ARTICLE pubs.acs.org/JPCC
Atomic Layer Deposition of Tantalum Nitride Using A Novel Precursor Shikha Somani,† Atashi Mukhopadhyay,† and Charles Musgrave*,‡ † ‡
Department of Chemical Engineering, Stanford University, Stanford, California 94305, United States Department of Chemical and Biological Engineering, University of Colorado, Boulder, Colorado 80309, United States ABSTRACT: We use B3LYP hybrid density functional theory to investigate atomistic mechanisms for the atomic layer deposition (ALD) of tantalum nitride (TaN) grown using tert-butylimidotris(diethylamido)tantalum [(t BuN)(NEt2)3Ta, TBTDET], and ammonia (NH3) as precursors. Our calculations examine various possible mechanisms for TaN growth by ALD and metal organic chemical vapor deposition (MOCVD). In particular, we identify low barrier (10.6 and 27.6 kcal/mol) ligand exchange mechanisms with NH3 that lead to incorporation of NH3’s nitrogen into the film. Ligand exchange with NH3 is thermodynamically and kinetically favored over competing mechanisms that incorporate nitrogen from the metal precursor including: β-hydrogen elimination of isobutene or ethene; and NH3 catalyzed β-hydrogen elimination of isobutene or ethene. β-hydrogen elimination of isobutene or ethene is found to proceed through a barrier of 76.0 kcal/mol. However, our results indicate that ammonia or diethylamine produced by precursor reaction with surface amine groups can also catalyze β-hydrogen elimination of isobutene with a predicted barrier of 64.3 kcal/mol, thus making MOCVD reactions kinetically active above ∼600 °C. In addition to providing a fundamental understanding of the chemistry of TaN ALD from (tBuN)(NEt2)3Ta and NH3, the set of mechanisms analyzed provide new insights into the principles governing the ALD processes of other metal nitride films using imido or amido ligand transition metal complexes and ammonia as precursors.
’ INTRODUCTION As the components of microelectronics shrink to increasingly smaller dimensions, new materials and processes to produce these structures are required. An area that has been of particular interest is the fabrication of metal interconnects that link the separate devices in integrated circuits and the materials that are used to make them. Copper has become the material of choice for interconnects due to its good mechanical properties and low resistivity compared to the previously dominant interconnect material, aluminum. However, Cu exhibits a high diffusivity, making it prone to diffusion into neighboring structures. As a result, much research has focused on exploring new materials to be used as Cu diffusion barriers. In particular, transition metal nitrides such as tantalum nitride, tungsten nitride, and titanium nitride, have all been identified as good candidates for copper diffusion barriers in microelectronics.13 Tantalum nitrides are particularly promising because they are stable up to extremely high temperatures and are unreactive toward Cu.4 Scaling of interconnects to current line widths has led to severe requirements for barrier materials. For instance, because barrier materials are poor electrical conductors, their thicknesses must be kept to a minimum in order to maintain copper interconnect’s resistance advantage over Al. Furthermore, in cases where via and trench liner materials are deposited between metal layers, for example at the bottom of a via connected to a lower metal level, the barrier material should be thin to reduce electrical resistance by allowing electron tunneling across the barrier interlayer. While ultrathin thicknesses are desired, barrier films must be deposited r 2011 American Chemical Society
so as to avoid pinholes that could lead to shorting of the interconnect. One method capable of depositing ultrathin, conformal and pinhole-free films is atomic layer deposition (ALD). ALD (sometimes called atomic layer epitaxy) has been shown to be an effective method of depositing TaN and has been shown to have several advantages over competing techniques, such as physical vapor deposition (PVD) and chemical vapor deposition (CVD).510 PVD is a vacuum deposition process that involves creating vapor phase species, usually either by evaporation or sputtering, that deposit on a substrate by condensation, rather than chemical reaction. Although PVD is effective in depositing films at reasonably low temperatures, it is a line of sight process and can have difficulty in achieving satisfactory uniformity, film thickness and film quality for the demanding industry requirements for next-generation microelectronics. CVD involves vaporizing precursors that deposit on a substrate by chemically reacting with it. Although CVD is better able to fulfill the strict constraints on the film than PVD, it does not achieve the same uniformity, conformality, ultrathin thicknesses, and absence of pinholes as ALD because unlike ALD, the surface reactions are not self-limiting. Consequently, ALD has proven a viable alternative when atomic layer thicknesses and thickness control of pinholefree films are desired.
Received: June 28, 2010 Revised: March 23, 2011 Published: May 19, 2011 11507
dx.doi.org/10.1021/jp1059374 | J. Phys. Chem. C 2011, 115, 11507–11513
The Journal of Physical Chemistry C The ALD process requires that precursors not self-react and that reactions between precursors are self-limiting. This results in uniform film deposition over large areas and high aspect ratio features11 with deposited films characterized by excellent conformality and thickness control. In the ALD process, the surface onto which a film is to be deposited is exposed to a dose of vapor of one precursor followed by a reactor purge of the unreacted vapor and byproducts. Next, a vapor dose of a second reactant is introduced and allowed to react with the substrate after which the reactor is again purged. This cycle is repeated until a film of the desired thickness is obtained. Because reactants do not self-react under ALD conditions and the ALD process introduces different reactants into the reactor separately particles are not formed. Furthermore, because the surface reactions are self-limited, at most one monolayer of material is deposited during each cycle, resulting in exquisite control of the film thickness and superb conformality. Both binary metal halides, for example TaCl5 together with NH3,12,13 and metal organic TaN precursors5,6,9,1318 have been used for TaN deposition. In this work, the detailed atomistic mechanisms for the ALD of TaNx using tert-butylimidotris(diethylamido)tantalum [TBTDET, (NC(Me)3)(NEt2)3Ta, Et = ethyl and Me = methyl] and NH3 as coreactants are investigated. TBTDET is a commercially available metal organic precursor for TaN deposition. One of its advantages as a precursor over TaCl5 is its ability to produce a Cl-free diffusion barrier. Furthermore, because it is a nitrogen-coordinated tantalum compound, it includes both components of the binary target material making it a possible single-source MOCVD precursor with the option of depositing TaN with or without additional reactive gases (e.g., NH3, hydrazines).7 For the single-source deposition ability of TBTDET to be realized, its surface reactions must involve competitive pathways that remove both ethyl and t-butyl groups of the precursor, while leaving N on the surface. One such mechanism is β-hydride elimination of ethene and butene. Analogous mechanisms for single source metal oxide ALD and CVD precursors using alkoxides with butoxy and ethoxy ligands have also been examined.19,20 TBTDET was also selected because of its several advantages over other commercially available metalorganic Ta sources. In particular, TBTDET is a stable, but relatively volatile liquid at room temperature, which permits its controllable and reproducible delivery in vapor form into the ALD reactor, providing easier handling than solid precursors.10 The ALD process consists of alternate pulses of TBTDET and NH3 separated by a reactor purge with an inert gas. To achieve the ideal ALD film characteristics, the reactants must not selfreact so that the surface reactions are self-terminating. However, TBTDET acts as a single-source TaN precursor in MOCVD7,21 by self-reacting at the MOCVD process temperature of 600 °C. Hence, for TBTDET to be suitable as an ALD Ta source it must react with NH3 through ALD reactions that are significantly more competitive than TBTDET self-reactions at temperatures well below the TBTDET single source MOCVD process temperature of 600 °C. Becker et al. successfully deposited TaN by ALD at 250 to 350 °C using TBTDET and NH3 as precursors and achieved good ALD film characteristics,5 whereas at higher temperatures, such as the 600 °C used for single-source MOCVD with this precursor, this process would not be expected to be self-limiting. More recently, George and co-workers deposited TaN by ALD using TBTDET with hydrazine6 as well as with H radicals.17
ARTICLE
Detailed studies of the surface reaction chemistry of the ALD of transition metal nitrides have been rare. Although experiments can provide insights from which simple mechanisms can be proposed, the actual detailed mechanisms are often too complicated to be unambiguously developed from experiment alone. This is especially true for understanding the thermodynamics of the short-lived intermediates and the transition states of multistep mechanisms as well as the kinetics of the individual steps of the reaction mechanism. However, experiment can provide observations that can invalidate a mechanism when the proposed mechanism predicts behavior that may be inconsistent with experiment. For example, Elam et al.22 employed infrared spectroscopy and a quartz-crystal microbalance to study the surface reactions occurring during TiN ALD from tetradimethylamidotitanium (TDMAT) and NH3 and proposed that the ALD mechanism involves a transamination exchange reaction, although several other possible reaction pathways may be active that cannot all be ruled out. Experimental work by Becker et al.5 using isotopically labeled ammonia, 15NH3, showed the presence of a small amount of 15N in the films, suggesting that at least one of the TaN single bonds undergoes some type of transamination reaction. Both of these examples illustrate how experiment provides valuable information related to an ALD chemical mechanism, but is unable to definitively prove the detailed mechanism. While it is often difficult to experimentally identify the multiple steps of a detailed mechanism, theoretical calculations of these reactions using quantum chemical methods can provide valuable insight into the surface chemistry of ALD not easily obtainable by other means. In the present work, we used density functional theory (DFT) to investigate possible surface reactions for the ALD of TaN using (tBuN)(NEt2)3Ta and NH3 as precursors. Our objective was to compare the predicted activation barriers of the possible reaction mechanisms to determine which possible reactions are most kinetically competitive. This analysis is expected to elucidate the active pathway by which the reaction takes place. From these results, one may also approximate the kinetic feasibility of these reactions at the experimental conditions to understand the atomistic mechanism for this ALD process at, for example, different temperatures. Pathways include reaction of NH3 with (tBuN)(NEt2)3Ta, which mimics both the reaction of NH3 with products of (tBuN)(NEt2)3Ta reactive adsorption, for example SNTa(tBuN)(NEt2)2*, and reaction of (tBuN)(NEt2)3Ta with surface amine sites (S-NH* and S-NH2*). Here, S represents the substrate and * indicates a surface reactive site. In particular, this work examined the four different mechanisms illustrated in Scheme 1: (a) β-hydride elimination of butene; (b) NH3 catalyzed β-hydride elimination of butene; (c) two sequential ligand exchange reactions starting with ligand exchange between NH3 and SNTa(tBuN)(NEt2)2* to produce SN Ta(NH)(NEt2)2* and t-butylamine; and (d) ligand exchange between NH3 and SNTa(tBuN)(NEt2)2* to produce SN Ta(tBuN)(NEt2)* and diethylamine. Note that in the model system, ethyl groups are replaced by methyl groups so that Scheme 1d shows dimethylamine as the product. The β-hydride elimination self-reaction shown in Scheme 1a is possibly active in single source MOCVD of TaN using (tBuN)(NEt2)3Ta. Although NH3 catalyzed isobutene elimination as shown in Scheme 1b is not possible in the single source MOCVD process because NH3 is not dosed or likely produced, diethylamine produced by reaction of the precursor with surface amine groups might catalyze the analogous reaction. 11508
dx.doi.org/10.1021/jp1059374 |J. Phys. Chem. C 2011, 115, 11507–11513
The Journal of Physical Chemistry C
ARTICLE
Scheme 1. (a) Isobutene elimination by β-hydrogen transfer, (b) NH3 facilitated isobutene elimination, (c) t-butylamine elimination via two sequential ligand exchange reactions with NH3, and (d) dimethalyamine elimination by ligand exchange with NH3. Methyl groups replace ethyl groups in the model and for clarity are represented as Me when not directly involved in the reaction
’ CALCULATIONAL DETAILS All calculations were performed using the B3LYP23 DFT method and a mixed Gaussian basis set where Ta atoms are described using the LANL2 effective core potential and the LANL2DZ basis set,2326 while the 6-311þþG(d,p) all-electron basis set is used to describe all other atoms. A similar basis set scheme was previously used to simulate the analogous ALD process for growing WN.27 Vibrational frequencies were calculated at molecular geometries corresponding to stationary points along the reaction
coordinate to determine zero-point energies, which are included in all reported energies. Furthermore, the calculated frequencies and associated normal modes are used to verify that structures correspond to reactants, transition states (TS), intermediates, and products. Furthermore, intrinsic reaction coordinate28 calculations were used to confirm that calculated TS’s connect the reported structures in cases where the normal-mode analysis is inconclusive. To reduce the computational expense, the precursor is modeled by (tBuN)(NMe2)3Ta where the ethyl groups of 11509
dx.doi.org/10.1021/jp1059374 |J. Phys. Chem. C 2011, 115, 11507–11513
The Journal of Physical Chemistry C
Figure 1. (Color online) Schematic PES for isobutene elimination by β-hydride transfer from the product of the precursor reacting at =NH sites; Scheme 1a: direct isobutene elimination (solid line) and Scheme 1b: isobutene elimination where NH3 acts as a proton transfer mediator (dashed line). (a) Adsorbed precursor after elimination of dimethylamine, (tBuN)(NMe2)(NH2)2Ta*; (b) TS for β-hydrogen transfer from the t-butyl group to the imido N of the ligand; (c) (NH)(NMe2)(NH2)2Ta* þ isobutene; (d) Ta precursor and NH3; (e) NH3•(tBuN)(NEt2)3Ta complex; (f) TS for the simultaneous proton transfer from NH3 to the imido N of the ligand and from the t-butyl group to NH3; (g) (NH)(NMe2)(NH2)2Ta* þ isobutene þ NH3.
(tBuN)(NEt2)3Ta are replaced by methyl groups. An analogous approximation is used to reduce the number of atoms in the model of the reacting Ta surface species where the ethyl groups of SNTa(tBuN)(NEt2)2* are replaced by methyl groups in SNTa(tBuN)(NMe2)2*. Because methyl and ethyl groups are electronically similar this will have little effect on the calculated reaction energetics. Although this approximation does preclude β-hydride elimination of ethylene from the diethylamine ligands, this reaction should possess similar energetics to the analogous β-hydride elimination of isobutene from the t-butylamine ligand, which we discuss below. We have previously used analogous approaches to simulate ALD reactions for a variety of systems.19,2934 Calculations were all performed using the GAUSSIAN 03 quantum chemistry suite.35
’ RESULTS AND DISCUSSION β-Hydrogen Transfer in the Absence of Ammonia. The solid line in Figure 1 shows a schematic potential energy surface (PES) for the β-hydrogen transfer decomposition reaction illustrated in Scheme 1a. Because this reaction eliminates isobutene (isobutylene) and produces an NH surface group it could enable TBTDET to act as a single-source precursor, consistent with the ability of TBTDET to act as a single-source precursor in the MOCVD of TaN. The initial surface Ta species results from the elimination of two dimethylamine ligands from the adsorbing precursor reacting with two SNH* surface groups and is represented by (tBuN)(N(CH3)2)(NH2)2Ta*. The analogous surface species that results from TBTDET reacting at a single S-NH* surface site to produce (tBuN)(N(CH3)2)2(NH2)Ta* will have nearly identical reactivity for β-hydride transfer as this
ARTICLE
reaction will not be significantly affected by replacing an NH2 terminating group with N(CH3)2. β-hydride elimination of isobutene proceeds through a fourcenter TS involving β-hydrogen transfer from the t-butyl group to the imido-nitrogen. The calculated activation barrier is 76.0 kcal/mol and the reaction is endothermic by 16.3 kcal/mol. Because of the high activation barrier, this reaction pathway does not allow the precursor to act as a single-source at temperatures much below ∼600 °C. In fact, the calculated barrier of ∼76 kcal/mol predicts kinetics that are relatively slow even at 600 °C. However, TaN MOCVD using TBTDET as a single-source precursor is carried out at 600 °C.7 Consequently, the ALD process likely involves an alternative, more competitive deposition mechanism. Furthermore, because the reaction is endothermic, enthalpy favors the reactants at equilibrium. However, the entropy associated with the production of the gas phase isobutene byproduct significantly lowers the relative product free energy and because the process is run under nonequilibrium flow conditions in which the byproducts are exhausted during ALD exposures, this step becomes essentially irreversible. Isobutene Elimination Facilitated by Ammonia or Diethylamine Acting As a Proton Transfer Mediator. We have also investigated a mechanism whereby NH3 or diethylamine acts as a proton shuttle between a tertiary butyl group and the imidonitrogen of (tBuN)(N(CH3)2)(NH2)2Ta* as illustrated in Scheme 1b. In the ALD process, NH3 is introduced into the ALD reactor during alternate half-cycles as a coreactant whereas diethylamine results from ligand exchange between the diethylamine ligands of the adsorbing precursor and surface amine groups and is bound to the Ta center via a TaN dipolar bond directly after being produced. Here, we only examine the ammonia catalyzed process as the analogous diethylamine catalyzed process, with diethylamine produced from the ligand exchange reaction with the surface, is expected to behave similarly. NH3 facilitated isobutene elimination proceeds via a TS involving simultaneous transfer of a proton from the t-butyl group to ammonia and transfer of a proton from ammonia to the imido nitrogen. The schematic PES for the proton shuttle facilitated isobutene elimination mechanism is illustrated in Figure 1 as a dashed line. Here, the simultaneous transfer of two protons proceeds through a six-center TS where one transferring proton is shared between the methyl of the tBu group and the NH3 lone-pair and the other transferring proton is shared between the NH3 and the developing lone-pair of the imido N. Examination of the normal mode corresponding to the reaction coordinate at the TS reveals that both protons are transferring concurrently. The calculated activation barrier for this step is 64.3 kcal/mol. Because NH3 is neither consumed nor produced in the process, NH3 indeed acts as a catalyst for isobutene elimination. Although this barrier is still relatively high, it reduces the barrier relative to the uncatalyzed β-hydrogen transfer by 11.6 kcal/mol, which corresponds to an increase in the rate by a factor of ∼2000 at 600 °C, ignoring entropic effects associated with the different TS’s. Consequently, this mechanism suggests a route whereby the metal precursor acts as the N source and ammonia (or diethylamine) acts as a catalyst. Moreover, unless lower barrier pathways exist whereby N is incorporated from NH3, this mechanism is consistent with the experimental observation5 that when TBTDET was used to grow TaN films under ALD conditions at temperatures between 200 and 400 °C with isotopically labeled 15NH3, the resulting films contained mainly 14N and a small amount of 15N, which suggested a catalytic 11510
dx.doi.org/10.1021/jp1059374 |J. Phys. Chem. C 2011, 115, 11507–11513
The Journal of Physical Chemistry C
Figure 2. (Color online) PES of two proton transfers from NH3 to the imido N with the elimination of tBu-amine as illustrated in Scheme 1c. The stationary points correspond to (a) (tBuN)(NMe2)(NH2)2Ta þ NH3; (b) the (tBuN)(NMe2)(NH2)2Ta•NH3 complex; (c) the TS for proton transfer from NH3 to one of the imido nitrogens of the complex; (d) the (tBuNH)(NMe2)(NH2)2Ta•NH2 complex; (e) the TS for proton transfer from NH2 to the amido nitrogen of the complex; (f) physisorbed tBu-amine with (NMe2)(NH2)2(NH)Ta; and (g) tBuamine and (NMe2)(NH2)2(NH)Ta.
effect of NH3 and transamination of at least one of the TaN bonds (vide infra). In the case where ammonia is not introduced as a coreactant, the predicted NH3 catalyzed pathway suggests that the diethylamine byproduct (produced upon TBTDET adsorption by ligand exchange with surface NHx groups and bound to the surface through a dipolar bond) analogously catalyzes isobutene elimination. This resolves the conflict between the experimental observation that TBTDET acts as a single-source precursor in MOCVD at 600 °C, but that the β-hydrogen transfer pathway for this reaction has a predicted barrier that is too high to be sufficiently active at that temperature. However, the barrier for the catalyzed β-hydrogen transfer is consistent with a growth temperature of 600 °C, which is the expected temperature necessary to activate an ∼65 kcal/mol rate limiting step. As this is a catalyzed pathway, the endothermicity of the reaction is equivalent to the uncatalyzed mechanism discussed above. Elimination of t-Butyl Amine by a Two-Step Ammonia Ligand-Exchange Reaction. Both mechanisms discussed above involve the formation of STadN-H* surface species where nitrogen derives from the metal precursor and the barriers are greater than 64 kcal/mol. However, Scheme 1c illustrates an alternative mechanism where ammonia may also act as the N source. The predicted schematic PES for this reaction is shown in Figure 2. Here, NH3 adsorbs at (tBuN)(NMe2)(NH2)2Ta* with an adsorption energy of 0.7 kcal/mol to form the (tBuN) (NMe2)(NH2)2Ta•NH3* complex where the NH3 adduct is bound to Ta through a TaN dipolar bond. From the adduct, one proton of NH3 transfers to the imido-nitrogen, breaking the TadN π-bond and converting the imido N to an amido nitrogen. The TS lies 21.3 kcal/mol above the entrance channel making the reverse step of NH3 desorption faster than proton transfer. The resulting species is (NH2)a(tBuNH)(N(Me)2)(NH2)2Ta
ARTICLE
Figure 3. (Color online) Schematic PES for proton transfer from NH3 to the dimethylamine ligand of the adsorbed precursor for the reaction of NH3 with (NMe2)(tBuN)Ta* as illustrated in Scheme 1d. The stationary points correspond to (a) (tBuN)(NMe2)(NH2)2Ta* þ NH3; (b) the (tBuN)(NMe2)(NH2)2Ta•NH3* complex; (c) the TS for the proton transfer from NH3 to dimethylamine; (d) physisorbed (NH2)(tBuN) 2Ta* þ dimethylamine; and (e) (NH2)(tBuN) 2Ta* þ dimethylamine.
where (NH2)a represents the NH2 derived from NH3 and the remaining two NH2 groups represent the N linkages to the surface. Next, a proton from the (NH2)a group and the amido N of the tBuNH group combine through a four-center TS lying 27.6 kcal/mol above the entrance channel to produce physisorbed t-butylamine. The overall reaction is endothermic by 6.2 kcal/mol and the rate-limiting step involves a TS that lies significantly below either β-hydrogen barrier. Consequently, if NH3 is introduced to catalyze β-hydrogen elimination of isobutene, it will simultaneously also directly remove t-butylamine by two sequential ligand exchange reactions. Because this reaction is much faster than the NH3 catalyzed β-hydrogen transfer pathway, N will be incorporated from NH3 and to a lesser extent, from TBTDET. However, this disagrees with the experimental observation that 15N was not significantly incorporated into ALD grown TaN when deposited using 15NH3 and (tBuN)(NEt2)3Ta. One might expect that a similar ligand exchange reaction with diethylamine produced by reaction of the adsorbing TaN precursor with a surface NH* would also be kinetically favored in the case of the single source TaN precursor. However, closer inspection reveals that although reaction between diethylamine and the adsorbed precursor can convert the imide ligand to an amine, diethylamine does not have a second hydrogen for a second ligand exchange reaction. Furthermore, the subsequent β-hydride elimination reactions to remove isobutene or ethylene will have the high barriers characteristic of the uncatalyzed β-hydride elimination reactions. Consequently, the diethylamine-catalyzed β-hydride elimination is kinetically favored over the diethylamine ligand exchange mechanism for single-source MOCVD at 600 °C. Hence, at high temperatures, diethylamine-catalyzed β-hydrogen transfer becomes significant such that the TBTDET ALD pulse is not expected to be self-terminating and consequently the process will exhibit MOCVD-like growth characteristics. Diethylamine Elimination by Ligand-Exchange with Ammonia. Ligand exchange between TBTDET and ammonia can also exchange a diethylamine ligand with NH2. Because we 11511
dx.doi.org/10.1021/jp1059374 |J. Phys. Chem. C 2011, 115, 11507–11513
The Journal of Physical Chemistry C model the precursor with methyl groups replacing the ethyl groups of the amine ligands, we model this reaction as dimethylamine elimination and expect that the reaction energetics will be nearly identical. Figure 3 illustrates the PES for a second alternative mechanism shown in Scheme 1d in which NH3 acts as the N source. Here, NH3 adsorbs at the surface Ta site with an adsorption energy of 0.7 kcal/mol. Subsequently, one proton from the NH3 adduct transfers to the N of the dimethylamine ligand, transforming the TaN covalent bond into a dipolar bond between the surface Ta and the N of the now physisorbed dimethylamine. The TS for this step lies only 10.6 kcal/mol above the entrance channel. The resulting complex is (NH2)(tBuN)(NH2)2Ta* and the reaction has very low endothermicity (1.3 kcal/mol). This direct ligand-exchange pathway between NH3 and the dimethylamine ligand (diethylamine in the actual precursor) possesses an activation barrier significantly lower than the other mechanisms considered. Consequently, although NH3 may be introduced to catalyze removal of the precursor tBu-amine ligand, and it is in fact predicted to do this, NH3 is much more likely to be involved in ligand exchange reactions with the diethylamine ligands and therefore act as the primary N source. In either case, our results predict that NH3 greatly accelerates the TaN deposition rate when using TBTDET as a Ta source. Because the NH3 ligand-exchange barriers are substantially lower than the uncatalyzed β-hydrogen transfer reactions, the TBTDET/NH3 ALD process exhibits ALD-like growth behavior. Also, the competitive NH3 ligand exchange reactions predict incorporation of N from NH3 and explain the presence of 15N in TaN films using isotopic ammonia (15NH3) in experiments by Becker et al.5 However, our results predict that 15 N incorporation should be significant, rather than only a small fraction as observed experimentally. Several possibilities could explain this discrepancy. The predicted barriers for the catalyzed β-hydrogen elimination reactions could be overestimated. However, this would require that the predicted barrier for β-hydrogen elimination be overestimated relative to the barrier for ligand exchange by at least 50 kcal/mol, which is unlikely. Another possibility is that lower barrier pathways for catalyzed β-hydrogen elimination reactions exist. While many attempts at locating low barrier pathways to β-hydrogen elimination were made, none succeeded in finding more competitive pathways. It is also possible that the barriers to the NH3 ligand exchange reactions are significantly underpredicted. While B3LYP tends to underpredict barriers, it usually does so by ∼35 kcal/mol.36 We note that the ∼27 kcal/mol barrier for the NH3 ligand exchange reaction to remove the t-butylamine group is consistent with the observation that WN deposition by the analogous (tBuN)2(NMe2)2W/NH3 ALD process was not significant below temperatures of about 300 °C.5 Another possibility is that in the experiment the 15NH3 ALD pulse is introduced through vacuum lines that have previously carried 14NH3 and thus the pulse partially displaces NH3 adsorbed on the walls of the ammonia supply lines to dilute the isotopically labeled species. At present, we are unable to unambiguously determine the specific source of the discrepancy between our results and experiment and leave this for future study. The tungsten nitride process using the analogous W precursor bis-(t-butylimido)-bis(dimethylamido)tungsten [(tBuN)2(Me2N)2W] has also recently been investigated.22 The nature of the PES and relative barrier heights are remarkably similar to those reported here, indicating ammonia as the main source of nitrogen
ARTICLE
in both ALD grown films. In the final product, Ta is not reduced from the þ5 to the þ3 oxidation state and hence the deposited TaN phase is closer to Ta3N5, which is insulating in nature. This is in line with the experimental finding that the TBTDET/NH3 ALD process produces extremely high-resistivity films.8
’ CONCLUSIONS The calculated energetics reported herein examine different chemical mechanisms for the ALD of TaN using (tBuN)(NEt2)3Ta and NH3. Ligand exchange reactions between TBTDET and NH3 are predicted to be significantly more kinetically and thermodynamically favorable than both NH3 (or diethylamine) catalyzed and uncatalyzed β-hydrogen elimination of isobutene or ethylene. The rate limiting barriers for ligand exchange reactions to eliminate t-butylamine and diethylamine are 27.6 and 10.6 kcal/mol, respectively. The 27.6 kcal/ mol barrier to remove t-butylamine is similar to the barrier (35.6 kcal/mol) predicted for rate limiting step of ligand exchange for WN ALD using the analogous tungsten precursor and consistent with the experimental observation that WN deposition was significant for temperatures of ∼300 °C and greater.5,7,21 In contrast, the catalyzed and uncatalyzed βhydrogen transfer reactions involve barriers of 64.3 and 76.0 kcal/mol, respectively. Consequently, these results predict that the nitrogen in the TaN ALD films primarily originates from ammonia rather than from the metal precursor. However, although NH3 catalysis of the β-hydrogen transfer elimination of isobutene is much slower than the ligand exchange reactions, the catalytic effect lowers the barrier for isobutene elimination by 11.6 kcal/mol, suggesting that diethylamine produced from surface adsorption of TBTDET will similarly catalyze the elimination of isobutene and ethylene, thus enabling TBTDET to act as a TaN single-source precursor at 600 °C, consistent with experimental observations for TaN MOCVD from TBTDET.7,8,21 This effect also predicts that as the TBTDET/ NH3 ALD growth temperature is increased to temperatures approaching 600 °C, a larger fraction of the incorporated N will originate from the precursor and the growth will become more CVD-like. ’ AUTHOR INFORMATION Corresponding Author
*E-mail:
[email protected].
’ ACKNOWLEDGMENT The authors would like to thank Alice Wong and Dr. Paul Zimmerman for their assistance in performing the quantum chemical simulations and Prof. Roy Gordon of Harvard University and Prof. Steve George and Dr. Roberto Bianco of the University of Colorado for their helpful discussions. ’ REFERENCES (1) Cho, S. L.; Kim, K. B.; Min, S. H.; Shin, H. K.; Kim, S. D. J. Electrochem. Soc. 1999, 146, 3724. (2) Lu, J. P.; Hsu, W. Y.; Luttmer, J. D.; Magel, L. K.; Tsai, H. L. J. Electrochem. Soc. 1998, 145, L21. (3) Park, K. C.; Kim, K. B. J. Electrochem. Soc. 1995, 142, 3109. (4) Rossnagel, S. M.; Nichols, C.; Hamaguchi, S.; Ruzic, D.; Turkot, R. J. Vac. Sci. Technol. B 1996, 14, 1819. 11512
dx.doi.org/10.1021/jp1059374 |J. Phys. Chem. C 2011, 115, 11507–11513
The Journal of Physical Chemistry C
ARTICLE
(5) Becker, J. S.; Suh, S.; Wang, S. L.; Gordon, R. G. Chem. Mater. 2003, 15, 2969. (6) Burton, B. B.; Lavoie, A. R.; George, S. M. J. Electrochem. Soc. 2008, 155, D508. (7) Lemberger, M.; Baunemann, A.; Bauer, A. J. Microelectron. Reliab. 2007, 47, 635. (8) Lemberger, M.; Thiemann, S.; Baunemann, A.; Parala, H.; Fischer, R. A.; Hinz, J.; Bauer, A. J.; Ryssel, H. Surf. Coat. Technol. 2007, 201, 9154. (9) Sandhu, G. S.; Doan, T. T. Advanced Metallization for ULSI Applications; Rana, V., Joshi, R., Ohdomari, I., Eds.; Materials Research Society: Pittsburgh, PA, 1992; pp 323. (10) van der Straten, O.; Zhu, Y.; Dunn, K.; Eisenbraun, E. T.; Kaloyeros, A. E. J. Mater. Res. 2004, 19, 447. (11) Kim, H. J. Vac. Sci. Technol. B 2003, 21, 2231. (12) Sherman, A. J. Electrochem. Soc. 1990, 137, 1892. (13) Yokoyama, N.; Hinode, K.; Homma, Y. J. Electrochem. Soc. 1991, 138, 190. (14) Eizenberg, M.; Littau, K.; Ghanayem, S.; Mak, A.; Maeda, Y.; Chang, M.; Sinha, A. K. Appl. Phys. Lett. 1994, 65, 2416. (15) Kim, H.; Detavernier, C.; van der Straten, O.; Rossnagel, S. M.; Kellock, A. J.; Park, D. G. J. Appl. Phys. 2005, 98. (16) Raaijmakers, I. J.; Yang, J. Appl. Surf. Sci. 1993, 73, 31. (17) Rayner, G. B.; George, S. M. J. Vac. Sci. Technol. A 2009, 27, 716. (18) Sun, S. C.; Tsai, M. H. Thin Solid Films 1994, 253, 440. (19) Mui, C.; Musgrave, C. B. J. Phys. Chem. B 2004, 108, 15150. (20) Ritala, M.; Kukli, K.; Rahtu, A.; Raisanen, P. I.; Leskela, M.; Sajavaara, T.; Keinonen, J. Science 2000, 288, 319. (21) Tsai, M. H.; Sun, S. C.; Chiu, H. T.; Tsai, C. E.; Chuang, S. H. Appl. Phys. Lett. 1995, 67, 1128. (22) Elam, J. W.; Schuisky, M.; Ferguson, J. D.; George, S. M. Thin Solid Films 2003, 436, 145. (23) Becke, A. D. J. Chem. Phys. 1993, 98, 5648. (24) Hay, P. J.; Wadt, W. R. J. Chem. Phys. 1985, 82, 270. (25) Hay, P. J.; Wadt, W. R. J. Chem. Phys. 1985, 82, 299. (26) Wadt, W. R.; Hay, P. J. J. Chem. Phys. 1985, 82, 284. (27) Mukhopadhyay, A. B.; Musgrave, C. B. Appl. Phys. Lett. 2007, 90, 173120. (28) Fukui, K. Acc. Chem. Res. 1981, 14, 363. (29) Han, J. H.; Gao, G. L.; Widjaja, Y.; Garfunkel, E.; Musgrave, C. B. Surf. Sci. 2004, 550, 199. (30) Heyman, A.; Musgrave, C. B. J. Phys. Chem. B 2004, 108, 5718. (31) Kang, J. K.; Musgrave, C. B. J. Appl. Phys. 2002, 91, 3408. (32) Kelly, M. J.; Han, J. H.; Musgrave, C. B.; Parsons, G. N. Chem. Mater. 2005, 17, 5305. (33) Mui, C.; Widjaja, Y.; Kang, J. K.; Musgrave, C. B. Surf. Sci. 2004, 557, 159. (34) Xu, Y.; Musgrave, C. B. Chem. Phys. Lett. 2005, 407, 272. (35) Frisch, M. J.; Trucks, G. W.; Schlegel, H. B.; Scuseria, G. E.; Robb, M. A.; Cheeseman, J. R.; Montgomery, J., J. A. ; Vreven, T.; Kudin, K. N.; Burant, J. C.; Millam, J. M.; Iyengar, S. S.; Tomasi, J.; Barone, V.; Mennucci, B.; Cossi, M.; Scalmani, G.; Rega, N.; Petersson, G. A.; Nakatsuji, H.; Hada, M.; Ehara, M.; Toyota, K.; Fukuda, R.; Hasegawa, J.; Ishida, M.; Nakajima, T.; Honda, Y.; Kitao, O.; Nakai, H.; Klene, M.; Li, X.; Knox, J. E.; Hratchian, H. P.; Cross, J. B.; Bakken, V.; Adamo, C.; Jaramillo, J.; Gomperts, R.; Stratmann, R. E.; Yazyev, O.; Austin, A. J.; Cammi, R.; Pomelli, C.; Ochterski, J. W.; Ayala, P. Y.; Morokuma, K.; Voth, G. A.; Salvador, P.; Dannenberg, J. J.; Zakrzewski, V. G.; Dapprich, S.; Daniels, A. D.; Strain, M. C.; Farkas, O.; Malick, D. K.; Rabuck, A. D.; Raghavachari, K.; Foresman, J. B.; Ortiz, J. V.; Cui, Q.; Baboul, A. G.; Clifford, S.; Cioslowski, J.; Stefanov, B. B.; Liu, G.; Liashenko, A.; Piskorz, P.; Komaromi, I.; Martin, R. L.; Fox, D. J.; Keith, T.; Al-Laham, M. A.; Peng, C. Y.; Nanayakkara, A.; Challacombe, M.; Gill, P. M. W.; Johnson, B.; Chen, W.; Wong, M. W.; Gonzalez, C.; Pople, J. A. Gaussian 03; Revision C.02 ed.; Gaussian, Inc.: Wallingford, CT, 2004. (36) Kang, J. K.; Musgrave, C. B. J. Chem. Phys. 2001, 115, 11040.
11513
dx.doi.org/10.1021/jp1059374 |J. Phys. Chem. C 2011, 115, 11507–11513