Reusability, Long‐Life Storage and Highly Sensitive Zirconium Nitride (ZrN) Surface‐Enhanced Raman Spectroscopy (SERS) Substrate Fabricated by Reactive Gas‐Timing Rf Magnetron Sputtering

Transition metal nitrides (TMN) are promising material alternative to replace noble metals in the field of plasmonic applications, especially surface‐enhanced Raman spectroscopy (SERS). Here we demonstrate a practical surface enhanced Raman spectroscopy (SERS) substrate using zirconium nitride (ZrN) thin films grown by reactive gas‐timing (RGT) rf magnetron sputtering. The tailored properties of ZrN thin film exploited for SERS activity could be achieved to obtain a highly sensitive ZrN thin film SERS substrate with the enhancement factor (EF) of 1.24 × 106 and 4.8 %RSD at 1626 cm‐1 toward methylene blue (MB) analyte which are comparable to the optimized Au sputtered thin films (EF=1.18 × 106 and with 5.1%RSD). We find that the spatial plasmonic hotspots on the surface of ZrN SERS substrate controlled by the turn‐on timing of Ar:N2 sputtered gas sequence, leading to the discrete conductive surface profile, strongly relates to non‐stoichiometric composition and the degree of (200)‐oriented texture at the surface of ZrN thin film. Furthermore, ZrN thin film SERS substrates exhibit an excellent recyclability more than 30 cycles with simple cleaning process and a storage time longer than 6 months. The detection and reusability of ZrN SERS substrate on the low concentration of trinitrotoluene (TNT) for homeland security are also performed.


Introduction
[3] Consequently, SERS has greatly received attention as an alternative analytical method with highly accurate, sensitive, and rapid detection for molecular quantification at trace levels of various sensing platforms. [1,2,4]10] Although, recently, semiconductor materials such as transparent conductive oxides (TCOs), alkali metals, metal silicides, metal carbides, and metal sulfides have been proposed as alternative plasmonic materials beyond noble metals, alkali metals manifest a high chemical reactivity with surrounding environments whereas TCOs and silicides materials are only active for ultraviolet and infrared region. [11,12]These drawbacks restrict them for SERS applications in various fields.][13][14][15][16] Among TMN materials, zirconium nitride (ZrN) gains an attention as an alternative plasmonic material, exhibiting metallic behavior at a wavelength shorter than that of titanium nitride (TiN) (i.e., ZrN = 390 and TiN = 510 nm) which is comparable to that of Au (330 nm) . [9]In addition, ZrN shows electric field intensity amplification similar to Au, resulting that ZrN may be a viable substitution of a widely-used Au as a plasmonic material. [9,13]esides, ZrN possesses not only high mechanical hardness of 22-24 GPa, the excellent biocompatibility and the lowest electrical resistivity (13.6 μΩ cm) in bulk among all TMN but also the better thermal stability than TiN (i.e., ΔH ZrN = −365.3kJ mol −1 and ΔH TiN = −336.4][19] To the best of our knowledge, although recently the plasmonic ZrN thin film has been utilized in wide range applications including spectroscopy signal enhancement platforms, light harvesting, thermoplasmonics, and CMOS-integrated devices, the exploitation of ZrN for SERS application is still scarce. [14,18,19]This might be due to a lacking comprehensive understanding to control the desired stoichiometry, crystallinity, morphological structure, and texture orientation of ZrN for achieving high performance SERS substrate through gaining hot spot density on the surface of ZrN SERS substrate.Such critical challenge is crucial not only for ZrN but also for other plasmonic TMN materials which will be enable us to regulate the LSPR characteristic via rational design of plasmonic TMN materials. [3,13]][22][23][24] The tailored properties of ZrN thin film exploited for SERS activity could be achieved by controlling an amount of sputtered atom from the target and enriching sputtered energy through a turn-on timing of Ar sequence variation. [5,15,20,22]We find that the spatially plasmonic hotspots on the surface of ZrN SERS substrate, resulting from the discrete conductive surface profile, strongly relate to non-stoichiometric composition and the degree of (200)-oriented texture at the surface of ZrN thin film.High SERS performances with the enhancement factor as high as 1.24 × 10 6 and 4.8% RSD at 1626 cm −1 toward a methylene blue (MB) are manifested from ZrN thin film SERS substrate which is comparable that to the optimized Au thin film SERS substrate. [6]In addition, ZrN thin film SERS substrates exhibit an excellent recyclability more than 30 cycles and a storage time longer than 6 months.Finally, we exhibit the utilization of reusable ZrN SERS substrate to distinguish a concentration of 2, 4, 6-trinitrotoluene (TNT).This finding can provide an important guide to design the effective TMN SERS substrate beyond noble metals which enables us to deliver guidance for designing plasmonic materials applied for a wide range of applications.

Results and Discussion
Figure 1a shows GIXRD patterns of the sputtered ZrN thin films fabricated via reactive gas-timing technique (RGT) as a function of turn-on Ar timing at room temperature (RT) and reference sputtered Zr film (black line).Note that the x-ray grazing incident angle employed during GIXRD measurement was fixed at 1°.The turn-on Ar sequence is varied from 20 s to 300 s (i.e., 20 s: purple line, 50 s: light blue line, 100 s: green line and 300 s: red line) whereas the turn-on N 2 sequence is fixed at 6 s (the details of growth condition can be seen in Figure S1 and Table S1, Supporting Information).The thickness of ZrN films is kept at 200 nm.The results show that three diffraction peaks corresponded to (111), (200), and (220) planes reflection of facecentered cubic ZrN can be obtained when the turn on Ar sequence is below 100 s.These indicate that crystalline-ZrN thin film could be fabricated using the RGT technique without any additional energy source, i.e., heating substrate and/or applying bias on substrate. [17,25]Furthermore, the occurrence of the (200) texture can be attributed to a metallic behavior of the transition metal nitride (TMN) films. [17]It should be noted that an amorphous ZrN occurs when the turn-on Ar timing sequence is below 20 s (not show here) which might result from a lack of Zr atom/cluster from the target. [22]Above 100 s, on the other hand, a mixture phase including (111) plane of (ZrN) and (200) plane of metallic Zr appears.This result reveals that there is an upper limit for the turn-on Ar sequence to obtain a single phase ZrN.In addition, we find that the FWHM of (200) peak slightly decreases from 1.3 to 1.1 when the turn-on Ar timing sequence decreases from 100 to 20 s (Figure S2, Supporting Information).We also observe that the position of (200) peak was shifted to a higher angle when the turn-on Ar timing sequence is decreased.Such redshift of the (200) peak might be from the compressive stress event due to the atomic peening effect (i.e., enrichment of sputtered energy) induced by the RGT technique. [22,25]Note that such similar trend of the increase of FWHM and redshift of (111) peak is also investigated when the turn-on Ar timing sequence increases from 20 to 100 s (Figure S2, Supporting Information).
Figure 1b-d exhibit the X-ray photoelectron spectroscopy (XPS) analysis of ZrN thin films fabricated using the RGT technique as a function of turn-on Ar sequence.Surface measurement was performed by using 2 kV of Ar + with rastering on an area 2×2 mm 2 for sputtering. [15]The chemical state and electron spectra were collected after 15 min etching time to avoid surface reaction after thin film growth.Figure 1b demonstrates the Zr 3d XPS spectra as a function of turn-on Ar sequence.Below the turn-on Ar timing sequence of 100 s, the Zr 3d XPS spectra are distinguished for two peaks of Zr3d 3/2 and Zr3d 5/2 which are corresponding to ZrN with a binding energy of 179.3 and 181.7 eV, respectively. [19,26]In order to verify the chemical bonding of such XPS peaks, the deconvolution of Zr3d 3/2 and Zr3d 5/2 was applied as shown in Figure S3 (Supporting Information).The results exhibit that three pairs of Zr-N, Zr-ON, and Zr-O are investigated.Above turn-on Ar sequence of 100 s, on the other hand, the appearance of the two XPS peaks at 178.70 and 181.10 eV is observed which can be assigned to the binding energy of Zr 0+ and Zr-Zr bonding (Figure S3, Supporting Information) . [26]These results imply that a phase transformation from Zr-Zr to Zr-N compound could be tailored by using the RGT technique.Note that the occurrence of oxygen profile as oxide species dominated for all samples might be from the residual oxygen during deposition process and/or measurement process. [17]Figure 1c exhibits N1S spectra of ZrN thin film prepared from various turn-on Ar timing sequences.It can see that a sharp dominated peak at 397.30 eV is observed from all samples which is corresponding to ZrN. [26,27] The deconvolution results show that three peaks corresponded to Zr─N bonding (397.30eV), Zr─O─N bonding (399.2628] On the other side, the sample prepared at the turn-on Ar timing sequence higher than 100 s exhibits only two peaks of ZrN and ZrO x N y , indicating to the lower N content in the sputtered thin films.The O1s XPS spectra is also investigated as shown in Figure 1d.The result shows a broad peak at a BE range of 530.70 -531.39 eV for all samples which could be assigned to oxide in oxide state. [26]Such oxide state is possible from the small amount of Zr-O compound which is possibly originated from residual oxygen in sputtering system. [18,20]To identify the elemental content of ZrN thin film, Auger electron spectroscopy (AES) was utilized to evaluate the depth profile and the quantitative evaluation data of ZrN thin film as shown in Figure 1e and Note S5 (Supporting Information).Noted that the quantitative measurement data was collected at the ion-beam etching from 20 to 60 min to avoid the natural oxide layer on the top surface and fluctuation of signal near the interfaces between ZrN and Si substrate (Figure S5 (e), Supporting Information).It can see that nitrogen content in the sputtered ZrN thin films increases from 21.8% to 52.9% whereas zirconium content decreases from 50.5% to 29.3% when the turn-on Ar timing sequence decreases from 300 to 20 s.These results indicate that controllability on N/Zr chemical ratio from 0.33 to 1.88 could be performed via the RGT technique.In addition, the occurrence of O content is observed in all samples which is corresponding to XPS results.Generally, it has been known that the fabrication of ZrN thin films from a conventional reactive magnetron sputtering could be performed by using reactive Ar+N 2 mixing gases.The Ar gas mainly acts as the sputtered ion, which bombards on the target surface to supply the sputtered atoms/clusters whereas the N 2 gas is the reactive gas, which reacts with sputtered atoms/clusters from the target to form the compound thin films. [5,15,20,22]However, the additional energy sources during reactive sputtered thin film growth process were frequently required to enhance the activation energy and/or for controlling the amount of Zr from the target to obtain a single phase of ZrN films couple with chemical compositional adjustability. [17,25,28,29]Such requirement of the additional energy sources could be confirmed from the obtained result of an amorphous ZrN film which was fabricated via the conventional reactive magnetron sputtering technique at RT as shown in Figure S2 (Supporting Information).In the aspect of the RGT technique, on the other hand, the turn-on Ar timing sequence can be considered to the generation of the number of sputtered Ar ions bombarded on the target surface whereas the turn-off Ar sequence can be attributed to the reduction of the working pressure. [5,22]uch turn-on Ar timing sequence enable us to adjust the number of sputtered atoms/clusters obtained from the target whereas the turn-off Ar sequence allows us to gain the thin film formation due to the increment of energy per atom. [5,15,22]On the other hand, the switched turn-on N 2 timing during the turn-off Ar sequence can be attributed to the reactive sputtering process for metal-nitride formation. [15,20,22]Since it has been known that the forming energy of sputtered compound thin films strongly relates to the amount of supplied sputtered atom/cluster from the target and the amount of the reactive-ion gas (i.e., flux of material species or precursor concentration) , [30] an emergence of the upper bound of turn-on sequence of Ar to obtain single phase of crystalline ZrN films as shown in Figure 1a should be from the upper limit of the number of sputtered atoms/clusters and the amount of reactive-ion gas of N correlated with the operated sputtered energy of the RGT conditions. [15,22]In the other words, the over excess of Zr sputtered atom/cluster from target at the longer turn-on Ar sequence of 300 s leads to the emergence of the mixed phase. [22]On the other hand, the shorter turn-on Ar timing not only allow us to reduce the amount of Zr sputtered atom/cluster from target and enrich sputter energy but also increase periodic number of turn-on and turn-off sequence over the growth time ((i.e., the energy modulation frequency (EMF)), supporting information S6).Such increase in the EMF value from 12 to 138 round h −1 could provide a larger an amount of total nitridation time from 71 to 831 s h −1 (Figure S6, Supporting Information) couple with forming energy enrichment which might be responsible for the increase of N content in the single phase ZrN thin films as shown in Figure 1e.Recently, HfN thin films fabricated via RGT technique through the variation of N sequence was demonstrated our research group. [15]Although the sputter energy enrichment could be obtained by the shorter N sequence, the controllability on chemical composition and texture orientation could not be achieved.These results highlight that capability to control the amount of Zr sputtered atom/cluster from target, enrich sputter energy and increase periodic number of turn-on and turn-off sequence over the growth time via Ar sequence is crucial for controlling crystallinity and chemical composition of ZrN thin films.
To investigate the optical properties of ZrN thin films, a spectroscopic ellipsometry (SE) is utilized.It should be noted that the thickness of ZrN thin films is set at 50 nm which is calibrated from step profiler and thickness monitor.The measurement of Ψ and Δ spectra is performed at the incident angle of 55°, 65°, and 75°in the range of 0.75 -5.00 eV with 0.025 eV interval.The spectra are fitted with the proposed triple layer model which consists of Si substrate, native silicon oxide and ZrN thin film as shown in Figure 2a.The thickness of the native oxide on silicon surface possibly originated from moister and oxygen in air is fixed at 2 nm. [21,23,24]The effective thickness (d eff ) and the effective dielectric function ( eff ) related to localized plasmonic resonance properties of ZrN films are determined through a combination of Drude model and Gaussian formula as presented in Figure 2b which is an analyzed result of optimized Au films (Figure S7, Supporting Information) . [31,32]Note that the optimization of the sputtered Au thin film related to spectroscopy enhancement can see elsewhere. [33,34]In more details, it should be noted that the infrared absorption related to the metallic behavior of ZrN thin films, especially the conductivity due to free electron, is determined via Drude function. [31,32]The absorption referred to the localized surface plasmon resonance (LSPR) and inhomogeneous broadening LSPR (i-LSPR) in the visible and near infrared regions are identified by two Gaussian functions. [31,32]The out-ofplane LSPR (o-LSPR) and interband transition of ZrN thin films in the ultraviolet region are examined by Gaussian and Lorentz functions. [33] 3a).This indicates that the surface of ZrN thin films might not be conductive below the turn-on Ar sequence of 50s. [6]Likewise, such lower conductive surface of ZrN 20:6 compared with the other Ar sequence is confirmed by 4-probe conductive measurement   technique (Figure 5h).The inter-band transition (orange color) deconvoluted by Gaussian function increases when the turn-on Ar sequence is decreased.This result implies that the chemical composition of ZrN thin films might be changed via the RGT conditions which is corresponding to the red-shift of the absorption spectra of ZrN due to the increase of the N/Zr ratio at the shorter turn-on Ar sequence (Figure 1d) as shown in Figure S8 (Supporting Information) and Figure 1e.On the other side, the Gaussian functions related to LSPR (dark blue) and i-LSPR (light blue) are observed when the ZrN thin films are fabricated below the turn-on Ar timing of 100 s.Such emergence of LSPR and i-LSPR fitting curves confirm that the LSPR properties of sputtered ZrN thin films could be performed via the RGT technique.In addition, we find that the maximum peak of LSPR and i-LSPR fitted curves of the sputtered ZrN 20:6 sample is larger than that of ZrN 50:6, suggesting that the sputtered ZrN 20:6 thin film is the most appropriate sample for LSPR applications. [31,32]Since the i-LSPR is the calculated function which is utilized for broadening compensation in the curve-fitting process to determine a major Gaussian component of plasmonic absorbance (LSPR) , [31,32] the LSPR Gaussian fitting curve of the ZrN 20:6 sample is further determined and compared that to the optimized sputtered Au thin film (Figure 2b) at different wavelength in order to illustrate the comparable LSPR properties of both materials as shown in Figure 4e.It can see that the maximum value of  2 of the ZrN 20:6 sample is observed at 580 nm whereas the maximum value of  2 of the optimized Au thin film is investigated at 735 nm.Surprisingly, we find that there is the overlapping between LSPR Gaussian fitting curve of the ZrN 20:6 and the optimized sputtered Au thin film in the visible wavelength which intersects together at the wavelength of 633 nm.Note that the maximum  2 value at 580 nm is also observed from the ZrN 50:6 sample (blue line).In addition, we also find that the absorption spectra in visible region of ZrN 20:6 thin film is comparable to that of the optimized sputtered Au thin film (Figure S8, Supporting Information).These results suggest that the comparable spectroscopy signal enhancement by using the ZrN 20:6 and the optimized Au thin films as enhancement substrate (i.e., SERS substate and fluorescence enhancement (FE) substrate) might be equated when measure with the excitation wavelength of 633 nm.To clarify such issue, here we have investigated the comparison of SERS performance between ZrN thin film grown via the RGT technique and the optimized sputtered Au thin film as shown in Figure 4.
Figure 4a exhibits Raman spectra of MB at the concentration of 10 −3 M obtained from ZrN thin film grown by the RGT technique (red line), the optimize sputtered Au thin film (blue line) and blank substrate (Si, red line).It should be noted that the thickness of ZrN thin film is fixed at 50 nm.The peak at 520 cm −1 corresponds to the Raman scattering of the crystalline Si substrate.The result shows that the Raman resonance peak at 450, 770, 1154, 1398, and 1632 cm −1 of MB related to C─N─C skeletal deformation mode, the in-plane bending mode of C─H, C─N symmetrical stretching and C─C ring stretching, respectively, [7] are clearly observed on the ZrN and optimize sputtered Au thin films.Note that the full width of half-maximum (FWHM) related to Raman resonance peak at 450, 770, 1154, 1398, and 1632 cm −1 of MB of ZrN is 20.9, 15.7, 23.9, 18.1, and 19.0 cm −1 whereas FWHM related to Raman resonance peak at 450, 770, 1154, 1398, and 1632 cm −1 of MB of optimize sputtered Au thin films is 16.7, 16.0, 27.2, 18.3, and 19.1 cm −1 , respectively.In addition, we find that the Raman intensity of MB obtained from ZrN thin film seems to be comparable to the optimize sputtered Au thin film.For further information, we have investigated the comparison of SERS performance between of the ZrN and optimize sputtered Au thin films through the dependence of MB concentration on SERS intensity as shown in Figure 4b.Note that the SERS intensity of MB is obtained from Raman shift at 1632 cm −1 which is collected from 10 spot points to calculate the relative standard deviation of the Raman intensity at each concentration (Figure S9, Supporting Information).It can see that the SERS intensity of the ZrN and optimize sputtered Au thin films decreases when the concentration of MB decreases from 10 −3 to 10 −7 M and after that cannot detect below 10 −7 M, indicating that the limit of detection (LOD) of both SERS substrates is ≈10 −7 M.Moreover, we find that the SERS intensity obtained from ZrN thin film and the optimized sputtered Au thin films at each concentration is in the same range value.For more details, the enhancement factor (EF) is calculated to compare the SERS enhancement between of the prepared ZrN and optimized sputtered Au thin films.Note that the method employed to determine the EF value can be seen elsewhere. [5,7,8,15]By collecting the intensity of Raman shift characteristic at 1632 cm −1 of MB at 10 −3 M, the EF of ZrN thin film is 1.24 × 10 6 with 4.8%RSD whereas the EF of the optimized Au sputtered thin film is 1.18 × 10 6 with 5.1%RSD.In addition, the comparable uniformity of SERS signals of ZrN thin film substrate to the optimize sputtered Au thin film is demonstrated by SERS mapping measurement as shown in Figure S10 (Supporting Information).Thus, these results highlight that the SERS performance of ZrN thin film could be comparable to the optimize sputtered Au thin film.Figure 4c exhibits the dependence of excited laser wavelength on the SERS intensity of ZrN thin films measured from Raman shift characteristic at 1632 cm −1 of MB at a concentration of 10 −3 M. The laser wavelength of 432, 532, 633, and 785 nm is utilized for SERS performance investigation.It can see that the maximum intensity of SERS signal is obtained when ZrN SERS substrates are activated by the excited laser wavelength of 633 nm.On the other hand, the other excited laser wavelengths show the lower SERS intensity.By changing the analyte target to the rhodamine 6G (R6G), in addition, the similar trend of SERS signal related to the excited laser wavelength is also observed (Figure S11, Supporting Information).Such obtained results are corresponding to the dependence of wavelength on  2 profile as displayed in Figure 3e, confirming that the proposed fitting model based on SE measurement is possibly employed for LSPR material determination.In addition, the similar trend of the dependence of excited laser wavelength on the SERS intensity is also found in ZrN 50:6 sample (Figure 4c and Figure S11, Supporting Information).,7] The former relates to a consequence of amplifying electric fields through the excitation of LSPR of SERS materials (e.g., Au, Ag, Cu) whereas the latter is a product of charge transfer (CT) mechanism of either plasmon-free SERs materials or adsorbed analyte on active SERS surface. [8]Since the maximum SERS intensity is attained by using the excited laser wavelength of 633 nm even though changing the analyte substance from MB (major absorption peak at 665 nm) to R6G (major absorption peak at 530 nm), the major SERS mechanism of ZrN thin film should be from the EM mechanism rather than that of CM mechanism. [3]ere we question that what causes the maximum obtained result of LSPR (Figure 3) and SERS intensity (Figure 4) of ZrN 20:6 sample.Regarding to the EM enhancement of SERS thin film substrate, the performance of SERS thin film based on noble materials (i.e., single element: Au, Ag, Cu) is commonly explained through the presence of the localized hot spot between nanostructures related to the physical features of materials which is not only based on the optical properties of material but also surface topography and surface charge distribution. [6,9]Despite ZrN is the compound materials, the dominant on the presence of the hot spot effect on the surface related to LSPR and SERS activities of transition metal nitride (TMN) thin film have been also confirmed from the previous reports. [9,10]We have recently used a conductive atomic probe microscope (CAFM) to investigate the local electronic characteristics correlated to SERS generation of noble thin films. [6]To clarify such hot spot effect issue, CAFM is therefore employed to explore the surface topography and the corresponding electronic distribution on the surface of the ZrN SERS thin films as shown in Figure 5. [6] Figure 5a-c show the morphology of the ZrN thin films fabricated via the RGT technique as a function of turn-on Ar sequence of 20, 50, and 100 s, respectively, which was investigated from AFM measurement.The surface grain size and root mean square (RMS) roughness are collected from AFM images as shown in Figure 5g.The results demonstrate that the surface grain size as a function of the turn on Ar timing of 20, 50, and 100 s are 37.9, 38.7, and 43.5 nm while the surface roughness as a function of the turn on Ar timing of 20, 50, and 100 s are 0.91, 0.94, and 0.99 nm, respectively.Note that the decrease of surface grain size at the shorten of turnon Ar sequence is also confirmed by FE-SEM measurement as shown in Figure S12 (Supporting Information).Such decrement of the surface grain size with the constant RMS roughness related to the shorter turn-on Ar sequence should come from an atomic peening effect which is the characteristic event of the RGT technique. [5,21]Although the surface grain size values trend to be deceased with the shortened turn-on Ar timing, such tiny decreased values, especially at the turn-on Ar timing of 20 and 50 s, might be not significant for the enhancement SERS signal.Thus, the surface topography solely cannot be employed to explain the maximum obtained result of LSPR and SERS intensities.By referring to that C-AFM can resolve the spatial distribution of surface conductivity which does not exhibit noticeable features the corresponding topography image, [35,36] we have considered the current maps simultaneously acquired at the same surface areas.Figure 5d-f exhibits CAFM current images of the ZrN thin films fabricated via the RGT technique as a function of turn-on Ar sequence of 20, 50, and 100 s, respectively.Noted that the supplied voltage of -3 V referred to n-type behavior of ZrN thin films is used to observe current images. [37]Surprisingly, it can see that an emergence toward dark color in CAFM images is clearly observed throughout the surface of ZrN thin films, especially at the turn-on Ar timing of 20 and 50 s.Such dark color spots can be attributed to the loss electrical contact between tip and ZrN thin film due to the non-conductive surface of the samples, depending on the distribution of composition and crystallinity.Furthermore, we find that the amount of dark spot (i.e., non-conductive area) distributed on the surface of ZrN increases when the turn-on Ar timing is decreased, leading to an occurrence of the discrete electrical profile on the surface of ZrN thin films correlated to the decrease and absence of Drude function of ZrN 50:6 and ZrN 20:6 samples as shown in Figure 3 due to the decrease of conductivity on the surface of ZrN thin film.Recently, C.P. Liu et.al. reported that the conductivity of ZrN thin films certainly depended on the composition and crystallinity. [17]The maximum conductivity could be accomplished when the maximum (200)textured orientation and stoichiometric phase are performed.On the other hand, the low conductivity is obtained when the N-rich ZrN phase and low (200)-textured orientation of ZrN thin film are observed. [17]Since the obtained results presented in Figure 1d show that the N/Zr ratio as a function of the turn-on Ar timing of 20, 50, and 100 s is in N-rich regime for  and 4-probe measurement.Below the turn-on Ar sequence of 50 s, the textured coefficient of (200) plane increases from 0.13 to 0.29 when the turn-on Ar sequence is increased from 20 to 50 s and after that slightly decreased to 0.24 above the turn-on Ar sequence of 100 s.On the other side, the resistivity of ZrN thin films rapidly decreases when the turn-on Ar sequence increases from 20 s (153.4Ω cm) to 50s (39.3 Ω cm) and slightly decreases above the turn-on Ar sequence of 50 s (e.g., 36.6 Ω cm at the turnon Ar sequence of 100 s).These obtained results indicate that the composition and textured coefficient of (200) plane controlled by the RGT technique strongly influence on the conductivity of ZrN thin film which possibly induces the emergence of discrete electrical profile on the surface of ZrN thin films.Note that the lower conductivity of ZrN 50:6 compared to ZrN 100:6 might be from the higher oxygen concentration in ZrN thin film (Figure 1e) which is often observed for the N-rich ZrN phase. [17]On the other hand, the lower (002)-textured coefficient of ZrN 100:6 compared to ZrN 50:6 might be from the lower forming energy enrichment than that of ZrN 50:6 due to the smaller of EMF (Figure S6, Supporting Information).Since the discrete electrical surface profile of ZrN thin films can be considered to the nanoislands and/or percolated film structures of noble metal film deposited on the substrate which strongly affect to the enhancement of surface plasmonic properties localized at the surface through hot spot effect, the corresponding electronic distribution on the surface should be responsible for the maximum obtained result of LSPR and SERS intensity of ZrN 20:6 sample.Moreover, we find that the over increase of non-conductive area distributed on the surface of ZrN (i.e., the deterioration in the textured coefficient of (200) plane) controlled by increasing the thickness of ZrN 20:6 film leads to the low performance of SERs activity due to the lack of hot spot effect on the surface (Figure S13, Supporting Information).Thus, our results highlight that controllability on the chemical composition and textured orientation of ZrN thin film via the RGT technique to provide the discrete electrical profile on the surface is an important key to perform the high performance ZrN SERS thin film.
Figure 6a exhibits the recyclability performance of the ZrN SERS substrate plotted as the intensity of Raman shift characteristic at 1632 cm −1 of MB at a concentration of 10 −3 M which is collected from 10 randomly selected places on the ZrN SERS substrate.For the recycling process, the tested ZrN SERSs substrates (1 st -MB detection) were directly cleaned by water rinsing coupled with wiping it off by cotton bud after the measuring SERS process.After removing MB from the surface of the ZrN SERs substrate, the Raman signal of MB is vanished, leaving the Raman spectrum identical to that obtained prior to dropping MB (1 st -MB cleaning).When the MB was dropped onto the same substrate again, a similar Raman signal was observed, as indicated by the spectrum labelled as 2 nd -MB detection.The completed Raman spectral results during the recyclability performance are shown in Figure S14 (Supporting Information).It can see that the SERS intensity of ZrN substrate is almost unchanged even though the ZrN SERS substrate is reused >30 times.The corresponding relative standard deviation (RSD) of recyclability data is found to be ≈5.0%,indicating to an excellent recyclability performance of the fabricated ZrN SERS substrate.The stability of ZrN SERS substrate under ambient air condition is investigated as shown in Figure 6b.The ZrN SERS substrate is still highly sensitive after storage in an air ambient >6 months.The RSD of storage data is ≈4.5%.10] Figure 6c demonstrates the reproducibility in 10 batches (n = 10) of ZrN SERS substrate plotted as the intensity of Raman shift characteristic at 440, 1180, and 1632 cm −1 of MB at a concentration of 10 −3 M which is collected from 10 randomly selected places on ZrN SERS substrate.The RSD of Raman shift characteristic at 440, 1180, and 1632 cm −1 obtained from 10 devices is 4.8%, 3.8%, and 4.7%, respectively, confirming that the ZrN SERS substrate has an excellent reproducibility.Additionally, we have demonstrated the utilization of the reusable ZrN SERS substrate to detect 2, 4, 6-trinitrotoluene (TNT) as shown in Figure 7.In this study, 2.0 μL of TNT at a concentration of 10 −1 -10 −7 M was deposited on ZrN SERS surface and left to dry at room temperature in an air atmosphere. [7]The Raman data collections were performed at 10 different locations on each substrate.Figure 7a exhibits the dependence of TNT concentration on SERS intensity.It can see that the major Raman peaks, especially at 792, 823, 1210, 1533, and 1616 cm −1 are clearly observed which is in good agreement with previous literatures. [7,38]The SERS signals of TNT can be observed only when the TNT concentration is above 10 −6 M, implying that the lowest concentration for TNT detection of of ZrN SERS substrate for TNT detection is ≈10 −6 M. Furthermore, the comparability of SERS performance between of the ZrN SERS substrate and optimize sput-tered Au thin films are examined by collecting SERS intensity of Raman shift characteristic at 1359 cm −1 related to the dependence of TNT concentration as shown in Figure 7b.The result demonstrates that the lowest concentration for TNT detection of the optimize sputtered Au thin film is also ≈10 −6 M. In addition, the same range in SERs intensity and RSD values between the ZrN SERS substrate and optimize sputtered Au thin films clearly are investigated, indicating to a good comparability between the ZrN and optimize sputtered Au thin films.Noted that the RSD value larger than 10% obtained from both SERS substrates might be from the coffee ring effect due to the evaporation of aqueous solution (i.e., acetone) on a flat surface during preparation process of the analyte. [39]The recyclability performance of ZrN SERS to detect TNT at a concentration of 10 −3 M is illustrated as shown in Figure 7c,d.The reusable ability >20 times could be achieved, indicating to the excellent recyclability performance.Thus, our results highlight that high sensitivity coupled with reusability, long-life storage and reproducibility could be accomplished by ZrN SERS substrate fabricated by the RGT rf magnetron sputtering.To achieve a higher SERS performance based on ZrN, the increase of the hot spot density on the surface is the major issue which could be possibly done by using nanoarrays, nano template and nanostructure synthesis [7,9,[40][41][42] to increase the amount of LSPR on surface of ZrN SERS substrate.However, the ability to precisely control the desired stoichiometry, crystallinity, morphological structure, and texture orientation of ZrN prepared via such methods is the critical challenge to gaining hot spot density on the surface of ZrN SERS substrate for achieving high performance SERS substrate.Further investigation is in progress.We believe that our work not only paves the way to control and design a practical TMN-SERS substrate but also guide for designing of new LSPR devices for a wide range application.

Conclusion
In this work, we have tailored the properties of ZrN thin films via the reactive gas-timing (RGT) rf magnetron sputtering technique to obtain a high sensitivity coupled with a reusable and long-life storage TMN-SERS thin film SERS substrate.By controlling an amount of sputtered atom and enrich the forming energy of sputtered films through reducing the turn-on Ar timing sequence, the morphological microstructure, chemical composition, and optical properties of ZrN thin films could be tuned to attain a highly sensitive ZrN thin film SERS substrate with the enhancement factor (EF) of 1.24 × 10 6 and 4.8%RSD at 1626 cm −1 of methylene blue which could be comparable to the optimized Au thin film SERS substrate.Furthermore, ZrN thin film SERS substrate exhibits an excellent recyclability >30 cycles and a storage time longer than 6 months due to their mechanical hard surface of 21.3 GPa and an inherent chemical stability.We find that the spatial plasmonic hotspots on the surface of ZrN SERS substrate strongly relates to non-stoichiometric composition and the degree of (200)-oriented texture at the surface of ZrN thin film, leading to the discrete conductive surface profile.In addition, the utilization of reusable ZrN SERS substrate to detect the low concentration of trinitrotoluene (TNT, 0.1 -10 −6 M) with recyclability more than 20 times for homeland security is also demonstrated.Our results highlight that the practical high sensitivity coupled with reusability, long-life storage of TMN-SERS thin film substrates could be performed by the RGT rf magnetron sputtering technique.
The Si (111) and glass slide substrates were cleaned via an alcohol process in which the Si (111) and glass slide substrates were ultrasonically cleaned with acetone, isopropanol alcohol and deionized water for 10 min, respectively, and then dried under a flow of nitrogen.The samples were mounted on a rotational substrate holder (rotation speed of 10 rpm) and transferred to the sputtering chamber in which the substrate-target distance was 70 mm.Then, the sputtering chamber was pumped down until the base pressure of 1.0×10 −6 mbar was performed in which the sputtering pressure was measured by Pirani and Penning gauges.Note that the quality of the vacuum pressure at 1.0 × 10 −6 mbar is monitored and calibrated using the residual gas analysis (RGA, INFICON transpector 2) before thin film deposition.High purity (99.999%) argon (Ar) as bombard gas and high purity (99.999%) nitrogen (N 2 ) as reactive gas were sequentially introduced to the sputtering chamber via mass flow controller (MKS) at a constant flow rate of 10 and 1 sccm.During the sputtering deposition, the rf generator power and the sputtering working pressure were kept at 100 W and 5.0×10 −3 mbar, respectively.With the RGT technique, ZrN thin films were prepared via varying the alternate on-off timing sequences between the sputtered Ar and reactive N 2 gases to control an amount of sputtered atom from the target and enrich the reaction energy of ZrN thin films during deposition process. [15,20]In this work, turn-on Ar timing was varied in the range of 20 -300 s whereas turn-on N 2 timing was fixed at 6 s.Details of all the conditions are summarized in Figures S1 and Table S1 (Supporting Information).We denote the RGT parameters of the Ar:N 2 timing sequence (s:s) as 20:6, 50:6, 100:6, and 300:6, respectively.Note that the thickness of ZrN thin film was varied from 10 to 200 nm.The sputtered ZrN samples were then investigated for the changes in structure, morphology, chemical composition, electrical and optical properties based on the gas-timing technique.It should be noted that the deposition processes were carried out at room temperature without post-thermal treatment process.

Characterization of the Zirconium Nitride Thin Films:
The crystal orientation of ZrN thin films was examined by grazing incident x-ray diffractometer (GIXRD, Smartlab, Rigaku) using Cu K radiation.Atomic force microscope (AFM, Hitachi Hi-Tech, SPA400 and JPK, NanoScience) was employed to investigate the morphology and conductive surface profile of ZrN thin films via contact conductive-AFM (CAFM) mode.The conductive PtIr probe tip (SCM-PIT-V2) with the tip radius 25 nm and spring constant 3 N m −1 were utilized.The set point force of < 20 nN, scanning velocity of 1.26 um −1 s and pixel resolution of 256×256 are set for data collection (the details can see elsewhere [6] ).The thickness of ZrN thin films was measured by using step profiler (MITUTOYO SURFTEST).Chemical state near the top surface region of ZrN thin films was characterized by x-ray photoemission spectroscopy (XPS, ULVAC-PHI, PHI Quantera II, Chigasaki, Japan) with monochromatized Al K radiation at 1486.6 eV and spot size of 100 μm 2 .The depth profiles and chemical compositions of ZrN thin films were analyzed using Auger electron microscopy (AES), which was operated at 10 kV acceleration voltage, 10 nA beam current, and 40 × 40 mm 2 beam diameter.The sample surface was pre-sputtered for 30 s using Ar + ions with a fixed scan speed at 1 eV per step before measurement.More details of AES operation were previously described in Khemasiri et al. [22] Absorption measurements were performed using a UV/Vis spectrophotometer (Lambda 650, Perkin Elmer) in the energy range of 0.8 -3.2 eV energy.The dielectric function of the ZrN thin films were analyzed via a spectroscopic ellipsometer (SE; J.A. Woollam VASE2000).The ellipsometric spectra including Ψ and Δ were performed at various incident angles of 55°, 65°, and 75°within the photon energy range of 0.75 -5.50 eV at 0.025 eV intervals.The polarized light's reflection, , is defined as follows: where R p and R s were the complex reflection coefficients of the polarization component in parallel and perpendicular axes, respectively.The optical model based on Drude, Gaussian and Tauc-Lorentz functions, related to free electron, localized surface plasmon resonance (LSPR) and interband transition absorption, was proposed and used to extract the properties of interest through the fitting process. [43]The imaginary part of dielectric function ( 2 ) of the ZrN film was firstly defined then the real part ( 1 ) was determined through the Kramers-Kronig relations to complete its complex dielectric function. [23,24]The electrical properties were examined utilizing four-probe measurement (Jandel, RM3 Test unit).The hardness of thin film was determined by nanoindentation (Micro Materials), with the load range of 0 to 1000 μN using the Berkovich tip with the indent probe area of 100 × 100 nm 2 (the values were extracted from the load displacement curve at the maximum load of 1 mN).
SERS Measurements: To investigate SERS performance of the sputtered ZrN thin films, methylene blue (MB) and trinitrotoluene (TNT) explosive solution were used as probing molecules.Raman measurements were conducted with Raman spectrometer (NTEGRA; NT-MDT) equipped with a confocal optical microscopy system.Prior to each experiment, a tiny drop of aqueous solution MB (water) and TNT (acetone) at 10 −3 -10 −8 M and 10 −1 -10 −7 M concentration, respectively with a volume of 2.0 μL was deposited on each SERS surface and left to dry at room temperature in an air atmosphere.During SERS measurement, various excitation sources at the wavelength of 532, 633, and 785 with 1 μm spot size diameter were employed while the laser power output was maintained at 0.35 mW and each spectral data was acquired in 10 s.The charge-couple device (CCD) with a resolution of 4 cm −1 , was employed to record the SERS spectra.The Raman data collections were performed at ten different locations on each substrate, in order to confirm the stability and reproducibility of the fabricated SERS samples.The recyclability of the fabricated SERS substrates was performed with MB and TNT as the adsorbed molecules on to the SERS samples, both before and after being cleaned via washing by cotton buds.The cleaning process was repeated for 30 times to ensure recyclability of ZrN SERS substrate.Furthermore, the long-life retention time of samples was investigated for 27 weeks.

Figure 1 .
Figure 1.a) GIXRD pattern of sputtered ZrN thin films fabricated via reactive gas-timing technique (RGT) as a function of turn-on Ar sequence at room temperature (RT) and reference sputtered Zr film (20 s: purple line, 50 s: light blue line, 100 s: green line, 300 s: red line and Zr: black line).XPS spectra of b) Zr 3d, c) N 1S, and d) O 1S as a function of turn-on Ar sequence.e) Zr and N atomic concentration measured by Auger electron spectroscopy (AES) of ZrN thin films as a function of turn-on Ar sequence.
Figure 2c,d present the proposed model fitting results of ZrN 20:6 which are performed at difference incident angles (55°, 65°, and 75°) as a function of photon energy.The results demonstrate that the proposed model and measured result of Ψ and Δ spectra exhibit a well-fitting all incident angles.To verify the practical utilization of the fitting model, the comparison between d eff evaluated by SE measurement and the film thickness measured by cross-sectional FE-SEM images (d FE-SEM ) of ZrN thin films fabricated via the RGT technique as a function of turn-on Ar sequence is performed as shown in Figure 2e.It can see that the d eff values evaluated by SE of ZrN thin films as a function of the turn-on Ar timing of 20, 50, 100, and 300 s are 52, 50, 45 and 44 nm whereas the d FE-SEM values are 46, 46, 42 and 42 nm, respectively.These indicate that the de ff values measured from SE are in the same range with those determined from the cross-sectional FE-SEM images (d FE-SEM ).Thus, the proposed fitting model could be suitability to further investigate the  2 component functions related to localized surface plasmon resonance (LSPR) of the ZrN thin films as presented in Figure 3.
Figure 3a-d exhibits the  2 component function curvefitted results of sputtered ZrN thin films fabricated via the RGT technique as a function of turn-on Ar timing sequence.It can see that the Drude function (green color) trends to decrease when the turn-on Ar timing is decreased.After that the Drude function disappears at turn-on Ar sequence of 20 s (Figure

Figure 2 .
Figure 2. Spectroscopic ellipsometry optical model represented as the a) physical structure, b) dielectric function of optimized Au thin film layer, c) Ψ, and d) Δ spectra of ZrN 20:6 thin films performed at difference incident angle of 55°, 65°, and 75°as a function of photon energy.e) Comparison between d eff evaluated by SE measurement and the thickness values measured by cross-sectional FE-SEM images (d FE-SEM ) of ZrN thin films fabricated via RGT technique as a function of turn-on Ar sequence.

Figure 4 .
Figure 4. a) Raman spectra of methylene blue (MB) droplets at concentration of 10 −3 M dried on Si (black), optimized Au thin films (red) and ZrN 20:6 thin films grown via GT technique (purple).b) The comparison of SERS performance between of ZrN 20:6 thin films (purple) and optimize sputtered Au thin films (red) through the dependence of MB concentration on SERS intensity collected at Raman shift of 1632 cm −1 .c) Dependence of excited laser wavelength on the SERS intensity of ZrN thin film measured from Raman shift characteristic at 1632 cm −1 of the MB concentration of 10 −3 M (ZrN 20:6 (purple) and ZrN 50:6 (blue)).
ZrN 20:6 (1.8) and ZrN 50:6 (1.33) and almost stoichiometry regime for ZrN 100:6 (0.95), respectively, we have investigated the dependence of turn-on Ar sequence on a textured coefficient of (200) plane and resistivity of ZrN thin films as shown in Figure 5h.It should be noted that the textured coefficient of (200) plane and the resistivity of ZrN thin films are obtained from GIXRD measurement in Figure 1a

Figure 5 .
Figure 5. Morphology of the ZrN thin films fabricated via RGT technique as a function of turn-on Ar sequence of a) 20, b) 50, and c) 100 s.CAFM current images of the ZrN thin films fabricated via RGT technique as a function of turn-on Ar sequence of d) 20, e) 50, and f) 100 s. g) Surface grain size (red) and RMS roughness (green) collected from AFM images as a function of turn on Ar sequence.h) Dependence of turn-on Ar sequence on texture coefficient of (200) (purple) and resistivity (blue) of ZrN thin films.

Figure 6 .
Figure 6.a) Recyclability performance of ZrN SERs substrate plotted as the intensity of Raman shift characteristic at 1632 cm −1 of the MB concentration of 10 −3 M. b) Stability of ZrN SERS substrate under ambient air condition.c) Reproducibility in ten batches (n = 10) of ZrN SERS substrate plotted as the intensity of Ramana shift characteristic at 440, 1180, and 1632 cm −1 of the MB concentration of 10 −3 M.

Figure 7 .
Figure 7. a) Dependence of TNT concentration on SERS intensity of ZrN thin films.b) The comparison of SERS performance between of ZrN thin films (purple) and optimize sputtered Au thin films (red) through the dependence of TNT concentration on SERS intensity collected at Raman shift of 1359 cm −1 .c) The complete Raman spectral results during the recyclability performance.d) Recyclability performance of ZrN SERs substrate plotted as the intensity of Ramana shift characteristic at 1359 cm −1 of the TNT concentration of 10 −3 M.