LASER DESIGN: Software models laser diodes with no iteration

A laser-diode-simulation software package introduced by Nonlinear Control Strategies (NLCSTR; Tucson, AZ) is quite a bit more intricate than the laser-diode modeling software that most laser engineers are used to; one reason is that it is the summation of 15 years of research at institutions such as the College of Optical Sciences at the University of Arizona (Tucson, AZ) and Philipps Universität at Marburg (Marburg, Germany).

Jul 1st, 2009
Th Lasers D 01

A laser-diode-simulation software package introduced by Nonlinear Control Strategies (NLCSTR; Tucson, AZ) is quite a bit more intricate than the laser-diode modeling software that most laser engineers are used to; one reason is that it is the summation of 15 years of research at institutions such as the College of Optical Sciences at the University of Arizona (Tucson, AZ) and Philipps Universität at Marburg (Marburg, Germany).

Jörg Hader, senior scientist at Nonlinear Control Strategies and an associate research professor at the University of Arizona, describes the need for software complexity in the design of semiconductor emitters.

“The processes that determine the light generation in a semiconductor are microscopic interactions between electrons and photons,” says Hader. “Because there is an incredibly high number of electrons present in the semiconductor and all are interacting with each other as well as with the surrounding lattice (phonons) and photons, these processes are complicated microscopic many-body interactions.” Existing commercial software models macroscopic heat, carrier, and light dynamics; however, describing the generation (or loss) of light in the laser’s active region has been done only at a general level, he notes.

One problem with a simplified model is that it has to rely on pre-existing data against which the calculations are fitted to determine the parameters. “One cannot use these values for devices and situations that are different from the ones for which the values were obtained,” says Hader. “Thus, one cannot make predictions.”

Developing a semiconductor light emitter requires many steps: coming up with a design, growing the designed structure, processing the wafer into actual devices, testing the device, analyzing the test results and adjusting the design, and growing the adjusted structure.

If an already working device needs to be optimized, then the engineer will simply change some small aspect of the original design; this will cost very little. But if one is developing a truly optimized device or a fairly new design, then these steps generally have to be repeated multiple times; as a result, the costs can multiply by 5 to 20 times, Hader explains.


Modeled and experimental output powers and wavelength shifts for a 1040-nm-emitting VECSEL agree closely. (Courtesy of Nonlinear Control Strategies)
Click here to enlarge image

“Quite often, one has to grow several test samples to adjust the growth conditions in order to really hit the design specs,” he says. For a complicated structure like a vertical external-cavity surface-emitting laser (VECSEL)—which might require diamond heat spreaders, heat sinks, external mirrors and possibly antireflection coatings—this can take a technician days before even the first growth of the complete structure can be attempted.

The new software cuts down these iterations by telling the engineer how quantitatively correct the crucial aspects of the proposed device will be, including lasing wavelength, threshold, tuning, and so on. Its abilities have been verified by taking existing complex semiconductor-laser designs, entering their parameters into the software, and confirming that the simulation matches the actual laser performance, all without iteration.

Experimental confirmation

In one example, an optically pumped VECSEL pumped at 808 nm, emitting at 1040 nm, and operating at room temperature is modeled, showing excellent agreement with experimental data.1 The laser consists of a distributed Bragg reflector with 20 pairs of quarter-wavelength layers of gallium arsenide (GaAs) and aluminum arsenide (AlAs), and an active resonant periodic gain region with 10 InGaAs wells separated by barrier layers; the chip is attached to a diamond heat spreader mounted on a copper heatsink.

Fully detailed many-body calculations were done for gain, absorption, reflectivity, spontaneous emission, carrier losses, and so on. Among many examples of theoretical and experimental data matching up are those for output powers and wavelength shifts as a function of pump power (see figure).

A cost comparison based on the experiences of a German client who eliminated some iterative steps in the design and fabrication of a new laser diode (from 20 steps down to five steps) showed an 80% reduction in cost to bring the laser from design to working model; in addition, there was a time savings, which would directly affect time to market for a new product.

—John Wallace

REFERENCE

  1. Jörg Hader et al., IEEE Journal of Quantum Electronics, submitted to IEEE Journal of Quantum Electronics, May 2009.

More in Software