1. Introduction

Owing to the classical or quantum electron confinement, nanorod (NR) possesses peculiar optical properties that is different from the corresponding bulk material [1, 2]. Collective excitation of electrons in metal NRs gives rise to the (localized) surface plasmon resonance (SPR) phenomenon, which could produce a strong resonant optical absorption as well as near-field and scattering enhancements [2]. These optical features of metal NR have been widely applied to surface enhanced Raman scattering (SERS), photothermal therapy, bio-sensing [3–7], etc., and even enable it be a promising component for quantum information devices [8]. The SPR spectra of metal NR is strongly sensitive to several structural parameters, such as the NR’s size, shape, and environment, in addition to the optical nature of the metal [9, 10], which can be used to tune or even design the SPR properties of metal NRs [10]. Moreover, when coating semiconductors on the surface of metal NR, its SPR would co-exist and interact with the semiconduction, providing the opportunity to improve the performances of opt-electrical devices, including solar cells, photodetectors and so on [11]. So far, most optical studies about metal NRs have been devoted to Ag and Au because of their strong SPR signals in the visible or near-infrared [12]. However, to some extent, their applications are still hindered by the high cost of the noble metals. As a cost-effective choice for SPRs, Nano-bismuth becomes increasingly attractive in recent years [13, 14]. It shows satisfying excitation efficiency that is close to Au in the spectral range corresponding to the interband transitions of Au [14], thus might be a potential substitute of Au to produce moderately-damped SPRs in this range [14]. More importantly, the SPRs of nano-bismuth is tunable over wide wavelength range [14]. Therefore, in principle, its SPRs can be flexible-designed to match the band gap of various semiconductors, enhancing their opt-electrical performance.

The nanodot (ND) of semiconductor is quite valuable because its band gap can be tuned by quantum confinement over wide range [15]. Hence, it is perspective that the hybridized structures based on Bi NR with covered semiconductor NDs could be a promising candidate for opt-electrical devices with wide spectra response. Usually, the synthesis and coating of the NDs over the NRs is carried out in solution phase via chemical bath deposition or successive ionic layer adsorption and reaction [16–18], etc. In these routes, the existence of the surface stabilizers or contaminants may affect the performance of the final products negatively. For the fabrication of coated NDs with-the most important semiconductor nowadays-silicon [19], thermal annealing of SiO_x (x<2) covered on NRs could be an alternative method to overcome the disadvantages of solution-phase routes [20]. It directly induces the solid-phase separation from SiO_x to Si and SiO₂ in nanoscale and thus forms Si NDs. However, the high temperature (>750 °C) during the annealing is still not desired. Hence the issue of this work is exploring the surface stabilizers-/contaminants- free and low-temperature route to prepare Bi-NR/Si-ND hybridized structure (BSHS) and investigating its SPR properties.

Surface dewetting is an interesting phenomenon which could produce apparent surface/interfacial tension and thus spontaneously nano-pattern the metal or polymer [21, 22]. In this work, utilizing the dewetting induced by poor inter-solutability of Bi-Si system, we develop a route based on physical vapor deposition (PVD) to prepare BSHS, which is realized simply by alternated depositing Bi and Si at low rate. The vacuum-growth process of NR/ND hybridized structure avoids the shortcomes of solution-phase methods produced by surface stabilizers or contaminants. Also its operation temperature (200 °C) is much lower than that of annealing SiO_x. The SPR spectra of BSHSs are numerical-studied over the wavelength from near ultraviolet to near infrared range under transverse and longitudinal excitation. Comparing with ordinary Bi NRs, the SPR modes of BSHS are shifted and splitted because of the covered Si-NDs. The longitudinal resonance in near infrared range could match the band gap of silicon. High order mode of longitudinal resonance is observed in BSHS. The size dependence of longitudinal resonance of BSHS is unable to be explained by classical Gans theory, but successfully understood by analogizing the longitudinal excitation of electrons in BSHS to the resonance of a LC circuit whose parameters are extracted from the size of BSHS, in addition to the dielectric function of Bi and Si. The size dependence of SPRs in BSHS provides us a guideline to tune its resonance in the whole near-ultraviolet, visible, and near-infrared range. Considering the unique properties of bismuth as well as the great importance of the silicon in semiconductor industry, BSHS should be useful in sensors, thermoelectric, photodetector, photocatalyst, battery electrodes, etc [23–25].

2. Experimental details

The PVD growth of BSHS was carried out on magnetron sputter (ACS-4000-C4) by alternated depositing Bi and Si at 200 °C with the durations of 600 s and 200s, respectively, for four cycles. During the deposition of Bi, the sputter power was 5 W. 20 sccm of Ar and 10 sccm of N₂ were vented as work gas and scattering gas, respectively, providing a low deposition rate of 0.04 nm/s [23, 26]. While in the Si-deposition with a sputter power of 30 W, only 20 sccm of Ar was vented, and accordingly, the deposition rate is 0.01 nm/s. Through these processes, the BSHS with SPRs matching the band gap of Si (E_g = 1.12 eV) could be fabricated, while the route of directly depositing Bi for 2400 s and then Si for 800 s would produce the final products whose resonant energy become lower than E_g owing to their larger size [27]. As a comparing, purely bismuth deposition without inserted silicon-deposition was performed under similar condition (Ar/N₂: 20/10 sccm; sputter power: 5 W and substrate temperature: 200 °C) with the deposition times ranging from 900 s to 4800 s. On the other hand, to study its initial nucleation stage of the deposition, which should be closely related to the wetting properties of Bi-Si system [28], an non-fully-covered bismuth film was prepared on silicon surface by depositing the bismuth on the silicon wafer for a shorter time of 300s with the sputter power of 50 W, Ar/N2 flow of 25/15 sccm.

The thicknesses of the sample were measured by stylus profiler (Vecco Dektak150) to estimate the deposition rate. The morphology and crystal structure of the samples was studied by scanning electron microscopy (SEM, Hitachi S-4800), transmission electron microscopy (TEM, FEI Tecnai G² F-20) and high resolution TEM (HR-TEM, FEI Tecnai G² F-20) and Energy Dispersive X-ray spectra (EDX). To investigate the SPR properties of BSHS, we numerically calculate their extinction, absorption and scattering cross-section over the wavelength from 300nm to 1700 nm by finite difference time domain (FDTD) method. The optical constant of Si and Bi is obtained from Palik and/or determined by spectroscopic ellipsometry [14,29].

3. Results and discussion

3.1. Morphology and structure of BSHS

Figures 1(a) and 1(b) show the products of alternated-deposition of Bi and Si. Unlike a multilayer thin film regularly obtained from the alternated deposition of two different materials, several rod-like products appear on the surface of the sample as shown in Figs. 1(a) and 1(b). The magnified SEM image presented in the inset of Fig. 1(b) shows individual rod-like product is actually a NR/NDs hybrid structure [23, 26]. The size distribution of the hybrid structure shown in Fig. 1(c) presents that most samples range in the diameter of 54~90 nm with aspect ratio (AR) 1.62~5.3. The statistics to their size shows that the mean diameter (d), length (l) and AR are 70.8 ± 17.5 nm, 238.8 ± 128.8 nm and 3.46 ± 1.84, respectively.

 

Fig. 1 (a) schematic illustration and (b) SEM images of BSHSs formed on the substrate, and the inset of Fig. 1b is the morphology of single BSHS ; (c) statistics of the size of BSHS.

Download Full Size | PDF

Figure 2(a) shows the TEM image of BSHS with flat endcap and covered by NDs whose typical size is ~10 nm or more. EDX results shown in Fig. 2(b) reveal that the BSHS contains Bi and Si, while the Cu and C signal originates from the copper grid with amorphous carbon film, which supports the sample. The BSHS is further studied by the electron diffraction (ED) as shown in Fig. 2(c). Although the ED patterns might be difficult to index unambiguously because of the overlap of the diffractions from both NR and NDs, and contributed by either Bi or Si, they are basically in accordance with the X-ray diffraction (XRD) data from ICDD (international centre for diffraction data) database, verifying the composition of BSHS is Bi and Si as revealed by EDX. To understand the crystalline nature of BSHS, we study the structure of ND and NR, respectively. Figure 2(d) is the HRTEM image of the ND, which clearly shows its lattice fringe of 2.23 angstrom with hexagonal-symmetry, agreeing with the atomic model of Si single crystal when view along its [111] direction as shown in Fig. 2(e). Accordingly, in the fast Fourier transfer (FFT) of HRTEM image presented in Fig. 2(f), six hexagonal symmetrical spots are indexed as $(1\overline{2}1),(2\overline{1}\overline{1}),(11\overline{2}),(\overline{1}2\overline{1}),(\overline{2}11)\text{and}(\overline{1}\overline{1}2)$ planes of silicon, respectively. These results indicate that NDs are composed of silicon. Accordingly, the NR serving as the inner-core must be the bismuth because the Bi-Si phase diagram shows Bi and Si are highly un-intersoluable below ~770 °C, and thus tend to separate with each other in our deposition condition (see Fig. 3(a), SGTE2007 alloy database), resulting in the formation of BSHS. Such feature allows us to deduce the structure nature of the NR indirectly through the studies to the formation mechanism of BSHS, which will be discussed latter in section 3.2.

 

Fig. 2 (a) TEM (b) EDX and (c) ED results of BSHS; (d) HRTEM of Si ND on the surface of BSHS; (e) atomic structure of Si view along [111] direction; (f) FFT of HRTEM image of Si ND with indexed lattice plane. (Note: the lattice spacing of (009) plane of Bi is very close to that of (300) plane, and the lattice spacing difference between (306) and (223) planes of Bi is also rather few. On the other hand, the diffraction from (220), (400) and (422) planes of Si could be respective overlapped with (202), (018) and (306)/(223) planes of Bi, etc.)

Download Full Size | PDF

 

Fig. 3 (a) Bi-Si phase diagram; (b) temporal evolution of the growth of Bi NRs; (c)schematic illustration to the formation of BSHS ; (d) initial stage of the deposition of Bi on Si.

Download Full Size | PDF

3.2. Formation mechanism of BSHS

It is amazing that the alternant deposition of Bi and Si could produce the BSHS, highly differing from the multilayer structure which is usually formed in the alternant deposition process [30]. Hence, the formation mechanism of BSHS is of significance to study. Here the morphology and structure features of BSHS imply that its formation in the alternant deposition of Bi and Si might be originated from the co-impacts of the preferential 1D growth of bismuth and the generation of Si-dots on the surface of bismuth [30]. To verify this assumption, we investigate the role of Bi deposition and Si deposition by comparing the aforementioned alternant deposition with the case of purely depositing Bi. Figure 3(b) reveals that, in this case, as the deposition time increases, the products gradually turn from Bi particles to NRs. The low-magnified SEM images show the morphology of the rod-like products is quite similar to that of alternant deposition, especially the case of depositing for 2400s, which equals to the equivalent bismuth-deposition time in the aforementioned alternated deposition ( = 4 × 600s). Such similarity implies that the mechanism of driving the preferential 1D growth of bismuth during alternant deposition of Bi and Si should be the same as that takes effect in purely depositing Bi, although unlike BSHS, the surface of Bi NRs is smooth as shown in the high-magnified SEM images and the size is relative bigger with typical diameter of ~100 nm and length of 3~5 μm (see Fig. 3(b), and note the cases of depositing for 2400s, 3000s and 4800s). This mechanism has been known as the anisotropic corner crossing (ACC) of Bi atoms [23, 26], which induces the preferential growth of Bi along [012] direction. Hence we argue that the inner NR of the BSHS originates from ACC-induced 1D growth of Bi is preferentially along [012] direction during the Bi-deposition step as shown in Fig. 3(c). Furthermore, owing to the low solubility between Bi and Si, the deposited Si atoms would be poor wetting on the surface of Bi and thus become difficult to sprawl over the NR but shrinking to separated dots as illustrated by Fig. 3(c), which has been also observed in several other systems [28]. Therefore, BSHS could form in the alternated deposition of Bi and Si as shown in Fig. 3(c). Additionally, the dewetting in Bi-Si system can further explain that the ACC process of Bi atoms could still work in alternant deposition of Bi and Si. Exactly, the Bi atoms, if deposited on the surface of Si NDs, will rapidly move away from it and onto the surface of the NR because the dewetting indicates that they are difficult to attach on the surface of Si. Hence, the inserted Si deposition should not forbid (but does obviously disturb) the 1D growth of Bi driven by its anisotropic surface diffusion [23], resulting in relative smaller size of BSHS in compare with Bi NR.

The surface dewetting in Bi-Si system can be further identified by checking the initial nucleation stage of the bismuth thin film deposited on the silicon wafer. As we known, the poor wetting of the materials deposited on the substrate would induce a spherical nucleation and Volmer-Weber growth [28], whereas the well-wetted substrate is beneficial to deposited materials to spread over the surface and form a relative smooth film. Figure 3(d) shows the initial stage of the bismuth-deposition on Si wafer follows Volmer-Weber growth mode [28], implying the bismuth is poor-wetting on the Si surface. Moreover, it is discovered recently that because of the poor wetting in Bi-V system, together with the porous surface structure, single-crystalline bismuth nanowires could vertically grow on nanoporous vanadium thin film even at room-temperature via spontaneous and continuous expulsion of nanometer-sized bismuth domains from the vanadium pores [31]. Considering the Bi-V phase diagram is similar to that of Bi-Si system in our deposition temperature [31], this result also suggests that the dewetting in the Bi-Si system could be an non-ignorable factor contributing to the formation of BSHS.

3.3. Plasmon resonance of BSHS with typical size (d = 70.8 nm and AR = 3.46)

It is expected that the fabricated BSHS could possess interesting SPR properties. However, the optical response corresponding to the BSHS with a certain size seems difficult to obtain experimentally via the transmittance or reflectance spectra of as-prepared sample [14], owing to the interference of the Bi-Si mixture layer and some particles over the surface as shown in Figs. 1(a) and 1(b), as well as the relative broad distribution of the size of BSHS on the surface of the fabricated sample. A more reliable approach might be transferring the BSHSs from as-prepared materials to a “clean” substrate like fused silica, silicon, etc., and then probing the spectrum of individual BSHS by techniques like dark field scattering microspectroscopy [32], but the absence of transfer routes is still a limitation. As a compromise, we numerically study the SPR properties of BSHS via FDTD simulation [33]. Figure 4 shows that the SPR properties, including the extinction, absorption and scattering cross-section (σ_ext(λ),σ_abs(λ) and σ_scat(λ)), of typical BSHS (d = 70.8 nm, AR = 3.46, and in surrounding medium with refractive index 1), is polarization-dependent as illustrated by Figs. 4(a) and 4(b). Although, as observed in SPRs of noble metals [33], there are slight difference on the detail sites of the resonant peaks which is derived by σ_ext(λ),σ_scat(λ) and/or σ_abs(λ) (see Figs. 4(c)-4(h), solid lines), in Figs. 4(c)-4(e) we can see that the peak around 380~390 nm is produced by light polarized perpendicular ($\perp$) to the longitudinal axes, while the light parallel ($\parallel$) the longitudinal axes stimulates two strong peaks at 876.3 nm and 941.4 nm as well as a weak one at 385.2 nm (see Figs. 4(f)-4(h), and the inset of Fig. 4(g) presents the details of the spectra from 800 to 1000 nm). To reveal the origin of these peaks, SPRs of BSHS are compared with that of Bi NR with the same size (in the same surrounding medium with index of 1, but without covered Si-NDs). As shown by dash line in Figs. 4(c)-4(h), Bi NR has a supposed transverse SPR (TSPR) mode near (maybe less than) 300 nm and a longitudinal SPR (LSPR) mode at 807 nm with intensity distributions of the electrical field respective plotted in Figs. 5(a) and 5(b). It can be found that in compare with the tail of the supposed sub-300 nm transverse resonance contributed by Bi NR itself, the (net) intensity of the peak emerging in BSHS around 380~390 nm is quite weak. Hence we suggest its generation should be still dominated by the strong “tail” of transverse resonance (STTR), which exactly exists in both Bi NR and BSHS at sub-300 nm. This point is also supported by the similarity of the field profile of BSHS at 385.2 nm shown in Fig. 5(c) to that of Bi NR at 300 nm shown in Fig. 5(a). On the other hand, the peaks of BSHS in near infrared region (876.3 nm and 941.4 nm) under parallel stimulation as shown in Figs. 5(c)-5(e) have the field profiles with the typical feature of LSPR: the field concentrated in its two end-faces. Therefore, it is rational to ascribe them to the LSPRs of BSHS. These results show two noteworthy features of SPRs in BSHS: (1) Mode shift: LSPR of BSHS red-shifts, in respect to that of Bi NR. This is induced by the variation of the effective medium dielectric constant (ε_m) [27], which originates from the covered Si-NDs shown in Fig. 5(f); (2) Mode split: Comparing with Bi NR, LSPRs of BSHS split to two modes (${\lambda }_{{\text{LSPR}}_1}$ = 876.3 nm and ${\lambda }_{{\text{LSPR}}_2}$ = 941.4 nm). This is because the Si-NDs randomly distribute over its surface. Such randomness breaks the mirror-symmetry along the transverse axes and thus lifts the degeneracy of LSPR mode originally existing in Bi NR [34, 35]. As a result, the LSPR mode splits as shown in Fig. 5(g).

 

Fig. 4 (a) schematic illustration of perpendicular-polarized stimulation and (b) schematic illustration of parallel-polarized stimulation; (c) extinction, (d) absorption and (e) scatter cross section of BSHS (solid line) and Bi NR (dash line) under perpendicular-polarized stimulation; (f) extinction, (g) absorption and (h) scatter cross section of BSHS (solid line) and Bi NR (dash line) under parallel-polarized stimulation.

Download Full Size | PDF

 

Fig. 5 Electrical field intensity distribution of SPRs of BSHS/Bi NR: (a) STTR of Bi NR under perpendicular-polarized stimulation; (b) LSPR of Bi NR under parallel-polarized stimulation; (c) STTR of BSHS under perpendicular-polarized stimulation; (d) LSPR₁ and (e) LSPR₂ of BSHS under parallel-polarized stimulation; Schematic illustration for the generation of (f) shift and (g) split of SPR modes in BSHS. (The black arrows represent the direction of the electrical field of stimulation light; the red arrows represent the direction of the electrical field of SPR).

Download Full Size | PDF

However, sum up, the origin of the weak resonance at 385.2 nm under parallel stimulation is still not clear. We suggest that it could be the 3rd longitudinal mode of BSHS and explained within the physical picture of Fabry-Perot (F-P) resonance. In our case, since the incident light is normal to BSHS, even modes are forbidden by conservation of the parity symmetry [36]. Hence the 3rd longitudinal mode should be the first observable high-order resonance in the spectrum. According to the phase matching condition of F-P resonance [37], the resonant wavelength λ_m (m is the order value) is roughly dependent on the effective mode index n_eff (λ) and the length l:

 

Fig. 6 (a) The field profile of 3rd longitudinal resonance between BSHS and Bi NR (the black arrows represent the direction of the electrical field of the stimulation light, the red arrows represent the direction of the electrical field of SPR. The field profile near the nodes (marked by red rectangular boxes) in BSHS is replotted with tuned color bar to provide better eyes-view); (b) extinction, absorption and scattering cross-section of Bi NR in the background refractive index of 1.35.

Download Full Size | PDF

3.4. Tunability of the SPRs of BSHS

As we known, the SPRs of noble metal NR are sensitive to their size [27]. Here we explore the tunability of the SPRs of BSHS by controlling its AR and diameter. Figure 7 shows that, whatever under perpendicular- or parallel-polarized stimulation, the increase of diameter (AR) of BSHS at fixed AR (diameter) would produce stronger extinction as well as absorption and scattering cross-section. The enhancement of σ_ext induced by the increase of diameter or AR can be understood by Gans theory, which predicts an extinction cross-section in proportional to the volume of NR (V = π × diameter³ × AR/4) [27]:

 

Fig. 7 FDTD simulation to SPRs of BSHS. (a) extinction, (b) absorption and (c) scattering spectra depending on AR under perpendicular-polarized stimulation (the diameter is fixed as 71 nm); (d) extinction, (e) absorption and (f) scatter spectra depending on diameter under perpendicular-polarized stimulation (AR is fixed as 3.46); (g) extinction, (h) absorption and (i) scatter spectra depending on AR under parallel-polarized stimulation (the diameter is fixed as 71 nm); (j) extinction, (k) absorption and (l) scatter spectra depending on diameter under parallel-polarized stimulation (AR is fixed as 3.46).

Download Full Size | PDF

Besides AR- and/or d- dependence of the extinction, absorption or scattering cross-section, Fig. 7 also indicates that the variation of AR and d could shift the resonance wavelength. At fixed d ( = 71 nm), the increase of AR (from 1.62 to 5.30) could slightly blue-shift the maximum of STTR band as shown in Figs. 7(a)-7(c) and remarkably red-shift the maximum of LSPR band as shown in Figs. 7(g)-(i), which can be seen more clearly in Fig. 8a. The contour plot presented in Fig. 8(a) reflects that the red-shift (blue-shift) of LSPR (STTR) band can be also derived from the spectra calculated based on aforementioned Gans theory (Eq. (3) at different AR from 1.5 to 5.5 [27]. However, the quantitative accordance between FDTD results and Gans theory is only achieved at STTR band, whereas the red-shift-rate of LSPR predicted by Gans theory is obviously smaller than the FDTD simulations, resulting in the discrepancy between the maximum of LSPR band extracted from FDTD results and Gans theory, especially when AR is large. This can be ascribed to the co-impacts of the breaking of the quasi-static approximation at large lengths and the flat endcap of BSHS. On one hand, as a non-retarded description of the SPRs of nanorod with dipolar approaching [39, 40], Gans theory seems inevitable to underestimate the resonant wavelength of LSPR of the rods with larger length (AR), because in these situations, the retardation effect, as well as the contribution of the higher order multipoles would become non-ignorable [39, 40]. On the other hand, Gans theory is actually an analytical expression to the extinction of ellipsoids, thus could just approximately explain the plasmon resonance of the nanorod with hemispherical endcaps [39], but here the endcap of BSHS is flat. The difference of the endcap shape can also impact the resonant wavelength of LSPR [41], causing apparent red-shifts from where one would expect the resonance based on AR.

 

Fig. 8 SPR wavelength of BSHS depending on (a) AR and (b) diameter predicted by Gans theory (dash line, extracted from the contour plot is the SPR spectra of BSHS with AR from 1.5 to 5.5) and/or LC model (solid line). The results of FDTD simulation (symbols) are presented as a comparison.

Download Full Size | PDF

The diameter is another factor impacting the resonance as shown in Figs. 7(d)-7(f) and Figs. 7(j)-7(l). At fixed AR ( = 3.46), the larger diameter of BSHS would produce apparent red-shift of LSPR (from 748 nm to 1005.3 nm), while the wavelength of STTR is almost a constant as shown in Fig. 8(b). Similar to the case of AR-dependence, the feature of STTR band (insensitive to the diameter) can be well explained by Gans theory which implies that the wavelength of the (transverse) resonance (and accordingly its strong “tail” in our cases) is only the function of AR and the medium permittivity [27]. However, the fact that Gans theory is independent on diameter indicates the red-shift of LSPR produced by the increase of diameter is totally out of its framework.

To break the deficiency of conventional Gans theory on revealing the red-shift of LSPR induced by the increase of AR and/or diameter, we employ the LC model developed by Huang, et.al [42], which analogies the longitudinal mode of the metal rod to the resonance of the LC circuit whose electrical parameters are described by the size of BSHS as well as the dielectric functions of Bi and Si [42]. In this model, λ_LSPR is dependent on both AR and diameter (d):

4. Conclusion

In summary, the high crystalline Bi-nanorod/Si-nanodot hybrid structure (BSHS) has been directly grown via alternated sputtering of Bi and Si. The formation mechanism of BSHS results from the co-impacts of the preferential 1D growth of Bi induced by low rate bismuth-deposition and generation of the Si nanodots on the surface of Bi owing to the dewetting in Bi-Si system. The numerical study shows BSHS could be promising for SPR enhanced opt-electrical applications. Considering the rich and unique properties of bismuth and the extreme importance of silicon in modern electronics and semiconductor industry, the BSHS could be useful in thermoelectric, magnetic-sensor, battery electrodes, photodetector, etc.