Deep-UV nitride-on-silicon microdisk lasers

Deep ultra-violet semiconductor lasers have numerous applications for optical storage and biochemistry. Many strategies based on nitride heterostructures and adapted substrates have been investigated to develop efficient active layers in this spectral range, starting with AlGaN quantum wells on AlN substrates and more recently sapphire and SiC substrates. Here we report an efficient and simple solution relying on binary GaN/AlN quantum wells grown on a thin AlN buffer layer on a silicon substrate. This active region is embedded in microdisk photonic resonators of high quality factors and allows the demonstration of a deep ultra-violet microlaser operating at 275 nm at room temperature under optical pumping, with a spontaneous emission coupling factor β = (4 ± 2) 10−4. The ability of the active layer to be released from the silicon substrate and to be grown on silicon-on-insulator substrates opens the way to future developments of nitride nanophotonic platforms on silicon.


Supplementary Note 2: Estimation of the β-factor
In order to estimate the global emission coupling factor , we use the theoretical model developed in the textbook "Optoelectronics" by E. Rosencher  with the transparency carrier density, = ℎ is the ratio between the carrier densities at threshold and at transparency, and is the spatial overlap between carriers and the photonic mode.
is here the product of the spontaneous emission coupling factor, and the quantum efficiency of the active medium ( ).
Under the monomode approximation of the model, the prefactor of is equal to , where is the quantum well radiative decay rate and is the cavity loss rate.
Inverting these two equations provides the power-dependence ( ) of the photon density, that is compared to the experimental input-output characteristics of the microlaser, thus determining the microlaser parameters.
As shown in the Supplementary Note 4, the precise estimation of the quantum efficiency of the active layer is an intricate issue, so that we prefer to provide the global -factor of the investigated microlaser. Moreover, the precise eigennumbers of the lasing modes being unknown, we assume for simplicity an overlap coefficient = 1. For the A1 mode presented in figure 2, the fitted -factor ( = (4 ± 2). 10 −4 ) and the threshold pump power density (Pthr=17±2 nJ per pulse) are calculated for a value = ℎ = 1.7, and a pump power density at transparency Ptr=10 nJ per pulse. Those values are compatible with the observation of the mode narrowing for this mode at P=Ptr.

Supplementary Note 3: Influence of the quantum confined Stark effect and comparison between microdisks embedding GaN/AlN quantum dots and quantum wells
Due to the strong internal electric field existing in GaN/AlN heterostructures, and the induced Quantum Confined Stark Effect (QCSE), the overlap between electrons and holes wavefunctions can be drastically reduced for QDs/QWs thicker than 5MLs that emit below 3.9 eV 8 . This results in a strong decrease of the oscillator strength of the optical transition.

Supplementary Note 4: Internal quantum efficiency of the GaN/AlN quantum wells
The estimation of the internal quantum efficiency (IQE) through photoluminescence experiments is an intricate issue. Figure S2.a presents the temperature-dependent PL spectra of a single GaN/AlN QW equivalent to the ones of the microdisk active layer under CW excitation at 5.07 eV (244nm). The spectra consist in a single peak which energy roughly follows the shift predicted from the Varshni law of the AlN barrier. As shown in Figure S2.b., the latter is close to the QW emission. The IQE is usually obtained from the ratio (300 )/ (5 ) of the PL intensities at We conclude from these measurements that the IQE is not an intrinsic feature of the active layer, and it strongly depends on the excitation conditions (laser energy, power density, …). Assuming that the nonradiative processes are negligible at T=5K and the highest excitation power density, the measured ratio (300 )/ (5 ) is probably an optimistic estimate of the IQE of the active layer, ranging between 10% at low power and 40% at the highest accessible cw power ( = 6.6 10 4 . −2 ). The same estimate of the IQE reaches 80% under the pulsed excitation identical to the microdisk laser operation. In a recent work on AlGaN/AlN QW ridge lasers 6 an IQE of 10% was obtained in the linear regime, also measured under strong excitation power density. This value appeared as the minimum IQE to reach laser action in the device; it required a drastic improvement of the dislocation density, down to 5. 10 8 −2 , i.e. 100 times lower than in our case.