A thermoelectric control system using thermoelectric cooler devices (TEC) combined with an aluminum heat dissipater and a fan heat extractor allows improving considerably the optical performance of deep UV LEDs (285 nm) operating at desired temperatures. A proportional, integral, and differential controller (PID) control technique was implemented to control the voltage in the TEC devices, and therefore, the desired range of temperatures can be achieved. The PID parameters are obtained with computational simulations based on physical models and experimental data recordings of the temperature, using a thermistor sensor for the temperature measurements and SiC photodiode with UV enhanced system for the optical power measurements. The experimental data show that decreasing the temperature of the UV-C LED light source using a TEC increases the optical output power, while it has been shown that the lifetime of the LED devices can be improved.
- thermoelectric cooler
- light-emitting diode
Manufactures of new solid state light sources have focused the attention to deep UV light sources (200–300 nm), where UV-C LEDs (or deep UV-LEDs) have shown great potential in many technological applications, such as water and surface disinfection processes, biomedical instrumentation systems, high density optical recording, lithographic micro fabrication, and biophotochemical research [1, 2, 3, 4]. The complexity of a reliable UV-C LED performance depends upon optical efficiency, adequate lifetime, and wavelength accuracy that all concern technological applications. It is known that the UV-C LEDs have lower external quantum efficiency (EQE) with a value less than 10%, compared with conventional LED light sources, for example, blue LED (400–450 nm) has an EQE > 70%. In microbiological research, the shift of the peak wavelengths would result in an inaccurate interpretation in microbial studies .
2. Efficiency droop and optical performance of UV-LEDs
The optical performance of the UV-LED devices has been studied by Kheyrandish et al. showing that higher solder point temperatures (TS) resulted in lower optical output power, shift of the peak wavelength, and an increase of the full width at half maximum (FWHM) . The achieved internal quantum efficiency (IQE) of UV-C LEDs can be quite high, that is, somewhere between 15 and 70%, supported by photoluminescence studies and ray-tracing calculations. However, the external quantum efficiency and the radiant efficiency of UV-C LEDs are below 2% (highest for 280 nm LEDs) . Nevertheless due to increasing development efforts, the UV-C LED devices have much lower output power from single chip than the widely used blue LEDs.
The optical performance of a UV-C LED device decreases, due to a variety of key factors such as quantum-confined Stark effect, carrier delocalization, and others discussed by Versellesi et al. . It is possible to make a classification of droop mechanisms in two levels. The first classification level reflects nonradiative losses taking place inside the optical active region and the second level reflects nonradiative losses taking place outside the optical active region.
Until recently, the principal cause of efficiency droop was unclear in AlGaN LEDs (from blue to UV-C sources). The Auger effect has been proposed as the main cause of efficiency droop in InGaN LEDs. Thanks to Iveland’s contribution with direct measurement of Auger electrons emitted from LEDs under electrical injection of high currents, it was possible to conclude that the efficiency droop phenomenon finalized a long-standing controversy .
2.1 Thermal stress and lifetime performance
The excessive self-heating of the UV-C LED reduces the lifetime of the devices, the optical performance, and yield spectral shift of the emission. Lifetime limiting issues of UV-C LED occur with high thermal stress in the chip, resulting from the higher electrical current (100–350 mA). The device lifetime may be difficult to assess, in a rigorous fashion. Nevertheless, one recent published result has shown that the lifetime decreases more with a current of 350 mA than with 100 mA in UV-C LEDs by Nikkiso Co, LTD . The thermal extraction is one of the challenges to improve the performance of the LED devices . To resolve this challenge, the thermal conductivity of the package must be considered. The relation between thermal conductivity and thermal resistance is known, and therefore, thermal resistance provided by the manufacturer will be considered.
2.2 LED package and junction temperature
Various indirect techniques and models have been reported for junction temperature TJ measurement techniques including micro Raman spectroscopy, threshold voltage, thermal resistance, photo thermal reflectance microscopy (PRM), electroluminescence, and photoluminescence . In this work, the junction temperature TJ has been obtained applying a thermal resistance technique and direct measurements of the solder point temperatures TS . The UV-C LED manufacturer allows to compute the optical efficiency of our 285 nm UV-C LED model: VPS171 by Nikkiso Co. The maximal operational solder temperature value can be determined using the maximal junction temperature provided by the manufacturer. The solder point temperature TS and the junction temperature TJ are linearly related, as shown in (1), where ILED is the electric LED current and VLED is a forward LED voltage. Wall plug efficiency (WPE) will be defined in the next section.
3. Wall plug efficiency and temperature dynamics
The wall plug efficiency (WPE) is defined as the ratio of electrical power consumption PLED and optical output power of the UV-C LED device denoted by PLIGHT. Kneissl related the WPE with external quantum efficiency (EQE) and the rate between optical and electrical energy  as shown in the next equation.
where h is the Planck’s constant, ν is the frequency of the UV light, e is the elementary electric charge, V is the voltage, and η EQE is the external quantum efficiency factor.
The ratio of the number of photons emitted from the LED per second to the number of carriers (electrons) injected into the device per second is precisely EQE. The data sheet of our 285 nm UV C LED showed that at 350 mA and 6 V, the UV-C LED can provide an optical output power of 30 mW. This implies that the WPE is 1.43% and EQE is 1.97%, that is, that 1.97% of the injected electrons is converted to UV-C light. Eq. (2) shows that the optical output power PLIGHT is related with PLED and the temperature, and its relation with EQE is shown in Eq. (3)
Eq. (3) allows to relate the EQE with the junction temperature. Since the junction temperature can be controlled, the EQE factor (η EQE) can be improved.
Joule’s first law shows the relation between the heat generated by an electric current passing through a conductor. Eq. (4) shows that the electrical power is converted to light and heat.
The light power PLIHT can be determined using the electrical efficiency factor WPE, obtained with the parameters of the UV-C LED provided by the manufacturer as a linear relation, PLIGHT = WPE PLED.
As shown in Figure 1, PHEAT is divided into two heat transfer rates. P1 is the heat transfer rate between the junction surface and the ambient environment, and P2 is the heat transfer rate between the junction surface and the cooling system. For operations at ambient temperature, P1 is negligible compared to P2.
3.1 Newton’s law of cooling
The thermal performance of the UV-LED device shows that the solder temperature dynamics TS (t) can be modeled with Newton’s law of cooling, where TEST represents the temperature in the steady state.
Taking into account that the cooling process begins at ambient temperature TAMB, commonly in the range of 19–25°C, to finally achieve the steady state of the temperature TEST. The analytical solution for the solder temperature is
Commonly, the inverse of the constant K is referred to as time constant of the system. The determination of the constant K of Newton’s law of cooling is important and has to be determined experimentally for each LED device (including thermoelectric cooler system). The K value depends on the predefined stationary temperature value TEST. The value of the cooling temperature constant K will be able to model the thermal performance of the system. Therefore, the UV-C LED device temperature can be controlled applying a corresponding strategy.
4. Experimental setup
Our setup is composed by three essential parts: the UV-C LED device, the optical sensor (SiC photodiode model: SG01D-C18, from SGLux), and the temperature sensor (calibrated thermistor). The UV-C LED system is mounted on an optical table (Figure 2B). The light output is directed toward an optical sensor at a predefined distance, and finally, a temperature sensor is mounted at the solder point. Experimental data are acquired using a computational interface.
The thermoelectric cooler cell (TEC) is composed of a large number of PN and NP semiconductor arrays. TEC modules can convert voltages (VTEC) into a temperature gradient, based on a phenomena discovered by Peltier in 1834. In this study, a TEC model number TEC1-12706 by Hebei I.T. Co. Ltd. was used.
The experimental setup for the measurements of solder temperature dynamics TS consists of a 10 kΩ thermistor model NTC Murata, which is directly put in contact on the solder point. The temperatures are measured at the edge of the LED package shown in the right-hand side of Figure 1. The thermistor is powered by a stabilized voltage source device, model: LM2596 DC-DC. The signal of the thermistor is connected to an A/D converter to acquire and record (every second) the temperature data.
The accuracuty of our thermistor is 1%, specified by the manufacturer, our operation range is from −40 to 60°C, therefore the maximum absolute error is ε1 = 1°C. The computational interface error is ε2 = 0.5°C. It is possible to estimate the standard error of the mean (SEM) for the temperature measurements, using the standard deviation of 20 repetitions of the data recorded for a fixed value of temperature, monitored by a thermal camera model FLIR E 63900 with an emissivity setting of 0.95. The standard deviation was S = 0.73°C, and the total error for the temperature recording can be estimated, that is, ε3 = 0.16°C. Finally, to complete the total error of the solder point temperature measurements TS, we sum all previously mentioned errors Δ = ε1 + ε2 + ε3 = 1.66°C. Previous work shows that the standard deviation between the modeled and true junction temperatures for blue LEDs was only 3.5°K .
5. Thermal performance and empirical modeling
5.1 Thermal resistance
The thermal partial resistances correspond to the various functional groups in the system or heat path and characterize their individual thermal behavior (Figure 3). The total thermal resistance R of the system is then equal to the sum of the individual thermal resistance components.
RJ-S is the thermal resistance of the UV-C LED chip and describes the transfer of heat within the LED housing (junction-to-solder-point), RS-B is the thermal resistance of the solder point to the metal core printed circuit board (MCPCB), RB-TEC is the thermal resistance of the MCPC and aluminum plate to the thermoelectric cooler device (TEC), RTEC-Al is the thermal resistance of the TEC to the aluminum heat dissipater, and finally RAL-AMB is the thermal resistance of the aluminum heat dissipater to the ambient environment. The thermal resistance between board and TEC (RB-TEC) is not strictly a thermal resistance, but it is technically a heat engine, which can be considered a peculiar thermal resistance with negative value . It is clear that this is an advantage of the TEC, since it can be observed in (7), that the negative value of RB-TEC reduces the total thermal resistance of the device. As a general rule, the smaller the thermal resistance of a LED, the faster the heat can be dissipated from the housing.
5.2 Thermal performance
To obtain the stationary temperature of the system TEST and its relationship with TEC voltage, we apply a pre-cooling turning on of the LED. First, the TEC is turned on with a predefined constant voltage value. Once the stationary temperature is reached, the UV-C LED source is turned on. The system will then reach its final stationary temperature.
TEST is related with VLED and TAMB. It is possible to achieve the stationary temperatures in the system making the cooling process with the TEC before turning on the LED. It is known that the relation between TEC voltage and gradient temperature is linear. The linear relation is described in the following equation.
If TAMB is constant, it is possible to obtain the value of slope m, relating the empirical recording of the cooler process dynamics, for different values of VTEC. The value of m is a thermal property of each system, in a specific range of temperatures. For example, the data displayed in Figure 3, allow to confirm the linear relationship stipulated in Eq. (8), that is, TEST = 19.5–1.59VTEC. In the six curves shown in Figure 3, VLED was maintained constant.
6. Temperature control using TEC and PID controller
The control technique implemented in this application has been developed applying the previously described physical model, the results of the simulations, and the recorded experimental temperature data.
According to Eq. (1), the solder point temperature can be obtained knowing the junction temperature. We can define a reference junction temperature TJ-REF and obtain a stationary reference solder point temperature TEST-REF. The controller operates comparing TEST-REF to TS (t) to obtain an error signal that is fed to the PID controller. The PID controller applies VTEC to the system to yield TEST-REF.
The block diagram displayed in Figure 4 is implemented with the MATLAB Simulink software, considering the empirical values of the time constant, the ambient temperature TAMB, the increment ∆T, the linear relation between VTEC and TEST shown in Eq. (8), and the set values for the junction temperatures TJ-REF.
A computational tuning based on a simulated plant is used to obtain the PID parameters, which were then applied and verified experimentally. For each steady-state temperature TEST, the PID parameters can be obtained in the simulations to be consequently applied in the controller (Figure 5).
Using these gain parameters for the PID controller applied to our system, we can fit the simulations of the temperature dynamics for three different steady-state operations. When the UV-C LED is turned on, the control of the system increases the TEC voltages to recover the starting temperature states as displayed in Figure 6.
One of the major advantages of the temperature control is the improvement of the optical output power. This result is displayed in Figure 7. The relation between junction temperature and UV intensity shows that as the temperature drops, the UV intensity increases. The plotted data demonstrate that reduction of the junction temperature TJ from 59 to 44°C increases the UV irradiation by 6%.
The result shows that the TEC can effectively reduce the thermal resistance of the high power UV-C LED device and therefore substantially increase the optical performance. In addition, the lifetime might be highly improved . Since the system temperature affects the optical performance, it is essential to control it. The TEC device makes it possible to implement an appropriate control strategy. Each LED device has different thermal performance, depending on many factors, for example, thermal conductivity of the heat dissipaters, the fan extractor velocity, and operation ambient temperature. At lower temperature operations, optical performance is improved. The operation of the UV-C LED at lower temperatures implies that the intrinsic lifetime could be substantially improved. The implementation of the TEC device for the thermal control of the UV-C LED light source allows an efficient optimization of the optical performance (Figure 8).
Portions of this work were presented at the OSA Light, Energy and Environment Congress in November 2017, titled “Improving the Optical Performance and Life Time of Deep UV LEDs by Thermoelectric Control”. This work was supported in part by for the Hydraluvx Spa, Santiago, Chile and Mrs. Andrea Bravo, Hydraluvx’s Graphics Designer, Santiago Chile is gratefully acknowledged for the rendering of Figures 2 and 8.