# waveguide analysis

TRANSCRIPT

Advanced Silicon Photonics Waveguide Analysis

Factors affecting the propagation of an optical mode in a waveguide

Charles Baudot

24th of July 2014

Simulation Method 2

• The device is made up of silicon (Palik).

• The device surroundings is defined as SiO2 (Palik).

• The device is simulated using 3D FDTD.

• The simulation boundary is made up of a 16 layer PML.

• Mesh refinement:

• 2nm for the waveguide width and length,

• 10nm for the waveguide height.

• The fundamental TE mode is injected in the waveguide.

• Monitors are placed at the waveguide tips and the differential intensity

is measured.

Model

• Device architecture:

3

Simulation region bounded by PML

I/O Monitors

Source

Strip Waveguide

Measured PML

tip reflectivity

= - 76.81 dB

Roughness introduced on

waveguide sidewalls

Roughness Introduction

• Model parameters:

• Roughness period

• Maximum protrusion

• Maximum cavity

(Roughness is introduced only on sidewalls. Top and bottom layers are perfect)

4

Period

Max. Protrusion

Max. Cavity

Top view of waveguide sidewall

TE Mode

Propagation

• Model Routine:

• For each Roughness Period, issue a random number

between [-1,1]

• If random number > 0,

• random bias = random number X Max. Protrusion

• else

• random bias = random number X Max. Cavity

• Waveguide boundary = (nominal width / 2) + random bias

• Operation is done independently on each sidewall to avoid

periodicity

Simulated Example

• Parameters:

• Roughness Period = 4nm

• Max. Cavity = 2nm

• Max. Protrusion = 2nm

5

Top view of generated sidewall* Top view of meshed sidewall*

*Pic

ture

s n

ot ta

ken a

t th

e s

am

e s

idew

all

location

Intrinsic Loss

• Strip waveguide

• Height = 306nm

• Width = 350nm

• Wavelength = 1310nm

• Intrinsic Insertion Loss = 1.97 dB/cm

• Reflected fraction for 1µm propagation = -39 dB

6

2µm

2µm

350nm

306nm

Waveguide

Simulation Region

Cavity vs. Protrusion Analysis

• Roughness Period = 4nm

• Max. Cavity = 2nm

• Max. Protrusion = 0nm

7

• Roughness Period = 4nm

• Max. Cavity = 0nm

• Max. Protrusion = 2nm

• Roughness Period = 4nm

• Max. Cavity = 2nm

• Max. Protrusion = 2nm

• Insertion Loss = 2.45 dB/cm • Insertion Loss = 2.12 dB/cm • Insertion Loss = 2.32 dB/cm

It appears that notches have a bigger impact than protrusions on insertion loss

A combination of the two defects gives a median effect

Cavity Depth Analysis 8

• We observe that cavities of 1nm result in an insertion loss smaller than the intrinsic one. This phenomenon is

to be understood.

• We also notice a non-monotonous trend for an increase in notch depth. This may be explained by a periodic

property that should be determined.

• Frozen parameters

• Roughness Period = 4nm

• Max. Protrusion = 0nm

• Variable

• Max. Cavity = [0, 4]nm

1

2

3

4

5

0 1 2 3 4 5

Tra

nsm

issio

n L

oss

(dB

/cm

)

Defect size (nm)

Protrusion Size Analysis 9

• Frozen parameters

• Roughness Period = 4nm

• Max. Cavity = 4nm

• Variable

• Max. Protrusion = [0, 8]nm

• For a notch size and a period fixed at 4nm, we observe that a protrusion size of 4nm results in a sudden rise in

propagation loss.

• The inter-relation between notches and protrusions seems conclusive.

2

3

4

5

6

7

8

9

10

0 2 4 6 8 10 12 14T

ran

sm

iss

ion

Lo

ss

(d

B/c

m)

Defect size (nm)

Periodicity Analysis 10

• Frozen parameters

• Max. Protrusion = 2nm

• Max. Cavity = 0nm

• Variable

• Roughness Period = [4, 23]nm

• By applying only one kind of defect and varying the repetition period, we observe that there is indeed a trend

indicating that there may be a periodic effect in the losses

1

2

3

0 5 10 15 20 25Tra

nsm

issio

n L

oss

(dB

/cm

) Roughness Period (nm)

Mode Propagation

• Fundamental TE mode propagation in rough waveguide

11

• Fundamental TM mode propagation in rough waveguide

We observe that the TE mode is strongly dependent on the sidewall roughness

The TM mode on the other hand, oscillates parallel to the defects

Roughness Profile Comparison

• Does the model replicate consistently process conditions?

12

Model Experimental observation

• Experimental observations show that during silicon patterning a non-linear outline is transferred to the

sidewall all the way down to the etch limit. This may be due to the granularity of the deposited hard mask

or the grain size of the polymer constituting the photoresist.

• Thus, the model is a good replica of process conditions.

Impact of BEOL on propagation loss

• Simulation of a waveguide with metal on top of it

13

Multimode rib

waveguide

306nm

165nm

1.2µm

Roughness on the

top part of the rib

Patterned metal

Metal Patterning

• Full metal layer

14

• Parallel pattern • Perpendicular pattern

• Insertion Loss = 0.09 dB/cm • Insertion Loss = 0.22 dB/cm • Insertion Loss = 3.93 dB/cm

• We observe that non-patterned metal results in no signal loss whereas patterned metal has a big impact on the

propagation loss. Depending on the height and pattern parameters (line width, line space), the consequence may be

dramatic as shown for the perpendicular pattern.

• Frozen parameters

• Distance between metal and top silicon = 0.694µm

• Metal pattern: Line 200nm / Space 200nm

• Variable

Note: Fundamental TE mode injected in waveguide

Metal Patterning 15

• Even at 1.1µm a patterned metal layer will have an impact on the propagation loss of a waveguide.

• The effect of roughness is small for a multimode waveguide.

• It should be noted that the larger the waveguide, the bigger will be the interaction with the top metal layer

• Frozen parameters

• Perpendicular pattern

• Metal pattern: Line 200nm / Space 200nm

• Variable

• Metal distance to top silicon

• Silicon roughness

Roughness Space (µm) Insertion Loss

(dB/cm)

Yes 0.694 3.93

yes 1.094 2.77

No 1.094 2.68