eigenmode analysis of vertical-cavity lasershiptmair/org/macsinet/slides/...• semiconductor stack...
TRANSCRIPT
ZürichIntegrated Systems LaboratoryOptoelectronics Modeling Group Zürich
Eigenmode Analysis ofVertical-Cavity Lasers
Andreas Witzig, Matthias Streiff, Wolfgang Fichtner
Integrated Systems LaboratorySwiss Federal Institute of Technology,
ETH Zurich Switzerland
ZürichIntegrated Systems LaboratoryOptoelectronics Modeling Group Zürich
Overview
• Motivation
• Solution of the Optical Modesin Laser Cavities➜ Discretization of Vectorial Helmholtz Equation
by linear and quadratic Nédélec Elements
➜ Solution of the Algebraic Eigenvalue Problemby Jacobi-Davidson QZ Algorithm
• Spectral Portrait➜ Identification of Optical Eigenmodes
➜ Target Values
➜ Tracking of Optical Modes
• Computational Effort➜ Linear vs. Quadratic Edge-Elements
• Conclusion / Outlook
ZürichIntegrated Systems LaboratoryOptoelectronics Modeling Group Zürich
d / 2
λ-cavity
15 Bragg pairs
20 Bragg pairs
GaAs Substrate
metal contact
Si3N4 isolation
ApertureMetal Contact
Si3N4 isolation25 Layer Pairs
λ-Cavity, Active Region
35 Layer Pairs
Vertical-Cavity Surface-Emitting Lasers (VCSELs)
• Application➜ Promising Light Source for Optical
Communication Systems
➜ Semiconductor Laser (PiN-Diode)
➜ Light Generation by RadiativeRecombination of Electrons and Holes
➜ Optical Resonator (Layered Medium)
• Challenge➜ Calculate Optical Eigenmodes
➜ Calculate Resonance Frequency
➜ Simulation of Opto-Electro-ThermalDevice Characteristics
10µm
Ligh
t Out
put
Standing Wave
Metal Contact
GaAs Substrate
ZürichIntegrated Systems LaboratoryOptoelectronics Modeling Group Zürich
Problem Formulation:
Calculation of Optical Eigenmodes
➜ Frequency-Domain Solution ofHomogeneous Maxwell ’s Equations(=Helmholtz Equations)
➜ Radiation Boundary Condition
➜ Non-Hermitian ComplexGeneralized Eigenproblem
➜ Real Part of Eigenvalue= Resonance Frequency(Free-Space Resonance Wavelength λ)
➜ Imaginary Part of Eigenvalue= Field Decay Constant(related to Photon Lifetime τph)
Optical Field
20µm
+ Radiation Boundary Condition
ZürichIntegrated Systems LaboratoryOptoelectronics Modeling Group Zürich
Temperature
Optical Field
20µm
Peak Change in
by ≈ 2%Refractive Index
Application:
Coupled Opto-Electro-ThermalVCSEL Simulation
Refractive Index Depends on Temperatureand Carrier Density
➜ Need Volume Discretization (Finite-Ele-ment Type, Full-Vectorial)
Solution of Optical Cavity Problem is Iteratedwith Electronic Equations
➜ Need a Fast Solver (in Compensation,allow to use a lot of memory)
Current Density
ZürichIntegrated Systems LaboratoryOptoelectronics Modeling Group Zürich
Discretization
• Discretization by N édélec Elements➜ Unstructured Grid (typ. order 500 ’000)
➜ Full-Vectorial Complex Optical Field
➜ "curl"-Conforming Discretization
➜ Linear and Quadratic Elements
• Solution in 2D or "2.5D"➜ Body-of-Revolution
Fourier Series Expansion in φ-Direction
ρ
edgesnodes
ϕρ
z
z
ZürichIntegrated Systems LaboratoryOptoelectronics Modeling Group Zürich
r�
�[�r]�1
� r ��
�� k20["r]� = 0:
["r] =2
4� 0 0
0 � 0
0 0 �
�1
35[�r] =
24
� 0 0
0 � 0
0 0 ��1
35
Solution ofThe Open Cavity Problem
• Radiation to In finity➜ Essential Result of the Simulation
(Directly proportional to Photon Life Time,Laser Threshold, Filter Linewidth, etc)
• Perfectly Matched Layer (PML) Concept➜ Arti ficial Material for Absorption
➜ Outside of PML, use Dirichlet BoundaryConditions.
Very good Results, Problem with Layered Media PML
ZürichIntegrated Systems LaboratoryOptoelectronics Modeling Group Zürich
Solution ofAlgebraic Eigenproblem
• Jacobi-Davidson QZ Algorithm➜ Low-Dim. Search Subspace (e.g. order 10)
➜ High-Order Correction Equation (BiCGstab)
➜ Preconditioner: Matrix Inversion,using Direct Solver PARDISO
• In Practice...➜ Find a few inner Eigenmodes (1..5)
➜ Need Good Target Value (know λ within 0.1%)
➜ Need a Lot of Memory for Computation (6GB)
➜ Quick Method to Find Eigenvalue (20min)
ZürichIntegrated Systems LaboratoryOptoelectronics Modeling Group Zürich
101
102
103
104
105
106
10−10
10−8
10−6
10−4
10−2
rela
tive
err
or
matrix order
linear
quadratic
Reference I:Metal Box Resonator
• Analytical Solution Known
• Comparison of Linear and QuadraticElements
• Symmetry similar to VCSEL(Body-of-Revolution, Fourier Expansion of 3DFields)
ZürichIntegrated Systems LaboratoryOptoelectronics Modeling Group Zürich
Reference II:Dielectric Pill
Resonator
• Cylinder with n = 5.9in Vacuum
• Includes Radiation➜ Good Test for Absorbing
Boundaries
➜ Non-Hermitian Matrices, Com-plex Eigenvalue
• Symmetry similar to VCSEL➜ Body-of-Revolution, Fourier
Expansion
Real Part of the Electric Field
m = 1 m = 0
ZürichIntegrated Systems LaboratoryOptoelectronics Modeling Group Zürich
2676 2680 2684 2688 2692
54
55
56
57
linear
quadratic
1
2
31
23Reference [1]
Resonance Wavelength [nm]
Ph
oto
n L
ifet
ime
[ps]
Reference II:Dielectric Pill
Resonator
• Accuracy Dependent onMesh Refinement:
• 1: 10 Cells/λ (8x8 in Pill)linear: order 9’000quad.: order 36’000
• 2: 20 Cells/λ (16x16 in Pill)linear: order 35’000quad.: order 140’000
• 2: 40 Cells/λ (32x32 in Pill)linear: order 139’000quad.: order 556’000
[1] Tsuji et. al. "On the Complex Resonant Frequency of Open Dielectric Resonators",IEEE Trans. on Microwave Theory and Techniques, Vol. 31, No. 5 May 1983
ZürichIntegrated Systems LaboratoryOptoelectronics Modeling Group Zürich
Example I:Airpost VCSEL
• Semiconductor Stack➜ Low Refractive Index Contrast
➜ Many Layers (30..40)
• How to Find Fundamentaland Higher Order Modes?➜ Spectral Portrait
Fundamental Mode
ZürichIntegrated Systems LaboratoryOptoelectronics Modeling Group Zürich
730 740 750Resonance Wavelength [nm]
0
4
8
12
16
20
Pho
ton
Life
Tim
e [p
s]
m=0m=1m=2m=3m=4m=5
Example I:Airpost VCSEL
• Spectral Portrait➜ Mode Selection??
➜ Mode with Smallest ImaginaryPart and Large Overlap withActive Region...
• Note: Different ExpansionNumber m
Spectral Portrait
ZürichIntegrated Systems LaboratoryOptoelectronics Modeling Group Zürich
Example I:Airpost VCSEL
• Spectral Portrait➜ Mode Selection??
• Mode with SmallestImaginary Part andLarge Overlap withActive Region...
Spectral Portrait
ZürichIntegrated Systems LaboratoryOptoelectronics Modeling Group Zürich
730 740 750Resonance Wavelength [nm]
0
4
8
12
16
20P
hoto
n Li
fe T
ime
[ps]
Spectral Portrait
Inve
rse
Ph
oto
n L
ife
Tim
e [1
/ps]
Resonance Wavelength [nm]
ZürichIntegrated Systems LaboratoryOptoelectronics Modeling Group Zürich
7.56 7.57 7.58 7.59x 10
−7
0
500
1000
1500
2000
2500
lambda
A
B C
D E
alph
a
8
9
11
11
11 9
9
6
6
6
7
4
4
4
9
4
3
3
3
7
4
2
1
5
2
3
4
4
2
2
5 3
3 3
4
2
4
4
5
4
5
5 4
3
3
4 3
4
3
3
4
3
2
2
3
6
5
5
4
4
4
4
6
4
4
3
6
3
7
7
4
4
4
4
3
2
2
3
A
BC
D E
2.5
2.0
1.5
1.0
0.5
0
744.6 744.8 745.0 745.2
Resonance Wavelength [nm]
How to Find Complete Spectrum?
• "Brute Force": Scan Target Value➜ Faster Convergence for Better Target Value
(Number of Jacobi Correction Equations Solved)
➜ New Preconditioner for Each Eigenvalue?
• Repeated Solution of Perturbed Problem➜ Track Eigenvalue
Inve
rse
Ph
oto
n L
ife
Tim
e [1
/ps]
ZürichIntegrated Systems LaboratoryOptoelectronics Modeling Group Zürich
Example II:Tunable Filter
• Air-Gap Mirrors➜ High Refractive Index Contrast
➜ Need Only 3..4 Layers
• Fabrication:➜ Selectively Ethching Sacrifice Layers
➜ Air Bridges Hold the Layers
• Concept Allows Tuning
• Filter: Incident Light from Top➜ Resonances in the Cavity Cause Sharp
Spectral Dip in the Reflected Beam
ZürichIntegrated Systems LaboratoryOptoelectronics Modeling Group Zürich
Example II:Tunable Filter
• Charge Center Layers to ControlAir-Gap Distance
• Critical Design Issue: Bending of theLayers
• Sensitivity Analysis by Simulation
Tuning
Bend
< 20nm
ZürichIntegrated Systems LaboratoryOptoelectronics Modeling Group Zürich
1..20nm
ZürichIntegrated Systems LaboratoryOptoelectronics Modeling Group Zürich
1580 1585 1590 1595 1600 1605
1500
1550
1600
1650
1700
1750al
ph
a [1
/m]
lambda [nm]
20 nm
0 nm
Convex Mirrors
➜ Trajectory for 0nm..20nm bend
➜ Mode Relations are Preserved
➜ Better Confinement (Smaller Decay Const.)
➜ Shift to Smaller Resonance Wavelengths
ZürichIntegrated Systems LaboratoryOptoelectronics Modeling Group Zürich
1580 1590 1600 1610 16201000
1500
2000
2500
3000
3500
alp
ha
[1/m
]
lambda [nm]
20 nm
0 nm
previous slide
Concave Mirrors
➜ Trajectory for 0nm..20nm bend
➜ Resonances
➜ Shift to Higher Resonance Wavelengths
ZürichIntegrated Systems LaboratoryOptoelectronics Modeling Group Zürich
Example III:VCSEL Simulation
• Separation between Longitudinal andTransverse Direction Impossible➜ Diffraction at Oxide Confinement or Mesa
Structure
➜ Remedy for Rotationally Symmetric Struc-tures: Expand by Fourier Series (Body-of-Revolution)
• Solution of Large Eigenvalue Problem➜ Finite-Element Discretization
➜ Need to Resolve Bragg Mirrors
➜ Jacobi-Davidson Method
20µm
ZürichIntegrated Systems LaboratoryOptoelectronics Modeling Group ZürichZürichIntegrated Systems LaboratoryOptoelectronics Modeling Group Zürich
1.
2.3.
4.
Zoom-In:
• Standing Waves➜ High Field Intensities
in the Centre
• Traveling Waves1.Laser Output
2.Bragg Losses
3.Radiating Waves
4.Guided Surface Waves
• Diffraction at theOxide Confinement➜ Important to Calculate Pho-
ton Lifetime and ThresholdCurrent
ZürichIntegrated Systems LaboratoryOptoelectronics Modeling Group Zürich
Typical Simulation Tasks
• Device Optimization➜ Mode Discrimination➜ High Output Power➜ Good Fiber Coupling
• Design of Top Contact➜ Aperture Diameter➜ Material
• However...➜ Results Questionable for
Optics-Only Simulation➜ Need To Simulate Fully-Coupled
Opto-Electro-Thermal Physics
2µm
3µm
4µm
5µm
Aperture
0.6 1.81.2 2.4
Distance from Centre [µm]
Op
tica
l Po
wer
Den
sity
[a.
u.]
3.0 3.6 4.20.0
ZürichIntegrated Systems LaboratoryOptoelectronics Modeling Group Zürich
1 2 3 4 5 6 70
0.02
0.04
0.06
0.08
0.1
0.12
0.14
dielectric aperture radius [um]
(QH
E11
−QH
E21
)/Q
HE
11
Simulation Results (Optical): Mode DiscriminationFurukawa Electric 850 nm VCSEL with oxide aperture:Optimal single-mode performance (cold cavity)
➜ Adjust dielectric aperture to obtain maximum separation betweenfundamental HE11 and next higher order HE21 optical mode.
HE11 (m=1)
HE21 (m=2)
1 2 3 4 5 6 70
2
4
6
8
10
12
dielectric aperture radius [um]
mo
dal
loss
[1/
cm]
total loss HE11top radiation HE11total loss HE21top radiation HE21
ZürichIntegrated Systems LaboratoryOptoelectronics Modeling Group Zürich
Electronics Optics
n, p: Electron/Hole DensityV: Electrostatic PotentialT: Lattice TemperatureG, R: Optical Gain/Recombinationf: Lasing FrequencyA: Optical Wave AmplitudeS: PhotonNumberK: Quantum Mechanical Wavefunction
Poisson EquationV, n, p, T, K
Quantum Carrier TreatmentK,n,p,V
Continuity Equationn, p, V, G, R, S, A, T
Temperature EquationT, V, n, p
Complex Wave EquationA, n, p, f, G
Photon Rate EquationS, G, R, f, n, p
Gain ModelG, n, p, f, T
Self-Consistent Solution
ZürichIntegrated Systems LaboratoryOptoelectronics Modeling Group Zürich
Example III:Opto-Electro-Thermal
VCSEL Simulation
• Separate meshes and interpolateelectro-thermal: ~ 10’000 FEoptical: ~ 200’000 FE
• Newton scheme to solve coupled non-linear set of electro-thermal equations.Parallelised direct solver (PARDISO) tosolve linear equations in Newtonscheme
• Photon Rate Equation Accounts forOptics
• Photon Life Time (Imaginary Part ofEigenvalue) as a Variable in the Equa-tion
r � (�r�) = �q�p� n +N+D �N�
A�
�r � S = H + ctot@tT
r � jn = q (R + @tn)
�r � jp = q (R + @tp)
S = ��thrT
jn = �q (�nnr��Dnrn+ �nnPnrT )
jp = �q (�ppr�+Dprp+ �ppPprT )
ddt
S�(t) = (G� � L�)S�(t) +Rsp
ZürichIntegrated Systems LaboratoryOptoelectronics Modeling Group Zürich
2.0x
1e2
2.0x
1e4
2.0x
1e3
[A/c
m2]
300
350
400
[K]
PML
PML
PM
L
contact
top DBR
bottom DBR
GaAs substrate
air
oxide confinement
Simulation Results (Coupled)
Electro-thermo-optical simulation:
➜ Distribution of current density
➜ temperature and optical intensity at 16.3 mA.
ZürichIntegrated Systems LaboratoryOptoelectronics Modeling Group Zürich
current [mA]0 5 10 15 20
op
tica
l po
wer
[m
W]
0
5
10
15
0.0
1.0
2.0
ano
de
volt
age
[V]
anode voltageoptical power
880
885
900
890
895
300
310
320
330
340
350
0 5 10 15 20
wav
elen
gth
[n
m]
aver
age
cavi
ty t
emp
erat
ure
[K
]
current [mA]
average cavity temperaturewavelength
Simulation Results (Coupled)Electro-thermo-optical simulation:
➜ I/P_opt and I/V curve
➜ Wavelength tuning of HE11 mode
➜ average cavity temperature versus laser current.
ZürichIntegrated Systems LaboratoryOptoelectronics Modeling Group Zürich
0 1 2 3 4 5radius [um]
op
tica
l in
ten
sity
(a.
u.)
0 [mA]5 [mA]
15 [mA]10 [mA]
20 [mA]
0 5 10 15radius [um]
loca
l op
tica
l mat
eria
l gai
n [
1/cm
]
2000
-1000
-500
0
500
1000
15001 [mA]3 [mA]7 [mA]
Simulation Results (Coupled)Electro-thermo-optical simulation:
➜ Normalised optical intensity of HE11 mode in active region versus current (thermal lensing)
➜ Optical material gain in active region versus current (spatial hole burning).
ZürichIntegrated Systems LaboratoryOptoelectronics Modeling Group Zürich
All benchmarks performed on Compaq AlphaServer ES45 1250 MHz (1 CPU)
Numerical Performance
Simulation Results (Optical):
• # optical vertices: 90’000
• job size: ~ 2 GBytes
• Total CPU Time Usage: ~3 m
Simulation Results (Coupled):
• # optical vertices: 90’000# electro-thermal vertices: 7’729
• job size: ~ 2.2 GBytes
• Total CPU Time Usage: ~8 h➜ 98 bias points➜ run time without self-consistent
optics: ~5 h
ZürichIntegrated Systems LaboratoryOptoelectronics Modeling Group Zürich
Summary
• Solve Helmholtz Equation➜ Open Resonator Requires Absorbing Boundaries
➜ Non-Hermitian Problem, Complex Eigenvalue
➜ Discretization: Linear vs. Quadratic
• Spectral Portrait➜ Difficult to Find Lasing Modes Automatically
➜ Track Modes for Perturbed Problem
• Applications➜ Tunable Air-Gap Filter
➜ Coupled Opto-Electro-Thermal VCSEL Simulation
ZürichIntegrated Systems LaboratoryOptoelectronics Modeling Group Zürich
Conclusions / Outlook
• Eigenmode Analysis of Microcavity Lasers➜ Perfectly Matched Layer (PML) Boundary Condi-
tion is a Good Choice for Absorbing Boundary
➜ Need Full-Wave Solution to Account for Diffrac-tion Losses and Calculate Photon Life Time Accu-rately
➜ Trade-Off Between Accuracy and Problem Size
• Improve Preconditioner➜ Possibility: First Order Elements for Precondi-
tioner and Second Order Elements for Jacobi Cor-rection Equation (Requires Hierarchical FiniteElements)
ZürichIntegrated Systems LaboratoryOptoelectronics Modeling Group Zürich
Acknowledgements
• Oscar Chinellato, Peter ArbenzInstitute of Scientific Computing, ETH Zurich
University of Kassel
• Avalon PhotonicsHardware Partner
• ISE Integrated Systems EngineeringSoftware Partner
• Swiss Funding Agency➜ Commission for Technology and Innovation (CTI)
➜ TOP NANO21 Programme.