Photonic Crystal VCSEL simulations

Vertical-cavity surface-emitting lasers (VCSELs) require lateral mode control for high performance optical communication and sensing systems. This can be achieved by utilizing 2D PC cavity patterns etched approximately halfway into the top DBR of the VCSEL. We will simulate two different structures here: in the first part we analyze in detail the geometry explained in ref. [1] and then we will show 3D FDTD simulation results of the geometry in ref. [2]. We will use a new saturable gain material plugin model to excite the cavity. You can find more details on how to setup and run the simulation as well as example files in a detailed example that we can provide to interested users.

Step 1. Planar VCSEL design

The planar structure is based on the structures shown in the ref. [1]. We use our stackrt and stackpurcell script functions to design top and bottom DBR mirrors and an active region, and then perform several calculations to calculate the mirror reflectivities, Purcell factors, the dipole power, cavity Q factor, and the E and H fields in the cavity at resonance.

Step 2. Transverse eigenmode analysis:

We use the finite difference eigensolver (FDE) of MODE Solutions to look at the types of modes that are guided by the PC structure in the VCSEL.

Step 3. FDTD simulations

In FDTD simulations, it is useful to first consider a 2D approximation of the structure before moving to 3D.

  1. 2D simulations: We run two simulations, once with no gain material to capture the resonant modes of cold cavity and once with the saturable gain material. Using different symmetry boundary conditions (BCs) we can separate two degenerate modes. The plot below shows the resonant mode profiles with a gain material when symmetric (left) and antisymmetric (right) BCs are used:


    When lasing, there will be mode competition between these modes resulting in a superposition of these two modes. Ideally, we want to design a VCSEL that is single mode with a large suppression ratio.

  2. 3D simulations: A key challenge in running 3D simulations is computational time while making sure that results are reliable. Mesh overrides are selected assuming that light will travel at an angle of less that 55\(^o\) (13.5\(^o\)) in air (VCSEL). The figure below shows the spectrum and mode profile inside the VCSEL when a linear material is used:


    These simulations can be repeated with the saturable gain material centred at the resonant wavelength.

We also run 3D FDTD simulations for the geometry explained in ref. [2] for three different etch depths.The figures below shows the spectrum at various simulation times as well the near and far field profile for a 60% etch depth where the device is singlemode.

References:

[1] Yokouchi et al., “Vertical-cavity surface-emitting laser operating with photonic crystal seven-point defect structure”, Appl. Phys. Lett. 82, 3608 (2003) http://aip.scitation.org/doi/pdf/10.1063/1.1577835

[2] Choquette et al., “High Speed Photonic Crystal Vertical Cavity Lasers” (2011), Optical Fiber Communication Conference and Exposition (OFC/NFOEC), 2011.

Please let us know if you wish to learn more about this example.

6 Likes

Hi bkhanaliloo,

Could you please share simulation files you used for Transverse eigenmode analysis and FDTD simulations?
In particular I would like to learn how you build the MODE simulation files and what is the specific boundary conditions you use in FDTD simulation to excite symmetric and anti-symmetric modes.

Best,
Shimon

1 Like

I provided you a response with private messaging.

Thanks

Thank you very much,
Shimon

1 Like

Hi bkhanaliloo,

I am trying to simulate bandstructure of 2D photonic crystal (square lattice with rods having gain).
As expected, the simulation is diverging out due to gain. I tried out by disabling “Divergence” in FDTD Advanced Tab, but it didn’t help. Could you please help me out as how to calculate bandstructure with gain rods?

Could you please share this simulation file of PhC VCSEL so that I could take a look if it helps me?

Thanks

When gain exists in a material, the time domain signal can quickly reach a very large value, and out of the largest value that a 64 bit computer can handle. To avoid this problem, you may need to reduce the source amplitude .

I guess you are using a linear gain model

if it is other nonlinear material, such as Chi3, you can keep Chi3*abs(E)^2 constant, increasing Chi3 while decreasing E.