1. Introduction

Plasmonic research has enabled the realization of nanoscale optical components and continues to expand the range of applications for plasmonic materials and devices, which includes surface enhanced Raman spectroscopy, bio- and gas-sensing, optical waveguiding, catalysis and photodetection [1–6]. Plasmonic applications requiring high operational temperatures, such as nanotherapeutics, heat-assisted magnetic recording (HAMR), and thermophotovoltaics [7–10], are also of considerable interest. However, typical plasmonic materials such as gold and silver are unsuitable for such high temperature plasmonic applications, due to their poor thermal stability. Refractory plasmonic materials are therefore of increasing importance, presenting the potential to improve the performance and operational lifetimes of high temperature plasmonic devices.

Several studies have investigated refractory plasmonic materials, including traditional refractory metals, such as tungsten, niobium, and molybdenum, transition metal nitrides (TMN), and transition metal oxides [11–13]. Of the transition metal nitrides, titanium nitride (TiN) is of significant interest because of its high melting-point (2930 °C) and has thus been thoroughly studied. In addition to high thermal stability, it enables a spectrally tunable, relatively low loss optical response [14,15]. Although less efficient than gold in terms of optical losses and plasmon lifetimes, TiN has been highlighted as a promising material for thermoplasmonic applications [16]. Regarding its thermal stability, TiN thin films were reported to retain metallic behavior up to 1,200 °C in vacuum [17]. However, when annealed in air, substantial deterioration of the optical response was reported at temperatures above 500 °C as a result of oxidation. This is below the temperature at which the optical response of gold thin films was reported to degrade, with morphological changes reported at temperatures of >600 °C [18]. The difference in degradation mechanism is noteworthy here, as changes in morphology become increasingly significant when materials are patterned on the nano- and micro-scale.

Indeed, just as the range of practical plasmonic applications has increased, so the scale of plasmonic devices continues to decrease. The consistent miniaturization of plasmonic devices has been facilitated by evermore sophisticated characterization methods and fabrication techniques, in addition to the ongoing exploration of new materials platforms [19]. Patterning and testing of alternative plasmonic materials, including refractory TMNs, has become more prevalent as it is essential to determine the suitability and limitations of thermoplasmonic materials when patterned on the nanoscale. In addition to the development of theoretical figures of merit for high temperature plasmonic applications [20,21], recent work has investigated the thermal stability of transition metal nitride arrays based on TiN and NbN [22]. Notably, the work of Gadalla et al. reported greater thermal stability of TiN nano-bowtie arrays compared to gold [23].

Whilst these previous reports investigate the plasmonic response and stability of plasmonic thin films and refractory nanoparticle arrays, to our knowledge, an assessment of the thermal stability of individual nanoparticles has not yet been completed. In previous studies of nanoparticle arrays the stability with temperature was assessed by averaging over an ensemble of many nanoparticles. In contrast, in this work we investigate the stability of individual nanodiscs and track the evolution of the optical and morphological changes of the same nanoparticle with temperature. Such an investigation of the stability of individual nanodiscs is made possible by our use of cathodoluminescence (CL) spectroscopy and is particularly relevant when considering applications that require discrete plasmonic components. For instance, HAMR heads typically utilize a plasmonic nanoparticle to focus light into a sub-wavelength region to locally heat a magnetic medium, and hence it is critical that it is thermally robust [9].

In this work, we investigate the temperature stability of nanostructured plasmonic materials at the level of individual nanodiscs, tracing both the shape modification and the optical response modification of the same nanostructure utilizing cathodoluminescence (CL) spectroscopy. In contrast to smooth plasmonic films, for which primarily the modification of the optical constants with annealing is important, for nanostructures, both morphology and optical constants can be modified at high temperatures. The electron beam of the CL microscope can be readily used to directly excite plasmonic modes with nanoscale resolution. Collection of the light emitted upon excitation can be used to identify plasmon resonances and hybrid modes, and also allows for the polarization and angular emission profile to be resolved [24,25]. This technique has successfully been used to measure emission from nanoantennas and individual metal nanoparticles [26–28].

Here, we compare the thermal stability of single nanoscale titanium nitride and gold discs, patterned on silicon substrates. Samples are annealed in air for 30 min each at temperatures ranging from 150 °C to 325 °C. Following each annealing stage, CL spectra and SEM images are collected to assess variations in the optical response and disc morphology. Experimental spectral data for the CL response of plasmonic discs are supplemented by simulations, identifying the excited plasmonic modes and confirming the oxidation-related mechanism of the change in the optical response of TiN nanodiscs with the annealing temperature.

2. Methods

2.1 Sample fabrication

Individual discs with diameters ranging from 150 nm to 450 nm and heights of 50 nm were patterned from gold and titanium nitride using electron beam lithography (EBL). Au and TiN nanodiscs measured in this study had diameters of 175, 220, 310 nm and 170, 260, 350 nm, respectively. Gold discs were prepared by thermal evaporation and patterned using a lift-off method. TiN discs were patterned using a reactive ion etching-based method [29,30]. A complete description of the fabrication processes is included in the Supplementary Information. In brief, TiN nitride thin films (50 nm) are deposited by RF magnetron sputtering, coated with a Cr hard mask prepared via EBL patterning and lift-off, and then etched with chlorine based Reactive Ion Etching (RIE). The Cr mask is then removed with Cr etchant solution revealing the TiN nanodiscs.

2.2 Annealing study

A thermal annealing study was completed to assess the thermal stability of individual TiN discs in comparison with gold discs. For each material, SEM micrographs and CL spectra were initially collected for discs of comparable diameter using a Tescan-SEM Delmic-cathodoluminescence instrument. When collecting CL data, a 30 keV electron beam is rastered over an individual disc, exciting modes which subsequently radiate into the far-field. This emitted light is collected using a parabolic mirror and directed to a spectrometer and CCD camera. Substrate background emission is corrected for in the CL spectra, which are spatially averaged for each disc.

For the annealing study, the samples were then heated in air for 30 min before collecting further CL spectra and SEM images. Samples were sequentially heated at 100, 200, 300, and 325 °C. Following annealing, spectra and micrographs were collected from the same disc after each anneal, to ensure that any variations in the disc diameter and morphology arising from the fabrication process did not influence the spectra obtained. This also allows specific variations in the disc shape and spectra to be monitored. A schematic of the CL measurement process is included in Fig. 1. The spatial variation in the CL spectra is evident from Fig. 1(b), with further insight provided by simulations.

 

Fig. 1. Schematic of CL measurements. (a) E-beam irradiates the nanodiscs, with the resulting CL collected by a mirror and directed into a spectrometer. Inset: SEM image of an Au nanodisc (b) CL spectra, normalized to 1, for the e-beam placed at the edge (red line) and center (blue line) of an Au nanodisc (not annealed) with a diameter of 280 nm. Inset: Corresponding CL map at a wavelength of 580 nm.

Download Full Size | PDF

2.3 Numerical simulations

Numerical modeling of the cathodoluminescence signal from Au and TiN nanodiscs was performed using a finite element method (COMSOL Multiphysics software). Generally, there are two mechanisms of cathodoluminescence, coherent and incoherent, but for plasmonic nanostructures, a coherent mechanism is dominant [31–33]. To model it, the CL source was implemented in simulations as a vertically-oriented point dipole at the nanodisc interface where the e-beam enters the structure, representing a collapsing dipole produced by the electron and its image charge. The dipole source was set to have a 1/ω frequency dependence obtained from a Fourier transform of the time-space trajectory of the moving electron in the beam. Five positions of the dipole along the nanodisc radius were studied, accounting for a CL signal from ring-shaped (or circular in the case of the central region) areas, altogether producing the overall measured signal (see Supplementary Information). The areas of these regions were set to be equal, while the position of the dipole inside each region was set in such a way that a circle passing through it divides the region into equal areal parts. The simulation domain was surrounded by perfectly matched layers to ensure the absence of back-reflection from the outer boundaries. The produced CL signal was integrated over a hemisphere surrounding the structure on the air side. The optical properties of Au, titanium oxynitride (TiON), and Si were taken from [34–36], respectively, while for TiN data collected using spectroscopic ellipsometry were used.

3. Results and discussion

3.1 Annealing study

Figure 2 displays the CL spectra of individual TiN and Au discs with respective diameters of 170, 220, 350 nm and 175, 260, 310 nm, collected throughout the annealing cycle. The cathodoluminescence spectra collected at room temperature (blue curves) display broad resonance peaks at ∼550 nm for Au and at ∼550-600 nm for TiN. The observation of TiN peaks at longer wavelengths than Au is expected as the plasmon resonances for TiN occurs at lower energy than Au, consistent with the literature [20,37,38]. This is further corroborated by spectroscopic ellipsometry measurements for TiN and Au thin films (Fig. S3.), in which the screened plasma energy for TiN thin films is observed at longer wavelength than those of Au. Additionally, the CL peaks for TiN peaks are broader than the Au peaks, which can be attributed to the higher losses.

 

Fig. 2. Spatially averaged CL spectra and SEM images collected during the 30 min annealing study for (a,c,e) Au discs with diameters 170, 260 and 350 nm and (b,d,e) TiN discs with diameters 175, 220 and 310 nm. Each spectrum is normalized to 1 independently and vertically offset for visibility.

Download Full Size | PDF

SEM images of each disc, collected at RT and after each annealing step, are shown in the insets of Fig. 2 and in Figs. S4-S5. Upon inspection, it is observed that the gold discs of all diameters display obvious changes in structure and morphology from ∼300 °C. The discs become deformed, likely due to melting and redistribution of the Au. The change of morphology in the Au nanoparticles is observed systematically in repeated experiments and is in agreement with the degradation mechanism previously reported for gold thin films, in which increased surface roughness results in film discontinuities. This increased surface roughness leads to greater scattering of the conduction electrons on the nanostructure surface and an increase in the magnitude of the imaginary dielectric permittivity, and, therefore, greater losses [39]. The previously reported deformation temperature for gold thin films in ambient conditions is 600 °C. Beyond this temperature the surface morphology of the films changes significantly, resulting in a loss of connectivity across the film and deterioration of the optical properties [18]. In contrast, the TiN discs retain their shape to 325 °C with no obvious changes in morphology or the disc shape. Indeed, TiN bowtie nanoantennas were previously reported to retain their shape when heated in air for 10 min at temperatures up to 800 °C with little change in microscopic appearance [23]. In a thin film form, TiN was reported to display a far less pronounced increase in roughness compared to gold when heated for one hour in air, so this lack of variation in surface morphology of the patterned discs is not unexpected [18].

Although the TiN discs do not display any deterioration in morphology or shape at elevated temperatures, there are significant changes in the CL spectra for both TiN and Au discs from 200 °C for all disc diameters measured (Fig. 2). For Au, we attribute this to changes in surface morphology observed in the SEM micrographs. However, in the TiN case, the spectral evolution is not due to morphological changes as evidenced by the SEM images. As the samples are annealed in air, it is likely that the degradation of the optical response arises from sample oxidation. This is consistent with the degradation mechanism previously reported for TiN thin films and TiN nanoparticle arrays, as discussed above. In our case, we see broadening of the out-of-plane resonant mode, as would be expected with a reduction in metallic character. The rate at which the plasmon peak broadens is dependent upon the extent of oxidation within the nanoparticles after annealing. For smaller TiN discs, the loss of the plasmon resonance peak at 325 °C is indicative of complete oxidation of the sample. Larger disc diameters retain some degree of plasmonic character at 325 °C, however, peak broadening and redshift of the resonant peaks in both the 220 nm and 350 nm TiN discs indicates significant oxidation of the samples.

3.2 Simulations

To understand the measured CL spectra, the mode profiles of Au and TiN nanodiscs were found performing numerical simulations with plane wave excitation for both s and p polarizations. This approach is more convenient, as in the direct simulation of the CL process the mode profiles are significantly disturbed by the point-like excitation source (see Methods). The obtained extinction spectra are presented in Fig. S9 in the Supplementary Information. The plasmonic mode observed in the CL experiments can be identified as an out-of-plane dipolar mode, which could be excited in the simulations only by the p-polarized illumination. The nature of the mode was additionally confirmed by plotting the corresponding charge maps on the surface of the nanodiscs (see insets, Fig. S9). In the case of TiN, the mode is red-shifted and broadened in comparison to Au, which agrees well with the experimentally observed CL results and indicates less metallic behavior of the material together with somewhat higher losses. In-plane resonant modes excited by s-polarized illumination found in the simulations are barely observed in the experimental CL spectra presumably due to a symmetry mismatch with the excitation CL source, which can be represented as vertically-polarized collapsing dipole.

After the identification of the peaks present in the CL spectrum, we proceeded with the modeling of the CL measurements themselves with a particular goal of checking the hypothesis that the modification of the spectrum with temperature in the case of TiN nanodiscs is related to their oxidation. Two possible scenarios of the oxidation process were considered. One is gradual expansion of a TiON 25% oxidized layer from the surface towards the nanodisc interior (which is assumed to be 10% oxidized already during the fabrication process), and another is uniform oxidation of the entire nanodisc characterized by the increase of the oxygen content from 10 to 40%. The results are presented in Fig. 3(a) and (b), respectively. Both oxidation models result in the red-shift and broadening of the plasmonic peak, correctly reproducing the experimental trends and thus supporting the oxidation-related interpretation of the change in the CL spectra with temperature. It can be logically assumed that the actual oxidation process is a cooperative action of the above model scenarios when the increase of the oxygen content and deeper penetration of oxygen atoms into the interior happen simultaneously.

 

Fig. 3. Simulations of a coherent CL signal from (a) TiN/TiON core/shell disc, diameter 260 nm, with increasing TiON shell thickness and (b) TiON disc, diameter 260 nm, with increasing uniform oxidation of the disc. Spectra are vertically offset for visibility.

Download Full Size | PDF

3.3 Discussion

From the experimental annealing study, we observe that TiN discs display greater morphological stability when compared to Au discs. However, both materials display deterioration of the optical properties and plasmonic response from 200 °C. Mechanistically, the optical degradation for Au and TiN is distinct, resulting from morphological changes and oxidation, respectively. This agrees with the previously reported degradation mechanism for Au and TiN plasmonic thin films. Additionally, this diminishment of the plasmonic character occurs at lower temperatures than reported for plasmonic thin films based on Au and TiN, which is expected for nano-patterned features due to the increased surface-volume ratio. Although the deterioration in plasmonic behavior reported here occurs at significantly lower temperatures than that of TiN nano-bowtie arrays (c.f. 200° and 600°C) [23], this can be explained by the significantly shorter annealing time of 10 min used in that work, compared to the 30 min anneal used here (see Supplementary Information). The degradation of optical properties at relatively low temperatures, therefore, indicates that plasmonic devices containing bare TiN nanoparticles are unsuitable for extended high-temperature operation in ambient conditions.

To overcome this, the application of small-scale TiN-based plasmonic components requires vacuum conditions, inert gas environments, or the use of an encapsulation layer to limit oxidation, as has previously been suggested [17,40,41]. However, similar encapsulation layers are also reported to improve the thermal stability of gold nanoparticles [42–44]. In this sense, the advantages of using TiN rather than gold are limited if both require a protective layer yet Au displays superior plasmonic properties, with lower loss and greater lifetimes. Nevertheless, the benefit of using encapsulated TiN rather than gold could arise from its spectral tunability and CMOS compatibility of the TMN. Additionally, TiN may be better suited than Au to specific plasmonic applications, such as local heating [20,29].

4. Conclusions

We have examined and reported the thermal stability of individual nanoscale TiN discs in comparison with gold. SEM imaging and CL spectroscopy indicate that although TiN displays superior thermal stability in comparison with Au when considering morphological changes and disc deformation, deterioration in the optical properties of both materials occurs at comparable temperatures. The degradation of individual discs occurs at significantly lower temperatures than previously reported for plasmonic thin films. From these results, it is apparent that oxidation of TMNs remains a limitation to the integration of TiN components within plasmonic devices to be used in ambient conditions. However, we suggest that TiN could display improved performance compared to gold when used in inert or vacuum environments or when coated with an optically transparent protective layer, particularly when used for targeted plasmonic applications such as local heat generation.