Record-breaking mid-wave infrared LEDs
Adding a resonant cavity to mid-wave infrared LEDs featuring an interband cascade structure results in record spectral intensities, valued for chemical sensing and infrared scene projection.
BY JEREMY MASSENGALE, CHUL SOO KIM, CHADWICK CANEDY, IGOR VURGAFTMAN AND JERRY MEYER FROM THE U.S. NAVAL RESEARCH LABORATORY, ALONG WITH MIJIN KIM FROM AMENTUM AND COLIN IVERSEN AND VIJAYSEKHAR JAYARAMAN FROM PRAEVIUM RESEARCH
Emerging as a strong contender for a compact, efficient source emitting in the mid-wave infrared, which is the spectral domain spanning 3-5 µm, is the interband cascade laser (ICL). Now it is the laser of choice in many battery-driven chemical sensing applications operating on a low power budget.
Behind this success is a marriage between the interband optical transitions of a conventional diode laser and the cascaded stage architecture of a quantum cascade laser. Now an established source, the ICL realised the key milestone of room-temperature continuous-wave (CW) operation in 2008, and since then this laser concept has spawned a host of related technologies exploiting derivative staged designs. They include the IC infrared photodetector, the IC thermophotovoltaic device, and the ICLED.
For some chemical-sensing applications, including pipeline leak detection and industrial process control, ICLEDs are an attractive alternative to ICLs, as their manufacture is more straightforward and less expensive. Additional strengths include vertical emission, so no waveguide structure is needed, and the opportunity to qualify growth/fabrication in bulk before individual devices are diced, cleaved, and mounted with facet coatings.
Another application for ICLEDs is infrared scene projection, to support laboratory testing of imaging infrared focal-plane arrays. Creating an IR scene requires the generation of a high-fidelity synthetic image, which may include features with widely varying temperatures, such as missile plumes, engine heating, vehicles, and personnel. It’s critical to use realistic, scenario-specific conditions, including atmospheric attenuation.
Adding a resonant cavity
To build on the significant strengths of the ICLED, our team – a collaboration between the Naval Research Laboratory, Amentum, and Praevium Research – is improving this device through the introduction of a resonant cavity.
One may ask why this addition is needed, given that conventional ICLEDs are already emitting powers of up to several milliwatts, which is sufficient for many spectroscopic chemical-sensing applications. But the spectral bandwidth is quite broad – it is typically of the order of 1-2 µm, which is far wider than the linewidth of molecular fingerprint absorption features. With the introduction of a resonant cavity, the emission from the ICLED is narrowed by well over an order of magnitude, and when the wavelength is tuned to match a given chemical absorption feature, the upshot is a higher absorption contrast.
When narrow-linewidth RCLEDs are employed for infrared scene projection, they can simultaneously produce signals at multiple wavelengths, with minimal crosstalk between sub-bands. However, to simulate a high apparent temperature, it’s essential to maximise emittance, defined as the output-power per emission area, and also the radiance, which is the emittance per unit solid angle. For applications at very specific wavelengths the spectral intensity, which is the emittance per unit wavelength, is key. For devices with high emittance, introduction of a resonant cavity can greatly enhance the spectral intensity.
Efforts to maximise emittance are impaired by internal losses. When electrons and holes are injected into a mid-wave IR LED they are more likely to recombine non-radiatively via the so-called Meitner-Auger process, than produce light. As the Meitner-Auger decay rate increases exponentially with wavelength, this reduces the output beyond 4 µm – and it explains why the highest reported mid-wave IR emittances occur in the 3-4 µm range. That makes it more challenging to succeed in maximising the emittance and spectral intensity of RCICLEDs operating at longer mid-wave IR wavelengths.
Design, growth and fabrication
To optimise the performance of our RCICLEDs, we have performed a carefully-choreographed dance between design/simulation, MBE growth, cleanroom fabrication, and optical characterisation – when viewed in conjunction, they form a feedback loop. Key to our success is drawing on several decades of development at NRL, where researchers have fine-tuned expertise related to all aspects of interband cascade emitter production. In order to implement the most favourable distributed Bragg reflector (DBR) mirrors for this latest project, NRL joined forces with Praevium Research, a pioneer in the exploitation of heterogeneously-bonded DBRs, to form high-performance resonant-cavity optoelectronic devices.
Figure 1. Profiles of the refractive index (orange) and simulated optical mode intensity (black) along the vertical axis within the RCICLED cavity.
The foundation for the recent success is NRL’s ICLEDs, emitting near 4.5 µm. Those devices employ type-II quantum wells in 20 active cascade stages, split into two groups of 10. This division allows both groups to centre on peaks of the optical field in the resonant cavity (see Figure 1), providing a modest improvement in the output. Growth of this optimised epitaxial design takes place at NRL, using MBE and a GaSb substrate.
Following initial testing at NRL, wafer materials are sent to Praevium. Here, the epitaxial side has been wafer-bonded to a 33-period GaAs/Al0.92Ga0.08As DBR that has a reflectivity of more than 99 percent and forms the bottom mirror of the RCICLEDs. The GaSb substrate is then removed with a chromic acid-based etch, before standard ICLED fabrication techniques are applied, including patterning LED mesas, to form circular apertures with a diameter of 140 µm, and the deposition of annular top and bottom contacts.
We have produced two forms of emitter in the current study. Device A has a top mirror with a single-period SiO2/germanium DBR that has a reflectivity of more than 85 percent (see Figure 2), and Device B has top mirror formed simply by the semiconductor/air interface, which has a reflectivity of around 30 percent. It is expected that the higher reflectivity top mirror of Device A will result in a narrower linewidth in the resulting emission relative to Device B.
Figure 2. Side-view schematic of the epi-down-mounted RCICLED with 20 active stages split into two groups of 10. The two DBR mirrors defining the cavity are 33 periods of GaAs/Al0.92Ga0.08As on the bottom and a single-period of SiO2/germanium on the top (Device A only).
Both classes of RCICLED have been characterised in CW mode at 25 °C. Measurements of output power and emission spectrum involved collecting light with an f/1 optic, which corresponds to a collection angle of 26 °. For cavities with a high quality factor (Q factor), which is typical in RCLEDs, the measured linewidth is actually highly dependent on the collection angle – set by the collection optic used in the measurement!
Spectra of devices A and B reveal resonance linewidths for these LEDs of 40 nm and 107 nm, respectively – that’s dramatically narrower than the more typical value of around 1 µm for conventional ICLEDs processed without a cavity. Peak resonances for devices A and B are 4.47 µm and 4.46 µm, respectively, close to the design value of 4.5 µm.
These results are encouraging. To the best of our knowledge, they are the first reports of room-temperature CW operation of resonant cavity LEDs emitting beyond 4 µm. What’s more, device A has the narrowest linewidth to date for any LED with a peak wavelength beyond 3.4 µm, thanks to its high-reflectivity top mirror. This linewidth, which is limited by collection angle rather than the quality factor of the cavity, is a valuable asset for chemical sensing and multi-spectral IR scene projection applications.
Plots acquired at 25 °C for the CW output power as a function of current and voltage reveal maximum output powers of 0.22 mW and 0.30 mW for device A and B, respectively (see Figure 3). These values, realised for a drive current of 60 mA (near rollover), indicate an emittance of 1.94 W cm-2 for device B, based on emission area. That’s the highest ever value reported for a mid-wave IR LED operating beyond 3.4 µm, whether driven in CW mode or with a high pulsed duty cycle (we ignore results for a device with out-coupling, enhanced by an immersion lens on its emission surface).
Figure 3. Continuous-wave light-current-voltage characteristics (left) and wall-plug efficiency (right) of Devices A (violet) and B (green) at room temperature.
Values for wall-plug efficiency at 25 °C for both devices peak at 0.32-0.35 percent, recorded at low drive current. At higher currents efficiency gradually declines. Although these values may seem low, they are characteristic of mid-wave IR LEDs. While low efficiencies stem partly from non-radiative decay, they take a bigger hit from total internal reflection, which prevents most generated radiation from exiting the device.
A significant strength of our devices is their excellent temperature stability between 25 °C and 75 °C, according to plots of output power as a function of current and voltage for device A (see Figure 4). At the highest measured temperature, wall-plug efficiency still peaks at nearly 0.24 percent. Spectral data indicate an almost negligible widening of the resonance linewidth with temperature (from 40 nm to 42 nm), along with a very gradual red-shift in the centre wavelength, which increases at a rate 0.31 nm °C-1.
A key figure of merit for spectroscopic applications and multi-sub-band infrared scene projection is spectral intensity, which is defined as the emittance per unit wavelength. The most appropriate approach to determine this quantity involves some calculus of the emission spectrum (obtained by determining the spectrum’s derivative as a function of wavelength, when the integration over dl segments is normalised to the total output power). However, there are many reports in the scientific literature where spectral intensities are crudely approximated, simply by taking the ratio of total emittance to the full-width-at-half-maximum linewidth – this is only accurate for a top-hat emission profile.
For our devices, emission includes significant tails (see Figure 4). We ‘removed’ them when remeasuring spectra, using a narrow-bandpass filter that excludes wavelengths far from the peak. This determined that around just 70 percent of the total output emitted by both RCICLEDs fell within the cavity resonances.
Figure 4. Temperature-dependent continuous-wave light-current-voltage characteristics (left), wall-plug efficiency (middle), and spectral intensity (right) of Device A over the range 25-75 °C.
It should be noted that when the width is limited by collection angle, as is the case for these devices, only linewidths measured with the same collection angle can be compared meaningfully. For example, the difference in apparent linewidths measured with an f/1 and an f/2 collection optic can exceed a factor of 3!
We have determined the dependences on wavelength of properly-calibrated spectral intensities for both devices (see Figure 5). For device A, the peak value is 10 mW cm-2 nm-1. This is more than four times the previous best for a mid-wave IR RCLED.
Figure 5. Continuous-wave spectral intensities of Devices A (violet) and B (green) under moderate current injection at room temperature. The legend indicates values at the resonance peaks.
Enhancing output
For mid-wave IR LEDs with a flat, high-index output surface, such as that of a GaSb substrate, most radiation is trapped by total internal reflection, due to a critical angle of around just 15 °. This limits the maximum out-coupling efficiency to about 4 percent, even when ignoring the reflection of some of the radiation falling within this cone (some portion of which may still be emitted following one or more bounces between the top and bottom surfaces). Adding a single-layer anti-reflection coating increases the maximum out-coupling, but only a little, with this figure rising to about 5.3 percent.
These values highlight the considerable headroom for a higher wall-plug efficiency, which may be realised by enhancing the low outcoupling efficiency. The most straightforward way to accomplish this is to replace the flat output surface with an immersion lens or micro-lens. However, that’s less practical when emitters need to be deployed in tightly-packed arrays with small pitch.
Simple alternatives are to roughen the output surface on the wavelength scale, or introduce a specially-designed ‘motheye’ pattern. We have pursued the latter, realising a modest output enhancement (see Figure 6 for a scanning electron microscopy image of a preliminary motheye structure). Efforts are ongoing to exploit methods that deliver high efficiencies in visible LEDs.
Figure 6. (a) Optical micrograph of the moth-eye (ME) ICLED mounted epi-down on a C-mount. The greenish-blue emission from the device is attributed to diffraction by the ME surface pattern. (b) Tilted-view SEM image showing the ME nanopillar array. Scale bars: 1 mm and 1 µm for (a) and (b), respectively.
Towards the future
Our RCICLEDs, emitting a high spectral intensity in a bandwidth that’s as narrow as 40 nm, will be compelling candidates for chemical sensing, multi-spectral IR scene projection, and a number of other applications. Prior to commercialisation, challenges must be addressed, related to controlling the efficiency and directionality of the light escaping the semiconductor. Future designs, incorporating engineered output surfaces, such as motheye, microlens, and integrated metalens structures, will need to increase radiance while preserving the vertical-emission geometry that makes ICLEDs attractive for arrays and scalable manufacturing. In parallel, we will continue to pursue optimisation of the active-stage, cavity, and DBR designs, as this promises further improvements to the emissivity, linewidth, and spectral intensity of these sources.
Praevium Research acknowledges support by SBIR contract number N6833524C0298 funded by Navair, and NRL acknowledges support by a CRADA with Praevium Research.
