Length Sensing and Control for an Advanced Gravitational Wave Detector Robert Ward PhD Candidacy Caltech, 17 Jan 2006 Jan 17th 2006 Ward Candicacy 1 introduction lock acquisition modeling DC readout Jan 17th 2006 Ward Candicacy 2 A New Window on the Universe • Once Gravitational Waves are detected, a new field of Gravitational Wave Astronomy will open up. – GW stochastic background can tell us about cosmology (Big Bang, Inflation) – Cosmic Strings – Compact binary inspirals – GRBs – Supernova collapse – Black Hole ringdowns • GW Astronomy will allow us to listen to what we cannot see. Jan 17th 2006 Ward Candicacy 3 The Michelson Interferometer as a Gravitational Wave Detector Gravitational Waves act on freely falling masses: mirrors laser Beam splitter Dark port photodiode Antenna pattern Suspend the masses Jan 17th 2006 Ward Candicacy 4 Upgrading the Michelson Gravitational Waves are tiny: they interact very weakly with matter. Need more than a simple michelson to have a chance of detection. LASER OOM 1 km arms 10W NdYAG h = 10^-21 1 photon Jan 17th 2006 GWD Ward Candicacy 5 Upgrading the Michelson • Fabry-Perot Arm Cavities (Like having longer arms) 2 Ga 1 rf GWD Jan 17th 2006 Ward Candicacy 6 Upgrading the Michelson • Fabry-Perot Arm Cavities (Like having longer arms) 2 Ga 1 rf • Power Recycling (Like having a bigger LASER->lower shot noise) 1 Gr 1 rr rarm GWD Jan 17th 2006 Ward Candicacy 7 Upgrading the Michelson • Fabry-Perot Arm Cavities (Like having longer arms) 2 Ga 1 rf • Power Recycling (Like having a bigger LASER) 1 Gr 1 rr rarm • Signal Recycling (Reshape G g( f ) the SIGNAL) Jan 17th 2006 Ward Candicacy GWD 8 Why Signal Recycle? Why RSE? • • • Problem 1: If the current Initial LIGO optical configuration (power-recycled Michelson with Fabry-Perot arms) is retained in AdLIGO, the increased laser power (needed for better sensitivity in the high-frequency shot-noiselimited regime) will put intolerable thermal load on the transmissive (absorptive, lossy) optics in the power recycling cavity (BS, ITM substrates). Solution 1: increase the finesse (optical gain) of the F-P arms, decrease the gain in the PRC. Problem 2: Increasing the finesse of the arms causes the cavity pole frequency to decrease, leading to reduced Laser bandwidth for GW signal. Solution 2: resonant sideband subtraction! Jan 17th 2006 FP cavity • PRM Ward Candicacy Power FP cavity BS GW signal SEM 9 The Reason for AdLIGO: Initial and Advanced LIGO • Factor 10 better amplitude sensitivity – (Reach)3 = rate • • Factor 4 lower frequency bound NS Binaries: for three interferometers, – Initial LIGO: ~20 Mpc – Adv LIGO: ~300 Mpc • BH Binaries: – Initial LIGO: 10 Mo, 100 Mpc – Adv LIGO : 50 Mo, z=2 • Stochastic background: – Initial LIGO: ~3e-6 – Adv LIGO ~3e-9 Jan 17th 2006 Ward Candicacy 10 Improvement of reach with Advanced LIGO Improve amplitude sensitivity by a factor of 10x, and… Number of sources goes up 1000x! Virgo cluster LIGO I Jan 17th 2006 Ward Candicacy AdLIGO 11 AdLIGO noise curve Fight the Fundamental Noise Sources: 1) Seismic 2) Thermal 3) Quantum Quantum noise Seismic noise Suspension thermal noise Silica Brownian thermal noise Coating Brownian noise (1/f) Gravity Gradients Total noise -21 10 Bench Active Seismic Isolation External Seismic PreIsolation Quadruple pendulum suspensions 40 kg, fused silica Test Masses 125W Laser Strain [1/ Hz] -22 10 -23 10 -24 10 1 2 10 Jan 17th 2006 3 10 Ward Candicacy 10 Frequency [Hz] 12 Caltech 40 meter prototype interferometer Objectives • Develop a lock acquisition procedure for suspended-mass detuned RSE interferometer with power recycling, preferably one that will be applicable to Advanced LIGO PRM Characterize and optimize optical configuration (for robust control and sensitivity) Bright port Characterize noise mechanisms Develop DC readout scheme Test QND techniques Extrapolate to AdLIGO via simulation Prototyping will yield crucial information about how to build and run AdLIGO Jan 17th 2006 BS SRM Dark port X arm Y arm Ward Candicacy 13 Bench: 40m Sensitivity -17 10 40 Meter Strain Sensitivity -18 h(f) [1/ Hz1/2] 10 m ic Seis Not very likely that we’ll actually detect any gravitational waves here, but hopefully we’ll learn some things about operating interferometers, especially about the quantum noise. -19 10 Total noise -20 10 quantum si en sp Su Tes t M -21 10 on Residual Gas as s In ter nal Therm al m er Th Bench al -22 10 1 10 2 3 10 10 4 10 f [Hz] Jan 17th 2006 Ward Candicacy 14 40m DARM Optical Plant UGF The 40m operates in a detuned RSE 190 Mag (dB) configuration, which gives rise to two peaks in the DARM transfer function: 210 170 150 130 180 1) Optical Resonance 2) Optical Spring Phase (deg) 90 0 -90 -180 0 10 1 10 2 3 10 10 4 10 5 10 Frequency (Hz) Jan 17th 2006 Ward Candicacy 15 Detune Cartoon •IFO Differential Arm mode is detuned from resonance at operating point IFO DARM/CARM 500 200 0 slope related to spring constant 100 SRC LSB -10000 USB -5000 0 5000 10000 frequency offset from carrier [Hz] •Responses of GW USB and GW LSB are different due to the detuning of the signal recycling cavity. Jan 17th 2006 DARM •IFO Common Arm mode is detuned from resonance at intial locking point fsig 50 0 0 FWHM Carrier frequency Sideband amplitude [a.u.] 1000 Ward Candicacy 0 0 PRC CARM 16 Signal Extraction Scheme Carrier -f2 -f1 • Single demodulation • Arm information f1 f2 PRM • Double demodulation • Central part information • • • Arm cavity signals are extracted from beat between carrier and f1 or f2. Central part (Michelson, PRC, SRC) signals are extracted from beat between f1 and f2, not including arm cavity information. Only +f2 sideband resonates in combined PRC+SRC Jan 17th 2006 Ward Candicacy 17 5 DOF for length control Signal Extraction Matrix (in-lock, DC) 40m ETMy Phase Modulation f1=33MHz f2=166MHz Ly=38.55m Finesse=1235 Port Dem. Freq. L L l l ls SP f1 1 0 -0.001 0 0 AP f2 0 1 0 0.001 0 SP f1 f2 -0.002 -0.001 1 -0.032 -0.100 AP f1 f2 -0.001 0.002 0.750 1 0.070 PO f1 f2 0.004 0.003 0.460 -0.023 1 ITMy PRM Laser ly lx lsy BS ETMx Lx =38.55m Finesse=1235 lsx SRM PO SP ITMx Common of arms : L=( Lx Ly) / 2 Differential of arms : L= Lx Ly Power recycling cavity : l=( lx ly) / 2 =2.257m Michelson : l= lx ly = 0.451m Signal recycling cavity : ls=( lsx lsy) / 2 =2.15m AP Jan 17th 2006 Ward Candicacy 18 Lock Acquisition Jan 17th 2006 Ward Candicacy 19 What does it mean to be locked? • GW IFOs are actively-nulled instruments with narrow linear operating ranges. – Locked: All degrees of freedom are within linear operating range, held there by an active control system -3 4 x 10 40M single arm cavity: = 1200 → less than 0.1% of available space offers good control signals 2 0 less than 1 nm 1 ms at 1µm/s -2 -4 0 45 90 135 180 Pound-Drever-Hall error signal for a single cavity Jan 17th 2006 Ward Candicacy 20 Lock Acquisition • Gravitational Wave Interferometers do not come ‘ready to use’ – Natural state is totally uncontrolled (with nonlinear, heavily coupled signals) • Lock Acquisition is the process by which an IFO is brought from an uncontrolled state to the controlled operating point. – Should be considered during the DESIGN phase of an IFO • Money = commissioning + runtime – Can have a very large impact on duty cycle • duty cycle = events Jan 17th 2006 Ward Candicacy 21 From a Bunch of Swinging Mirrors to a Gravitational Wave Detector • AdLIGO will be much harder to lock than LIGO-1 – 4 DOFs to 5 DOFs + SRM scramble – factor 10000 smaller actuation potential – all signals come with offsets • Prototyping can address: – Bootstrapping problem – LIGO I set itself a difficult problem by deciding to lock ONLY at the operating point. • It’s better to cheat (offsets, misalignments, etc). – GOAL: A robust, reliable, and easily diagnosable LA procedure. • Less time spent locking = more time for science! Jan 17th 2006 Ward Candicacy 22 40m Lock Acquisition part I: Off-resonant lock scheme for a single cavity Transmitted light is used as Resonant Lock 1 offset Transmitte d power Off-resonant Lock point 10x higher finesse than LIGO Jan 17th 2006 Ward Candicacy 23 40m Lock acquisition procedure (v 1.0) Start with no DOFs controlled, all optics aligned. ITMy 166MHz ITMx 13m MC BS 33MHz PRM SP33 PO DDM SRM SP166 SP DDM AP166 AP DDM Jan 17th 2006 Ward Candicacy 24 40m Lock acquisition procedure (v 1.0) 1/sqrt(TrY) DRMI + 2arms with offset Average wait : 3 minute (at night, with tickler) ITMy 166MHz ITMx 13m MC 33MHz BS 1/sqrt(TrX) PRM T =7% SP33 SP166 I SP DDM Q SRM T =7% PO DDM AP166 AP DDM Jan 17th 2006 Ward Candicacy 25 40m Lock acquisition procedure (v 1.0) Short DOFs -> DDM DARM -> RF signal CARM -> DC signal 1/sqrt(TrX)+ 1/sqrt( TrY) CARM -> Digital CM_MCL servo + ITMy 166MHz -1 DARM + ITMx 13m MC 33MHz CARM BS PRM SP33 SP166 SP DDM PO DDM SRM To DARM AP166 AP DDM Jan 17th 2006 Ward Candicacy AP166 / sqrt(TrX+TrY) 26 40m Lock acquisition procedure (v 1.0) Reduce CARM offset: 1. Go to higher ARM power 2. Switch on AC-coupled analog CM_AO servo, using REFL DC as error signal. 3. Switch to RF error signal (POX) at half-max power. 4. Reduce offset/increase gain of CM_AO. -1 DARM ITMy 166MHz ITMx 13m MC BS SP166 33MHz PRM PO DDM SRM SP33 SP DDM REFL To DARM AP166 AP DDM Jan 17th 2006 Ward Candicacy AP166 / (TrX+TrY) 27 DARM TFs as CARM offset is reduced Jan 17th 2006 Ward Candicacy 28 Other Lock Acquisition Schemes Alternative Locking Schemes are on the way! • Deterministic Locking: – Locking occurs in stages, with each stage having robust control – Each stage can (and should) lock on the first ‘fringe’, or be robust to fringes. – Transitions between stages are smooth and robust. • Advantages: – Easier to diagnose problems – Should require less actuation potential • If we can lock a single arm cavity, we can lock the IFO. 40M: 7 mN 1.3 kg test mass AdLIGO f/m = 5 Jan 17th 2006 20 µN 40 kg test mass f/m =5e-4 Ward Candicacy 29 Digital length control system D/A mixer Jan 17th 2006 Ward Candicacy Output to suspensions A/D Feedback filters AP166 Demodulated signal from PD D/A 30 Compensating the resonances Compensation Filters for the two resonances associated with the signal cavity: UGFs ~ 250Hz Optical DARM CARM Jan 17th 2006 Opto-mechanical 4kHz >> UGF no compensation AdLIGO: 180 Hz ~ UGF 40Hz < UGF no compensation AdLIGO: 70Hz? 1kHz -> 100Hz ~ UGF dynamic compensation 0->100Hz ~ UGF Not yet coherently compensated Ward Candicacy 31 Dynamic compensation filter for CARM servo Open loop TF of CARM Optical gain of CARM • Optical gain (normalized by transmitted arm power) shows moving peaks due to reducing CARM offset. • We have a dynamic compensative filter having nearly the same shape as optical gain except upside down. Designed using FINESSE. • Open loop transfer function has no phase delay in all CARM offset. Jan 17th 2006 Ward Candicacy 32 CARM optical springs CARM optical springs at different CARM offsets 140 Arm power = 6 Arm power = 8 Arm power = 10 •Solid lines are from TCST •Stars are 40m data •Max Arm Power is ~80 •Also saw CARM anti-springs, but don’t have that data 130 CARM optical response (dB) 120 110 100 90 80 2 3 10 10 f (Hz) Jan 17th 2006 Ward Candicacy 33 Mode healing/injuring at Dark Port Negative spring constant with optical spring Positive spring constant with no optical spring Carrier power at DP is 10x smaller • Repeatable • The same alignment quality Jan 17th 2006 Ward Candicacy 34 Modeling Jan 17th 2006 Ward Candicacy 35 What’s modeling all about? • With 5 DOFs and detuned cavities, Advanced LIGO will have a very complicated optical configuration, with a rich frequency response. We need good modeling tools, and we need to use them in order to understand AdLIGO, before it is built. • We already rely heavily on modeling at the 40m because the configuration is so complicated. • Building a prototype, and then using modeling to extrapolate, is a good way to understand AdLIGO in advance! Jan 17th 2006 Ward Candicacy 36 Optickle: Frequency Domain IFO Simulation • Optickle is a new frequency domain IFO modeling tool: – Written in Matlab • Matlab allows easy integration to other modeling efforts (a frequency-domain e2e, like LinLIGO) • Easily Extensible • Uses Matlab classes for generality – Uses the methods outlined in T. Corbitt et al: “Mathematical framework for simulation of quantum fields in complex interferometers using the two-photon formalism” (LIGO-P030071-00R) to calculate the IFO opto-mechanical frequency response. – Designed for concrete units (Watts, meters, Hz) Jan 17th 2006 Ward Candicacy 37 Optickle example: detuned FP cavity • Response of front mirror to back mirror ‘excitation’ • 1 nm detune • finesse ~ 1200 0 Mag (dB) -50 -100 -150 -200 180 1 100 10000 100000 100000000 135 Phase (deg) 90 45 0 -45 -90 -135 -180 0 10 Jan 17th 2006 1 10 Ward Candicacy 2 10 Frequency (Hz) 3 4 10 10 38 Optickle Example: AdLIGO • Normalized row of DC readout signal matrix 40 CARM PRC SRC MICH 20 0 -20 dB mag • Easy to create a frequency dependent coupling matrix, useful for, e.g., estimating the contribution of loop noise to DARM. This plot is Open Loop. Closed loop coming soon! -40 -60 -80 -100 -120 -2 10 Jan 17th 2006 0 10 Ward Candicacy 2 10 f (Hz) 4 10 6 10 39 Optickle v. the 40m DARM Response DARM Response 80 80 70 70 60 50 dB (a.u.) dB (a.u.) 60 50 40 40 30 20 10 30 0 40m Data Optickle 20 10 1 10 2 3 10 10 40m Data Optickle -10 4 10 -20 1 10 2 3 10 4 10 10 f (Hz) f (Hz) Optickle Modeling Jan 17th 2006 Ward Candicacy 40 DC Readout Jan 17th 2006 Ward Candicacy 41 Quantum Noise: Heterodyne vs Homodyne -20 Quantum noise curves plotting using formulas in: 10 A. Buonanno, Y. Chen, N. Mavalvala, “Quantum noise in laser-interferometer gravitational-wave detectors with a heterodyne readout scheme” PHYSICAL REVIEW D 67,122005 2003 10 -21 GREEN = RF RED = HOMODYNE -22 h 10 -23 10 -24 10 -25 10 1 2 10 3 10 10 f (Hz) Jan 17th 2006 Ward Candicacy 42 What is DC Readout and how does it relate to Homodyne Detection? DC Readout is Homodyne detection, using light circulating in the interferometer as a local oscillator. Advantage: LO light has been filtered by the <1Hz coupled cavity pole Disadvantage: limited ability to control homodyne phase OMC Jan 17th 2006 Ward Candicacy 43 Technical noise sensitivity Noise Source Laser frequency noise Laser amplitude noise Laser pointing noise Oscillator phase noise Jan 17th 2006 RF readout DC readout ~10x more sensitive Less sensitive since carrier is filtered Sensitivity identical for frequencies below ~100 Hz; both driven by technical radiation pressure 10-100x more sensitive above 100Hz Carrier is filtered Sensitivity essentially the same -140 dBc/rtHz at 100 Hz Ward Candicacy NA 44 RF vs DC o Phase modulate the input light o RF sidebands act as local oscillator for GW signal, after passing through (unstable) recycling cavity(ies) o GW signal is an audio frequency sideband of RF photocurrent o Mix GW signal down to nearDC o Acquire GW signal at DC with ADC Jan 17th 2006 Eliminate the RF sidebands at Dark Port with an Output Mode Cleaner Eliminate junk light at the Dark Port with Output Mode Cleaner Carrier light acts as a local oscillator GW signal is an audio frequency photocurrent Acquire GW signal at DC with ADC Ward Candicacy 45 Making the DC local oscillator • Two components – Carrier field due to loss differences (not LIGO I GW parallel to DC offset controllable? TCS?) • • fringe – Carrier field due to dark fringe offset offset (controllable) – An output mode cleaner should take care of the rest. (RF sidebands, junk light) Loss mismatch component – Average arm round trip loss: 200 ppm – Difference between arms: 50 ppm – Output power due to mismatch: 20 µW Detection angle, β – Tuned by adjusting fringe offset β Loss mismatch • Can tune from 0-80 deg with 0-10pm of DARM offset • 1 mW LO – Angle of GW is frequency dependent in detuned RSE Jan 17th 2006 Ward Candicacy Detuned RSE: GW signal gets fdependent phase shift in SRC Some linear component No slope 46 Laser Intensity Noise • • calculated using rsenoise 10 pm DARM offset for DC 1e-13m residual L- RF: noise sidebands of RF sidebands beat against residual length offset DC: dark port power proportional to input power Radiation pressure effects not included Jan 17th 2006 -5 10 Ampl. noise equiv shot noise 1/rtHz • RF -6 DC 10 -7 10 -8 10 -9 10 0 10 1 10 2 10 3 10 4 10 f (Hz) Ward Candicacy 47 Laser Frequency Noise • • calculated using rsenoise 10 pm DARM offset for DC 1e-13m residual LRF: frequency noise sidebands of RF sidebands beat against static carrier contrast defect DC: Arm cavity pole imbalance couples carrier frequency noise to dark port -1 10 Freq. noise equiv shot noise Hz/rtHz • -2 10 RF -3 10 DC -4 10 -5 10 -6 10 -7 10 -8 10 -9 Radiation pressure effects not included Jan 17th 2006 10 0 10 1 10 2 10 3 10 4 10 f (Hz) Ward Candicacy 48 OMC Properties The Output Mode Cleaner filters the light coming out of the dark port, cleaning away the junk and transmitting the GW-signal containing TEM00 mode of the carrier Transmission of HOM versus g-factor, Max Mode Number = 3 2 10 Transmission % 10 11 12 23 1 10 3 0 31 32 43 22 23 31 01 12 33 30 22 11 32 23 31 0 10 -1 10 -2 Maxmimum Unwanted Transmission Transmission of higher order TEMs Transmission of RF Transmission of TEM00 10 -3 10 Jan 17th 2006 02 20 32 0 0.1 0.2 0.3 Ward Candicacy 0.4 0.5 g-factor 0.6 0.7 0.8 0.9 1 49 OMC design in SolidWorks • • • • • • Small number of pieces HV compatible – some glue near the PZT mirror Mirrors mounted mechanically, on silver washers (no glue) ALGOR FEA: lowest mech resonance at ~770 Hz Construct out of well-damped material, to minimize effect of resonances: Copper All high-quality (REO super-polished and coated) mirrors available from LIGO lab spares (the 4th HR mirror, 0o incidence, came from Newport) Mike Smith Jan 17th 2006 Ward Candicacy 50 The Vacuum Compatible DC Photodiode Ben Abbott DC Readout Jan 17th 2006 Ward Candicacy 51 Summary & Future Directions Things I may spend significant time on Things I’ve already spent significant time on, and will spend more on • Lock Acquisition • Modeling • DC Readout Jan 17th 2006 • • • • Ward Candicacy Revamping the LSC Scheme QND Techniques SPI Data Analysis 52