The junction temperature, one of the major parameters that strongly affect the performance of light-emitting diodes (LEDs), increases during operation because of the power dissipated as heat within an LED device. Therefore, LED devices with poor characteristics are expected to have higher junction temperatures for the same driving conditions. In this study, an observation contrary to this expectation is presented: a deep-ultraviolet LED device with superior electrical characteristics shows a higher junction temperature at the same input electrical power than a device with poor characteristics. A simple equivalent circuit comprising a diode, a series resistor, and shunt components is employed to elucidate this counter-intuitive observation by considering the possible heat sources inside the LED device. It is found that the junction temperature is mainly dominated by the power dissipated at the diode instead of the other possible heat sources including the Joule heating effect of the resistive components.
Temperature is one of the most important parameters determining the performance of semiconductor devices, including III-nitride-based light-emitting diodes (LEDs). In general, the light output power (LOP) of an LED decreases as the temperature increases because of the stimulated Shockley–Read–Hall recombination,1 also called the thermal droop.2–4 In addition, the emission wavelength is red-shifted by bandgap energy reduction,5 and the device reliability6,7 and lifetime are deteriorated via encapsulant degradation at elevated temperatures.8,9 Therefore, thermal management has been a major part of reliable LED technologies; it has become increasingly important with the development of high-power LEDs and deep-ultraviolet (DUV) LEDs, which consume greater electrical power for device operation than conventional blue LEDs do.
The temperature of an operating LED is represented by the junction temperature which indicates the actual highest temperature in the device. During operation, the junction temperature increases relative to the ambient temperature because of the power dissipated as heat in the device. In this regard, the junction temperature of an LED device is generally expressed by
where Tj is the junction temperature, TA is the ambient temperature, θJA is the junction-to-ambient thermal resistance, and Pd is the total power dissipation of the device, defined as the difference between the input electrical power (Pin) and the LOP. The equation predicts a higher junction temperature for a device with worse characteristics, which agrees well with our intuition.
In this paper, we study the junction temperatures of AlGaN-based DUV LEDs with different characteristics and present an unexpected junction temperature behavior, wherein a higher junction temperature is observed for the DUV LED device with better electrical properties under the same electrical input power. To understand this behavior, heat sources inside the LED device are studied based on a simple equivalent circuit comprising a diode, a series resistor, and shunt components. Among the possible heat sources, it is found that the power loss at the diode component in the equivalent circuit dominates the junction temperature, but the Joule heating effect of the resistive components does not significantly affect the junction temperature. Consequently, the junction temperature is strongly related to the device parameters of external quantum efficiency, injection efficiency, junction voltage, and injection current.
Two types of AlGaN-based DUV LEDs were grown by metal–organic chemical vapor deposition (MOCVD) on a 4-in. c-plane sapphire substrate; they were labeled DUV LED I and DUV LED II in this study. Both DUV LEDs have epitaxial structures; each consists of an n-type Al0.55Ga0.45N cladding layer, Al0.4Ga0.6N/Al0.55Ga0.45N multiple quantum wells (MQWs), an electron blocking layer, a p-type cladding layer, and a highly p-doped GaN p-type contact formation layer. They are differentiated by the Si doping profiles in the n-type cladding layers. Graded Si doping was applied only to the n-type cladding layer of DUV LED I to improve its electrical properties including the contact resistance and sheet resistivity.
DUV LEDs of 1 × 1 mm2 in size were fabricated by conventional processes including mesa etching and n- and p-type contact formation based on photolithographic patterning and lift-off. Each chip was diced and transferred onto a sub-mount with a flip-chip configuration. A fabricated device is shown in the inset of Fig. 1(a).
The electrical properties of the DUV LEDs were measured by using a parameter analyzer (Keithley 4200), and their emission spectra were obtained using a spectrometer (Black C-50, StellarNet, Inc.) equipped with a UV-enhanced optical fiber (F1000-UV-VIS-SR). The junction temperatures of the DUV LEDs were acquired by observing the changes in the operation voltage depending on the temperature. This method is known to provide accurate junction temperatures compared to other methods, including the peak emission wavelength method.10 The operation voltage was measured by the pulsed-source injection mode with a 0.2% duty cycle for various ambient temperatures from room temperature (RT) to 200 °C, and for various injection currents from 10 mA to 100 mA with a 10 mA step.
The current–voltage (I–V) characteristics of both DUV LEDs at RT are shown in Fig. 1(a). Heating effects are excluded by employing a pulsed voltage sweep scheme with a 0.2% duty cycle. The operation voltages at 100 mA are 8.34 V and 12.9 V for DUV LED I and II, respectively. The improved I–V characteristic of DUV LED I is mainly due to the graded Si doping scheme adopted in the n-type Al0.55Ga0.45N cladding layer.
Figure 1(b) shows the normalized emission spectra of DUV LEDs measured at RT with the injection current of 50 mA. Both LEDs show similar peak emission wavelengths of 287 nm and 285 nm for DUV LED I and II, respectively. This small difference in the peak emission wavelength may originate from the different doping profiles in the n-type cladding layers, which affect not only the strain but also the growth behavior of the subsequent epitaxial layers, including the MQWs.11 The full-width-at-half-maximum (FWHM) of the DUV emission peak is approximately 10 nm for both LEDs. A broad parasitic emission peak at approximately 325 nm may arise from deep-level states, such as a deep-level acceptor-like state, in the p-type Al0.55Ga0.45N cladding layer.12 The LOPs of the DUV LED are approximately 0.8 mW and 0.2 mW at an injection current of 100 mA for DUV LED I and II, respectively.
The temperature-dependent I–V characteristics of DUV LED I and II are shown in Figs. 2(a) and 2(b), respectively. The ambient temperature is regarded as the junction temperature under the pulsed-source injection with a 0.2% duty cycle; this allows enough time to completely dissipate the heat generated during biasing. As the temperature increases, the threshold voltage and operation voltage at a given injection current are decreased owing to the decrease in the bandgap energy and series resistance.13
The temperature- and injection current-dependent operation voltages are summarized in Figs. 2(c) and 2(d) for DUV LED I and II, respectively. The operation voltage is the average of the repeatedly measured voltages at a given injection current under the pulsed mode. The injection current is varied from 10 mA to 100 mA with a 10 mA step. This yields a guide map for obtaining the junction temperatures of the DUV LEDs through their operation voltages under actual DC-driven conditions.
Figure 3(a) shows the junction temperature as a function of the input power. The dissipated power in Eq. (1) can also be expressed as a function of the input power with the consideration of a series resistor by13
where η is the external quantum efficiency, IF is the injection current, and Rs is the series resistance of the device. When η of the LED device is much smaller than unity, the input power can be assumed equal to the total dissipated power in a device. Hence, the slope of Fig. 3(a) indicates the thermal resistance according to Eq. (1), which is determined by the thermal conductivities of the materials and the device architecture. Because both LEDs are equal in material systems and device structures, their similar thermal resistances (57.0 °C/W for DUV LED I and 58.4 °C/W for DUV LED II) coincide with expectations.
Interestingly, the junction temperature of DUV LED I is higher than that of DUV LED II for the same input power. As discussed in the previous paragraph, the input power is assumed equal to the total dissipated power in the device. Therefore, the junction temperatures of the two DUV LEDs should be similar for equal input powers because the ambient temperature is maintained at approximately 21 °C for all measurements. This observation indicates that Eqs. (1) and (2) are not sufficient to understand the junction temperature behavior. One possible approach for elucidating the observed discrepancy between the expected and actual results is to consider power dissipation that does not increase the junction temperature of the device.
Figure 3(b) shows the current-dependent junction temperatures of the DUV LEDs. DUV LED I shows a higher junction temperature up to an injection current of approximately 55 mA, even though less electrical power is applied to it than to DUV LED II at equal injection currents. Meanwhile, the junction temperature increases more rapidly with increasing current in DUV LED II (0.733 °C/mA) than in DUV LED I (0.465 °C/mA); DUV LED II, thus, reaches a higher temperature than DUV LED I with an applied current exceeding 55 mA. This is due to the increase in the input power with increasing current, which is larger in DUV LED II, as shown in the supplementary material. However, the higher junction temperature of DUV LED I in the low-current region cannot be understood via conventional approaches. Therefore, it is necessary to develop a more accurate understanding of power dissipation behaviors inside the LEDs and their influences on the junction temperature.
In order to understand the observations, the power dissipation in an LED device was directly calculated by analyzing the electrical properties based on a theoretical equivalent circuit, which comprises a diode, a series resistor, and a shunt resistor. Figure 4(a) shows this equivalent circuit, while Fig. 4(b) shows the measured and calculated I–V characteristics of DUV LED I in both linear and semi-logarithmic scales. For DUV LED II, an additional shunt component is simultaneously considered [see Fig. 4(c)] to reproduce its forward leakage. The additional shunt current has a power relation to the applied bias, implying that the current has a nature of space-charge limited conduction with trap states.14–16 With the addition of the shunt component, the I–V characteristics are well-reproduced, as shown in Fig. 4(d). The device parameters applied for I–V reproduction are summarized in the supplementary material, S2. Based on the equivalent circuits employed for I–V analysis, the total power dissipation in an LED device can be expressed as follows:
where Id is the current through the diode component, Vj is the junction voltage, and Ish is the shunt current. In DUV LED II, Ish is equal to the sum of Ish,1 and Ish,2, corresponding to the currents through the parallel resistor and additional shunt component, respectively, as shown in Fig. 4(c). On the right-hand side of Eq. (3), the first term indicates the power dissipated at the diode component due to non-radiative recombination, while the other terms are mainly related to the Joule heating effect of undesired resistive components in a device.
The heat generation (power dissipation) of each component in the equivalent circuit can be calculated using Eq. (3) based on the device parameters that accurately reproduce the I–V characteristics of the LED device. The calculated heat is provided in the supplementary material, S3, as a function of the input power and injection current. The junction temperature behavior is mainly dominated by the heat generated by the diode component (Pd,diode), as shown in Fig. 5 and the supplementary material. In this calculation, 1 − η is assumed to be unity because the EQEs of the DUV LEDs are estimated to be 0.185% and 0.046% for DUV LED I and DUV LED II, respectively, at 100 mA (10 A/cm2) based on the measured LOP (see the supplementary material, S3). First, the power dissipated at the junction of DUV LED I is higher than that of DUV LED II for the same input power, as shown in Fig. 5(a), which is consistent with Fig. 3(a). In addition, the current-dependent Pd,diode in Fig. 5(b) shows a trend identical to that of the current-dependent junction temperature in Fig. 3(b). Pd,diode of DUV LED I is greater than that of DUV LED II up to the injection current of 55 mA; at higher injection currents, it is lesser. On the other hand, the Joule heating effect does not seem to effectively increase the junction temperature, contrary to the general assumption. This is in line with a previous report claiming that the exact relation between the power conversion efficiency and junction temperature of a green LED can be calculated by excluding the power loss in the series resistor component.17
The observed junction temperature behavior can be elucidated by determining Pd,diode [the first term on the right-hand side of Eq. (3)] of the device, which can be simplified as follows with the assumption of 1 − η ≈ 1,
where IE is the injection efficiency defined by the ratio between the diode current (Id) and the injection current, Id/IF. For the same injection current, Vj is higher for DUV LED II owing to its larger ideality factor (14.0 for DUV LED I and 20.5 for DUV LED II, as listed in Table S1 in the supplementary material). The lower junction temperature of DUV LED II for the low-current regime consequently indicates worse IE in this device. Indeed, DUV emission was not observed in our measurement setup for DUV LED II up to an injection current of 10 mA, which supports its limited IE. This shows that an LED device with superior electrical properties can have a higher junction temperature by an effective suppression of the shunt current, thus enhancing the current passing through the diode component. The junction temperature of DUV LED II increases faster than that of DUV LED I as the current increases, which is due to the fact that the IE of DUV LED I approaches unity even under very low injection currents, while the IE of DUV LED II is gradually increased with increasing current injection as depicted in the supplementary material, S2. In addition, Vj of DUV LED II is larger in all current regimes and increases faster with increasing current than that of DUV LED I.
The model could be valid for all sorts of semiconductor emitters. The conventional model expressed in Eq. (1) is a special case of the proposed model when there is no series resistance (Rs = 0, thus, VF = Vj, where VF is the operation voltage) and no shunt components (Ish = 0, thus, IE = 1 and IF = Id). Therefore, the conventional model can be applied to the LED devices with effectively suppressed resistive and shunt components. For example, commercialized GaInN blue LEDs is the case. However, in the case of LED devices that are severely influenced by resistors or shunt components, such as high power LEDs and micro-LEDs, their junction temperature behavior will not be explained readily by the conventional model, but can be explained by the proposed model. In addition, the model is not material-dependent, so it can be applicable to the other material-based LEDs including emerging halide perovskite LEDs.
An operation voltage under DC-driven conditions is in a steady-state, which is guaranteed by allowing enough time before it is measured. Therefore, the measured junction temperature is also in the steady-state. On the other hand, the device parameters used in the calculations were obtained at room temperature, neglecting their temperature dependence. The device parameters including the ideality factor, saturation current, and series resistance will continue to change until the actual device temperature reaches steady-state. A more accurate analysis can be possible by considering temperature-dependent device parameters.
In conclusion, we measured the junction temperatures of two DUV LEDs with different electrical characteristics and observed a higher junction temperature in the higher-quality LED device than in the lower-quality LED device. According to the I–V analyses based on equivalent circuits, this counter-intuitive behavior is explained by considering the power dissipated at the diode component and excluding other power losses, such as the Joule heating effect of resistor components.
In this regard, the junction temperature is mainly determined by the external quantum efficiency, injection efficiency, junction voltage, and injection current. The higher junction temperature observed in the higher-quality DUV LED device arises from its high IE, which provides more chances to dissipate power at the diode component. On the other hand, the fast increase in the junction temperature of the worse LED device is attributed to its high Vj, which is caused by the large ideality factor of the device as well as the increase in IE with increasing current.
See the supplementary material for the relationship between the current and the input power of DUV LEDs, I–V analysis based on the equivalent circuits, and the power dissipation by the various heat sources in DUV LEDs.
This work was partially supported by the basic science research program through the National Research Foundation (NRF) of Korea funded by the Ministry of Education (Grant No. 2017R1D1A1A09000684).
The data that support the findings of this study are available from the corresponding authors and Dr. D. Y. Kim upon reasonable request.