1. Introduction

GaN-based diode lasers are key photonic power sources in ultraviolet and blue spectral regions. Their invention by Nakamura [1] followed considerable research and developing work resulting in increasing emission powers and reliability [2–6] without, however, having ever met the benchmarks in terms of absolute powers set by GaAs-based high-power diode lasers. In order to identify the ultimate causes, it is of principal interest to compare properties and actually acting microscopic mechanisms in devices made of the two different III-V material systems.

Except for their primary emission, GaAs-based diode lasers show a number of additional spontaneous emission bands. While such secondary emissions at higher photon energies might arise from second-harmonic-generation at the facets or spontaneous emission from barriers [7] or waveguides, low-energy emission bands are generated by transitions via states generated by defects being located in the active region [8, 9], waveguide [10], or the substrate [11]. Planck’s black-body radiation appears in the mid-infrared and is practically employed for thermography [12, 13]. The shortwave infrared (SWIR) emission found practical application, as well [13]. It peaks at 1300 nm wavelength, and originates from defect states located in the waveguide of the devices [10, 13]. The SWIR intensity turned out to be a measure of the non-equilibrium carrier concentration in the waveguide, allowing for a non-destructive waveguide mapping in spatially resolved detection schemes.

Recently the application of thermography to GaN-based high-power diode lasers has been reported [14, 15]. Information on secondary SWIR emission from such devices is scarce. Infrared photoluminescence (PL) from GaN:Pr samples has been observed by Birkhahn et al. [16]. Garter et al. [17] discussed the emission from GaN:Er-doped material and discovered even 1550 nm electroluminescence (EL) from Schottky-barrier LEDs. All these emissions are line spectra caused by inner-shell electronic transitions in the Pr- and Er-dopant atoms. SWIR PL bands from nitrides have been reported by Sedhain et al. [18] and Feng et al. [19]. In concert with investigations on the yellow PL in GaN [20], Sedhain was able to assign the observed 1010 nm emission band to transitions involving a deep-level complex consisting of a gallium vacancy and an oxygen atom sitting on one of the neighboring nitrogen sites [18, 21]. Feng et al. have shown that the development of this emission band depends on the GaN growth pressure [19]. This confirms Sedhain’s assignment to transitions via levels created by native defects [18, 21]. Moreover, there is one theoretical paper by Leung et al. [22], who discussed the possibility of defect levels due to charge accumulation on threading edge dislocations in GaN. In concert with native defects, and depending on the actual charge state, such defect levels possess a variety of energies. Thus, when exciting an ensemble of defects in a macroscopic focus, the resulting defect-related PL may eventually appear as band being spectrally located in the SWIR. Altogether, these findings show that GaN-based structures are very likely to contain native defects, which give rise to SWIR emissions. Notice that these reports are either on PL or on special infrared emitting devices [19]. Up to now reports on SWIR bands from blue-green emitting devices are not available.

All these reports on secondary SWIR PL from blue-green emitting GaN-based materials do not concern the field of primary SWIR emissions from GaN-based structures employing intersubband transitions [23], which are not subject of this report.

Analysis of secondary SWIR emissions from GaN-based devices, however, is relevant because of several reasons. First, it is desirable to learn about their microscopic nature. Second, it should be tested whether this radiation could be used for non-destructive analysis, for instance by employing imaging techniques as earlier demonstrated for GaAs-based devices [10]. Third, SWIR-radiation is a potential source of systematic error in infrared thermography [13]. In this letter, we report the results of a study of the SWIR emission in the 900-1700 nm range from GaN-based diode lasers by imaging and PL and EL spectroscopy. We find three emission bands, named A, B, and C, which are generated predominantly in the GaN-substrate. Band B (1130-1350 nm) is entirely due to transitions via deep-level defects located in the substrate. Band A (up to 1130 nm) and C (beyond 1350 nm) involve additionally genuine defect-related EL from the active region and Planck’s black-body radiation, respectively. EL from the active region and the thermal radiation have the potential to be used in imaging applications.

2. Experimental

The test devices analyzed in this study stem from three different lots. The epitaxial structures are based on InGaN multi quantum well active layers, GaN waveguides, and AlGaN cladding layers. The chips are 1.2 mm in length and 200 µm in width. The emitter stripe width is 15 µm. Laser threshold and slope efficiency are ~0.2 A and ~1.6 W/A, respectively. Continuous-wave power levels are on the order of several Watt [14, 24]. The chips are mounted by hard solder in a p-side up configuration either on TO56 sockets or C-mounts; see scheme in Fig. 1(a).

 

Fig. 1 (a) Schematic of a diode laser (side view) and the coordinate system used. (b) SWIR EL pattern from the side at I = 100 mA (i.e. below the laser threshold). Image of the device under external illumination by a light bulb and an integration time of 10 ms. (c) The same view without extra illumination and an extended integration time of 3.3 s. (d) SWIR EL pattern at I = 1.5 A from the front facet of a device. The laser chip geometry (dimension 200 × 100 µm²) is highlighted by dashed lines, while the vertical dashed line indicates the position from where the profiles have been taken. (e) Profiles of primary emission at 450 nm (full) and SWIR emission (dotted) as taken from cuts along growth direction (x).

Download Full Size | PDF

SWIR imaging is accomplished by an EHD IK1523 camera with a NAVITAR 12X zoom objective (spatial resolution: 8.2 µm/pixel). Alternatively, the camera is mounted (without objective) to a Leica DM4000M optical microscope. The system works in the wavelength range 950-1700 nm, but the short-wave detection range has been further limited by 1100 nm long-pass filters. Thus SWIR imaging is accomplished in the 1100-1700 nm range.

EL and PL spectra are measured at ambient temperature using an Acton SP2760i monochromator with Princeton Instruments 7525-001 camera or a BRUKER IFS 66v Fourier-spectrometer, the latter being equipped with a LN₂-cooled D418-9 Ge-detector. It allows for detection in the wavelength range from the visible to ~1700 nm. The PL is excited by a diode-pumped frequency doubled Nd:YVO₄ laser emitting at 457 nm. EL is monitored, except for the external excitation, under identical conditions as PL. ZnSe- and GaAs-wafers are employed as longpass-filters in order to eliminate any primary emission. Most PL measurements with the Fourier-spectrometer are performed as quasi-DC-measurements with unmodulated excitation. Notice, that a standard measurement with a Fourier-spectrometer is eventually also an AC-measurement, because the motion of the interferometer mirror modulates the signal, as well. In an extra set of experiments, the excitation laser beam is modulated (f = 1.5 kHz) and a Stanford SR830 lock-in serves as preamplifier between detector and spectrometer-electronics. The latter approach corresponds to an AC-measurement, which damps ‘slow’ contributions, i.e., such ones which decay-time-constants exceeding 1/f ~670 µs. Comparable modulation techniques have been originally introduced into mid-infrared Fourier-spectroscopy in order to get rid of thermal backgrounds that distort DC mesurements [25].

3. Results

Figure 1(a) depicts a scheme of a packaged device including the coordinate system. Figures 1(b)-1(d) show SWIR emission images from devices packaged on C-mounts. Figure 1(b) and 1(c) are side views of the same operating device, taken with 10 ms and 3.3 s integration time, respectively. Under additional external illumination by a light bulb, Fig. 1(b) clearly reveals the geometry including submount and bond wires on top. Figure 1(c) is taken without any extra illumination and proves the SWIR emission to emerge along the entire cavity (z-direction). Figure 1(d) gives the front view (x-y-plane). Figure 1(e) shows normalized cuts taken along the dashed line (x-direction) in Fig. 1(d) for both SWIR and primary emissions.

Figure 2 shows SWIR emission spectra in the 900-1700 nm range as taken from the front facet of an operating device. Obviously, there are 3 distinguishable spectral bands, which we name A, B, and C. Figures 2(b)-2(d) show their dependencies on the operation current as taken by integrating the signals in the wavelength intervals that are shaded in Fig. 2(a). The resulting curves differ substantially, and the laser thresholds at ~200 mA represent pronounced turning points; see dashed lines. Figure 2(e) shows temperature dependent spectra.

 

Fig. 2 (a) SWIR EL spectra in the 900-1700 nm range monitored from the front facet of an operating device for different operation currents. The 3 distinguishable emission bands are named A, B, and C. (b-d) operation current dependencies of bands A, B and C. The ordinate values are obtained by integrating the signals in the wavelength intervals that are shaded in (a). The laser thresholds at ~200 mA are indicated by vertical dashed lines. (e) Temperature dependent spectra. Notice that at 100°C the laser operation terminates and exclusively spontaneous emission remains

Download Full Size | PDF

Figure 3 presents SWIR PL spectra in the 900-1700 nm range taken from the GaN-substrate at the front facet of the device at different temperatures. Note the PL excitation wavelength of 457 nm, which is close to the emission wavelength of the device of 450 nm. As already seen in the SWIR EL spectra in Fig. 2, there are again 3 distinguishable spectralbands, which we call A-C, as well. The temperature dependencies of these bands are given in Figs. 3(b)-3(d). Strikingly different behaviors are observed. Figure 3(e) shows the result of an excitation power dependent measurement, which shows band C to rise disproportionally. In Fig. 3(f), we present spectra as obtained by DC- and AC-measurements; cf. consideration in the preceding paragraph on the methodology used in Fourier-spectroscopy. Thus, Fig. 3(f) demonstrates band C to feature a time-constant exceeding 670 µs, while band B has a time constant of this order of magnitude. Even band A seems to involve ‘slow’ contributions.

 

Fig. 3 (a) SWIR PL spectra in the 900-1700 nm range as monitored from the GaN-substrate at the front facet. The PL excitation power is 50 mW and the 457 nm laser is not focused. There are 3 separated spectral bands A-C. (b-d) Temperature dependencies of bands A-C. (e) Normalized PL spectra taken in an excitation power dependent measurement. (f) Normalized PL spectra obtained by a standard quasi-DC-measurement and an AC-measurement with modulated f = 1.5 kHz excitation.

Download Full Size | PDF

4. Discussion

It is worthwhile to address the implications of these experimental results. Figure 1 clearly reveals the fact that SWIR emission emerges from the active region, in both cases in below- and above-threshold operation; see Fig. 1(c) and 1(d), respectively. Moreover, we observed that even completely failed lasers still show SWIR emission. This proves the emission to be spontaneous and not, for instance pumped by the devices laser emission. Figures 1(c)-1(e) clearly show that SWIR is detected from the substrate as well, and not exclusively from the active region. This could be either SWIR light from the active region that passes through the substrate only, or emission originating from defects in the substrate. Photoexcitation by spontaneous primary radiation could be the root source of energy for the SWIR emission, provided an effective excitation transfer mechanism from defects absorbing at 450 nm spontaneous emission to such emitting in the SWIR. The absorption coefficients in highly doped GaN as used for diode laser substrates are α_GaN(450 nm) ~10 cm⁻¹ and result in absorption lengths (1/e-decent) of ~1000 µm [26, 27]. The SWIR profile shown in (d) shows an 1/e-decent at ~20 µm corresponding to ~500 cm⁻¹. This raw estimate points rather to generation of the SWIR emission via defect-related transitions in the substrate than to pure EL originating predominantly from the active region.

Spectral analysis of this emission is presented in Figs. 2 and shows the presence of three pronounced emission bands. Obviously, the two ones named B and C contribute to the images given in Figs. 1, which are actually taken in the 1100-1700 nm range.

We now discuss the microscopic nature of these emission bands. Although the excitation mechanisms, device operation and PL, as well as the origins of the signals, active region and substrate, are totally different, the spectra in Figs. 2(a) and 3(a) look surprisingly similar including the presence of the bands A, B, and C. This suggests the direct conclusion that a substantial fraction of the SWIR emission from the operating device is indeed defect-related PL from the GaN-substrate, as already assumed when addressing the SWIR images. This is further supported by the current dependence of band B, see Fig. 2(c), which resembles the rate of spontaneous primary EL generated in the active region including the Fermi-level pinning above threshold. In conclusion, band B is formed by transitions involving deep levels, being located in the substrate and being photo-excited by spontaneous 450 nm emission.

A very similar behavior is observed below the laser threshold for emission band A; see Fig. 2(b). Above threshold, however, there are distinctions: With increasing currents, there is a change in lineshape from a top-hat- (below threshold) to a tail-shape (above threshold) in Fig. 2(a), which looks at elevated currents like an extension of more powerful emission at wavelengths shorter than 900 nm that are actually blocked by the longpass-filter. Obviously, this contribution grows with current, and this growth makes band A, in contrast to B, also growing further beyond the laser threshold; see Fig. 2(b). As this ‘tail-contribution’ is not observed in the PL from the substrate, see Figs. 3, it is likely a contribution from the active region that is not excited in the PL experiment. Since the ‘tail-contribution’ is still present after lasing has terminated at 100°C, see Fig. 2(e), it must be genuine EL and nothing caused by the lasing. Based on this, it is fair to assume that its contribution is proportional to the operation current. This allows estimating (extrapolation to I = 0) that at the laser threshold both contributions to band A, genuine EL from the active region and deep-level contributions from the substrate, are almost similar in magnitude. During operation, however, the EL dominates band A. In conclusion, band A consists of at least two contributions. The nature of first one is very much similar to the one of band B, while the second one is an EL that originates from the active region.

We now address band C. In contrast to bands A and B, both EL and PL of band C increase with temperature; see Figs. 2(e) and 3(a-d). In PL experiments, increasing excitation power leads to increasing contributions from band C; see Fig. 3(e). The same holds for increasing currents in EL experiments; see Figs. 2(a) and 2(d). In both cases, the increasing excitation, either laser power or current, causes heating of the region, from where the spectra are taken. In the case of EL experiments, we were even able to quantify this by thermocamera measurements, which were performed with devices being packaged on the TO56 sockets. At 1 A operation current, a temperature increase of up to 70 K is disclosed [14, 28]. At a total temperature of 370 K substantial contributions of Planck’s blackbody-curve [29] enter the spectral sensitivity range of Ge-detectors. This is illustrated by the curve (black line) in Fig. 3(d), which was calculated by integrating Planck’s curve in the 1500-1600 nm range and adding a temperature-independent offset, which stands for deep-level contributions. Thus, there are four completely independent experiments which show that increasing temperature promotes band C. These facts make plausible an interpretation of band C at least in part as thermal radiation. The use of Fourier-spectroscopy offers another option for verifying the nature of band C. This experiment has been addressed already, when presenting the results shown in Fig. 3(f). While the DC-measurement shows the bands A-C, the lower cutoff frequency of 1.5 kHz in the AC-measurement removes band C and partly band B. Even band A gets slightly modified. Thermal processes are known to be ‘slow’, compared to the time-constants relevant in semiconductor recombination kinetics. Thus, the substantial reduction of band C in the AC measurement is an independent indication of the thermal nature of band C, but not unambiguous evidence. Deep-level related recombination time constants might be as ‘slow’ as thermal effects are. In conclusion, band C includes contributions of Planck’s thermal radiation.

All investigated devices from all three batches show the emission bands A, B, and C in both the PL and EL spectra. Their relative weights, however, differ slightly. Comparing the PL spectra in Fig. 3(a) from a device of one batch, and (e,f) from a device of another batch, we find that band A is by 20-30 percent more pronounced in the second one (compared to B and C). The lineshapes of the bands are, however, in all cases almost identical. Since the PL reveals only the deep-level contribution to band A, we assign the different weights of band A to slightly different deep-level concentrations in the substrates of the two batches.

5. Summary

We have presented a study of the SWIR emission from 450 nm emitting GaN-based diode lasers in the 900-1700 nm range. We find three emission bands, named A, B, and C, which are generated predominantly in the GaN-substrate by photoexcitation of deep-level defects by spontaneous primary emission. While band B is entirely due to transitions via deep-level defects located in the substrate, band A and C involve additionally EL from the active region and Planck’s black-body radiation from the entire device, respectively. Filters allow for separation between these bands, for instance the EL from the active region (A) and the thermal radiation (C). SWIR thermography could become practically feasible, in particular, if SWIR cameras with extended infrared range to 2350 or 2500 nm are used. In such applications, however, care must be taken in order to avoid residual effects of deep-level emissions, which would result in overestimated temperatures. If one is interested in the deep-level emission, care must be taken to avoid residual effects of EL and thermal radiation. The short wavelength limit of our investigations was 900 nm. We suggest proceeding with analysis of the EL in the spectral region between the primary emission and 900 nm.