16544
J. Phys. Chem. B 2005, 109, 16544-16553
Boundary Layer Chemistry Probed by in Situ Infrared Spectroscopy during SiO2 Deposition at Atmospheric Pressure from Tetraethylorthosilicate and Ozone Lucio D. Flores† and John E. Crowell* Department of Chemistry and Biochemistry, UniVersity of California at San Diego, La Jolla, California 92093-0314 ReceiVed: April 11, 2005; In Final Form: June 22, 2005
The deposition of silicon dioxide films at 450 °C was studied in quasi real time by probing the thermally activated boundary layer region near the growing surface during atmospheric pressure chemical vapor deposition (APCVD). Potential tetraethylorthosilicate (TEOS)/O3 reaction products have been investigated in an attempt to clarify the reaction mechanism leading to the observed silanol deposition intermediates and delineate the film formation process. The organic products formed during the TEOS/O3 reaction are acetic acid, formic acid, formaldehyde, carbon monoxide, carbon dioxide, and water. Quantitative methods are developed using FT-IR (Fourier transform infrared) spectroscopy during ozonation of TEOS at elevated temperatures. The measurement of gaseous alcohols of silicon alkoxides by FT-IR is demonstrated by application of an in situ methodology that probes the high-temperature region within the CVD environment. Partial least squares (PLS) Beer’s law absorption models are used in determining relative TEOS, ozone, and ethoxysilanol levels during the reaction. The reaction order in TEOS is measured at 1.65 ( 0.02 over a 0.9 Torr pressure range. Similarly, the ratio of ethoxysilanol formed versus the amount of ozone consumed is ca. 1:3. A radical chain oxidative mechanism involving direct reaction of TEOS and ozone is proposed for formation of highly reactive silanol film growth intermediates.
1. Introduction The chemical vapor deposition (CVD) of tetraethylorthosilicate [TEOS, Si(OC2H5)4, also known as tetraethoxysilane] and ozone (O3) has been used to deposit SiO2 films suitable for use in VLSI (very large scale integration) device applications.1-3 The TEOS/O3 reaction system is unique in that the “as deposited” film profile can be varied from a conformal to flowlike step coverage depending on the initial ozone to TEOS concentration ratio.4-7 The flowlike property is key to the deposition of high-quality dielectrics for shallow trench isolation with high aspect ratios and finds use in multilevel interconnect devices. Studies have focused on deposition at 100 g/m3) using high flow rates for oxygen as given
J. Phys. Chem. B, Vol. 109, No. 34, 2005 16547
Figure 3. Infrared spectra for the lower portion of the C-H stretch region, comparing reaction products at 450 °C of TEOS/O2 (a) to that of TEOS/O3 (b). The sharp features observed for TEOS/O3 are identified as the individual υ1 and υ5 ro-vibrational transitions of formaldehyde (c).
in Supporting Information table S6. Beer’s law UV measurements for ozone concentrations are given relative to oxygen when measured under known pressure and temperature conditions, at STP, and as such, the absolute concentration of molecular oxygen is implied by conversion using molar volumes. The ozone concentration was measured relative to a fixed oxygen flow rate, and all calibration data showed Beer’s law linear trends based on UV ozone meter measurements taken at the ozonizer (see Supporting Information table S7 and graph S8). Relative ozone concentrations measured during CVD reaction are found by referencing the conditions under which the ozone standards are generated. Relative concentrations of ozone are therefore determined by calculation of a single ozone flow rate that accounts for dilution of nitrogen without accounting for ozone decomposition. This method allows for direct comparison of ozone measurements by a single quantitative model since variation of the initial ozone concentration at the ozonizer is not performed. Similarly, the variation in ozone concentration resulting from TEOS dilution cannot be applied in the determination of the reaction order in ozone since the initial concentrations only span the error in ozone delivery ((5 g/m3) and generate a nearly “flat” log-log plot. 3. Results and Discussion 3.1. In Situ Product Analysis and Spectral Assignments. Absorbance spectra of the intermediates produced during the CVD process by probing above the wafer surface gave species similar to those observed by other researchers.15,27,37 Our studies are unique in that (i) we are probing the deposition region in a flow mode at atmospheric pressures upon initial injection of the separated gases (rather than the downstream exhaust region or after substantial mixing or for a static system) and (ii) we have carried out our studies at a resolution higher than that used previously to utilize each individual component’s ro-vibrational band structure for spectral identification. The C-H stretching region given in Figure 3 shows the measured and reference spectra of formaldehyde (CH2O) with the expected υ1 and υ5 features centered at 2780 cm-1 consistent with accepted literature values.38 At lower resolutions, previous researchers have attributed these TEOS oxidation product features primarily to acetaldehyde (CH3CHO).14,39 However, the reaction spectra given here are assigned by using product standards (reference spectra) prepared and measured in the laboratory, and by investigating the oxidation of simple organic molecules using
16548 J. Phys. Chem. B, Vol. 109, No. 34, 2005
Flores and Crowell TABLE 4: Infrared Assignments for Species Observed during Acetaldehyde Ozonation with both Static and Flow Reactionsa species
band (cm-1)
acetic acid 3581, 642 peracetic acid 3300, 1243 formaldehyde 2900-2642, 1745 formic acid 1776, 1105 methanol 1033
CH3CHO/O3 CH3CHO/O3 TEOS/O3 static flow flow w + +
w + +
+ +
+ +
+
+ -
a Comparison is made to the products observed during the TEOS/ O3 flow reaction during CVD (w means weakly observed, + observed, and - not observed).
Figure 4. Comparison at 450 °C of TEOS/O3 (a) and TEOS/O2 (b) during flow reaction in the CVD reactor. The observed features are labeled A-H and are given in Table 3. Figure 3b has absorption features due solely to TEOS whose assignments are given by Mondrago´n et al.48
TABLE 2: Representative TEOS/Ozone CVD Reaction Conditionsa flow rate (slm) TEOS ∼100 g of ozone/m3 at 20 °C separator vent shield left N2 plenum right N2 plenum window purge a
2 6.5 8 10 3 3 5
Twafer ) 450 °C. Tbubbler ) 65 °C.
TABLE 3: TEOS/O3 Reaction Products in Figure 4a A
H2O
B C D E F G H
SiOH CH3COOH CH2O CO2 CO HCOOH O3 SiOH CO2
υ3 (antisymmetric) υ1 (symmetric) υOH υOH υ1 and υ5 υ3 (antisymmetric) υ υ3 υ3 υbend υ2
3756 cm-1 (band center) 3652 cm-1 (band center) 3737 cm-1 3581 cm-1 2642-2900 cm-1 2350 cm-1 (band center) 2140 cm-1 (band center) 1776 cm-1 (Q-band) 1042 cm-1 905 cm-1 667 cm-1
ozone. All reference spectra were collected either (a) in situ in the CVD reactor in a flow mode or (b) in a sealed gas cell in a static mode. It is notable that use of any room-temperature gas cell standards for spectral subtraction has a limited account of the integrated area due to rotational broadening and precludes quantitative analysis of the resultant spectra. Studies involving potential reaction products with ozone were performed in a flow mode at CVD temperatures or in a static mode using the sealed gas cell; these latter studies were made for comparison under conditions involving longer mixing reaction times (see Supporting Information spectra S9 and S10). 3.2. TEOS/O3 versus TEOS/O2 Reaction at 450 °C. Spectra comparing the reaction of TEOS/ozone (O3) versus that of TEOS/oxygen (O2) are given in Figures 3 and 4. The TEOS/ ozone CVD reaction conditions that were used are listed in Table 2. Assignments of the IR features observed in Figure 4 are given in Table 3. In Figure 4b, it is evident that TEOS does not significantly react with O2 under these conditions. However, new features are observed in the aliphatic and carbonyl infrared regions for the TEOS/O3 system at this temperature (Figure 4a). We can distinguish absorbance features from silanol, H2O, CO, CO2, formaldehyde, formic acid, and acetic acid in our spectra.
Note that the assignments for the TEOS/O3 reaction are for those features that are additionally observed versus the TEOS/O2 reaction spectra. Only molecular TEOS features are evident in the lower spectrum (Figure 4b), and these molecular TEOS assignments are not repeated or labeled. 3.3. Comparison to Potential Product Chemistry: Acetaldehyde Ozonation. The change in the oxidation products formed during CH3CHO/O3 reactions in our static gas cell experiments compared to those measured under flow conditions is presented in Table 4. In particular, we are unable to measure the level of formic acid in the APCVD flow environment during acetaldehyde ozonation in contrast to the results obtained in a static gas cell (see Supporting Information spectra S9 and S10). However, in addition to CO, CO2, and formaldehyde production, we note that methanol and peracetic acid are produced during acetaldehyde ozonation in both our static gas cell and APCVD flow reactor studies.40,41 We note that we did not observe performic acid species when formic acid was produced in our gas cell during CH3CHO/O3 reaction.42 3.4. Implications from Acetaldehyde Ozonation Chemistry. In the reaction of acetaldehyde and ozone, we have noted differences in the oxidation products detected during our flow measurement at atmospheric pressure compared to those measured in our static gas cell (Table 4). We find that in a gas cell during ozonation of acetaldehyde, formic acid (1105 cm-1) and methanol (1033 cm-1) are produced (see Supporting Information spectra S9 and S10). We note that formic acid is only found in the static gas cell measurement and not in the flow measurement. The ozone-assisted autoxidation of acetaldehyde in solution29 has been reported to occur via hydroperoxide and peroxide intermediates; however, we have not detected performic acid42 in our studies. This reaction sequence leads to peracetic acid [CH3C(O)OOH] which has been detected in our CH3CHO/O3 studies as evidenced by the weak absorption band at 3300 cm-1 and the relatively strong band at 1243 cm-1 (see Supporting Information spectra S9 and S10). Additionally, we find that methanol is a major product of the ozonation of acetaldehyde in both our flow and gas cell studies.40,41 Methanol, having not been detected to date in any TEOS/O3 flow process, is a critical difference in the two systems. The generation of acetic acid during TEOS/O3 reaction coupled with the lack of methanol production suggests that acetaldehyde production is not part of the primary mechanism during TEOS ozonation and subsequent SiO2 deposition. 3.5. Assignment of TEOS/O3 Reaction Products. Previous researchers investigating the TEOS/O3 process with FT-IR were able to observe gas-phase byproducts without interference from TEOS and identified acetic acid, formic acid, H2O, CO, and CO2.27 In those studies, there were no aldehydes (i.e., acetaldehyde and formaldehyde) detected; however, the authors
Boundary Layer Chemistry of TEOS and Ozone by FT-IR
J. Phys. Chem. B, Vol. 109, No. 34, 2005 16549
Figure 6. Nineteen pairs of concentrations measured for TEOS/O3 used with the PLS model. The numbers next to the data points correspond to the number of calibration and validation measurements, respectively. Figure 5. Comparison of acetaldehyde/O3 (top spectrum) and TEOS/ O3 (middle spectrum) product spectra to specially generated gas-phase silanol spectra (bottom spectrum) and low-resolution CO2 reference spectrum (boxed). The numerous sharp absorption lines in the top and middle spectra are due to H2O (antisymmetric υ3 ) 3756 cm-1 and symmetric υ1 ) 3652 cm-1) and CO2 [where υ(1,00,1)1 ) 3714 cm-1 and υ(1,00,1)2 ) 3612 cm-1]. The low-resolution CO2 spectrum that is included (boxed) shows characteristic P, R band structure. The dashed vertical lines from left to right show the band centers for the isolated hydroxyl groups of ethoxysilanols (band center at 3737 cm-1) and carboxylic acids (band center at 3580 cm-1), respectively.
suspected that the reason the aldehyde products were not observed was that they were being oxidized into the observed acetic acid and formic acid products.27 Assignment of the 1745 cm-1 band to formaldehyde rather than acetaldehyde (Figure 4) is supported by comparing the relative intensity of the Fermishifted C-H ro-vibrational features centered near 2780 cm-1 (see Figure 3). The approximate 3:1 intensity ratio for υ(C-O) and υ(C-H) bands is similar to that observed in our gas-phase reference spectra of formaldehyde. Since silanol species and their condensation products (ethoxysiloxanes) are suspected in the film formation process, we attempted to generate measurable spectroscopic gas-phase silanol species by known preparative methods43-45 without success. Apparently, any attempt to transfer alkoxysilanols to the vapor phase at room temperature was accompanied by solvent evaporation and precipitation of silicon dioxide; i.e., no vaporphase FT-IR spectra of ethoxysilanols are generated by static or flow methods from known preparative methods involving (EtO)3SiCl. This behavior is consistent with the literature where ethoxysilanols are only observed in solution. We have found a convenient means of generating isolated ethoxysilanol standards by comparing TEOS/O3/TMPi (TMPi ) trimethyl phosphite) reaction spectra with TEOS/O3 spectra at 450 °C.30 By taking the 1:1 spectral difference of both systems, we find the reduction in the amount of silanol species (by a factor of
