Monolithically integrated green-to-orange color InGaN-based nanocolumn photonic crystal LEDs with directional radiation beam profiles

Monolithic integrated red, green, and blue (RGB) micro-sized light-emitting diodes (LEDs) have gained interest for their potential applications in next-generation full-color micro (μ)-LED displays, applicable to head-mounted displays, smart watches, and μ-LED projectors. ^1–3) InGaN-based LEDs emit the entire range of visible light by changing the composition of In. To integrate RGB-LEDs on the same wafer in the vicinity of the micrometer scale, different technologies have been reported. ^4–7) In addition to solving various fabrication problems, the optical properties should be improved. This concerns the light extraction efficiency, wavelength stability against temperature and injection current, and wavelength accuracy; the emission spectrum should be narrow for applications. Photonic crystal (PC) LEDs are expected to have improved optical properties. ^8–12)

GaN nanocolumns are one-dimensional columnar nanocrystals, the diameters and positions of which are precisely controlled using Ti-mask selective area growth (SAG), ¹³⁾ fabricating ordered InGaN/GaN nanocolumn arrays in a triangular lattice. The emission colors of the InGaN/GaN nanocolumn arrays prepared on the same substrate can be changed from blue to red by controlling the nanocolumn diameter (D). ¹⁴⁾ Using this emission color control scheme, the monolithic integration of 5 × 5 μm² emission area nanocolumn μ-LED pixels of four colors was successfully realized. Therein, the LED pixels spaced at a distance of 10 μm from each other were two-dimensionally (2D) arranged and independently driven with matrix wiring electrodes. ¹⁵⁾ The periodic arrangement of the nanocolumns induced the PC effect, which contributed to the directional radiation of the light beam from the nanocolumn LEDs, resulting in an improvement in the light extraction efficiency. ^8,9,16) The photonic band diagram of a two-dimensional PC with a triangular lattice (lattice constant: L) provides several band edges, at which different types of wave coupling occur through 2D Bragg diffraction. Through light diffraction at photonic band edges, the light beam is extracted directionally, which prevents the mixing of emissions from neighboring LEDs and also improves the extraction efficiency. ^17–19) In addition, the photonic band edge shifts caused by changes in temperature and injection current are negligibly small, and the full width at half maximum (FWHM) of the directionally extracted light beam is narrow because the light beam is extracted through 2D diffraction. ^8,9) In the nanocolumn system, the photonic band-edge wavelength (λ_B) is determined by the designed parameters of the nanocolumn diameter (D) and L. At the same time, the emission wavelength from the ordered InGaN/GaN nanocolumn arrays can also be controlled by D and L. ¹⁴⁾ However, for nanorod PCs prepared through top–down etching, ⁶⁾ the composition of InGaN is given by the underlying crystal and does not change with D and L.

In this letter, we demonstrate the monolithic integration of nanocolumn PC LEDs with different emission colors (green, yellow, and orange), exhibiting narrow radiation beam angles (<±30°). We evaluated the radiation beam and spectral properties of the nanocolumn LEDs based on the PC effect. Using the experimental result of the map of normalized photonic bandedge wavelength (L/λ_B) against filling factor (FF) defined by (D/L)², we suggested control methods of the radiation beam characteristics and emission colors by D and L.

Figure 1(a) schematically shows the monolithically integrated multicolor nanocolumn PC LED structure. Four pn-junction InGaN/GaN ordered nanocolumn arrays, in which nanocolumns were arranged in a triangular lattice with L = 320 nm, were prepared next to each other on the same substrate. A cross-sectional schematic diagram of a nanocolumn LED is shown in Fig. 1(b). The ordered GaN nanocolumn arrays were prepared by Ti-mask SAG using RF plasma molecular beam epitaxy (RFMBE); ¹³⁾ on top of the Si-doped GaN nanoclumns (500 nm thick), subsequently InGaN active layers (120 nm), Mg-doped p-type GaN cladding (270 nm), and Mg-doped InGaN contact layers (30 nm) were grown. A 300 nm thick indium tin oxide (ITO) film was used to prepare circular electrodes with a diameter of 100 μm on the top of the grown crystals [see Figs. 1(c) and 1(d)]. In the experiment, we interrupted the nanocolumn growth before the growth of the InGaN active layer and measured the n-GaN nanocolumn (D_n-GaN) to be 140, 180, 240, and 260 nm for LED-1, 2, 3, and 4, respectively, where the D_n-GaN is controlled by the nanohole size of the Ti-mask SAG pattern. After the growth, the diameters at the p-type GaN region (D_p-GaN) increased to be 280, 280, 310, and 310 nm, respectively. However, cross-sectional TEM observations in our previous study indicated the nanocolumn diameters in the InGaN region were closer to the D_n-GaN than the nanocolumn diameter in the p-GaN region. ⁹⁾ Comparing the SEM images of LED-1 and 2 with those of LED-3 and 4 in Fig. 1(d), we note that the hexagonal prismatic nanocolumn crystals of LED-1 and LED-2 were rotated by 30° around the c-axis against the underlying n-GaN nanocolumns. It was frequently observed that the Mg doping in p-GaN growth induces the crystal rotation for the n-GaN nanocolumn arrays having a low FF such as LED-1 and 2, though there is no space of rotation for high-filling factor nanocolumns such as LED-3 and 4; this phenomenon occurs probably because the Mg doping enhances the lateral growth along the M axis at the side wall.

Zoom In Zoom Out Reset image size

The nanocolumn LEDs were evaluated under DC current injection at room temperature by a micro-electroluminescence (micro-EL) system, observing from green to orange emissions. Figures 2(a)–2(d) show the normalized emission spectra of LED-1, 2, 3, and 4, respectively, at an injection current of 10 mA. Here, we have three series of spectra, for which the emission intensity was detected using different magnifications of 40× and 4× objective lenses. For LED-2, the spectrum was measured additionally when a pinhole of a 3 mm diameter was placed in the 4× objective lens barrel (see solid red line). The 40× objective lens with an NA of 0.60 (acceptance angle 2θ_max = 73.6°), enables the detection of a widely spread radiation beam (see dotted black lines), which is suitable for detecting the Lambert profile of escaping light from ordinary LEDs. At the same time, the 4× objective lens possesses an NA of 0.16 (2θ_max = 18.4°), and further use of a pinhole reduces the acceptance angle to less than 1°. Thus, using these, a directional light component in the vertical direction from the surface of LEDs can be preferentially detected.

Zoom In Zoom Out Reset image size

The emission spectra detected with the 40× objective lenses are considered to exhibit the band-to-band emission of the InGaN active layer, although small modifications in the spectrum were observed owing to the PC effect. The peak wavelengths (λ_P) of the dotted black lines corresponding to the bandgap wavelength were estimated by fitting with Gaussian beam profiles to be 551, 543, 573, and 597 nm, respectively. The bandgap wavelengths of the ordered InGaN nanocolumn arrays shift to larger values, which is caused by the increased In composition of InGaN as the D_n-GaN increases. ¹⁴⁾ The FWHM of values of LED-1, 2, 3, and 4 were 57.8, 58.8, 58.4, and 62.7 nm, respectively.

The spectra obtained with the 4× objective lens, which exhibit the behavior of directional components of the radiation beam, are composed of two components: a sharp dominant peak caused by the PC effect and a weak broadened emission component at the bottom of the spectrum. The latter is caused by the original Lambert-type broadened radiation of the LED. The sharp emission peaks originate from the wave coupling through the light diffraction at the photonic band edge of the Γ points for the nanocolumn PC. ¹⁷⁾ In Fig. 2, strong and sharp emission lines appear in the spectra of LED-2, 3, and 4, with dominant peaks at 544, 594.5, and 607.5 nm, respectively. For LED-3, the second dominant peak is observed at 568.4 nm. These wavelengths correspond to the photonic band edge wavelength λ_B, as discussed in previous papers in Refs. 8, 9. The spectral linewidths of the dominant peak for LED-2 are 4.5 nm for the detection using the 4× objective lens with the pinhole, as shown in Fig. 2(b); the spectral linewidths of the nanocolumn PC LEDs dramatically decrease in order of 58.8, 28.8, and 4.5 nm with decreasing acceptance angle of the detection system in order of 73.6°,18.4°, and 1°, respectively. The wave coupling through the light diffraction at the photonic band edge generates a directional radiation beam conducive to such small linewidths. ⁸⁾ Such dominant sharp emission peaks are not observed for LED-1, even though an intensity enhancement at approximately 520 nm [see the arrow in Fig. 2(a)] was observed for the 4× objective lens observation, but several fine peaks appear, which are probably caused by Fabry–Perot resonance between the boundaries at the underlying sapphire substrate and the upper ITO layer.

The emission spectra of LED-4 were measured at an injection current of 1 mA as a function of the radiation angle (θ) measured from the vertical direction to the nanocolumn surface, where the light intensity was detected through a fiber probe system with an optical coupling lens that can detect the beam diameter of 3 mm. The detection arm with the fiber probe was vertically rotated in a circle from 0° to 90° at an increment of 1.0° along the a-axis and m-axis of the hexagonal-shaped nanocolumn system [see the inset in Fig. 3(b)]; here, both the TE and TM modes were detected. The distance from the fiber probe to the LED was 28 cm, and the lens diameter was 3.0 mm; thus, the angular resolution of the measurement system was 0.6°. Figure 3(a) shows the spectrum at θ = 0^o, where the light intensity is enhanced in the wavelength range from 590 to 612 nm, with the dominant emission peaks related to the photonic band edges at 601, 604, and 608 nm. Reflecting a small acceptance angle of detection (0.6^o), the spectrum component based on the ordinary LED behavior is more suppressed, as compared with the case shown in Fig. 2(d).

Zoom In Zoom Out Reset image size

The photonic band diagram was obtained for LED-4 using the angular dependence of the emission spectrum, as shown in Fig. 3(b). Here, the spectral light intensity is plotted by the image contrast for the vertical axis of the normalized frequency L/λ (λ: wavelength) and the horizontal axis of the normalized in-plane wave vector k_∥ L $\left({k}_{\parallel }=\tfrac{2\pi }{\lambda }{\rm{s}}{\rm{i}}{\rm{n}}\theta \right).$ The point of L/λ_B = 0.53 [λ_B = 604 nm; β in Fig. 3(b)] corresponds to the dominant peak of LED-4, indicating a clear photonic band edge, where two bands seem to degenerate. Near here, another photonic band edge appears at L/λ_B = 0.532 (601 nm; α) and 0.526 (608 nm; γ), respectively. Five photonic bands related to the three TE and two TM mode blanches are calculated to be nearly degenerated at the Γ point (i.e. k_∥ L = 0), as discussed for triangular lattice nanocolumn arrays. ^7,20) Note that the next higher-order band edge possesses the one TE and two TM mode blanches.

Next, the angular profiles of the radiation beams were evaluated for LED-2, 3, and 4. The light intensity for individual θ values was evaluated by integrating the spectrum over the wavelength range of 450–700 nm, and plotted as a function of rotation angle, as shown in Fig. 4(a). Note that the radiation beam profile of ordinary planar-type LEDs spreads widely in the Lambert radiation profile of ${I}_{\theta }={I}_{0}{\rm{c}}{\rm{o}}{\rm{s}}\theta$ with a radiation angle of ±60°, and exhibits a broad emission spectrum. In contrast, nanocolumn LED-2, 3, and 4 exhibit directional radiation beam profiles with a radiation angle of approximately ±30°, which is attributed to the nanocolumn and PC effects. The light strongly interacts with the nanocolumn PC system at specific wavelengths (i.e. the photonic band edges, λ_B), directionally extracting the specific wavelength components. The emission intensities at the dominant peak components at 601, 604, and 608 nm corresponding to the photonic band edges [see Fig. 3(b)] are plotted as a function of θ, as shown in Fig. 4(b), indicating that their light intensities are concentrated within a radiation angle of ±15°. The diffraction at the band edge of the Γ point radiates the light vertically from the top surfaces of the nanocolumns, creating a directional radiation beam profile. Next, we integrate the light intensity observed in the angle and wavelength ranges of θ < ±15° and 590–612 nm, which corresponds to the directional enhancement of the radiation beam profile caused by the PC effect; the ratio of the integrated intensity caused by the PC effect to the total emission intensity can be estimated as 16.7%. In Fig. 4(c), the blue solid line shows the subtraction of the intensity from 590 to 612 nm (red solid line) from the total integrated intensity for the entire wavelength range (black solid line), which indicates the component excluding PC effects, though the intensity enhancement from the Lambert profile (blue shadowing region) indicates the nanocolumn effect.

Zoom In Zoom Out Reset image size

Figure 5(a) shows λ_P and λ_B for LEDs 1, 2, 3, and 4, which are plotted as a function of the D_n-GaN of each LED (λ_P is the bandgap wavelength of InGaN and λ_B is the photonic band edge for the LED crystals) Here, the λ_B was plotted basically for the dominant peaks with the highest intensity (main-peak) in Fig. 2, although the second highest point (sub-peak) was additionally plotted for LED-2 with smaller star symbol for indicating the interband transition. For LED-1, no sharp dominant peaks were observed and, thus a clear PC effect was not observed, which is because the InGaN bandgap wavelength was somewhat apart from the photonic band edge. λ_HM is the half maximum point of the 40× objective lens spectrum of LED-3, which are 537 and 602 nm at the shorter and longer points, respectively [see Fig. 2(c)].

Zoom In Zoom Out Reset image size

In this experiment, different InGaN nanocolumn arrays were integrated side by side, decreasing D_n-GaN from one to another under a constant L(320 nm). The FF of the n-GaN nanocolumn region, here defined by (D_n-GaN/L)² , decreases, lowering the effective refractive index of the nanocolumn system, thus shortening the λ_B value. However, the abrupt wavelength jump of λ_B occurs, as shown in Fig. 5(a), which is caused by the transition between different photonic band edges. Figure 5(b) shows the FWHMs of LED-2 to 4, which exhibited a clear PC effect, as a function of the difference between λ_P and λ_B. Note that the FWHM of an ordinary film InGaN LED decreases with decreasing In composition; ²¹⁾ however, the FWHM becomes narrower when the ∣λ_P–λ_B∣ decreases in Fig. 5(b), which is attributed to the emphasized emission peak by the PC effect.

The structural dependency of λ_B was evaluated for InGaN/GaN nanocolumn LED crystals with different L = 320, 300, and 280 nm through the 4× objective lens observation of photoluminescence spectra [see Fig. 5(c)]. In the figure, the vertical axis is the normalized frequency L/λ, and the horizontal axis is the FF of the n-GaN nanocolumn region. The dominant peak wavelengths for the spectra of LED-2, 3, and 4 observed by the 4× objective lens are plotted by symbols of large blue stars, showing the sub-peaks with the small ones. As can be seen in Fig. 5(c), LED-2, 3, and 4 interacted with the higher-order bands, and the transition of the dominant peak related to λ_B from one band to another higher band appears from LED-4 and 3 to LED-2. For LED-3, double peaks, which belong to different higher-order band edges separated by ∼30 nm, were observed [see the solid line in Fig. 2(c)]. The emission spectrum from InGaN spreads covering both band edges as indicated by λ_HM points.

LED-1 and 2, interacting with the second higher-order band edge, exhibited a lower PC effect compared with LED-3 and 4, which is probably attributed to the band structure at the edge, that is, the difference in the number of nearly degenerated bands. However, L = 300 nm was taken to design LED-1 and 2, all LEDs could operate on the same series of first higher-order band edge. The light coupling phenomena dependent on the different band edge will be clarified. Note that the wavelength of the nanocolumn PC can be appropriately designed based on the Γ₁₁ band edge because the normalized frequency L/λ_B is stable at ∼0.5 for different D_n-GaN and L. In this study, λ_P and λ_B were controlled by changing the D_n-GaN under a constant L(320 nm); however, the approach of changing both D_n-GaN and L could expand the capability of controlling the operation wavelength over a wide range, such as from blue to red.

In summary, we demonstrate the monolithic integration of PC nanocolumn LEDs with different emission colors (λ = 543, 573, and 597 nm, from green to orange) with the directional radiation properties of radiation beam angles of approximately ±30°. The controlled L and D of the nanocolumn arrays change both the InGaN bandgap and the photonic band edge wavelengths of the nanocolumn arrays prepared on the same substrate. Although the change mechanism is different for the two, matching the bandgap wavelength of InGaN to λ_B can introduce a highly directional radiation beam as well as reduce the spectral FWHM. Note that an effective method for the complete matching between them in a wide range of wavelengths is still not fully understood. In the experiment, the operation wavelength transited between different photonic bands, which was attributed to an insufficient control of L and D. Therefore, we should develop a more specific design scheme for matching λ_B to the InGaN wavelength. The experimental result of the map of L/λ_B against FF suggests that the design scheme of λ_B based on the Γ₁₁ band edge is suitable for controlling λ_B at the bandgap wavelength of InGaN. In this experiment, only D was changed, and L was constant. Therefore, controlling L of nanocolumn LEDs in addition to D as design parameters would effectively contribute to broadening the operation wavelength range, for example, from blue to red.

Acknowledgments