T_j may be indirectly measured using an LED’s inherent optical characteristics. The emission spectrum of a semiconductor device is influenced by temperature variations due to the temperature dependence of the energy band gap [60]. This behavior motivates researchers to use spectral power distribution (SPD) characteristics such as the peak wavelength and spectral bandwidth of LEDs, which are known as temperature sensitive optical parameters (TSOPs), to estimate T_j [61]. The TSOP measurement method is non-destructive and does not interrupt the electrical performance of LEDs [62,63,64,65,66] (e.g., in case of an alternating current (AC) LED [67], TSOP-based T_j measurements have shown successful implementation without altering their electrical performance [68]).

The TSOPs are unique for each electrical working condition, and light output changes depending on the internal quantum efficiency and band gap characteristics of an LED at each specific temperature and input current [69]. A summarized illustration of the SPD response to temperature and input current induced changes of an arbitrary white LED (WLED) is shown in Figure 2.

It is known that with increasing T_j, SPD shows a red shift and broadening [70]. The redshift phenomenon is primarily due to the band gap reduction. This was explained by Wang et al. [71,72] for GaN-based blue LEDs in low temperatures, and recently, similar results were reported at high temperatures for high brightness GaN on sapphire blue LEDs [73]. It should be noted that SPD displays a blue shift and broadening with increasing input current. Li et al. [74] studied the effect of input current and temperature on the spectral behavior of green InGaN/GaN multi-quantum well LED and showed that the excitation source could alter the carrier dynamics in the active region. A large blue shift was observed in high input power levels, mainly due to the carrier screening effect as a result of a weakened piezoelectric field that causes the quantum-confined Stark effect [75]. This issue can be verified by the findings of Kim et al. [76], which analyzed carrier leakage of GaN based on photoluminescence properties of LEDs both at forward biased and intentionally formed leakage path conditions. Increased current leakages were observed in low series resistance for LEDs, which led to a blue shift of the SPD.

Temperature-induced full width at half maximum (FWHM) broadening is due to the thermal broadening [77]. On the other hand, current-induced FWHM broadening is due to the combined effect of the screening to the piezoelectric field and band filling effect as discussed by Lin et al. [70]. With input currents of 150 to 850 mA for GaN-based blue LED at temperature range of 273 to 338 K, the slope of the center of mass wavelength per T_j was below 0.034 nm/K, while for FWHM, broadening was above 0.052 nm/K. Assuming high precision temperature measurements by taking precise optical measurements within 0.1 nm, the accuracy of the T_j measurement based on wavelength shift was around 3 K, while the FWHM measurement could reach below 2 K. Based on this, we suggest that FWHM calibration can yield high accuracy in T_j evaluation.

It should be noted that the spectral width and shape of the LED emission can also be associated with carrier distribution, growth procedure and structure, the density of states, and the successful pairing of electrons and holes, etc. [78]. For instance, an inconsistent temperature-dependent shifting behavior in wavelength (red–blue–red shifting) was observed for the peak energy of an InGaN based MQW LED and mainly attributed to the non-uniformity (variation in layer thickness and defects) and carrier localization in quantum wells [79]. Considering these issues, the accuracy of the wavelength shift method is stated to be only 5–10% of the FWHM of the emission line [80]. However, in devices with a narrow emission, the accuracy of the wavelength shift method increases and can be a suitable approach for T_j prediction [81].

Chhajed et al. [77] calibrated the peak wavelength of ultraviolet (UV), blue, green, and red GaN-based LEDs for a forward current range of 10 to 100 mA at temperatures of 22 to 120 °C. Similarly, a high temperature coefficient of the spectral widths was determined in comparison to the wavelength shift. Temperature coefficients of the dominant wavelength for the blue, green, and red LEDs were determined as 0.0389, 0.0308, and 0.1562 nm/K, respectively. The slopes increase to 0.0466, 0.0625, and 0.1812 nm/K for the FWHM-based T_j evaluations. Similarly, a strong red shift of the red AlGaInP LED was also seen in another study by the authors for a trichromatic white LED system [82]. Blueshift was negligible for the red LED and slightly higher for the UV LED, while the highest blueshift effect was recorded for the green LED. The authors stated that the uncertainty of T_j estimation based on the wavelength shift method is higher in comparison to the FWHM method. Furthermore, the authors preferred the forward voltage method (FVM) with a reported accuracy of ±3 °C over the TSOP method.

Chen et al. [83] conducted T_j estimation experiments and obtained peak wavelength shifts for three different AlGaInP LED arrays. Shifting characteristics investigated at longer, central, and shorter wavelengths showed that the center wavelength is the most suitable method to calculate T_j of an LED array. Tamura et al. [84] analyzed the wavelength shift of InGaN-based white LEDs at temperatures from 20 to 160 °C and their investigations showed that blue light emission from the active layer and yellow light emission from Ce:YAG phosphor formed two different electroluminescence bands while each band displayed a distinct behavior with the temperature change. However, similar to previous cases, T_j was stated to be successfully calibrated to the blue emission of the chip. Chen et al. [85] showed a simplified peak wavelength shift variation at different T_j for white LEDs under different drive currents. Their findings showed that the temperature dependence of peak wavelength is lower for direct current (DC) LEDs compared to bilevel drive, while the thermal energy needed for correlated color temperature (CCT) stabilization is also less for a DC LED. Gu et al. [86] selected the point of interest as the lowest energy in the SPD between the peaks of blue and yellow emissions. The ratio of the total radiant energy of white LEDs to the radiant energy within the blue emission in a different T_j has shown a linear relationship. The authors claimed that with a ratio of 0.005, the temperature prediction accuracy of 1 K can be achieved in commercial white LEDs with this relation. In constant forward current density, Azarifar et al. [87] performed four machine learning regressions including k-nearest neighbor (KNN), radius near neighbors (RNN), random forest (RF), and extreme gradient booster (XGB) on temperature sensitive optical data from over 500 commercial white LED packages and tested the accuracy of prediction with experimental measurements. With near unity in R² scores and small root mean square deviation values, the XGB regressor showed close-to-perfect correlation capability to assess T_j based on SPD behavior. Their recent findings demonstrated that white LED brightness and color characteristics, irrespective of the package’s thermal resistance, can offer a real-time T_j prediction capability.

TSOP-based methods have also been shown to be a practical approach for measuring phosphor temperature in an operating white LED. Based on the total emission division of a white LED to a sum of the spectrum of the blue chip and two spectrums from phosphor with a short and a long wavelength band, Yang et al. [88] examined the fitting peak wavelengths and FWHMs of the short and long wavelength bands at the different phosphor temperatures. They stated that phosphor temperature can be precisely measured by checking the variations of its related emission spectrum. Similar to the LEDs, redshift at higher temperatures was observed. Linear relationships were seen for the FWHM and peak wavelength of the phosphor at different temperatures. However, this model is practical for the same phosphor only, and it cannot be used for mixtures of different phosphor combinations.

Recently, TSOP-based temperature measurement methods have extended to two-dimensional (2D) thermal mapping. Based on new microscopic hyperspectral imaging (MHI), the 2D spectral power distribution can be obtained from light emitting surfaces and can be used to incorporate the TSOP for surface temperature measurement of the LEDs. Jin et al. [89] used the MHI-based centroid wavelength method to study the 2D temperature distribution of the blue, green, and red LEDs. After calibration of the centroid wavelength coefficient, the authors reached as low as 3 μm resolution for the surface temperature measurement of the LEDs, which is claimed to be capable of reaching submicron level accuracy.

As the next generation micro-LEDs are gaining popularity [90], practices for performance characterization of micro-LEDs are also becoming a topic of interest. Although previous discussions suggest that heat accumulation in micro-LEDs is lower due to enhanced current spreading in active layers with size reduction [91], the studies of Feng et al. [92] and Yu et al. [93] emphasized the importance of T_j for micro-LEDs by stating that the internal series resistance of micro-LEDs is expected to increase with their decreasing size. Therefore, it is important to extend the practical application of TSOP-based temperature measurement to the next generation micro-LED devices. Recently, a thermal study on GaN-based micro-LEDs was conducted by Feng et al. [92]. In their study, a band gap reduction-related redshift of 0.024 eV and FWHM broadening from 27 to 35 nm were seen in the temperature range of 25 to 80 °C. Although they proposed that micro-LED band gap behavior with T_j follows a semi-empirical Varshni relation, it is not possible to determine an accurate coefficient of temperature for the FWHM or band gap for T_j calibration from the presented experimental results.

A parametric summary of TSOP-based experimental LED T_j measurement reports is presented in Table 1.