Highly Plasmonic Titanium Nitride by Room-Temperature Sputtering

Titanium nitride (TiN) has recently emerged as an attractive alternative material for plasmonics. However, the typical high-temperature deposition of plasmonic TiN using either sputtering or atomic layer deposition has greatly limited its potential applications and prevented its integration into existing CMOS device architectures. Here, we demonstrate highly plasmonic TiN thin films and nanostructures by a room-temperature, low-power, and bias-free reactive sputtering process. We investigate the optical properties of the TiN films and their dependence on the sputtering conditions and substrate materials. We find that our TiN possesses one of the largest negative values of the real part of the dielectric function as compared to all other plasmonic TiN films reported to date. Two-dimensional periodic arrays of TiN nanodisks are then fabricated, from which we validate that strong plasmonic resonances are supported. Our room-temperature deposition process can allow for fabricating complex plasmonic TiN nanostructures and be integrated into the fabrication of existing CMOS-based photonic devices to enhance their performance and functionalities.


Results and Discussion
show respectively the real (ε 1 ) and imaginary (ε 2 ) part of the extracted dielectric functions of TiN films (thickness ~50 nm) deposited on various substrate surfaces at Ar:N 2 = 80%:20% (except for the TiN films on quartz deposited at Ar:N 2 = 95%:5%) using Drude-Lorentz model. All TiN films are clearly plasmonic in the spectral region of interest. In particular, the large negative values of ε 1 for TiN deposited on hafnium dioxide (HfO 2 ), silicon (Si), and quartz indicate that these films are highly plasmonic. Both of their ε 1 and ε 2 exhibit similar characteristics, and the variation of their λ ps is less than 20 nm (λ ps = 480, 470, and 460 nm for TiN deposited on HfO 2 , Si, and quartz, respectively), showing that our room-temperature sputtered TiN films have a weak dependence for these substrate surfaces as those prepared by ALD 22 . Note that λ ps of these films occurs at wavelengths shorter than that of the ALD plasmonic TiN 16,22 , therefore enabling a broader plasmonic response in the visible region. TiN films on poly(methyl methacrylate) (PMMA) possess a quite distinct dielectric function from others and are less plasmonic, which may have been caused by the surface roughness of the PMMA films induced by the ion bombardment during the sputtering process. As can be seen in the atomic force microscopy (AFM) images (see Fig. 2), the TiN films deposited on PMMA are fairly rough, possessing a root-mean-square roughness value of 26.4 nm, much greater than that for the films deposited on HfO 2 (0.953 nm) and Si (0.888 nm) surfaces. It is expected that the optical properties of TiN films on PMMA are different since their roughness is essentially on the same order of the film thickness. Nevertheless, our results show that the room-temperature sputtering process can be employed to fabricate plasmonic TiN on unconventional polymeric or elastomeric substrate for flexible plasmonics and nanophotonics applications 23 , which is not possible by either high-temperature sputtering 2,3 or low-temperature ALD 16 . We plot for comparison in Fig. 1(a,b) the ε 1 and ε 2 of TiN films deposited by high-temperature sputtering, low-temperature ALD, and high-temperature epitaxial growth, in dashed lines denoted as ref 1 3 , ref 2 16 , and ref 3 5 , respectively. Epitaxial TiN films grown on lattice-matched substrate to our knowledge possess the largest negative value of ε 1 among all plasmonic TiN films reported so far due to their high crystalline quality and low roughness, although they need to be grown at a fairly high temperature (750 °C). Our room-temperature sputtered TiN films are apparently comparable to the epitaxial TiN and superior to films prepared by both high-temperature sputtering and low-temperature ALD in plasmonic character. We believe that the strong plasmonic character of our room-temperature sputtered TiN films is enabled by the close proximity of our samples to the plasma during deposition providing the necessary high energy reactive ions and activated species in lieu of substrate heating or biasing 24 . Note that the observed strong plasmonic character of our TiN is in sharp contrast to the dielectric properties of titanium oxynitride (TiO x N y ) 25 , which usually exhibits double epsilon-near-zero (ENZ) characteristics without substrate heating and special chamber preparation. We also plot in the same figures the optical data of gold (Au) reported by Johnson and Christy 26 . Despite its several disadvantages for plasmonic applications 4 , Au still has better plasmonic character (more negative ε 1 ) and a smaller loss (smaller ε 2 ) compared to TiN, which can be further verified by examining the general figure of merits (FOMs) for localized surface plasmonic resonance (LSPR) defined as − ε 1 /ε 2 27 , as shown in Fig. 1(c). However, for plasmonic induced hot-carrier generation in photothermal or photoelectric applications 28,29 where plasmons are resonantly excited by incident light and subsequently decay nonradiatively into hot carriers on an ultrafast time scale, it is desirable to have plasmonic materials possessing not only a large negative ε 1 but also a large ε 2 value. Our TiN films are clearly most suitable for hot-carrier applications in comparison with Au and all the other TiN films, allowing for both strong plasmonic coupling (due to larger negative ε 1 ) and fast plasmon decay (due to larger positive ε 2 ). In Fig. 1(d) we show the x-ray diffraction (XRD) spectra of TiN films sputtered on different substrate surfaces. All TiN films possess essentially the same crystalline structure, exhibiting a predominant peak at ~42.5° along with two weak peaks at ~36.2° and ~61.6°, corresponding to the (200), (111), and (220) orientation, respectively. The observation of these peaks in XRD suggests a polycrystalline nature of our room-temperature sputtered TiN films. Their grain size (L) in the (200) orientation estimated using the Scherrer's formula 30 regardless of substrate surface is comparable to that for the reactive sputtered TiN films reported elsewhere 19,31 : L = 7.33, 7.89, 9.76, and 7.89 nm for the films deposited on HfO 2 , Si, quartz, and PMMA, respectively. In Fig. 3(a) we show the ε 1 (solid lines) and ε 2 (dashed lines) of the TiN films (thickness ~50 nm) deposited on Si at various Ar:N 2 gas flow ratios (95%:5%, 85%:15%, 80%:20%, 70%:30%, and 60%:40%). As can be seen, the optical properties of the TiN films can be tuned by adjusting the Ar:N 2 ratio in the sputtering process. The value of ε 1 becomes more negative for the films deposited at higher N 2 flow and is most negative when Ar:N 2 reaches 80%:20%, but further increasing the N 2 content in the reactive gas instead makes TiN films less plasmonic, essentially similar to the trend observed in the earlier work 2 . These findings are also in good agreement with high-temperature sputtered TiN films on Si and magnesium oxide (MgO) substrate, but are different from those on sapphire as well as TiN prepared in a room-temperature sputtering with a substrate bias voltage 20 . It could therefore be inferred that in addition to the Ar:N 2 ratio, the plasmonic character of sputtered TiN films is strongly influenced by numerous other processing parameters including substrate material, deposition temperature, external biasing, and so on. In fact, we find in our experiments that the Ar:N 2 = 80%:20% TiN films possess the most negative value of ε 1 as compared to those deposited at other gas flow ratios for all substrate surfaces but quartz, for which the plasmonic character of the TiN films is optimized when Ar:N 2 is 95%:5%. We also find that the λ ps of our TiN films (see the inset to Fig. 3(a)) redshifts from 450 nm for the Ar:N 2 = 95%:5% films, to 470 and 490 nm for the Ar:N 2 = 80%:20% and the Ar:N 2 = 60%:40% films, respectively. The observed redshift of λ ps suggests that increasing the N 2 flow leads to stronger dielectric screening of the free carriers in Ti 3d bands resulting from the interband transitions from N 2p to Ti 3d bands 32 . Moreover, the λ ps value has been shown to be a reliable measure of the stoichiometry of TiN: λ ps is ~468 nm (or screened plasma energy E ps is ~2.65 eV) when the ratio of N and Ti in TiN films is near unity 33 . This suggests that our Ar:N 2 = 80%:20% films are most likely to be stoichiometric, although others should still have close stoichiometry due to their comparable λ ps values. In Fig. 3(b) we compare the corresponding XRD spectra of these TiN films, revealing that the Ar:N 2 gas flow www.nature.com/scientificreports www.nature.com/scientificreports/ ratio has an impact on their crystalline structure. The Ar:N 2 = 95%:5% films exhibit a predominant peak for the (200) orientation along with two small peaks for (111) and (220) orientations. The (111) peak intensity gradually increases with increased N 2 flow, although both Ar:N 2 = 85%:15% and Ar:N 2 = 80%:20% films still show a (200) preferred orientation. Further increasing the N 2 flow dramatically changes the crystalline structure of TiN, making the Ar:N 2 = 70%:30% and Ar:N 2 = 60%:40% films preferentially grown in the (111) orientation. These observations seem to suggest that our TiN films in the (200) preferred orientation possess better plasmonic character than (111) oriented films; however, other factors such as oxygen impurity content 20 also need to be considered to fully elucidate the relationship between their structural properties and the corresponding plasmonic response.
We then investigate the thickness dependence of the dielectric function and crystalline structure of the TiN films deposited on Si at Ar:N 2 = 80%:20%, as shown in Fig. 3(c,d), respectively. All TiN films show desirable plasmonic character which appears to be enhanced for thicker films, and they all exhibit clear crystalline order in the (200) preferred orientation. We do not observe any growth defects nor film cracks induced by the strain relaxation due to the lattice mismatch between TiN and Si when increasing the film thickness in our experiments. Recent studies 34,35 have shown that the thickness dependent optical properties of plasmonic thin films arise from changes in surface-to-volume ratio, grain boundary, and carrier scattering from substrate surface as thickness varies. In particular, when films are within the ultrathin regime (thickness ≤10 nm), their plasmonic character significantly degrades due to island formation 36 , low mobility interface layers 37 or quantum confinement effects 38 , and special thin film preparation methods such as high-temperature epitaxial growth as well as substrate treatment are thus required. However, in most of the practical plasmonic structures and devices, the thicknesses of their constituent plasmonic layers are within the thickness range of the present study (i.e., thickness ~20-100 nm). Our room-temperature sputtered TiN films can therefore be employed for a wide variety of practical applications for plasmonics due to their high plasmonic character. It is also worth noting that Sugavaneshwar et al. 39 recently claimed the realization of the TiN films with the best plasmonic character reported to date by pulsed laser deposition (PLD). Despite the demonstrated high plasmonic character (see the gray solid and dashed lines in Fig. 3(c) for ε 1 and ε 2 , respectively), their TiN films are in fact much thicker (thickness = 180 nm) than those reported in literatures. The plasmonic quality of their thinner films (thickness ~30-50 nm) remains to be verified to allow for a fair comparison.
To verify the chemical compositions of our room-temperature sputtered TiN films, we perform x-ray photoelectron spectroscopy (XPS) measurements on the Ar:N 2 = 80%:20% films deposited on Si substrate at three different depths by in situ Ar ion milling: at the film surface, in the middle of the film, and near the substrate. The measured XPS spectra for Ti 2p, N 1 s, O 1 s, and C 1 s are plotted in Fig. 4(a-d), respectively. At the film surface, the Ti 2p doublet lines (i.e., 2p 3/2 and 2p 1/2 ) at the binding energies of ~455 and ~461 eV along with the nitrogen www.nature.com/scientificreports www.nature.com/scientificreports/ peak at 397.2 eV confirm the characteristic TiN phase 40 . The surface oxidation of the TiN films results in the formation of titanium dioxide (TiO 2 ), manifesting as the other two peaks at 458.4 and 464.1 eV in the Ti 2p spectrum as well as the oxygen peak at 530 eV. The expected carbon signal detected at 284.6 eV accompanied by a small peak at 288.5 eV originates from the carbon contamination of the surface of the TiN films, as they are fully exposed to the atmosphere after growth. Moreover, the filling of the valley between the Ti 2p 3/2 and 2p 1/2 transitions near the binding energy of 457 eV indicates the existence of an intermediate chemical state between TiO 2 and TiN, i.e., TiO x N y , which also appears as the shoulder at 395.7 and 531.4 eV in the N 1 s and O 1 s spectra, respectively. After etching halfway through the films, the characteristics of the TiN phase become clearly dominant while the features of the surface oxidation and carbon contamination are reduced, including the significant enhancement of the Ti 2p signals as well as the N 1 s peak for TiN, the strong suppression of both the TiO 2 peaks in the Ti 2p spectrum and the oxygen peak in the O 1 s spectrum, and the complete elimination of the two carbon peaks in the C 1 s spectrum. The signals representing the TiO x N y state are also much weaker, suggesting that the titanium oxynitride forms mostly near the film surface. By comparing the measured Ti 2p spectra including the spectral line shape and peak positions to those reported in the previous XP spectroscopic studies of reactive sputtered TiN films 41 , we find that the bulk of our Ar:N 2 = 80%:20% TiN films is likely to possess high purity with a small amount of oxygen impurity. This is evidenced by the elemental compositions of the films (see Table 1) extracted from the XPS spectra: the oxygen content is ~10% in the bulk despite a higher concentration of oxygen (~26%) probed at the surface. The XPS spectra near the substrate reproduce all of the important characteristics observed in the bulk with expected reduced intensity and a strong Si signal, indicating uniform chemical compositions of  www.nature.com/scientificreports www.nature.com/scientificreports/ the TiN films and a robust sputtering deposition process. These XPS measurement results further confirm that our room-temperature sputtered TiN differs strongly from TiO x N y (even for the TiO x N y films deposited at high temperature after careful chamber preparation) 25 , showing much stronger TiN characteristic signals and much less distinct features for TiO x N y .
In Fig. 5 we show the unpolarized reflectance and transmittance spectra of 45 nm-thick TiN films deposited on quartz substrate at Ar:N 2 = 95%:5%, obtained by averaging the measured p-and s-polarized reflectance/transmittance. The reflectance R first exhibits a minimum (R = 24%) at λ ~400 nm, and then significantly increases to R ≥ 70% as moving towards near-infrared wavelengths. The corresponding transmittance T instead reaches its maximum (T = 22%) at λ ~400 nm, and then decreases with increasing wavelength and remains <5% for λ > 900 nm. The observed characteristics of both reflectance and transmittance closely resemble those for Au films, validating that our TiN films are highly plasmonic/metallic. The reflectance minimum (or transmittance maximum) of TiN occurs on the high energy side where strong interband transitions prevail, whereas reflectance maximum (or transmittance minimum) in the longer wavelength region results from the high metallicity of TiN 1 . The high plasmonic quality of our TiN films is further confirmed by their Au-like luster 42 , as evident in the photograph inset to the Fig. 5.
We fabricate two-dimensional periodic square arrays of TiN nanodisks using our room-temperature sputtered TiN films (Ar:N 2 = 95%:5%; thickness = 45 nm) on quartz substrate. The nanodisk diameter d is varied from 100 to 180 nm while the period is fixed to p = 250 nm. In Fig. 6(a) the unpolarized transmittance spectra of the TiN nanodisk arrays with d = 100, 140, and 180 nm are shown (solid curves), obtained by averaging the measured p-and s-polarized transmittance. A pronounced transmittance dip is observed for all d values, corresponding to a plasmonic resonance mode excited in the TiN nanodisk arrays. For d = 100 nm, the transmittance dip occurs at λ = 630 nm with T = 71%. When d is larger, the observed transmittance dip clearly redshifts: for  d = 140 nm, the transmittance dip occurs at λ = 690 nm with T = 46%; for d = 180 nm, the transmittance dip occurs at λ = 760 nm with T = 27%. The redshift suggests that the excited resonance is a localized surface plasmon resonance (LSPR) mode. We perform full-wave numerical simulations for the transmittance of TiN nanodisk arrays using CST Microwave Studio to gain insight into our experimental results. The simulations are carried out in frequency domain with periodic boundary conditions using the dielectric functions of TiN extracted from variable angle spectroscopic ellipsometry (VASE) measurements and online optical data 43 for quartz. The simulated transmittance spectra are shown in Fig. 6(b), revealing overall good agreement with the experimental results. For d = 100 nm, the transmittance dip occurs at λ = 615 nm with T = 66%. Increasing the diameter d to 140 and 180 nm shifts the dip to λ = 663 and 711 nm and reduces the corresponding transmittance to T = 39% and 20%, respectively. Note that all of these resonance dips appear to be slightly stronger (i.e., lower transmittance) and occur at shorter wavelengths as compared to those observed in the experiments. This discrepancy can likely be reduced by minimizing the fabrication imperfections (see inset SEM to Fig. 6(a)) and by incorporating a thin surface oxide layer (TiO 2 or TiO x N y ) in the structural model for the fitting of the ellipsometry data of TiN. The simulated electric field intensity at the plasmonic resonance for arrays with d = 100 nm is shown in the inset to Fig. 6(b), indicating that the resonance is indeed an LSPR mode. Both experiments and numerical simulations therefore demonstrate that TiN based plasmonic nanostructures can be successfully realized by our room-temperature, bias-free reactive sputtering process. We further compare experimentally the plasmonic properties of the fabricated TiN nanodisk arrays to those of Au nanodisk arrays of identical geometric parameters fabricated on quartz. It is evident that the resonance dips for the fabricated TiN nanodisk arrays are similar to those for the fabricated Au nanodisk arrays (dashed curves in Fig. 6(a)) but cover a wider spectral range, proving that our TiN not only has comparable plasmonic performance to that of Au, but is more favorable for applications in the visible and near-infrared region where a broad plasmonic response is required 11 .

conclusions
In conclusion, we have demonstrated highly plasmonic TiN thin films and nanostructures by a room-temperature, low-power and bias-free reactive sputtering process. We investigate the optical and structural properties of our TiN films and their dependence on a number of processing parameters including substrate material, reactive gas flow ratio, and film thickness. We find that our TiN possesses one of the largest negative values of ε 1 as compared to all other plasmonic TiN films reported to date. We further fabricate periodic square arrays of TiN nanodisks using room-temperature sputtered TiN films, and validate that strong localized plasmonic resonances are supported in the arrays. Our room-temperature deposition process provides an easier and more cost-effective route to realizing plasmonic TiN than other high-temperature growth methods, allows for the fabrication of more complex TiN-based plasmonic nanostructures and devices, and could potentially be integrated with the existing CMOS process technologies to enable more practical applications of plasmonics.

Methods
Sputtering of TiN thin films was carried out at room temperature in a radio-frequency (RF) reactive sputtering system (Kurt J. Lesker) using three-inch titanium targets (99.995%) on different substrate surfaces including Si, quartz, HfO 2 , and PMMA, where HfO 2 and PMMA thin films were first deposited on Si substrate by ALD and spin-coating, respectively. All substrates were thoroughly cleaned using either buffered oxide etchants or solvents to remove surface oxide or contaminants before loading into the chamber. The substrate-to-target distance was kept at 5 cm to ensure a high sputtering yield and an enhanced density of reactive nitrogen ions near the substrate surface. The system was then pumped down to a base pressure of 2 × 10 −7 Torr. The Ar:N 2 gas flow ratio was varied (Ar:N 2 from 95%:5% to 60%:40%), while the pressure and RF power were held constant at 3 mT and 275 W respectively for all TiN films. Note that we found increasing the N 2 content would reduce the film deposition rate, and the plasma in the sputtering process could not be sustained when further increasing the nitrogen content in the gas flow ratio to Ar:N 2 = 40%:60%. It should also be noted that, although no intentional substrate heating was introduced, the substrate temperature during the sputtering process may have been slightly higher than room temperature, due mainly to the energy transfer from the sputtered species such as ions and neutrals from the target and the electrons from the plasma discharge. However, we expect this temperature increase should be insignificant, and the substrate temperature was much lower than that used in the low-temperature ALD work 16 . The dielectric function, thickness, transmittance (normal incidence), and reflectance (incident angle 20°) of the sputtered TiN films were measured using VASE at wavelengths from 300 to 2500 nm. XRD, XPS (VGS Thermo K-Alpha), and AFM (Bruker Innova) measurements were conducted to characterize the crystalline structure, chemical composition, and surface roughness of the TiN films, respectively. Arrays of TiN nanodisks (array area = 1 mm 2 ) on quartz substrate were fabricated by electron-beam (e-beam) lithography and inductively coupled plasma (ICP) etching. Au nanodisk arrays of the same surface area were fabricated on quartz by e-beam lithography, metal evaporation, and liftoff. All of the fabricated nanodisk arrays were examined using high-resolution scanning electron microscopy (SEM), and their transmittance was measured at normal incidence using VASE.