To reap the benefits of quasi-phasematching across the 2-5 μm and 8-12 μm wavelength regions, AFRL is therefore developing the next generation of nonlinear optical devices. The new technology employs gallium arsenide (GaAs) and similar zincblende crystal structure semiconductor materials, which have several properties particularly suited to IRCM applications. GaAs, for instance, is transparent across the 2-12 μm range and has a high nonlinear coefficient for efficient frequency conversion. Furthermore, GaAs—like all quasi-phase-matched materials—has the capacity to be engineered within broad limits to obtain a desired output wavelength from an available pump source. Engineers pattern the material so as to reverse the sign of the nonlinear coefficient with a periodicity that compensates for the phase mismatch between the input and output waves. Since GaAs is isotropic in its refractive index, quasi-phasematching is the only way to exploit its nonlinear potential.

The main obstacle to developing GaAs for nonlinear devices has been finding a practical method of patterning the material. GaAs is not ferroelectric; therefore, it is impossible to pattern it by electric field poling. An initial patterning approach, pursued by Stanford University researchers over a decade ago, involved polishing GaAs wafers to the necessary thickness and then cutting, inverting, and manually stacking them.¹ Although this early technique produced samples that successfully doubled carbon dioxide laser output, the fabrication process was not only prohibitively labor intensive, but was clearly incapable of producing the material thicknesses and tolerances necessary for reaching optical parametric oscillator (OPO) thresholds.

Several years later, motivated by the significant potential of GaAs as a nonlinear optical material, other Stanford researchers developed a twostep fabrication technique that could produce device-quality samples.² In this technique, growth of the desired pattern occurs on a GaAs substrate using molecular beam epitaxy (MBE). MBE growth is slow, however, and yields patterned growth just a few microns thick. Reaching thicknesses in the hundreds of microns needed for device demonstration requires the deposit of additional material onto the MBE-grown template. Researchers accomplished this using 1960s-vintage technology: hydride vapor phase epitaxy (HVPE).