Surface Chemistry of Copper(I) Acetamidinates in Connection with Atomic Layer Deposition (ALD) Processes

The thermal chemistry of copper(I)- N,N' -di- sec -butylacetamidinate on Ni(110) single-crystal and cobalt polycrystalline surfaces was characterized under ultrahigh vacuum (UHV) conditions by X-ray photoelectron spectroscopy (XPS) and temperature-programmed desorption (TPD). A complex network of reactions were identified, starting with the dissociative adsorption of the precursor, from its dimeric form in its free state to a monomer once bonded to the nickel surface. The dissociation of a C–N bond in the acetamidinate ligand at ~200 K leads to the formation of adsorbed 2-butene and N - sec -butylacetamidinate. Some of the latter intermediate hydrogenate around 300 K to release N - sec -butylacetamidine into the gas phase, while the remaining adsorbed species dissociate further around 400 K, as the copper atoms become reduced to a


Introduction
Given the high stability of amidines, they have long been considered good choices as bidentate ligands in organometallic compounds. Early uses of metal amidinates have been reported in catalysis, to promote polymerizations and other related reactions. [1][2][3] More recently, metal amidinates have been developed as promising precursors for the deposition of solid thin films. [4][5][6][7][8] They have proven particularly useful in atomic layer deposition (ALD) processes, where the surface chemistry of film growth is split into two self-limiting and complementary half-reactions in order to control the deposition at a monolayer level. [9][10][11] Ideally, the metal-based precursor in metal ALD processes should adsorb on the substrate until saturation of a monolayer, preferably retaining the structure of most if not all of its ligands intact, after which a second reactant is introduced to remove those ligands and to activate the surface for the next ALD cycle. Q. Ma, R. G. Gordon, and F. Zaera, Surface Chemistry of Copper(I) Amidinates in... This is often undesirable, because the new adsorbates may bind strongly to the surface and remain there even after exposures to the second reactant. Such side reactions are the most common source of impurities in the films grown by chemical means, and the reason why stable ligands are sought for these applications. In the case of amidinates ligands, for instance, the hope is that those moieties survive the first half of the cycle so they can be hydrogenated to the corresponding amidine in the second half.
To date, though, not much is known about the chemistry of either metal amidinates or amidines on solid surfaces. An infrared absorption study on the chemistry of lanthanum(III)-tris-N,N'-diiso-propylacetamidinate on a hydrogen-terminated Si(111) surface indicated the incorporation of acetate/carbonate and hydroxyl impurities in the growing films. 12 It was also established in that work that deposition at 573 K, a temperature low enough to prevent the formation of interfacial SiO 2 , leads to decomposition of the adsorbed ligands and the formation of cyanamide or carbodiimide surface species. Analogous chemistry has also been proposed for the gas-phase chemistry of a related copper guanidimate compound. 13 A more recent report on the surface chemistry of copper(I)-N,N'-di-sec-butylacetamidinate on SiO 2 shows a cleaner surface chemistry, with initial adsorption occurring via displacement of one of the ligands at a surface hydroxide site. 14 According to that work, subsequent exposure to molecular hydrogen leads to the hydrogenation of most of the remaining N,N'-di-secbutylacetamidinates to free N,N'-di-sec-butylacetamidine vapor. Reattachment of some of the Q. Ma, R. G. Gordon, and F. Zaera, Surface Chemistry of Copper(I) Amidinates in... -4 -released acetamidine to the SiO 2 surface was suggested as the source of the carbon contamination seen during the initial cycles of growth. Finally, our initial work on the deposition of the same copper precursor on a Ni(110) single-crystal surface identified a temperature window between approximately 350 and 450 K for the deposition of the precursor on the surface: lower temperatures are insufficient for activation of the dissociative adsorption, and higher temperatures lead to continuous decomposition beyond Cu monolayer saturation. 15 That study also suggested more complex surface chemistry than on SiO 2 . Here, we provide further details on the thermal conversions involved.

Experimental Details
The experiments were conducted in an ultrahigh vacuum (UHV) system turbo-pumped to a base pressure of 1×10 −10 Torr and equipped with an UTI mass quadruple for temperature programmed desorption (TPD), a concentric hemispherical analyzer (VG 100AX) and an Al K α /Mg K α dual anode X-ray source for X-ray photoelectron spectroscopy (XPS), and a Kratos rasterable rare-gas ion gun for sample cleaning. 16 The XPS data were taken by using the Al K α X-ray source (hν = 1486.6 e V), with a total resolution of approximately 1.0 eV. No evidence of X-ray damage was observed: similar results were obtained regardless of the time of exposure to the X-rays, and the same molecular desorption behavior was observed in TPD before versus after radiation. The raw data were analyzed by using deconvolution software in order to identify the different components of the overall signals in the spectra of each element. Shirley background subtraction was carried out first when needed, after which Gaussian peaks were fitted. The binding energies Q. Ma, R. G. Gordon, and F. Zaera, Surface Chemistry of Copper(I) Amidinates in...
-5 -and widths for a given element were fixed across the data sets versus annealing temperature in order to minimize the number of adjustable parameters.
A polished Ni(110) single-crystal was used as the substrate for most of the experiments, except for those reported in Figure 8, which were performed with a cobalt polycrystalline foil. Both samples were used in the form of disks approximately 10 mm in diameter and 1 mm in thickness, and were mounted on a vacuum manipulator via 0.5 mm Ta-wires spot-welded to the edge of the crystal and fixed to the ends of copper vacuum feedthroughs. Cooling was accomplished by using a liquid-nitrogen reservoir in direct contact with the copper feedthroughs, and resistive heating was used to reach any desired temperature between 90 and 1200 K. A K-type thermocouple was spot-welded to the edge of the crystal to monitor the temperature of the surface, and a homemade temperature controller was used to provide linear temperature ramps for the TPD experiments and to maintain the crystal to within ±1 K of any specified temperature.
The heating rate for the TPD measurements was set to 10 K/s. Annealing of the adsorbed layers to the temperatures indicated in the XPS data was accomplished by heating the crystal to the desired temperature at a rate of approximately 20-30 K/s and then holding that temperature for a couple of minutes. Both the Ni(110) crystal and the polycrystalline cobalt foil were cleaned before each experiment by repeated cycles of Ar + ion sputtering and annealing at 1100 K until the surface was deemed clean by XPS.
The copper(I)-N,N'-di-sec-butylacetamidinate precursor was synthesized by sequential reactions of N,N'-di-sec-butylacetamidine with CH 3 Li and CuCl; the free N,N'-di-sec-butylacetamidine was prepared by reaction of acetonitrile with N-sec-butylamine. 17 All gases were purchased Torr·s), uncorrected for differences in ion gauge sensitivities. The pressure in the main UHV chamber was measured using a nude ion gauge.

Results
The thermal chemistry of copper(I)-N,N'-di-sec-butylacetamidinate on the Ni(110) single-crystal surface was first explored by using X-ray photoelectron spectroscopy (XPS). The raw data obtained after dosing 50 L of the precursor at 90 K is shown as a function of annealing temperature, all the way to 800 K, in Figures 1 (Cu 2p 3/2 XPS, left, and Cu L 3 VV AES, right) and 2 (N 1s XPS, left, and C 1s XPS, right). Figure 4 summarizes the data in terms of peak intensities versus annealing temperature for the different features deconvoluted from the raw data, and Tables 2 and 3 summarize the parameters used in the fits for the N 1s and C 1s traces, respectively. Additional reference spectra are provided in Figure 3, as discussed below.
In terms of the behavior of the copper signal (Figure 1), only two types of copper atoms were identified here. The first type is clear in the spectra obtained after annealing at temperatures below 300 K, with the Cu 2p 3/2 XPS peak centered at 934.1 eV and the Cu L 3 VV AES signal at 913.8 eV, which amounts to an Auger parameter (BE(Cu 2p 3/2 ) + KE(Cu L 3 VV) -hν) of 361.3 Q. Ma, R. G. Gordon, and F. Zaera, Surface Chemistry of Copper(I) Amidinates in...
-7 -eV, a value typical for Cu(I). 18 That signal is therefore easily identified with the original copper precursor, although we believe that the dimeric form that exists in the gas phase does dissociate into monomeric units upon adsorption (see below). The second state is seen above 500 K, and is characterized by Cu 2p 3/2 XPS and Cu L 3 VV AES signals at 932.5 and 919.0 eV, respectively, and an Auger parameter of 364.9 eV, the value expected for metallic copper. 18  The changes seen in the N 1s and C 1s XPS peaks upon annealing of the surface are somewhat more complex. In the case of the nitrogen signal, the XPS trace obtained at 150 K, presumably due to the molecularly adsorbed precursor, is quite broad, and likely composed of two or more peaks ( Figure 2  This suggests that the ligands adopt a preferential geometry in the first monolayer, with the amido and amidine carbon more exposed relative to the alkyl carbons. To analyze the N 1s and C 1s XPS spectra obtained after annealing at temperatures above 200 K, the combined peaks for the acetamidine moieties were first subtracted from the data by using the fits shown in Figure 3, appropriately scaled. A monotonic decrease in that contribution is seen for both nitrogen and carbon atoms from 200 K to approximately 500 K, at which point only about ~5% of that signal is left ( Figure 4). This trend matches the decrease in Cu(I) signal seen in the copper XPS data, an observation that indicates that the reduction of the metal center occurs upon elimination or decomposition of the acetamidinate ligands. Displacement of the copper Q. Ma, R. G. Gordon, and F. Zaera, Surface Chemistry of Copper(I) Amidinates in...
-10 -atoms by nickel atoms from the surface in the copper complex immediately upon adsorption, and coordination of the ligands to those new atoms, is ruled out by these data.
Two new features develop in the N 1s XPS spectra upon annealing of the adsorbates at 200 K and above ( Table 1). The first is centered at values between 399.0 and 398.6 eV, and is detected at temperatures as low as 200 K (drifting from the high value to the low number with increasing temperature). This binding energy is close in value to that of the N 1s XPS signal for the original copper acetamidinate, and may therefore originate from a fragment of the original acetamidinate adsorbed on the nickel surface or a closely related species. We propose that this signal signifies the onset of formation of N-sec-butylacetamidinate, a species detected in the TPD experiments (see below). A second feature then starts to develop around 397.2 eV after annealing to 300 K that peaks at 400 K. This peak is not likely to correspond to the formation of any metal nitride, the binding energy of which is typically lower, 19 and is most possibly due to a new organic moiety, perhaps a nitrile (-C≡N) surface species. 20 The N 1s XPS peak of acetonitrile adsorbed on Ni(111) 21 and Pt(111), 22 single-crystal surfaces has been reported at 397.7 and 397.2 eV, respectively, within the range of our observation, and a C≡N surface species has been identified in the infrared spectra of a thermally-rearranged acetamidinate ligand on lanthanum oxide. 12 The previous work has associated this new species with a flat adsorption geometry, with the carbonnitrogen bond rehybridized and di-sigma bonded to the metal surface. We therefore suggest, tentatively, that the 397.2 eV N 1s XPS peak observed here may reflect the formation of adsorbed acetonitrile, even though the evidence is not compelling. It is interesting to note that the total intensity of the N 1s XPS signal decreases monotonically throughout the whole annealing temperature range explored here, and that the intensities of both 398.6 and 397.2 eV Q. Ma, R. G. Gordon, and F. Zaera, Surface Chemistry of Copper(I) Amidinates in...
-11 -features reach their maxima at 400 K and decrease afterwards; no N 1s XPS signal at all is detected after annealing to 800 K (within our instrumental sensitivity). In other words, all the nitrogen appears to be removed via the desorption of nitrogen-containing compounds.
The high-temperature behavior of the C 1s XPS spectra is again more complex than what is observed for the N 1s XPS data. Because of limitations in both resolution and signal intensity, only a broad analysis of these spectra can be performed here. Two new peaks are seen to evolve in the spectra starting at temperatures as low as 200 K, the first of which is likely to represent a series of evolving species on the surface, since it shifts from 286.0 eV at 200 K to 285.5-285.4 eV at 300 K or above (  Table 2). The 284.0 eV feature appears already around 200 K, reaches maximum coverage at 600 K, and almost disappears by 800 K (Table 2), and displays a binding energy consistent with dehydrogenated hydrocarbon fragments, perhaps even graphitic carbon Q. Ma, R. G. Gordon, and F. Zaera, Surface Chemistry of Copper(I) Amidinates in...
-12 -(but not nickel carbide, the binding energy of which appears at 283.0 eV). 23 Overall, annealing to 800 K removes all but about 5% of the carbon signal (and all the nitrogen, as mentioned above). It should be stressed that the XPS data, specially the C 1s XPS results, are useful to identify changes on the surface, but could not be used alone to unequivocally identify individual adsorbates.
Further insight into the surface chemistry of the copper(I)-N,N'-di-sec-butylacetamidinate precursor on Ni(110) single-crystal surfaces was obtained from temperature-programmed desorption (TPD) data. Figure  seen around 300-320 K. Although the shape of that peak changes somewhat with coverage, it remains approximately the same after exposures of 2.0 L or above. This temperature is close to that seen for H 2 desorption after dosing H 2 on Ni(110), 16,24 and is therefore likely to be kinetically limited by the recombination of hydrogen surface atoms into H 2 , that is, it may correspond to hydrogen produced from dehydrogenation of surface species at lower temperatures. A small low-temperature shoulder is seen in some instances, probably from weakly-bonded hydrogen coadsorbed with other species. 25, 26 A third feature shifts to higher temperatures with increasing coverages, from 320 to 410 K, reflecting the increasing stability of the initial intermediates as the surface becomes more crowded. The main peak at saturation is detected at 480 K, but that feature also shifts to higher temperatures, starting from the value of about 440 K seen in experiments with low copper(I)-N,N'-di-sec-butylacetamidinate doses.
Finally, the small shoulder at 580 K could originate from dehydrogenation at defect sites on the surface, and most likely indicate full dehydrogenation of the ~5% of carbon-containing species that remain on the surface after heating to high temperatures (see XPS data in Figures 2 and 4).
Butene desorption is only detected after exposures of 2.0 L or above (Figure 7, second panel from the left). This means that any butene (or other adsorbed species) that may form at lower coverages dehydrogenate fully on the surface instead of desorbing into the gas phase. This is a quite common behavior with hydrocarbons on transition metal surfaces. [27][28][29] The butene TPD peak shifts to slightly higher temperatures with coverage, from 450 K for 6.0 L to 485 K at - 16 -signal in addition to what can be assigned to the original copper acetamidinate is seen as early as 200 K, at a position that starts around 286.5 eV but shifts to lower values upon annealing to higher temperatures. A second feature also grows at 284.0 eV with increasing temperature, and dominates the spectra above 400 K. All these trends match those observed on Ni(110).

Discussion
The experiments reported here indicate that the surface chemistry of copper acetamidinate complexes on metal surfaces may be quite complex, involving a number of intermediates and displaying a conversion in several stepwise reactions over a wide range of temperatures. The data in this report may not be sufficient to fully account for all of that chemistry, but can provide some guidelines on its main features. In the paragraphs below a discussion is provided on the thermal chemistry of copper(I)-N,N'-di-sec-butylacetamidinate on Ni(110) single-crystal surfaces. To aid in that discussion, Figure 9 provides a schematic representation of the main reactions proposed.
The first conclusion derived from the data presented in this paper is that the chemisorption of the copper acetamidinate complex on the nickel surface is dissociative. The free complex has been and (2) no methane desorption was seen in any of the TPD experiments.
Finally, further decomposition of the species remaining on the surface at high temperatures is seen around 500 K. Dehydrogenation of the sec-butyl fragment in the sec-butylamido or secbutylcyanamidate species formed at 400 K at its β position may produce the butene detected in TPD, which is accompanied by the production of significant amounts of H 2 . This butyl moiety most likely remains bonded to the nitrogen atom until its conversion to butene, otherwise it Q. Ma, R. G. Gordon, and F. Zaera, Surface Chemistry of Copper(I) Amidinates in...
-20 -would desorb or fully dehydrogenate at much lower temperatures. 27,38 It should be indicated that decomposition on the nickel surface of the sec-butylamido plus acetonitrile species (or the secbutylcyanamidate species in the alternative mechanism) at 500 K would leave some nitrogen atoms behind on the surface, and no such nitrogen was detected in the XPS data. It may be that those nitrogen atoms hydrogenate to form ammonia, a reaction that could not be confirmed by TPD because of interference from other species, or, more likely, recombine and desorb as N 2 . 39,40 If what forms on the surface is a sec-butylcyanamidate species, butene elimination would leave a cyanamidate fragment (HN(Metal)-C≡N) behind, 12 , which, again, could either hydrogenate and desorb as cyanamide (although no evidence for this was obtained in the TPD experiments) or decompose to yield surface nitrogen and then N 2 . It should be said that a very different decomposition pathway has been reported recently for a related copper guanadinate precursor in the gas phase, involving the sequential formation of an oxidized guanidine first and a carbodiimide afterwards; 13 no evidence for either type of products was obtained here with the acetamidinate. Lastly, in contrast to the case of nitrogen, some carbon is still seen on the surface until annealing temperatures of 700 K, with a C 1s XPS binding energy of 284.0 eV, a value typical of CH x fragments or graphitic carbon. 23 Interestingly, most of that carbon is removed from the surface upon annealing at 800 K.
It may be valuable to place the surface chemistry uncovered here in context in terms of the use of metal amidinates as precursors in thin film deposition by chemical means, in particular in atomic layer deposition (ALD) processes. 4 As mentioned in the introduction, the ligands in the metal precursors used in ALD should ideally remain intact upon adsorption at the temperatures used for film growth, and be able to react with the second reducing agent and removed cleanly from Q. Ma, R. G. Gordon, and F. Zaera, Surface Chemistry of Copper(I) Amidinates in...
-21 -the surface in the second half-cycle of the ALD. Metal amidinates have been identified as potential precursors for these processes precisely because of the stability of the amidinate ligands. However, the results from our study indicate that such compounds are in fact not as stable when adsorbed on metal surfaces, where they decompose at temperatures well below those used in ALD (typically between approximately 390 and 530 K). 41 On the other hand, the initial decomposition reported here, which starts at temperatures as low as 200 K, yields surface byproducts that may still be eliminated cleanly upon subsequent reactions with molecular hydrogen, ammonia, water, or hydrogen sulfide (the second reactants used in these ALD processes). Indeed, it was shown above that copper(I)-N,N'-di-sec-butylacetamidinate at 350 K, a temperature above that require for N-secbutylacetamidine desorption, than at 300 K. 15 It should be noted that self-limiting uptake still occurs at 350 K, and that temperatures above ~400 K are still needed for the continuous deposition of copper by using the copper amidinate alone. 15 Less clear is what may happen if the temperature used in ALD is high enough to promote the second set of reactions reported here. As indicated in Figure 4, the copper atoms start to be reduced on the surface at temperatures slightly above 300 K, a change that is quite extensive by 400 K and complete by 500 K. It should be highlighted that such reduction in this case occurs in the absence of any reducing agent, in what would be the first half-cycle in ALD. It is clear from our results that metal acetamidinates have the ability to undergo reduction of the metal center by themselves. In fact, self-reduction of ALD precursors has also been seen with TiC 4 , 19,42,43 Ti In view of this emphasis of the need to use a hydrogenation agent in the second half of the ALD cycle, it is particularly important to identify the thermal chemistry of the ALD precursor and to evaluate the feasibility of being able to remove any possible intermediates that may form on the Q. Ma, R. G. Gordon, and F. Zaera, Surface Chemistry of Copper(I) Amidinates in...
- 23 -surface. In connection with that, our studies with the copper acetamidinate precursor indicate that heating to ~400 K already results in the dissociation of the smaller N-sec-butylacetamidinate moiety formed at 200 K, possibly into nitrile and alkylamido surface species (Figure 9). It could still be possible to displace the nitrile from the surface by H 2 in the second half of the ALD process, a reaction that is perhaps more likely on smother low-Miller-index planes, 37 and also to hydrogenate the alkylamido species to the free alkylamine, but these reactions may be more difficult and may not occur to completion. However, by approximately 485 K, further dissociation of the alkylamido surface species occurs, releasing an olefin (butene in our case) and leaving NH x species on the surface, and the latter adsorbates would be quite difficult to remove.
Furthermore, additional dehydrogenation occurs above ~500 K, and at that stage it would be certain that some carbon would be left behind on the growing metal film. We have already reported that by 480 K the deposition of copper films using the copper amidinate precursor along is no longer self-limiting, and continues indefinitely with increasing exposures. 15 That is a definitive threshold temperature to avoid the deposition of carbon impurities.
Based on the surface chemistry described above, it could be concluded that the ALD of copper films using copper amidinates should be conducted at as low a temperature as possible.
Certainly, a monolayer of copper can be deposit this way at temperatures as low as 350 K. 15 However, other considerations, including the crystallinity of the films and their electrical properties, may demand the use of higher temperatures. 41  conversion of the copper acetamidinate on the surface, which means that deposition would not be strictly via an ALD process above that temperature. 15 In any case, the use of temperatures above 500 K would almost surely lead to the deposition of some carbon in the growing films.

Conclusions
The thermal chemistry of N,N'-di-sec-butylacetamidinate on Ni (110)  -25 -the remaining adsorbates around 500 K. All but the latter stage produce surface species that can potentially be removed by hydrogenation with H 2 . On the other hand, the occurrence of the final decomposition of the surface species at 500 K sets an upper limit for the use of this compound as a precursor in the deposition of clean copper films by ALD processes. Tables   Table 1. Fitting parameters for the N 1s XPS peaks shown in the left panel of Figure 2. Up to four Gaussian peaks were fitted to each trace after background subtraction. The binding energies (BE) and peak widths (w) were fixed for the complete set, and only the amplitudes (A) were fitted to each individual spectrum. The optimized parameters are indicated by bold characters.  Table 2. Fitting parameters for the C 1s XPS peaks shown in the right panel of Figure 2. Up to five Gaussian peaks were fitted to each trace after background subtraction. The binding energies (BE) and peak widths (w) were fixed for the complete set, and only the amplitudes (A) were fitted to each individual spectrum. In addition, the integrated areas of the first three peaks, assigned to the original copper amidinate, were fixed to follow ratios of 7