Researchers at the University of California at Berkeley made monolithic multiple-wavelength arrays of vertical-cavity surface-emitting lasers using a method that creates consistent wavelengths and yields between wafers and within a single wafer. A single such array could provide the multiple wavelengths needed for wavelength-division multiplexing (WDM) in optical-fiber communications—individually packaged devices are required now for each wavelength. A method that could be used for mass-production would drive the price of the devices down, making low-cost WDM possible.
As early as 1991, Connie Chang-Hasnain and her colleagues reasoned that, as she explains, "a VCSEL is a great laser structure for multiwavelength array fabrication" because a VCSEL resonates in only one Fabry-Perot mode, and the output wavelength could be changed by changing the length of the cavity. Thus, if a monolithic VCSEL array could be grown in which there was a variety of cavity lengths, then the lasers in the array could emit varying wavelengths. By creating a physical slope of gallium arsenide, which is transparent at the near-IR wavelengths of these lasers, VCSELs grown at different points along the slope will emit a range of wavelengths.
Several methods have been investigated for creating this slope, but previous methods could not produce reliable and reproducible wavelengths and wavelength ranges. Now, says Chang-Hasnain, "we feel we’ve got the problem solved." The researchers created monolithic VCSEL arrays with output wavelengths ranging from 924 to 954 nm and several milliwatts of output power.
Structure growth
Wupen Yuen, Gabriel Li, and Chang-Hasnain made a gallium arsenide (GaAs) substrate with a patterned backside in which areas are etched away before being put in a molecular-beam-epitaxy (MBE) growth chamber. The bottom mirror, an n-doped distributed Bragg reflector (DBR) made of aluminum arsenide (AlAs) and GaAs, is grown on the substrate. On top of this DBR, a layer of GaAs is grown. During this normal MBE process, the substrate is heated to around 580°C, but after the GaAs layer is grown, the temperature is increased to more than 640°C.
At this point, there is a temperature differential of as much as 50°C between the areas of the substrate that are in contact with the substrate heater and areas in which the substrate has been etched. In the hotter regions, the GaAs re-evaporates (desorbs) off the wafer, while in the coolest areas the GaAs remains; between the two temperature extremes, a physical surface slope is formed corresponding to the temperature. The temperature is lowered again, and a wavelength adjustment layer, the active region, and top p-doped mirror are grown on top of the GaAs slope region.
Control of the desorption process is vital to reproducing the same wavelengths from wafer to wafer. The Berkeley researchers use in situ laser reflectometry to tell when the GaAs is completely removed from the hottest region, which signals the time to lower the temperature again. The desorption process thus guarantees the wavelength span of the monolithic array, whereas the wavelength-adjustment layer shifts the absolute VCSEL wavelengths to the designed values. The monitoring laser beam from a diode laser operating in the near-IR can be focused to a 1-mm-square spot on the 2-in. wafer.
The researchers produced several wafers of the arrays with wavelength spans within 2 nm of the desired values. The CW performance of an array with 4.8 × 4.8-µm apertures showed a threshold current of 450 µA and output powers of more than 2 mW for the shortest wavelength (924 nm). The longest wavelength—954 nm—had a threshold current of below 3 mA and output power of more than 4.5 mW. All the growth steps are conducted without removing the substrate from the MBE chamber’s high-vacuum environment, which avoids the possibility of contamination.
Ready for technology transfer
The laser arrays can be made with the slope as long as a few centimeters across to as short as about 300 µm. Shortening the distance is not a high priority because the space is needed for contact pads. So far, the researchers have shown wavelength changes of 40 nm between VCSELs 1 mm apart, and Chang-Hasnain expects that wavelength changes of up to 80 nm could be made in a straightforward fashion in the same distance—this is planned for the future.
Other plans include increasing the uniformity of the arrays, including temperature insensitivity of threshold currents and optical powers. Chang-Hasnain would also like to integrate the devices with silicon integrated circuits. "We’re fairly ready for technology transfer," she says.