Probing plasma/surface interactions
10/01/1997
Eray S. Aydil, University of California Santa Barbara, Santa Barbara, California
Richard A. Gottscho, Lam Research Corporation, Fremont, California
Understanding plasma-surface interactions in plasma etching, deposition, and cleaning is a prerequisite for achieving process goals. Attenuated total reflection Fourier transform infrared (ATR-FTIR) spectroscopy is a valuable surface-sensitive diagnostic technique that can be used in situ during plasma process development to detect adsorbed species on surfaces and to monitor the composition of thin films. This article briefly describes the ATR-FTIR technique, and summarizes selected results from silicon surface cleaning and plasma-enhanced chemical vapor deposition (PECVD).
Plasma processes are crucial for the etching and deposition of thin films in IC manufacturing. While much of the initial progress in plasma processing was empirically derived, the most recent advances resulted from fundamental research in plasma diagnostics and plasma-surface interactions. Controlling surface reactions in plasma processing is crucial to the microelectronics industry, though it has been difficult to arrive at an understanding of plasma-surface interactions. Phenomena occurring on surfaces during plasma etching, deposition, and cleaning remain among the least understood aspects of plasma processing.
Some of the most sensitive and powerful conventional surface science techniques cannot be used in the chemically harsh and higher-pressure environments common to plasma reactors. However, in the last decade, several in situ optical surface diagnostic techniques such as ellipsometry [1], interferometry [2], laser-induced thermal desorption [3], and photoluminescence [4–6] have been used to obtain information on surfaces and thin films during process development. One of these techniques is ATR-FTIR spectroscopy [7, 8].
ATR-FTIR spectroscopy has been used as a surface analysis technique to obtain infrared spectra of thin films and adsorbates both in ultra-high vacuum and under ambient conditions for the past 30 years. However, its use in a plasma environment to study problems relevant to thin-film etching and deposition is relatively recent [4–6, 9–13]. This article briefly describes the ATR-FTIR spectroscopy technique and summarizes the results of recent R&D studies where some of the key issues in plasma processing have been addressed.
ATR-FTIR spectroscopy: The principles
Probing a surface with infrared radiation provides information on the vibrational spectra and the chemical nature of the overlayers on a substrate. ATR-FTIR spectroscopy is one of the possible arrangements (Figs. 1 and 2) for performing surface infrared spectroscopy, and is ideally suited for studying adsorbates and thin films on crystalline semiconductor substrates that are transparent to at least some portion of the infrared spectrum (400–4000 cm-1; 2.5–25 µm). For each semiconductor, the spectral range across which ATR-FTIR can be used is determined by a lower frequency below which the ATR-crystal becomes opaque (Table 1).
 |
Specially prepared ATR-FTIR substrates are shaped to trap and transmit IR radiation across the semiconductor, much like an optical fiber, by multiple total internal reflections (Fig. 1). A substrate with a trapezoidal cross section works well for most geometries, though a parallelogram cross section also enables multiple internal reflections.
 Figure 1. Schematic of an ATR-FTIR crystal and the multiple internal reflection geometry. |
The trapped infrared beam undergoes multiple total internal reflections from both the top and the bottom surfaces until it reaches the bevel on the opposite side of the substrate. The number of total reflections from the upper surface, typically 50–100, is a function of the substrate thickness and length. The large number of total internal reflections makes ATR-FTIR two orders of magnitude more sensitive than transmission infrared spectroscopy, allowing detection of overlayers on the substrate as thin as a monolayer.
When the top surface of the crystal totally reflects the infrared radiation, the electric field of the reflected electromagnetic wave interacts with surface overlayers, leading to absorption of the infrared radiation and attenuation of the transmitted wave at characteristic absorption frequencies of the overlayers. For silicon, penetration depths range from 0.18–1.8 µm in the infrared, so ATR-FTIR spectroscopy can probe overlayers up to a few microns thick. In thick films, the exponential variation of the electric field above the semiconductor-film interface must be taken into account for quantitative analysis, and the spectra must be corrected for decreasing sensitivity as the films get thicker [8].
If there are no adlayers on the substrate that can absorb the infrared radiation, the intensities as a function of frequency at the entrance and exit bevels are equal and the transmission is defined as 100% (T% = 100 × I/Io). However, an infrared absorbing layer on the sample attenuates the intensity exiting the sample at the characteristic frequencies at which this layer absorbs. Spectra are presented either as transmission or absorbance (A = -log10 T). The most common acquisition and presentation mode is "differential," referencing the spectrum of everything in the beam path during processing (including the semiconductor substrate) to a spectrum of everything in the beam path before the process.
In situ ATR-FTIR spectroscopy
A typical in situ ATR-FTIR apparatus (Fig. 2) was used with a helical resonator plasma source and a substrate on a platen in a chamber directly below the plasma. Rectangular samples (~1 × 5 cm) were cut from commercially available double side polished wafers (~0.03–0.07 cm thick) and their short sides beveled to produce a trapezoidal cross section (Fig. 1).
 Figure 2. Schematic of an in situ ATR-FTIR spectroscopy apparatus installed on a plasma reactor with a helical resonator discharge source. (Reprinted from [13]) |
Although there is an optimum internal angle of incidence for a given semiconductor, in most cases the bevel angle is determined by the geometry of the plasma reactor feedthroughs. Practical bevel angles are between 30° and 60° — usually 45°.
 |
Focused infrared radiation from a source, after passing through a Michaelson interferometer, impinges upon one of the beveled edges of the sample (preferably at normal incidence to the facet). The infrared radiation leaving the spectrometer is typically 5–10 cm in dia. and approximately collimated. Lenses and mirrors then direct the beam to a viewport on the reactor. The viewports and the lenses must be transparent in the infrared (Table 2).
 |
In the reactor depicted in Fig. 2, all the lenses and windows are made of KBr. An infrared detector (Table 3) collects the emerging radiation. Fourier transforms convert the signal to obtain the absorption spectrum, which includes absorptions by infrared active gas phase molecules such as H2O and CO2. The beam path outside the reactor must be enclosed and purged with dry N2 to eliminate these gas phase absorptions.
Applications of in situ ATR-FTIR spectroscopy
The best way to illustrate the use of ATR-FTIR spectroscopy is by discussing its applications in plasma processing. Selected examples are described briefly in the following sections; the reader is referred to the referenced papers for details.
Silicon surface cleaning. Vapor phase and plasma-based cleaning of Si surfaces is an alternative to wet-cleaning technology and is a potential solution to the escalating cost and environmental concerns associated with handling and recycling large volumes of liquid-based chemicals. In situ cleaning of wafers without removing them from vacuum also allows for the integration of cleaning with other processing steps in cluster tools.
ATR-FTIR spectroscopy was used to study the removal of native oxide from Si surfaces and subsequent H passivation of the surface using H2 plasmas (Fig. 3) [10]. Spectra were taken after plasma treating the native oxide and hydrocarbon contaminated Si surfaces with a) H atoms only, b) H atoms and ions under unoptimized conditions, c) H atoms and ions under optimized conditions, and d) a nonplasma wet-HF dip for comparison (the last step of a typical RCA clean). In each case, the reference spectrum was that of a native-oxide-covered Si surface.
An increase in transmission, above 100%, represents the characteristic absorption frequency of any removed bonds. Conversely, a decrease in transmission below 100% represents the absorption frequencies of any formed chemical bonds. For example, the top spectrum in Fig. 3 shows an increase in transmission around 2900 cm-1, which corresponds to the C-H stretching absorption of hydrocarbons on the surface that were removed by H exposure. On the other hand, transmission decreased at 2250 cm-1, corresponding to Si-H stretching absorption of species such as OySi-Hx, indicating that such species formed as a result of H atom exposure.
Based on the infrared spectra of the Si surface obtained after H atom exposure, we conclude that H reacts with the native oxide to form an OySi-Hx layer that stays on the surface: the native oxide cannot be removed using H atoms alone. However, hydrocarbon contamination on the Si surface is reduced by exposure to just H atoms.
Other experiments showed that additional energy (ion bombardment for example) must be supplied to the surface in order to remove the OySi-Hx layer (Fig. 3b). The absorption peak at 2100 cm-1 corresponds to Si-H stretching absorption of surface hydrides where the Si is backbonded to Si. An increase in this peak represents a native-oxide-free surface that is passivated with H (compare Fig. 3d). Thus the surface treated by a combination of ions and H atoms is at least partially hydrogen terminated.
The comparison of the IR spectrum of this surface to the IR spectrum after HF dipping shows that the two surfaces are very similar. However, a careful re-examination of the Si-H stretching absorption in Fig. 3b finds a shoulder characteristic of residual OySi-Hx on the surface. Thus, although the surface could be partially cleaned, the conditions are not yet optimized. Other experiments showed that cleaning was strongly dependent on the ratio of H atoms to the ion energy flux, and that the process can be optimized by adjusting this ratio (Fig. 3c) [10]. The optimized plasma clean conditions produce a surface that is identical to the HF-cleaned Si surface.
In situ ATR-FTIR experiments demonstrated that three competing processes play important roles in plasma cleaning of the native-oxide-contaminated Si surfaces: ion bombardment damage, ion bombardment assisted removal of SiO2 and OySi-Hx on the surface, and reformation of SiO2 and OySi-Hx [10]. Which process dominates depends on the ratio of neutral-to-ion fluxes, and water and OH contamination in the reactor (which can be produced by H atoms reacting with quartz walls) [6, 10].
PECVD of silicon dioxide from SiH4/O2/Ar plasma. Silicon dioxide films produced by PECVD are important IC interlevel dielectrics. SiOH and SiH content in these films must be minimized to maintain optimal material properties, which requires understanding the fundamental processes that control surface concentrations of hydroxyls and hydrides (the reaction pathways that lead to SiO2). The surface reaction mechanisms and the composition of the growing film surface during PECVD from a mixture of SiH4, O2, and Ar in a helical resonator plasma reactor on a GaAs substrate were recently studied by ATR-FTIR (Fig. 4) [6, 10, 12, 13].
 Figure 4. Infrared spectra of silicon dioxide film and its surface during growth on a GaAs ATR substrate. (Reprinted from [12]) |
As the oxide grows, the Si-O band around 1080 cm-1 increases with time, as does the SiOH absorption band at 3000–3750 cm-1. The OH band includes absorption due to hydroxyls in the bulk oxide and on the surface; additional experiments could distinguish the surface and near surface contribution from the bulk [8]. Approximately 70% of the SiOH absorption is due to the surface SiOH, while the rest is due to bulk SiOH. Thus, during the deposition in an O2-rich SiH4/O2/Ar plasma, the surface species are predominantly hydrogen-bonded SiOH and isolated SiOH. Surface silicon hydrides that may be expected to be present on the surface are not detected.
The low SiH4:O2 ratio (R) results in the absence of silicon hydrides on the surface. The excess O in the plasma instantaneously oxidizes adsorbed silicon hydrides, yielding an SiOH covered surface; the lifetime of the SiHx species on the surface is very short. However, as the ratio of the SiHx-to-O flux increases, the residence time of silicon hydrides on the surface also increases and they can be detected by ATR-FTIR (Fig. 5).
We begin to detect SiH on the surface and in the film at a ratio of approximately 0.8. Absorptions due to silicon hydrides appear at 2250 cm-1 (Fig. 6), the characteristic frequency of monohydrides backbonded to three oxygen atoms (O3Si-H). SiOH content in the film rapidly declines as R is increased above 0.8. Reduced SiOH levels correspond to a reduction in the dielectric constant.
 Figure 6. Schematic showing how the surface changes qualitatively as the SiH4/O2 precursor ratio is changed in silicon dioxide PECVD. (Reprinted from [13]) |
Further increase in R results in the appearance of a band at 2150 cm-1, characteristic of silicon hydride absorption where the Si atom backbonds to other Si atoms (e.g., O2SiSi-H). As surface SiH replaces SiOH with increasing R, the probability of gas phase silane fragments finding silicon hydrides rather than surface hydroxyl increases and silicon hydrides backbond to surface Si atoms. Films then become Si and SiH rich above R>1 and the refractive index rises as more Si-Si bonds replace Si-O bonds. An optimum operating window exists where minimal SiOH and SiH content results in low-dielectric-constant SiO2 films.
PECVD of silicon dioxide with TEOS/O2 plasma. ATR-FTIR was also used to investigate SiO2 PECVD from TEOS and O2 plasma [11]. These studies found evidence that TEOS chemically adsorbs onto the SiO2 surface, forming ethoxysiloxanes, and the adsorption proceeds through a physically adsorbed state of the TEOS molecule. The existence of a physisorbed state is consistent with the common observation that the deposition activation energy is "negative."
As temperature increases, the residence time of the weakly (physically) adsorbed TEOS precursor state on the surface decreases, leading to a lower deposition rate and the observed "negative activation energy." Ethoxysiloxanes are formed on the surface (upon adsorption of TEOS and TEOS fragments) and subsequently oxidized to yield SiOH, H2O, and CO. The oxide may form by reaction of two nearby SiOHs to form SiO2 and H2O.
If the removal rate of the ethoxysiloxanes and SiOH is low relative to the adsorption rate of TEOS, hydroxyls and ethoxy species can be incorporated into the film leading to porous and poor quality oxides. Good quality, dense, and SiOH-free films can be obtained when the deposition is limited by TEOS flux to the surface so that oxidizing species such as O have plenty of time to oxidize the adsorbed TEOS fragments.
α-SiN:H PECVD for flat panel displays. Bailey and Gottscho used ATR-FTIR to study amorphous silicon and silicon nitride deposition from SiH4/N2 plasma [14]. They showed that ATR-FTIR is sufficiently sensitive to monitor processes at relatively high deposition rates and demonstrated that PECVD of α-SiN:H at 10 nm/min can be monitored with monolayer sensitivity (Fig. 7). They projected that improvements in software would allow for equal sensitivity monitoring of higher deposition rates, making ATR-FTIR applicable to technologically relevant conditions.
The ATR-FTIR investigation of a-SiN:H PECVD on crystalline silicon showed that the nitride layer grows through a reaction layer between the gas phase and the silicon nitride film (Fig. 8a). The reaction layer is enriched in -N-Si-H bonds and is approximately 25-nm thick under the conditions studied. The underlying nitride film grows by addition of material from this reaction layer.
The spectra collected as a function of growth time (Fig. 7) indicate the existence of an amorphous hydrogenated silicon layer at the substrate-nitride interface. The evidence for such an amorphous silicon layer is the Si-H stretching absorptions observed at 2040 cm-1 (the absorption frequency for SiH where the Si is backbonded to other Si atoms). This absorption is most pronounced at the earliest stages of the deposition and is overshadowed by the NSi-H absorption as the nitride film gets thicker.
The sequence of α-Si:H and α-SiN:H film depositions in manufacturing thin-film transistors (TFTs) for active matrix liquid displays determines final TFT electrical performance. TFTs produced using the "bottom-gate" configuration (α-SiN:H precedes α-Si:H deposition) have superior characteristics compared to "top-gate" TFTs (α-SiN:H deposited after α-Si:H). However, the top-gate configuration is simpler to fabricate, with fewer processing steps. The differences in TFT performance result from the differences in the quality of the α-Si:H–α-SiN:H interface, since the current channel lies in the semiconductor side at this interface.
A reaction layer, like the one on the α-SiN:H film, is included on top of the α-Si:H film. The reaction layer on the α-Si:H ends up at the dielectric-semiconductor interface in the top-gate configuration (Fig. 8c), but away from this interface in the bottom-gate configuration (Fig. 8b). Could this reaction layer, expected to have higher concentrations of defects, be responsible for poor electrical characteristics of top-gate TFTs? Or is the amorphous silicon surface damaged during the deposition of α-SiN:H on top of α-Si:H? More work remains to relate the film structures to internal plasma properties such as the SiHx, H, N, and ion concentrations, and to properties of TFTs.
Conclusion
ATR-FTIR can be used for rapid process development and optimization, as well as for obtaining fundamental information on surface and interface chemistry in plasma systems. However, relatively complicated equipment and the need for special substrates to obtain multiple internal reflections limit the technique to R&D labs. Both data acquisition and analysis can be complicated, particularly for thick inhomogenous films and for spectral overlap in the absorption bands of several species.
With improvements in data acquisition and analysis, increases in sensitivity would allow the technique to investigate plasma processes that more closely mimic IC production conditions (i.e., high-deposition-rate PECVD). Nevertheless, much can be learned even if manufacturing conditions cannot be exactly duplicated. ATR-FTIR has already provided valuable information on a number of key processes relevant to microelectronics manufacturing.
Acknowledgments
The authors acknowledge funding from the National Science Foundation (ECS 9457758), Advanced Research Project Agency (DAAH04-95-1-0059) and Lam Research Corp. for research on use of ATR-FTIR in plasma processing. Thanks are also due to A. Bailey, S. Deshmukh, S. Han, D. Marra, and Z. Zhou, who contributed to the example results discussed in this article.
References
- B. Drevillon, Applied Surface Science 63, p. 27, 1993.
- I. Tepermeister, et al., Solid State Technology 39, March 1996, p. 63, 1996.
- I.P. Herman, V.M. Donnelly, K.V. Guinn, C.C. Cheng, Physical Review Letters 72, Vol. 17, p. 2801, 1994.
- E.S. Aydil et al., Appl. Phys. Lett. 62, p. 3156, 1993.
- E.S. Aydil, R.A. Gottscho, Y.J. Chabal, Pure & Appl. Chem. 66, p. 1381, 1994.
- E.S. Aydil, Z.H. Zhou, R.A. Gottscho, Y.J. Chabal, J. Vac. Sci. Technol. B 13, p. 258, 1995.
- Y.J. Chabal, Surf. Sci. Repts. 8, p. 211, 1988.
- N.J. Harrick, Internal Reflection Spectroscopy, Wiley, New York, 1967.
- K. Kawamura, S. Ishizuka, H. Sakaue, Horiike, Jpn. J. Appl. Phys. 11B, p. 3215, 1991.
- Z. Zhou, E.S. Aydil, R.A. Gottscho, Y.J. Chabal, R. Reif, J. Electrochem Soc. 140, p. 3316, 1993.
- S.C. Deshmukh, E.S. Aydil, J. Vac. Sci. Technol. A 13, p. 2355, 1995.
- S.M. Han, E.S. Aydil, J. Vac. Sci. Technol. A 14, p. 2062, 1996.
- S.M. Han, E.S. Aydil, Thin Solid Films 291, p. 427, 1996.
- A.D. Bailey, R.A. Gottscho, Jpn. J. Appl. Phys. 34, p. 2172, 1995.
Eray S. Aydil received BS degrees in chemical engineering and materials science from the University of California, Berkeley, in 1986, and his PhD degree in chemical engineering from the University of Houston in 1991. After postdoctoral work at AT&T Bell Laboratories, he joined the faculty at the University of California Santa Barbara in 1993, where he is assistant professor in the chemical engineering department. Aydil is the author and coauthor of over 40 papers, and holds four patents on plasma processing and diagnostics. He has received the National Young Investigator Award of the National Science Foundation, the Norman Hackerman Award of the Electrochemical Society, and, most recently, the Camille-Dreyfus Award of the Dreyfus Foundation for his research in this field. University of California Santa Barbara, Santa Barbara, CA 93106; ph 805/893-8205, fax 805/893-4731, e-mail [email protected], www.eci2.ucsb.edu/ce/faculty/aydil/.
Richard A. Gottscho received his BS and PhD degrees in physical chemistry from Pennsylvania State University and MIT, respectively. He is director of plasma chemical vapor deposition for Lam Research Corp. Previously, he served as director of corporate research at Lam, and before that, he headed the display research and electronics packaging research departments at AT&T Bell Laboratories, where he worked from 1980 through 1995. Gottscho has authored numerous papers and lectures and holds a number of patents in plasma processing.