IEEE JOURNAL OF QUANTUM ELECTRONICS, VOL. 45, NO. 5, MAY 2009 531 Single-Shot, High-Speed, Thermal-Interface Characterization of Semiconductor Laser Arrays Nicholas G. Usechak, Member, IEEE, and John L. Hostetler Abstract—Through a detailed characterization of thermally induced output power degradation it is possible to use junction heating as a tool to resolve thermal interfaces on s timescales using a single-shot characterization technique. In this work, the deleterious effect junction heating has on the optical output power of a laser array is characterized and then used to infer the time-dependent junction temperature in response to current pulses of varying widths. The extracted parameters are also used numerically to model the laser as a temperature-dependent heat source for thermal simulations. This treatment allows realistic packaging and emitter-placement studies to be parametrically performed by incorporating the relationship between temperature and output power/efficiency for each emitter. In this respect, once the tem- perature behavior of a single emitter is quantified, the operating temperature and output power performance can be accurately predicted for any realistic physical arrangement of laser array and packaging. The experimental method presented in this work is also compared to other techniques and numerical simulations using the nonlinear heat source; this demonstrates the utility of this approach and the convenience of using easily measured parameters in thermal simulations. Index Terms—High-power lasers, semiconductor device testing, semiconductor laser arrays, single-shot thermal measurements, thermal modeling. I. INTRODUCTION O NE OF THE largest problems facing high-power semi- conductor laser array development is thermal loading. High temperatures reduce electrooptic efficiency, stress solder interfaces, and ultimately reduce the lifetime of the laser. In fact, temperature-driven degradation is often used to perform accel- erated lifetime testing in the semiconductor industry. Of the variety of ways to measure junction temperature in a laser, arguably the most widely used approach relies on the linear relationship between the centroid of the optical spectrum and temperature [1]. This approach is favored in the semiconductor laser community since it does not require expensive equipment, is easy to apply, and does not suffer from contact issues [2]. Using this approach, CW measurements Manuscript received June 02, 2008; revised August 01, 2008. Current version published April 17, 2009. N. G. Usechak is with the Air Force Research Laboratory, Wright-Pat- terson AFB, OH 45433 USA (e-mail: nicholas.usechak@wpafb.af.mil; nick.usechak@gmail.com). J. L. Hostetler is with TRUMPF Photonics Inc., Cranbury, NJ 08512 USA (e-mail: john.hostetler@us.trumpf.com). Color versions of one or more of the figures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identifier 10.1109/JQE.2009.2013097 of laser junction temperatures provide one method to assess the efficiency of the laser and its packaging. Nevertheless, this approach is limited in that it provides no information on whether a packaging problem exists in the laser, solder inter- faces, or cooling structure; time-resolved junction temperature measurements can fill this void [2]–[11]. However, schemes that employ the use of monochromators or optical spectrum analyzers in conjunction with long-time averaging, temporal gating (optical or electronic) [2]–[7], high-speed detectors [11]–[13], or streak cameras [14]–[17] implicitly average over many pulses to build up a spectrogram. It is hard to quantify what impact this averaging will have on accuracy or to what extent a laser may degrade during these measurements, which can be time-consuming to make. For these reasons, techniques that operate in a single-shot capacity are attractive. To that end, approaches that use high-speed thermal imaging can be used but are generally limited by integration times of millisec- onds [10], [18]. Still, other approaches exist such as using grating-based spectrometers and acquiring spectra as quickly as the spectrometer’s CCD array can be swept ( 500 ns per pixel [2], [19]). 1 These approaches work well in the 100-ms regime but can only operate in the s regime at the expense of reduced spectral resolution and/or range [19] and do not seem likely candidates to work on s timescales for the characterization of high-power laser arrays in the near future since wavelength shifts of 10 nm are routine and resolutions of 0.1 nm are sought. Remote temperature-sensing techniques provide an alternate way to determine laser temperature. For example, using an auxiliary laser and an interferometer, real-time heating on picosecond time scales can be monitored [20]–[22]. Since this approach requires a reference laser and is influenced by charge-induced changes in the refractive index, the quantity measured is not purely thermal. A reflectance modulation tech- nique that exploits the change in refractive index has also been used to investigate temperatures in semiconductor lasers and provides one alternative [23], [24]. Other single-shot remote sensing techniques also exist such as focusing an auxiliary laser obliquely on a layered part of the device to be characterized and using Fabry–Perot interference as a tool to resolve the width of the etalon formed by the layer [25]. Using the coefficient of thermal expansion of the layer, the temperature can then be inferred with ps resolution. Of course, this technique requires accurate knowledge of the device structure and the material properties of both the substrate and the layer used to form the etalon. Still another approach uses an external laser of 1 This is related to the rate at which the CCD pixels are digitized 2 MHz for commercial linear CCD arrays. 0018-9197/$25.00 © 2009 IEEE