Interfacial reaction and thermal stability of Ta2O5/TiN for metal electrode capacitors
02/01/1999
FIRST IN A SERIES
Interfacial reaction and thermal stability
of Ta2O5/TiN for metal electrode capacitors
J.P. Chang, R.L. Opila, G.B. Alers, M.L. Steigerwald, Bell Labs, Lucent Technologies, Murray Hill, New Jersey, H.C. Lu, E. Garfunkel, T. Gustafsson, Rutgers, The State University of New Jersey, Piscataway
High dielectric constant (k) materials are being considered as a replacement for SiO2 and SiN in capacitors for DRAM and RF circuits as critical dimensions of ICs continue to decrease. The transition to a nonsilicon-based dielectric in these applications is not simply a change in materials. It also represents a fundamental change in processing toward deposited dielectrics and away from those that can be thermally grown on bulk or polycrystalline silicon. Dielectrics formed by thermal growth on a similar material have many advantages, including good step coverage, thickness uniformity, and control of interfacial properties. To retain these advantages with deposited materials, one must have good control of interfacial growth properties, such as with epitaxial film growth. We discuss here the importance of the high-k dielectric/metal interface in deciding electrical properties of deposited metal oxide thin films.
Tantalum pentoxide (k|25) has been proposed as a replacement for SiO2 or SiN that is compatible with silicon processing [1-4]. The preferred method for depositing Ta2O5 thin films is CVD, which has an advantageous low process temperature (<500?C), with excellent conformal step coverage properties. This makes Ta2O5 a good candidate for metal-oxide-metal (MOM) capacitors, where the metal electrode materials might include W, TiN, or TaN. Silicon electrodes have a problem: a silicon oxide layer can form at the Ta2O5/Si interface, lowering the specific capacitance [5]. Thus, a thin layer of SiN or SiOxNy is often grown or deposited at the Si surface to prevent interfacial diffusion during the CVD process and subsequent anneals [6]. Even a very thin layer of SiOxNy at the interface, however, will increase the dielectric`s effective thickness and reduce the benefit of a high-k material. Using metals for the electrode could circumvent this problem if they do not form their own oxides. Metal electrodes also increase capacitance densities (above 25 fF/?m2), compared to those obtained with Si-based electrodes (10 fF/?m2), with the same Ta2O5 thickness and maintaining low leakage currents [5, 7]. We will focus on one typical structure proposed for a MOM-based capacitor - a Al/TiN/Ta2O5/TiN/Ti/Si layered stack - where the interfacial properties between Ta2O5 and TiN are critical and may significantly affect the device`s electrical performance. Bulk properties, including the amount of carbon incorporated from the precursor in the Ta2O5 films during the deposition process, may also affect the device`s electrical properties.
While Ta2O5 with low electrical leakage may be deposited at relatively low temperatures (<500?C), it is crucial that the finished MOM capacitor withstand subsequent processing at higher temperatures. We performed experiments to investigate the thermal stability of Ta2O5 films grown in Ta2O5/TiN/Ti/Si and Ta2O5/TiN/Si stacks and found that the presence of metallic Ti was detrimental to the thermal stability of Ta2O5. TiN was apparently not an adequate diffusion barrier to prevent oxygen depletion from the Ta2O5 and Ti oxidation. We found that the temperature of the CVD deposition played a crucial role in the interfacial and electrical properties.
Deposition of Ta2O5 films
The structures we studied were Ta2O5/TiN/Ti/Si and Ta2O5/Si multilayers. We used an argon plasma to sputter-clean the silicon wafer substrates for 30 sec in a Novellus Systems M2000 tool, then Ti and TiN were deposited by sputtering a titanium target in situ at a substrate temperature of 300?C in argon and a mixture of argon and nitrogen, respectively. The nominal thicknesses of the deposited films were 30-nm TiN and 30-nm Ti. The Ta2O5 films were deposited using tantalum tetraethoxide dimethylaminoethoxide (TAT-DMAE) as a precursor in an Applied Materials` high-temperature film (HTF) tool, where the sample was lamp-heated and a mixture of N2 and O2 acted as a carrier gas for the precursor. We used a Gasonics International L3510 photoresist strip tool for post-deposition anneals in a remote O2 or N2/O2 microwave plasma at a substrate temperature of 300?C. An approximately 9-nm thick layer of Ta2O5 was deposited to fabricate capacitors for electrical testing with an Al/TiN/Ta2O5/TiN/Ti/Si structure.
We used thinner Ta2O5 deposited films (=5 nm comparable to the photoelectrons` escape length) to allow x-ray photoelectron spectroscopy (XPS) analysis to determine the compositions and chemical states of the Ta2O5 film and its interface with the TiN. These films, however, were too thin for reliable electrical measurements. We used a monochromatic Al Ka source (1486.6 eV) for XPS, with the pass energy typically set at 11.75 eV to obtain high-resolution photoemission spectra at a take-off angle of 75?. The sample was heated in situ to 900?C in the XPS vacuum chamber to determine the thermal stability of the Ta2O5 films. An ion source in the XPS chamber facilitated sputtering of the sample by Ar+ (4 keV) to determine the elemental concentration profiles. Preferential sputtering of light atoms (e.g., O) over heavier atoms (e.g., Ta) occurred on the surface, requiring careful interpretation of the experimental data.
We performed secondary ion mass spectroscopy measurements (SIMS) on a film stack of 50-nm Ta2O5 deposited at 400?C, followed by a plasma anneal treatment in a N2/O2 mixture and an additional 50-nm deposition of Ta2O5 on the surface. Analysis of this film stack allowed an investigation of the effects of plasma treatment on the surface of the Ta2O5, with less concern over surface contamination. We determined the microstructure and the surface roughness of the Ta2O5 films via transmission electron microscopy (TEM) and atomic force microscopy (AFM). Medium energy ion scattering (MEIS), a technique similar to Rutherford backscattering with incident H+ energies of 100 keV, was also used to determine the concentration profiles for elements in the film stacks to study the thermal stability of Ta2O5 deposited directly on Si [8, 9].
Interfacial reaction between Ta2O5 and TiN
The Ta(4f) photoemission spectrum with its (4f7/2) and (4f5/2) spin orbit pairs at 26.8 eV and 28.7 eV was clearly resolved, indicating that the tantalum was fully oxidized to form Ta2O5. Our results showed the Ta2O5 layer was nearly stoichiometric. The titanium photoemission spectra from below a 5-nm Ta2O5 film in a Ta2O5 /TiN/Ti/Si stack before and after an O2 anneal showed distinct multiple states overlapping with a broad Ta(4p1/2) photoemission line at 467.8 eV. The Ti(2p3/2) spectra were de-convoluted into three features: TiN, Ti-suboxide, and TiO2 at 455.5, 457.0, and 458.8 eV, respectively. For each of the chemical states, a Ti(2p1/2) spin orbital split component was resolved, assuming a spin-orbit splitting of 6.2 eV and an intensity ratio (2p3/2/2p1/2) of 2 (Fig. 1). This suggests that the top surface of the Ti was oxidized during the vacuum break and/or during the deposition of Ta2O5. An O2 anneal greatly reduced the detectable amount of suboxides by converting part of the suboxides to TiO2. A thicker layer of TiO2 at the interface further attenuated the Ti-N intensity. Furthermore, Fig. 2 shows both O2 and O2/N2 annealing significantly, reducing the leakage current [6].
From the SIMS spectra obtained for the thick Ta2O5 multilayer, we found that the oxygen plasma depleted the near surface (~25 nm) region of carbon, presumably due to volatile CO species forming at the surface, consistent with previous work [7]. We can thus conclude that for a 10-nm film, the oxygen from the plasma not only penetrated through the bulk of the film, but also reached the Ta2O5/TiN interface to oxidize the titanium suboxides.
The plasma anneal can have other effects on the film in addition to reduction of bulk traps and passivation of the interface. However, we found that interfacial interactions dominate the leakage properties beyond the effects of carbon. The amount of carbon from
 |
Figure 1. The Ta(4P1/2) and Ti (2p3/2) photoemission spectra in chemical states: TiN, Ti-suboxide, and TiO2 at 455.5, 457.0, and 458.8 eV, respectively, at a) no anneal and b) O2 anneal. The bright red line is the sum of the individual resolved features, which closely fits the raw data (black dots).
 |
Figure 2. A remote plasma anneal in O2 or O2/N2 gas at 300?C greatly reduced the leakage currents in Ta2O5 films while retaining a capacitance of 25 fF/?m2.
the precursor TAT-DMAE incorporated in the Ta2O5 sample depended strongly on the deposition conditions. Less carbon was incorporated in the film at lower pressure and higher deposition temperature, as shown by the XPS spectrum (Fig. 3a) obtained after sputtering the film for 30 sec in vacuum to remove surface carbon contamination. The C(1s) peak position suggests that carbon formed Ta-C structures in the film. Approximately 9 atomic percentage (at.%) of carbon was incorporated in the Ta2O5 film when deposited at 375?C, whereas a film deposited at 450?C contained 2 at.% carbon. However, the film deposited at 375?C had significantly less leakage than the film deposited at 450?C, shown in Fig. 3b, contrasting with the assumption that lower carbon content in the Ta2O5 films leads to lower leakage current. We find that the interfacial oxidation of TiN for the 450?C deposited films was much more severe than in the films deposited at 375?C, suggesting that the TiN/Ta2O5 interface may be more important in determining the electrical properties than carbon content. A subsequent treatment in a remote plasma of O2/N2 reduced leakage of both of these samples, but the post-annealed leakage of the film deposited at 375?C sample was the lowest. Careful studies are necessary, however, to resolve the relative importance of bulk and interfacial states on the device`s electrical performance.
Thermal stability of Ta2O5
To characterize the thermal behavior of typical MOM capacitors, we monitored the evolution of the elemental composition as a function of temperature of Ta2O5/TiN/Ti/Si substrate multilayers. While heating the unannealed sample in situ to 400? and 600?C, we determined the photoemission of Ta, O, C, and Ti at the end of each temperature ramp (20?C/min) to the desired temperature. Figure 4a shows the evolution of the Ta(4f) photoemission spectra at various annealing temperatures. At 30?C, the Ta (4f7/2) binding energy was approximately 26.8 eV, which is characteristic of Ta2O5. At 600?C, the Ta (4f) binding energy shifted to approximately 22.5 eV. The shift by roughly 4.5 eV to lower energy indicates that the highly oxidized Ta (as in Ta2O5) was reduced to metallic Ta. Evidence for the reduction of the Ta2O5 layer in this sample is the reduction in the oxygen XPS signal. The O(1s) photoemission intensity was greatly reduced after heating the sample to 600?C, (Fig. 4b), suggesting that elemental Ti could reduce Ta2O5 during vacuum annealing, even though separated from the Ta2O5 layer by a 30-nm TiN layer. More recent work has proven the reduction of Ta2O5 is induced by the metal layers below the barrier layer [7].
The O migrated from the Ta2O5 layer to the Ti layer during the heat treatment. We determined this by comparing the concentration of each element as a function of depth from the surface before and after heating. A 4-keV Ar+ beam was used to sputter the films in situ and to determine the concentration profiles of various species. Figure 5 shows the O, Ti, and N concentration profiles. The O concentration in the surface layer was high prior to heating and significantly lower after heating. Accompanying the O depletion from the surface layer was an increase in O concentration at the TiN/Ti interface and in the Ti layer. The increased O concentration at and near the TiN/Ti interface was accompanied by a decrease in the relative concentration of Ti in the same region. We can conclude that the Ti in this region was at least partially oxidized, although we cannot determine the exact local TiOx stoichiometry. Note that the N concentration profile did not change during the heating. Comparison of initial and final curves showed no discernible difference in the O concentration within the TiN layer. Thus the O moved through the TiN but did not remain there.
 |
Figure 3. a) The amount of carbon incorporated in the Ta2O5 sample depends strongly on the deposition conditions. b) The film deposited at 375?C had the lowest leakage current (with no anneal process) but the most carbon incorporated in the film.
 |
Figure 4. Evolution of the a) Ta(4f) photoemission and b) O(1s) photoemission lines from a Ta2O5/TiN/Ti/silicon sample during in situ vacuum heating.
Thermal gravimetric analysis measurements of bulk material showed direct thermal decomposition of Ta2O5 to give elemental tantalum and diatomic oxygen, rather than being reduced by Ti, did not occur at temperatures used in these experiments [10]. The stability of the Ta2O5 films in the absence of Ti confirmed the assumption that Ta2O5 did not lose oxygen to the vacuum.
A Ta2O5/TiN/SiO2 sample with no underlying Ti metal was heated at 400?, 600?, 700?, and 800?C. The tantalum remained in its high oxidation state at temperatures up to approximately 800?C. The concentration profiles of Ta, O, C, Ti, and N of a Ta2O5/TiN/SiO2 sample before and after vacuum annealing are similar.
The grain structure of the TiN barrier layer was determined by TEM. The columnar boundary structures may serve as the diffusion pathways for oxygen. The TEM image of a sample (not shown) indicates the formation of a TiOx layer at the TiN/Ti interface after being heated to 600?C.
For applications that require a thermal treatment >800?C, SiN electrodes are usually preferred over TiN due to the thermal stability issues. MEIS examination of the Ta2O5/Si interface showed stability problems for temperatures >800?C, with limited silicon out-diffusion into the Ta2O5. Figure 6 shows the concentration profiles of elements in a Ta2O5/Si bilayer before and after vacuum heating to 890?C, where MEIS and TEM studies of the film suggested a 7-nm Ta2O5/ 2-nm SiOx/ Si structure. After heating the system in ultra-high vacuum, we observed limited silicon diffusion into the region between Ta2O5 and substrate silicon and a decrease in net oxygen intensity, implying that oxygen was lost as H2O, or possibly SiO [11]. Recrystallization-induced roughness became significant at annealing temperatures >800?C, and complicated interpretation of both MEIS and XPS depth profiling. Interdiffusion and roughening at either Ta2O5 interface further justified the consideration of barrier layers. A thin Si3N4 buffer layer stabilized the Ta2O5/Si interface, with limited silicon out-diffusion and oxygen loss at a higher temperature of 970?C.
 |
Figure 5. Concentration profiles in a Ta2O5/TiN/Ti/Si multilayer before (open symbols) and after (solid symbols) heating to 600?C in vacuum.
 |
Figure 6. Medium energy ion scattering results indicate silicon out-diffusion and a decrease in net oxygen intensity, implying that oxygen is lost as H2O and possibly SiO. The open and solid symbols represent the concentration profiles before and after heating in vacuum to 890?C, respectively.
Discussion
We have discussed interfacial reactions and thermal stability between Ta2O5 and TiN in making MOM capacitors. Significant amounts of titanium suboxides were also seen at the Ta2O5/TiN interface. We believe the imperfect interface reduced the specific capacitance and increased leakage currents. An O2 plasma anneal converted the suboxides to titanium oxide and improved electrical performance. At higher deposition temperature, we found greater interaction between TiN and the Ta2O5, leading to greater leakage current. Increased carbon content in the low-temperature deposited films, however, did not appear to be detrimental.
We also discovered that the thermal behavior of Ta2O5 was dramatically affected by the presence of nearby Ti metal. It was not surprising that Ti could reduce Ta2O5 at thermal equilibrium. We did not, however, expect this reduction to occur quickly at temperatures as low as 600?C when we used a barrier layer of 30-nm TiN. If elemental Ti was eliminated and the Ta2O5 film was deposited on TiN without an underlying Ti layer, the oxide would have been thermally stable to 800?C. Recent work has indicated that TaN and WN are superior barrier materials to oxygen diffusion, and Ta and W are better materials for metal electrodes because they do not reduce Ta2O5 at elevated temperatures [12]. Studies showed that a Ta2O5/Si structure was thermally stable up to 800?C without decomposition of the Ta2O5. Silicon out-diffusion and a decrease in net oxygen intensity, however, were observed and a Si3N4 buffer layer could stabilize the interface. n
Acknowledgments
We would like to acknowledge R. Urdahl at Applied Materials for depositing Ta2O5 and other thin films used in this work, and N. Fernandes at Gasonics International for the post-deposition annealing. We are grateful to C.Y. Sung, Y.H. Wong, R.M. Fleming, R.B. van Dover, and D.V. Lang for useful discussions, and D. Werder and J. Sapjeta for the TEM and AFM measurements.
References
1. T. Aoyama et al., J. Electrochem. Soc., Vol. 143, No. 3, p. 977, 1996.
2. F. Chiu, et al., J. Appl. Phys., Vol. 81, No. 10, p. 6911, 1997.
3. R.J. Cava, J.J. Krajewski, J. Appl. Phys., Vol. 83, No. 3, p. 1613, 1998.
4. M. Lee et al., 1996 IEDM, 27.3.1, IEEE, San Francisco, CA, 1996.
5. G.B. Alers et al., Appl. Phys. Lett., Vol. 73, p. 1517, 1998.
6. G.B. Alers et al., Appl. Phys. Lett., Vol. 16, p. 1308, 1998.
7. S. Kamiyama et al., J. Elecrochem. Soc.,Vol. 141, p. 1246, 1994.
8. J.F. van der Veen, Surf. Sci. Rep., Vol. 5, p. 199, 1985.
9. E.P. Gusev, et al., Phys. Rev., Vol. B 52, p. 1759, 1995.
10. R.L. Cava, private communication.
11. R.E. Walkup, S.I. Raider, Appl. Phys. Lett., Vol. 53, p. 888, 1988.
12. J.P. Chang et al., "Oxygen Diffusion in Ta2O5 Metal-Oxide-Metal (MOM) Capacitor Structures," Bell Labs Technical Memorandum.
For further information, contact R.L. Opila, Bell Labs, Room ID-352, 600 Mountain Ave., Murray Hill, NJ 07974; ph 908/582-3390, fax 908/582-3957, [email protected].