Mechanisms of Thermal Decomposition of Organic Monolayers

Dec 5, 2006 - Michael K. F. Lo , Matthew N. Gard , Bryan R. Goldsmith , Miguel A. .... H. Sabbah , J. P. Conde , V. Chu , A. Zebda , S. Ababou-Girard ...
0 downloads 0 Views 166KB Size
Download PDF
1326

Langmuir 2007, 23, 1326-1332

Mechanisms of Thermal Decomposition of Organic Monolayers Grafted on (111) Silicon A. Faucheux,† A. C. Gouget-Laemmel, P. Allongue, C. Henry de Villeneuve, F. Ozanam, and J.-N. Chazalviel* Laboratoire de Physique de la Matie` re Condense´ e, CNRS-EÄ cole Polytechnique, 91128 Palaiseau-Cedex, France ReceiVed May 5, 2006. In Final Form: July 28, 2006 The thermal stability of different organic layers on silicon has been investigated by in situ infrared spectroscopy, using a specially designed variable-temperature cell. The monolayers were covalently grafted onto atomically flat (111) hydrogenated silicon surfaces through the (photochemical or catalytic) hydrosilylation of 1-decene, heptadecafluoro1-decene or undecylenic acid. In contrast to alkyl monolayers, which desorb as alkene chains around 300 °C by the breaking of the Si-C bond through a β-hydride elimination mechanism, the alkyl layers functionalized with a carboxylic acid terminal group undergo successive chemical transformations. At 200-250 °C, the carboxyl end groups couple forming anhydrides, which subsequently decompose at 250-300 °C by loss of the functional group. In the case of fluorinated alkyl chains, the C-C bond located between CH2 and CF2 units is first broken at 250-300 °C. In either case, the remaining alkyl layer is stable up to 350 °C, which is accounted for by a kinetic model involving chain pairing on the surface.

1. Introduction Functionalization of the silicon surface has been a topic of growing interest in recent years.1-8 Grafting of organic moieties to silicon through direct covalent Si-C bonding is a most attractive route, especially in view of the good electronic quality and fair chemical stability of the obtained surfaces. Potential applications involve chemical and biochemical sensors,9-12 but also the elaboration of thin dielectric layers or primer layers for use, for example, in microelectronics.13 For such purposes, the thermal stability of the layers may become an important issue. A few studies of the stability of alkyl monolayers grafted on a silicon surface have been reported in the literature.14-16 We have recently * Corresponding author. E-mail: [email protected]; tel: +331 6933 4663; fax: +331 6933 3004. † Also at STMicroelectronics, 850 rue Jean Monnet, 38926 Crolles-Cedex, France. Present address: MaCSE, Universite´ de Rennes, Campus de Beaulieu, 35000 Rennes, France. (1) Buriak, J. M. Chem. ReV. 2002, 102, 1271. (2) Wayner, D. D. M.; Wolkow, R. A. J. Chem. Soc., Perkin Trans. 2 2002, 23. (3) Linford, M. R.; Fenter, P.; Eisenberger, P. M.; Chidsey, C. E. D. J. Am. Chem. Soc. 1995, 117, 3145. (4) Sieval, A. B.; Demirel, A. L.; Nissink, J. W. M.; Linford, M. R.; van der Maas, J. H.; de Jeu, W. H.; Zuilhof, H.; Sudho¨lter, E. J. R. Langmuir 1998, 14, 1759. (5) Bateman, J. E.; Eagling, R. D.; Worrall, D. R.; Horrocks, B. R.; Houlton, A. Angew. Chem., Int. Ed. 1998, 37, 2683. (6) Boukherroub, R.; Morin, S.; Bensebaa, F.; Wayner, D. D. M. Langmuir 1999, 15, 3831. (7) Buriak, J. M.; Stewart, M. P.; Geders, T. W.; Allen, M. J.; Choi, H. C.; Smith, J.; Raftery, D.; Canham, L. T. J. Am. Chem. Soc. 1999, 121, 11491. (8) Cicero, R. L.; Linford, M. R.; Chidsey, C. E. D. Langmuir 2000, 16, 5688. (9) Wagner, P.; Nock, S.; Spudick, J. A.; Volkmuth, W. D.; Chu, S.; Cicero, R. L.; Wade, C. P.; Linford, M. R.; Chidsey, C. E. D. J. Struct. Biol. 1997, 119, 189. (10) Strother, T.; Cai, W.; Zhao, X.; Hamers, R. J.; Smith, L. M. J. Am. Chem. Soc. 2000, 122, 1205. (11) Pike, A. R.; Lie, L. H.; Eagling, R. A.; Ryder, L. C.; Patole, S. N.; Connoly, B. A.; Horrocks, B. R.; Houlton, A. Angew. Chem., Int. Ed. 2002, 41, 615. (12) Voicu, R.; Boukherroub, R.; Bartzoka, V.; Ward, T.; Wojtyk, J. T. C.; Wayner, D. D. M. Langmuir 2004, 20, 11713. (13) Chen, R.; Kim, H.; McIntyre, P. C.; Porter, D. W.; Bent, S. F. Appl. Phys. Lett. 2005, 86, 191910. (14) Sung, M. M.; Kluth, G. J.; Yauw, O. W.; Maboudian, R. Langmuir 1997, 13, 6164. (15) Yamada, R.; Ara, M.; Tada, H. Chem. Lett. 2004, 33, 492.

undertaken a systematic investigation of the stability of such layers, using infrared spectroscopy in a specially designed variable-temperature setup.17 In a preliminary report, we studied the thermal desorption of alkyl monolayers on (111) Si surfaces.17 These layers were found to be resistant upon heating up to 250 °C. Above this temperature (in the range 250-300 °C), thermal desorption of the alkyl chains was evidenced, accompanied by silicon oxidation. The desorption temperature was not found to be significantly affected by the chain length from C18 to C6; however a special behavior was observed for very short chains (C2, C1), which exhibit a higher stability.17 The present work is aimed at investigating the thermal behavior of functionalized monolayers grafted on (111) Si. Up to now and to our knowledge, no such study has ever been performed. 2. Experimental 2.1. Sample Preparation. The silicon substrates were taken from double-sided polished, float zone, 500 µm thick, n-type (111) silicon wafers, misoriented by 0.2° toward the [112h] direction (in order to obtain optimum control on terrace morphology18). The samples were cut as 15 × 18 mm2 platelets, and the shorter sides were beveled at 45°, in order to obtain attenuated total reflection (ATR) prisms (36 internal reflections). The sample surface was prepared atomically flat and hydrogenated by chemical etching in NH4F solution, then grafted with organic groups using the hydrosilylation reaction with alkenic precursors, according to the reaction tSi-H + CH2dCH-R f tSi-CH2-CH2-R

(1)

In practice, the Si(111) sample was initially cleaned in a 3:1 H2SO4/H2O2 piranha solution at 100 °C for 30 min and then rinsed copiously with ultrapure water. The atomically flat hydrogenated silicon surface was prepared by chemical etching in oxygen-free (16) Hunger, R.; Fritsche, R.; Jaeckel, B.; Jaegermann, W.; Webb, L. J.; Lewis, N. S. Phys. ReV. B 2005, 72, 045317. (17) Faucheux, A.; Yang, F.; Allongue, P.; Henry de Villeneuve, C.; Ozanam, F.; Chazalviel, J.-N. Appl. Phys. Lett. 2006, 88, 193123. (18) Munford, M. L.; Corte`s, R.; Allongue, P. Sens. Mater. 2001, 13, 259.

10.1021/la061260i CCC: $37.00 © 2007 American Chemical Society Published on Web 12/05/2006

Thermal Decomposition of Organic Layers on (111)Si

Figure 1. Typical ATR-FTIR spectra (p-polarization) of a decyl monolayer, a heptadecafluoro-1-decyl monolayer, and a carboxydecyl monolayer (the last two curves have been shifted downward for clarity). The reference of each spectrum is the hydrogen-terminated silicon surface. Notice the sharp negative peak at 2083 cm-1 (SiH loss) and the positive bands corresponding to grafted organic species. 40% NH4F (∼0.05 mol‚L-1 ammonium sulfite was added to the etching solution).18 Grafting of organic species was achieved using the hydrosilylation reaction of 1-decene, 3,3,4,4,5,5,6,6,7,7,8,8,9,9,10,10,10-heptadecafluoro-1-decene, and undecylenic acid [reaction 1, where -R stands for -(CH2)7CH3, -(CF2)7CF3, or -(CH2)8COOH, respectively]. For that purpose, the neat precursor reagent was introduced into a Schlenk glass tube, heated at 90 °C for 30 min under argon flushing to remove O2 and traces of water, then allowed to cool to room temperature. The freshly prepared H-Si(111) substrate was then introduced into the Schlenk tube, which was still purged with argon bubbling for 30 min before being hermetically closed. The tube containing undecylenic acid was irradiated for 3 h in a UV reactor (6 mW‚cm-2, 312 nm).19 Decyl and heptadecafluoro-decyl monolayers were obtained by heating at 100 °C for 20 h, with 10 vol % of ethylaluminum chloride catalyst being added to activate the reaction.6 After the reaction, the acid-terminated layer was rinsed with hot acetic acid.19 The decyl and heptadecafluorodecyl layers were rinsed with CF3CO2H (3% in tetrahydrofuran) and finally with HPLC-grade tetrahydrofuran and dichloromethane. 2.2. Surface Characterizations. Cleanliness of the obtained surfaces was controlled by atomic force microscopy (AFM) using a Molecular Imaging SPM in contact mode with a silicon nitride cantilever. The AFM images show atomically smooth terraces separated by straight parallel atomic steps (3.1 Å height) (see Supporting Information). This topography is identical to that of a H-Si(111) surface,18 which indicates a homogeneous coverage of the Si surface. The surface chemistry of the grafted organic monolayers was further characterized by Fourier transform infrared spectroscopy (FTIR) in ATR geometry in a Bomem MB100 spectrometer, equipped with a liquid-nitrogen-cooled Hg1-xCdxTe photovoltaic detector. The spectrum recorded after surface modification was compared to that taken before modification, providing the absorbance change with respect to the H-terminated surface. Figure 1 shows the ATR-FTIR spectra of a decyl layer, a heptadecafluorodecyl layer, and a carboxydecyl layer grafted on Si(111). In all cases, a negative peak at 2083 cm-1 indicating the loss of SiH bonds is evidenced, accompanied by the appearance of characteristic bands corresponding to the grafted organic species. For each spectrum, the bands at 1467, 2850, and 2920 cm-1 are assigned to methylene groups. For the case of the heptadecafluoro-decyl layer, a νCF massif appears below 1250 cm-1, where prominent modes are observed at 1149, 1205, and 1243 cm-1. For the carboxydecyl layer, the peak at 1715 cm-1 is consistent with the CdO stretching mode of the (19) Faucheux, A.; Gouget-Laemmel, A. C.; Henry de Villeneuve, C.; Boukherroub, R.; Ozanam, F.; Allongue, P.; Chazalviel, J.-N. Langmuir 2006, 22, 153.

Langmuir, Vol. 23, No. 3, 2007 1327 carboxyl group. The peaks at 1285 cm-1 and 1410 cm-1 correspond to IR bands related to the COH group. The weakness of the band around 1000-1100 cm-1 (especially for the case of carboxydecyl) indicates a very low concentration of Si-O-Si groups.19 For the decyl and carboxydecyl monolayers, quantitative analysis of the νCH spectrum was made by fitting the spectrum as a linear baseline plus the superimposition of several Voigt functions (for the fluorinated chains, no similar analysis could be made either for the νCH or the νCF spectrum, due to the weakness of the CH bands and the complex background in the CF region). For the decyl monolayer, five Voigt functions were used, representing the four symmetric and antisymmetric stretching modes of CH2 and CH3 groups and the main Fermiresonance-enhanced combination band at 2900 cm-1.20 For carboxylterminated layers, only three Voigt functions were used (only CH2 groups are present). The fits were found to be consistent with expectation. Especially, there was no detectable amount of unwanted species such as hydrocarbon chains substituted with oxygen (alcohols, ethers, or peroxides). A quantitative determination of the surface concentration of grafted chains was obtained from the intensity of the νSCH2 mode, after calibration of the IR absorption,19 yielding values of (3.2 ( 0.2) × 1014 cm-2 for decyl chains and (2.5 ( 0.2) × 1014 cm-2 for carboxydecyl groups. 2.3. Thermal Stability Investigations. The thermal stability of the organic layers was investigated in situ with a special IR cell. The description of the cell and the principle of the measurements have been detailed elsewhere.17 In brief, the cell consists of a leakproof chamber, which can be placed under the reduced pressure of a controlled atmosphere (air, nitrogen, or argon/hydrogen mixture), and where a resistive wire allows for heating the sample from ambient temperature up to 500 °C. The temperature scale was calibrated against the fusion points of LiNO3 (255 °C) and NaNO3 (307 °C). An essential point of the experiment is that the whole thermal treatment is performed without moving the cell from the spectrometer. In addition, in order to avoid unwanted effects associated with temperature-induced variations of the IR absorption (of either the Si lattice or the organic layer), the IR measurements were always made at a fixed temperature. A temperature of 40 °C was chosen for that purpose, since temperatures above 100 °C tend to saturate the photovoltaic detector. This experimental procedure allows for detecting very weak absorption changes by minimizing baseline problems. In practice, a first reference spectrum was recorded at 40 °C for the freshly grafted surface, then the sample was heated to 100 °C for 15 min and cooled at 40 °C; a new spectrum was recorded and so on by increasing the heating temperature in increments of 50 °C. Because thermal desorption is progressive, the spectra actually represent the cumulatiVe changes at the surface, which depend on the step duration and increment, as shown in our preliminary experiments on alkyl chains.17 Therefore, using a fixed experimental protocol (15 min plateaus in 50 °C increments) in all the experiments allows for a direct comparison of the thermal behavior of the various organic monolayers investigated in this work.

3. Results Series of spectra, illustrating the thermal behavior of the organic monolayers, are shown in Figures 2-4. The normalized changes in absorbance per reflection, obtained by using the freshly grafted surface for the reference spectrum, are plotted in Figures 2a, 3a, and 4a. The same data are also plotted in a “differential” manner in Figures 2b, 3b, and 4b, by using as the reference for each spectrum the spectrum recorded after treatment at the preceding temperature. The former type of plot allows for a quantitative analysis of the organic matter that has disappeared at a given temperature, and the latter type of plot shows what occurs at a given temperature in a very direct way. All of the results presented here correspond to silicon samples heated under argon/hydrogen atmosphere since, as reported earlier for alkyl layers,17 the thermal (20) Ward, R. N.; Duffy, D. C.; Davies, P. B.; Bain, C. D. J. Phys. Chem. 1994, 98, 8536.

1328 Langmuir, Vol. 23, No. 3, 2007

Faucheux et al.

Figure 2. IR spectra (p-polarization) recorded at 40 °C after successive heating of a catalytically decylated surface to increasing temperatures for durations of 15 min, under argon/hydrogen atmosphere (curves shifted vertically for clarity): (a) the common reference for each spectrum is the freshly grafted surface at 40 °C; (b) for each spectrum, the reference spectrum is the one taken after heating at the preceding temperature (“differential” presentation).

Figure 3. Same as Figure 2 for a catalytically heptadecafluoro-decylated surface: (a) the common reference for each spectrum is the freshly grafted surface at 40 °C; (b) “differential” presentation.

Figure 4. Same as Figure 2 for a photochemically carboxyl-terminated surface: (a) the common reference for each spectrum is the freshly grafted surface at 40 °C; (b) “differential” presentation.

decomposition of the organic layers is not significantly dependent on the type of atmosphere. 3.1. Decyl Monolayers. The thermal stability of decyl monolayers has been largely discussed elsewhere.17 From Figure 2a, nothing happens up to 200 °C. Above this temperature, we observe the simultaneous disappearance of the CHs and the growth of an oxide band around 1000-1100 cm-1, indicating that the chains have begun to desorb. At the same time, the broad Si-H band disappears at 2080 cm-1 and reappears at 2200-2260 cm-1

(SiH species with oxidized Si backbonds21,22). These changes occur in the range 200-350 °C, with the largest decrease in the CH intensity occurring between 250 and 300 °C (Figure 2b). The spectrum is essentially unmodified between 350 and 400 °C, indicating that alkyl monolayers have nearly completely desorbed at 350 °C (Figure 2b). As reported earlier, a limited oxidation (21) Schaefer, J. A.; Frankel, D.; Stucki, F.; Go¨pel, W.; Lapeyre, G. J. Surf. Sci. 1984, 139, L209. (22) Lucovsky, G. Solid State Commun. 1979, 29, 571.

Thermal Decomposition of Organic Layers on (111)Si

Langmuir, Vol. 23, No. 3, 2007 1329

Figure 5. Quantitative analysis of the data in Figures 2a and 3a. Temperature dependence of the integrated intensities (corresponding to the species remaining on the surface) for (a) a decyl monolayer and (b) a heptadecafluoro-decyl monolayer. The SiO band intensity has been obtained by integration from 950 to 1250 cm-1 and is accurate only within ( 10 cm-1, due to uncertainties on the baseline. Note the stability of the CHs up to higher temperatures in panel b compared to those in panel a, although the CFs have disappeared.

of the silicon surface is present, even under reducing atmosphere. The amount of oxide present after annealing at 400 °C is estimated to be on the order of 2-3 Å.19 This oxidation is probably due to the high fraction of water vapor in the base pressure of our system. Faster oxide growth was observed in oxidizing atmosphere.17 3.2. Fluorinated Layers. Figure 3 shows the spectra obtained after thermal treatment of a fluorinated layer. The spectra remain unchanged up to 200 °C. At higher temperatures, we observe the disappearance of the CF2 and CF3 groups (negative peaks at 1149, 1205, and 1243 cm-1) and the growth of an oxide band around 1000-1100 cm-1. One also notices that the SiH band disappears at 2080 cm-1 and reappears at 2200-2260 cm-1, as observed in Figure 2a. The largest CF decrease occurs in the range 250-300 °C. However, the CHs do not significantly disappear up to ∼400 °C (Figure 3b). 3.3. Carboxyl-Terminated Monolayers. Figure 4a represents spectra recorded after heating a carboxyl-terminated monolayer at different temperatures. The spectra remain unchanged up to 150 °C. In the range 150-250 °C, the peak at 1715 cm-1 decreases and two peaks appear at 1756 and 1822 cm-1 (Figure 4b), indicating the partial transformation of the carboxyl groups into anhydrides,23 according to the reaction

tSi-R-COOH + HOOC-R-Sit f tSi-R-(CO)-O-(CO)-R-Sit + H2O (2) This is accompanied by the decrease of two peaks at 1285 and 1410 cm-1, assigned to the loss of COH groups. The anhydrides in turn decompose in the range 250-300 °C, although very few methylene groups have disappeared even at 300 °C. Above 150 °C, the chemical transformations go together with the growth of an oxide band around 1000-1100 cm-1 (note that a narrower band near 1040 cm-1, attributed to the antisymmetric νCOC mode of the anhydrides, is present superimposed on the oxide band for the spectra in the 200-300 °C range; see Figure 4b). In the CO region, some loss of the 1715 cm-1 peak still occurs between 300 and 350 °C, but the spectrum is essentially unmodified between 350 and 400 °C (Figure 4b), indicating complete loss of carboxyl and anhydride groups. The decrease of the CH band is most important in the range 350-400 °C (Figure 4b). (23) Lin-Vien, D.; Colthup, N. B.; Fateley, W. G.; Grasselli, J. G. The Handbook of Infrared and Raman Frequencies of Organic Molecules; Academic: San Diego, CA, 1991; pp 142-143.

3.4. Thermal Stability of Functional Layers Compared to Decyl Layers. The thermal decomposition of functional layers (Figures 3 and 4) was quantitatively characterized by integrating the different IR bands, and the results were compared to those obtained for a decyl layer (Figure 2). For that purpose, we did not use the absolute quantitative analysis method described in Section 2.2, since it can be applied only in favorable cases. We simply determined a linear baseline by fitting the p-polarized spectrum in two regions adjacent to the spectral range under consideration (e.g., the νCH region), subtracted the fitted baseline, and integrated over that range. Some special cases occurred when overlapping bands were present (νCdO and νSiO+νC-O regions for the carboxydecyl layer and the νSiO+νCF region for the heptadecafluoro-decyl layer). The spectrum was then fitted with a linear baseline plus several Voigt functions, and the integrated area of the relevant bands was deduced from the fit. The results are presented as integrated absorbances obtained by using the ungrafted surface for the reference spectrum. Hence, they represent the amount of species left at a given temperature, with the integrated absorbances of the spectra of Figure 1 giving the initial values for the fresh monolayer. Figure 5 presents the results obtained from the FTIR spectra of the decyl and the heptadecafluoro-decyl layers. CF desorption (Figure 5b) is seen to occur in a similar temperature range as decyl desorption (Figure 5a), with surface oxidation occurring in parallel. However, the CH part of the heptadecafluoro-decyl chains appears to desorb at a temperature significantly higher than that for the decyl monolayer. In correspondence, the amount of oxide is somewhat lower at 400 °C than for the case of the decyl layers. The desorption of the whole fluorinated part of the grafted group is consistent with recent observations of the thermal behavior of perfluoroalkylsiloxane monolayers on oxidized Si.24 However, in our case, this desorption takes place in a much narrower temperature range. For the carboxydecyl monolayer, we studied in detail the νCHand νCO-related bands. Expanded views of these spectra are presented in Figure 6, together with the results of the corresponding analysis. Figure 6c shows the detail of the transformation of the acid terminal groups into anhydrides and the subsequent desorption of the anhydride groups, and Figure 6d shows the results of the quantitative analysis for the CH and SiO bands. Figure 6d clearly evidences that the alkyl skeleton of the carboxydecyl monolayers is desorbed more slowly than the alkyl (24) Fre´chette, J.; Maboudian, R.; Carraro, C. Langmuir 2006, 22, 2726.

1330 Langmuir, Vol. 23, No. 3, 2007

Faucheux et al.

Figure 6. (a,b) Zoom of the data in Figure 4b (differential form) in the νCO and νCH regions. (c,d) Quantitative analysis of the same data (for SiO, same remark as in Figure 5). Note the partial transformation of the carboxyl groups into anhydrides around 250 °C and the increased thermal stability of the remaining alkyl chains. (The dashed line in panel d is a copy of the CH curve from Figure 5a. The higher initial value is mainly due to the higher packing density of the alkyl monolayer).

chains of the decyl monolayers (dashed line). At 350 °C, the decyl layer has been decomposed nearly totally, while only about 25% of the carboxyl-terminated layer has been desorbed. Here again, the amount of oxide is lower in the case of the functional layers than that for the decyl layers, an effect that may be associated with the improved stability of the CHs.

4. Discussion 4.1. Desorption Mechanisms. For the decyl monolayer, we showed elsewhere that the shape of the negative CH bands always appears to be identical in all of the spectra and differs from the initial νCH (Figure 1a) just by a scaling factor.17 This suggests that alkyl chains are desorbed, as a whole, by cleavage of the silicon-carbon bond. Maboudian et al. showed that this cleavage is accompanied by regeneration of the alkene.14

tSi-CH2-CH2-R f tSi-H + CH2dCH-R

(3)

The similar behaviors observed in our experiments under various atmospheres and in the UHV experiments suggest that the same desorption mechanism may be operating.17 However, the appearance of a weak band around 1250 cm-1, which could be assigned to the δSCH3 mode of Si-CH3 groups, might suggest the presence of a competing desorption pathway.17

tSi-CH2-CH2-CH2-R f tSi-CH3 + CH2dCH-R (4) The respective contributions of reactions 4, which leaves a SiCH3 group on the surface, and 3 can be tested by carefully examining the νCH band. For a Cn alkyl chain, according to reaction 3, 1 CH3 group and (n - 1) CH2 groups are lost, hence

an expected CH3/CH2 ratio of 1/(n - 1) in the experimental spectra. According to reaction 4, (n - 1) CH2 groups are lost, and no variation in the number of CH3 takes place. However, since the IR cross-section of the νCH3 band of an Si-CH3 group is about 5 times weaker than that of an alkyl CH3,25 the ratio is expected to be (1 - 1/5)/(n - 1) for reaction 4. This ∼20% difference is rather tenuous indeed. However, after careful analysis of the νCH spectra, especially those of the shorter alkyl chains, we can ascertain that the (lost CH3)/(lost CH2) ratio deviates by, at most, a few percent from that expected from reaction 3. This is a strong indication that reaction 4 is a minority mechanism. The above data clearly demonstrate that fluorinated layers and carboxyl-terminated layers undergo chemical transformations prior to the final desorption of the residual hydrocarbon fragments. In the case of carboxyl-terminated layers, acid groups are first transformed into anhydrides. In a second step, the C-C bond between the alkyl skeleton and the anhydride group will break, again due to bond polarization. This may occur through a release of CO + CO2, leaving a saturated hydrocarbon chain bonded with two silicon atoms, as depicted in Scheme 1.In the case of fluorinated chains, we infer that the carbon-carbon bond between the alkyl skeleton and the first fluorinated carbon, which is a polarized bond, will break first. The initial CH2-CF2 bond breaking might result in the formation of surface tSi-CHd CH2 groups through the desorption of HF2C-(CF2)6-CF3, by analogy with reaction 3. However, such a scheme appears to be incompatible with the weak change in νCH observed below 350 °C (see Figure 5). As explained in the next subsection, the enhanced thermal stability of the residual hydrocarbon fragment suggests that chain pairing occurs according to the mechanism (25) Fide´lis, A.; Ozanam, F.; Chazalviel, J.-N. Surf. Sci. 2000, 444, L7.

Thermal Decomposition of Organic Layers on (111)Si

Langmuir, Vol. 23, No. 3, 2007 1331

Scheme 1

Figure 7. Scheme of the thermal desorption model for two-end grafted chains.

Table 1. Characteristic Temperatures and Activation Energies for Various Desorbing Speciesa T1/2 (K) Ea (eV)

alkyl CHs

CFs

COOH

280 1.34

280 1.34

225 1.20

CHs of paired chains 400 1.62

a Activation energies are obtained by fitting the experimental points to the simple first-order kinetic model mentioned in section 4.2, with a characteristic attempt-to-escape frequency, ω0, of 109 s-1.

below:

to regraft (form a second bond) at its free end is K′ (Figure 7). Two-end grafted chains can therefore be transformed into oneend grafted chains with a probability per unit time of 2K because either end can be broken. One-end grafted chains can either desorb with a probability per unit time K or be transformed into two-end grafted chains with a probability per unit time K′. Desorbed chains are assumed to be irreversibly lost. For the sake of simplicity, we assume that there are no steric restrictions on chain grafting, such as, for example, the availability of a free site within a suitable distance from the first anchoring point. Within this simplified framework, the time evolution of the concentrations N1 and N2 can be written as

2 tSi-(CH2)2-(CF2)6-CF3 f tSi-(CH2)4-Sit + F3C-(CF2)12-CF3 (5) 4.2. Kinetic Analysis. The thermal desorption of the alkyl chains, the fluorinated fragments, and the carboxyl groups was simulated assuming first-order desorption kinetics [desorption rate K ) ω0 exp(-Ea/kBT)] with the exact temperature sequence used experimentally. Such an analysis for the case of alkyl chains was reported earlier.17 The attempt frequency ω0 was taken as 109 s-1, and the activation barrier Ea was adjusted in order to fit the experimental results. The resulting values for Ea are given in Table 1, together with the T1/2 values defined as the temperature where a smooth curve drawn through the experimental points crosses the half-initial-coverage value. In the two former cases (for alkyl and fluorinated chains), the activation energy is found to be almost the same (1.34 eV), a result consistent with the similarity of the two associated microscopic processes (the cleavage of a Si-C or a C-C bond). In the later case (acid chains), the activation energy is found to be significantly lower (1.2 eV), which corresponds to the more facile elimination of a water molecule between two acid groups. This analysis could not be made for the anhydride groups, which appear and disappear in the same temperature range. 4.3. Why Are Paired Chains More Stable? In the case of the fluorinated chains or the carboxyl-terminated chains, the decomposition of the functional groups according to Scheme 1 and reaction 5 occurs at intermediate temperatures and leaves an alkyl chain that bridges two silicon sites. This new structure accounts for the increased thermal stability of the residual alkyl fragments compared to the alkyl monolayers. As shown in Table 1, the fit of the onset of the CH desorption plotted in Figure 5b to a simple first-order kinetic model would correspond to an activation energy of ∼1.62 eV, much larger than that characteristic of alkyl chain desorption.17 This value is only apparent, as the rupture of each individual Si-C bond is expected to exhibit an activation energy of 1.34 eV, as seen for the alkyl chains. This difference can be explained with the simple kinetic model that is exposed hereafter. Let us consider chains bonded to a surface through both ends. At time t, the surface concentration of one-end grafted chains is N1, and the concentration of two-end grafted chains is N2. We assume that the probability per unit time of breaking a bond with the surface is K, and the probability for a one-end grafted chain

dN2 ) -2KN2 + K′N1 dt

(6)

dN1 ) 2KN2 - (K + K′)N1 dt

(7)

The total number of grafted chains (N1 + N2) at time t can be obtained by solving this linear system. The general solution is of the form

N1 + N2 ) Ae-λ-t + Be-λ+t

(8)

1 λ( ) [(3K + K′) ( x(3K + K′)2 - 8K2] 2

(9)

with

Assuming initial conditions N1 ) 0 and N2 ) N0, the solution finally writes as

N 1 + N2 )

N0 [λ e-λ-t - λ-e-λ+t] λ+ - λ- +

(10)

The rate constants may be taken as K ) ω0 exp(-Ea/kBT) and K′ ) ω′0 exp(-E′a/kBT), where Ea and E′a are activation energies and ω0 and ω′0 are characteristic attempt frequencies. Since desorption is endothermic, one expects E′a < Ea. Hence, at moderate temperatures, K will be much smaller than K′. The above expressions then become

λ+ ≈ K′ and λ- ≈ 2K2/K′ , λ+

(11)

N1 + N2 ≈ N0 exp(- 2K2t/K′)

(12)

The characteristic desorption time is now K′/2K2, which is much larger than the corresponding time 1/K for one-end grafted chains. This simple result confirms that two-end grafted chains may exhibit a superior stability compared to that of one-end grafted chains. For a quantitative comparison, the exact experimental protocol, consisting of 15 min anneals by temperature increments of 50 °C, has been simulated using reasonable values for Ea, E′a, ω0, and ω′0. The values Ea ) 1.34 eV and ω0 ) 109 s-1 (see Table 1) were taken in order to fit the experimental data for one-end

1332 Langmuir, Vol. 23, No. 3, 2007

Figure 8. Simulation of chain desorption for one-end grafted chains (desorption rate K) and two-end grafted chains (bond breaking rate K and regrafting rate K′), with Ea, E′a, ω0, and ω′0 as given in the text. The lines are smooth curves drawn through the calculated points. The simulation for a surface with 80% two-end grafted chains and 20% one-end grafted chains has also been represented. Note the qualitative similarity of the two-step structure with the CH decrease in Figure 6d.

grafted chains.17 The orders of magnitude for the regrafting parameters could be deduced by using the detailed balance principle, namely, K′/K ) a exp(∆H/kBT), where ∆H represents the reaction enthalpy for one-end desorption (∼1 eV)17 and a accounts for the smaller pre-exponential factor in K′, due to the small probability that the free end of the chain is properly positioned with respect to the surface for chain regrafting. Because of the lack of an accurate value of ∆H, we chose a ) 10-3 (hence, ω′0 ) 106 s-1) and have determined E′a in order to fit the experimental results (Figures 6d and 8). From the obtained value of E′a ) 0.6 eV, we estimate ∆H ) 0.74 eV, a very plausible value. Note, however, that this may be an underestimate if other (26) Yan, L.; Marzolin, C.; Terfort, A.; Whitesides, G. M. Langmuir 1997, 13, 6704. (27) Chi, Y. S.; Choi, I. S. Langmuir 2005, 21, 11765. (28) Perring, M.; Dutta, S.; Arafat, S.; Mitchell, M.; Kenis, P. J. A.; Bowden, N. B. Langmuir 2005, 21, 10537.

Faucheux et al.

mechanisms leading to desorption of the layers are present (for example, oxidation of the silicon backbonds, which may occur above 400 °C when traces of oxygen are present in the atmosphere). Also, the observed progressive character of the desorption in the range 250-350 °C, not accounted for by this simple model, may be attributed to the survival of unpaired chains upon carboxyl elimination, as indicated by the presence of single carboxyl groups up to temperatures where the anhydrides become unstable. A simulation taking into account the presence of such unpaired chains has been made, and the results (dotted curve in Figure 8) yield a somewhat better fit to the experimental data. Despite these limitations, we think that the above simple model provides conclusive evidence of the beneficial effect of two-end grafting on the thermal stability of the layers.

5. Conclusion Our study shows that, when temperature is increased, the two functional groups considered here are more fragile than the alkyl skeleton and the Si-C bond, an effect which may become the limiting factor in the use of silicon surfaces functionalized with organic material. However, the appearance of new functionalities in the course of the decomposition process can be of interest. For example, anhydrides are used as a starting point for anchoring new organic or biological species.26-28 Finally, an interesting output of the above study is the fact that elimination of the functional group may lead to an end-pairing of the grafted chains, which results in an improved thermal stability, opening the way to applications at temperatures higher than previously foreseen. Acknowledgment. The authors are indebted to P. Roca i Cabarrocas and F. Kail for help in assessing the vacuum system. Supporting Information Available: AFM image of a carboxylterminated (111) Si surface. This material is available free of charge via the Internet at http://pubs.acs.org. LA061260I