Experts raise doubts about Northwestern results

Hopes that a dramatic breakthrough has been achieved in nitride lasers appear to have been dashed. Preliminary results of this work, published earlier this year in the MRS Internet Journal of Nitride Semiconductor Research suggested that lasers with very low threshold current sensities could be achieved by eliminating the Al from the cladding layer of the laser diode structure. The group achieved excellent p-type ohmic contacts and low series resistance, both problems which have bedeviled most of the other laser groups. At the SPIE meeting in January, the Northwestern group claimed to have observed even lower threshold currents in devices measured at room temperature. But the Northwestern group's suggestion that they had also achieved lasing in these structures appears to have been premature.

Experts who examined the results, which have recently been presented in more detail at technical meetings, are expressing doubt that the devices were anything more than (perhaps superluminescent) light-emitting diodes(LED's). According to Shuji Nakamura, of Nichia Chemical Industries, the world's leader in nitride laser development, "The data is consistent with similar data we obtain for InGaN LEDs. The electroluminscence spectra does not show any sign of spectral narrowing. We always see very narrow spectra at the currents just above lasing." The key piece of evidence appears to be the polarization ratio the Northwestern group is reporting, about 7:1 (TE/TM). "A large polarization TE/TM ratio greater than 100:1 is typically observed in GaN lasers, a ratio of 5-10 is typical for quantum well LEDs" says Dr. Nakamura.

Part of the Northwestern group's problem seems to be that the equipment used to measure the devices had poor sensitivity at the short emission wavelength of GaN. Professor Manijeh Razeghi, the leader of the Northwestern group, responded to Dr. Nakamura's comments as follows:

The equipment we used to measure the emitting spectra was designed for longer wavelengths (> 0.8 microns), and only less than 5 % of light is analyzed by the interferometer. Thus the actual stimulated emission signal to noise ratio is 95 % higher than what was shown in the spectrum that we partially showed in Figure 1c of the paper we submitted to SPIE. The same reason explains the TE:TM mode ratio, which is much higher than 10.

However, for us, as the emission spectrum was not the only criteria necessary to show lasing operation, we carried other routine measurements such as:

Far-field measurements which showed very narrow beam divergence (~ 20° in perpendicular direction, and 13° in parallel direction to the epilayer). Such a narrow far-field cannot be obtained from the spontaneous emission.
Another experimental result that strongly indicated lasing operation was the optical power vs. current curve (i.e., P-I curve), which shows very sharp change of slope by an order of 2 near threshold. Although some have suggested that it could come from band-to-trap type non-radiative recombination, to my knowledge, this model can never explain such a large and sudden change of slope in the P-I curves observed in our lasers. LED's fabricated with the same materials in our laboratory never showed such a "threshold behavior".
The lasers also showed the dependence of J_th on cavity length as expected in semiconductor lasers, that is longer cavity length lasers yield lower Jth. Certainly such a behavior cannot be explained from LED's with any non-radiative recombination models.
We also measured near-field patterns of lasers and LED's fabricated in our laboratory. Both LED's and lasers consisted of exactly the same structures except that LED's did not have polished mirrors. We showed the dependence of near-field of a laser on current. A 100 nm-aperture broad-area laser was used in this experiment, and we could observe the formation of a filament as expected. With an increase in current, only the emission intensity in the filament area increased while the rest of the active area showed much slower increase or almost saturation of the emission intensity, as described in my presentation during the SPIE. Such a behavior, however, has not been observed in LED samples. In LED's, the light intensity increased more or less uniformly in the entire width of the metal contact with an increase of current injection. This difference can be explained by stimulated emission in the filament. Filamentation in the laser results from self-focusing of light which stimulates the recombination of carriers, further increasing refractive index and causing even more self-focusing. Thus the formation of filaments can only be sustained when stimulated emission exists. Without stimulated emission, the light can be non-uniform along the laser width because of composition inhomogeneity, but it can never make filamentation as observed in our lasers.
Manijeh Razeghi
Northwestern University
m-razeghi@nwu.edu

Lasers are distinguished from normal light sources by the modal collapse and the resulting coherence of the light. You can recognize coherence by the ability of the light to interfere with itself. When you shine a laser on the wall, you see a speckle pattern from the interference between different parts of the laser beam. You don't see a speckle pattern in light from from an LED. When you look at an LED, you're seeing emission in many optical modes. As a laser is driven above threshold, a few optical modes, sometimes just one, will "win out". This is because the stimulated emission lifetime drops for the best modes, and they suck all of the gain out of the gain medium. This results in a narrowing of the emission spectrum. In an LED, the spontaneous emission from all the modes increases. Similarly, if emission in one polarization is preferred in an LED, the polarization ratio will dramatically increase when the device begins to lase, as one polarization steals all the gain from the other polarization.

Self-focussing and filamentation are usually not considered to be desirable characteristics in semiconductor lasers. It appears now that even if the Northwestern group is right about whether their devices are lasers, the low threshold current densities they report do not constitute a breakthrough for practical applications.

Eric Hellman, February 25, 1998. Updated February 27, 1998

 MIJ-NSR    News