Engineered telecom emission and controlled positioning of Er³⁺ enabled by SiC nanophotonic structures

2.1 Fabrication of self-aligned ultrathin NW arrays

Self-aligned 1D NW arrays grown at predetermined positions were fabricated on silicon substrates via thermal CVD. An abridged fabrication process of the NW-PC is schematically depicted in Figure 1. First, a hydrogen silsesquioxane (HSQ) negative-tone resist layer was spin-coated onto Si. We used electron beam lithography to expose line patterns, and the resulting wafer piece was developed in a chemical solution bath yielding an HSQ ribbon array (Figure 1A-i) [24]. After the development process, we conformally deposited ≤20-nm-thick SiC (with an index of refraction, n=2.7) or SiC doped with oxygen, SiC:O (SiC_0.5O_0.8, n=2.2), onto the HSQ template using CVD processing (Figure 1A-ii), described elsewhere [25], [26].

Figure 1:

SiC nanostructured materials through on-demand placement of NW and Er³⁺.

(A) Fabrication process steps for realizing self-aligned 20-nm-width NW-PC structures and representative cross-section SEM images. i. Spin-coating and electron beam lithography of HSQ (blue) to create a ribbon template on a substrate (gray). ii. Self-aligned deposition of conformal SiC or SiC:O ultrathin layer (green). Geometry control using a conformal oxide (purple) and SiC layer. iii. Anisotropic RIE. iv. Wet-etch removal of HSQ. (B) On-demand integration of Er ions in the NW-PC. i. Deposition of sacrificial oxide layer (orange) and CMP. ii. Combination of RIE and wet-etch to expose the top of NWs. iii. Deposition of thin oxide layer (yellow) based on targeted ion implantation depth. The corresponding top-down SEM images of step ii and iii are shown. iv. Er ion (red circles) implantation and wet-etch to remove oxide layers. (C) Schematic cross-section diagram of the resulting Er-doped 1D NW-PC structures with lattice periodicity P₁. P₂: sub-lattice periodicity; H: height of the NWs; W: width of the NWs. (D) Low magnification top-down and (E) high magnification cross-section SEM images of the resulting Er-doped NW-PC structures (W=20 nm, P₁=400 nm, P₂=100 nm, H=130 nm).

Post-deposition thermal treatment was carried out in forming gas ambient (95% argon [Ar] and 5% hydrogen [H₂]) at 1100°C for 1 h. Following the annealing of the ultrathin chemically grown layer, we performed anisotropic reactive ion etching (RIE) to remove the top SiC or SiC:O layer, leaving the sidewall self-aligned layer (NWs) intact and exposing the HSQ template (Figure 1A-iii). For control over the geometry of the NWs, the SiC sidewalls are encapsulated with a conformal oxide and SiC layer. This helps protect the self-aligned sidewalls during the ensuing RIE and enables control over the shape and height of the resulting NWs. A wet etch, buffered hydrofluoric (BHF) acid, was used to remove the HSQ, yielding ultrathin NW arrays synthesized in a self-aligned manner (Figure 1A-iv).

This synthesis route utilizes new precursor chemistry, TSCH (1,3,5-trisilacyclohexane) from Gelest Inc. (see Methods), resulting in highly crystalline matrix with good surface roughness, enabling the fabrication of NWs with width (W) of 10 nm (examples of more structures are shown in Figure S1). The critical dimension, width, of the NWs solely depends on the thickness of the conformal sidewall layer. The height (H), lattice periodicity (P₁), and sub-lattice periodicity (P₂) of the NW array are also shown in Figure 1C.

2.2 On-demand placement and integration of Er³⁺ into NW-PC

Panel b of Figure 1 presents the major process steps employed to engineer the position of Er³⁺ ions in the NW-PC. First, we deposited a sacrificial thick oxide layer onto the NW array using CVD [24], followed by chemical mechanical planarization (CMP) to flatten and thin the oxide layer. We then performed controlled RIE and wet-etch steps to recess the oxide layer and to expose the top of the NWs. A thin (15 nm) conformal CVD-grown oxide layer was then deposited to encapsulate the NWs. We can modulate the thickness of the oxide layer based on the targeted ion-implantation depth in the NWs. Accordingly, the remaining thicker oxide layer between the NWs prevents ion implantation into the substrate.

Samples were then subjected to a 150 keV beam of erbium ions targeting ions to be embedded into 30-nm depth from the top of the NW arrays, as determined by Transport of Ions in Matter simulations. Various doses of Er ions – 1×10¹³ cm⁻² (low), 5×10¹³ cm⁻² (medium), and 1×10¹⁴ cm⁻² (high) – were implanted into the NW-PC structures and (thin film) samples by ion implantation. We then carried out a wet-etch to remove the oxide layers yielding the controlled-Er-doped NW-PC structures (Figure 1D–E). This process was followed by a postimplantation Ar annealing (1 h, 900°C) and a forming gas (95% nitrogen [N₂] and 5% H₂) thermal treatment (1 h, 850°C) to activate the Er ions optically and for surface passivation [27]. The integration scheme dictates the final positions of the implanted ions to be within the 10 nm width of the NWs, which is an improvement to the accuracy of a state-of-the-art deterministic ion implantation process for single atom devices in [28] and the single ion implantation process in [29].

3.1 Modeling of a dipole emission in NW-PC structures

The extraction efficiency of the spontaneous emission from a dipole in 1D NW-PC and a uniform slab without a PC structure was explored using finite-difference time-domain (FDTD) calculations (Figure 2). We used a monitor parallel to the xy-plane to calculate the extraction efficiency of the emitted radiation power from the dipole (see Methods and Figure 2A). The extraction efficiency, n_ext, is defined as the fraction of the radiation power transmitted through the monitor to the total emitted power from the dipole [19], [30].

Figure 2:

FDTD modeling of the spontaneous emission extraction efficiency for the NW-PC and reference (Ref.) thin-film slabs.

(A) i. Schematic diagram of the geometry used in the FDTD computation with the relative position of the monitor used for extraction efficiency calculation. ii. Top-down schematic of the geometry showing the orientation of y-polarized dipole. (B) The emission extraction efficiencies, η_ext, from the NW-PC structures and reference as a function of wavelength. Gray data points refer to SiC:O structure with n=2.2. (C) Calculated spatial distribution of |E|² at 1540 nm in the zx plane (y-polarized dipole, n=2.7) for i. the reference and ii. the 600-nm-P₁ NW-PC. The dashed lines in each map represent the outline of the structure. iii. Dispersion relation of the 600-nm-P₁ NW-PC structure (periodicity, a=P₁=600 nm), showing an incomplete photonic bandgap extending from 190 THz (1577 nm) to 229.2 THz (1308 nm).

FDTD calculations revealed that the emission extraction efficiency from the SiC slab was ~3.6% at most wavelengths across the simulated range (Figure 2B). The observed behavior is consistent with findings from other studies indicating a high density of guided optical modes within the high-refractive-index SiC over a broad wavelength range [19], [30]. Similarly, we performed FDTD calculations in SiC NW-PC structures with three different P₁ values that revealed two major differences. First, the extraction efficiency was enhanced throughout the whole investigated wavelength range with values up to ~40% (Figure 2B). Specifically, in the 600-nm-P₁ NW-PC sample, the Er³⁺-induced 1540-nm emission extraction efficiency was 39%, an order of magnitude higher than that in the slab with ~3.6%. Second, a modulation of the extraction efficiency was observed with NW periodicity P₁ as shown in Figure 2B. For example, at 1000 nm, n_ext. from the 200-nm-P₁ and 600-nm-P₁ SiC NW-PC structures are 24% and 37%, respectively. Similar trends were also observed in SiC:O systems (Figure 2B, gray data points).

The extraction efficiency enhancement of the dipole’s spontaneous emission in NW-PC structures is expected because of the suppression of in-plane emission modes (guided modes). By eliminating guided modes at transition frequencies, spontaneous emission can be coupled efficiently to free space vertical modes, resulting in substantially enhanced extraction efficiency [19], [30]. Therefore, radiation originating from NW-PC structures results in enhanced light emission in the vertical direction. This behavior is supported by our spatial distribution calculation of the extracted power radiation (~|E|²) at 1540 nm (Figure 2C). In the case of the thin-film slab, a significant portion of the radiation is confined within the slab (x-direction in this simulation) (Figure 2C-i). Conversely in NW-PC, guided modes are eliminated as it is presented in Figure 2C-ii and Figure S2 (Supplementary Material). For the 600-nm-P₁ structures, 1540-nm emission is inhibited along the x-direction more prominently than other NW-PC structures, as the 1540-nm wavelength falls within the computed incomplete photonic bandgap of the structure (Figure 2C-iii). Only the 600-nm-P1 structures among our fabricated arrays exhibited a photonic bandgap containing 1540 nm. Nevertheless, due to the ultrathin geometry of nanostructures, the extraction efficiency is enhanced due to lack of total internal reflection even if there is no bandgap present containing the frequency of interest [19].

3.2 Morphological and structural properties of materials

Fourier transform infrared (FTIR) spectroscopy measurements showed complete crystallization of the as-deposited amorphous SiC materials at temperatures ≥1000°C, in accordance with our previously reported results (Figure 3A-i) [31]. In this respect, the line shape of the Si-C stretching mode (~800 cm⁻¹) of the as-deposited material transformed from Gaussian to Lorentzian after annealing at 1100°C, which was accompanied by a substantial narrowing of the full width at half maximum (FWHM) (~30 cm⁻¹), comparable to values for high-quality crystal SiC [32]. The surface roughness of as-deposited and annealed materials was analyzed by atomic force microscopy. Both materials have extremely flat surfaces with a mean surface roughness of 1–2 Å (measured from several 1 μm² areas). X-ray diffraction and high-resolution transmission electron microscope (HR-TEM) analyses agreed with the FTIR results, revealing the formation of cubic 3C-SiC. The NWs were polycrystalline, with an average nanocrystal grain size of ~5 nm as presented in Figure 3A-ii.

Figure 3:

Structural and PL properties.

(A) i. FTIR absorption spectra of the Si-C stretching mode for the as-deposited (AD) and 1100°C-annealed 20-nm thin films showing the transformation from amorphous to polycrystalline phase. ii. Representative HR-TEM image of annealed NW-PC showing the average grain size of ~5 nm with 3C-SiC crystalline phase. (B) Visible PL spectra from NW-PC structures under 476 nm laser excitation. PL of the same wavelength range from outside of the NWs is shown in orange. (C) Characteristic room-temperature Er³⁺ emissions at ~1540 nm and ~1555 nm in NW-PC structures under 488-nm (solid lines) and 476-nm (dotted line) excitations. (D) Photoluminescence excitation (PLE) spectra at 1540-nm emission from NW-PC (square) and SiO₂ (circle symbols) under a range of laser excitation wavelengths from 458 nm to 514 nm. Error bars are not depicted as the errors are smaller than the symbol size.

3.3 Er³⁺ PL properties in SiC NW-PC structures

We investigated the room-temperature (RT) visible and Er³⁺ emission properties of undoped and Er-doped NW-PC structures, respectively, by PL spectroscopy using a micro-PL (μPL) system (Figure S3 in Supplementary Material) at different laser visible excitation wavelengths (e.g. 458 nm, 476 nm, 488 nm, and 514.5 nm). Figure 3B shows the matrix-related PL emission spectra of undoped NW-PC structures, with a peak PL emission at ~550 nm in concurrence with our previous studies [24], [25]. Pertaining to Er³⁺ emission, the 476-nm excitation is non-resonant excitation, while the 488-nm excitation corresponds to resonant transitions from the ground state ⁴I_15/2 to the ⁴F_7/2 excited state in Er³⁺. The characteristic Er³⁺ emission, due to intra-4f transitions ⁴I_13/2→⁴I_15/2, [18] at ~1540 nm and ~1555 nm were observed in both SiC and SiC:O NW-PC as shown in Figure 3C (Er dose: 10¹⁴ cm⁻²). Er³⁺-1540-nm PL emission was observed using various visible pumping wavelengths in NW-PCs, demonstrating the structures’ broadband excitation characteristics (Figure 3D). Conversely, for the SiO₂:Er system, only resonant transitions [33] to the excited states of Er³⁺ occurred around the excitation wavelengths of 458 nm, 488 nm, and 514.5 nm in agreement with other reports [34], [35].

NW-PC structures with larger P₁ periodicity, resulting in a lower density of NWs, would be expected to have lower visible PL due to their volumetric difference and also lower Er³⁺-1540-nm PL due to the corresponding decreased number of total Er³⁺ ions in the NW-PC. This volumetric effect was observed in the visible PL emission at 550 nm, where PL decreased for increasing P₁ periodicity (Figure 4A). However, the opposite trend was observed for the Er³⁺ PL. The structures with the smallest periodicity, P₁=200 nm, exhibited the lowest Er³⁺ PL among the studied NW-PC structures (Figure 4B). After accounting for the decrease in the number of Er³⁺ ions in the air region of the different NW-PC structures (Section S4 and S5 for more details), the Er³⁺ PL increases in a monotonic fashion by a factor of about five as P₁ increases from 200 to 600 nm (Figure 4C). Furthermore, the modulation of the Er³⁺ PL with P₁ periodicity presented a direct correlation with the FDTD-computed extraction efficiency in NW-PC (Figure 4B). This correlation was observed for equivalent SiC and SiC:O NW-PC structures, and we observed the highest Er³⁺ PL intensity in the 600-nm-P₁ SiC:O structures.

Figure 4:

Volumetric effects on PL properties of NWs.

(A) PL intensity (counts per second [cps]) at 550 nm as a function of P₁ periodicity. (B) Measured PL intensity at 1540 nm (green circle) and FDTD-computed extraction efficiency (pink triangle) for different NW-PC structures. Error bars are not depicted as the errors are smaller than the symbol size. Inset: PL intensity mapping of NW-PC with P₂×P₁=150×600. The scale bar is 500 nm. (C) Er³⁺ PL intensity in SiC NW-PC and SiC:O NW-PC as a function of P₁; the difference of implanted Er ions was accounted for among the NW-PC. Three different NW-PC structures were measured for each P₁. The gray data points represent the Er³⁺PL in SiC:O NW-PC without the volumetric normalization.

To further our understanding of the NW-PC effect on Er³⁺ emission, we studied a representative 600-nm-P₁ SiC:O NW-PC device (Er dose: 10¹³ cm⁻²). It is worth highlighting that with the experimental μPL setup used in this study, the Er³⁺ PL emission was not detected from the thin film without a PC structure (Ref.) at pumping power densities below 40 kW/cm². This power value can be defined as the threshold power for detecting Er³⁺ PL from the thin film. In contrast, Er³⁺ PL from the 600-nm-P₁ NW-PC was measurable with a pumping power as low as 0.9 kW/cm². This behavior suggests that, collectively, the pumping and emission-extraction efficiencies of Er³⁺ in NW-PC are higher compared to their thin-film analog. At the threshold power, and accounting for the same number of Er ions (Supplementary Material), the Er³⁺ PL collected from the 600-nm-P₁ NW-PC was found to be approximately 60 times higher than thin film (Figure 5A). Moreover, the PL intensity ratio between the NW-PC and thin-film samples increases monotonically with decreasing Er dose. For example, the PL ratio was approximately 17 and 35 for the 10¹⁴ cm⁻² and 5×10¹³ cm⁻² doses, respectively (inset in Figure 5A). The time-resolved PL (TRPL) of the 1540-nm emission revealed a luminescence lifetime, τ, of ~2.7 ms in 600-nm-P₁ NW-PC structures.

Figure 5:

Comparison of 1540-nm emission characteristics between Er³⁺ in 600-nm-P₁ NW-PC and thin film.

Photostability of Er³⁺ emission in 600-nm-P1 NW-PC. (A) Room-temperature steady-state Er³⁺ PL spectra in 600-nm-P₁ NW-PC and thin film (Ref.) under 488-nm excitation (pumping power density: 40 kW/cm²; Er dose: 1×10¹³ cm⁻²; the difference of implanted Er ions was accounted for between the NW-PC and thin film). The Er³⁺ PL intensity in thin film was multiplied by 20 for presentation purposes. Inset: Ratio of Er³⁺ PL intensity between NW-PC and thin film as a function of Er dose measured at their respective threshold pumping powers. (B) Er³⁺ PL emission peak intensity as a function of pumping power density from i. 600-nm-P₁ NW-PC and ii. thin film (Er dose: 10¹³ cm⁻²). The dashed lines indicate fits to the data using Eq. (1). (C) i. Peak position and ii. FWHM of Er³⁺ PL spectra as a function of pumping power. iii. Er³⁺ PL time traces with sampling bins of Δt=10 ms and Δt=100 ms (excitation power density: 250 kW/cm², Er: 1×10¹⁴ cm⁻²). (D) Polar plots of the integrated Er³⁺ PL peak area as a function of the emission polarization from i. 600-nm-P₁ NW-PC and ii. thin film. The polarization angle of the excitation source was kept parallel along to the NWs’ length.

To obtain more insight into these compelling Er³⁺ PL behaviors in the NW-PC, we measured the Er³⁺ PL emission as a function of excitation power (Figures 5B and S6). After subtracting the linear background and the detector dark counts, we fitted the 1540-nm emission data to the following equation [15], [36],

where I(P) is the Er³⁺ PL intensity at a given excitation power P, I_sat is the saturation intensity, and P_sat is the excitation power required to yield half of I_sat. Figure 5B-i presents the results from a representative 600-nm-P₁ NW-PC nanostructure, revealing I_sat=4244±114 cps and P_sat=~8.3 kW/cm². I_sat and P_sat were also extracted for a representative thin-film sample, giving I_sat=192±10 cps and P_sat=~150 kW/cm² (Figure 5B-ii). These values indicate an approximate 20-fold increase in both pumping efficiency and Er³⁺ emission output flux (I_sat) in NW-PC. Furthermore, Eq. (1) can be expressed as a function of the photon flux of incoming photons, φ, by [18], [35],

where σ is the effective excitation cross-section and τ is the luminescence lifetime that was experimentally determined in TRPL. I_sat is given by I_sat=N^*/τ_r, where τ_r and N^* are, respectively, the radiative lifetime and the concentration of the optically active erbium ions. We then fitted the experimental power dependence data to Eq. (2) using the observed Er PL lifetime, which yielded an effective cross-section value of 2.2×10⁻¹⁸ cm⁻² for the NW-PC. This value is one order larger than that of the thin-film sample (2.8×10⁻¹⁹ cm⁻²), and two orders of magnitude higher than the typical range of ~10⁻²⁰–10⁻²¹ cm² for Er-doped materials [35], [37], [38].

The matrix-related PL spectrum of the NW-PC devices is characterized by a broadband luminescence feature in the spectral range of 500–750 nm, with the dominant PL occurring around 550 nm wavelength. To this end, some of the Er³⁺ transitions, such as ⁴I_15/2→²H_11/2 (514–520 nm), ⁴I_15/2→⁴S_3/2 (550 nm), and ⁴I_15/2→⁴F_9/2 (660 nm), are in resonance with the broadband PL of the NW-PC devices [33]. The PL and PLE data thus demonstrate an efficient energy transfer mechanism between the nanostructured matrix and Er³⁺ ions, as the Er emission can be sensitized by the nanostructured matrix. Effective light trapping by multiple scattering events can also increase absorption probability of the incident light in the NW-PC devices [7].

Photostability under high pumping is important for photonic applications. The peak position and lineshape (FWHM) of the observed Er³⁺ PL emission in NW-PC remained unchanged with pumping power (Figure 5C-i,ii), showing good photostability of the Er-doped nanophotonic structures. Furthermore, we examined the photostability of Er³⁺ PL under high pumping power as a function of time. The Er³⁺ PL time-traces on two different timescales are shown in Figure 5C-iii. The Er³⁺ PL count rate remained constant over several minutes with no indication of photobleaching. Furthermore, the large geometric anisotropy of the ultrathin NW also results into anisotropy of the emission and absorption properties of the emitters [39], [40]. As presented in Figure 5D-i, the Er³⁺ emission from NW-PC was found to be predominantly polarized along the length of the NWs, with an emission polarization degree, ρ_em, of ~0.4. Even if the dipoles in a NW are randomly oriented, the emitted light from the NW will be preferentially oriented along the NW axis and ρ_em is expressed as [41]:

where I_‖ or I_⊥ is the intensity of the most (polarizer parallel to NWs) or least (polarizer perpendicular to NWs) intense PL, respectively. In contrast, Er³⁺ PL emission from the thin film appears completely isotropic (independent from emission polarization direction) (Figure 5D-ii).

3.4 Optically active Er³⁺ concentration calculation and defect density

To estimate the concentration of the optically active Er³⁺ ions in the NW-PC and thin film, we used the following approach described in more detail in Section S7. In the linear pumping regime (low pumping power), where the excitation rate constant, σφ, is much smaller than the de-excitation rate constant, 1/τ (σφτ<<1), Eq. (2) can be approximated by

At saturation (high pumping power), where σφτ>>1, the PL intensity is equal to

As discussed, at saturation the I_sat ratio between the NW-PC and thin film was ~22. We calculated the radiative lifetime (τr,NW=Γr,NW−1), reciprocal of the radiative PL decay rate Γ_r, of the NW-PC structure relative to the lifetime in the reference thin film (τr,ref=Γr,ref−1) using FDTD simulation [30], [42]. In this respect, the total power radiated by the dipole within the NW-PC (P_NW) is proportional to the radiative spontaneous emission rate, Γ_r,NW, and equivalently for the reference thin film. Thus, in the case of the 600-nm-P₁ NW-PC and thin film, we calculated the emission rates ratio as

Γr,NWΓr,ref=(τr,NWτr,ref)−1= PNWPref≈0.39

Using Eq. (5) and the extraction efficiency ratio between the NW-PC and thin film, as determined by FDTD computations, we estimated [43] the optically active Er³⁺ concentration in NW-PC to be ~12.4 times higher than that in the thin film. Considering the decrease in the number of Er³⁺ ions in the air region for the NW-PC structures (~0.083× decrease), the total number of Er³⁺ was found to be approximately ~3% more in NW-PC. Nevertheless, we observed that the Er³⁺ PL was saturated at much lower power in the NW-PC structures, indicating a more efficient mechanism of pumping Er³⁺ compared to the thin film. As mentioned in the Introduction, an enhanced pumping efficiency is important for solid-state photonics and quantum communications applications [9], [11].

The integration of rare-earth ions in wide-bandgap ultrathin NWs can offer supplemental benefits. It can be argued that Auger and energy back-transfer effects are appreciably reduced in wide-bandgap SiC and SiC:O due to a low concentration of free carriers and the large mismatch between their bandgap and the first excited state of Er³⁺, respectively [24], [35]. Confinement effects on electron-phonon interaction may be expected primarily due to the absence of low-energy acoustic phonon modes in ultrathin NWs, resulting in different luminescence dynamics of Er³⁺ at low temperatures [44]. Moreover, in ultrathin NWs, a decreased bulk defect density [45] and a strain-relaxed environment in the material is expected [46], [47] and, hence, a smaller probability of nonradiative sites within the ion’s recombination volume. Thus, in ultrathin NW-PC having well-passivated surfaces, excited Er³⁺ is expected to be exposed to a smaller number of bulk nonradiative recombination sites compared to its bulk analog, which can lead to an increased PL lifetime [45]. From deconvolution of the PL spectra at RT, we observed that spectral line width is narrower in NW-PC structures (~10 nm) compared to the thin-film sample (~18 nm), which can be due to reduction in defect density and/or reduction in the number of optically active ions.

To access the influence of nonradiative quenching processes in the NW-PC structures, we performed TRPL of the 1540-nm emission at 77 K. As can be seen in Figure 6A, the observed Er³⁺ luminescence lifetime did not change (2.7±0.1 ms) with respect to its 300 K value. The Er³⁺ PL dynamics of the NW-PC implies a good-quality material, with significantly low defect density and a well-passivated surface. It may be thus inferred that the higher concentration of optically active Er³⁺ in NW-PC can be the result of reduction of bulk defect-density, as compared to the Er³⁺ PL dynamics of the thin film (Figure S8), and a more effective thermal annealing process in forming optically active Er ions due to the NW geometry. Additionally, with decreasing Er dose, the number of Er³⁺ ions in NW-PC also decreases, which results in an even lower probability of a defect center being within the ions recombination volume. This may explain the trend of increased PL intensity ratio with decreasing Er dose between the NW-PC and thin film mentioned in Section 3.3 (inset of Figure 5A).

Figure 6:

Effect of NW-PC structure on Er³⁺ PL emission dynamics.

(A) Er³⁺ PL (~1540 nm) decay in the 600-nm-P₁ NW-PC structure under 488-nm pulsed excitation at 77 K and 300 K (red points: 300 K [RT]; green points: 77 K). The Er³⁺ PL lifetime in NW-PC, as obtained from single exponential fitting, was calculated to be ~2.7 ms for both 77 K and 300 K. The solid lines indicate fits to the data. (B) PL decay of 550-nm emission from the NW-PC structures (legend: P₁) with instrument response function (IRF) in gray. (C) Room-temperature Er³⁺ PL decay at 1540 nm from different NW-PC structures (legend: P₁). Solid lines represent single exponential fits. (D) Er³⁺ PL lifetimes for the different NW-PC structures and for a single NW (orange). (E) Computed relative radiative lifetime from NW-PC with different P1 and a single NW. Data points corresponding to the NW-PC with periodicity values beyond our fabricated structures are shown under the shaded region.

The much shorter Er³⁺ PL lifetime in thin films (Figure S8) can be attributed to an increase in nonradiative de-excitation paths for Er ions in the thin-film matrix in agreement with other reports [48], [49]. Furthermore, a reduction in the effective refractive index in NW-PC and confinement of the Er ions within the NW may reduce the photon density of states for emission and thereby reduce the transition rate and correspondingly increase the Er luminescence lifetime [35]. The effective refractive index in NW-PC is expected to be between that of air (n=1) and thin film (n=2.2).

As discussed, we observed that the Er³⁺ PL from NW-PC was modulated with P₁ periodicity and was correlated with the computed extraction efficiency trend at 1540 nm (Figure 4B). In conjunction to the different behavior between the steady-state PL at 550-nm and 1540-nm emissions from NW-PC structures, we also observed different trends in the luminescence dynamics at these wavelengths. Average lifetime at 550 nm was found to be the same, ~680±40 ps, for all the NW-PC structures (Figure 6B). However, luminescence lifetime of Er-PL at 1540 nm was found to be approximately the same, 2.3±0.1 ms, only for NW-PCs with 300-, 400-, and 500-nm-P₁ periodicity (Figure 6C,D). For the 600-nm-P₁ NW-PC structures, in agreement with the computed radiative lifetimes (Figure 6E), we observed a longer lifetime (2.7±0.1 ms), which suggests that an additional physical mechanism is contributing to the modulation besides nanostructuring and reduced defect density effects. The spontaneous emission rate of PC structures is predicted to decrease when the emission spectrum lies within the photonic bandgap [20], [30], as in the case of 600-nm-P₁ NW-PC structures, yielding longer luminescence lifetime. We observed a ~15% and ~30% relative difference in lifetimes from experimental (Figure 6D) and computational (Figure 6E) results, respectively, between 600-nm-P₁ and the three other NW-PC that do not have a photonic bandgap containing 1540 nm. As expected for our 1D 600-nm-P₁ NW-PC structures, the increase in PL lifetime was less pronounced compared to the increase observed in two-dimensional photonic bandgap materials (~80% in Ref. [30]).

The increase of the Er³⁺ PL lifetime in NW-PC, decrease of its PL emission rate as ΓNW=τNW−1=Γr,NW+Γnr,NW, is coupled to the above mechanisms (Γ_nr is the nonradiative decay rate). Γ_NW is influenced by the complex interplay of the radiating Er³⁺(Γr,NW=τr,NW−1) and its interactions with the nanophotonic structure (e.g. NW geometry, nanostructuring effect, photonic bandgap), as well as the surrounding defectivity (e.g. ion’s energy transfer to competing nonradiative decay channels, Γr,NW=τr,NW−1).

5.1 Synthesis of NW array structures

Silicon (100) wafers were spin-coated with HSQ negative-tone resist at 1000 rpm followed by a soft-bake, yielding an approximately 130-nm-thick HSQ layer. We exposed the HSQ layer using a Vistec VB300 (Vistec Electron Beam GmbH, Jena, Germany) and a Voyager Raith (Raith GmbH, Dortmund, Germany) electron-beam lithography tool using line patterns created in the Layout Editor. Following the exposure, development was performed in tetramethylammonium hydroxide, yielding HSQ ribbon arrays. We conformally deposited 20 nm of SiC/SiC:O onto the HSQ ribbon arrays using a home-built thermal CVD system at 800°C. CVD-742 (1,1,3,3-tetramethyl-1,3-disilacyclobutane) from Starfire Systems or TSCH (1,3,5-trisilacyclohexane) from Gelest Inc. (Morrisville, PA, USA) was used as the silicon and carbon source. After the SiC or SiC:O growth, anisotropic fluorine-based (combination of CHF₃ and CF₄ gases) reactive ion etch was performed using a Plasma-Therm Versalock 700 (Plasma-Therm, St. Petersburg, FL, USA) to remove the top SiC or SiC:O layer, leaving the sidewall layers intact and exposing the HSQ ribbon template. A wet-etch using BHF acid was then carried out to remove the HSQ, yielding ultrathin NWs synthesized in a self-aligned manner.

5.2 Device modeling

The calculations of the extraction efficiency and the spatial distribution of the dipole radiation were carried out using 3D FDTD simulations (Lumerical Inc.). A y-polarized single dipole source was positioned at the center of a uniform slab (thin film) and of a NW-PC structure (Figure 2A). The spatial dimension of the simulation window was varied to keep the total number of NWs (21 pairs) the same for all structures. We used a fixed grid size of 1 nm and a perfectly matched layer as the boundary condition in all directions [30]. We used a monitor (parallel to the xy-plane) to calculate the transmitted flux through the top surface of the slab. It consisted of one 2 μm×2 μm xy-plane at z=500 nm from the surface of the material (Figure 2A).

The position and dimensions of the monitor were kept the same for all simulations to ensure equivalent solid angle detection. We then calculated the total power injected by the dipole (P_in) by integrating the time averaged Poynting flux over the six surfaces of a cube enclosing the radiating dipole. The extracted power (P_out) was calculated from the power transmitted through the monitor plane. The extraction efficiency, η_ext, is defined as the ratio of P_out/P_in [19], [30]. The index of refraction, n, thickness, and geometry values for the thin film and NW-PC structures, used in the simulations, were experimentally determined by scanning electron microscopy (SEM) and spectroscopic ultraviolet-visible ellipsometry measurements [J.A. Woollam (Orlando, FL, USA) RC2 dual rotating compensator ellipsometer].

5.3 PL characterization

A home-built μPL system – composed of an argon laser (Beamlock 2065-7S, Spectra-Physics, Santa Clara, CA, USA) coupled to an FLSP920 spectrometer from Edinburgh Instruments (Livingston, UK), a Triax-550 spectrometer from Horiba Jobin Yvon (Kyoto, Japan), and liquid nitrogen cooled Ge/InGaAs detectors – was utilized for PL and power-dependence PL (PDPL) measurements. We performed the PL and PDPL measurements at room temperature using an argon laser (Beamlok 2065-7S, Spectra-Physics, Santa Clara, CA, USA), to optically excite the Er³⁺ ions, through a Zeiss 50× [numerical aperture (NA)=0.85] objective lens. TRPL studies were conducted in the FLSP920 spectrometer utilizing a multichannel scaling technique. An acousto-optic modulator (Gooch & Housego, Fremont, CA, USA) was used to create laser pulses with a ~500 μs FWHM and 50 Hz repetition rate for the NW-PC structures and 100 Hz for the thin film. For PL measurements at 77 K, a close-system cold-finger cryostat with Mitutoyo NIR 50× objective lens (NA=0.55) was used in the same spectrometer. For the emission-polarization anisotropy measurements, PL spectra were collected at different angles of a polarizer placed in front of the tube lens.