Enhanced Light Extraction of Nano‐Light‐Emitting Diodes with Metal‐Clad Structure Using Vertical GaAs/GaAsP Core–Multishell Nanowires on Si Platform

Nanometer‐scaled light sources using III–V compound semiconductor nanowires (NWs) on Si are expected as building blocks for next‐generation Si photonics, bioimaging, on‐chip microscopy, and light detection and ranging (LiDAR) techniques. This is, however, limited in a few materials systems due to complexity in integration of the vertical III–V NWs on Si and device process flow. Suppressing optical loss in the NW materials beyond the optical diffraction remains difficult in enhancing light extraction. Herein, the effect of the vertical metal‐clad architectures for the vertical nano‐light‐emitting diodes (LEDs) using GaAs/GaAsP‐related core–multishell NWs heterogeneously integrated on Si is investigated. The grown core–multishell NW is composed of a radial n‐GaAs/n‐GaAsP/p‐GaAs/p‐GaAsP double heterostructure. The vertical metal‐clad NW‐LEDs show suppression of carrier overflow effect and rapid enhancement of electrical luminescence.


Introduction
Vertical III-V compound semiconductor nanowires (NWs) on Si have attracted attention as building blocks for nanoscale light sources in such light-emitting diodes (LEDs) and laser diodes (LDs) in Si nanophotonics, [1] imaging technologies, [2] on-chip spectroscopy [3] and LiDAR systems. [4] This is because the NW structure geometrically shrinks the occupied area by one several thousand as compared to those of planar light sources like the smallest vertical surface-emitting laser (VCSEL). Core-multishell (CMS) structures are able to mount radial double heterostructures (DH) with quantum well (QW) and p-n junctions on the NW sidewalls and enlarge surface-to-volume ratio. This advantage simultaneously is satisfied with low-power and high-brightness properties that are required for the wearable display devices. The NW-LED are demonstrated mainly in nitride systems for visible color wavelength. [2,[5][6][7] In Si nanophotonic applications, nanoscale coherent light emitters and detectors are required for replacing metal-based interconnections with optical interconnections inside the circuit area. Furthermore, very small lasers would achieve thresholdless current with very small power consumption. With this regard, the NW-based light sources with telecommunication (1.3 and 1.55 μm) and the near-infrared (NIR) wavelength region have been explored in As-/P-related NWs on Si. The optically pumped stimulated emission has been demonstrated using single NWs [8][9][10] and NW arrays on Si. [11,12] Among the NWs, the light sources on Si operated at NIR would be more feasible for nanoscale fast optical communication between integrated circuit (IC) chips with a combination of Si-based avalanche photodiodes (APDs). This has been, however, limited in a few NW material in the NIR region and in a simple vertical diode structure toward NW-LD applications due to complexity in integration of the vertical III-V NWs on Si and 3D device process flow.
The materials for nanolight source in nanophotonics must emit coherent light by overcoming the diffraction limit. Recently, metal-clad cavity structures with surface plasmon resonance have demonstrated stimulated light emission from nanomaterials with current injection, overcoming the metal large ohmic loss at optical frequencies and increasing Purcell factor. [13][14][15] The light at NIR wavelengthregion was more feasible for small light sources because the reals part of the metal dielectric function became negative in the NIR region. Also, the nanolight source in the NIR wavelength has advantage for applications in on-chip spectroscopy, bioimaging technologies, and photoimmunotherapy. There have been, however, no reports regarding the effect of the metal-cladding layer for the III-V NWs on Si platforms in the NIR region. The NW light source is required for important techniques to realize electrically driven NW-LDs on the Si platform. First, the metal-cladding layer for confining light propagation along the radial direction against the NW's axial direction has to be made. Second, the cavity structure against light propagation along the NW axial direction is to be made. In this report, we clarify the effect of the metal-cladding layer for the NW light source. Thus, we demonstrate the heterogeneous integration of vertical GaAs/GaAsP CMS NWs with GaAs QW with a radial p-n junction on Si(111) substrates and vertical LED structures with vertical metal-clad architecture using process-available tungsten (W).

Growth and Optical Properties of GaAs/GaAsP-Related Core-Multishell NWs on Si
Vertical GaAs NWs were directly grown on n-Si(111) substrates by selective-area growth (see Experimental Section). The alignment of vertical <111> direction for the GaAs NW growth on nonpolar Si(111) was the same as in previous reports. [16,17] Then, the radial DH structure composed of n-GaAsP barrier/p-GaAs QW/p-GaAsP barrier/p-GaAs cap multishell layers is designed, as illustrated in Figure 1a. Figure 1b shows the representative 30°-tilted scanning electron microscopic (SEM) image showing the vertically aligned n-GaAs/n-GaAsP shell/p-GaAs QW/p-GaAsP/p-GaAs shell CMS NWs on n-Si(111) substrate. The NWs were %210 nm in average diameter and 1.6 μm in average height. The core n-GaAs NW was 70 nm in diameter, and the thickness for n-and p-GaAsP shell was 30 nm in each. Photoluminescence (PL) spectra at room temperature (RT) of the CMS NWs on Si regardless of impurity doping are shown in Figure 1c. The PL spectra were composed of several origins denoted as P1-P5 and band-edge emission (E G Γ , 1.423 eV), which were analyzed by peak deconvolution with Gaussian distribution functions (shown in Figure 1c). In this spectra, the PL from the bandgap of the core GaAs NW (1.423 eV) showed very weak PL intensity. This was because that almost all of excitation light was absorbed inside the GaAsP/GaAs shell layer because of the optical penetration depth. The photogenerated electron-hole pairs inside the CMS NWs were radiatively recombined in the GaAs QWs and GaAsP shells.
Considering the volume of each layers in the CMS, weak PL of core GaAs NWs (%1.423 eV) and optical transition in the DH structure indicated that the P1-P5 peaks originated from GaAsP shell layers and GaAs QW tubes inside the CMS NWs. The PL denoted as P1 (1.518 eV) presumably originated from the GaAs QW tubes. P2 (1.555 eV), P3 (1.543 eV), and P4 (1.629 eV) originated from the GaAsP shell layers according to the P composition in vapor phase. The P composition in the solid phase (shell layers) was estimated to 17.5% for P2, 16% for P3, and %27% for P4 as the strained GaAsP layer. The peak difference in P2 and P3 originated from alloy fluctuation of the GaAsP shell layers. The weak PL denoted as P4 was ascribed as localized P-rich GaAsP nanostructures due to GaAsP alloy phase segregation [18][19][20][21][22][23] or the thin GaAsP layer grown on the top (111)B facet. P5 at 1.482 eV originated from the mixing structure of zincblende (ZB) and wurtzite (WZ) phase in core GaAs NWs. [24] Assuming that P1 originated from the 1e-1hh quantized state in strained GaAs 0.83 P 0.18 /GaAs/GaAs 0.83 P 0.18 QWs, the thickness of the GaAs QW was calculated to be 4.0 nm that approximately corresponded to the growth rate of GaAs shell layer. Figure 2 exhibits low-temperature PL spectra for the GaAs/GaAsP CMS NWs and GaAs NWs in order to further characterize the luminescence origins. In addition to the recombination of the exciton-related level (X) at 1.511 eV and (D,X) at 1.505 eV, the luminescence from carbon impurity at As site-related levels [C(As)] at 1.494 eV and from donor-acceptor pair (DAP)-related level at around 1.48 eV measured from the bare GaAs NWs (black curve in Figure 2a) were involved in the radiative recombination processes in the GaAs/GaAsP CMS NWs. GaAs/GaAsP CMS NWs exhibited a strong PL peak at around www.advancedsciencenews.com www.adpr-journal.com 1.53 eV and the weak PL at around 1.640 eV and at around 1.850 eV. Assuming that the PL at around 1.652 eV originated from the GaAsP shell layers, the P-composition was estimated to be 17.5%, which corresponded to the P2 emission observed in Figure 1c. The origin of the PL at %1.850 eV was ascribed to be the same as P4 in Figure 1c. The P-composition in the solid phase estimated from the PL peak was around 43%, which was much higher than the estimated value (%27%) of P4 in Figure 1c. This difference is reasonably explained by the quantum size effect in the alloy phase segregated GaAs 0.73 P 0.27 quantum wires (QWRs), rather than the origin from the very thin GaAs 0.73 P 0.27 layer grown on GaAs (111)B top facets. The quantum size effect blueshifted the PL peaks of the GaAs 0.73 P 0. 27 QWRs as compared to that of planar GaAs 0.73 P 0.27 . Thus, the GaAs/GaAsP CMS NWs are thought to have GaAsP QWRs at the corner of the NWs by the alloy phase segregation of As/P atoms. [20] Peak deconvolution with Gaussian function revealed that the dominant PL at around 1.53 eV was composed of the edge optical transition (E g Γ ) of GaAs (1.519 eV) and strong PL at 1.528 eV, which was thought to be from GaAs QW tubes. The PL intensity of the edge transition in the GaAs/GaAsP CMS NWs was enhanced instead of the absence in that of the bare GaAs NWs in Figure 2a. The PL intensity at X-related emission in the GaAs/GaAsP CMS NWs was 7 times higher than that of the bare GaAs NWs. As for the PL from the GaAs QW tubes, the thickness of the GaAs QW tube was estimated to 3.4 nm from the peak position, which almost corresponded to the PL peak measured at RT. The QW thickness differed by 1.5 monolayer compared to that of PL at room temperature ( Figure 1c). This suggested that the dominant PL origin was the QW with thickness fluctuation. The PL that originated from the GaAs QW was dominant emission both at RT and at 4.2 K, which meant that the carrier confinement effectively occurred inside the GaAs QW tubes. Furthermore, the PL peaks at around 1.528 eV had small additional two peaks at around 1.528 eV. The peak difference was AE3 meV, as shown in Figure 3c, which originated from the thickness fluctuation [25] of the GaAs QW layer. The thickness difference was estimated to þ1 Å and À3 Å against 3.4 nm that originated from the monolayer-scaled thickness fluctuation of the GaAs QW tubes.

Characterization of Vertical Metal-Clad NW LEDs on Si
Figure 3a-c depicts the NW-LED structure using the vertical n-GaAs/n-GaAsP/p-GaAs QW/p-GaAsP/p-GaAs CMS NWs monolithically integrated on Si. The device procedures for the conventional NW-LEDs in Figure 3a were the same as the previous report. Figure 3b illustrates the device structure for the vertical metal-clad NW-LEDs. The difference structure was contact length (L C ) of the top electrode. The metal-clad structure shown in Figure 3b was achieved by increasing the L C to 1500 nm. Figure 3c shows the fabricated vertical metal-clad NW-LED structure where the NW sidewalls were wrapped with the Ti/Pt/Au/W multilayer. The thickness of Ti/Pt/Au/W was 10 nm/10 nm/50 nm/150 nm. The current density-voltage ( J-V ) properties for both vertical NW-LED structures indicated rectifying properties with a turn-on voltage of 0.4 and 1.3 V. The current-voltage properties were measured from the NW array and the measured current was normalized by the number of NWs and surface area of the revealed NW surfaces after the reactive ion etching (RIE) process. The turn-on voltages were estimated from extrapolation of the J-V curve. 0.4 V originated from the band discontinuity of the n-Si/n-GaAs NW junction (shown in Figure S1, Supporting Information) and 1.3 V originated from the n-GaAs/n-GaAsP/p-GaAs QW/p-GaAsP/ p-GaAs CMS junction. The ideality factors (n) for the conventional vertical NW-LEDs and vertical metal-clad NW-LEDs were 2.81 and 2.66, respectively. Since both LED structures had %3.8 Â 10 3 NWs which were parallelly contacted with top electrodes, the estimated n values indicated a larger value than 2 due to the device variations. From the estimated n values, the surface recombination process was dominant process for both NW-LED structures. This indicated that these NW-LEDs were based on the surface-emitting process across the junction in the CMS layers. The estimated series resistances for the conventional vertical NW-LEDs and vertical metal-clad NW-LEDs were 75 and 33 kΩ cm À2 , where the NW-LED with longer L C showed small series resistance. The surface area of the metal contact for the NW-LED with L C of 1500 nm was 5 times larger than that of the NW-LED with L C of 300 nm, which had a much smaller ratio  against the difference in the series resistance. This means that longer L C enlarged the surface area for the current path and effectively decreased the series resistance. Thus, the vertical metal-clad structure with longer L C illustrated in Figure 3b is beneficial for the reduction in the series resistance and their Joule heating effect. Figure 4 shows the electroluminescence (EL) properties for the vertical-metal clad NW-LEDs. Figure 4a shows the EL spectra measured from the NW-LED with different L C . The primary difference in the EL spectra for the NW-LED with short L C (that is the conventional NW-LEDs) and long L C (that is the vertical metal-clad NW-LEDs) was the EL intensity as the similar current injection. The EL intensity ratio was much higher than that observed in the J-V properties in Figure 3. This indicated that the difference in the EL intensity resulted from some reason far behind the difference in the series resistance. Next, the EL peak position was different between the conventional NW-LED and vertical metal-clad NW-LEDs. The metal-clad vertical NW-LED (L C = 1500 nm) showed four EL peaks at 1.568, 1.518, 1.431, and 1.366 eV. The EL at 1.568 eV was emitted from the GaAsP shell layer. The main peak at 1.518 eV originated from the GaAs QWs which corresponded to P1 shown in Figure 1c. On the other hand, the conventional NW-LED exhibited four peaks at around 1.579, 1.494, 1.431, and 1.366 eV. Similar to the EL in the vertical metal-clad NW-LEDs, the GaAsP shell and GaAs QW-related emission were observed at 1.579 and 1.494 eV. This indicated that the current injection for the conventional LED structure dominantly occurred in the unintentional GaAsP/GaAs layer grown on the top GaAs NWs due to the high series resistance. EL observed at 1.431 and 1.366 eV for both NW-LEDs was originated from the GaAs NW-related emission. Thus the Joule heating effect induced the redshift of the EL shown in Figure 4b. Also, the EL peaks were redshifted with increasing current injection. Figure 4c shows EL with various current injection for the vertical metal-clad NW-LEDs. The peak position of the dominant EL that originated from the GaAs QW layer had no shift with increasing current injection. This indicated that the NW-LED had no Joule heating in the junction temperature. Figure 4d shows the integrated EL intensity for the dominant EL peak as a function of injection current density. The integrated EL intensity for the conventional NW-LED structure showed sublinear increase with current density. This was because the carrier overflow occurred similar to the planar QW-type LED. [26] While in case of the vertical metal-clad NW-LEDs, the EL intensity increased with current density. Another NW-LED structure on Si exhibited similar properties in low current injection. [17] The metal-clad structures enhanced the superluminescence property under the large injection current density. The benefit of the vertical metal-clad structure was able to inject carriers into the whole QW layer which was formed around the NW sidewalls regardless of current spreading and to suppress the carrier overflow properties regardless of the Joule heating effect.
According to the J-V properties, the metal-clad NW-LED structure would improve light extraction. In order to verify this point, a comparison of the PL signals between the conventional NW-LED and metal-clad NW-LED would be effective to verify this point. However, light excitation was difficult in case of the metal-clad NW-LED structure. Thus, we calculated the light propagation inside the NWs. Figure 5 shows the electrical field of the NW-LED structures calculated by 3D finite-difference timedomain (FDTD) method. [27] In our fabricated GaAs/GaAsP CMS NWs, the active layer was formed around the GaAs NWs. As compared to air, light propagation from the GaAs QW active layer along the NW axial (vertical) direction reflected with a larger refractive angle than that of air across the GaAs/BCB due to their larger refractive index contrast. Also, tungsten (W) can reflect over half of the EL inside the W-wrapped NWs. [28] This means that the arbitrary point in the NW which is far from the EL source effectively reflects the light along the vertical axis direction. Below the refractive angle, light propagation perpendicular to the NW vertical facet www.advancedsciencenews.com www.adpr-journal.com or with angle below the refractive index was almost same both in case of GaAs/air and in case of GaAs/BCB. In addition, we believe that the contribution of the thickness of the multishell NW layer had no impact on the confinement of light because the thickness was thinner than optical wavelength of the electroluminescence.
In this calculation, we assumed a simplified single NW-LED structure composed of bare GaAs/GaAsP CMS NWs on Si and 100 nm-thick W-wrapped GaAs/GaAsP CMS NWs on Si. Both NWs were surrounded by the BCB whose refractive index was 1.57. The calculation neglected the effect on the very thin SiO 2 layer and other semitransparent Ti/Pt/Au multilayers. As for light source, point light sources with the wavelength of 820 nm were aligned at the center of the NW with a pitch of 200 nm inside the NW. Note that the EL was detected normal to the LED surface in the EL measurements. In the FDTD mappings of the electrical field for the bare GaAs NWs, which approximately corresponded to the situation of the conventional NW-LEDs, light was propagated at the NW whole structure and was extracted at the upper part of the NW and bottom part of the NW adjacent to the Si substrate. Note that the current injection of the conventional NW-LED would be localized on the upper part of the NW because of the short L C . While in case of the metal-clad NW-LED in Figure 5d-f, electrical fields were localized at the edge of the NW top surface. The electrical fields were widely extracted from the NW as compared to those of the bare GaAs NW in Figure 5a-c. The intensity of the electrical fields around the top part of the metalclad NW was also stronger than those of the bare GaAs NW, which indicated that light extraction along the vertical direction of the metal-clad NW-LED was higher than that of the bare NW LED. Note that each single NW in the NW array shown in Figure 1b had this specific EL behavior that originated from the vertical metal-clad architectures. This means that the metal-clad NW-LED structure can generate EL emission with high directivity, suggesting that the metal-clad structure would be essential for the realization of electrically pumped laser diodes using very thin NWs.

Conclusion
In conclusion, the GaAs/GaAsP-based CMS vertical NWs were heterogeneously integrated on Si and the PL at room www.advancedsciencenews.com www.adpr-journal.com temperature revealed that the luminescence at 1.52 eV was the dominant radiative recombination process, which originated from 4 nm-thick GaAs QW tube inside the CMS NWs. Then, we found that the metal-clad architecture for the vertical NW-LED enabled to suppress carrier overflow effect of the GaAs QW-based LED and enhance the light extraction with high directivity in the NW axis direction from the simulation. The suggested metal-clad architecture would play an important role for the application of the electrically pumped laser diode using NWs.

Experimental Section
Selective-Area Growth of Vertical GaAs/GaAsP NWs on Si: In the experiment, the substrates were n-type Si(111) with 20 nm-thick thermal oxide SiO 2 film. The carrier concentration of the substrates was 1 Â 10 18 cm À3 . The mask pattern was designed as a periodical opening whose diameter was 70 nm with a pitch of 800 nm. The NW growth was performed using low-pressure (76 Torr) MOVPE with pure hydrogen (H 2 ) gas that was purified by thermal Pd films. For NW growth, trimethylgallium (TMGa) was used as group-III precursor. Tertiarybutylphosphine (TBP) and arsine (AsH 3 ) were used as group-V precursors. The monosilane (SiH 4 ) gas was used for n-type doping, and diethylzinc (DEZn) was used for p-type doping. The formation of (111)B-polar surface and growth of low-temperature buffer layer were performed to directly grow the vertical Si-doped n-GaAs NWs on the Si(111) substrates xx . The vertical GaAs NWs were grown at 750°C for 60 min with a V/III partial pressure ratio of 250. The carrier concentration for the Si dopants was 2 Â 10 18 cm À3 . As for the CMS NW structure, DH structure n-GaAsP barrier/p-GaAs QW/p-GaAsP barrier/p-GaAs cap multishell layers were designed, as shown in Figure 1a. The Si-doped n-GaAsP, Zn-doped p-GaAsP barrier, and the p-GaAs QW were grown at 650°C, and the cap layer was grown at 700°C for 1 min. The growth time for the n/p-GaAsP shell layers was for 10 min each, and for GaAs QW layer it was for 3 min with V/III partial pressure ratio of 395. P-composition in vapor phase was set to 37%. The carrier concentration for the n-GaAsP was 2 Â 10 18 cm À3 and for p-GaAsP and p-GaAs was 2 Â 10 19 cm À3 .
Micro-PL (μ-PL) and EL Measurements: μ-PL was carried out at room temperature and 4.2 K. The excitation light was He-Ne laser (633 nm), which was focused on %2 μm spot on the substrate. Since the NW pitch was 4 μm, approximately single NWs were included in the excitation laser spot. The PL and EL were detected normal to the sample surface and taken with InGaAs CCD detector.
Device Processes for NW LED Structure: The NW LEDs were fabricated using the GaAs/GaAsP DH CMS NWs on Si. First, Al 2 O 3 (5 nm) was deposited on the NW surface to avoid RIE-induced damage. Al 2 O 3 was formed using atomic layer deposition (ALD). Next, the NW samples were buried with polyresin and etched with RIE to expose the top parts of the NWs for metal contacts. We revealed the top portion of the NWs. The revealed NW height corresponded to L C . After RIE, Al 2 O 3 was removed by BHF solution. Then, semitransparent Ti(2 nm)/Pt(2 nm)/Au(2 nm) multilayer was deposited onto the top parts of the RIE-etched NWs. In the metal deposition, we used the rotating sample holder to effectively www.advancedsciencenews.com www.adpr-journal.com coat the metals on the sidewalls of the CMS NWs. The rotation per minute was 50. Also, the wafer was tilted to about 3°-5°. In order to obtain metalclad architecture, tungsten (W) film was sputtered on the sample. Then the sample was buried with polymer resin and RIE etching was performed to etch the top portion of W for light extraction. Then, Ti(20 nm)/Au(100 nm) metal was deposited on backside n-Si wafer. Finally, the LED structures were annealed at 400°C to obtain ohmic contact.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.