LIGO-G040160-00-Z 1

LIGO

HWS

Lowering Mechanical Loss in Fused Silica Optics with

Annealing

Steve Penn

Alexander Ageev, Garilynn Billingsley, David Crooks, Andri Gretersson, Gregg Harry, Jim Hough,

Sheila Rowan, David Shoemaker, Peter Saulson, Peter Sneddon, Phil Willems

LIGO-G040160-00-Z 2

LIGO

HWSSapphire vs. Fused

Silica

High Q, ≈ 200 million (Willems, MSU, Glasgow)

Higher Young’s Modulus Higher Density Higher Thermal

Conductivity

Higher Optical Absorption High Thermoelastic Loss Less History as an Optical

Material Expensive

High Q (but not consistantly)• 200 million Ageev/Penn - rods• 120 million, Willems - LIGO I optic

Lower Young’s Modulus Lower Density Lower Thermal Conductivity

Lower Optical Absorption Negligible Thermoelastic Loss Extensive History as an

Optical Material Expensive

LIGO-G040160-00-Z 3

LIGO

HWS

Brief History of Silica Research

Research has been conducted over the past several years to understand the fundamental loss mechanism in fused silica and produce extremely low loss FS optics suitable for Advanced LIGO. (Syracuse, Glasgow, Caltech, MSU, HWS)

Experiments have been performed on fiber/rod samples over wide range of sizes reveal a clear surface loss dependence. Loss appears to be entirely in the surface.

For each sample, loss also increases with frequency.

Slowly “annealing” samples can lower loss, but not below the surface loss limit.

BOB

SAMPLE

EXCITER

LIGO-G040160-00-Z 4

LIGO

HWS

Surface Loss & the Effect of Annealing

8 mm fiber Q = 80 million, before annealingQ = 200 million, after annealing

LIGO Test mass, if surface loss limitedQ (predicted) = 2 billion

SURFACE LOSS

LIGO-G040160-00-Z 5

LIGO

HWS

What is the initial Q of LIGO masses?

LIGO-G040160-00-Z 6

LIGO

HWS

Annealing: Benefits & Challenges

Annealing can greatly lower the mechanical loss for samples above the surface loss limit, including superpolished samples.

Loss reduction from annealing depends on the peak temperature and the cool down rate. This parameter space should be explored, but doing so with high Q samples is very time consuming.

Low temperature anneals (600° C) yield large decrease in loss (≈ 10) for superpolished samples. (Standard Anneal temp. ≈ 11,000° C)

Cool down rate is geometry dependent. We may be able to increase rate. Otherwise the annealing could be quite long for Adv. LIGO masses.

Annealing could change surface figure, optical absorption, or silicate bonding to support structure.

LIGO-G040160-00-Z 7

LIGO

HWS

S312

S312SV

Q Dependence on Silica Type?

Preliminary measurementsshow a factor 3 difference between S312 & S312SV

LIGO-G040160-00-Z 8

LIGO

HWS

Differences between S312 & S312SV

Heraeus has provided limited insight into the differences between the Suprasil families: (S1, S2, S3), (S311, S312, S313), (S311SV, S312SV, S313SV).

» Manufacturing processes differ for each family (no details).

» No significant composition difference except for OH content.

» OH level affects the fictive temperature of the glass such that lowering OH raises the fictive temperature.

» Annealing temperature scales with fictive temperature. Heraeus suggests that a change in annealing temperature from 950 C upto 1050 C could be significant for S312SV.

LIGO-G040160-00-Z 9

LIGO

HWSThe Abbreviated

Silica Research Plan

Samples are in limited supply for our tight timescale and tight budget. We need to leapfrog using the few existing extra samples.

“Optimize” annealing procedure on mid-sized uncoated optics by testing an annealing curve scaled by sample geometry, x 1/3

Test Peak temperature by lowering peak temp by 200 C for 1 run.

Changes in surface figure is presently ignored though it must be known before we can anneal larger optics that we do not wish to damage.

The geometries of optics gathered for testing will allow test of predicted surface loss limit, frequency dependence and bulk Q.

LIGO-G040160-00-Z 10

LIGO

HWS

Surface Loss» Water adsorption (Braginsky, many others)» Alkali absorption (Marx and Sivertsen)» Degenerate States for Surface Oxygen Bonds (Bartenev)» Microcracking

Additional Loss» Additional loss for V/S > 1 mm to be stress-induced loss arising

from larger thermal gradients during manufacturing. Annealing shown to decrease of stress-related loss (Numata, Lunin, Harry, Penn)

Bulk loss» Bulk loss at 400 Hz estimated as 2.5 x 10-9 (Q = 4 x 108)

(Wiedersich, et al., Roessler group)– Extrapolation down from the GHz regime.– Loss arises from Asymmetric double-well potential,

Theories of Silica Loss

Q 1 A f , 0.77

LIGO-G040160-00-Z 11

LIGO

HWS

Surface Loss» “Constant” Surface loss:

Additional Loss: stress, adsorbed impurities/water» Anneal has been shown to bring loss to or within “a few” of surface limit» At large geometries this loss is very low < 5e-9

Bulk loss» Loss » Bulk loss extrapolated from GHz regime down to LIGO frequencies,

estimated as 1–2.5 x 10-9 (Q = 0.4–1 x 109) (Wiedersich, Roessler et al.,)

Estimating Silica Loss

S f 2kBT 3 2 f

1 2

wYsubstrate

2

1 2 1

d

wsurface

dsurface 5.210 12

Q 1 A f , 0.77

LIGO-G040160-00-Z 12

LIGO

HWS

LIGO-G040160-00-Z 13

LIGO

HWS

LIGO-G040160-00-Z 14

LIGO

HWS

LIGO-G040160-00-Z 15

LIGO

HWS

LIGO-G040160-00-Z 16

LIGO

HWS

LIGO-G040160-00-Z 17

LIGO

HWS

Q = 120 million, NSB Range = 185 Mpc

LIGO-G040160-00-Z 18

LIGO

HWS

Q = 200 million, NSB Range = 196 Mpc

LIGO-G040160-00-Z 19

LIGO

HWS

Q = 600 million, NSB Range = 210 Mpc