birefringence

Sparked by a talk by Brian Keating at "testing gravity" meeting in Vancouver, Jan 2015:

A variety of phenomena might lead to birefringence for propagating photons. Numerous efforts have looked for rotation of polarization as a function of wavelength, which is very sensitive since it tests equality of arrival times  with a resolution that is the period of the electromagnetic wave. But this requires high SNR detection of existing linear polarization, which is typically small. An alternative is to look at synchronicity of arrival times for L and R circularly polarized light, which amounts to looking at photons of different helicity. 

We could use any transient or time-variable source, and cross-correlate the R(t) and L(t) signals to constrain arrival time difference. Can measure difference of path integral of refractive index with a fraction precision that is delta-t/t_propagation, of order (1 day)/6 x10⁹ years or a part in 4x10¹³ or so. Not bad!

Magellan with circular polarizer in front of focus:

• magellan plate scale is about 3 arcsec/mm so a 60 mm disk subtends 60 x 3 = 180 arcsec = 3 arc min. 
• f/11 beam has half-opening angle of arcsin(1/22) = 2.6 degrees, so it's damn near collimated. 
• We should be able to put a circular polarizer (linear polarizer plus broadband QWP) in the beam and still retain good performance. 
• Achromatic True zero-order waveplates have angular acceptance of +-7 degrees. 

wave plate bands and grizy overlap:

380-570 g

400-600 g

450-675 NA

500-750 r

600-900 i

700-1050 z,y

800-1200 NA

can get these from astropribor company, Ukraine

What about a broadband high performance polarizer? 

Newport has decent 50 mm broadband polarizers

http://www.newport.com/Precision-Linear-Polarizers/141147/1033/info.aspx#tab_Specifications

UBB polarizers look even better, see

http://www.lot-qd.de/files/downloads/moxtek/eu/Ultra_Broadband_Polarizers_UBB_series_300-4000nm_eu.pdf

Although sensitivity is down by huge factor (number of EM periods in 1 day, about (lambda/c)/1day = (500 THz ) * seconds per day = 500 E 15* 86400 = 4 E 22 !! , this is a differential test that does have some advantages:

• independent of acutal polarization state (as long as detectable amount of light exists in both L and R states). 
• prospect of probing volume-dependence since we can use sources at various RA, Dec, z. 
• avoids instrumental calibration issues with subtle linear polarization features. 
• Can be done in a single band
• works for any EM system that can determine circular polarization (Radio, optical, IR)
• any temporally variable source works: SNe, GRB's, millisecond pulsars, QSO's, AGN, 
• differential, null test, which is always the best way to do things. 
• no prior knowledge of polarization needed. 
• independent of wavelength dependence of birefringence, although could look for that too
• sensitive to various new-physics effects: Lorentz invariance, WEP violation (torsion?), axion-like particles. 
• no phase-winding ambiguity, cumulative effect. 
• simple to do with a pair of filters. 

competing methods either look for rotation of some known polarization or look for wavelength dependence of linear polarization (requires knowledge of source effects). 

FLWO 1.2 meter has keplercam instrument. Uses 4 inch square filters. 

I think it's about f/8. that still works for the broadband quarter wave plate. 

Any temporally variable source can be used for this: microlensing, QSO's, SNe, etc. 

Also can use msec pulsars, gain of 1000*3600*24= . 

prior expectation, or energy/wavelength dependence of polarization orientation.