In-situ observation of self-cleansing phenomena during ultra-high vacuum anneal of transition metal nitride thin films : Prospects for non-destructive photoelectron spectroscopy

Self-cleansing of transition metal nitrides is discovered to take place during ultrahigh vacuum annealing of TiN, NbN, and VN thin films. Native oxide layers from air exposure disappear after isothermal anneal at 1000 °C. Also, for TiN, the Ti 2p and N 1s X-ray photoelectron spectra (XPS) recorded after the anneal are identical to those obtained from insitu grown and analyzed epitaxial TiN(001). These unexpected effects are explained by oxide decomposition in combination with N-replenishing of the nitride during recrystallization. The finding opens up new possibilities for true bonding assignments through non-destructive XPS analyses, thus avoiding artefacts from Ar etching.

The importance of X-ray photoelectron spectroscopy (XPS) in materials science cannot be overestimated.Over last decades XPS became an essential research technique for assessing the surface chemistry and composition of compounds in bulk or thin film form.Many classes of samples require some form of cleaning prior to spectra acquisition to remove surface oxides and contaminants, which is typically done in-situ by etching with 0.5-4 keV Ar + ions.The latter is performed despite well-known destructive effects of ion bombardment like cascade mixing, chemical reduction, recoil implantation, segregation, surface roughening, preferential sputtering, redeposition, and forward implantation of surface species, to name the predominant. 1This destructive type of surface treatments is particularly adherent to XPS studies of refractory ceramic thin films grown by physical vapor deposition (PVD), for applications such as protective layers on high-speed cutting tools 2,3 engine parts, 4,5 as well as diffusion barriers in electronics. 6,7,8Interpretation of XPS results obtained in such way always poses a number of questions as to what extent the actual XPS spectra reflect the native material to be studied instead of the ion-beam modified surface layer, which thickness is comparable to the XPS probing depth.
Here, we report on a surprising observation that allows for a non-destructive acquisition of high-quality XPS spectra characteristic of a native material from samples that have been exposed to atmosphere.High-temperature anneal of transition metal (TM) nitride (TM = Ti, V, Nb) thin films under ultrahigh vacuum (UHV) conditions leads to surface self-cleansing as evidenced by XPS.Heat treatments performed in the vacuum vessel with the base pressure better than 1.5×10 -10 Torr (2×10 -8 Pa) and in the temperatures range 200 to 1000 °C cause gradual removal of the native oxide layer following even 2-years-long storage in air.The effect is believed to be triggered by the recrystallization process, the latter evidenced by an increased grain size and reduced residual stress from defect annihilation.The accompanying surface reconstruction leads to the release of CO, CO2, ONH, and H2O species.The Ti 2p and N 1s XPS spectra acquired from TiN samples after the 1000 °C anneal are identical to those obtained from in-situ grown epitaxial TiN/MgO(001).The oxygen concentrations following the heat treatment are lower than those recorded from Ar + -etched surfaces, without destructive effects from ion bombardment.Polycrystalline (TM)N thin films of thickness 200±10 nm are grown on Si(001) substrates at 410 °C by reactive high power pulsed magnetron sputtering (HIPIMS) 9,10,11 in a CC800/9 CemeCon AG system using rectangular 8.8×50 cm 2 target and Ar:N2 (4:1) gas mixture.HIPIMS is operated at the average power of 1.3 kW, the pulsing frequency of 600 Hz, and the duty cycle of 12%.Substrate bias is applied in the form of 200-µs-long pulses synchronized with HIPIMS cathode and the amplitude of -60 V. 12,13 The target-to-substrate distance is 6 cm, while the total pressure during deposition is 3 mTorr (0.4 Pa).Following film growth, the samples are allowed to cool down to 180 °C before the deposition chamber is ventilated, which allows for a better control of surface chemistry upon air exposure. 142θ x-ray diffraction scans reveal that all films possess single-phase NaCl-crystal structure.TiN layers exhibit no preferred orientation, while NbN and VN films are 002-and 111-oriented, respectively.Rutherford backscattering spectroscopy (RBS) gives N/Ti = 0.93±0.01,N/Nb = 1.13 and N/V = 0.85, while from time-of-flight energy elastic recoil detection analyses (ToF-E ERDA) 15 the O bulk concentrations are ≤ 0.5 at% in all layers.
Annealing employing an e-beam heater is performed in the UHV chamber directly connected to the XPS instrument, with a base pressure <1.5×10 -10 Torr (2×10 -8 Pa), which raises during the treatment to a maximum of 3.8×10 -8 Torr (5.1×10 -6 Pa) due to sample outgassing.
After the anneal, samples are allowed to cool for ca.0.5 h and then transferred in UHV to the analysis chamber for XPS characterization.
Core level XPS spectra are acquired using an Axis Ultra DLD instrument from Kratos Analytical, with a base pressure of 1.1×10 -9 Torr (1.5×10 -7 Pa) and monochromatic Al Kα source (hν = 1486.6eV).The binding energy (BE) scale is calibrated by examining the sputtercleaned Au, Ag, and Cu samples according to the recommended ISO standards for monochromatic Al Kα sources and setting the Fermi level cut-off recorded from TM(N) surfaces at zero energy. 16,17Spectra deconvolution and quantification is performed using CasaXPS software employing Shirley-type background, 18 and manufacturer's sensitivity factors. 19e set of normalized Ti 2p core-level spectra acquired from air-exposed polycrystalline TiN (poly-TiN) film as a function of the in situ annealing temperature Ta is shown in Fig. 1(a).
We focus here on the stronger 2p3/2 peaks of the spin-orbit split doublet, located in the BE region from 455.0 to 460.0 eV, which give the clearest signature of Ti chemical states compared to their 2p1/2 counterparts.Since the sample has been stored in air for longer period of time, the spectrum of the as-deposited film, apart from the Ti-N component at 455.0 eV, possess a clear signature of Ti-OxNy and Ti-O2 peaks at 456.6 and 458.2 eV, respectively. 20,21,22,23,24The formation of native oxides and oxynitrides is regularly encountered in the XPS practice as insitu XPS capability is rare and for vast majority of samples exposure to ambient atmosphere during transport to spectrometer cannot be avoided.Surprisingly, already after 1 h anneal at 200 °C a significant decrease of Ti-OxNy and Ti-O2 signals is observed, which continues with further increasing Ta up to 1000 °C (a maximum temperature available in our setup).
Eventually, Ti 2p signal obtained after the 1000 °C in-situ anneal is essentially identical to that of epitaxial stoichiometric TiN/MgO(001) films (epi-TiN) grown and analyzed in-situ in a UHV XPS system, as discussed below. 25,26In particular, the satellite features on the high BE side above the primary peaks, 27,28,29 exhibit high intensities comparable to those obtained from epi-TiN.We have also established in control experiments that (i) a direct 1 h long anneal at 1000 °C gives essentially the same result as the temperature ramp, and (ii) a prolonged treatment at 200 °C (3 h) leads to marginally small change as compared to the 1 h anneal.
The corresponding set of normalized N 1s spectra from as-deposited poly-TiN films shown in Fig. 1(b) is fully consistent with the evolution of the Ti 2p signals.The spectrum consists of a main peak at 397.2 eV due to Ti-N, a pronounced feature at 396.1 eV assigned to Ti-OxNy formation, and an asymmetrical tail on the high BE side.Following a 200 °C anneal, the intensity of the Ti-OxNy peak is drastically reduced, in accordance with the evolution of the Ti 2p spectrum.In addition, there is a shoulder on the low BE side of the main N 1s peak, which we assign to Ti-OxNy species with an y/x ratio significantly higher than for the as-deposited sample, thus possessing stoichiometry closer to that of TiN.This interpretation is supported by changes observed upon further anneals, which lead to gradual decrease of this feature, until complete disappearance after the Ta = 1000 °C treatment.With increasing Ta the presence of a small satellite peak at ~399.8 eV emerges, in agreement with previous reports. 20complementary Ta sets of O 1s and C 1s spectra are shown in Figures 1(c) and 1(d), respectively.The anneal at 200 °C, leads to a drastic decrease in the Ti-OxNy, O-C=O, C-O, and C-C/C-H signal intensities.This is expected as these C-containing species are likely due to adventitious carbon accumulating on the surface upon air exposure and typically weakly bonded to the oxides. 30With increasing Ta, the O 1s spectra confirm a gradual loss of TiO2, and eventually only residual amounts on oxygen are found on the surface following the 1000 °C anneal, likely due to redeposition of released oxygen-containing species during the relatively long acquisition time of 2 h.C 1s spectrum recorded after the final anneal step indicates, in addition to residual amounts of hydrocarbons, also a small contribution at ∼282.1 eV, assigned to TiC unintentionally formed during film growth due to reaction with residual gases. 31Very low intensity of this peak accounts for the fact that no TiC contribution is indicated in the corresponding Ti 2p spectrum.
In Figures 2(a) and 2(b) Ti 2p and N 1s spectra recorded from UHV-annealed (Ta = 1000 °C) poly-TiN films are compared to those acquired from (i) in-situ grown and analyzed epi-TiN from Ref. 32, and (ii) poly-TiN in-situ capped with a 1.5-nm-thick Al layers in the deposition system immediately after the film growth and prior to air-exposure. 33The latter provides a good barrier towards oxidation and allows non-destructive XPS with data quality characteristic of films grown and analyzed in-situ. 34In addition, spectra recorded from poly-TiN films treated in a conventional way, i.e., with Ar + ion etch, are included to illustrate detrimental effects of ion bombardment. 35Clearly, the Ti 2p spectrum from UHV-annealed poly-TiN is essentially identical to that obtained from epi-TiN samples and in agreement to Alcapped poly-TiN films.In particular, the intensity of the satellites features is similar, but higher than after sputter etching with low energy Ar ions   + = 0.5 keV incident at a shallow angle of ψ = 70° with respect to the surface normal.The latter case constitutes a good example of Ar + -induced sputter damage, for a modified surface layer thickness by the sputter beam 36 that is comparable to the XPS probing depth. 37These destructive effects are even more pronounced for   + = 4 keV and ψ = 45°, with almost complete smearing out of the satellite peaks.Changes in the corresponding N 1s spectra, confirm destructive effects of Ar + etch, which are manifested by broadening of the Ti-N peak and smearing out of the satellite feature on the high BE side (better visible in the inset of Fig. 2(b)).In fact, the spectra obtained after 1000 °C UHV anneal have a superior quality to that seen for in-situ Al-capped poly-TiN.This is because inelastic scattering in nm-thick capping layers is avoided, which results in a noticeably lower background level on the high BE side of both Ti 2p and N 1s signals, as well as narrower peaks.
To get insight into the mechanism behind the intriguing removal of surface native oxides by high-Ta UHV annealing, we examined other members of the TM(N) family, focusing on the widely-applicable group IVb-VIb.Representative examples are included in Figures 3(a)-3(c) where the strongest metal core level signals obtained from polycrystalline NbN, VN, and ZrN films are shown.Spectra are recorded in the as-deposited state after 1000 °C UHV anneals, and following an Ar + -etch (  + = 0.5 keV, ψ = 70°).The 3d spectrum of NbN (Fig. 3(a)) is of particular interest as, in contrast to the Ti 2p spectra of poly-TiN, no satellites features are observed on the high BE side of the main peaks 38,34 hence, any residual oxide left on the surface after UHV anneal may be directly observed.Clearly, Nb 3d5/2 and Nb 3d3/2 oxide peaks, originally present at 207.3 and 210.2 eV in the spectrum from the as-deposited film, are completely gone after the 1000 °C anneal.The remaining doublet with the 3d5/2 -3d3/2 peaks at 204.0 and 206.8 eV, corresponds well to nitride spectra obtained from stoichiometric NbN films grown in situ by N2 + ion implantation. 38Corresponding data for Ar + -etched samples reveal sputter damage with low BE peaks shifted with respect to the original nitride peaks, and confirms the benefits of our non-destructive XPS by means of UHV anneal.
The V 2p spectra recorded from VN layers (Fig. 3(b)) are interesting for their proximity to the O 1s peak.UHV-annealed films exhibits main spin-split 2p3/2-2p1/2 peaks at 513.4 and 520.9 eV, with no traces of oxygen, which is spectacular taking into account that the O-free surface is not obtained by sputter-etching, in which case remaining O 1s signal is still detected at ∼531 eV, due in large part to recoil mixing or redeposition of the sputtered oxygen.In analogy to the Ti 2p signal of poly-TiN, the intensity of the satellite peaks of UHV-annealed VN is significantly higher than for the sputter-cleaned samples and agrees with the results published for oxide-less films mechanically cleaned in situ in the UHV system. 39wever, not all TM(N)'s undergo self-cleansing upon 1000 °C UHV annealing.An exemption is ZrN, where Fig. 3(c) shows Zr 3d spectra that are dominated by 3d5/2 -3d3/2 oxide peaks at 182.4 and 184.7 eV, with low intensity of corresponding nitride peaks at 180.0 and 182.4 eV.The spectrum is similar to that of as-deposited ZrN, thus different from oxide-free spectra of sputter-cleaned films.As the O/Zr ratio decreases by only 10% upon anneal, the identical heat treatment to that performed on poly-TiN, has no effect on ZrN.Other examples where the oxide was not removed by 1000 °C anneal in UHV are TaN and HfN.
In previous experiments designed to investigate the TiN surface chemistry upon contact with oxygen-containing atmospheres (dry air or O2), 23,24,40,41 XPS studies confirmed a severe loss of N and TiO2 formation at temperatures as low as Ta ∼ 400 °C42,24 according to TiN+O2 → TiO2+1/2N2.Typical XPS spectrum after such few hours long treatment is characteristic of a pure TiO2, indicating that the thickness of the reacted layer exceeds the XPS probing depth.
In contrast, our results show that the UHV anneal of air-exposed poly-TiN leads to an extensive oxygen loss, with O/Ti ratio decreasing from 0.83 for the as-deposited film to 0.06 after 1000 °C anneal, which is accompanied by insignificant N loss.The XPS-derived N/Ti ratio following the 1000 °C treatment is 0.84, much higher than 0.74 obtained from poly-TiN etched with lowenergy (0.5 keV) Ar + ions, and only 10% lower than the bulk N/Ti ratio of 0.93.Furthermore, as shown in Figure 4, x-ray diffraction (XRD) measurements following UHV anneal indicate substantial reduction in the full-width-at-half-maximum (FWHM) accompanied by an increased intensity of 001 and 111 Bragg reflections for TM(N) films, where self-cleansing is observed (TiN, NbN, and VN).For example, FWHM of the TiN(001) decreases upon anneal from 0.89° to 0.36°, accompanied by an increase in peak intensity by a factor of 3.5.This corresponds to recrystallization where the average crystallite size increases from 11 to 26 nm, using Scherrer's formulae. 43Simultaneously film recovers from a growthinduced compressive stress state (indicated by the shift of 001 and 111 XRD peaks towards higher diffraction angles, see Fig. 4), from -2.8 GPa measured on the as-grown films to -0.3 GPa for the Ta = 1000 °C sample, as estimated by the sin 2 ψ method. 44XRD shows no, or very little change for layers that retain native oxide, such as ZrN (see Fig. 4(d)), in which case a stoichiometric signature in both bonding and lattice parameter is preserved.
Based on the striking correlation between the removal of surface oxides and recrystallization, we propose that the latter process triggers the self-cleansing.Diffusion of N to the film surface together with C from residual gases or adventitious C results in reactions that disintegrate the native oxide by the release of gaseous species CO/N2 (28 amu), ONH (31 amu), H2 (2 amu), H2O (18 amu), and CO2 (44 amu), as indicated by residual gas analyzer (RGA).The out-diffusing N also replenishes the nitride surface for possible reconstruction, hence the N/Ti ratio in the surface region of annealed films is comparable to that in the bulk.Thus, the activation energy for recrystallization of a given nitride is a key parameter that steers the effectiveness of the UHV anneal treatment in removing surface oxides.
The type and stoichiometry of the surface oxide is correspondingly expected to play a crucial role as there is a clear correlation between the effectiveness of the 1000 °C UHV anneal and the heat of oxide formation Δ   0 .All oxides studied here that were not affected by the high-Ta treatment have Δ   0 in the range -1098 kJ/mol (ZrO2) to -2046 kJ/mol (Ta2O5), which is significantly higher than -420 kJ/mol (NbO), -431 kJ/mol (VO), and -944 kJ/mol (TiO2), where oxides are effectively removed. 45In the case of poly-TiN film the fact that native oxide is not a pure-phase stoichiometric compound, but rather a mixture of TiO2 and TiOxNy (cf.Fig. 1(a)), with the latter compound located closer to the TiN interface, 14 may facilitate the cleansing process.
The scenario presented above is corroborated by annealing experiments performed with epi-TiN(111) and epi-TiN(001) films (not shown).Although the Ti 2p spectra quality improved after 1000 °C treatment, qualitative analyses indicated that removal of surface oxygen is not as efficient as for the poly-TiN samples, with the O/Ti ratio varying from 1.44 to 0.53, and from 1.04 to 0.31, for epi-TiN(111) and epi-TiN(001), respectively.Greatly reduced number of grain boundaries in the case of epitaxial layers, removes the more easily thermally-activated sites for inter-diffusion and recrystallization, considering that 1000 °C is only 1/3 of the melting temperature of TiN.
In summary, while performing in-situ annealing experiments of magnetron-sputtered TM(N) films in UHV we unexpectedly observe that a few hours at 1000 °C yields oxide-free