Ultraviolet Light

From oemagazine June 2003
01 June 2003
By John Carrano, Asif Khan, Michael Kneissl, Noble Johnson, Geoffrey Wilson, and Richard DeFreez

No longer just a concern of the military establishment, the threat of biological attack has become a real issue to civilian populations throughout the world. Stopping such attacks requires the fast detection of pathogens. Ultraviolet laser-induced auto-fluorescence (UVLIF) of biological material markers provides a real-time technique for detecting airborne pathogens such as aerosolized bacterial spores and viruses.

High-power, short-wavelength UV sources are key to the development and deployment of UVLIF bio-agent detection systems. Participants in the year-old Semiconductor Ultraviolet Optical Sources (SUVOS) Program have made substantial progress in the development of UV-emitting semiconductor diode lasers and the biological agent detection systems made practical with the advent of these devices. Sponsored by the Defense Advance Research Projects Agency (DARPA; Arlington, VA), the SUVOS program has focused on the development of UV LEDs and diode lasers operating in the 280- to 340-nm spectral region. These optical sources are based on III-V semiconductor alloys of gallium nitride (GaN), aluminum nitride (AlN), and indium nitride (InN). Principal concerns have been output power, optical intensity, device lifetime, and device cost.

nifty new nitrides

Using aluminum gallium nitride (AlGaN) pn-junction structures on sapphire substrates, researchers at the University of South Carolina (Columbia, SC) are developing high-power deep-UV LEDs with peak emission at 280 nm, 320 nm, and 340 nm (see figure 1).1-4 In addition to the AlxGa1-xN multiple quantum well (MQW) active region (x=0.1 to 0.42 respectively for the 280- to 340-nm devices), the devices incorporate AlxGa1-xN barrier layers (x=0.15 to 0.5 for 340- to 280-nm devices). For this design, the major factors limiting the LED internal quantum efficiency are the quality of the AlxGa1-xN layers, the sheet resistivity of the n-type AlxGa1-xN barrier layer, and the low doping efficiency of the p-type AlGaN barriers. The sheet resistivity of the n-type AlxGa1-xN barrier layer in turn depends on its doping and thickness, and this resistivity controls the current crowding in the lateral-conduction, mesa-type device structures that are currently under development. This device geometry is dictated by the use of the insulating sapphire substrates, which are necessary for deep-UV light extraction.

Figure 1. Normalized electroluminescence spectra of the South Carolina team's deep UV LEDs at a pulsed pump current of 100 mA show strong emission peaks at 278 nm (black), 325 nm (red), and 338 nm (green) with spectral FWHM of 9 nm, 8.5 nm, and 10.2 nm, respectively. The weak emission band at 330 nm (black) is caused by the recombination via deep levels in the p-AlGaN layers.

To overcome the above barriers, the South Carolina team has used several innovative approaches. They reduced AlxGa1-xN dislocation densities to about 5 * 108 cm-2 by incorporating an AlN buffer layer5 and an AlN/Al0.5Ga0.5N superlattice deposited using pulsed atomic-layer epitaxy. Strain reduction from the superlattice also allows the deposition of 2-µm-thick n-type AlxGa1-xN (for x>0.35) barrier layers with sheet resistivity as low as (250 Ω/□). The lower sheet resistivity also reduces the current crowding, which would otherwise give rise to excessive non-uniform device self-heating under high DC operation.

Self-heating in 325-nm-emission UV LEDs can increase the device temperature by about 70°C for a DC bias current of 50 mA. This overheating results in premature power saturation at high current. To mitigate this effect, the South Carolina group uses a flip-chip packaging technique. The diced chips are mounted on the AlN carriers with p-side down. The carriers are then bonded to Al-coated copper headers, and the UV light is collected through the sapphire substrate.4

To overcome the problem of low doping efficiency for magnesium-doped p-AlGaN, the South Carolina group used a piezo-doping approach with a p-AlGaN/p+-GaN heterojunction to create a hole accumulation layer from the spontaneous polarization at the interface.5 This process increased the p-doping and hole injection into the active region. The p+-GaN layer was also beneficial for p-type ohmic contact deposition, since the contact to high Al-content p-AlGaN is not required.

The room-temperature emission spectra of the devices at a pulsed pump current of 100 mA contain strong emission peaks at 278 nm, 325 nm, and 338 nm. An additional very weak emission band at 330 nm in the case of the 280-nm LED is caused by the recombination via deep levels in the p-AlGaN layers. The group is seeking a solution.

The current-voltage (I-V) characteristics for 200 * 200-µm LED devices emitting at 278 nm, 325 nm, and 340 nm show sharp turn-on voltages ranging from 4.2 V to 5.2 V and a total series resistance of 15 to 28 Ω (see figure 2). For 1 A of pumping current, the devices exhibited pulsed powers as high as 3 mW, 10 mW, and 13 mW, respectively. The external quantum efficiencies of about 0.1% (278 nm) and 0.45% (325 nm) represent record high values and the shortest emission wavelengths to date. The group measured DC powers as high as 0.47 mW (at 260 mA), 1 mW (at 100 mA), and 1.2 mW (at 100 mA) for the 278-, 325- and 338-nm devices, respectively. The devices operating at 325 nm displayed lifetimes in excess of 200 hours for an accelerated life-tests under 20 mA DC pump current at 85° C.

Figure 2. Emitted power versus wavelength for pulsed (circles) and DC-current pumping (squares) show good performance. The pulsed devices used 500-ns-long current pulses at 10 kHz, and the devices achieved external quantum efficiencies of about 0.1% (278 nm) and 0.45% (325 nm). The current-voltage (I-V) characteristics for 340- and 280-nm devices (inset) feature a sharp turn-on voltages ranging from 4.2 V to 5.2 V.
utterly ultraviolet

At the Palo Alto Research Center (PARC; Palo Alto, CA), researchers are also focused on UV diode lasers. PARC's initial efforts concentrated on InGaN quantum-well (QW) devices operating from 365 to 380 nm. The group produced heterostructures constructed from the InGaAlN alloy system and grown on sapphire substrates. Devices featured with threshold current densities near 5 kA/cm2 and light output powers greater than 400 mW at wavelengths between 368 and 378 nm. PARC also achieved room-temperature, continuous-wave (CW) operation of ridge-waveguide devices with threshold currents of 85 mA and output power of more than 5 mW.

While the PARC team continues to improve the performance characteristics of ridge-waveguide diode lasers in the 370-nm range, their main interest is in developing UV lasers that emit at wavelengths as short as 280 nm (see figure 3). The bandgap of AlGaN alloys can be tuned over a wide wavelength range. With GaN QWs, lasers emit at about 363 nm. Decreasing the wavelength further into the UV spectral region requires the introduction of increasing amounts of Al; for example, at 280 nm, the Al concentration in the AlGaN QWs is close to 50%. The semiconductor bandgap energy is increased by raising the aluminum content of the AlGaN alloy. This simultaneously reduces the number of free carriers in the material. The carrier reduction, in turn, leads to increased electrical resistance with consequent device heating, lowered overall device efficiency, and reduced device lifetime. Demonstrating lasers at this wavelength will require resolving a number of materials, growth, and device design issues. The development of efficient active regions in the UV region is an engaging challenge by itself; the PARC approach includes investigating both AlGaN and quaternary InAlGaN MQW active regions.

Figure 3. Leveraging new violet and UV diode laser sources, this biosensor system analyzes aerosols using auto-fluorescence to detect bioagents and direct detection of elastic scatter to size particles.

Besides growth on sapphire substrates, the group is also exploring alternative substrate technologies, which offer lower dislocation densities. These include quasi-bulk GaN and bulk AlN substrates. Last year the group demonstrated the first UV LED on bulk AlN substrates (Crystal IS; Latham, NY). Other materials/device efforts include the development of short-period superlattices for strain compensation and enhanced carrier concentrations in high-Al-containing AlGaN cladding and waveguide layers, as well as the investigation of new laser device concepts such as asymmetric waveguide laser structures that the group has successfully demonstrated with InGaN MQW lasers emitting at violet wavelengths.

busting bio-bugs

It is not enough to develop the UV sources; we need to develop the systems that use them. Pacific Scientific Instruments (PSI; Grants Pass, OR) has been developing an aerosol sampling system that incorporates UV optical sources developed by the SUVOS program.

To date, most UVLIF bio-aerosol sensors have relied on harmonic generation of UV emission using pulsed visible and IR solid-state lasers. Such lasers are costly, bulky, inefficient, and fragile. A less obvious drawback is the aerosol concentration limitation imposed by the laser pulse duty cycle: the sensor sees no aerosols between shots. An even less obvious, but significant, drawback is that self-Q-switched diode-pumped solid-state lasers featured in current bio-aerosol sensors preclude the use of the excitation beam for particle sizing (particle respirability determination). Traditional CW UV sources such as mercury lamps, however, are not bright enough to sufficiently excite the weak fluorescence of tiny bio-aerosol particles for detection against the stray light background.

UV diode lasers have long been recognized as a potential solution to this predicament, but until recently, such devices didn't exist. In mid-2001, PSI developed a compact UVLIF bio-aerosol sensor—the BioLert system—based on a UV diode laser. Using an excitation source emitting at 370 nm, this system induced auto-fluorescence centered around 450 nm while simultaneously determining particle respirability through direct detection of elastic scatter with the 370-nm wavelength. Although this system has proven itself as a fast and sensitive trigger, its ability to sense bio-aerosols is limited by its single-wavelength excitation and single-fluorescence-band detection.

The current generation BioLert system uses simultaneous excitation and detection at multiple wavelengths (see figure 4). The resultant low-resolution spectrum of the aerosol fluorescence will allow improved distinction of biological compounds from inert fluorescent particles such as paper dust, minimizing false positives. Thus far, a BioLert testbed equipped with two fluorescence channels has detected small differences in the spectra of individual spores of Bacillus globigii, a harmless bacteria used to simulate anthrax; fine dust particles; and polystyrene microspheres doped with fluorescent dye. The system is currently configured to excite bio-aerosols at two first-generation SUVOS diode laser wavelengths: 370 nm (nicotinamide adenine dinucleotide) and 410 nm (flavins). The next-generation unit will incorporate 280-nm (tryptophan) and 340-nm (NADH peak) diode lasers being developed and optimized under the SUVOS program.

Figure 4. Room-temperature laser emission spectra for InGaN, GaN, AlGaN, and InAlGaN MQW active regions.

In the present system, a current supply modulates both diode lasers at different frequencies so that their aerosol scatter signals may be distinguished. The beams are combined and intersected with a stream of aerosols to be tested. The power in each beam is monitored as a diagnostic. The baseband elastic scatter signal provides information about aerosol size and, thus, respirability. Aerosol fluorescence is collected in two spectral regions by separate optical ports. Each detected signal contains wavelength components generated by UV and violet sources; the signal is demodulated to provide a total of four fluorescence measurements for each aerosol. The fluorescence and elastic scatter data are coordinated by an event counter, which decides whether potentially hazardous bio-aerosols are present.

Semiconductor UV optical sources offer the potential of low cost, small size, low power, and high reliability needed to enable several important systems of U.S. Department of Defense interest. Current research into the III-nitride systems is critical to this effort. Team members of the DARPA SUVOS program have made substantial progress in the areas of materials growth, semiconductor physics, device design, fabrication, and packaging. The first demonstrations of aerosol biodetection systems with this technology have been highly encouraging. oe

Software: imaging for defense

Those who have delved into systems development know that delays are part of the process. Digital camera simulators can supply simulated real-time digital camera image data, allowing researchers to test their systems even when sensors or camera electronics are not available. On a recent Naval Research Laboratory-sponsored satellite program, for example, engineers used a digital camera simulator (DCS; Nova Biomimetics; Solvang, CA) to perform a wide variety of sensor integration tasks before the final prototype camera systems were even available.

In addition to playback of actual sensor data, engineers were able to create special-purpose test patterns (horizontal and vertical grayscales, walking bit patterns, bar patterns, etc.) to verify proper functionality of downstream signal processing circuitry in the absence of actual camera hardware.

In a recent Phase II Small Business Innovative Research program at Nova, the DCS produced large sequences of simulated IR imagery and sent the data at high frame rate to a SKYsds-1 computer system acting as a programmable digital signal processor for subsequent application of real-time image-processing algorithms. The DCS greatly expedited algorithm development by helping confirm the accuracy of key downstream processing functions.

A DCS can download full-size frames of consecutive images from the system to be simulated, then play them back to downstream electronics as though the actual camera were producing the data. Frame synchronization can be used to control the precise output frame rate produced. The output data rate from a single 7.0 in. * 3.0 in. DCS board can be as high as 40 Mb/s, and several boards can be operated in parallel to produce any desired effective output rate.

In addition to using actual digitized image data, the DCS can feed test patterns or arbitrary digital data to equipment requiring "closed loop" playback of the data for system evaluation and testing purposes. Researchers can use accompanying software to create custom sequences to tailor specific testing and system evaluation requirements.

-Mark Massie, Nova Biomimetics


1. S. Nakamura, S. Pearton, et al., The Blue Laser Diode: the Complete Story, Springer-Verlag, Berlin (2000).

2. A. Chitnis, J. Zhang, et al., Jpn. J. Appl. Phys. 41, L450-L451 (2002).

3. J. Zhang, M. Asif Khan, Appl. Phys. Lett. 81, 4392-4394 (2002).

4. A. Chitnis, S. Jason, et al., Appl. Phys. Lett. 81, 3491-3493 (2002).

5. M. Shatalov, G. Simin, et al., IEEE Electron Device Lett. 23, 452-454 (2002).

John Carrano
John Carrano is with DARPA.

Asif Khan
Asif Khan is with the University of South Carolina.

Michael Kneissl, Noble Johnson
Michael Kneissl and Noble Johnson are with Palo Alto Research Center.

Geoffrey Wilson, Richard DeFreez
Geoffrey Wilson and Richard DeFreez are with Pacific Scientific Instruments.
Recent News