In Situ Monitoring of Hafnium Oxide Atomic Layer Deposition

In Situ Monitoring of Hafnium Oxide Atomic Layer Deposition J.E. Maslar*, W.S. Hurst, D.R. Burgess, Jr., W.A. Kimes, N.V. Nguyen, and E.F. Moore National Institute of Standards and Technology 100 Bureau Drive, Stop 8360, Gaithersburg, MD 20899 Abstract. Atomic layer deposition (ALD) is increasingly being utilized as a method of depositing the thin (nanometerscale), conformal layers required for microelectronics applications such as high κ gate dielectric layers and diffusion barriers. However, significant process development issues remain for implementation of this technology in many applications. One potential solution to some process development issues is in situ monitoring. In situ monitoring of atomic layer deposition processes has the potential to yield insights that will enable efficiencies in film growth, in the development of deposition recipes, and in the design and qualification of reactors. However, demonstrations of in situ monitoring of actual atomic layer deposition processes are limited. In this work, the species present in the gas phase during atomic layer deposition of hafnium oxide were investigated in an attempt to gain insight into the chemistry of this system and evaluate potential in situ gas phase optical monitors. Hafnium oxide was deposited on a silicon substrate using tetrakis(ethylmethylamino) hafnium (TEMAH) and water as the hafnium and oxygen sources, respectively. In situ infrared absorption spectroscopic measurements were performed near the growth surface in a research-grade, horizontalflow reactor under a range of deposition conditions. Density functional theory quantum calculations of vibrational frequencies of expected species were used to facilitate identification of observed spectral features. Keywords: Atomic Layer Deposition; In Situ Diagnostics; Infrared Spectroscopy. PACS: 82.33.Ya; 77.84.Bw; 82.80.Gk; 31.15.Ew

i.e., the range in which this technique can be used to help optimize gas injection conditions rather than simply monitor precursor delivery. Gas phase Atomic layer deposition (ALD) is increasingly chemistry of concern includes gas phase dissociation being utilized as a method of depositing the thin and gas phase reaction with other precursors, e.g., (nanometer-scale), conformal layers required for residual water in the chamber. This investigation was microelectronics applications such as high κ gate prompted in part by reports of gas phase dissociation dielectric layers and diffusion barriers. However, of the alkyl amide compounds tetrakis(dimethylamino) significant process development issues remain for zirconium[1] and tetrakis(dimethylamino) titanium implementation of this technology in many and tetrakis(diethylamino) titanium[2-4] and gas phase CREDIT LINE (BELOW) TO BE INSERTED ON THE FIRST PAGE OF EACH applications. One potential solution to some process mixing of tetrakis(ethylmethylamino) hafnium PAPER EXCEPT THE PAPERS ON PP. 121 – 125, 146 – 150, 151 – 155, 156 – development issues is in situ monitoring. In situ (TEMAH) and water.[5] All of these reports 160, 161 – 167, 168 – 172, 185 – 190, 209 – 215, 287 – 291, 308 – 314, monitoring of atomic layer deposition processes has employed FTIR467 spectroscopy for472 in situ–observation of 386, – 401, – 412, 428 – 431, – 471, 382 – to the potential yield397 insights that 407 will enable gas – phase However, none of the 485,in525 – 529, and 534chemistry. 476, in483 efficiencies film –growth, the development of 530 investigations were conducted in ALD chambers deposition recipes, and in the design and qualification capable of growing high quality films. of reactors. However, demonstrations of in situ CP931, Frontiers of Characterization and Metrology for Nanoelectronics, This report involves an effort to determine the monitoring of edited ALDbyprocesses in actual ALD reactors D. G. Seiler, A. C. Diebold, R. McDonald, C. M. Garner,toD. which, Herr, R. P. E. M. Secula extent if Khosla, at all,andTEMAH reacts prior to are limited, especially in the case of American gas phase Fourier © 2007 Institute of Physics 978-0-7354-0441-0/07/$23.00 reaching the growth surface in an ALD reactor under transform infrared (FTIR) spectroscopy realistic gas flow conditions, i.e., relatively short In this work, gas phase FTIR spectroscopy was residence times and relatively low gas temperatures. investigated as a potential in situ diagnostic for CREDIT LINE (BELOW) TO BE INSERTEDMeasurements ONLY ON THE FIRST PAGE THE performed in aOF single-wafer, hafnium oxide ALD. A PAPERS: major goal of PP. this work to 146 – 150, 151– were FOLLOWING 121 –was 125, 155, 156 – 160, 161 – 167, warm-wall, horizontal-flow reactor. Additional design determine the 172, process range – under which 168 – 185parameter – 190, 209 215, 287 – characteristics 291, 308 – of 314, 382 – 386, 397 –access this reactor include optical FTIR401, spectroscopy to gas–phase 412, 428 431,chemistry, 467 – 471, 472 – 476, 483 – 485, 525 – 407 –is sensitive

INTRODUCTION

529, and 530 – 534 CP931, Frontiers of Characterization and Metrology for Nanoelectronics, edited by D. G. Seiler, A. C. Diebold, R. McDonald, C. M. Garner, D. Herr, R. P. Khosla, and E. M. Secula 2007 American Institute of Physics 978-0-7354-0441-0/07/$23.00

121

contamination from the heater occurs. An aluminum plate was used to hold a substrate and was mounted on the front of the chuck base chamber. A mineralinsulated metal-sheathed (MIMS) type K thermocouple inserted into a thermocouple well in the face of the base chamber was used for temperature control. Similarly, a MIMS thermocouple mounted into a thermocouple well in the aluminum plate was used for substrate temperature monitoring. The standard measurement uncertainty for these thermocouples is 2.2 °C at 250 °C. To reduce the amount of water adsorbed to the delivery line walls, the lines were heated to ≈110 °C. Microelectronics-grade TEMAH (Air Products/Schumacher)[7] was purchased in a stainless steel bubbler and employed as received without further purification. For delivery, the TEMAH bubbler was heated to ≈80 ºC to achieve a TEMAH vapor pressure of ≈13.3 Pa (0.1 torr).[8] Water was purified using a reverse osmosis membrane, an ultraviolet oxidizer, and an ion exchange resin bed. The resulting purified water had a resistivity of 18 MΩ · cm or greater and a pH of 5.76, at 21 °C. The carrier and purge gas was helium (from the same high pressure cylinder) that was obtained as ultra-high-purity grade gas (99.999 % pure) and further purified with a bulk purifier. Deposition was performed on 50 mm diameter (100) silicon wafers. Prior to deposition, wafers were dipped in a 2 % HF solution for 30 s and then exposed to ozone for 4 min.

near the wafer surface, good gas flow characteristics (facilitating reproducible high quality film growth), and an aluminum body (facilitating maintenance of a uniform wall temperature). FTIR spectroscopic measurements were performed under a range of deposition conditions. Density functional theory (DFT) quantum calculations of vibrational frequencies of expected species were performed and were used to facilitate identification of observed spectral features.

EXPERIMENTAL METHOD ALD Reactor System Deposition was performed in a custom designed and built single-wafer, warm-wall, horizontal-flow reactor and pulsed gas injection system. The reactor incorporates optical access near the wafer surface with good gas flow characteristics. The performance and operation of this deposition system have been previously described.[6] Two precursor lines and one extended purge line (with purge gas injection into the chamber through two ports from this line) were employed. Helium was used as the purge/carrier gas for all measurements. For all spectra presented in this work, the nominal flow rate (at standard temperature and pressure) was 75 mL/min (sccm) through each precursor line and 150 mL/min through the purge line for a total nominal flow rate of 300 mL/min in the chamber. Helium gas was always flowing through all three lines, but a precursor was not always entrained in the gas flow. All valves were fast-switching pneumatic diaphragm valves. The nominal chamber pressure was ≈133.3 Pa (1 torr) as measured with a capacitance manometer. The reaction chamber was a custom-fabricated aluminum tube with a ≈102 mm internal diameter. Optical windows measuring 50 mm in diameter were mounted on opposite sides of the reactor such that the optical path was across the wafer surface. The windows were sealed to the reactor using O-rings that provided a clear aperture measuring ≈41.9 mm in diameter. The distance from the inside surface of the reactor wall to the window inside surface varied from ≈1.5 mm to ≈5 mm. A 50 mm diameter substrate was mounted vertically on a custom-designed chuck such that the gas stream impinged on the substrate surface parallel to the surface normal. The distance from the gas injection port to the wafer surface was ≈253 mm. The chuck consisted of two parts. The base was a stainless steel, sealed chamber containing a 50 mm diameter etchedfoil resistance heater patterned to give uniform temperature over its heated area. Isolation of the heater from the deposition chamber insures that no

Infrared Absorption Measurements In situ IR absorption measurements were performed using a commercial FTIR spectrometer configured as shown in Fig. 1. The instrument module containing the mid-IR source and interferometer was decoupled from the detector module and three uncoated ellipsoidal aluminum mirrors were used to collect the IR radiation converging out of the interferometer module, transfer the radiation through the ALD reactor chamber, and focus the radiation onto a mercury cadmium telluride detector. A potassium bromide beamsplitter was used for all measurements. The instrument module was purged with dry nitrogen; however, the beam path between the instrument module and the deposition chamber and between the chamber and the detector was not enclosed. The direction of propagation of the IR beam was perpendicular to the wafer surface normal. Uncoated zinc selenide windows were employed on the ALD chamber. FTIR spectra presented in this investigation were the result of 220 to 650 co-added scans each of which were ratioed to a corresponding background spectrum. Background spectra were obtained just

122

heated. The monitor thermocouple temperature was 102 °C, with the corresponding wafer surface temperature estimated to be 102 °C ±5 °C. Spectra recorded during ALD runs were obtained under nominally identical gas injection conditions except for the duration of PWater which was varied from 15 s to 0.1 s. The TWater, TTEMAH, and PTEMAH values were 0.1 s, 5 s, and 5 s, respectively. During ALD cycles, power was supplied to the wafer chuck heater such that the monitor thermocouple temperature was ≈255.5 °C, with the corresponding wafer surface temperature estimated to be 237 °C ±5 °C (K=1).

prior to a data scan and were recorded with the same spectral conditions and under the same gas injection conditions except with only helium flowing into the chamber. All spectra were recorded with 4 cm-1 spectral resolution. Gas Injection

Ellipsoidal Mirrors

HgCdTe Detector

RESULTS AND DISCUSSION

FTIR

ZnSe Windows

Reference Spectra

ALD Reactor

Measured reference spectra were compared to spectra calculated with DFT using B3LYP theory [9,10] with LANL2DZ basis sets [11] and assuming a temperature of 25 °C. The calculations were performed using commercial software (Gaussian 03, Gaussian, Inc)[7] and the high-performance computational capabilities of the Biowulf Linux cluster at the National Institutes of Health.[12] The final peak wavenumbers in the DFT spectra were obtained by scaling the peak wavenumbers in the C-H stretching spectral region, ≈ 2800 cm-1 to 3150 cm-1 (before scaling), by 0.95 and peaks at other wavenumbers by 0.961. Figure 2(a) shows the measured (denoted “Measured”) and calculated (denoted “Calculated”) TEMAH reference spectra. Figure 2(b) shows the measured and calculated MEA reference spectra. The FTIR instrument lineshape function was not determined, so the DFT spectra are shown with infinitely narrow spectral linewidths. Although a direct comparison of the measured and calculated spectral intensity profiles is difficult, the agreement between the experimental and calculated spectra is good for both TEMAH and MEA. The agreement between the measured spectrum in Fig. 2(b) and the calculated MEA spectrum and the lack of any clear TEMAH-related spectral features in the experimental spectrum in Fig. 2(b) forms the basis of the identification of this spectrum as that of MEA. While not all IR modes from the DFT calculations can be unambiguously attributed to specific molecular vibrations, a few modes can be identified with a relatively high degree of certainty. From the DFT calculations for TEMAH, the features observed in the measured TEMAH spectrum at ≈980 cm-1 and ≈1154 cm-1 are attributed to N-C stretching modes while the feature observed at ≈875 cm-1 is attributed to a C-C-N-C backbone mode. From the DFT

FIGURE 1. The experimental configuration for in situ IR measurements.

Reaction precursor and product reference spectra were obtained by recording spectra during TEMAH injection and (nominally) complete precursor mixing conditions, respectively. The TEMAH spectra were obtained during injection of 20 cycles of 5 s of TEMAH (denoted TTEMAH) followed by 5 s of helium purge (PTEMAH). The same reactor was used for TEMAH reference spectra measurement as for measurements during ALD. Spectra of (nominally) completely mixed precursors were obtained in a longresidence time chamber, as previously described.[5] These spectra were obtained during injection of 26 cycles with 3 s TTEMAH, 0.1 s PTEMAH, 1 s water injection time (TWater), and 0.1 s helium purge following water injection (PWater). TEMAH and (excess) water were injected in rapid succession, i.e., short purge times, so as to promote gas phase precursor mixing and complete consumption of TEMAH in the gas phase. Assuming that gas-phase TEMAH and water are present at the same time in the deposition chamber and that TEMAH reacts completely the gas phase reaction can be described by: Hf[N(C2H5)(CH3)]4 + 2H2O → HfO2 + 4HN(C2H5)(CH3)

(1)

where HfO2, in this case, is a particulate formed in the gas phase and HN(C2H5)(CH3), subsequently denoted MEA, is a gas phase reaction product. During generation of these reference spectra, no power was supplied to the wafer chuck heater but the TEMAH bubbler, gas delivery lines, and reactor walls were still

123

is unreacted TEMAH. This result indicates that laminar flow is probably maintained in the reactor near the wafer surface.

calculations, the Hf-N stretching modes are expected at ≈560 cm-1. From the DFT calculations for MEA, the feature observed in the measured spectrum at ≈1145 cm-1 is attributed to N-C stretching modes. However, it is difficult to deduce the actual lineshape of this feature due to interference from the bending modes of laboratory ambient water that were not removed in the background subtraction process. In both measured spectra, the C-H stretching modes are observed from ≈2750 cm-1 to 3000 cm-1.

(/2)

Absorbance

1 x 10

(a)

0.1 s PWater

5 s PWater 10 s PWater

Calculated

15 s PWater

1 x 10

-3

TEMAH

Absorbance

Measured

800

1000 1200 1400

2600 2800 3000 3200 -1

Wavenumber (cm

(b)

1 x 10

Despite the similarity of all ALD spectra with the TEMAH spectrum, the 0.1 s PWater spectrum and possibly the 5 s PWater spectrum do exhibit features that can not be attributed to unreacted TEMAH. In addition to TEMAH-related features, the 0.1 s PWater spectrum exhibits a sloping background in the low wavenumber region and two additional peaks in the C-C/N-C spectral region at ≈1050 cm-1 and Differences between the 5 s PWater ≈1115 cm-1. spectrum and the TEMAH spectrum are slight. However, the 5 s PWater spectrum exhibits some very weak peaks at ≈1050 cm-1 and ≈1115 cm-1, wavenumbers at which more intense features were observed in the 0.1 s PWater spectrum. No intense MEA features are observed in any spectra, although TEMAH signal may interfere with observation of weak MEA signal. In addition, the intensity of the ≈2870 cm-1 feature relative to the intensities of the other C-H-related features is significantly higher in the 0.1 s PWater spectrum than in the other ALD spectra. This is in contrast to the MEA spectrum in which the intensity of the ≈2870 cm-1 feature relative to the intensities of the other C-H-related features is relatively low. Further, the MEA spectrum does not exhibit features at ≈1050 cm-1 and ≈1115 cm-1. Hence, the observed differences in the 0.1 s PWater spectrum and the 5 s PWater spectrum compared to the other ALD spectra are probably not due to MEA. It is uncertain

-3

Measured

1000 1200 1400

2600 2800 3000 3200 -1

Wavenumber (cm

)

FIGURE 3. In situ IR spectra recorded during ALD cycles as a function of purge time after water injection. The spectra are offset on the absorbance axis for clarity.

Calculated

800

-3

)

FIGURE 2. The measured and calculated IR spectra of (a) TEMAH and (b) MEA. The spectra are offset on the absorbance axis for clarity.

In Situ Spectra During ALD Figure 3 shows the in situ IR spectra obtained during a series of ALD runs as well as the TEMAH spectrum for comparison. All gas injection conditions for the ALD runs were nominally identical except for PWater values which were varied from 15 s to 0.1 s. The C-C-N-C feature (≈875 cm-1), N-C features (≈980 cm-1 and ≈1154 cm-1), and TEMAH-like C-H features (≈2750 cm-1 to 3000 cm-1) are observed in all ALD spectra and correspond closely to the same features in the TEMAH spectrum, although the shorter duration PWater spectra exhibit some differences from the TEMAH spectrum. The similarity between the TEMAH and ALD spectra indicates that the primary contribution to the observed signal in all ALD spectra

124

what species are responsible for the observed differences in spectra, however, some tentative estimations can be made. The origin of the ≈1050 cm-1 and ≈1115 cm-1 features in the 0.1 s and (weakly) 5 s PWater spectra are unknown, as is the origin of the relatively intense ≈2870 cm-1 feature in the 0.1 s PWater spectrum. It is known that MEA is not responsible for these features and, hence, it seems likely that these features are due to partially reacted or dissociated TEMAH. The flow rates and partial pressures of reactants are nominally the same for all measurements. However, the 0.1 s PWater spectrum exhibits a significant increase in overall signal intensity. This suggests that the species responsible for this increase in intensity are not entrained in the gas flow. It seems likely that these species are deposited on the optical windows. Also, the sloping background in the 0.1 s PWater spectrum is attributed to absorption and/or scattering by HfO2 particles formed in the gas phase that are either deposited on the reactor windows or entrained in the gas in the optical path during measurements. Hence, it is possible that the species deposited on the windows are particles that were formed in the gas phase by partial reaction or dissociation of TEMAH and subsequently deposited on the windows.

R. Pavin for helpful discussions and R. R. Fink and M. J. Carrier for technical assistance.

REFERENCES 1.

I. W. Kim, S. J. Kim, D. H. Kim, H. Woo, M. Y. Park, and S. W. Rhee, Kor. J. Chem. Eng. 21, 1256-1259 (2004). 2. L. H. Dubois, B. R. Zegarski, and G. S. Girolami, J. Electrochem. Soc. 139, 3603-3609 (1992). 3. J. Y. Yun, M. Y. Park, and S. W. Rhee, J. Electrochem. Soc. 145, 2453-2456 (1998). 4. J. Y. Yun, M. Y. Park, and S. W. Rhee, J. Electrochem. Soc. 146, 1804-1808 (1999). 5. J. E. Maslar, W. S. Hurst, D. R. Burgess, W. A. Kimes, and N. V. Nguyen, ECS Transactions 2, 133-143 (2006). 6. J. E. Maslar, W. A. Kimes, E. F. Moore, W. S. Hurst, and N. V. Nguyen, Rev. Sci. Instrum., submitted (2007). 7. Certain commercial equipment, instruments, and materials are identified in this publication to adequately specify the experimental procedure. Such identification in no way implies approval, recommendation, or endorsement by the National Institute of Standards and Technology, nor does it imply that the equipment, instruments, or materials identified are necessarily the best available for the purpose. 8. D. M. Hausmann, E. Kim, J. Becker, and R. G. Gordon, Chem. Mater. 14, 4350-4358 (2002). 9. A. D. Becke, J. Chem. Phys. 98, 5648-5652 (1993). 10. C. T. Lee, W. T. Yang, and R. G. Parr, Phys. Rev. B 37, 785-789 (1988). 11. P. J. Hay and W. R. Wadt, J. Chem. Phys. 82, 270-283 (1985). 12. Biowulf Linux Cluster. http://biowulf.nih.gov (accessed 2006).

CONCLUSIONS Hafnium oxide ALD from TEMAH and water was investigated using in situ FTIR spectroscopy measurements performed in a research-grade, horizontal-flow reactor under a range of deposition conditions. The species present in the gas phase during atomic layer deposition of hafnium oxide were observed. Density functional theory quantum calculations of vibrational frequencies of TEMAH and the primary expected product, MEA, were performed and compared to measured spectra. Unreacted TEMAH was the primary component of spectra under all injection conditions, consistent with laminar gas flow near the wafer surface. However, with water purge times of 0.1 s and possibly 5 s, contributions to the spectra other than unreacted TEMAH were observed. No indications of MEA formation were observed. The observed differences in spectra were attributed to partially reacted or dissociated TEMAH deposited on the optical windows, possibly in the form of particles that originated in the gas phase.

ACKNOWLEDGMENTS The authors gratefully acknowledge the NIST Office of Microelectronics Programs for funding,

125