Understanding Initial LIGO and Possible Influences on Enhanced LIGO Sam Waldman and Rai Weiss February 2008
Motivations Outline Improve understanding of initial LIGO, f < 200Hz Improve chance of detections in enhanced LIGO Reduce probability of surprises in advanced LIGO Reduce commissioning time in advanced LIGO Known and suspected noise contributions Evidence for the noise sources and their effect on the enhanced LIGO spectrum in the low frequency band. Noise sources considered in the original enhanced LIGO plan included. Concepts to reduce the noise
Noise sources Sufficient understanding to propose fix Charge fluctuations on the dielectrics Up-conversion: magnetic moment fluctuations in control magnets driven by control current Broad band noise in bias and coil driver electronics: Johnson and excess noise Insufficient understanding to propose fix Auxiliary length and angular control noise coupling to main interferometer Excess dissipation in the test mass suspensions leading to increased thermal noise
Horizon Distance Mpc curve NS/NS /BH 30/30BH 60/60BH H1 S5 32 160 169 57 L1 S5 31 157 215 83 SRD 34 170 219 127 Rana S6 71 349 443 208 SUM 92 450 638 209 Detection Rate relative to SRD curve NS/NS /BH 30/30BH 60/60BH H1 S5 0.84 0.83 0.46 0.09 L1 S5 0.79 0.79 0.94 0.28 Rana S6 9.1 8.7 8.2 4.4 SUM 20 19 23 4.5 Estimates by Ilya Mandel, Richard Shaughnessy, Vicky Kalogera Jan 2008 NS/NS detection rates using 0 NS/NS mergers per Myr in MWEG and 0.01 MWEG/Mpc 3 H1 S5 => 0.012/year L1 S5 => 0.011/year SRD => 0.014/year Rana S6 => 0.13/year SUM => 0.28/year
Charging Noise Charging event from earthquake stop Viton contacting test mass. Figure bottom right shows fit of difference in displacement spectra (top right) to 1/f 3 spectrum. Charge transfer measurements between Viton and fused silica are 0 times larger than by silica against silica, Moscow State University measurements. Fix: new earthquake stops Rupal Amin
Up-Conversion Observation of reduction in NS/NS range and increased noise below 0Hz with increase in low frequency (f < Hz) noise. Investigation into causes of up-conversion RF saturation Coil driver saturation and non-linearity relation to test mass displacement Positive correlation with coil current Mitigation Reduction in coil current by redistribution of control Feed forward technique for low frequency contribution (proposal) Source of the up-conversion is magnetization noise in the magnets Barkhausen noise test rig results In-situ direct measurements of the up-conversion Projection for the residual noise under quiescent conditions Proposal to replace noisy NdFeB by quiet SmCo magnets of equal strength and geometry
null RF saturation measurement null coil driver measurement ) 1/2 Magnitude (m/hz for ref 8 exciting H2:SUS-ETMX_LSC_EXC f=1.73 Hz,A=300-6 -7-8 -9 - -11-12 1 Frequency (Hz) *T0=20/02/2006 02:32:52 *Avg=13 BW=0.1875 Power spectrum H2:LSC-DARM_ERR H2:LSC-DARM_ERR(RMS) H2:LSC-DARM_ERR(REF2) H2:LSC-DARM_ERR(REF2) H2:LSC-DARM_ERR cts/rthz 3 Input digital excitation 2 1 0!1!2!3!4!5!6!4 0 1 2 Measured analog signal Normal Spectra 7! 12 Hz exc. 2.5! 7 Hz exc. 1! 2.5 Hz exc. -17 1/2 Magnitude (m/hz ) -18 V rms /rthz!5-19 30 40 50 60 70 80 90 0 1 RA 02/19/2006 LHO Frequency (Hz) 1!6 1 2 3 Frequency SW 04/07/2006 LHO
noise independent of DARM motion noise correlate to coils ) Mid-Y ground injections: Black: none, Red: 0.75 Hz, Blue: 1.2 Hz -6 H1:LSC-DARM_ERR(REF0) H1:LSC-DARM_ERR(REF9) H1:LSC-DARM_ERR 16 15.5 200 days of H1 performance 1/2 Magnitude (m/hz -7-8 -9 H1 Range [Mpc] 15 14.5 14 13.5 13-0.5 1 1.5 2 2.5 3 3.5 4 Frequency (Hz) *T0=03/12/2005 Same color scheme 23:00:55 *Avg=1 *BW=0.046866 ) 1/2-17 H1:LSC-DARM_ERR(REF0) H1:LSC-DARM_ERR(REF9) H1:LSC-DARM_ERR 12.5 12 0 20 30 40 50 ETM Drive voltage [V pkpk ] 200 days of L1 performance 16 15 14 Magnitude (m/hz -18 L1 Range [Mpc] 13 12 11-19 20 40 60 80 0 120 140 160 180 200 RS 02/19/2006 LHO Frequency (Hz) 2 9 8 0 20 30 40 50 ETM Drive voltage [V pkpk ] SW 03/29/2007 LSC
Coil Drive [LLCOIL_IN2 cts] L1:IOO!MC_F [cts/rhz] ETMY Pringle mode comparison 3 2 1 0 0 1 Frequency [Hz] 1 0!1 0 1 Frequency [Hz] Calibrated DARM_ERR [m/rhz] L1:ASC!WFS1_QY [cts/rhz]!8!9!!11 0 1 Frequency [Hz] 4 3 2 1 0 0 1 Frequency [Hz] Pringle mode excitations show noise without optic motion dependent on coil drive Induced noise spectrum similar to drive with pos and microseism Noise spectra consistent with 70-0 Hz excess. + - Calibrated DARM_ERR [m/rhz]!15!16!17!18 Baseline DARM Into ETMY_LSC_EXC Into Pringle drive - +!19 1 2 Frequency [Hz] 3 SW 08/2006 LLO
Coil reduction w/ damping filters Coil reduction w/ tidal feedback Power spectrum ) 1/2 Magnitude (m/hz -8-7 -9 - -11-12 -13-14 -15-16 -17-18 -19 H1:LSC-DARM_ERR H1:LSC-DARM_ERR(REF0) -1 1 Frequency (Hz) *T0=17/09/2006 Power spectrum 00:30:40 Avg=20/Bin=2L BW=0.0468742 2 Mag [db] 20 0!!20!30!40 180 data fit -15 H1:LSC-DARM_ERR H1:LSC-DARM_ERR(REF0) 135 90 ) 1/2 Magnitude (m/hz -16-17 -18 Phase [deg] 45 0!45!90!135-19 Frequency (Hz) *T0=17/09/2006 00:30:40 Avg=20/Bin=2L BW=0.0468742 SW 09/17/2006 LHO 2 4!180!3!2!1 0 1 Frequency [Hz] irish 01/30/2007 LLO
reduction w/ feedforward (first effort) Power spectrum reduction w/ adaptive feedforward (predictive) MISO Wiener Filter based subtraction Magnitude 1!3-1 H1:SUS-ETMX_SUSPOS_IN H1:SUS-ETMX_SUSPOS_IN(REF0) H1:SUS-ETMX_SUSPOS_IN(REF3) 1 Frequency (Hz) *T0=21/11/2007 01:39:05 *Avg=20 BW=0.18 Transfer function 20 Amps or Amps/!Hz!4 Magnitude (db) 0 - -20-30 -40 H1:SUS-ETMX_SUSPOS_IN(REF2) / H1:SEI-ETMX_GS_2(REF2) H1:SUS-ETMX_SUSPOS_IN(REF1) / H1:SEI-ETMX_GS_2(REF1) 1 RA 11/21/2007 LHO Frequency (Hz) 2 5!5!6 DARM_CTRL DARM_CTRL (rms) STS 24 STS 24 (rms) STS + Guralp 24 STS + Guralp 24 (rms) STS 2048 STS 2048 (rms) RA /16/2007 LLO!1 0 1 Frequency [Hz]
Barkhausen Noise Fluctuations in the magnetization of ferromagnetic materials due to internal friction of domain rotations when driven by time varying magnetization currents. Occurs in permanent magnets that are not saturated Fluctuating force F(f) "!(f) db dz Fluctuating magnetic moment of control magnet varying as the I 3/2. Has both coherent (repeatable) and incoherent components. Gradient of the magnetic field acting on the control magnet. The gradient comes primarily from the PAM magnets with a small part from the bias current in the coil
Upper figure: Repeatable component of the Barkhausen noise near H = 0 and the magnetization discontinuity at H extremum. Measured in bridge test rig in NdBFe magnets. The repeatable part depends on the direction of the exciting current and the magnetic moment. SmCo does not exhibit this behaviour because the magnetization is saturated. Lower figure: Both incoherent and coherent components Spectrum in test rig varies as 1/(f-fo). This spectrum corresponds after conversion to B field (1/f) and displacement (1/f2 ) to the up-conversion spectrum B field gradients from the PAM magnets as a function of magnet separation
Fix:reduce current, increase PAM/control magnet spacing, change NdFeB to SmCo magnets Assume PAM/control magnet spacing is 8mm Assume PAM/control magnet spacing is 8mm Note: Estimate from bridge experiment underestimates Barkhausen noise due to assumption of 8mm magnet spacing. Estimate from in situ measurments most likely overestimates due to simplicity in convolution with quiescent current spectrum. Barkhausen noise scaling Drive current: I 3/2 Spectrum: convolution with 1/(f-fo)f 3 fo = drive frequency
Broadband noise in bias and coil driver electronics Considered in the original plan for enhanced LIGO Both bias modules and coil drivers to include filters to remove excess noise Johnson noise could become dominant noise Variety of ideas and mitigation strategies: Use larger series resistors, may require new bias module and will require higher voltage coil drivers Operate series resistors at 30K using cryogenics acquired for beamtube bakeout. Place cryogenics in maintenance area. Use an inductor as the series impedance for the bias Use electronic cooling with low noise FET Less urgency, external to the instrument and can be modified during S6 The noise estimated in the figure comes from Johnson noise in the series resistors, 7.5K ohms in the bias modules and 4K ohms in the coil drivers using a coil drive of 1.6x -2 Newtons/Amp. Excess electronics noise is typically larger by 1.5.
Auxiliary Length and Angle Noise A mixture of technical noise sources in the initial LIGO detector has been grouped as auxiliary length noise and angle noise, shown in figure at left. Auxiliary length noise is the coupling of differential Michelson (BS) and common mode Michelson (RM) sensing noise into the darm error signal. This noise is expected to reduce with increased injected laser power. The angle noise is the coupling of the WFS sense and control signals into the darm error signal. The best guess is that this noise will not change for enhanced :LIGO. In either group,there is significant work expected to make any changes in these contributions. Consult Rana Adhikari and Peter Fritschel for estimates of the intransigence of these noise terms.
EXCESS DISSIPATION IN TEST MASS WIRE SUSPENSIONS Evidence for the excess loss Inconstant and low values of wire violin mode Q in all three interferometers Direct measurements in the suspension test rig give similar inconstant and low violin mode Q Needs resolution May significantly limit enhanced LIGO spectrum Needs to be understood (and fixed) for Advanced LIGO signal recycling mirror suspension Do not currently have a reliable fix
What is known Wire structure damping loss is low, #$"$%$x -5 (S. Penn free wire measurement) #$&$'$x -4 (S. Penn guitar measurement). BASIS FOR SUM CURVE IN ENHANCED LIGO PROJECTIONS Upper wire clamp is most likely not implicated. Lower grooved cylindrical wire standoff is most likely cause. sharp edged prism gives factor ~ increase in Q (guitar) prism improves Q in suspension test rig but not consistently Wire vibration polarizations experience different loss in the LIGO suspension, FB > RL. Guess that wire below the standoff vibrates due to slope coupling across standoff and rubs on the test mass in FB motions. More consistent results replacing standoff with a small hardened steel clamp.
D. Malling G. Harry Large amplitude D. Malling Small thermal amplitude Test mass wire suspension violin modes have significantly lower Q than expected from knowledge of the material properties and the Q vary from measurement tp measurement Test mass wire suspension dissipation
Fix: Establish a well defined boundary condition at the standoff in the suspension test rig. Program: Need to do more research Need to be ready with a tested solution when (if) enhanced LIGO gets stuck at this noise. Thermal noise line drawn for wire structural loss 1 x -3, the average of the frequency domain in situ wire loss measurements