Low temperature (Ts/Tm<0.1) epitaxial growth of HfN/MgO(001) via reactive HiPIMS with metal-ion synchronized substrate bias

Low-temperature epitaxial growth of refractory transition-metal nitride thin films by means of physical vapor deposition has been a recurring theme in advanced thin-film technology for several years. In the present study, 150-nm-thick epitaxial HfN layers are grown on MgO(001) by reactive high-impulse magnetron sputtering (HiPIMS) with no external substrate heating. Maximum film growth temperatures Ts due to plasma heating range from 70-150 {\deg}C, corresponding to Ts/Tm = 0.10-0.12 (in which Tm is the HfN melting point in K). During HiPIMS, gas and sputtered-metal ion fluxes incident at the growing film surface are separated in time due to strong gas rarefaction and the transition to a metal-ion dominated plasma. In the present experiments, a negative bias of 100 V is applied to the substrate, either continuously during the entire deposition or synchronized with the metal-rich portion of the ion flux. Two different sputtering-gas mixtures, Ar/N2 and Kr/N2, are employed in order to probe effects associated with the noble-gas mass and ionization potential. The combination of x-ray diffraction, high-resolution reciprocal-lattice maps, and high-resolution cross-sectional transmission electron microscopy analyses establish that all HfN films have a cube-on-cube orientational relationship with the substrate, i.e., [001]HfN||[001]MgO and (100)HfN||(100)MgO. Layers grown with continuous substrate bias, in either Ar/N2 or Kr/N2, exhibit a relatively high mosaicity and a high concentration of trapped inert gas. In distinct contrast, layers grown in Kr/N2 with the substrate bias synchronized to the metal-ion-rich portion of HiPIMS pulses, have much lower mosaicity, no measurable inert-gas incorporation, and a hardness of 25.7 GPa, in good agreement with results for epitaxial HfN(001) layers grown at Ts = 650 C (Ts/Tm = 0.26).

heating, was 420 °C (Ts/Tm ≃ 0.22, in which Tm is the TiN melting point, 3220 K). 31 While Ts was much less than for conventional epitaxial TiN(001), typically ≥ 700 °C (Ts/Tm ≃ 0.30), 26 it is still too high for many applications, including film growth on temperature-sensitive substrates such as polymers and light-weight metals (e.g., Li, Mg, and Al).
Here, we investigate the possibility of using reactive high-power impulse magnetron sputtering (HiPIMS) to grow epitaxial TM nitrides at much lower temperatures, Ts/Tm ≃ 0.1.
HiPIMS allows ionization of up to 90% of sputtered metal atoms, depending upon pulsing conditions and the choice of metal target. 32 Thus, HiPIMS offers the opportunity to increase momentum transfer to the growing film, and enhance adatom mean-free paths due to irradiation by low-energy ionized sputtered metal atoms, rather than gas ions, during bias deposition. 33 Greczynski et al 34 probed the effect of metal-versus gas-ion irradiation during Ti1-xAlxN film growth in mixed Ar/N2 atmospheres using synchronized pulsed substrate bias in a hybrid HiPIMS/DCMS co-sputtering configuration. The results demonstrated that synchronizing the substrate bias with the metal-rich portion of HiPIMS pulses provides film densification, microstructure enhancement, surface smoothening, and decreased film stress with no measurable Ar incorporation.
We report the LTE growth of HfN on MgO(001), in the absence of applied substrate heating, using reactive HiPIMS in Ar/N2 and Kr/N2 discharges with low-energy ion irradiation of the growing film. The two gas mixtures are selected to provide significantly different momentum transfer from gas ions to the growing film (mAr = 39.96 amu, mKr = 83.79 amu, while mHf = 178.49 amu), and different ionization potentials (the first ionization potential of Ar, Ar 1 , is 15.76 eV, while Kr 1 = 14.00 eV). 31 HfN/MgO(001) is chosen as a model system for these experiments since: (1) HfN has the highest melting point among the TM nitrides (Tm = 3310 °C [3583 K]), 31 (2) the HfN/MgO(001) system has a large lattice mismatch (7.46%), 28 and (3) the growth of high-quality HfN/MgO(001) single-crystals has been demonstrated at elevated temperatures. 28,35 We isolate the effect of the nature of the incident ions (gas vs. metal) during film growth by comparing the results of two different modes of applied substrate bias: (i) continuously applied bias during the entire deposition and (ii) bias applied only in synchronous with the metal-rich portion of each HiPIMS pulse. The maximum substrate temperature, due to plasma heating during film growth, is 150 °C (Ts/Tm < 0.12) with continuously-applied bias and 70 °C (Ts/Tm < 0.10) using synchronized bias. maps (HR-RLMs), and high-resolution cross-sectional transmission electron microscopy (XTEM) analyses establish that all HfN films have a cube-on-cube relationship with their MgO (001) substrates. The epitaxial layers with the highest quality, grown with synchronized bias in Kr/N2 discharges, are essentially fully-relaxed with no measurable concentration of trapped gas atoms and, hence, low film stress. They also have much lower room-temperature resistivity, 70  In situ time-dependent mass and energy spectroscopy analyses of ion fluxes incident at the substrate plane are carried out using a Hiden Analytical PSM003 instrument to determine relative compositions, charge states, and energies as a function of the choice of inert gas and bias scheme.
The mass-spectrometer orifice is electrically grounded during these experiments and the ion energy is scanned from 1 to 80 eV in 0.5 eV steps. Further details regarding the mass spectrometer can be found elsewhere. 39 The separation between the mass spectrometer orifice, located below the substrate position, and the center of the target is 21 cm. Thus, measured energy distributions underestimate ion-impact energies at the substrate due to additional gas-phase collisions occurring between the substrate plane position and the entrance to the mass spectrometer. Ion times-of-flight (TOF) are corrected, following the procedure of Bohlmark et al, 39 for the difference in position and listed in Table 1 for the lowest and highest incident kinetic energies.
Since the quadrupole mass analyzer has a higher transmission at lower mass, 40 and because of the large mass differences between metallic and gaseous ions in these investigations (ranging from 14 to 180 amu, see Table 1), the spectrometer settings are separately tuned for each species. consecutive pulses such that the total acquisition time per data point is 2 ms. To synchronize the measurements with the target-pulse onset, the mass-spectrometer circuit is triggered by a transistortransistor logic (TTL) pulse sequence generated at the output of the HiPSTER synchronization unit.
The gate width is 5 µs, with a 1 µs delay following the pulse onset.
Compositions of as-deposited HfNx layers are determined by time-of-flight elastic recoil detection analyses (TOF-ERDA) employing a 36 MeV 127 I 8+ probe beam incident at 67.5° with respect to the sample normal; recoils are detected at an angle of 45°. The results are analyzed using CONTES software. 41 Uncertainties in reported x values for HfNx films are less than +0.025.
A high-resolution PANalytical X'pert XRD diffractometer with a CuKα1 source (λ = 1.540597 Å) and a four-crystal Ge(220) monochromator is used for determining film orientation.
Lattice parameters perpendicular a⊥ and parallel a∥ to the film surface, and residual strains ε, are obtained from high-resolution reciprocal lattice maps (HR-RLMs) around asymmetric 113 reflections. The RLMs consist of a series of ω-2θ scans acquired over a range of ω offsets. Highresolution transmission electron microscopy (HR-TEM) analyses are carried out in a FEI Tecnai G 2 TF 20 UT instrument operated at 200 kV. Cross-sectional specimens are prepared using a twostep procedure consisting of mechanical polishing followed by Ar + -ion milling at a shallow incidence angle of 6° from the sample surface, with gradually decreasing ion energies: 5 keV initially, followed by 2.5, and 1 keV.
Room-temperature electrical resistivities ρ, as a function of film-growth conditions, are obtained from four-point probe results, with each reported ρ value averaged over ten separate measurements. Nanoindentation analyses of as-deposited HfN films are performed using a Hysitron TI-950 Triboindenter equipped with a sharp Berkovich diamond probe calibrated using a fused-silica standard and a single-crystal TiN(001) reference sample. 28

A. Plasma characterization
Prior to initiating film-growth experiments, reactive-HiPIMS discharges in mixed Ar/N2 and Kr/N2 environments are characterized in order to design substrate bias strategies as described in subsections III.B and III.C. Figure 1 presents typical target current IT(t) and voltage VT(t) waveforms acquired during HIPIMS sputtering of Hf in Ar/N2 and Kr/N2 atmospheres. Both voltage waveforms are rectangular in shape with the maximum Kr/N2-discharge voltage 100 V higher than that of Ar/N2.
IT(t) waveforms exhibit delays in the target current rise onset of ~10 µs for Kr/N2 and ~25 µs for Ar/N2, after which the current increases rapidly to reach a maximum value at 50-60 µs, and then decreases to a saturation value at ~90 µs, before falling to zero at the end of the HiPIMS pulse.
The reason for the earlier current increase in the Kr/N2 discharge during pulse onset is due to a significantly higher Kr vs. Ar electron-impact-ionization cross section, 44 which decreases the time required to accumulate a sufficiently high electron density for gas breakdown (i.e., rapid discharge current increase). However, the differences in T ( ) during the remainder of the discharge pulse arise due to significant recycling of the ionized process gas (primarily Ar or Kr), 45,46 since the sputter yield of Hf by Ar + ions ( Hf Ar + ≃ 0.4) and by Kr + ions ( Hf Kr + ≃ 0.8) are well below unity, thus limiting self-sputter recycling. 46 A slightly lower maximum discharge current is expected for Kr/N2 discharges, since the ionization of Kr required for gas recycling is reduced compared to Ar by the lower electron temperature arising due to: (1) a higher concentration of Hf (for which Hf Ar + > Hf Kr + ) available in the ionization region with a much lower first-ionization potential, Hf + 1 = 6.83 eV compared to the process gas, 46 and (2) Kr + having a lower secondary-electron-emission yield than Ar + . 47 In situ time-averaged IEDFs for Ar + , N + , N 2+ , Hf + , and Hf 2+ gas and metal ions incident at the substrate plane during reactive HiPIMS sputtering of Hf in Ar/N2 and Kr/N2 atmospheres are shown in Figures 2(a) and 2(b), respectively. Ar 2+ and Kr 2+ ions are also detected, but not shown here since the sum of their contribution to the total measured ion intensity is < 1%. All gas-and metal-ion IEDFs exhibit a narrow low-energy peak located at 1.4 eV in Ar/N2 and 2.0 eV in Kr/N2 discharges. These values are consistent with previous mass spectrometry investigations of HiPIMS discharges, [48][49][50] and represent the potential drop between the bulk plasma potential and the grounded orifice of the mass spectrometer, across which thermalized ions are accelerated. 48 Hf + and Hf 2+ metal-ion IEDFs, for both Ar/N2 and Kr/N2 discharges, exhibit an intense broad peak with a maximum near 10 eV, followed by a high-energy tail extending to 80 eV, which we attribute to a superposition of a Sigmund-Thompson sputtered-species energy distribution, 51,52 ion acceleration by the combination of plasma and floating potentials, and HiPIMS plasma instabilities due to collective plasma effects. [53][54][55] Ar/N2 and Kr/N2 gas-ion IEDFs at the substrate plane display significant differences. As depicted in Figure 2(a), Ar + IEDFs possess a broad high-energy peak centered at ~65 eV. This differs dramatically from Kr + IEDFs (a typical result is shown in Figure 2(b)), for which only the low-energy peak is observed. The high-energy Ar + ions observed in Ar/N2 HiPIMS discharges arise primarily from the high probability of Ar + ions incident at the target being neutralized and reflected from heavy Hf target atoms toward the substrate plane with a significant fraction of the incident energy retained. 56 for which Ei is the kinetic energy of the incident Ar + ion, 350 eV, corresponding to acceleration across the cathode sheath. Equation (1)   IEDFs suggests that a significant fraction of the energetic N + ions also originate from sputterejected N atoms. Modeling studies of reactive HiPIMS discharges have confirmed the importance of the target as a source of reactive gas atoms in Ar/O2 discharges, 59 and we expect a similar effect for Ar/N2.
The gas-ion IEDF intensities in Figure 2 are significantly higher in Ar/N2 than in Kr/N2 discharges. In the former case, the total energy-integrated ion intensity J, with energies above 5 eV, is dominated by gas ions ( gas / metal = 2.78); while for Kr/N2 discharges, J is controlled by Hf + and Hf 2+ metal ions ( gas / metal = 0.54). Furthermore, the doubly-to-singly charged Hf ion intensity ratio is significantly higher in Ar/N2, Hf 2+ / Hf + = 1.7, than in Kr/N2 HiPIMS discharges, 3. This is consistent with previous results for non-reactive HiPIMS sputtering of Ti in Ar/N2 versus Kr/N2 discharges. 60 Hf + and Hf 2+ ion intensities begin to increase rapidly as gas-ion intensities decrease at t ~ 50 µs, in both Ar/N2 and Kr/N2 discharges, due to a strong gas rarefaction and the increased availability of Hf neutrals that leads to a decrease in the average plasma electron temperature. The latter effect is the result of Hf having a lower first ionization potential than those of the two noble gases and nitrogen (N2 and N) as shown in Table 2. Figure 3(a) reveals that measured gas-and metal-ion intensities are of comparable values at t between ~100 and 170 µs in Ar/N2 HiPIMS discharges; no clear metal-ion dominated mode is established. At t = 120 µs, gas / metal reaches a minimum value of ~1.8, with Hf 2+ ions contributing the major fraction, accounting for ~65% of the total metal-ion intensity. In distinct contrast, for Kr/N2 discharges, the ion flux shifts from being gas-dominated to metal-dominated at t ~ 55 µs (Figure 3(b)). At 100 µs, metal ions (Hf + + Hf 2+ ) represent ~90% of the total ion intensity with the major contribution, ~70%, being singly-charged Hf + ions. At longer times, t > 200 µs (100 µs after pulse termination), gas rarefaction has decreased due to gas refill and noble-gas ions again dominate the total ion flux in both Ar/N2 and Kr/N2 discharges.   Synchronizing the substrate bias to the metal-rich portion of the HiPIMS pulses during HfN film growth in Kr/N2 yields 002 peak intensities, I002 = 8.710 4 cps, which are approximately an the 002 peak FWHM intensity Γ2θ decreases by more than a factor of 2.3 to 0.30°. The synchronized-bias 002 peak position is 39.73°.  HfN(001) in-plane || and out-of-plane ⏊ x-ray coherence lengths, which are directly related to the sample mosaicity, a measure of crystalline quality, are obtained from the widths of the 002 diffracted intensity distributions perpendicular Δ ⏊ and parallel Δ || to the diffraction vector using the relationships 37

B. HfN film growth
and for which Γ2θ and Γω are the FWHM intensities, after correction for instrumental broadening, of the 002 Bragg peak in the ω and 2θ directions, respectively, and θ is the Bragg angle. Using the data presented in Figures 4 and 5 the use of primarily metal-ion, rather than gas-ion, irradiation during the growth of HfN (001) results in films with much lower mosaicity and higher crystalline quality. and in which E is the Ewald sphere radius, given by E = 1/λ. For a 113 reflection from a 001-oriented NaCl-structure sample, the in-plane || and out-of-plane ⏊ lattice parameters are obtained from the relationships || = √2/ || and ⏊ = 3/ ⏊ . 27 The results reveal that the lattice constants of   with the HR-RLM results. The dark speckles in the image are due to local strain-fields associated with point-defect clusters, which are more clearly observed in the lattice-resolution image in Figure   8(b). Figure 8(a) also reveals the presence of threading dislocations extending from the film/substrate interface along the growth direction; the associated strain fields appear as dark lines in the higher-resolution image, Figure 8(c). The presence of threading dislocations is expected due to the very large film/substrate lattice mismatch, 7.46%, which gives rise to a high density of misfit dislocations.

C. Film properties
Room-temperature resistivities ρ of epitaxial HfN(001) layers grown in Ar/N2 and Kr/N2 discharges with a continuous bias are 130 and 100 µΩ-cm, respectively; both significantly higher than for HfN grown in Kr/N2 with synchronized bias, 70 µΩ-cm. However, all three values are much lower than those reported for polycrystalline-HfN films deposited at 400 °C for which ρ ranges from 225 to 750 µΩ-cm. 65

IV. DISCUSSION
The results presented in section III demonstrate very-low-temperature epitaxial growth (Ts/Tm < 0.10 -0.12) of HfN on MgO(001) in the absence of applied substrate heating using reactive HiPIMS, with low-energy ion irradiation of the growing film, in both Ar/N2 and Kr/N2 discharges.
Mass-spectroscopy analyses show that the measured ion flux incident at the growing film during reactive HiPIMS contains both metal ions, Hf + and Hf 2+ , originating from the target, and gas ions, primarily N + , N2 + , and either Ar + or Kr + . Based on the energy-resolved measurements shown in Figure 2 and the applied substrate bias, the average incident energies for the metal ions are Hf + = 115 eV and Hf 2+ = 228 eV. For the gas ions, Ar + , N 2 + and Kr + = 100 eV, while N + = 112 eV. Bombardment with energetic ions and fast neutrals during HfN film growth results in overlapping shallow collision cascades which help to anneal out defects and desorb trapped noble-gas atoms. Low-energy bombardment also provides additional kinetic energy to the surface and near-surface region of the growing film, enhancing surface adatom mobilities 67 necessary for LTE.
Comparing the results for film growth with continuous vs. synchronized bias reveals that selecting the metal-rich portion of the ion flux during HiPIMS pulses is an efficient approach for improving the crystalline quality of LTE films. Sputtering Hf in Ar/N2 discharges with continuous bias yields an ion intensity at the film-growth surface which is dominated by gas ions -Ar + (26.8%), N2 + (12.2%), and N + (34.6%) -and doubly-charged metal Hf 2+ ions (16.6%), with a lower contribution from singly-charged Hf + ions (9.8%). In contrast, ion irradiation conditions during sputter deposition of HfN in mixed Kr/N2 discharges with continuous bias are quite different; the relative contribution of metal ions to the total ion intensity is five times larger (~50.9% for Hf + and 14.3% for Hf 2+ ), while the metal-charge-state ratio, Hf 2+ / Hf + , is approximately six times lower than values measured in Ar/N2 discharges.
The observed changes in the composition of the ion flux between Ar/N2 and Kr/N2 discharges are directly related to differences in the sputtered metal-atom first and second ionization potentials compared to the first-ionization energy of the noble gas (see Table 2). The primary ionization mechanism during HiPIMS is via inelastic collisions with energetic electrons. 68,69 The upper limit of the electron-energy distribution is, to a first approximation, determined by the first-ionization potential of the noble gas ( g 1 ), which dominate the initial part of the HiPIMS pulses ( Figure 3). Thus, the production rate of Hf 2+ and Hf + metal ions depends strongly on whether g 1 is larger or smaller than Hf 2 and Hf 1 , respectively. For Hf sputtered in Ar/N2, a significant fraction of the plasma electrons have an average energy Ee in the range Hf 2 < e < Ar 1 , which is too low to ionize Ar, but high enough to produce Hf 2+ ions. The situation is different for HiPIMS discharges in Kr/N2, for which ionization of Kr atoms at the beginning of the pulse tends to shift the electron-energy distribution to lower values which, in turn, results in a dramatic reduction in the Hf 2+ ion density, as shown in Figure 3. Since both Ar 1 and Kr 1 are much higher than Hf 1 , the effect on the Hf + production rate of switching noble gases is smaller. The shift in the electronenergy distribution between Ar/N2 and Kr/N2 discharges also explains the reduction of N + and N2 + densities in Kr/N2 discharges.
Changes in ion bombardment conditions at the growing film due to switching from Ar/N2 to Kr/N2 are reflected in the XRD results. HfN layers grown in Kr/N2 discharges exhibit a small, but clear, improvement in crystalline quality compared to films grown in Ar/N2. In-and out-ofplane x-ray coherence lengths in Kr/N2 layers ( || = 28 Å; ⏊ = 134 Å) are slightly higher (i.e., lower mosaicity) than those deposited in Ar/N2 ( || = 26 Å; ⏊ = 114 Å). In addition, the trapped oxygen concentration in Kr/N2 layers (1.8 at%) is considerably lower than for Ar/N2 films (3.6 at%) indicating higher film density.
Films grown in Kr/N2 discharges with a pulsed substrate bias synchronized to the metalion-rich portion of each HiPIMS pulse are subjected to intense Hf + bombardment (with greatly reduced noble-gas-ion irradiation) of the growing film. During the remaining portion of each pulse, Vs is only ~10 eV. Based upon TRIM 43 Monte-Carlo simulations, the average Hf + ion penetration depth is estimated to be ~14 Å, with overlapping collision cascades producing shallow, nearsurface dynamic atomic intermixing and noble-gas desorption, thus resulting in films with lower residual compressive stress.
The average projected ranges for Hf and N primary recoils are 10 and 8 Å due to Hf + bombardment, and 4 Å for both Hf and N recoils resulting from Kr + irradiation. Since the penetration depth for Kr + is ~6 Å, the film forming species -both the primary Hf + metal ions and the recoiled Hf lattice atoms -penetrate deeper into the near-surface region of the growing film than the inert-gas ions and can be directly incorporated into the lattice by filling residual vacancies arising due to the low growth temperature (Ts/Tm < 0.10). 70 Energetic metal-ion bombardment also enhances Hf adatom mobility and, hence, leads to increased crystalline quality, due to the mass match which gives rise to more effective energy transfer. XRD ω-2θ ( Figure 4) and ω-rocking curve ( Figure 5) results both clearly show that the epitaxial Kr/N2 synchronized-bias films have much lower mosaicity with a larger x-ray coherence length, by a factor of ~3, and lower roomtemperature resistivities than layers grown with continuous bias.
The reactive HiPIMS results presented here, Kr/N2 vs. Ar/N2 and continuous vs.
synchronized bias, clearly illustrate the importance of the nature of the low-energy species bombarding the film. Selecting plasma conditions favoring bombardment by singly-ionized filmforming Hf + species results in a significant increase in the crystalline quality of LTE Hf layers. The strategy is general and can easily be transferred to the growth of other TM-nitride (as well as carbide and oxide) systems and metallic films. Each material system will require careful experimental design of HiPIMS pulse parameters (maximum target current density and duty cycle), time-dependent ion intensities and substrate bias (Vs amplitude and time), and choice of noble gas for a given target material in order to obtain the maximum metal-ion/noble-gas-ion ratio and the lowest doubly-to singly-charged metal-ion ratio incident at the film-growth surface.

IV. CONCLUSIONS
The results presented in Section III demonstrates very-low-temperature (Ts/Tm < 0.10) epitaxial growth of HfN thin films on MgO(001) using reactive HiPIMS with the substrate bias synchronized to the metal-ion-rich portion of the discharge. The HfN/MgO system was chosen since HfN has the highest melting point, 3310 °C, 31 of all the TM nitrides and the film/substrate lattice mismatch is large, 7.46%, both providing challenges for LTE growth.
The highest-crystalline-quality LTE HfN films are achieved in Kr/N2 discharges using a substrate bias synchronized with the metal-rich portion of each HiPIMS pulse, which also decreases plasma heating such that the film growth temperature is reduced to < 70 °C (Ts increases essentially linearly with deposition time to reach a maximum of value of 70 °C at 40 min into 60 min deposition runs.). The synchronized bias results in increased Hf + self-ion irradiation and reduced inert gas-ion bombardment, both leading to a significantly decreased film mosaicity, and correspondingly lower resistivity values. The layers grow with a cube-on-cube orientation, Layers grown in Ar/N2 discharges under continuous bias (Vs = 100 V) reach a maximum substrate temperature of 120 °C (Ts/Tm = 0.12) due to plasma heating and have lower crystalline quality than those grown in Kr/N2 with continuous bias. This is manifested in higher mosaicity, trapped noble-gas concentration, and porosity. Switching to a Kr/N2 discharge, with continuous bias, suppresses the formation of high-energy gas ions, particularly N + .