1. Introduction

There has recently been significant progress in bonding III-V active material onto Si to form lasers [1–3]. Here we demonstrate photonic crystal cavities on silicon on insulator (SOI) system. Most of demonstrations for photonic crystal defect cavities are in suspended membranes [4–6], however this type of laser is not able to operate under continuous wave (CW) conditions because of the lack of efficient heat dissipation. There are only a few photonic crystal cavities on sapphire [7, 8] and AlO_x[9] substrates that can be operated under CW pump conditions. Silicon photonics has received a great deal of attention because of its transparency at optical communication wavelengths, 1.3 and 1.5 μm, and compatibility with existing electronic integrated circuits. Compact size chip-scale lasers on silicon or SOI platforms [1–3, 10] might be a key element of a silicon photonic integrated circuit and other integrated elements such as high speed modulators [11] and low-loss waveguides [12] have been demonstrated.

In this paper, we discuss our demonstration of photonic crystal laser cavities on a SiO₂/Si substrate. This work was done with two-dimensional photonic crystal defect cavities formed in InGaAsP membranes bonded directly to a SiO₂/Si substrate. We also demonstrate tuning of the lasing wavelength by varying the photonic crystal lattice constant. For chip-scale photonic integrated circuits, the thermal properties of the elements are important parameters since high density integration of optical elements requires increasing the thermal dissipation per unit area. By characterizing the shift of the lasing wavelength from our photonic crystal lasers with pumped power, the thermal impedance of this compact light source is evaluated.

2. Device fabrication and characterization

Figure 1 shows an illustration of the structure of the demonstrated photonic crystal cavity. The photonic crystal cavities patterned in an InGaAsP layer are on the top of a SiO₂/Si substrate. The SiO₂ layer has a low index of refraction, approximately 1.46, and a thickness of 1 μm. This is enough to reduce the energy radiated from the cavity into the substrate to a tolerable level. The 240 nm thick InGaAsP epitaxial layers were grown by metalorganic chemical vapor deposition (MOCVD) on InP substrates. These layers contain four compressively strained quantum wells that were designed to support gain at 1.55 μm. The InGaAsP wafer was then bonded directly to a SiO₂/Si substrate at 500 degrees Celsius in a hydrogen chamber, and the InP substrate was then etched by a selective etching procedure performed with HCl acid. A silicon nitride layer was deposited and spin coated with a 3% polymethylmethacrylate (PMMA) resist on the top of the bonded structure. The photonic crystal cavities were patterned using electron beam lithography. The photonic crystal patterns were then transferred into the InGaAsP layers using a reactive ion and electron cyclotron (ECR) dry etching steps with CF4 and CH₄/H₂/Ar chemistries, respectively. These defect photonic crystal lasers were formed in a region in which 37 holes (D₄) were missing from a two-dimensional triangular photonic crystal lattice. We label this cavity D₄, as the 37 missing holes corresponds to a hexagon with a “radius” of 4 holes. The defect region of these cavities is approximately 3.2 μm in diameter. In this study, we fabricated an array of devices which have different photonic crystal lattice constants. Figure 2 shows a scanning electron microscope (SEM) image of a D₄ photonic crystal cavity on a SiO₂/Si substrate.

These laser cavities were optically-pumped at room temperature using an 850 nm diode laser at normal incidence with a 1% duty cycle and an 800 ns pulse width. The pumping spot was focused by a 100x objective lens to a spot about 2.5 μm in diameter. The output power was collected by a multi-mode fiber connected to an optical spectrum analyzer.

 

Fig. 1. Illustration of a two-dimensional photonic crystal defect cavity on a SiO₂/Si substrate.

Download Full Size | PDF

 

Fig. 2. Scanning electron micrograph (SEM) of a photonic crystal cavity on a SiO₂/Si substrate from an angled view. The cavity is formed in the region in which 37 holes are missing from the photonic crystal lattice.

Download Full Size | PDF

3. Lasing data

Figure 3 shows the optical spectrum and input power-output power characteristic for a D₄ photonic crystal laser with a lattice constant of 392 nm. The lasing wavelength of the cavity is 1541.7 nm. This device has a 13 dB side-mode suppression-ratio under these pulsed pumping conditions. The incident threshold pump power of the cavity is about 1.5 mW and the estimated absorption threshold power is approximately 730 μW. One of the advantages of photonic crystal lasers is their lithographically defined resonant wavelength. Figure 4(a) shows the lasing spectra from cavities with lattice constants of 392, 396, 398 and 400 nm. These four cavities are all lasing in the same mode with a normalized frequency (a/λ) of approximately 0.255. In Fig. 4(b), we plot the lasing wavelength of the photonic crystal D4 cavities versus their lattice constants. The lasing wavelength of the D₄ cavities in this data shifts by approximately 13 nm by increasing the lattice constant by 4 nm. The wavelength tuning rate is about 3.2 nm for a 1 nm variation in lattice constant.

 

Fig. 3. (a)The lasing spectrum of a D4 photonic crystal cavity on a SiO₂/Si substrate. The lasing wavelength is 1541.7 nm. (b)The incident power versus the output power (L-L) curve for this cavity. The threshold power is about 1.5 mW.

Download Full Size | PDF

 

Fig. 4. (a)The lasing spectra of four D4 photonic crystal cavities on a SiO₂/Si substrate with lattice constants of 392, 396, 398 and 400 nm. (b)The plot for lasing wavelength versus lattice constants of the D4 cavities. The grey line is from a linear fitting for the data.

Download Full Size | PDF

4. Comparison of experiment results and 3-D model

To better understand the lasing modes of the D₄ photonic crystal cavity on a SiO₂/Si substrate, a three-dimensional (3-D) finite-difference time-domain (FDTD) method was used to simulate this cavity. Figure 5 shows the calculated quality factors (Q) of the resonant modes versus normalized frequency. Superimposed on this data is an experimentally obtained lasing spectrum. We do not expect to obtain resonance modes which have Q values over 10,000 from the D₄ cavity on a SiO₂/Si substrate since the index contrast of core layer and cladding layers is reduced compared to a suspended membrane cavity with air as the lower cladding layer. In fact, there are only a few modes with Q over 1,000 in our simulation results. The two highest Q modes within the simulated region are labeled A and B in the spectrum. This lasing spectrum in Fig. 5 shows lasing at mode A. Good agreement between the measured lasing spectrum and the calculated Q spectrum was obtained not only for lasing in mode A, but also at other high Q resonance modes like mode B. There is approximately a 1% difference between the experimental resonance wavelengths and the theoretically predicted wavelengths. We attributed this small shift to imperfections of the fabrication and a slight inaccuracy in the indices of materials used in the 3-D FDTD simulation. It is also worth noting that the lasers operated at mode A instead of mode B even through mode B is expected to have a higher Q value. This occurred because the gain profile of the InGaAsP quantum wells from fabricated cavities was better aligned to mode A. Mode B, which is more than 45 nm away from the gain peak of the quantum wells had much less gain due to this spectral mismatch.

 

Fig. 5. Comparison of the measured spectrum and the modeled spectrum. The blue line is a lasing spectrum from a D4 photonic crystal laser on a SiO₂/Si substrate, and the grey spectrum is the calculated resonance wavelengths with Q values for this cavity.

Download Full Size | PDF

5. Thermal impedance

To characterize the thermal impedance of these cavities, we monitored the shift of the lasing wavelength with changing pumping power. Figure 6 shows the lasing wavelength of a photonic crystal laser versus the incident optical pumping power. The rate at which the lasing wavelength shifts with increasing pump power,$\frac{\Delta \lambda }{\Delta P}$ is approximately 0.55 nm/mW. The thermal impedance of this bonded photonic crystal cavity can be determined by the following equation [13]

where R_Th is the thermal resistance, ΔT is the temperature change, ΔP is the absorbed power from optical pumping, and Δλ is the lasing wavelength shift of the device. The lasing wavelength shift due to temperature changes, $\frac{\Delta \lambda }{\Delta T}$, is about 0.05 nm/K [5]. After substituting these two values into equation (1), we obtained a thermal impedance of this photonic crystal laser that is approximately 11 K/mW. This value for photonic crystal lasers is larger than values reported for vertical-cavity surface-emitting lasers (VCSELs), which about 1 K/mW [13, 14]. We also obtained a thermal impedance value of this D₄ cavity on a sapphire substrate of approximately 2.6 K/mW. We attribute this lower value to the fact that the sapphire substrate has no oxide layer between the InGaAsP membrane and the high thermal conductivity substrate.

 

Fig. 6. The lasing wavelength of a D4 photonic crystal laser on a SiO₂/Si substrate versus the optical pumping power. The slope is 0.55 nm/mW for this cavity.

Download Full Size | PDF

6. Summary

In summary, two-dimensional D₄ photonic crystal defect lasers have been demonstrated in InGaAsP membranes bonded to a SiO₂/Si substrate. The lasing wavelength is around 1.55 μm and the incident lasing threshold pump power is about 1.5 mW. The lasing wavelength can be fine-tuned by varying the lattice constant of the photonic crystals. The thermal impedance of this D₄ cavity is also evaluated.