Effect of the Nature of the Substrate on the Surface Chemistry of Atomic

Transcription

1 Effect of the Nature of the Substrate on the Surface Chemistry of Atomic Layer Deposition Precursors Yunxi Yao, 1 Jason P. Coyle, 2 Seán T. Barry, 2 and Francisco Zaera 1,* 1 Department of Chemistry, University of California, Riverside, California 92521, United States 2 Department of Chemistry, Carleton University, Ottawa, Ontario, K1S5B6, Canada * Corresponding author. zaera@ucr.edu Abstract The thermal chemistry of Cu(I)-sec-butyl-2-iminopyrrolidinate, a promising copper amidinate complex for atomic layer deposition (ALD) applications, was explored comparatively on several surfaces by using a combination of surface-sensitive techniques, specifically temperatureprogrammed desorption (TPD) and x-ray photoelectron spectroscopy (XPS). The substrates explored include single crystals of transition metals (Ni(110) and Cu(110)), thin oxide films (NiO/Ni(110) and SiO 2 /Ta), and oxygen-treated metals (O/Cu(110)). Decomposition of the pyrrolidinate ligand leads to the desorption of several gas-phase products, including CH 3 CN, HCN and butene from the metals and CO and CO 2 from the oxygen-containing surfaces. In all cases dehydrogenation of the organic moieties is accompanied by hydrogen removal from the surface, in the form of H 2 on metals and mainly as water from the metal oxides, but the threshold

2 for this chemistry varies wildly, from 270 K on Ni(110) to 430 K on O/Cu(110), 470 K on Cu(110), 500 K on NiO/Ni(110), and 570 K on SiO 2 /Ta. Copper reduction is also observed in both the Cu 2p 3/2 XPS and the Cu L 3 VV Auger (AES) spectra, reaching completion by 300 K on Ni(110) but occurring only between 500 and 600 K on Cu(110). On NiO/Ni(110) both Cu(I) and Cu(0) coexist between 200 and 500 K, and on SiO 2 /Ta a change happens between 500 and 600 K but the reduction is limited, with the copper atoms retaining a significant ionic character. Additional experiments to test adsorption at higher temperatures led to the identification of temperature windows for the self-limiting precursor uptake required for ALD between approximately 300 and 450 K on both Ni(110) and NiO/Ni(110); the range on SiO 2 had been previously determined to be wider, reaching an upper limit at about 500 K. Finally, deposition of copper metal films via ALD cycles with O 2 as the co-reactant was successfully accomplished on the Ni(110) substrate. 2

3 1. Introduction Atomic layer deposition (ALD) has become a necessary approach for the growth of thin solid films. 1 It s a chemical method that relies on pairing surface-mediated reactions in order to make them complementary and self limiting, thus offering better control on the thickness of the films down to the Ångstrom level. 2-4 One added benefit of this approach is that mild conditions, in particular low temperatures, may be used. There is typically a temperature window where ALD is operative, limited on the low-temperature side by the reactivity of the precursor used and on the high-temperature end by the triggering of decomposition leading to continuous deposition. The desire to perform ALD at low temperatures often drives the choice of precursors. 5-7 The chosen chemicals need to be volatile but also thermally stable so that they can be delivered easily to the solid substrate and then made to react easily and cleanly upon adsorption on the surface. 3,6,8 Importantly, the surface chemistry driving these processes depends not only on the precursors themselves, but also on the nature of the substrate, 4 a fact that is being exploited recently to carry out film depositions selectively for the nanopatterning of surfaces. 9 This matter is complicated further by the fact that the surface evolves during the film growth process, starting with the initial substrate but transitioning to the material being deposited. 10 The reactivity of the ALD precursor may be significantly different on the surfaces of those two materials, a fact that perhaps provides a reason why some ALD processes display an induction process before the beginning of the deposition 11 and why in some instances with metals on oxides in particular three-dimensional rather than layer-by-layer growth is observed. 12 3

4 In this report we discuss the effect of the substrate on the thermal chemistry of ALD precursors. The example addressed here is that of Cu(I)-sec-butyl-2-iminopyrrolidinate (Scheme 1, shown in its gas-phase tetramer form), 13,14 a type of precursor being considered for the deposition of copper interconnects in the microelectronics industry. 15 The metal may be deposited on silicon or silicon oxide directly, but more often it is grown on top of a thin diffusion barrier made out of metals or metal nitrides. 16 We have in the past highlighted the differences that the change in the nature of the substrate can bring to the corresponding surface chemistry by contrasting the adsorption of similar copper precursors on nickel versus copper surfaces. 17,18 Furthermore, for the copper iminopyrrolidinate reported here, we have already published surface-science studies of its thermal chemistry on Ni(110) 19 and on silicon oxide thin films. 20 Here, we expand that data set by adding results from work on Cu(110) surfaces, clean and oxygen-predosed, and on a thin NiO film grown on top of Ni(110). Significant differences in reactivity were observed, manifesting themselves in large variations in the threshold for surface conversion. For instance, the elimination of hydrogen by dehydrogenation of the iminopyrrolidinate ligands on the surface starts at temperatures as low as 270 K on Ni(110), but only at 430 K on O/Cu(110), 470 K on Cu(110), 500 K on NiO/Ni(110), and 570 K on SiO 2 /Ta. The byproducts produced also differ on the different surfaces, and the threshold temperature at which the copper ions become reduced to their metallic state follow a similar trend. Below we provide more details on these changes. 2. Experimental The experiments reported here were all carried out in situ inside an ultrahigh vacuum (UHV) 4

5 chamber equipped with several surface-sensitive techniques. The details of the chamber, spectroscopic techniques, and operational protocols have all been reported in the past. 21 Briefly, the temperature programmed desorption (TPD) data were acquired by using an Extrel C-50 quadrupole mass spectrometer with an electron-impact ionizer enclosed in a stainless-steel box with a 7-mm-diameter aperture for the selectively sampling of the gases that desorb from the front surface of the solid samples. The spectrometer is interfaced to a personal computer programmed to collect data for up to 15 different masses in a single TPD run. The x-ray photoelectron spectroscopy (XPS) data were acquired by using a 50-mm radius hemispherical electron energy analyzer (VSW HAC 5000), set at a 50 ev constant pass energy, and an aluminum-anode (h = ev) x-ray source. The binding energy scale was calibrated against the Ni 2p and Cu 2p peaks from the metallic samples. The samples, Ni(110) and Cu(110) polished single crystals approximately 10 mm in diameter and 1 mm in thickness, were mounted on an on-axis manipulator capable of x-y-z- motion and of resistive heating and liquid-nitrogen cooling. They were either spotwelded to (Ni) 22 or wedged in between (Cu) 23 thin tantalum wires attached to copper rods, which were in turn connected to the main body of the manipulator. The temperature of the crystals was monitored continuously with the aid of a K-type thermocouple spot-welded/wedged to the side of the crystals, and controlled by a homemade proportional integral derivative (PID) circuit; a constant heating rate of 5 K/s was used for the TPD experiments. The surfaces of the crystals were cleaned before each TPD and XPS experiment by sequential cycles of Ar + ion sputtering (using a Perkin Elmer PHI ion gun and an ion energy of 2 kv) and annealing at 1100 K. Both the ~3 monolayers (ML) NiO film 24 on Ni(110) and the atomic oxygen overlayer on 5

6 Cu(110), reported to form an ordered (2 x 1) superstructure, 25, were produced by treatment of the respective clean crystals with O 2 (1 x 10 6 Torr x 10 min) at 500 K. The copper(i)-n-sec-butyl-2-iminopyrrolidinate precursor was synthesized by reaction of the corresponding N-sec-butyl-2-iminopyrrolidine (made by a published method) 26 with butyl lithium and copper chloride. The compound was characterized by NMR and liquid-injection field desorption ionization mass spectrometry (LIFDI-MS), as described in detail in previous publications. 14,19,20 The latter technique identified tetramers as the most stable units of this compound in the gas phase (with perhaps one third being in the form of dimers), 14 a result that was supported by DFT quantum mechanics calculations 14 and by the fact that the analogous copper(i)-n-iso-propyl-2-iminopyrrolidinate precursor was shown by X-ray crystallography to be in tetrameric form in the solid state. 13 Dosing was carried out by backfilling of the chamber using a leak valve, and controlled by warming up the precursor and gas lines (but not the UHV chamber where the dosing was done) to 353 K in order to increase its vapor pressure. The final exposures were determined by the dosing time, which was in the range of tens of minutes; they amounted to approximately 10 L (1 L = 1 x 10 6 Torr s) in most cases, sufficient for the deposition of a few monolayers of the precursor on the surface. 3. Results and Discussion The general aspects of the thermal chemistry of the adsorbed Cu(I)-sec-butyl-iminopyrrolidinate precursor were first surveyed by TPD. Figure 1 displays the results for the most representative 6

7 masses from experiments carried out on four surfaces, namely, on Ni(110) and Cu(110) both clean and after treatment with oxygen. Several observations derive from these data. First, the nature of the products that desorb and the temperatures at which this occurs are somewhat different in each case. Three products are seen in all cases: molecular Cu(I)-sec-butyliminopyrrolidinate, molecular hydrogen, and molecular nitrogen. Molecular desorption is seen at low temperatures, around K, as evidenced by the sharp peaks detected for most of the traces plotted in Figure 1 including those for 84 amu (which corresponds to the iminopyrrolidine fragment missing the sec-butyl group). H 2 (2 amu) evolves into the gas phase in all cases as well, even if the TPD profiles are quite different; this will be addressed in more detail below in connection with Figure 2. N 2 desorption, from recombination of the atomic nitrogen produced by ligand decomposition, is indicated by the peaks in the 14 and 28 amu traces seen at 800 K or higher temperatures; the yields vary because of the different degrees of ligand decompositions that take place on each surface. On Ni(110) (Figure 1, left panel), peaks are also seen in the 40 and 41 amu traces at 490 K (not shown here but reported before), 19 identified with acetonitrile (CH 3 CN) formation; assignment of these signals to propene was ruled out by the absence of the corresponding signal in the 42 amu trace. Similarly, the feature at 570 K in the traces for 26 and 27 amu was determined to originate mainly from HCN desorption, although a limited amount of ethylene may be produced, since a small peak is also seen at that temperature in the 28 amu trace (another earlier ethylene feature is seen in the 27 and 28 amu trace at ~420 K). The TPD data for the case of metallic Cu(110) are much cleaner, indicating less decomposition, but a small amount of HCN/C 2 H 4 is also seen around 450 K. Some butene (C 4 H 8, 56 amu) may also have been detected at 300 K, although 7

8 that signal may overlap with desorption of the hydrogenated/protonated ligand, as has been seen with other analogous copper amidinates before. 17,20 Significant differences are seen on the "oxide" surfaces, the chief of which is the fact that the dehydrogenation steps lead mainly to the production of water (18 amu) instead of hydrogen (2 amu). The presence of oxygen atoms on the surface also facilitates the oxidation of any residual carbon left by the decomposition of the ligands, producing CO (28 amu, corroborated by the small matching features in the 12 and 16 amu traces) and CO 2 (44 amu, together with 12 and 16 amu). On NiO/Ni(110) this happens around 650 K (and also, to a lesser extent, at 320 K), on O/Cu(110) at about 500 K. In exchange, no organic molecules such as alkanes, alkenes, nitriles, or cyanides desorb from these surfaces. The TPD traces for the hydrogen-containing products, namely, for molecular hydrogen with Ni(110) and Cu(110) and for water for NiO/Ni(110), O/Cu(110), and SiO 2 /Ta (this latter one from our previous study), 14 are plotted again, together, in Figure 2 to facilitate their comparison. The onset of the H 2 /H 2 O desorption TPD peaks provides an indication of the threshold temperature above which extensive ligand decomposition occurs, and therefore speaks to the degree of reactivity of the different surfaces. In that context, it is clear that the most active surface is that of the nickel metal, on which hydrogen desorption starts at approximately 270 K. Interestingly, the production of H 2 there takes place in a stepwise fashion, since its evolution is seen in stages covering a large range of temperatures, all the way to ~ 650 K. It is likely that a number of small organic fragments form on the surface at different stages, following a complex decomposition mechanism analogous to what has been seen with Cu(I)-sec-butyl-acetamidinate 22 8

9 and Cu(I)-N,N-dimethyl-N',N"-diiso-propyl-guanidinate. 19 The kinetics of the dehydrogenation of the iminopyrrolidinate ligand is slower, and also simpler, on the other surfaces. On Cu(110), the onset for H 2 desorption is seen at about 470 K. Curiously, the oxygen-pretreated surface (O/Cu(110)) shows earlier activity, with water desorption starting at about 430 K and displaying a single sharp peak centered at 480 K. It has been established in previous surface-science experiments that the ability of copper surfaces to promote dehydrogenation steps can be aided by surface oxygen, via the formation of OH surface groups (these later disproportionate to make H 2 O + O(ads)). 27 The same promoting effect is not seen on Ni(110), however, since on that surface the production of water starts above 500 K; this behavior is more akin to that of a regular oxide, not an oxygen-modified metal substrate. 28 The silicon oxide surface proved to be the most unreactive of them all, with a hydrogen desorption onset temperature of approximately 550 K (and no water production at all). 20 Additional information on the thermal chemistry of Cu(I)-sec-butyl-2-iminopyrrolidinate was obtained by XPS. Such studies on Ni(110) and SiO 2 /Ta 20 have been reported before, and are complemented here with similar data for Cu(110) 19 and NiO/Ni(110). The C 1s and N 1s XPS traces as a function of annealing temperature after saturation at cryogenic temperatures are displayed in Figures 3 and 4 for those two latter substrates, respectively. In both cases, condensation of the molecular precursor below 200 K is indicated by C 1s and N 1s XPS peaks at binding energies (BEs) of approximately and ev, respectively. Those features are in fact composed of several components, 22 but the signal-to-noise of the data are not sufficient to extract such information. Some molecular desorption occurs above 200 K, as indicated by the 9

10 TPD data discussed above, and that leads to decreases in all XPS signals, after which covalent binding to the surface is evidenced by red shifts in both peaks, to and ev on Cu(110) and to and ev on NiO/Ni(110). The subsequent evolution of the C 1s and N 1s XPS features diverge for the two surfaces after that. On Cu(110), not much else happens until 400 K, but between 500 and 600 K most of the signal is lost, and the XPS features shift to and ev, respectively (Figure 3). This coincides with the desorption of hydrogen seen in the TPD, and most likely corresponds to the formation of highly dehydrogenated surface species. By 900 K the majority of the surface species are gone from the surface, but some nitrogen (~15%) and significant carbon (~25%) remain. On NiO/Ni(110) more transitions are observed (Figure 4), not all of them accompanied by the desorption of byproducts in TPD. Specifically, a shift is seen in the C 1s signal between 250 and 300 K, back to approximately ev (less obvious in the N 1s XPS), a change possibly associated with the elimination of some of the ligands to the gas phase in the form of the protonated iminopyrrolidine, as seen on other surfaces. 19,20 Another transition was detected between 350 and 400 K, as some CO 2 desorbs from the surface (Figure 1), after which the C 1s and N 1s features are centered at and ev, respectively. The next and final transition happens between 500 and 600 K, and results in desorption of most of the adsorbates, in the form of CO, CO 2, and N 2 according to the TPD data (Figure 1). Less residual carbon and nitrogen are left behind in this case, approximately 10-15% of each. The corresponding spectroscopic data for copper is reported in Figure 5. The left panel shows the evolution of the Cu 2p 3/2 XPS peak as a function of annealing temperature on Cu(110). 10

11 These traces were obtained by subtraction of the trace for the clean surface, since the substrate contributes to the total signal in this case, which is why negative peaks are reported in some cases (signifying lower signal intensity than from the untreated substrate; this is also the reason why AES signals are not reported in this case, as is done for NiO/Ni(111), see below). A peak is seen at a BE of ev after condensation at 170 K, but that feature shifts to approximately ev after multilayer condensation, and remains about the same until the metal is reduced between 500 and 600 K. For the thermal conversion of the Cu(I)-sec-butyl-2-iminopyrrolidinate on NiO/Ni(110), data are provided for both the Cu 2p 3/2 XPS and the Cu L 3 VV Auger (AES) electrons, to better estimate the oxidation state of the copper centers (Figure 5, center and right panels, respectively). The first transition shifts slowly over a wide range of temperatures, as will be discussed next, but the second shows to distinct features, at AES kinetic energies (KEs) of and ev. Using the corresponding extreme Cu 2p 3/2 XPS peak positions of and ev, values for the Auger parameter (the sum of the XPS and AES values, BE(XPS 2p 3/2 ) + KE(AES L 3 VV)) of and ev are calculated, typical of Cu(I) and Cu(0) species, respectively. 29 One other thing to note from the copper XPS and AES data for the ALD precursor on NiO/Ni(110) is that, after molecular desorption at ~250 K, the signal intensities do not change much. This means that all of the copper initially deposited by the metallorganic compound remains adsorbed, and that only the elements from the ligand (C, N, H) are removed upon heating of the surface. A summary of the Cu 2p 3/2 XPS binding energies recorded in experiments such as those in Figure 5 are summarized in Figure 6. Average binding energies are shown there as a function of temperature for adsorption on Ni(110), NiO/Ni(110), Cu(110), and SiO 2 /Ta. Red shifts in BE 11

12 with increasing annealing temperature are seen in all cases, but the details are unique in each case. On Ni(110), the transition to values around ev, associated with metallic copper (the BE of Cu 2p 3/2 for clean Cu(110) is ev), occurs right after molecular desorption, by 300 K. 19 On Cu(110), the Cu 2p 3/2 XPS BE value after molecular desorption is lower than that observed for the condensed precursor (934.0 vs ev) but still within the range of oxidized copper, suggesting a significant interaction of the intact precursor with the substrate, and only transitions to the reduced Cu(0) final state between 500 and 600 K. The case of NiO/Ni(110) is somewhat intermediate: the BE drifts progressively with increasing temperature, until reaching the Cu(0) value at 600 K. As indicated below, based on the Cu AES data in Figure 5, it can be safely concluded that this is not because of a gradual reduction of the copper ions, but rather due to two, Cu(I) and Cu(0), states coexisting on the surface over a wide range of temperatures, starting at 200 K. It seems that some of the copper precursor is reduced immediately upon adsorption, even at low temperatures, but that the reactivity is inhibited by crowding of the surface as the coverage increases. Finally, on SiO 2 /Ta, partial reduction also occurs at 600 K, but in this case the copper atoms remain mostly oxidized, possible because they may form ionic (or highly polarized) bonds with the oxygen atoms on the surface. It is interesting to contrast the information regarding the reduction of the copper ions extracted from the XPS data in Figures 5 and 6 with the threshold temperatures for the thermal decomposition of the ligands identified by the TPD experiments in Figure 2. It can be seen that the two correlate quite well: the temperatures at which the elimination of hydrogen from the surface starts (in the form of H 2 or H 2 O) is approximately the same as that when the Cu(I) ions are reduced to the Cu(0) metallic state. The suggestion that derives from this coincidence is that 12

13 the decomposition of the ligands may not take place while coordinated to the copper centers, but rather requires their prior transfer to the surface; it seems that it is the substrate the one that promotes the subsequent dehydrogenation steps. It is interesting to speculate on how this chemistry evolves as the thickness of the deposited films increases. It could be thought that decomposition would be suppressed at that point, but this cannot be entirely true, because hydrogen evolution is seen even in the case of the deposition on a copper (Cu(110)) substrate (the film being grown is made out of copper as well). We think that the secret to understanding this activation may be related to the nature of the initial adsorbates. Our recent studies on this subject have pointed out that this type of precursors exist in dimeric or tetrameric form in the gas phase, and that adsorption may favor the retention of a dimer structure, at least on SiO Furthermore, the DFT calculations suggest that the initial bonding to the surface may take place through one of the nitrogen atoms, not directly via the copper ion. It may be that the limiting step in the activation of the molecularly adsorbed precursor involves access of the copper ion to the surface, at which point the ion may be reduced to its metallic state as the ligand becomes more reactive and dehydrogenates on the surface. These ideas will need further experimental and theoretical confirmation. Next, experiments were carried out to probe the chemisorption of the Cu(I)-sec-butyl-2- iminopyrrolidinate precursor at higher temperatures, closer to those used in ALD, as we have done in the past with other precursors. 23,30 On SiO 2 films, self-limiting adsorption was previously determined to occur for substrate temperatures of up to 500 K; by 550 K, continuous copper deposition takes place instead. 20 The tests applied here to the characterization of the uptake of the copper precursor on Ni(110) and NiO/Ni(110) surfaces are illustrated in Figure 7. 13

14 The top panels display Cu 2p 3/2 XPS traces recorded after a fixed exposure time (60 min) while holding the substrate at the indicated temperatures, ranging from 300 to 500 K. It is seen there that, on the metal, the peaks are all centered around a BEs of 932 ev, which corresponds to metallic copper, although a slight red shift is observed at higher temperatures. It would seem that the copper atoms deposited with Cu(I)-sec-butyl-2-iminopyrrolidinate at room temperature or above are reduced upon adsorption, as already suggested by the data in Figure 6. On NiO/Ni(110), by contrast, the transition is more gradual, with the Cu 2p 3/2 XPS peak shifting from a BE of ev at 350 K to ev at 500 K. This, again, is consistent with the annealing experiments, and reflects the coexistence of copper in ionic and metallic forms on the surface over a wide range of temperatures. The intensity of the XPS peaks can also be used to follow the uptake of copper. A summary of such uptake measurements as a function of dosing time is provided in the bottom panels of Figure 7. In both cases, no significant uptake is seen at 300 but it is already observed by 350 K. The leveling off of the Cu 2p 3/2 peak intensities after approximately 20 minutes of dosing at either 350 or 400 K attests to the self limiting character of the adsorption, as required for ALD. By 450 K, however, some continuous, albeit slower, copper addition is seen at longer times, and by 500 K the rate of Cu deposition seems to be constant and continuous well past the monolayer saturation point. This sets a temperature window for the copper ALD on either nickel or nickel oxide using Cu(I)-sec-butyl-iminopyrrolidinate at between approximately 300 and 450 K, narrower than on silicon oxide. Data from complementary H 2 TPD experiments are reported in Figure 8. These correspond to 14

15 the desorption of molecular hydrogen, from surface ligand decomposition, after deposition at different temperatures between 300 and 500 K on Ni(110) (left panel), Cu(110) (center), or NiO/Ni(110) (right). On nickel, hydrogen from the precursor dosed at low temperatures was seen to desorb in several stages, over a wide range of temperatures (Figures 1 and 2), and dosing at higher temperatures by and large basically cuts on the low-temperature end of the H 2 TPD profile (because, presumably, some hydrogen evolution takes place during adsorption). The one exception is the trace obtained after adsorption at 500 K, which shows a higher yield than the others (above the 500 K threshold) and an extended desorption all the way to 700 K. It would seem that the activated chemisorption of Cu(I)-sec-butyl-2-iminopyrrolidinate on Ni(110) at such high temperatures leads to extensive decomposition, a fact that explains the continuous CVDstyle copper film growth reported in Figure 7. On Cu(110), hydrogen desorption after lowtemperature desorption starts at 470 K (Figures 1 and 2), and no changes are seen after adsorption at either 300 or 400 K. Deposition at 500 K, by contrast, leads again to enhanced high-temperature H 2 production, which should again reflect CVD-type deposition. Finally, on NiO/Ni(110), only a small amount oh H 2 is ever seen in the TPD data, and that peaks above 600 K regardless of the adsorption temperature. Testing of the surface coverages obtained by dosing the copper iminopyrrolidinate ALD precursor on solid surfaces can also be probed indirectly by chemical titration of the remaining open adsorption sites afterward, as shown in Figure 9. The TPD from a saturated layer of carbon monoxide adsorbed on the clean Cu(110) surface at cryogenic temperatures shows one main feature at approximately 200 K, together with a low-temperature tail starting at around 120 K. 31 Dosing of the Cu(I)-sec-butyl-iminopyrrolidinate ALD precursor on that surface at 400 K blocks 15

16 many, but not all, of the CO adsorption sites: subsequent CO titration/tpd experiments still show a small peak corresponding to approximately 25% of the yield measured on the clean substrate. It could be speculated that the iminopyrrolidinate moieties (or the organic ligands/adsorbates that form upon thermal activation of those on the surface) are bulky and do not pack well, leaving open sites not large enough for further uptake of the copper complex but sufficient for CO adsorption. Annealing of the copper-dosed surface at temperatures as high as 700 K do not change this picture, but by 800 K the surface sites are completely blocked. At this stage the newly deposited copper atoms may be incorporated into the solid, but carbon and nitrogen impurities may remain and prevent further chemical activity. Fortunately, this undesirable transition occurs at high temperatures, much higher to those relevant to ALD processes. Finally, a few ALD cycles were attempted to test the possibility of building up a copper film on Ni(110) using the Cu(I)-sec-butyl-2-iminopyrrolidinate precursor. Molecular oxygen was chosen here as the co-reactant instead of the more commonly-used reducing agents (like H 2 ) because the former is more effective for removing organic matter from surfaces. The idea of incorporating oxidizing cycles in ALD to deposit metallic films has in fact been explored successfully with other metals, 7 although not with copper; the only reports we are aware of are for the deposition of copper oxide films using water or "wet" oxygen (water + O 2 combinations) followed by reduction using formic acid or other agents. 32 Figure 10 reports Cu 2p 3/2 (left), C 1s (center), and N 1s (right) XPS data obtained after representative ALD half cycles carried out by Cu-precursor adsorption at 400 K for 20 min followed by annealing in 1 x 10 6 Torr of O 2 at 650 K for 30 min. The high temperature used in the second half-cycle was chosen to ensure the total 16

17 removal of the organic ligands, a fact that was corroborated by the flattening of the C 1s and N 1s XPS spectra seen in the center and right panels of Figure 10 after each O 2 exposure, following the growth of those peaks during the previous deposition of the iminopyrrolidinate precursor. It is also seen, in the left panel of Figure 10, that the Cu 2p 3/2 XPS signal increases (in an approximately linear way) with the number of ALD cycles performed, attesting to the success of the deposition. The position of these XPS peaks (and the accompanying L 3 VV AES features, not shown) is consistent with the growth of metallic films. 29 This is somewhat surprising, since the electrochemical potential of Cu ions is not as positive as those of the elements for which this oxygen-based ALD process have been reported in the past; it is not obvious why the oxidized copper expected from the O 2 treatment can be reduced afterwards. One possible explanation for the observed behavior is that the metallic copper may be stabilized by the nickel substrate that lies underneath, and that it may be the nickel that becomes oxidized instead. However, although a thin NiO film forms on top of the Ni(110) surface in our ALD experiments after the first exposure to O 2, 24 no additional changes in the Ni 2p 3/2 XPS signal were observed during the subsequent exposures to the Cu precursor or to the O 2 gas in the reported ALD cycles (data not shown). We cannot tell at this point if the successful deposition of metallic Cu reported here using O 2 as a co-reactant can be sustained upon the growth of thicker copper films, or if it can be reproduced on other substrates. It is also possible for the copper to nucleate in clusters in the initial stages of the deposition, as perhaps suggested by the slight shifts in binding energies in the Cu 2p 3/2 peaks with increasing number of ALD cycles, in which case the growth mechanism may change as the film becomes thicker and the nickel substrate is fully covered. Finally, it remains to be seen if the effectiveness of the oxidation of the ligands can be sustain once thicker Cu films 17

18 are grown, since copper is less effective than nickel in promoting the oxidation of carbonaceous deposits. More studies are needed to address all these issues. 4. Conclusions The effect of the nature of the substrate on the thermal chemistry of ALD precursors was probed for the specific case of a Cu(I) iminopyrrolidinate metallorganic compound. Significant differences were observed on nickel, nickel oxide, copper, oxygen-covered copper, and silicon oxide surfaces. Perhaps the most relevant results to support this conclusion are provided in Figures 2 and 6. The first shows the temperature programmed desorption profiles for hydrogen elimination from decomposition of the organic ligands on the surface. It is seen there that the threshold for such conversion varies widely, from approximately 270 K on Ni(110) to 470 K on Cu(110) and 550 K on SiO 2. On oxides (or oxygen-predosed surfaces), most of the atomic hydrogen generated on the surface may evolve as H 2 O instead of H 2. This is what was seen for the oxygen-pretreated copper and nickel substrates, in which cases water evolution was determined to start at ~430 and 500 K, respectively. A qualitative difference between these two "oxides" is that, while on copper oxygen is atomically adsorbed on the metal and aids hydrogen abstraction steps from the copper ligands (leading to decomposition of the precursor at temperatures lower that on clean Cu), on nickel a thin oxide film passivates the activity of the chemical metal. It should be also indicated that on metals some of the decomposition products include organics such as olefins and nitriles, whereas with coadsorbed oxygen oxidation to carbon monoxide and carbon dioxide dominates instead. 18

19 The decomposition of the ligands associated with the Cu(I)-sec-butyl-2-iminopyrrolidinate precursor upon thermal activation on surfaces is accompanied by reduction of the Cu ion to metallic copper. That transformation is also affected by the nature of the solid, and roughly follows the same trends as those seen for the decomposition of the ligands. Specifically, it can be seen in Figure 6 that on Ni(110) reduction to Cu(0) occurs immediately after molecular desorption of any excess precursor at 300 K. On Cu(110), by contrast, the reduction takes place between 500 and 600 K. Curiously, on NiO/Ni(110), both oxidation states coexist over a wide range of temperatures, between approximately 200 and 500 K. Finally, on SiO 2 /Ta, a conversion is seen about 600 K, but the copper center is never fully reduced, possibly because it retains its ionic character while bonded to individual oxygen or silanol surface groups. The implications of these differences in surface chemistry for ALD processes were explored as well. Figure 7 shows that self-limiting adsorption on either Ni(110) or NiO/Ni(110) switches into a CVD-like regime by 450 K, a lower temperature than that seen with SiO 2 /Ta (on which the CVD-type uptake starts between 500 and 550 K). The continuous deposition of copper with increasing dosing seen at high temperature, past the first monolayer, is explained by the increased decomposition rates seen in the H 2 TPD data in Figure 8. Finally, a successful start of the ALD of metallic copper on nickel was demonstrated using molecular oxygen as a coreactant. Acknowledgements 19

20 Financial support for this project was provided by a grant from the U.S. Department of Energy, Office of Science, Basic Energy Sciences, Materials Sciences and Engineering (MSE) Division, under Award No. DE-FG02-03ER

21 References 1 R. W. Johnson, A. Hultqvist, and S. F. Bent, Mater. Today 17 (5), 236 (2014); G. N. Parsons, J. W. Elam, S. M. George, S. Haukka, H. Jeon, W. M. M. Kessels, M. Leskelä, P. Poodt, M. Ritala, and S. M. Rossnagel, J. Vac. Sci. Technol., A 31 (5), (2013) S. M. George, Chem. Rev. 110 (1), 111 (2010). K. Knapas and M. Ritala, Crit. Rev. Solid State Mater. Sci. 38 (3), 167 (2013). F. Zaera, Coord. Chem. Rev. 257 (23 24), 3177 (2013). S. E. Potts and W. M. M. Kessels, Coord. Chem. Rev. 257 (23 24), 3254 (2013). R. G. Gordon, in Atomic Layer Deposition for Semiconductors, edited by S. C. Hwang (Springer US, Boston, MA, 2014), pp J. Hämäläinen, M. Ritala, and M. Leskelä, Chem. Mater. 26 (1), 786 (2014). F. Zaera, J. Mater. Chem. 18, 3521 (2008); F. Zaera, J. Phys. Chem. Lett. 3 (10), 1301 (2012). 9 A. J. M. Mackus, A. A. Bol, and W. M. M. Kessels, Nanoscale 6 (19), (2014); M. Fang and J. C. Ho, ACS Nano 9 (9), 8651 (2015) X. Qin, H. Sun, and F. Zaera, J. Vac. Sci. Technol. A 30 (1), 01A112 (2012). J. Kim and T. Kim, JOM 61 (6), 17 (2009). R. Methaapanon, S. M. Geyer, S. Brennan, and S. F. Bent, Chem. Mater. 25 (17), 3458 (2013); A. J. M. Mackus, M. A. Verheijen, N. Leick, A. A. Bol, and W. M. M. Kessels, Chem. Mater. 25 (9), 1905 (2013). 13 J. P. Coyle, P. J. Pallister, A. Kurek, E. R. Sirianni, G. P. A. Yap, and S. T. Barry, Inorg. Chem. 52 (2), 910 (2013). 21

22 14 B. Chen, Y. Duan, Y. Yao, Q. Ma, J. P. Coyle, S. T. Barry, A. V. Teplyakov, and F. Zaera, J. Vac. Sci. Technol. A, submitted (2016). 15 J. Rickerby and J. H. G. Steinke, Chem. Rev. 102 (5), 1525 (2002); A. Bhatnagar, M. Naik, S. Ramaswami, M. Spuller, M. Armacost, R. Perry, J. Van Gogh, J. Shu, and G. Miner, Solid State Technol. 52 (10), 16 (2009); J. Gambino, in Handbook of Thin Film Deposition (Third Edition), edited by K. Seshan (William Andrew Publishing, Oxford, 2012), pp Z. Tőkei, K. Croes, and G. P. Beyer, Microelectron. Eng. 87 (3), 348 (2010); D. Priyadarshini, S. Nguyen, H. Shobha, S. Cohen, T. Shaw, E. Liniger, C. K. Hu, C. Parks, E. Adams, J. Burnham, A. H. Simon, G. Bonilla, A. Grill, D. Canaperi, D. Edelstein, D. Collins, M. Balseanu, M. Stolfi, J. Ren, and K. Shah, presented at the Interconnect Technology Conference / Advanced Metallization Conference (IITC/AMC), 2014 IEEE International, 2014 (unpublished) Q. Ma, F. Zaera, and R. G. Gordon, J. Vac. Sci. Technol. A 30 (1), 01A114 (2012). Q. Ma and F. Zaera, J. Vac. Sci. Technol., A 31 (1), 01A112 (2013). T. Kim, Y. Yao, J. P. Coyle, S. T. Barry, and F. Zaera, Chem. Mater. 25 (18), 3630 (2013) Y. Yao, J. P. Coyle, S. T. Barry, and F. Zaera, J. Phys. Chem. C 120 (26), (2016). F. Zaera, J. Vac. Sci. Technol. A 7 (3), 640 (1989); T. Kim and F. Zaera, J. Phys. Chem. C 116 (15), 8594 (2012) Q. Ma, H. Guo, R. G. Gordon, and F. Zaera, Chem. Mater. 23 (14), 3325 (2011). Q. Ma and F. Zaera, J. Vac. Sci. Technol., A 33 (1), 01A108 (2015). H. Öfner and F. Zaera, J. Phys. Chem. B 101 (44), 9069 (1997); H. Guo and F. Zaera, 22

23 Nat. Mater. 5 (6), 489 (2006) B. E. Hayden and C. L. A. Lamont, J. Phys., Condens. Matter 1 (SB), SB33 (1989). Y. A. Wasslen, A. Kurek, P. A. Johnson, T. C. Pigeon, W. H. Monillas, G. P. A. Yap, and S. T. Barry, Dalton Trans. 39 (38), 9046 (2010). 27 P. R. Davies and M. Bowker, Catal. Today 154 (1 2), 31 (2010); Y. Yao and F. Zaera, Surf. Sci. 650, 263 (2016); Y. Yao and F. Zaera, Surf. Sci. 646, 37 (2016). 28 N. R. Gleason and F. Zaera, J. Catal. 169, 365 (1997); F. Zaera, N. R. Gleason, B. Klingenberg, and A. H. Ali, J. Mol. Catal. A 146 (1-2), 13 (1999). 29 Handbook of X-Ray Photoelectron Spectroscopy, edited by C. D. Wagner, W. M. Riggs, L. E. Davis, J. F. Moulder, and G. E. Muilenberg (Perkin-Elmer Corporation, Eden Prairie, MN, 1978) Q. Ma, H. Guo, R. G. Gordon, and F. Zaera, Chem. Mater. 22 (2), 352 (2010). T. Schneider and W. Hirschwald, Catal. Lett. 16 (3), 335 (1992). S. Mueller, T. Waechtler, L. Hofmann, A. Tuchscherer, R. Mothes, O. Gordan, D. Lehmann, F. Haidu, M. Ogiewa, L. Gerlich, S. F. Ding, S. E. Schulz, T. Gessner, H. Lang, D. R. T. Zahn, and X. P. Qu, presented at the Semiconductor Conference Dresden (SCD), 2011, 2011 (unpublished); D. Dhakal, T. Waechtler, S. E. Schulz, T. Gessner, H. Lang, R. Mothes, and A. Tuchscherer, J. Vac. Sci. Technol., A 32 (4), (2014). 23

24 Figure Captions Figure 1. Temperature programmed desorption (TPD) data for saturation layers of copper(i)-nsec-butyl-2-iminopyrrolidinate on four different surfaces, namely, Ni(110) (left panel), a ~3 ML thin NiO film grown on Ni(110) (second from left), Cu(110) (second from right), and a Cu(110) surface saturated with atomic oxygen (right). Traces are displayed for masses representative of the main desorbing products, which for the metals include H 2 (2 amu), C 2 H 4 (26, 27, 28 amu), C 4 H 8 (56 amu), HCN (26, 27 amu), and N 2 (14, 28 amu) and for the oxygen-containing surfaces H 2 O (18 amu), CO (28 amu) and CO 2 (44 amu). Molecular desorption was followed in all cases by using the signal for 84 amu. Figure 2. H 2 (for the metals and for SiO 2 ) or H 2 O (for NiO/Ni(110) and O/Cu(110)) TPD traces for the five surfaces used in this study. Detection of these hydrogen-containing products reflects the kinetics of the surface dehydrogenation steps associated with the decomposition of the iminopyrrolidinate ligand. The threshold temperature at which point such conversion is initiated varies widely, from 270 K on Ni(110) to 550 K on SiO 2 /Ta. Figure 3. C 1s (left panel) and N 1s (right) XPS data for copper(i)-n-sec-butyl-2- iminopyrrolidinate adsorbed on Cu(110) as a function of annealing temperature. Figure 4. Similar set of C 1s (left panel) and N 1s (right) XPS traces as a function of annealing 24

25 temperature for the iminopyrrolidinate adsorbed on NiO/Ni(110). Figure 5. Corresponding Cu 2p 3/2 XPS temperature-dependent data for the Cu(110) (left panel) and NiO/Ni(110) (center) surfaces, and Cu L 3 VV Auger (AES) traces for the case of the NiO/Ni(110) surface (right), to accompany the results reported in Figures 3 and 4. The traces for the Cu(110) are reported in differential mode, after subtracting the spectrum for the clean surface. Figure 6. Cu 2p 3/2 XPS peak binding energies (BEs) as a function of annealing temperature for the copper iminopyrrolidinate adsorbed on Ni(110), NiO/Ni(110), Cu(110), and SiO 2 /Ta. The data for the second and third surfaces were estimated from the raw spectra in Figure 5, the ones for the first and last were taken from our previous publications. 19,20 The reduction of the copper atoms, signified by the shifts in BEs from values above ev to the ev range, occurs at different temperatures on the different substrates. Figure 7. Top: Cu 2p 3/2 XPS traces obtained after 60 min exposures of the Ni(110) (left panel) and NiO/Ni(110) (right) surfaces to Cu(I)-sec-butyl-2-iminopyrrolidinate at the indicated temperatures. Bottom: Corresponding uptake curves versus exposure time, estimated from the peak intensities of data sets such as those on the top. Self-limiting deposition is seen at 350 and 400 K, whereas continuous film growth is observed at 450 K and above. 25

26 Figure 8. H 2 TPD traces for the copper iminopyrrolidinate precursor adsorbed on Ni(110) (left panel), Cu(110) (center), or NiO/Ni(110) (right) as a function of deposition temperature. Increases in hydrogen production in the high-temperature end of these traces are seen in the first two cases after deposition at 500 K. Figure 9. Results from titration experiments on Cu(110) after dosing Cu(I)-sec-butyl-2- iminopyrrolidinate at 400 K and annealing to the indicated temperatures. Those surfaces were saturated with carbon monoxide at 80 K, after which the reported CO (28 amu) TPDs were acquired. The detection of approximately 25% of the CO that adsorbs on the clean surface after annealing to 400 or 700 K indicates the availability of some adsorption sites after the copper dosing. However, those disappear after annealing to 800 K, rendering the surface totally unreactive. Figure 10. Cu 2p 3/2 (left panel), C 1s (center), and N 1s (right) XPS traces after each half cycle of a copper ALD sequence with Cu(I)-sec-butyl-2-iminopyrrolidinate and molecular oxygen on Ni(110). Each Cu-precursor dose was done for 20 min at 400 K, and the subsequent oxidation was carried out by annealing the surface in 1 x 10 6 Torr of O 2 at 650 K for 30 min. Data are reported for the first, second, third, and eighth cycles. Copper is continuously deposited in metallic form, in an approximately linear fashion as a function of the number of ALD cycles, and the carbon and nitrogen of the ligands are fully removed in each oxidation half cycle. 26

27 Scheme 1 27

28 Figure 1 28

29 Figure 2 29

30 Figure 3 30

31 Figure 4 31

32 Figure 5 32

33 Figure 6 33

34 Figure 7 34

35 Figure 8 35

36 Figure 9 36

37 Figure 10 37

38

39

40

41

42

43

44

45

46

47

48