Colour-crafted phosphor-free white light emitters via in-situ nanostructure engineering

Colour-temperature (Tc) is a crucial specification of white light-emitting diodes (WLEDs) used in a variety of smart-lighting applications. Commonly, Tc is controlled by distributing various phosphors on top of the blue or ultra violet LED chip in conventional phosphor-conversion WLEDs (PC-WLEDs). Unfortunately, the high cost of phosphors, additional packaging processes required, and phosphor degradation by internal thermal damage must be resolved to obtain higher-quality PC-WLEDs. Here, we suggest a practical in-situ nanostructure engineering strategy for fabricating Tc-controlled phosphor-free white light-emitting diodes (PF-WLEDs) using metal-organic chemical vapour deposition. The dimension controls of in-situ nanofacets on gallium nitride nanostructures, and the growth temperature of quantum wells on these materials, were key factors for Tc control. Warm, true, and cold white emissions were successfully demonstrated in this study without any external processing.


Results and Discussion
In-situ SiN x nanomask layer deposition. To investigate the effects of the in-situ SiN x nanomask deposition, we grew n-type doped GaN (n-GaN) templates on a c-plane sapphire substrate using metal-organic chemical vapour deposition (MOCVD). The in-situ SiN x nanomask layers were deposited on the n-GaN templates for variable time lengths. These experimental conditions are further detailed in the "Methods" sections. Finally, the n-GaN layer was regrown at 850 °С for 3 s without coalescence of the n-GaN nanostructure with each other. As shown in the scanning electron microscopy (SEM) images of Fig. 1, the four templates show different surface morphologies depending on the SiN x nanomask deposition time. These templates were labelled based on deposition time as reference (0 s), sample-A (1500 s), sample-B (2000 s), and sample-C (2500 s). As shown in Fig. 1, while the reference (no SiN x nanomask) features a flat surface, samples-A, B, and C feature high-density n-GaN nanodots (G.D.). These results confirm that the regrown n-GaN layer was selectively grown on the open sites of the in-situ SiN x nanomask. Moreover, samples-A, B, and C have different n-GaN nanodot densities of 4.7 × 10 10 /cm 2 , 3.7 × 10 10 /cm 2 , and 2.9 × 10 10 /cm 2 , respectively. These results indicate that the density of openings on the SiNx nanomask changes based on deposition time. The size of the n-GaN nanodots on each sample is inversely proportional to the opening density of the corresponding nanomask. Sample-A features nanodots approximately 10-30 nm in diameter, with the narrowest nanodot spacing among the three samples. Sample-B and sample-C show nanodot diameters of approximately 15-40 nm and 20-55 nm, respectively. Sample-C shows that the average size of n-GaN nanodots is the largest among these samples, indicating that more Ga adatoms were incorporated into the open areas of the SiN x nanomask layer than in sample-A and sample-B. As a result, we could control the size and the distribution density of the n-GaN nanodots by changing the deposition time of the SiN x nanomask. 3D n-type GaN nanostructure growth. To fabricate the n-GaN nanostructures, we re-grew the 3D n-GaN structure on in-situ SiN x nanomask-embedded templates for 600 s at 850 °С. Figure 2 shows a schematic of the growth process for the 3D n-GaN nanostructures. As shown in Fig. 2, the shape of the n-GaN nanostructure surface depends strongly on the coverage of the SiNx nanomask. For SiN x nanomasks with low coverage (sample-A), the vertical growth rate is relatively slower than that of high coverage nanomasks (sample-C) because many of the neighbouring n-GaN nanodots are laterally merged with each other. This accelerates the formation of the low trapezoidal shape of the n-GaN nanostructure with flat top-plane [P: polar plane (0002) on GaN] and inclined-plane [SP: semipolar planes {11-22} or {10-11} of GaN] surfaces. On the other hand, the shape of the n-GaN nanostructure grown on high coverage SiN x nanomask (sample-C) template features a narrower top-plane than that of the low-coverage SiN x template, due to the fast growth rate of n-GaN due to the low density of openings on this SiN x nanomask ( Fig. 2(b)). To confirm these changes of 3D nanostructures, we conducted additional (d) Sample-C (SiN x deposition time: 2500 s). The regrown n-GaN density on sample-A, sample-B, and sample-C is 4.7 × 10 10 /cm 2 , 3.7 × 10 10 /cm 2 , and 2.9 × 10 10 /cm 2 , respectively. experiments of 3D n-GaN nanostructures grown at higher growth temperatures and shorter growth times. These demonstrated very clear changes to the 3D GaN multifacets depending on the deposition time of the in-situ SiN x nanomask (see Supplementary Fig. S1). The evolution of the 3D GaN growth mode depended on the mask coverage, in line with a previous study using an ex-situ mask 18 . Finally, we fabricated three types of nanostructure templates by controlling the in-situ SiN x nanomask density during MOCVD. Figure 3 shows a schematic and the surface morphologies of these three n-GaN nanostructure templates. The schematic indicates the nanofacets (blue colour) on top of the 3D n-GaN nanostructures. Based on SEM analysis, the shape of the nanofacets on top of n-GaN nanostructure obviously changed depending on the SiN x nanomask deposition time, described by the 3D GaN growth mechanism shown in Fig. 2. Each inset image shows the difference of the top surface shape of each GaN nanofacet and the facet ratio between P and SP for each. To measure the quantitative aspect ratio of the nanofacets of three templates, we also conducted the atomic forced microscopy (AFM) analysis. Figure 4 shows AFM surface images of the nanofacets on each template; the root-mean-square roughness (R rms ) of these three templates were 31 nm, and 34 nm, and 49 nm for sample-A, B, and C respectively. To clarify the change of aspect ratio among the three types of nanofacets, we measured AFM line scans on a representative nanostructure in each sample; the aspect ratios (h/d) were approximately 0.55(low), 0.69(middle), and 0.78(high), respectively. The differences between the facet ratios (P/SP) are clearly distinguished on each sample. The facet ratio is a very important factor for the adjusting the wavelength in PF-WLEDs using a 3D GaN multifacet technique 15,16 . Consequently we successfully fabricated various 3D n-GaN nanofacet templates using only in-situ SiN x nanomasks.

Indium composition analysis in the PF-WLED.
To fabricate PF-WLEDs, we conducted the growth of LED structures on these n-GaN nanostructure templates with five pairs of InGaN/GaN MQWs and P-type GaN layers in order. All samples were simultaneously grown using the same growth conditions. Each LED was labelled as LED-A, LED-B, and LED-C, based on the previous n-GaN nanostructure templates sample-A, sample-B, and sample-C, respectively. To investigate the indium distribution in the MQWs, we conducted scanning transmission electron microscopy (STEM) and energy-dispersive x-ray spectroscopy (EDS) on LED-A (sample-A: 1500 sec) and LED-C (sample-C: 2500 sec). Figure 5(a,b) show a schematic of two LEDs and the analysis position for each nanofacet. LED-A and LED-C consisted of different nanofacet structures with low aspect ratio and high aspect ratio, respectively. Figure 5(c,d) show the STEM, EDS-mapping, and EDS spot results of the MQWs in each LED structure, respectively. The STEM and EDS-mapping images show that the InGaN/GaN MQWs were grown with different thicknesses on both the P and SP planes of the n-GaN nanofacets, indicated with white arrows in each STEM image. Moreover, the indium EDS spot intensity and spectrum on each facet in both samples obviously indicate that the indium incorporation in the top plane (P) is nearly twice as high as in inclined-plane (SP), and there is no indium intensity at the GaN nanofacet. This is comparable to previous reports that utilize as ex-situ multifacet method 11 . It was reported previously that the diffusivity of indium adatoms on the SP-plane is larger than that on the P-plane. The surface energy on the P-plane is lower than that of the SP-plane because the SP-plane consisted of nitrogen terminated sites, whereas the P-plane is terminated with gallium sites; therefore, more indium adatoms migrate into the P-plane. Moreover, the diffusivity of gallium atoms is also larger at the SP-plane than the P-plane, similar to indium atoms as mentioned previously. As a result, the total thickness of the active region on the P-plane is larger than that of SP 11,19 . Therefore, InGaN QW thickness and indium composition depend on the polarity of the 3D  GaN nanofacet. From these results, the low aspect ratio nanofacet LED (LED-A) has more in-rich sites (P-plane) than that of the high aspect ratio nanofacet LED (LED-C).
Additionally, the cross-sectional TEM analysis was also performed to confirm the presence of indium segregation at each crystal plane (P-plane and SP-plane) of the nanofacet. As shown in Fig. 6(a-d), both LED-A and LED-C show the indium segregation regions on the P-planes. These indium segregations on the P-planes are very similar to the results of previous studies 10,20,21 . However, it is important to note that segregation area of LED-A (low aspect ratio) is wider than that of LED-C (high aspect ratio). On the other hand, indium segregation on the SP plane was hardly observed in both samples. These characteristics are consistent with the results in Fig. 5(c,d). Figure 7 shows the EL results of three PF-WLEDs at the injection current range of 20-100 mA. The reference LED's peak wavelength (λ p ) is 466 nm (blue emission) at 100 mA with a narrow width, as indicated in Fig. 8(b). As shown in Fig. 7(a-c), the spectra of all three LEDs show broad emissions caused by indium fluctuations in the MQWs grown on their nanofacets. EL spectra of these LEDs also show more intense blue emissions (λ p : 435-490 nm) than that of the longer wavelength region (λ p : 520-650 nm) because the nanofacets have more SP-planes (low indium incorporation) than the P-planes (high indium incorporation), as shown in both Figs 4 and 5. Figure 7(a-c) show the obvious differences of in the long wavelength regions (λ p : 570-660 nm) in the EL spectra of the three LEDs, marked by dotted line-circles and described in Fig. 7(d-f). These wavelength variations, which are dependent on the multifacets, were also reported in ex-situ multifacet structures 12,16 . The EL images of each LED at an injection current of 100 mA are shown in Fig. 7(g-i). The T c of each PF-WLED changed with the aspect ratio of the 3D nanofacet. Their colour coordination is indicated using the Commission Internationale de 1'Eclairage (CIE) 1931 chromaticity diagram shown in Fig. 7(j). The T c of the three LEDs was also changed from 6000 K (true white) to 10000 K (cold white), demonstrating that T c control is possible by in-situ nanofacet engineering.

Optical characteristics of the nanofacet-controlled PF-WLEDs.
Optical characteristics of the QW temperature-controlled PF-WLED. It is well known that indium composition in MQWs crucially depends on the growth temperature. So we tried to change MQW's growth temperature only using LED-A structure (low aspect ratio PF-WLED: true white emission). The EL results are shown in the Fig. 8. Figure 8(a-c) and (d-f) show the EL spectra of three reference LEDs and three PF-WLEDs. The quantum well growth temperature (T QW ) of each reference LED is 755 °С (low T QW ), 765 °С (middle T QW ), and 775 °С (high T QW ), respectively. As mentioned previously, all PF-WLEDs have the same nanofacet structure as LED-A [low aspect ratio PF-WLED in Fig. 7(d)]. As shown in Fig. 8(a-c), the centre wavelengths (λ c ) of each reference LED is 499 nm (green), 466 nm (blue), and 456 nm (blue), respectively at an injection current of 100 mA. The indium incorporation rate in InGaN QWs is also inversely proportional to T QW . The EL spectrum of the low T QW reference LED (Fig. 8(a)) shows the lowest optical intensity and the broadest full width at half maximum (FWHM) among the three reference LEDs. It is likely that there are a few indium segregation sites in InGaN QWs caused by partial strain relaxation due to the high indium incorporation at low T QW . On the other hand, the EL spectrum of the high T QW reference LED (Fig. 8(c)) shows the highest optical intensity and narrowest FWHM value. Similar results are also observed in the PF-WLEDs. Figure 8(d-f) show the T QW controlled white spectra of three PF-WLEDs. The EL spectrum of the low T QW PF-WLED shows the highest intensity in the long wavelength region (λ p : 570-660 nm) compared to those of the middle and high T QW PF-WLEDs. Moreover, the λ c of the short wavelength region is 488 nm, which is longer than those of the other LEDs due to high indium incorporation at low T QW . Both the short and long spectral regions show blue-shifts of the peak wavelength with increasing injection current level, which indicates the band filling effect of the local potential minimum in the potential fluctuation of the In-rich InGaN well layer 22 . On the other hand, the EL spectrum of high T QW PF-WLED shows the lowest intensity in the long wavelength region compared to that of the low and middle T QW PF-WLEDs. Moreover, λ c is 459 nm in short wavelength region here, the shortest among the three PF-WLEDs due to the low indium incorporation rate in InGaN QWs at high T QW . Figure 8(d-f) show the λ c variation in the short wavelength region by T QW was similar to that of the reference LEDs, whereas the variations in the long wavelength based on T QW was more remarkable because the indium incorporation rate of In-rich QWs on the narrow top region (P-plane) is more sensitive in regards to T QW than the SP-plane. EL images of each PF-LED at an injection current of 100 mA are shown in Figure 8(g-i), and their colour coordination is indicated by the Commission Internationale de 1'Eclairage (CIE) 1931 chromaticity diagram shown in Fig. 8(j). As shown here, T c of the low T QW PF-WLED decreased to 4500 K, compared to that of the middle T QW PF-WLED (6000 K) due to the high intensity in the long wavelength region (Fig. 8(d)). On the other hand, T c of the high T QW PF-WLED increased to 8500 K due to the dominant short wavelength spectral region of the low T QW PF-WLED. These results indicate that T c of the PF-WLEDs could be controlled from the warm white region to the cool white region by changing the T QW condition on the 3D GaN nanostructures in PF-WLEDs.
The results also showed the obvious difference of light intensity between reference blue LED and suggested PF-WLEDs (see Supplementary Fig. S3). However, it is notable that the light intensity of LED-A (T QW : 775 °C) was significantly higher than those of other proposed PF-WLEDs, which suggest the possibility of light intensity enhancement of proposed PF-WELDs through further research.

Conclusions
We demonstrated colour-crafted PF-WLEDs using in-situ nanostructure engineering for the first time. Here we used only conventional blue or green wavelength QW growth conditions to change the white colour-temperature in our in-situ nanofaceted PF-WLEDs. This can increase the commercialization potential of our approach as most similar systems require more complicated growth techniques to protect against degradation of the In-rich InGaN QW region. Our technology uses a commercial MOCVD system without external photolithography for an in-situ one-step growth process to control the T c of the white emission in PF-WLEDs. Therefore, this practical method has a low technical barrier. Moreover, there was no wafer contamination due to the lack of ex-situ patterning. We believe that this in-situ method can be applied to large-diameter wafers to fabricate PF-WLEDs with various T c values (see Supplementary Fig. S2). Further work in necessary for improving optical performance, and is planned as future research. In this study, we preferentially focused on the possibility of T c control of PF-WLEDs using our new approach. Finally, we expect that this in-situ nanostructure variation technique with in-situ SiN x nanomask control will be useful for a variety of nanotechnology applications.