Wafer-bonding AlGaInP light emitting diodes with pyramidally patterned metal reflector

    Corresponding author: Zhiyuan Zuo, zuozhiyuan@inspur.com
  • 1. School of Physics and Engineering, Sun Yat-Sen University, Guangzhou 510275, China
  • 2. INSPUR GROUP Shandong Inspur Huaguang Optoelectronics Co., Ltd, Ji'nan 250100, China
  • 3. State Key Laboratory of Crystal Material, Shandong University, Ji'nan 250100, China

Key words: pyramidally patterned reflectorlight emitting diodeswafer-bondinglight extraction efficiency

Abstract: We demonstrate and introduce here a pyramidally patterned metal reflector into wafer-bonding AlGaInP light emitting diodes (LEDs) to improve the light extraction efficiency by using a photo-assisted chemical etched GaP:Mg layer. The pyramid patterns were fabricated employing a HF and H2O2 mixed solution in combination with a 532 nm laser on a GaP:Mg surface firstly, and then a gold reflector layer was evaporated onto the patterned GaP:Mg surface. After the whole chip process, the patterned gold reflector structure was confirmed to be efficient for light extraction and a 18.55% enhancement of the electroluminescent flux has been obtained by an integrating sphere, compared to the surface textured LEDs with flat reflectors.


1.   Introduction
  • During the past decade, light emitting diodes (LEDs) devices have witnessed a great evolution, which have already been widely realized as the next generation light sources[1] and display devices[2, 3]. However, conventional LEDs today still suffer from having a low light extraction efficiency due to the total internal reflection in their semiconductor film interfaces and the Urbach absorption[4] of the materials. For example, in the case of as grown (AS-type) AlGaInP LEDs with a GaP:Mg window layer, only $\sim $2.5 % of light generated in the quantum wells (QWs) can escape from the device[5] due to the high refractive index of GaP ($n$ $=$ 3.3 @ 650 nm)[6], and most photons are trapped inside the chip lost into guided modes. Extensive efforts over the past decade have yielded many techniques, such as encapsulation, shaped LEDs[7], patterned substrate[8], imbedded reflectors[9], resonant cavity[10], surface texturing[11, 12], and photonic crystals (PhC)[13, 14] to improve the light extraction efficiency of LEDs. Among these, imbedded reflectors can be easily integrated into the wafer-bonding LEDs and efficiently improve the light extraction efficiency. The improvement in AlGaInP LEDs grown on a GaAs substrate and transferred to the silicon substrate is much more remarkable due to the bottom absorption[15]. Although the imbedded metal reflector used in AlGaInP wafer-bonding LEDs is a mature technology, experimental results of a metal reflector with light extraction patterns fabricated by wet etching methods have seldom been reported.

    In general, pyramid patterns are highly desired for light extraction from LEDs. For example, pyramid structures created on laser lift-off (LLO) n-polar GaN by using potassium hydroxide[16] or phosphoric acid[17] under heating or ultraviolet (UV) radiation showed a remarkably high light extraction efficiency. However, a similar approach for GaP is still absent and highly demanded before our previous studies[18]. Although AlGaInP LEDs have a much higher internal quantum efficiency and surface texturing can improve the light extraction efficiency greater than the flat LEDs, AlGaInP LEDs still have low level wall-plug efficiency compared with the GaN/sapphire LEDs due to their much lower light extraction efficiency. The lower wall-plug efficiency restricts its applications such as the high brightness outdoor display, high color rendering illumination and RGB projection. Considering the discussions above, AlGaInP LEDs still need new techniques with advantages as low cost, highly controlled and uniform to promote its light extraction efficiency. To meet the requirements above, here we employ a photo-assisted chemical method to create pyramid patterns in the GaP:Mg spreading layer and then a gold reflector is fabricated on the patterned GaP surface to form the patterned reflection structures. It is verified that the light extraction efficiency of wafer-bonding AlGaInP red LEDs can be dramatically improved by 18.55 %. Finally, this method has also been proved in the mass production of LEDs.

2.   Experimental
  • The LED wafers were grown on a GaAs substrate by metal-organic chemical vapor deposition (MOCVD) and contained a substrate etching block layer, an n-type doped ohmic contact layer, an n-AlGaInP window layer, an n-AlGaInP cladding layer, 15 pairs of GaInP/AlGaInP quantum wells (QWs), a p-AlGaInP cladding layer and an 8 $\mu $m GaP:Mg spreading layer whose orientation is 15\r{ } off-axis from [100] to [111]. All the structures are shown in Figure 1. In the flat reflector chip process, firstly, we fabricated ohmic contact and current channels, and then grew a SiO$_{2}$ insulating layer with apertures by Plasma Enhanced Chemical Vapor Deposition (PECVD, Oxford, 800 plus) onto the ohmic contact structures. Finally, a gold reflector was evaporated onto the SiO$_{2}$ layer by electron-beam evaporation (EBE). The etchant contains 10 % hydrofluoric acid (HF) and 1 M hydrogen peroxide (H$_{2}$O$_{2})$. In the patterned reflector process, a photochemical reaction of GaP:Mg induced by a 532 nm laser had been carried out firstly as the photon energy is larger than the GaP band gap. The etching radiation power density of the laser was 5 mW/cm$^{2}$ during the reaction. The etching times are 15 min according to our previous optimization[18]. After the pattern fabrication, a normal wafer-bonding chip process has been carried out: a 300 nm gold reflector layer has been obtained on the patterned GaP:Mg surface by using electron beam evaporation (ULVAC EI-5Z) after the p-type ohmic contact finished. Then wafer-bonding, GaAs substrate removing and p/n electrodes are processed successively. Next, the surface textured n-AlGaInP window layer has been obtained by wet etching using hydrochloric acid in both samples. Finally, a Si$_{x}$O$_{y}$ passivation layer is made by PECVD. Finally, we obtain 10 $\times$ 10 mil$^{2}$ wafer-bonding AlGaInP chips of flat reflector and patterned reflector. A Hitachi S-4800 field emission scanning electron microscope (FESEM) was employed to study the surface morphology after the etching process. The electroluminescent (EL) spectra were obtained by using a fiber spectrometer (Avantes Spec-2048, Netherlands), a source-measure unit (Keithly 4200, USA), and a probe station (Suss PM5, Germany). All the measurements were carried out in a clean room at room temperature.

3.   Results and discussions
  • Figure 2(a) shows the SEM image of the GaP:Mg surface after the chemical etching process, but without reflector evaporation. The SEM image reveals that the GaP:Mg epilayer is highly etched because of the strong absorption of 532 nm laser radiation, and plentiful carriers (electrons and holes) are created in the GaP layer. These free carriers are very reactive, enabling chemical reactions with etchants. A pyramidal pattern with a density of $\sim $10$^{8}$ cm$^{-2}$ appeared, and the pyramid sizes are measured to be 1000 $\pm$ 100 nm after the etching process. As the GaP:Mg layer is employed in AlGaInP LEDs to spread current flow, the etch process should be controlled carefully to avoid over reaction. Figure 2(b) shows a cross section SEM image of the sample after etching. The image reveals that the remaining GaP:Mg layer is $\sim $5 $\mu $m thick at the lowest position. The pyramidal patterns have an average height of 1000 $\pm $ 100 nm. Figures 2(c) and 2(d) show the SEM image of patterned GaP:Mg with gold reflector. Figure 2(c) shows the total morphology and Figure 2(d) shows the cross section. Figures 2(e) and 2(f) show optical images of gold reflectors without and with pyramidal patterns obtained by an optical microscope; the gray dots patterns in both images are current channels of the chip made by the Be alloy, which can obtain ohmic contact to GaP:Mg. The colour of the patterned reflector turned darker, indicating that the pyramid pattern is more efficient for light scattering.

    Figure 3 shows the $I$-$V$ curves of the two samples with and without patterned reflectors. The threshold voltage of hetero-junction $V_{\rm th}$ in the LEDs is commonly approximated as:
    $V_{\rm th} =\frac{KT}{e}\ln \frac{N_{\rm A} N_{\rm D} }{n_{\rm i}^2 }\approx \frac{E_{\rm g}}{e}, $(1)
    \label{eq1} where $N_{\rm A}$, $N_{\rm D}$, and $n_{\rm i}$ are the acceptor, donor and intrinsic carrier density, respectively, and $E_{\rm g}$ is the band gap of the QWs. Although the thickness of the GaP layer is thinner due to the etching process, the threshold voltage of the patterned chips is $\sim $10 meV higher than the flat reflector sample. This is because the highly textured GaP layer enlarges the resistivity and reduces the current spreading efficiency. Firstly, hole transportation could be hindered by the pyramid pattern. Our tests reveal that the hole mobility in the patterned GaP layer decreases from 120 $\pm$ 10 to 95 $\pm$ 10 cm$^{2}$/(V$\cdot $s), and the resistivity increased from $\sim $2.0 $\times$ 10$^{-2}$ to $\sim $8.5 $\times$ 10$^{-1}$ $\Omega $$\cdot $cm. Secondly, surface damage induced by etching could also degrade the ohmic contact between the Be alloy and GaP. Despite the rise of threshold voltage, series resistance is also increased slightly after etching.

    Figure 4 shows the box {\&} whisker plot of light intensity (Figure 4(a)), voltage (Figure 4(b)), dominant wavelength (Figure 4(c)) and peak wavelength (Figure 4(d)) of flat reflector and patterned reflector samples obtained on wafer test, respectively. All the tests were carried out under 20 mA injection, which is the normal working current level. Figure 4(a) reveals that the light intensity of patterned reflector chips is $\sim $15.9 % lower than the flat reflector chips under 20 mA injections due to the change of the light output path caused by the patterned reflector. Further, we also find that the concentration ratio of flat reflector chips is better. Figure 4(b) shows the voltages under 20 mA injection; the flat reflector sample presents a lower voltage than the patterned reflector sample, the reason is analyzed in the $I$-$V$ discussion section. From Figures 4(c) and 4(d) we can find that the dominant wavelength of the flat reflector sample is longer than the patterned reflector one, but in the peak wavelength comparison, the result is inverse. This phenomenon is due to the nonuniformity of the component and structure in MQW, which can cause the single emission peak to split into multi-emission peaks. The multi-emission peaks will disturb the human eye sensitivity Gaussian conversion.

    Figure 5 shows the box {\&} whisker plot of luminous flux (Figure 5(a)), output power (Figure 5(b)), luminous efficiency (Figure 5(c)), and peak wavelength (Figure 5(d)) of the flat reflector and patterned reflector samples obtained after encapsulation respectively. All tests were carried out under 20 mA injection as wafer testing. Figure 5(a) reveals that the luminous flux of patterned reflector chips is $\sim $18.55 % higher than the flat reflector chips under 20 mA injections; this phenomenon proves our analysis in the wafer test to be correct. The patterned reflector changed the path of light output, vertical output was reduced while the horizontal output was enhanced, and the luminous flux was stronger in the total angle. Figure 5(b) shows the output light power under 20 mA injection, the patterned reflector sample presents a 21.88 % enhancement over the patterned reflector sample, despite wavelength influence. From Figure 5(c) we can find that in the luminous efficiency comparison, the patterned reflector sample presents a 21.15 % enhancement over the patterned reflector sample, which is matched with the luminous flux and output power results. Figure 5(d) shows the peak wavelength of the two samples, and there is a $\sim $0.5 nm difference between them.

4.   Conclusion
  • In summary, pyramidally patterned reflectors are fabricated on GaP by an etchant containing 10 % hydrofluoric acid and 1 M hydrogen peroxide in combination with 532 nm laser radiation. It is confirmed that the pyramid patterned reflector can be utilized to increase luminous flux and luminous efficiency by 18.55 % and 21.15 % respectively. This technique can greatly increase the efficiency of LEDs and can be easily integrated into the industry.

Figure (5)  Table (1) Reference (18) Relative (20)

Journal of Semiconductors © 2017 All Rights Reserved