Quantification of scattering loss of III-nitride photonic crystal cavities in the blue spectral range

The mechanisms contributing to experimental quality factors of short wavelength ($\lambda =440–480$ nm) III-nitride on silicon one-dimensional photonic crystal cavities were quantified. Fluctuations in fundamental and first-order cavity mode wavelength and quality factor were compared over sets of nominally identical cavities. Unlike at $\lambda =1.5\mu \mathrm{m}$, experimental quality factors were not limited by fabrication disorder modeled as smooth, normally distributed hole size and position variations; after ruling out absorption losses, additional scattering losses were found to predominate at short wavelengths. Experimental quality factors were sensitive to conformal deposition of few nanometer thin films on the photonic crystal surface, suggesting that the additional scattering losses were linked to the surface.

Figure 1

Highest reported photonic crystal cavity $Q$ for various materials plotted versus wavelength in material, where $n$ is the refractive index including dispersion. Although materials and cavity designs differ, a clear trend of decreased $Q$ with decreasing wavelength is evident [2, 4, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23].

Figure 2

(a) Schematic of nanobeam photonic crystal design employed in this study and corresponding electric field intensity profiles for fundamental (blue) and first-order (red) cavity modes simulated by 3D-FDTD. (b) SEM image of finished nanobeam suspended in air between anchored support pads. (c) Typical $\mu \mathrm{PL}$ spectrum of a single nanobeam displaying two modes with $Q>5000$ resonances. The far field $\mu \mathrm{PL}$ mapping with 2 $\mu \mathrm{m}$ spatial resolution of the relative intensities of the modes (top and bottom insets) for qualitative comparison with simulation: the blue-shifted resonance is the symmetric, fundamental cavity mode, whereas the red-shifted mode is the antisymmetric first-order cavity mode. Black lines show the nanobeam location during the scan.

Figure 3

Power dependence of the fundamental (blue) and first-order (red) photonic crystal cavity mode linewidths at room temperature. Vertical line shows the excitation power density used for $Q$ measurements.

Figure 4

(a) Amplified spontaneous emission and (b) net modal gain spectra observed from the cleaved facet of the InGaN/GaN single QW on a Si substrate at pump fluence of 5 mJ/${\mathrm{cm}}^2$ (dark red) and InGaN/GaN multiple QWs on a free-standing GaN substrate at 0.7 mJ/${\mathrm{cm}}^2$ (blue). The absorption losses of the GaN layers on silicon are bound at $<50{\mathrm{cm}}^{-1}$.

Figure 5

Critical dimension (hole size and position distribution) extraction from SEM images for modeling of fabrication disorder. (a) High magnification view of nanobeam holes with algorithmically detected hole (white solid) and best-fit hole profile (red line). (b) Histogram of extracted hole diameters exhibiting Gaussian distributions with standard deviation less than 1 nm. $N=85$, 128, 158, and 60 holes were analyzed for the four target diameters, 45, 55, 65, and 75 nm, respectively. (c) Histogram of extracted hole centroid distances from the hole center line. (d) Hole sidewall deviation of $2.0^{\circ }\pm 0.3^{\circ }$ deduced from top views with CD extraction (left) and crosssections of cleaved etch test samples (right).

Figure 6

Simulated effect of fabrication disorder. (a) 3D-FDTD computed wavelength and (b) $Q$-factor distributions for 100 disorder realizations (${\sigma }_x={\sigma }_y={\sigma }_r=1$ nm) of a single cavity geometry comprised of two 18-period mirror sections (period 119 nm/hole radius =35 nm) surrounding an 18-period quadratic taper section (period 119 nm/central hole radius 42 nm). Disorder simulations include a realistic $2^{\circ }$ sidewall taper.

Figure 7

Lithographic tuning and $Q$ statistics of (a) fundamental and (b) first-order nanobeam cavity modes. Each color represents a lithographic tuning data set for each of the four base designs described in Table 1. The vertical box plot details the experimental $Q$ over 12 nominally identical nanobeams per point. Simulated interquartile range and median $Q$ are depicted by the shaded regions and line, respectively. $Q_{\mathrm{off}}$ values of 6100 and 8400 are required to fit the 3D-FDTD simulations to experiment for fundamental and first-order cavity modes, respectively.

Figure 8

Fundamental mode degradation (decrease in $Q$ and red-shift) observed during $\mu \mathrm{PL}$ spectroscopy in ambient. The inset is a SEM image at 3 kV taken after the experiment, showing carbon deposition where the laser spot was focused.

Figure 9

Fundamental mode fabrication statistics for Design 2 before (dark blue) and after (light blue) ALD coating with 3 nm of ${\mathrm{HfO}}_2$. Twelve nanobeams were measured per point. A red-shift and decrease in $Q_{00}$ for the same corresponding nanostructures was observed after coating.