Si-EPIC Workshop: Silicon Nanophotonics Fabrication Fibre Grating Couplers June 30, 2012 Dr. Lukas Chrostowski
Outline Coupling light to chips using Fibre Grating Couplers (FGC, or GC). Grating coupler physics Tutorial on modelling 2
lower than -40 db [5] that is ignored in our simulation. calculated as functions of the coupling coe cients. The waveguide effective indices are calculated by a 2D finite-difference mode solver [9]. The coupling coe cients of the directional couplers using coupled390 Caryare Gunncalculated and Thomas L. Koch Fiber input Fiber output mode theory [10]. w Coupling coe cients are critical to the performance of microring resonators [2] and, as shown in Fig. 2, signifisi SiO magnitude, cladding cantlyl affect the shape, Q factor, and extinc2!m Input port tion ratio of the reflection spectrum. Therefore, we use h Are used to couple light in/out of the chip a racetrack shape to carefully control the coupling coefw via the top Buried SiO coupling condition, we scan ficients. To find the optimal Through port the reflection spectrum as a function of 12 and 34 and Input coupler similar electrical pads calculate Rp Rvtoand 10log(Rp /R v ), where Rp is the Si substrate maximum reflectivity and Rv is[28]. Reflection port Figure 11.7 typically Schematic of waveguide 10x10 with nano-taper coupler umthe minimum reflectivity. Based on the results shown in Fig. 3, we choose 12 Heat sink photonics [31, 32] have a number of unique traits a relatively small number of to be 0.84 and to be 0.77 in order to satisfy the dual34 tocan have chip grating teeth, a desire mode-match to the hundreds Gaussian output of a per single-mode Temperature controller fiber, and criteria a remarkably of widehigh optical reflectivity bandwidth. and high extinction ratio. A silicon-grating coupler exploits the high index contrast between silicon and The was fabricated by epixfab at IMEC using silicon dioxide, as welldevice as the sub-wavelength patterning capabilities of a modern DUV lithography process, to create a grating capable of creating a well-controlled schematic is 193 nm photolithography. The measurement Fig. 4. Measurement schematic with an inset showing optical mode with lateral dimensions equivalent to the core of an optical fiber, or shown in Fig. 4. Periodic grating couplers [11] are used to!10 mm in length. An example of a grating coupler is shown in Figure 11.8. an image of the Y-branch power splitter. couple light into and out of the waveguides. A Y-branch www.download-it.org/learning-resources.php?promocode=&partnerid=&content=story&storyid=1900 e reference: t9svae7agovn5h9q6o7tp6cpk6-2587, for 1 user on Jun 19, 2010 to Customize Your Account lukasc@gmail.com Grating Couplers w 2 t In pu tl ig ht t 2 etp://www.download-it.org/learning-resources.php?promocode=&partnerid=&content=story&storyid=1900 reference: t9svae7agovn5h9q6o7tp6cpk6-2587, for 1 user on Jun 19, 2010 to Customize Your Account lukasc@gmail.com 2 Luxtera Inc. Figure 11.8 Oblique view of a planar grating coupler structure employing curved gratings, nonuniform pitch, and non-uniform periodicity. This particular device is formed using curved polysilicon gratings formed on top of a planar silicon slab which is tapered down to a single mode silicon waveguide. Image courtesy Luxtera, Inc. r 11 Silicon Photonics From Optical Fiber Telecommunications - V Ivan P. Kaminow, Tingye Li and Alan E. Willner 3
Grating Fiber Coupler single-mode fibre, 10 adiabatic taper (>150µm) TE to integrated circuit grating 10µm wide waveguide intec 2008 UBC EECE - Photonics 584 / CMC Workshop Research Silicon Group Nanophotonics - http://photonics.intec.ugent.be Fabrication 4
1-D grating coupler Experimental results (λ=630nm,depth=70nm, TE pol.) 31 % efficiency (5.1 db coupling loss) 40nm 1dB bandwidth Also acts as a broadband filter shallow grating deep trench intec 2008 UBC EECE - Photonics 584 / CMC Workshop Research Silicon Group Nanophotonics - http://photonics.intec.ugent.be Fabrication 5
2-D grating + polarization splitter 10 Fiber-to-waveguide interface for polarization independent photonic integrated circuit 2-D grating, 2 waveguides couples each fiber polarization in its own waveguide TE in the waveguides the polarization is the same (TE) patented TE intec 2008 UBC EECE - Photonics 584 / CMC Workshop Research Silicon Group Nanophotonics - http://photonics.intec.ugent.be Fabrication 6
single-mode fiber Polarisation Diversity Circuit light in y x on-chip components are polarisation dependent fiber-to-fiber transmission is polarisation independent light out x-polarization 2-D grating split polarisations identical circuits y-polarization z y x combine polarisations patented 2-D grating intec 2008 UBC EECE - Photonics 584 / CMC Workshop Research Silicon Group Nanophotonics - http://photonics.intec.ugent.be Fabrication 7
Measurement!"#$%&'(# )*+,-./#0,1/2#3#4/5/6572# ;G98# ;,E/2,# G=627167>/# (H#0/+1/F# "G#I-/2#!7.F/2#!7.F/2# ;7E>*5/2# G7+=572# )?;# @AB#C,+7>71=D7+/2# 89:#;<=># )/E>/2,5*2/# ;7+527../2# 8
Semi-Automated Optical Probe Station 2 fibres All-band Agilent tunable laser (180 nm span, 1460-1640 nm) 1220 nm tunable laser Grating Grating 12.5 um 25 um 50 um Grating Grating Grating Grating 4 mm 9
Automated Probe Station Fibre Array 10
11
Automated Probe Station Fibre Array 127 µm fibre spacing 2 separate devices 127 µm 127 µm 12
13
14
15
16
FDTD Tutorial Grating Couplers
Grating Coupler Operation Case 1 Optical wavelength inside the grating matches its period, 0 n e = K = 2 = n e k 0 = 2 n e Vertical output (1 st diffraction order), plus back-reflection (from 2 nd diffraction order) 0 18
Grating Coupler Operation Detuned Case 2 Optical wavelength is smaller than the grating period, 0 n e < K = 2 = n e k 0 = 2 n e Vertical output at an angle, no 2 nd order back-reflection 0 19
Grating Coupler Bragg Condition =sin 1 k x k 0 n 1 =1 y k x = k 0 = 2 0 mk Grating, m=2 = n e k 0 = 2 n e waveguide propagation constant 0 x 2K =2 2 K = 2 n 2 = n SiO2 Grating, m=1 20
Gratings Bragg condition Bragg condition Grating s scattering modifies the light s wave-vector to be (in the direction of propagation, x): Slab s effective index in the region of the grating: for 220 nm thick ~ 2.875 for 150 nm thick ~ 2.574 k x = mk = m 2 Duty cycle is 50%, thus estimate average effective index to be 2.724 k x = K = 2 2.724 1.55µm 2 0.63µm = 11.04 9.97 = 1.069µm 1 21
Gratings Bragg condition We know the free-space wave-vector: k 0 = 2 0 =4.05µm 1 Estimated diffracted angle is: =sin 1 k x k 0 =sin 1 1.069 4.05 = 15.3 22
Detuned second-order gratings: A first generation of gratings was etched 40 to 50nm deep, with a 610 nm pitch and uniform 50% fill factor. These have a coupling efficiency of about 20% and a 60 nm 3 db bandwidth [16] without index matching material between grating and fibre. The second generation of couplers used has a 70 nm etch depth, a 630 nm pitch and a higher coupling efficiency of up to 35% when cladded with oxide, with an almost 60 nm wide 3 db bandwidth. [Pieter Dumon thesis] 23
Source: Dirk Taillaert, PhD Thesis, IMEC 24
Dirk Taillaert, PhD Thesis, IMEC 4.2.1 Vertical coupling Source: Dirk Taillaert, PhD Thesis, IMEC The case of vertical coupling ( =0) is very interesting from a practical point of view. Vertical coupling can be achieved when the grating period equals the wavelength divided by the refractive index. For a very shallow grating, this index is the effective index of the waveguide mode. As mentioned in chapter 3, this grating is called a second order grating. But for the grating coupler, the first order diffraction is used. The second order diffraction is reflecting back into the waveguide. To avoid any confusion, we will use the term coupler grating instead of second order grating in the rest of this work. Figure 4.2 shows the reflection R as a function of wavelength for dif- 4.2.2 Almost vertical coupling To avoid the reflection at the grating, we have to choose a working point away from the second order reflection peak. Either a shorter or longer wavelength can be chosen. As a result, light is coupled out not exactly vertical, but at a small angle with respect to the vertical direction. This grating is also called a detuned grating. Instead of changing the wavelength, the grating period can be changed. The grating can be negatively or positively detuned (figure 4.5). In a negatively detuned grating, the grating period is smaller (K is larger) or the wavelength is longer ( is smaller) compared to the case of vertical coupling. In a positively detuned grating, the grating period is larger ( is smaller) or 25
Source: Dirk Taillaert, PhD Thesis, IMEC 0.6 0.5 power up fibre 10 reflection 0.6 0.5 power up fibre 8 reflection 0.4 0.4 0.3 0.3 0.2 0.2 0.1 0.1 1500 1550 1600 1650 wavelength (nm) (a) with air on top 1500 1550 1600 1650 wavelength (nm) (b) with index matching layer Figure 4.8: Calculated coupling efficiency to fibre for an optimized uniform grating and near vertical coupling. =630 nm, ed=70 nm, ff=0.5, N=20, 26
Experimental Alignment Tolerances Source: Dirk Taillaert, PhD Thesis, IMEC 4 3 1dB z offset (µm) 2 1 0 1 2 3dB 1dB 0.5dB 0.5dB 1dB 3dB 3 4 3 2 1 0 1 2 3 4 x offset (µm) (b) measurement results Figure 6.9: Experimental alignment tolerances. 27
Resources - Grating Couplers Book chapter: David J. Lockwood and Lorenzo Pavesi, Silicon Photonics II Components and Integration, 2011, Online PDF Chapter 3 Interfacing Silicon Nanophotonic Integrated Circuits and Single-Mode Optical Fibers with Diffraction Gratings (IMEC) Chrostowski and Hochberg, Silicon Photonics Design, Ch. Optical I/O Thesis: Dirk Taillaert, PhD Thesis, IMEC Journal papers: Luxtera: A. Mekis et al. A Grating-Coupler-Enabled CMOS Photonics Platform. IEEE Journal of Selected Topics in Quantum Electronics, 17.3 (2011), pp. 597 608. issn: 1077-260X. doi: 10.1109/JSTQE.2010.2086049 28
Grating Coupler Modelling Approach: 1) Waveguide to air 2D FDTD Start with mode-source in the waveguide, measure output power in free-space far-field, check angle 2) Air to waveguide 2D FDTD Start from optical fibre Gaussian mode incident on grating. Measure power in the waveguide use previous angle, vary position Fibre mode MFD 10.5 3) optimize for 1550 nm 4) validate design 3D FDTD 29
IMEC Coupler Oxide 2 µm Cladding 2 µm check? Silicon 0.22 µm Period 0.63 µm Fill 0.32 µm (each tooth) Etch 70 nm (150 nm remaining) Oxide index 1.444 Si index: 3.47 constant, vs. Palik data dispersive? 30
1) Output Grating Coupler simulation Launch a mode in the slab Monitor the output far-field pattern vs. angle (for a specific wavelength) 31
3D FDTD Fibre Grating Coupler object 32
Setup 33
Far field projection Peak angle is between 10-20º (wavelength dependant) 34
Total output power 1545 nm peak wavelength Power out 56% 35
e.g., 20 15 Grating coupler, 2D FDTD simulation, accuracy=5 y Position [um] 10 5 0 0 2 4 6 8 10 12 14 16 x Position [um] 36
2) Input Grating Coupler simulation Gaussian beam input (waist diameter 10.5 µm) Measure transmission spectrum into slab waveguide Sweeps: angle, position 37
Setup 38
Setup FDTD 7: FDTD 8: 39
Setup Movie Move the movie monitor into simulation region. Setup script is automatically executed before simulation 40
Run Run simulation Run analysis script runanalysis; 41
Runanalysis; # Transmission at 1550 nm lambda=c/getdata("coupled","f"); Tspectrum=transmission("coupled"); T_1550=interp(Tspectrum,lambda,1550e-9); # Spectrum plot ( lambda, Tspectrum,"Wavelength","Power Coupled", "Grating coupler efficiency"); plot ( lambda, 10*log10(abs(Tspectrum)),"Wavelength","Power Coupled", "Grating coupler efficiency"); setplot("x min", 1.5e-6); setplot("x max", 1.6e-6); setplot("y min", -15); setplot("y max", 0); setplot("y label", "Coupling, db"); Tspectrum1=transmission("below"); plot ( lambda, 10*log10(abs(Tspectrum1)),"Wavelength","Power Coupled", "GC - into substrate"); setplot("x min", 1.5e-6); setplot("x max", 1.6e-6); setplot("y min", min(10*log10(abs(tspectrum1)))-1); setplot("y max", 0); setplot("y label", "db"); Tspectrum1=transmission("above"); plot ( lambda, 10*log10(abs(Tspectrum1)),"Wavelength","Power Coupled", "GC - refection"); setplot("x min", 1.5e-6); setplot("x max", 1.6e-6); setplot("y min", min(10*log10(abs(tspectrum1)))-1); setplot("y max", 0); setplot("y label", "db"); Tspectrum1=transmission("backwards"); plot ( lambda, 10*log10(abs(Tspectrum1)),"Wavelength","Power Coupled", "GC - backwards, waveguide"); setplot("x min", 1.5e-6); setplot("x max", 1.6e-6); setplot("y min", min(10*log10(abs(tspectrum1)))-1); setplot("y max", 0); setplot("y label", "Coupling, db"); # http://docs.lumerical.com/en/fdtd/cmos_angular_response.html 42
Simulation Results 43
Where is the coupling loss from? 44
Optimization Optimize the gaussian beam angle the position of the beam Done either with optimization or with sweep 45
Sweep Angle 46
After it is done, run the script to analyze all the data. # plot the results from an FDTD sweep on position. angles=getsweepdata("sweep, angle","angle"); lambda=c/getdata("coupled","f"); NUM=length(angles); T_data=getsweepdata("sweep, angle","tspectrum"); plot ( lambda, T_data,"Wavelength","Power Coupled", "Gaussian angles: "+num2str(angles(1))+" to " +num2str(angles(num))); legend (num2str(angles(1))); setplot("y label", "Coupling"); plot ( lambda, 10*log10(abs(T_data)),"Wavelength","Power Coupled", "Gaussian angles: "+num2str(angles(1))+" to " +num2str(angles(num))); legend (num2str(angles(1))); setplot("x min", 1.5e-6); setplot("x max", 1.6e-6); setplot("y min", -15); setplot("y max", 0); setplot("y label", "Coupling, db"); T_1550=getsweepdata("sweep, angle","t_1550"); plot ( angles, 10*log10(T_1550), "Coupling @ 1550"); setplot("x label", "Angle"); setplot("y label", "Coupling, db"); matlabsave("gc_in_sweep,angle",t_data,angles,lambda); # http://docs.lumerical.com/en/fdtd/cmos_angular_response.html 47
Gaussian input Angle Mesh accuracy = 2 (auto mesh, conformal) About 5-10 nm per degree tuning 48
Gaussian Input Angle Mesh accuracy = 4 (auto mesh, conformal) 49
Sweep Position 50
Gaussian Input Position 51
Sweep Buried Oxide Thickness Sweep the oxide thickness Achieved by overlapping the oxide on top of the silicon substrate, and changing the y-min of the oxide. 52
Oxide thickness 2 µm Oscillations as a function of thickness are a result of constructive/destructive interference from the oxide layer. 53
Optimized with 10 nm mesh-x 16º injection angle, optimized laterally: 10 µm peak is 1.547 nm -3.45 db coupling efficiency with auto-mesh, accuracy=4 54
Sweep Mesh accuracy Convergence test 55
Convergence Test Without Mesh Override 56
Convergence Test Using Mesh Override Conclusion: Slightly faster convergence 57
Sensitivity to Accuracy Using mesh overrides to ensure correctly-periodic mesh Error is < 10 nm, ~0.01 coupling error 58
Manual mesh Mesh override: 10 nm grid in the waveguide & grating 59
3D FDTD Grating Coupler 60
3D FDTD Grating Coupler Layout imported from GDS FDTD simulation region includes substrate and cladding: -2.4 µm < z < 3.0 µm Gaussian beam input, above the oxide Beam centre offset 5 µm from 1 st grating tooth. Power monitors: in the taper (faster simulation time) in the waveguide (include taper in the simulation, 2X longer) 61
GDS Grating coupler IMEC 62
GDS Import GDS: TE_Curved_Grating_coupler_right Layer 52, Silicon, edit to be 220 nm Layer 75, Silicon, edit to be 150 nm 63
3D FDTD Grating Coupler Mesh accuracy = 1 (several minutes) Mesh accuracy = 2 (several tens minutes) 64
3D FDTD Grating Coupler Core i7 ~$1000 linux Accuracy Time hh:mm:ss increasing accuracy 1 00:03:30 2 00:15:40 3 00:47:00 4 01:56:00 5 03:46:00 6 07:27:00 65