In the format provided by the authors and unedited. DOI: 10.1038/NPHOTON.2017.56 Hybrid indium phosphide-on-silicon nanolaser diode Guillaume Crosnier 1,2, Dorian Sanchez 2, Sophie Bouchoule 2, Paul Monnier 2, Gregoire Beaudoin 2, Isabelle Sagnes 2, Rama Raj 2, Fabrice Raineri 2,3* 1 STMicroelectronics SA, 850 Rue Jean Monnet, 38926 Crolles, France 2 Centre de Nanosciences et de Nanotechnologies, CNRS, Univ. Paris-Sud, Université Paris-Saclay, C2N- Marcoussis, 91460 Marcoussis, France 3 Université Paris Diderot, Sorbonne Paris Cité, 75207 Paris Cedex 13, France * To whom correspondence should be addressed. E-mail: fabrice.raineri@c2n.upsaclay.fr Composition of the InP-based NIP heterostructure Materials Thickness (nm) Doping (cm -3 ) InP 105 N (Si) 2-3x10 18 (x6) In 0.84 Ga 0.16 As 0.48 P 0.52 55 nid In 0.84 Ga 0.16 As 0.76 P 0.24 7 nid In 0.84 Ga 0.16 As 0.48 P 0.52 15 nid In 0.84 Ga 0.16 As 0.48 P 0.52 25 nid In 0.84 Ga 0.16 As 0.48 P 0.52 110 P (Zn) -2-3x10 18 InP 20 nid Supplementary Table 1: Composition of the InP-based heterostructure bonded to SOI. nid: not intentionally doped. NATURE PHOTONICS www.nature.com/naturephotonics 1
Design of the InP-based 1D photonic crystal cavity: The holes of the photonic crystal are etched through the InP-based rib waveguide such as the lattice constant is varied as indicated in Supplementary Fig. 1. Supplementary Figure 1: Lattice constant a i variation strategy in the x-direction. i=0 represents the central cell of the nanocavity. Here, the lattice constant is varied from 300 nm in the centre to 330 nm at both extremities of the nanocavity. NATURE PHOTONICS www.nature.com/naturephotonics 2
Evanescent wave coupling of the InP-based 1D photonic crystal cavity to the SOI waveguides: Supplementary Figure 2a) shows the calculated Q-factor (Q L ) of the InP nanorib cavity (including the absorptive doped layers) integrated on top of a w Si wide, 220nm thick SOI wire waveguide. The cavity is separated from the waveguide by a 40nm BCB layer and a SiO 2 layer of thickness, t ox, varying from 300nm to 600nm. For all the t ox considered, Q L finds a minimum for w Si =550nm, indicating a maximum evanescent coupling strength for this particular width. For example, for t ox =300nm, Q L can be tuned at will from 550 to 4.13x10 4 by varying w Si. To give an insight on the coupling efficiency (η) as a function of w Si and t ox, we plot on Supplementary Figure 2b) the fraction of the optical power not coupled to the waveguide, 1- η given by 1 = Where Q i is the intrinsic quality factor, here equals to 4.4x10 4. We show that η can be tuned by varying w Si and be as large as 99% for w Si =550nm and t ox =300nm. Such a large η comes at the price of a low Q L which may prevent the structure from lasing if the active material gain is not large enough. Our experimental operation set-point is displayed in both figures (Q L= 1.7x10 3 and =1. = 83%). 10 5 a 10 0 b Loaded Q-factor 10 4 10 3 600nm 500nm 400nm 300nm Experiment 10 2 400 500 600 700 800 SOI waveguide width (nm) 1- η 10-1 10-2 10-3 600nm 500nm 400nm 300nm Experiment 400 500 600 700 800 SOI waveguide width (nm) Supplementary Figure 2: (a) Calculated loaded Q-factors and (b) resulting fraction of optical power not coupled to SOI (1-η) for our hybrid nanorib cavity as a function of w Si and t ox. The experimental data (red star) is that of a passive nanocavity which presents a loaded and intrinsic Q-factors of 1.7x10 3 and 1x10 4 respectively. NATURE PHOTONICS www.nature.com/naturephotonics 3
Impact of the P contact size and position on the radiative carrier recombination profile: Supplementary Figure 3a) displays the maximum value of the radiative recombination rate (R rad ) (red line) and the associated full width at half maximum (FWHM) of its spatial profile (blue line) as a function of the P contact length (L) along the semiconductor strip drilled with the PhC. Here the P contact is 1µm away from the edge of the strip. While the FWHM increases rapidly from 5µm to 14µm with L varying from 1µm to 15µm, the R rad maximum is clamped at 4.1x10 27 cm -3.s -1 for L > 3µm. We plot on Supplementary Figure 3b) the same 2 quantities as a function, this time, of the distance (d) of the p-contact to the edge of the semiconductor strip. Here, L is fixed at 3µm. The R rad maximum diminishes as the P contact is distanced from the strip due to the increase of the electrical resistance. The FWHM increases softly from 3µm to 6µm for 0µm<d<1.5µm. Maximum of R rad (cm -3.s -1 ) 5x10 27 4x10 27 3x10 27 2x10 27 1x10 27 0 a 0 0 2 4 6 8 10 12 14 16 L (µm) 20 15 10 5 Rrad spatial profile FWHM (µm) Maximum of R rad (cm -3.s -1 ) 1.0x10 28 8.0x10 27 6.0x10 27 4.0x10 27 2.0x10 27 0.0 0 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 d (µm) Supplementary Figure 3: Optoelectronic properties of the hybrid nanolaser as a function of the P contact position and length. a) Maximum of radiative recombination rate (red line) and FWHM of its spatial profile (blue line) as a function of the P contact length (L) along the semiconductor strip. Here d=1µm. b) Maximum of radiative recombination rate (red line) and FWHM of its spatial profile (blue line) as a function of the P contact distance, d, to the edge of the semiconductor strip. Here, L=3µm. For all simulations, the bias voltage is fixed at 1V. b 10 8 6 4 2 Rrad spatial profile profile FWHM (µm) NATURE PHOTONICS www.nature.com/naturephotonics 4
Rate equation model for the InP-based nanorib laser The expected current laser threshold is calculated by solving in the steady state the following standard rate equation for the carrier density (N) and the photon number S: =. =. With the electrical injection efficient, I the current, q the electron charge, the active volume, the carrier non radiative lifetime, F p the Purcell factor, B the bimolecular recombination coefficient, g the gain coefficient, N tr the carrier density at transparency, the photon lifetime and the spontaneous emission factor. The parameters defining our system can be estimated from literature or measurements, as summed up in the following table: (cm -3 ) 0.15x10-12 (ns) 7 F p 2.3 B (cm 3 /s) 3 10-10 g (cm -3 /s) 1.38x10 25 N tr (cm -3 ) 10 18 (ps) 1.4 0.1 can directly be determined from our P(I) laser characterisation. From the obtained curve, we derive a slope efficiency P/ I= (0.28±0.07)W/A giving a differential quantum efficiency, = = (35±0.9)%. with, the optical coupling efficiency evaluated here to be 80%. This gives = (44 ± 1)%. The threshold current derives as with the carrier density at the threshold h = h h + 1 h = 1 NATURE PHOTONICS www.nature.com/naturephotonics 5
I th is calculated to be 64µA which is in the order of magnitude of the measured value. This relatively high value finds its origin in the rather low loaded Q factor of 1700 and to a lesser extent in the active volume which is larger than in references 22 and 25. To diminish the current threshold, it is necessary to increase the loaded Q factor of the cavity. This is possible by improving the processing technology and reducing the coupling strength of the cavity to the waveguide. Indeed, there is room for improvement as the calculated intrinsic Q factor is 4.4x10 4 while the measured one is of about 10 4. By maintaining the same optical coupling efficiency, the new loaded Q would be 7500. This would ensure the preservation of the same high differential quantum efficiency while the threshold current could be reduced to 14µA by lowering the number of QWs in the active region to 2. NATURE PHOTONICS www.nature.com/naturephotonics 6
Process flow of the hybrid nanolaser diodes: Supplementary Figure 4: Process flow of the hybrid nanolaser diode. 7 NATURE PHOTONICS www.nature.com/naturephotonics 7