of cloroethane, the much larger dose resulted in higher exposure of the mounting materials (tantalum cap) to the pumping gas, which contains CO. Although this procedure resulted in a relatively small amount of CO adsorbed (as confirmed by the fact that even at a dose of 5000 L of chloroethane the saturation curve for CO increased linearly), this component was clearly recorded by the mass spectrometer and interfered with the interpretation of the TPD spectra.

Fig. 7.

(Color online) (a) Integrated peak area folowing m/z = 28 as a function of chloroethane exposure and (b) the integrated peak area folowing m/z = 4 as a function of chloroethane exposure.

Fig. 7.

(Color online) (a) Integrated peak area folowing m/z = 28 as a function of chloroethane exposure and (b) the integrated peak area folowing m/z = 4 as a function of chloroethane exposure.

Close modal

Multivariate curve resolution has been applied to analyze the low-exposure and the high-exposure reaction systems to separate the thermal desorption data from background interference. In a high-exposure reaction system, when 5000 L of chloroethane-d5 was needed to produce a saturated monolayer on a clean Si(100)-2 × 1 surface, impurities in the form of pumping gas (predominantly CO) were introduced into the reaction vacuum chamber in a concentration that was sufficient to interfere with the interpretation of the TPD traces characterizing the target surface reaction. With the help of MCR analysis, it was possible to distinguish the impurities reacting with the metals of the mounting materials for sample attachment to the manipulator from the actual target sample surface reactions in an unbiased way. In this reaction system, the trace containing information about CO interaction with the tantalum metal of the sample holder was separated from information about the formation of ethyl groups following C-I or C-Cl dissociation of the corresponding alkyl halides with a clean Si(100)-2 × 1 surface, followed by hydrogen elimination processes occurring during TPD. This approach can be further developed to be applied to unbiased and straightforward interpretation of complex thermal desorption data.

This work was supported by the National Science Foundation (CHE 1057374 and CHE 1308118).

3.
E.
Habenschaden
and
J.
Küppers
,
Surf. Sci.
138
,
L147
(
1984
).
4.
L.
Pirolli
and
A. V.
Teplyakov
,
J. Phys. Chem B
110
,
4708
(
2006
).
5.
A. V.
Teplyakov
and
B. E.
Bent
,
J. Am. Chem. Soc.
117
,
10076
(
1995
).
6.
W. H.
Lawton
and
E. A.
Sylvestre
,
Technometrics
13
,
617
(
1971
).
7.
E. A.
Sylvestre
,
W. H.
Lawton
, and
M. S.
Maggio
,
Technometrics
16
,
353
(
1974
).
8.
C.
Ruckebusch
,
L.
Duponchel
,
J.-P.
Huvenne
, and
J.
Saurina
,
Vib. Spectrosc.
35
,
21
(
2004
).
9.
J.
Jaumot
,
V.
Marchán
,
R.
Gargallo
,
A.
Grandas
, and
R.
Tauler
,
Anal. Chem.
76
,
7094
(
2004
).
10.
J. C. F.
Rodríguez-Reyes
,
A. V.
Teplyakov
, and
S. D.
Brown
,
Surf. Sci.
604
,
2043
(
2010
).
11.
K. M.
Bulanin
,
A. G.
Shah
, and
A. V.
Teplyakov
,
J. Chem. Phys.
115
,
7187
(
2001
).
12.
K. M.
Bulanin
,
A. G.
Shah
,
D. R.
Fitzgerald
,
D. J.
Doren
, and
A. V.
Teplyakov
,
J. Phys. Chem.
106
,
7286
(
2002
).
13.
W.
Widdra
,
C.
Huang
,
G. A. D.
Briggs
, and
W. H.
Weinberg
,
J. Electron. Spectrosc.
64–65
,
129
(
1993
).
14.
“NIST Chemistry WebBook,” http://webbook.nist.gov.
15.
F.
Zaera
,
E.
Kollin
, and
J. L.
Gland
,
Chem. Phys. Lett.
121
,
464
(
1985
).
16.
R.
Klein
and
L. B.
Leder
,
J. Chem. Phys.
38
,
1866
(
1963
).
17.
H. H.
Hwu
,
B. D.
Polizzotti
, and
J. G.
Chen
,
J. Phys. Chem. B
105
,
10045
(
2001
).
18.
J. C.
Tully
,
M.
Gomez
, and
M.
Head-Gordon
,
J. Vac. Sci. Technol. A
11
,
1914
(
1993
).
19.
F. A.
Londry
,
A. J.
Slavin
, and
P. R.
Underhill
,
Surf. Sci.
140
,
521
(
1984
).
20.
T.
Kubo
,
T.
Aruga
,
N.
Takagi
, and
M.
Nishijima
,
Surf. Sci.
395
,
L246
(
1998
).
21.
T. W.
Haas
,
J. Chem. Phys.
46
,
3025
(
1967
).