Gallium nitride light emitting diode (GaN-LED) chips are considered as the promising element of display walls, flat panel displays, transparent displays, near-eye displays, and wearable displays due to their advantages of high brightness, high efficiency, and low power consumption [1–9]. Because GaN-LED chips are obtained by etching LED epitaxial wafers, a high-quality LED epitaxial wafer, including high uniformity of electroluminescence (EL) wavelengths and electrical characteristics, is the foundation for preparing high-quality GaN-LED chips [10]. Therefore, selecting LED epitaxial wafers with EL wavelengths and electrical characteristics that meet the display requirements is a prerequisite for obtaining high-performance displays with brightness and color uniformity. Thus, LED epitaxial wafer inspection is necessary before the etching process.

The first strategy for LED epitaxial wafer inspection is photoluminescence (PL) inspection. In the inspection process, the LED epitaxial wafer is illuminated with an ultraviolet beam, which excites the multi-quantum wells (MQWs) in the epitaxial layer to generate photoluminescence. One outstanding advantage of this technology is that the PL inspection is a non-destructive method. However, PL inspection can only detect the optical characteristics of the MQWs, and the effect of other functional layers, including n-GaN, p-GaN, and undoped-GaN, on future EL performance cannot be determined. In other words, the PL results cannot fully reflect the EL results when the LED chips are fabricated and used. As a result, the PL inspection accuracy is relatively low.

In addition to PL inspection, electroluminescence (EL) inspection is also used in LED epitaxial wafer inspection. This method involves inserting two metal probes into the p-GaN layer and n-GaN layer of the LED epitaxial wafer, respectively. Then, a forward DC bias is applied between the two probes to generate electroluminescence. By touching the probe to different p-GaN regions on the LED epitaxial wafer, the optical and electrical characteristics of different regions can be obtained. The optical and electrical parameters obtained by this method are accurate, but pretreatment of the wafer is necessary before inspection. The pretreatment includes using a high-precision diamond cutter or laser to etch the LED epitaxial wafer into the n-GaN layer, so that the one of the probe can contact the n-GaN layer [11]. In addition, small indium pads are usually evaporated on each area of the LED epitaxial wafer, so as to form a good ohmic contact between the probe and the p-GaN layer. Moreover, when the probe comes into contact with the p-GaN, it can also cause mechanical damage to the LED epitaxial wafer, and reduce the quality of the LED epitaxial wafer.

Recently, the non-contact operation and single-contact operation have been proposed and applied in various light-emitting device structures, including GaN-based LEDs and quantum-dot-based devices [12–23]. As for the non-contact operation mode, the ohmic contact between electrodes and LED is avoided. Two insulating layers are introduced between the electrodes and the LED [12–21]. Similarly, for the single-contact operation mode, an insulating layer is introduced between the electrode and the p-GaN terminal (or n-GaN terminal) [22–24]. Due to the presence of the insulating layer, DC voltage cannot drive the LED because the insulating layer would block the injection of charge carriers. Therefore, high-frequency AC power must be used to drive this non-contact or single-contact LED. The advantage of non-contact or single-contact mode is that it does not require complex pre-treatment on the LED surface to achieve ohmic contact. This advantage is particularly important in the inspection of LED epitaxial wafers. In this work, we propose a non-destructive inspection method and system for LED epitaxial wafers based on soft single-contact electroluminescence (SC-EL) operation mode. The proposed SC-EL inspection can detect the electrical characteristics as well as the EL properties of LED epitaxial wafers without introducing mechanical damage and without etching the LED epitaxial wafers. In addition, it is demonstrated that the SC-EL inspection has a higher EL wavelength accuracy than conventional PL inspection. Moreover, by using the SC-EL inspection, we can also get the electrical characteristics of LED epitaxial wafers, which cannot be obtained by using conventional PL inspection. The proposed SC-EL inspection can ensure high inspection accuracy without causing damage to the LED epitaxial wafer, which holds promising application in LED technology.

The SC-EL inspection system consists of an X–Y axis electric moving stage, a Z axis electric moving stage, and an optical-electrical probe, as schematically illustrated in Fig. 1(a). The electric moving stages have a step accuracy of 1 μm. An indium tin oxide (ITO) glass is set on the X–Y axis stage. The X–Y axis electric moving stage is used to control the movement of the ITO glass and the LED epitaxial wafer, while the Z axis moving stage is used to control the movement of the optical-electrical probe. The optical-electrical probe consists of a hollow copper tube (1 mm in internal diameter, 1.5 mm in external diameter) with optical fiber inside, and a conductive rubber tube wrapped around outside the copper tube. The conductive rubber is soft; thus, micro deformation would occur when in contact with the LED epitaxial wafer, thereby minimizing mechanical damage to the LED epitaxial wafer. One end of the optical fiber is used to collect the emitted light from the LED epitaxial wafer during the inspection process, and the other end is connected to a spectrometer (Ocean Insight Maya2000 pro). The spectrometer can collect spectral information. During the inspection, the LED epitaxial wafer is placed on the ITO glass, and an AC voltage (generated by RIGOL DG4162 and Aigtek ATA-2161) is applied to the optical-electrical probe while the ITO glass is grounded. The applied AC voltage and AC current are measured by using an oscilloscope (RIGOL DS7024). The picture of the SC-EL inspection system is shown in Fig. 1(b).

During the inspection process, an AC voltage (70Vpp, 100 kHz) is applied between the optical-electrical probe and the ITO electrode, and then the probe is slowly moving down until it comes into contact with the LED epitaxial wafer and generates electroluminescence. The light emitted from the LED epitaxial wafer is collected through a spectrometer. After finishing the detection of this area, the probe is automatically lifted up and the LED epitaxial wafer is automatically moved to another position. Finally, the probe is slowly moving down until it comes into contact with the LED epitaxial wafer again. Repeat the above steps until all areas of the entire LED epitaxial wafer have been inspected, and the entire inspection process is completed.

Additionally, the PL inspection and conventional EL inspection are also performed. During the PL inspection process, an ultraviolet light with a wavelength of 350 nm is used. As for EL inspection, a metal probe is used to touch the p-GaN of the LED epitaxial wafer, while another metal probe is used to contact the n-GaN of the LED epitaxial wafer. Then, a DC voltage between the two probes is applied. A spectrometer is used to collect the optical parameter during the PL inspection and the conventional EL inspection.

The working mechanism of the single-contact light emitting under AC voltage is schematically illustrated in Fig. 1. When the voltage is rising in a positive half cycle [Fig. 1(c)], the probe potential is positive and the ITO is always grounded, and the applied electric field follows the direction from p-GaN to n-GaN. Therefore, holes are injected into the p-GaN through the conductive rubber. Due to the existence of the sapphire layer, electrons cannot be injected into the n-GaN through the ITO glass. However, electrons remaining in the n-GaN can transport to the MQWs under the external electric field. Therefore, radiative recombination occurs in the MQWs region. At the same time, the number of pre-accumulated holes on the sapphire/ITO interface decreases and finally electrons accumulate on the sapphire/ITO interface. When the voltage rises to the maximum value [Fig. 1(d)], a “depletion region” with positive charge centers appears on the sapphire/n-GaN interface, similar to the depletion region in the PN junction, and the number of electrons accumulated on the sapphire/ITO interface also reaches the peak value.

When the applied voltage is in the reduction stage of the positive half cycle and in the rising stage of the negative half cycle, the number of electrons accumulated on the sapphire/ITO interface gradually decreases, and the depletion region on the sapphire/n-GaN interface gradually disappears, as illustrated in Fig. 1(e). When the negative half cycle voltage reaches its maximum, the number of holes accumulated on the sapphire/ITO interface reaches the peak value, and an equal number of electrons are induced on the sapphire/n-GaN interface [Fig. 1(f)]. Note that the above four processes can occur repeatedly under AC voltage. Therefore, even in the case of single contact, the electroluminescence from the LED epitaxial wafer can be observed.

In addition, we analyze the change of electron and hole concentration in a voltage cycle by using simulation. The simulation mode and details can be found from literature [16]. It should be noted that one voltage cycle is 10 μs. As shown in Fig. 2(a), from 0 to 2.5 μs the electron and hole concentrations in the MQWs region gradually increase under the applied electric field. In this case, a depletion region on the sapphire/n-GaN interface is formed because there is no electron supply from the external electrode. An induced field with opposite direction to the applied field is also formed. At 2.5 μs, the voltage reaches its maximum value, after which the electron and hole concentrations in the MQWs region gradually decrease, as shown in Fig. 2(b). It can be found that the electron concentration on the sapphire/n-GaN interface also gradually decreases after 2.5 μs, and finally returns to the initial level. The reason is that the electrons move toward the n-GaN region under the induced field. When the AC voltage switches to the negative half cycle [Fig. 2(c)], electrons continue to move toward the n-GaN region, leading to the increase of electron concentration on the sapphire/n-GaN interface. When the negative voltage reaches the maximum value (7.5 μs), the concentration of the aggregated electrons on the sapphire/n-GaN interface also reaches the maximum value. After 7.5 μs, the concentration of the aggregated electrons on the sapphire/n-GaN interface gradually decreases, and the device gradually returns to its initial state.

According to the LED equivalent circuit model, we further discuss its working mechanism. As shown in Fig. 3, the impedance of the entire circuit loop is as follows: Z=Rt+1jωCs+1jωCwafer×Rwafer1jωCwafer+Rwafer,(1)where Rt is the total resistance of the external circuit outside the LED epitaxial wafer (including probe resistance, external resistance, wire resistance, contact resistance, etc.), Cs is the equivalent capacitance of sapphire, Cwafer is the equivalent capacitance of the LED epitaxial wafer from the PN junction, and Rwafer is the resistance of the LED epitaxial wafer.

In the experiment, an external resistance (1 kΩ) is connected. Considering that the probe resistance, wire resistance, and contact resistance are relatively small, the value of Rt is considered about 1 kΩ.

According to the formulation of parallel plate capacitors, Cs=εrS4πkd,(2)where εr is relative dielectric constant of sapphire (∼9), S is the area of the sapphire layer (19.56cm2), k is the electrostatic force constant (8.9880×109Nm2/C2), and d is the thickness of the sapphire layer (430 μm). It can be calculated that the value of Cs is about 362.44 pF.

As for Cwafer and Rwafer, it is difficult to measure their accurate value. The reason is that the Cwafer changes with the electric field applied to the PN junction. Therefore, the value of Cwafer can change in a cycle of AC voltage. Similarly, the Rwafer also changes with the applied electric field. As is well known, the Rwafer under forward bias is much smaller than that under reverse bias. Due to the fact that the LED epitaxial wafer is working under SC operation mode, it is difficult to get the actual voltage applied to the PN junction, which makes it more difficult to measure the Cwafer and Rwafer.

The SC-EL inspection can obtain the EL wavelength of the LED epitaxial wafer, as shown in Fig. 4(a), which is similar to the conventional EL inspection. The used LED epitaxial wafer is a blue epitaxial LED with a central EL wavelength of about 443 nm. In addition, the SC-EL inspection can also obtain the AC current characteristics of the test area, as shown in Fig. 4(b). On the basis of the measured current, we can extract the electrical characteristics of the LED epitaxial wafer, which will be discussed in the following section. It should be noted that an AC voltage with 70Vpp and 100 kHz is applied in the experiment. As for the single-contact mode, there is an optimized driving frequency, at which the highest EL intensity can be obtained. In our experiment, we measured that 100 kHz is the optimized frequency. The reason to choose 70Vpp is that we hope the output EL in the single-contact mode is similar to that in the DC driving mode. This approach facilitates a more precise comparison of their respective emission wavelength.

As is well known, for the conventional EL inspection, it is important to make the metal probe well contact the p-GaN layer. Therefore, it is necessary to apply a sufficient pressure between the metal probe and the LED epitaxial wafer, which would result in mechanical scratches on the LED epitaxial wafer surface, as shown in Fig. 4(c). However, for our SC-EL inspection, the conductive rubber is selected as the electrical probe because it can reduce the LED epitaxial wafer mechanical damage such as scratches and stains during the inspection process, as shown in Fig. 4(d). This ensures that the quality of the LED epitaxial wafer would not be reduced before etching.

For LED epitaxial wafer inspection, obtaining the distribution of luminescent peak wavelength (PWL) is the key task. Figures 5(a)–5(c) show the PWL mappings obtained from SC-EL inspection, PL inspection, and conventional EL inspection, respectively. It can be found that the PWLs of LED epitaxial wafers obtained by the three inspection methods have the same distribution trend. That is, the region near the edge of the wafer has a shorter emission wavelength. The main reason for the uneven wavelength of the epitaxial film is substrate warping caused by stress [25].

During the growth of the epitaxial film, the stress generated by lattice mismatch interacts with the stress generated by thermal mismatch between the thin film and substrate, resulting in the substrate and thin film becoming concave or convex during the epitaxial growth process. Regardless of whether it is concave or convex, it will cause the distance between the center or edge of the substrate and the heating plate to be inconsistent, resulting in the differences in wavelength distribution at different regions of the LED epitaxial wafer [26–30].

As is well known, the conventional EL inspection has the highest accuracy because the LED epitaxial wafer is driven in the same way as the LED chips. We compare the accuracy of the SC-EL inspection and PL inspection based on the results of EL inspection. The error is defined as the PWL of SC-EL inspection (or PL inspection) minus the PWL of EL inspection in the corresponding region, as follows: Errori=Xi−Ei,(3)where Xi denotes the PWL of SC-EL inspection (or PL inspection) in a certain region, and Ei denotes the PWL of EL inspection in the corresponding region.

From the experimental results, it can be found that the error of SC-EL inspection [Fig. 5(d)] is significantly smaller than that of PL inspection [Fig. 5(e)]. Moreover, we can use the following equations to get the non-uniformity of central wavelength: Errorave=∑i=1114Errori/114,(4)σ=Errormax−ErrorminErrorave×100%,(5)where Errorave denotes average of the error between SC-EL inspection (or PL inspection) and EL inspection, and σ denotes the non-uniformity of PWL of SC-EL inspection (or PL inspection). It can be calculated that the σ of SC-EL inspection is 444.64%, and the σ of PL inspection is 504.78%. Therefore, the proposed SC-EL inspection can ensure higher inspection accuracy than PL inspection.

The better performance of SC-EL inspection is attributed to the different controls to electrons and holes that attribute to the luminescence process [31]. As for LED epitaxial wafers, the electron mobility is larger than that of holes, and the radiative recombination between electrons and holes mainly occurs in quantum wells close to the p-GaN, as schematically illustrated in Fig. 6(a). Therefore, the EL spectrum mainly reflects the properties of quantum wells close to the p-GaN. For PL inspection, the recombination between electrons and holes occurs in all quantum wells, as schematically illustrated in Fig. 6(b). As a result, the PWL of PL is a little different from the central wavelength of EL, leading to a decrease in the accuracy of PL inspection.

Although the accuracy of SC-EL inspection is higher than that of PL inspection, there is still a little discrepancy between the SC-EL inspection and the conventional EL inspection. We tend to attribute the observed discrepancies to the distinct quantum well behaviors governing electroluminescence under AC and DC driving modes. Traditionally, EL inspection is conducted under DC voltage, where electrons and holes are injected from both terminals of the LED and recombine within the multi-quantum well region. In this region, there exists a dynamic equilibrium of electron and hole concentrations. It is generally accepted that luminescence predominantly occurs in the quantum wells adjacent to the p-GaN region. However, in the SC-EL driving process, electrons and holes move toward the multi-quantum well region under an alternating electric field. Ensuring that the electron and hole concentration distributions match those in the DC-driven mode becomes challenging. As is well known, in the multi-quantum wells region, the emission wavelengths corresponding to different quantum wells vary slightly, resulting in discrepancies between the wavelengths obtained by conventional EL inspection and SC-EL inspection.

For LED epitaxial wafer inspection, obtaining the distribution of electrical characteristics, especially the reverse current, is also important. However, it is impossible for PL inspection to get the electrical characteristics. As for our SC-EL inspection, we can obtain the electrical characteristics of the LED epitaxial wafer by measuring the operation current. When AC voltage with a fixed amplitude and frequency [top panel in Fig. 7(a)] is applied during the SC-EL inspection, the related operation current can be obtained. The bottom panel of Fig. 7(a) presents the measured currents of different LED epitaxial wafer regions.

It can be found that the currents are changing with the applied voltage, and in different regions the current amplitudes are also different. In order to get the origin of the current difference during the SC-EL inspection, we measure the I-V characteristics by using the conventional EL method. As shown in Fig. 7(b), it can be found that when the reverse leakage current under DC bias is relatively larger, the current amplitude obtained by the SC-EL inspection is also higher. Therefore, the measured current obtained by our SC-EL inspection can reflect the electrical characteristics of the LED epitaxial wafer, which has an advantage over PL inspection. This phenomenon can be explained by Eq. (1). Due to the different manufacturing processes and defects, there may be differences in Rwafer across different regions. When the Rwafer is small, the reverse leakage current of LED epitaxial wafers under DC bias will be larger, and the corresponding sinusoidal current under AC drive will also be larger.

The characteristics of the existing LED epitaxial wafer inspection methods are summarized in Table 1. Meanwhile, a radar graph with a more intuitive description of the advantages and disadvantages of various inspection techniques is shown in Fig. 8. For the methods proposed in Refs. [11,32,33], etching and metal vapor deposition pre-treatment are necessary, resulting in damages to the LED epitaxial wafer. For the method proposed in Ref. [34], the localized epitaxial layers are also destroyed, and the diode-like I-V characteristics and emission spectra can be obtained at an injection current as high as 100 mA. For the method proposed in Refs. [35–37], only optical characteristics of the LED epitaxial wafer can be obtained. Our proposed method does not cause damage to the LED epitaxial wafer and does not require pre-treatment of the LED epitaxial wafer. In addition, the proposed method can simultaneously measure the electrical and optical characteristics of LED epitaxial wafers, providing higher accuracy and efficiency than other methods.Table 1.

Comparison of Existing Inspection Methods^a

In summary, we propose a non-destructive and accurate electroluminescence inspection method for LED epitaxial wafers, namely, SC-EL inspection. The SC-EL inspection uses a high frequency AC electric field to stimulate LED epitaxial wafers to generate electroluminescence. It is demonstrated that the SC-EL inspection method can obtain more accurate optical characteristics of LED epitaxial wafers than conventional EL inspection, and can also obtain electrical characteristics that cannot be detected by PL inspection. In addition, this SC-EL inspection method can reduce the mechanical damage caused by using conventional EL inspection, and avoid the pre-treatment that is necessary in conventional EL inspection. The proposed SC-EL inspection can ensure high inspection accuracy without causing damage to the LED epitaxial wafer, which holds promising application in LED technology.