LIGO II Photon Drive Conceptual Design

Transcription

1 LIGO II Photon Drive Conceptual Design LIGO-T R M. Zucker 10/13/00 ABSTRACT LIGO II will require very small forces to actuate the final stage test masses, due to the high isolation factor and hierarchical action of the multiple suspensions and active seismic isolators. To reduce possible extrinsic noise interactions, it may be desirable to actuate on the final stage through photon pressure alone. A simple configuration is outlined for further evaluation and discussion. Introduction Extrinsic sources of mechanical loss in LIGO II interferometer test masses must be avoided to achieve the required low internal-mode thermal noise. As a result attachment of ferromagnets for fine position and angular control, as done in LIGO I, is precluded for LIGO II test bodies. More sophisticated suspension and seismic isolation systems planned for LIGO II greatly reduce the peak and RMS correction forces needed on the test masses, by actuating principally on upper stages of the suspension. However it is likely that high-bandwidth control actuation capability will be required, especially for the cavity lengths, to enable the requisite high open-loop gains and to permit lock acquisition. Because of the intrinsic time delay in transmission of forces down the suspension fibers, actuation bandwidths greater than about 1/4 of the suspension fiber violin mode frequency (300 to 500 Hz typically) will require direct force actuation on the test mass. Two candidate non-magnetic actuation schemes are under consideration: electrostatic and photon-recoil drive. These may be used in combination or sequentially, e.g., for higher reserve during lock acquisition and lower noise during quiescent operation. The fringingfield dielectric electrostatic drive under development by GEO is likely to meet all relevant criteria, but must at some point must be proven to against an array of suspected spurious effects (patch effect, nonlinear upconversion, surface charge migration, electronic noise, etc.). The photon drive system presented here is intrinsically unable to provide as much peak force, but assuming seismic isolation and suspension developments are reasonably successful at reducing the residual forces required, it should be adequate. Furthermore it is so simple that (at least on paper) its performance limitations are comparatively easy to predict.

2 Briefly, a secondary laser of adjustable power is directed at the test mass and reflected into a beam dump. Assuming there is very little absorption, twice the momentum of each photon is transferred to the mirror. The resulting acceleration is just Pt () xt () = 2 Mc where Pt () is the instantaneous power reflected and M is the mass. To offer a numerical example, for sinusoidally varying power P RMS at frequency f the root-mean-square mirror displacement is x RMS 2PRMS PRMS = RMS 13 f Mc 10 Hz f 30 kg m 2 2 4π 10 W M 2. Noise and Dynamic Range Intensity noise Technical intensity noise on the drive laser must not compromise the displacement noise of the interferometer. Allowing no more than 0.5% degradation of strain sensitivity at any observation frequency implies that the total contributed noise must, for example, satisfy x n( f) ( 01. ) m/ Hz at f = 10 Hz. Taking ( ) P ( f) 2π 2 f 2 Mcx ( f)/ N where N is the number (probably 2) of test masses equipped with independent photon drives provides the spectral noise density limit for each drive laser; n (. 7 W 30 kg P 10 Hz) , Hz M (. 7 W 30 kg P 30 Hz) , and Hz M (. 6 W 30 kg P 100 Hz) Hz M Here at each frequency we have used the lower of the sapphire and silica cases from the August '00 reference design noise spectrum, i.e , and m/ Hz at 10, 30 and 100 Hz respectively. If four mirrors must be driven rather than two, these requirements become 2 tighter.

3 Expressing the power fluctuations as a fraction of the mean drive laser power, often called "relative intensity noise" or RIN ( RIN P ( f)/ P), translates this noise constraint into a maximum peak force, given any technically feasible RIN. As a baseline, suppose we are able to stabilize the drive laser intensity to the level determined by photon shot noise in a 100mW sample, giving ( P f)/ P = 2hν / P sample / Hz. 1 Without some form of optical filtering (a significant challenge at these low frequencies), this constrains the maximum drive laser power per mirror to about 290 W. 2 This power limit in turn constrains the maximum peak-to-peak drive force at F 2P / c N p p max implying a DC displacement stroke of approximately 1.7 nm (for a 30 cm pendulum of 30 kg mass). To determine the force required for normal quiescent operation in principle requires analysis of the seismic isolation and suspension isolation effectiveness, ambient excitation spectrum, and the actuation crossover and range allocation characteristics of upper stage suspension controls. It is comforting that the noise spectrum projected for LIGO II, translated into force units and integrated down to 10 Hz, corresponds to approximately Newton RMS. However the true determinant will be how well the larger residual motions at lower frequencies can be filtered away, consistent with a stable control characteristic at the crossover with upper-stage actuators. The issue must be probed through further simulation and analysis. Scattering interference Light scattered from the photon drive laser must not interfere with detection of the interferometer signals. This requirement should not be especially difficult. The drive laser power is a small fraction of the circulating power in the power recycling cavity, it can be directed well outside the diffraction cone of the interferometer, and it can be at a different wavelength which is rejected by selective relay optics. Thus, even if a DC readout is used (so there is no demodulation rejection) we expect sufficient isolation to be readily achievable. Optic heating The drive beam is highly reflected from the optic and carries a small fraction of the power in the main resonant interferometer beam. There is thus little potential for additional optic distortion due to the drive laser's absorbed power. 1 To our knowledge this RIN level has not as yet been achieved in a working high-power laser at 10 Hz, so this may be optimistic. 2 Note that the drive might be permitted higher noise at the extreme end of its force range; for example, the extremes may only be needed for acquisition transients. Here I am explicitly treating only the "operational" low noise mode.

4 Design Concept Stabilized laser source A TEM 00 laser, preferably though not necessarily single-longitudinal-mode, is probably the best candidate to achieve the requisite intensity stability with active feedback control. Several diode-pumped solid state lasers in the 10 W class are available commercially 3. We propose use a folded multibounce optical path (see below) to hit the test mass repeatedly, adding up discrete bounces to reach the optimum total drive power. The laser will require wideband active stabilization to suppress amplitude noise, potentially by a factor of 10 4 to 10 6 depending on the initial laser noise level. Prior work (e.g., Robertson et al. [noise eater]) indicates this degree of intensity noise suppression can only be achieved with significant attention to spatial inhomogeneity interactions between sensing photodetectors and perturbations of the laser spatial mode. While the most effective technique might be to pass the laser beam through a single-mode optical fiber, power handling and damage at this power could be an issue. A simpler alternative involving a spatial filter is offered in Figure 1; this is combined with a spatially uniform diffused detector (at the expense of some quantum efficiency) to further reduce the propensity for geometric and interference-related power sensing errors. The actuation technique will vary with the laser selected, but will likely be based on the successful current-shunt actuator developed by LIGO (Abbott et al. [current shunt]) Testing may indicate that noise suppression to the required level is not feasible in a single stage; we therefore also include in Figure 1 a possible independent second-stage suppressor, based upon an acoustooptic modulator. In this case the higher sense power would be directed to the second stage error detector, as its photocurrent shot noise would be the final determinant of achievable RIN. 3 e.g., Q-peak MPS-1047CW, Spectra Physics fcbar, or Coherent Compass series.

5 SPATIAL FILTER AOM SPATIAL FILTER DPSS LASER OUTPUT PUMP DIODES INT. SPHERE INT. SPHERE Σ FAST PD ~ RF OSC AM IN PD Σ COMMAND INPUT PUMP DIODE POWER SUPPLY SLOW Figure 1. Two-stage intensity stabilization for a diode-pumped solid state laser used in photon drive system. The laser is a 10W-class single mode oscillator operating at a non-interferometer wavelength (e.g., Nd:YLF or Nd:YVO 4 at 1047 or 1053 nm) to permit filter separation of optical interference if required. Initial stabilization is implemented by modulating the pump diode supply current; bandwidth is augmented by a high-speed current shunt bypassing the supply regulators. The second stage uses an acoustooptic modulator (AOM) to directly modulate the laser power. Spatial filtering of the beam and integrating spheres on both error point detectors reduce parasitic interactions with beam geometry and insure total power is stabilized. The drive force command (from the interferometer length controller) is introduced by offsetting the error signal of the second loop. Multipass beam folding & steering Several multipass geometries have been suggested to achieve the requisite number of bounces (between 10 and 30) on the interferometer test mass. A Herriott delay line (Herriott et al [delayline]) would be self-aligning but requires a large well-figured spherical optic, larger than the test mass, with a substantial clearance hole in the center for the main beam. A modified White cell arrangement (Figure 2), while mechanically more complex, requires no special or large optical components. Its distinct adjustable mirrors also afford greater flexibility for "tuning" the centroid of the beam pattern to the center of percussion of the optic, thereby eliminating parasitic torques. We currently envision making a few of these mirrors remotely controllable to allow tuning of the beam path under vacuum. To ease mechanical congestion, beam clearances and polarization issues it is best to locate the folding optics some distance (e.g., 3-5 m) from the test mass. However this

6 places the folding system in the beam manifold. Close integration with the manifold scattered-light baffles will therefore be required. MODEMATCH PBS/QWP POWER CMD LASER SYSTEM READBACK PD DUMP VAC WINDOW FL CC MAIN BEAM TEST MASS FL ~ 3 m Figure 2: Modified White cell multibounce system for photon drive. The modematching group magnifies the beam to approximately 15 mm diameter to minimize diffractive divergence over the full optical path. N=4 bounces are shown for clarity; for N=30 bounce system, 14 flat mirrors (FL) will be arrayed in an annular zone about the interferometer beam axis. Concave retro mirror (CC) returns the beam along its initial path, doubling the bounce number, and maintains the mode diameter. A polarizing beamsplitter (PBS) / quarter waveplate (QWP) circulator deflects the returning beam to a beam dump; a sample is measured by a photodiode for readback of the drive power.