Enhanced Conductivity and Microstructure in Highly Textured TiN1-x /c-Al2O3 Thin Films.

Titanium nitride thin films are used as an electrode material in superconducting (SC) applications and in oxide electronics. By controlling the defect density in the TiN thin film, the electrical properties of the film can achieve low resistivities and a high critical temperature (T c) close to bulk values. Generally, low defect densities are achieved by stoichiometric growth and a low grain boundary density. Due to the low lattice mismatch of 0.7%, the best performing TiN layers are grown epitaxially on MgO substrates. Here, we report for the first time a T c of 4.9 K for ultrathin (23 nm), highly textured (111), and stoichiometric TiN films grown on 8.75% lattice mismatch c-cut Al2O3 (sapphire) substrates. We demonstrate that with the increasing nitrogen deficiency, the (111) lattice constant increases, which is accompanied by a decrease in T c. For highly N deficient TiN thin films, no superconductivity could be observed. In addition, a dissociation of grain boundaries (GBs) by the emission of stacking faults could be observed, indicating a combination of two sources for electron scattering defects in the system: (a) volume defects created by nitrogen deficiency and (b) defects created by the presence of GBs. For all samples, the average grain boundary distance is kept constant by a miscut of the c-cut sapphire substrate, which allows us to distinguish the effect of nitrogen deficiency and grain boundary density. These properties and surface roughness govern the electrical performance of the films and influence the compatibility as an electrode material in the respective application. This study aims to provide detailed and scale-bridging insights into the structural and microstructural response to nitrogen deficiency in the c-Al2O3/TiN system, as it is a promising candidate for applications in state-of-the-art systems such as oxide electronic thin film stacks or SC applications.

Introduction

Titanium nitride (TiN) is an intensely studied electrode material for oxide and superconducting (SC) electronics. For example, the successful combination of TiN and dielectric materials has been demonstrated on hafnium oxide (HfO2) in complementary metal-oxide-semiconductor field-effect transistors and next-generation Hafnia-based resistive random access memories (RRAM).[1−3] Ferroelectric RAM (FeRAM), which requires symmetric electrodes, can also be realized with TiN as a bottom and top electrode.[4−8] In addition, TiN is a strong candidate as an electrode material in Josephson junctions and SC coplanar waveguides[9] with bulk transition temperature (Tc) of up to 6.0 K[10−12] and exhibits low losses at microwave frequencies.[13] In the search for the highest possible Tc, extensive research including machine learning-based growth parameter prediction is applied.[14] TiN has also gained significant interest in plasmonic and metamaterial applications in the visible and near-infrared wavelength range.[15] The fabrication of TiN electrodes is generally based on thin film deposition by either physical vapor deposition techniques, such as radio frequency (RF) sputtering, chemical vapor deposition (CVD) techniques like inorganic CVD or metal organic CVD (MOCVD), or by sol–gel based approaches.

These films exhibit resistivity values ranging from 14 to 1000 μΩcm for film thicknesses above 40 nm.[9,16−18] Sol–gel based approaches of depositing a TiO2 film with subsequent nitridation achieved 40 nm TiN films with a 1510 μΩcm resistivity.[19] In most cases, these TiN thin films exhibit a polycrystalline microstructure with a columnar grain growth[20−22] and grain boundaries (GBs) that run along the entire thin film thickness.[23] This microstructure is commonly discussed as the source for reduced conductivity in TiN thin films.[24] The use of highly epitaxial TiN thin films would be advantageous for thin film electronic applications.[25] Epitaxial or single-orientation growth of TiN has been previously reported in the literature using DC reactive sputtering,[26] pulse laser deposition (PLD),[27] atomic layer deposition (ALD),[28,29] and reactive molecular beam epitaxy (RMBE) methods.[9,30] High-quality epitaxy leads to very smooth thin film surfaces and low-defect densities. In the work by Krockenberger et al., TiN grown on MgO substrates by RMBE yields Tc of up to 5.25 K and resistivity in the range of 10–15 μΩcm at 40 nm film thickness.[9] The highest surface quality with a Tc of 5.25 K was obtained for thin films grown at 720 °C. Superconductivity was found to disappear at lower growth temperatures due to a decrease in the nitrogen content of the thin films. As demonstrated by Torgovkin et al., this can be reversed by annealing deficient TiN1– layers in a nitrogen atmosphere.[12] Chen et al. achieved RMBE-based quasi-epitaxial growth[31] TiN thin films on Si and Al2O3 substrates at elevated temperatures, resulting in Tc as high as 5.9 K for 130 nm thick films[30] (closest to bulk TiN, Tc = 6.0 K[11]). At the same time, the presence of nitrogen vacancies and GBs in TiN for RRAM applications has been reported to be beneficial for oxygen incorporation at the HfO2/TiN interface during resistive switching processes.[32]

To our knowledge, highest surface quality TiN (001) thin films were reported when grown on MgO (001) substrates due to their low lattice mismatch (0.7%). However, Narayan et al. reported the epitaxial growth of TiN (001) on Si(001) substrates by PLD, which was suggested to occur via a domain matching epitaxy.[33] This is unexpected because Si (001) as a substrate for TiN (001) thin films should not yield high epitaxial growth due the high lattice mismatch of 25%. Here, the epitaxy is induced by the common structure, which is composed of four TiN (001) unit cells matching on three Si (001) unit cells, resulting in a lattice mismatch of less than 4%. Reisinger et al. identified a recombination in the Si (001)/TiN (001) system with a “5-on-4-cube-on-cube” reconstruction (lattice mismatch 2.4%) for lower deposition temperatures.[34,35] For 100 nm thick films, Talyansky et al. reported epitaxial growth of TiN (111) film on c-cut sapphire substrates with resistivity as low as 13 μΩcm and surface roughness (RMS) of 0.6–1 nm.[36] A high lattice misfit of 8.75% was calculated considering a rhomb-on-rhomb growth on the c-cut sapphire substrate.

Although quasi epitaxial growth of TiN thin films was reported for both Silicon and c-cut sapphire substrates,[31,37] there is a lack of experimental evidence to clarify the specific nature and properties of the grown microstructures. The high relevance of microstructural effects in TiN is underpinned by studies reporting a total loss of SC for films thinner than 40 nm.[12,16] Most studies are limited by being confined to macroscopic characterization methods like resistivity and X-ray diffraction (XRD) techniques, which allow for functional assessment of the as-grown films, but provide an incomplete picture of the deposited thin films. This study aims at broadening the understanding of defect sources in TiN thin films, namely, nitrogen deficiency and grain boundary density, by combining macroscopic and microscopic investigations down to the atomic level.

Results and Discussion

TiN1–—Nitrogen Deficiency

Figure a shows the temperature-dependent resistivity measurements of the ∼25 nm thick TiN thin films. Films with increasing nitrogen deficiency are presented in the sample series number #1 to #5. The N deficiency was introduced by (a) lowering the substrate temperature and (b) increasing the Ti evaporation rate; see the Experimental Section for further details. Figure b shows the low-temperature regime of Figure a to reveal the transition temperature to the SC state. Only sample #5 does not show a critical temperature (Tc) above the lowest measurement temperature of 4.2 K. To investigate the influence of the growth temperature in more detail, out-of-plane d-spacing (111) values were retrieved from XRD 2θ/ω-scans, as shown in Figure S1. The TiN thin film with the largest d-spacing (2.451 Å, #1) shows the highest Tc value (4.9 K), as shown in Figure c. Tc drops to 4.4 K at a d-spacing of 2.440 Å (#4). The film with the lowest d-spacing (2.438 Å, #5) shows no SC. The d-spacings found for the different deposition conditions can be compared to literature data of known stoichiometry, here (111) TiN0.71[38] and TiN0.9[39] (dashed lines, Figure c,d), which allows for the estimation of the nitrogen deficiency. The most stoichiometric film yields a high Tc, whereas the most nitrogen deficient sample shows no Tc. Changing nitrogen deficiency directly translates to the formation of nitrogen vacancies in the crystal lattice.[11] To investigate the defect density of the TiN thin film, the residual resistance ratio (RRR)[12] was calculated (Figure d). RRR is defined as the quotient of the resistivity at room temperature (RT) and at 10 K [RRR = ρ(300 K)/ρ(10 K)]. The conductivity at 10 K is an indirect measure of defect density because the (thermal) phonon scattering contribution is greatly reduced at this temperature.[40] In the present system, this defect density is primarily governed by N deficiency shown by a decrease in RRR with smaller d-spacing.

Figure 1

With an increase in growth temperature, the room temperature resistivity ρ(300 K) decreases (a) while Tc increases (b), which is associated with an increase in the residual resistance ratio (RRR = ρ(300 K)/ρ(10 K)), and (c) increase in the TiN1– (111) d-spacing (d); dashed lines indicate literature references for N-deficient and stoichiometric TiN1– compositions.

Thin Film Microstructure

The doubling of resistivity cannot be only explained by the volume defects introduced by the increased nitrogen deficiency. Another source of defects that decreases electrical conductivity in a material is the presence of GBs. A first indication for the presence of GBs would be the existence of more than one in-plane orientation of the (111) oriented TiN grains. The out-of-plane texture that was extracted from the 2θ/ω scans is TiN (111)||Al2O3 (001). The in-plane alignment of the TiN grains can be achieved by azimuthal or ϕ-scan. The results are depicted in Figure S1 and show that TiN {111} peaks are separated by 60° and are offset by ± 30° to the Al2O3 {1̅02} peaks. This indicates the presence of two in-plane locked crystal orientations or domains, which are 60° in-plane rotated to each other. This was also previously reported for epitaxial TiN (111) thin films on c-cut Al2O3 and Si substrates[36,41] where the equivalent TiN (011̅) and (101̅) crystal orientations align along Al2O3 (100). The full epitaxial relationship can thus be summarized as follows: TiN (111)||Al2O3 (001) out-of-plane and TiN(011̅),(101̅)||Al2O3(100) in-plane.

With an electron mean free path (MFP) of ∼100 nm for TiN thin films as estimated by Krockenberger et al.[9] for a 40 nm TiN (001) film on MgO (001), a supposedly grain boundary free TiN film was grown epitaxially on the low lattice mismatch substrate. In the present samples, a grain size distribution analysis provides insights into the contribution of these boundaries in the resistivity.[10] The contribution of these GBs is higher if the distance between the boundaries is the same length scale as the MFP.[42] A common method to image GBs is backscatter electron (BSE) imaging, which has been applied to the present sample series and is displayed in Figure .

Figure 2

(a) Twin grain boundary of two 60° in-plane rotated TiN grains (red and green). (b–e) Top view BSE maps of the SC TiN thin films showing no visible GBs in contrast to the non-SC TiN thin film (f). False color is applied to enhance contrast.

Figure a shows a model for two 60° in-plane rotated TiN grains with GBs indicated by dashed lines. The GBs that are formed can be understood as twins (with boundaries of the (1̅21̅)/(2̅11)/(112̅) type). Figure b–f shows representative top-view BSE images of the TiN films. BSE contrast is susceptible to the crystal quality, grain orientation, and RMS. With the film fully aligned in the (111) orientation, the observed contrast should only indicate GBs and roughness. GBs reduce the number of backscattered electrons due to the presence of defects in the crystal lattice, thus giving a detailed view of the thin film microstructure. For the sample with the highest nitrogen deficiency (Figure f), GBs are visible as meandering dark lines. For the samples grown at 620, 660, and 690 °C and the stoichiometric sample grown at 750 °C, no distinct GBs can be observed. Dark spots observed for the low growth temperature samples (Figure b–d) can be correlated with the increase in RMS as indicated by the X-ray reflectometry (XRR) results (Figure S3a) down to a RMS of 1.5 nm for the films grown at 750 °C. The meandering of the boundaries in the sample shown in Figure f does not indicate a random behavior, but it follows a set of three dominating orientations. They correspond to the twin boundaries marked by dashed lines shown in Figure a. Consequently, this indicates the presence of a high number of twin boundaries for the sample shown in Figure f.

With BSE imaging, no GBs could be observed for the samples grown at lower temperatures. However, XRD ϕ-scans indicate that there are two orientations of the TiN grains present in each sample. To investigate the microstructure of all samples, ion channeling contrast (iCC) imaging[43] was performed, as shown in Figure . The imaging conditions were selected to give the strongest iCC between the two orientations. A sample tilt of 54° from the 30 keV Ga ion beam allowed the ions to penetrate one TiN orientation at the prominent (001) zone axis, while the other grain orientation would be oriented at the same time along the (1̅22) zone axis. The discrepancy between the index of the zone axis directly correlates with the secondary electron (SE) yield and thus the observed contrast in the images (prominent {001} zone axis, low SE yield; {1̅22} zone axis, high SE yield).[44] With the high contrast achieved in an otherwise featureless surface, this method proved to be ideal to identify grain sizes and microstructures in this system, which can also be optimized by a ϕ/azimuth rotation series. In the iCC images, large, over 100 nm sized grains can be observed for all samples.

Figure 3

(a) Ion channeling contrast (iCC) imaging geometry. (b–f) iCC images of the samples revealing an elongated grain structure with comparable grain sizes for all samples. Red lines (e,f) showing the orientation of the cross sections cut in the subsequent focused ion beam lamella preparation, perpendicular to the observed GBs.

The anisotropy of the grain size can be attributed to a miscut of the c-cut sapphire substrate, which was used for all samples. Cuccureddu et al. found that even for nominally 0° miscut “c-cut” samples, there are still surface steps present that are homogeneously distributed on the surface.[45] For a 0.15° miscut c-Al2O3 sample, the step width was found to be 120 nm. In addition, annealing leads to a “bunching” of the surface steps if they are held at elevated temperatures, meaning that each resulting step height is a multiple of c/6 = 2.2 Å of the Al2O3c-axis unit cell parameter (c = 13 Å). The step width and height converge to uniformity.[46] The (111) d-spacing value of TiN is 2.439 Å, which is a ∼10% deviation of the c/6 Al2O3 = 2.2 Å (see Figure S4) lattice constant along growth direction. A bunched surface step of 5*c/6 Al2O3 would introduce an overall mismatch of 56%, whereas a surface step of 9*c/6 Al2O3 would be a full TiN (111) d-spacing. Depending on the step height, the bonding configuration for the first adatom differ[47] and might trigger the respective TiN in-plane rotation (the out-of-plane direction fixed by the rhomb-on-rhomb epitaxy). By this high degree of ordering, the observed consistent alternation of in-plane orientations can be understood as an effect of uniform step widths and heights on the seeding conditions. The investigated TiN films have been grown on the same batch of substrates, which were heated during deposition and which will therefore yield to similar surface steps, as indicated by the uniform grain size distribution throughout the sample series. The four-point electrode geometry used in the resistivity measurement was not aligned with the observed elongation. The anisotropy, induced by the substrate miscut, dictates one crucial source for defects in the thin film: the grain boundary density. It is virtually the same for all samples and thus will have the same contribution for all samples. For the following high-resolution scanning transmission electron microscopy (HR-STEM) analysis, cross-sectional FIB lamellae have been cut from the samples grown at 750 °C. Cutting was performed perpendicular to the elongated grain structures (indicated in Figure e,f) to ensure that the GBs are aligned along the observation direction of the STEM.

Atomic Structure of (Dissociated) Twin GBs

Figure shows representative cross sections of GBs for the samples grown at 750 °C. The sets of neighboring grains are all oriented with (111) pointing upward along growth direction, which confirms the observations of the 2θ/ω scans (Figure ). Both of the imaged GBs can be indexed as twin boundaries of the {1̅21̅} type. For the sample grown at a Ti evaporation rate of 0.2 Å/s (Figure a), atomic resolution high-angle annular dark-field (HAADF)-STEM imaging shows that the defects present at the boundary are located at the boundary itself, where the joining of the lattices of the grains form periodic recombination sites along the boundary. This is in stark contrast to the boundary of the sample grown at 0.6 Å/s (Figure b). Here, an extended defect structure indicated by split atomic columns (indicated by blue arrows) can be observed. These variations in contrast can be attributed to stacking faults along the (111) axis for the left grain shown in Figure b. Twin boundary dissociation by stacking fault emission is observed for other systems[48,49] as well as for TiN, but up until this point, only under mechanical deformation.[50] Strain induced by the lattice mismatch in the present substrate/thin film stack in combination with the growth conditions of the RMBE deposition process was reported as the cause for the occurrence of stacking faults in the Si/GaAs system.[51,52] Here, the volume affected by stacking faults extends through the full film thickness up to the surface of the film, where it gives rise to the grain boundary contrast found in the BSE images (Figure f), as they pose a sample volume of high defect concentration, reducing the backscatter yield.

Figure 4

High-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) images of {1̅21̅} twin boundaries in the (a) SC and (b) non-SC TiN thin films. The grain boundary in (b) shows a high density of stacking faults in the atomic resolution image, indicated by blue arrows.

Strongly confined geometrical conditions for the microstructure developed during the thin film growth process have a direct impact on the physical properties. A high fraction of high symmetry boundaries are generally associated with reduced resistance in conductors.[41,53] This is the case in our study because the microstructure of the TiN films contains almost solely twin boundaries. If the average grain size is in the range of the MFP, the grain boundary can be considered as a main contributor to electron scattering, increasing the resistivity.[24,54] Due to the equal average grain size in the present sample series, the impact on resistivity should therefore be equivalent. Nonetheless, the defects associated with the extended GBs play a crucial role in the contribution to the resistivity. Extended scattering regions lead to higher resistivity, which is described as a reduction of “grain boundary transparency” in terms of electrical conduction across a grain boundary.[55−57] This is directly evident in the increase of conductivity (Figure a). There is also a higher concentration of oxygen[12] and other contaminants[58] expected at the boundaries, which could negatively affect electronic transport properties.

In their application as bottom electrodes for valence change memory RRAM devices, the ability of the electrodes to incorporate oxygen plays a fundamental role in the resistive switching process.[59−63] Nitrogen-deficient TiN thin films lower the energy necessary to incorporate oxygen ions in the electrode material.[32] As GBs form a localized high density of defects, especially for the dissociated GBs observed in this work, they would have the potential to locally facilitate electroforming and switching processes.[3] As the microstructure of the TiN electrode also defines the growth conditions—or conditions for crystallization by annealing for ALD grown films[64]—GBs present a predefined seeding location and thus will have a direct influence on the microstructure observed in subsequent layers of a thin film stack.[3] This influence can be further addressed by nanoscale orientation mapping techniques like automated crystal orientation mapping in a plan-view geometry.[65]

Conclusions

The impact of nitrogen deficiency on the room and low temperature conductivity as well as on the microstructure and type of evolving GBs was studied for the c-Al2O3/TiN system. By combining XRD 2θ/ω, ϕ-scans, BSE, iCC imaging in a multibeam FIB system and atomic resolution HR-STEM, the microstructure, grain size, and type of GBs were determined. The ∼25 nm thick TiN films exhibit a homogeneous (111) out-of-plane texture, with a set of two in-plane rotations, separated by 60°. This texture is defined by the quasi epitaxial growth of (111) TiN on c-Al2O3 where the large lattice mismatch of 8.75% is compensated by a rhomb-on-rhomb configuration. A fixed grain size as observed for the present sample series allowed us to separate the grain boundary density, as one contributor to defect density, from the effect of increasing nitrogen deficiency. The observed grain boundary density and the anisotropy of the grain size are directly related to the surface step density found on nominally 0° c-cut sapphire substrates. These steps also define the initial seeding conditions of the TiN grains and result in the alternating in-plane rotations observed in the thin film microstructure. Increased nitrogen deficiency has three effects on the film properties: (a) the loss of superconductivity, (b) an increase in resistivity, and (c) grain boundary dissociation by stacking fault emission. Nonetheless, the same morphological quality is achieved. Stoichiometric, ultrathin 25 nm films are achieved for the first time with a room-temperature conductivity of 10 μΩcm and a Tc of 4.9 K with a RMS of ∼1.5 nm. The observed properties match best in class SC epitaxially grown TiN films on MgO substrates.

Experimental Section

BSE imaging was performed on a Tescan Mira3 XMH at 15 kV, 320 μs dwell time and 434 pA specimen current using a Deben 4 quadrant detector. Ion beam channeling contrast (iCC) imaging and cross-sectional focused ion beam (FIB) sample preparation for scanning transmission electron microscopy (STEM) were performed on a JEOL JIB-4600F, and iCC imaging was conducted at 30 keV acceleration voltage. Atomic resolution high-angle annular dark-field (HAADF) STEM imaging was performed on a JEOL JEM-ARM200F operated at 200 kV with a semiconvergence angle of 25 mrad.

In this study, an electron-beam evaporation setup was used to evaporate elemental sources of titanium (99.99%, Chempur). Nitrogen radicals obtained using a RF plasma source (Oxford Applied Research, HD25) were used as the source of nitridation in our experiments. Mass flow controllers (MFC) allowed us to control the flow rates of nitrogen (99.995% purity) to the radical sources. The growth temperature was increased from 620 to 750 °C with a constant Ti evaporation rate of 0.2 Å/s, a nitrogen flow of 1.0 sccm, and a RF power of 200 W. For a sample with higher nitrogen deficiency, a Ti evaporation rate of 0.6 Å/s has been selected. Initial selection of the growth parameters included aspects like minimization of nitrogen deficiency[26] stoichiometry, RMS, and grain size.[17,66]

XRD and XRR were performed on a Rigaku (SmartLab) X-ray diffractometer by using Cu Kα1 (λ = 1.54057 Å) at 45/170 kV/mA and a parallel beam geometry and on a Seifert PTS by using Cu Kα1,2 (λ = 1.54057 Å) at 40/40 kV/mA and a parallel beam geometry with a graphite monochromator.

A Keithley 6221 model current source and Keithley 2182A model nanovoltmeter were used for the presented four-probe resistivity measurements, while the temperature was measured using a calibrated Lakeshore silicon diode sensor (model DT-670D-SD), placed in contact with the sample stage.