High-resolution mirror temperature mapping in GaN-based diode lasers by thermoreflectance spectroscopy

Nitride based, wide-bandgap semiconductors attract great interest as materials for light emitting devices in the blue to ultraviolet wavelength region.^1–3) The first demonstration of emission in GaN-based laser diode was reported in 1996 by Nakamura.^4,5) From this time GaN based laser diodes have undergone a rapid development, however, still their maximum optical power, lifetime and reliability are strongly limited by degradation processes induced mainly by increase of the temperature in the particular parts of the laser: facet, active layers or contacts.^6–8)

To improve laser reliability and performance, it is necessary to understand degradation mechanisms induced or enhanced by temperature increase. The first symptom of laser diodes degradation occurs at their facets. Mirror degradation is a phenomenon well known in case of arsenide devices.⁹⁾ We can roughly divide it into two classes: first one is a catastrophic optical mirror damage (COMD)^10–12) the second one is the formation of carbon deposits on the surface of the output mirror.^7,10) COMD is a process, which manifests itself through the sudden increase of the mirror's temperature leading to melting of the facets. Increase of the temperature is caused by high optical power density and nonradiative carrier recombination, which creates surface depletion region and results in reabsorption of emitted radiation in the near-facet area of a device. A positive feedback loop between the increase of device temperature and the increase of the optical absorption leads to device damage. The threshold of COMD was determined to be around 40–57 MW/cm^{2 13,14)} for InGaN laser diodes. These values are an order of magnitude higher than for GaAs counterparts, thus showing the potential of nitride lasers for high optical power emission. However, a mechanism of COMD is far from being well understood in the nitride system.¹⁵⁾

Investigation of the temperature distribution on the facet of the device, with high spatial and temperature resolution, is crucial to gain insight into thermally activated degradation modes in GaN based lasers. Sudden temperature increase on the laser facet is one of the main factors limiting their performance and lifetime. This paper undertakes the problem of experimental determination of temperature distribution on the mirror of the nitride lasers by applying thermoreflectance spectroscopy.

Until now, the techniques applied to thermal investigation of the nitride devices were electroluminescence or photoluminescence emission spectra (peak position of respective transition is a function of temperature)^16,17) and Raman measurements (the relative shift of the phonon frequency during operation, with respect to the frequency at zero current is temperature sensitive value).¹⁸⁾ All mentioned techniques have temperature accuracy not better than 5 °C and are time consuming which makes them not suitable for temperature mapping.¹⁹⁾ Commonly used technique applied to the thermal investigation of semiconductor lasers, including the nitride devices is thermal imaging.²⁰⁾ However, this approach is characterized by inherently low spatial resolution as well as the fact, that the registered image is averaged over the volume of the device, limiting the ability to observe the enhanced thermal processes occurring at the vicinity of the surface, e.g., front facet. The temperature of the junction can also be obtained from electrical measurements like forward voltage or device resistance.²¹⁾ Thermoreflectance (TR) spectroscopy, as the only thermometric technique, provides the possibility of registering of high spatial and temperature resolution images of the surface of the device operating in quasi-CW or pulsed mode.

This paper presents the unique experimental setup and procedure, devoted to thermal characterization of nitride lasers. Results originating from the experimental work can be directly applied to optimize thermal management in nitride lasers.

TR spectroscopy is an optical modulation technique, which relies on measurement of the relative change in reflectivity induced by periodic change of the sample's temperature. Focused laser beam TR spectroscopy, as one of the methods to determine temperature distribution on the facet of the semiconductor lasers, was used for the first time by Epperlein in 1992 for measuring facet temperature of laser diode (LD).²²⁾ This modulation technique relies on periodic laser temperature modulation induced by pulsed current supply of the device. The probe laser beam incidents perpendicularly to the facet and is reflected back. The periodic temperature change of the laser induces variations of the refractive index and consequently modulates probe beam reflectivity. The temperature of semiconductor lasers is modulated by operating the device in the pulse mode. The change of temperature is connected with the relative reflectance change by the following equation:

$$\begin{equation} \Delta T = \left(\frac{1}{R}\frac{\partial R}{\partial T}\right)\frac{\Delta R}{R}\equiv\kappa^{-1}\frac{\Delta R}{R}, \end{equation} \tag{ 1 }$$

where κ is the thermoreflectance coefficient, which varies strongly with the material being examined,²³⁾ as well as wavelength of probe beam.²⁴⁾ Therefore, it needs to be determined experimentally in case of each device.

The TR has been previously extensively applied to study facet heating of different type semiconductor devices (edge emitting lasers,²⁵⁾ laser bars,²⁶⁾ vertical external-cavity surface emitting laser — VECSELs²⁷⁾ and quantum cascade lasers — QCLs^28,29)) based on the GaAs or InP material system. The application of TR spectroscopy to investigated the temperature distribution on the front facet of GaN based devices is presented for the first time.

Several aspects of the TR spectroscopy implementation to thermal characterisation of GaN based devices had to be addressed in order to establish reliable, repeatable and accurate methodology of the measurements. Among them were: choice of the proper illumination wavelength and cutting-off device emission and appropriate optics used in the setup. Choice of appropriate probe beam wavelength is very important in case of TR method, because proper wavelength maximizes TR signal strongly enhancing the sensitivity of the technique to small temperature changes. The probe beam energy has to be above the fundamental bandgap of the investigated material, which in case of GaN is above 3.42 eV (362.6 nm) and in the range of maximal change of TR signal (ΔR/R). For practical reasons 355 nm (3.5 eV) laser [UV-diode pumped solid state laser (DPSS) laser with TEM₀₀ mode, low noise and output power ∼20 mW] was used for investigation of GaN-based devices. For this wavelength the TR coefficient was determined (Fig. 1).

Zoom In Zoom Out Reset image size

As in case of each thermometric technique, to recalculate TR signal into temperature variation a calibration procedure has to be performed. Temperature calibration procedure used in this work was performed for the laser diode heated by an external heat source (thermo-electric module used for temperature stabilization, operating in the heating mode). At each temperature setting a reflectance line-scan starting at the substrate and going towards the edge of the p-type contact is registered. This way, different reflectance of the materials of the laser structure can be taken into account by the calibration procedure. It is assumed that in the temperature range 15–60 °C the GaN based LD and the submount are uniformly heated by the thermoelectric cooler (TEC). The temperature was measured using a calibrated thermistor placed very close to the laser chip. Figure 1 shows the dependence of the reflectance signal R on the temperature at two positions at the laser facet, namely at the GaN substrate and at the heterostructure. From a linear fit of the data the (1/R)dR/dT is calculated. The thermoreflectance coefficients obtained by this procedure are κ_Epi = 1.36 × 10⁻³ K (epitaxial layers) and κ_Sub = 1.45 × 10⁻³ K (GaN substrate).

The choice of probe beam wavelength affected the choice of optics of TR setup, as it had to be adapted to ultraviolet light. Also, appropriate optical filters had to be used to cut off the lasing wavelength of the examined device. In case of nitride lasers, the emission is very close to the probe beam wavelength, requiring notch-type filters to be used, either to cut-off the emission or transmit only the probe beam. Due to the high intensity of the UV light, the filters used in the setup need to be of high optical density. Scheme of the thermoreflectance set-up is presented in the Fig. 2.

Zoom In Zoom Out Reset image size

The temperature of investigated lasers was modulated by operating them in pulse mode. During the measurement, the examined GaN based laser is mounted on a temperature-stabilized heat sink with water-cooled Peltier element. The whole assembly is mounted on a x–y–z piezotranslator stage which allows precise mapping of the whole facet area with a step width of 0.1 µm. The probe beam enters an optical microscope and is focused on the sample surface to a spot of diameter of about ϕ_1/e ∼ 0.4 µm when using a 0.65 numerical aperture 74× reflecting objective.

The typical probe beam power on the sample surface is below 100 µW. Such low power probe beam does not introduce additional heating of the sample.²¹⁾ The probe beam spot and the laser facet are simultaneously visualized by a CCD camera. The reflected beam is directed onto a Si photodiode and the resulting output is fed into a lock-in amplifier. The output signal from the detector is analyzed by lock-in amplifier (a signal is collected in-phase with the driving voltage generator frequency). A lock-in detection technique is necessary to achieve a sufficient signal to noise ratio. The signal (ΔR) is normalized by the simultaneously measured dc (R) component resulting in information on the relative reflectance change (TR signal). The thermoreflectance setup was optimized to provide high spatial (∼0.4 µm diffraction limited) and high temperature resolution (1 K).

The investigated structures were obtained by metal organic chemical vapor deposition (MOCVD) on bulk GaN crystals. After epitaxial growth, the devices were processed as oxide isolated, ridge waveguide devices with stripe dimensions of 3 × 700 µm². As active region, two undoped In_0.1Ga_0.9N quantum wells (QWs) and two layers of In_0.01Ga_0.99N barriers were used. Waveguide is constructed with thick AlGaN layers and InGaN layers.³⁰⁾ 200 Å thick AlGaN layer with 12% of Al acts as a electron blocking layer (EBL). The devices were characterized by the following parameters: emission wavelength of 417 ± 2 nm, threshold current density of 3 kA/cm², threshold voltage of 4,2 V and differential quantum efficiency of 0.8 W/A. Devices were mounted p-side up in 56 TO package. Laser diodes facets were coated with AR/HR films.

The facet temperature mapping for GaN-based lasers was performed by TR spectroscopy. Figures 3 and 4 present temperature distribution map and temperature line scans taken across the LD active region, registered at different experimental conditions. Heat load in the devices was changed by changing pulse width at constant frequency or amplitude of driving current. Figure 3 presents temperature distribution map for LD operated in pulse mode with pulse width τ = 800 µs and frequency f = 420 Hz. The laser was operated at I = 200 mA and the bottom side of the laser submount was temperature stabilized at T_HS = 25 °C using Peltier element. The amplitude of supply current was twice the threshold current. The registered map covers the area around the active stripe 90 × 100 µm². The maximal temperature rise equals ΔT = 6.5 K. As could be expected, the highest temperature increase takes place within LD active region layers. The heat generated in the active layer is conducted to the heat sink through the GaN substrate.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

More details on heat distribution and maximal facet temperature increases in the active region can be found in line scans taken at the center of the active region, perpendicularly to the epitaxial layers. Figure 4 shows temperature line scans for different pulse widths, ranging from 200 to 1000 µs, with other operating conditions kept the same as for the TR maps (f = 420 Hz/I = 200 mA). A photograph in the inset in Fig. 4 shows the scanned area with an arrow indicating the place and the direction in which the line scans were measured.

In temperature line scans presented in Fig. 4(a), it can be observed that the maximal temperature rise is localized within 2 µm thick epitaxial layers and, as could be expected, for longer pulses, higher temperature rise is observed. The maximal temperature increase equals ΔT = 6.5 K for the following driving conditions: I = 200 mA, τ = 1000 µs, f = 420 Hz. It can be seen in Fig. 4(b) that for duty cycle ranging from 8 to 25%, the maximal temperature increases linearly with the increase of pulse width. At duty cycle higher than 25% the temperature saturation is observed. This effect is explained by taking into account a specific feature of the thermoreflectance experiment; i.e., the fact that TR signal is proportional to the temperature modulation depth. In the case when the device is not cooling down to the heat sink temperature between pulses, amplitude of temperature modulation is lower, resulting in a lower registered signal of TR. In other words, the observed apparent temperature drop at higher duty cycle is not connected with lower device temperature but with inefficient heat dissipation (heat accumulation).

Additionally, it is observed that even for short pulse width (low duty cycle) almost uniform increase of the temperature in whole GaN substrate is registered, with only small "peak" of the temperature increase above GaN substrate (0.4–0.6 K).

Influence of driving current on temperature distribution on LD facet was also studied. Figure 5 shows temperature line scans for different driving current, ranging from 25 to 200 mA, with the following operating conditions: f = 420 Hz and τ = 400 µs. It can be observed that the maximal temperature rise is also localized in 2 µm thick epitaxial layers and, as could be expected, for higher driving current, higher temperature rise is observed. Maximal temperature increase equals ΔT = 3.9 K for driving conditions: I = 200 mA, τ = 400 µs, f = 420 Hz. Under such operating conditions maximal temperature rise increases linearly with the driving current [Fig. 5(b)].

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Measurement of surface temperature is very important as it provides information about the most stressed part of device, which is the laser facet. It is subject to high optical density (MW/cm²), potentially very prone to optical damage. It is expected that temperature profile along the resonator of laser is not uniform, with the facet region hotter due to reabsorption of emitted radiation. TR mapping of heat dissipation also indicates how to optimize mounting and processing steps to lower facet temperature to improve thermal performance of devices. The results obtained in the TR experiment could be very useful if coupled with numerical model describing temperature rise due to electrical current and carrier diffusion. Numerical model accounting for carrier diffusion and recombination at surface would show how the uniformity of temperature is affected by nonradiative recombination at the facet due to interaction of the carriers with surface defect states, possibly allowing to extract data relevant to specific surface treatment like optical coatings.

The paper describes works towards development of the instrumentation for high-resolution thermal imaging of GaN-based LD based on TR spectroscopy. It is shown that TR can be successfully applied to provide information on heat dissipation in these devices. Temperature distribution maps registered with TR can serve as complementary information to the traditional thermometry methods based on Raman spectroscopy, photoluminescence and thermocamera imaging, providing high spatial resolution information about surface temperature. The TR spectroscopy can be considered as a very useful tool for investigation of thermally induced degradation modes of GaN lasers.

Acknowledgment