Instrument Science Report NICMOS 98 - 017 Methodologies to Calibrating NICMOS Polarimetry Characteristics Lisa Mazzuca, Bill Sparks, Dave Axon November 24, 1998 ABSTRACT Astronomical imaging polarimetry commonly utilizes a sequence of images taken with a polarizing filter oriented at different angles. For example, each of the Faint Object Camera (FOC), Near Infra-Red Camera and Multi Object Spectrometer (NICMOS) and the Advanced Camera for Surveys (ACS) instruments on the Hubble Space Telescope use three separate polarizing spectral elements oriented at approximately 60 degrees to one another to obtain a set of images that may subsequently be processed to derive polarization information on the astronomical object. Here we present the general linear theory of analysis for the case of three non-ideal polarizers. There are several ways to express the calibration of such polarizing elements and a variety of commonly used conventions. We present a summary of three of the more commonly encountered conventions, along with transformations between the different calibration systems. Specific examples are also cited for reference based on the short wavelength NICMOS polarizer characteristics. We present the transformations between three images taken with independent polarizers and the corresponding set of Stokes parameter images. 1. Introduction In order to calibrate data obtained with a set of polarizers, there are three quantities needed. They are the throughput to unpolarized light, the efficiency of the element as a polarizer, and the orientation of the polarizer. These elements can be expressed in a polarization reduction algorithm to form a solution containing the polarization characteristics of the incoming beam (i.e., the Stokes parameters I, Q, and U). 1 The general form of the equation for polarimetry data reduction is expressed as1 ′ I k = A k I + ε k ( B k Q cos 2φ k + C k U sin 2φ k ) (1) where I’k is the emerging light intensity from the kth polarizer, Ak, Bk, and Ck are the transmission coefficients, εk is the polarizing efficiency, and φk is the position angle of the kth polarizer. This linear equation captures the observed signal from a polarized source of intensity I and linear Stokes parameters Q and U, which describe the state of polarization for each polarizer. The transmittance expressions for Ak, Bk, and Ck will differ depending on the characteristics of the polarizing elements. For a set of three polarizers, as used with the FOC, NICMOS, and ACS instruments, equation (1) becomes a straightforward linear system of three equations with three unknowns. After solving the system of equations to derive the Stokes parameters, the degree of polarization P and the position angle θ can be calculated as 2 2 Q +U P = ------------------------, I 1 U θ = --- atan ---2 Q (2) There are three conventions presented below, each with differing transmittance expressions, that can be used to characterize the polarizing elements. The differences in the expressions for the transmission coefficients Ak, Bk, and Ck are dependent on the method chosen to characterize the polarizer. Although each convention involves different expressions, the solution is the same for a given polarizing set. 2. Theory and Convention Any one polarizer is characterized by two calibration parameters which are the parallel light transmission and the perpendicular light transmission, and a third parameter which is its position angle. Both the transmission expressions and the polarizer efficiency expression are composed of these two quantities. The material presented below shows the transformations between three commonly used conventions. In the natural theoretical approach, the polarizer splits the unpolarized light into two orthogonally polarized beams, one of which passes straight through onto the detector. The representation for the transmission coefficients in this case is presented in Hines, et al 2 paper entitled “The Polarimetric Capabilities of NICMOS2”. The transmission coefficients, including the polarizing efficiency, are: (tk ) (tk ) A k = -------- ( 1 + l k ) , B k = ε k -------- ( 1 + l k ) cos 2φ, 2 2 (tk ) C k = ε k --------( 1 + l k )sin 2 φ ( 3 ) 2 where tk is the fraction of light transmitted for a 100% polarized input aligned with the polarizer’s axis, and lk is the fraction transmitted when the incoming light is perpendicular to the axis of the polarizer. The polarizer efficiency ε is defined as the following: ε = ( S par – S perp ) ⁄ ( S par + S perp ) (4) where Spar = tk and Sperp = tk lk. Therefore, the equation for εk becomes 1 – lk t k – ( t k ⋅ lk ) ε k = -------------------------- = ------------1 + lk t k – ( t k ⋅ lk ) (5) and the reduction equation (1) becomes: I ′ k tk 1 – lk t k tk = ---- ( 1 + l k )I + ------------- ⋅ ---- ( 1 + l k ) ⋅ cos 2φ k Q + ---- ( 1 + l k ) sin 2φ k U ( 6 ) 2 1 + lk 2 2 which reduces to tk 1 – lk ′ I k = ---- ( 1 + l k ) I + ------------- ( cos 2φ k Q + sin 2φ k U ) 2 1 + lk (7) The full expression for the transmission coefficients from equation (3) is then (tk ) 1 – lk A k = -------- ( 1 + l k ) , B k = ------------- cos 2φ, 2 1 + lk 1 – lk C k = ------------- sin 2 φ 1 + lk (8) The second convention contains transmission coefficient expressions that apply to a pair of identical polarizers in series. This convention can be used as a practical way of characterizing a polarizer set in the laboratory (e.g., Serkowski, 1962). 3 In this case the values for Ak, Bk, and Ck, including the polarizing efficiency, are the following3: Ak = T k + Lk T k + Lk T k + Lk -----------------, B k = ε k ----------------- cos 2φ, C k = ε k ----------------- sin 2 φ 2 2 2 (9) where Tk is the transmittance of unpolarized light by the two identical analyzers oriented parallel, and Lk is the transmittance by the two identical perpendicularly oriented analyzers. The polarizer efficiency ε k is defined as εk = T k – Lk ----------------T k + Lk ( 10 ) The reduction equation (1) now becomes T k + Lk T k + Lk T k – Lk T k + Lk T k – Lk ′ - I + ----------------- ------------------ ⋅ cos 2φ k Q + ----------------- ------------------ sin 2φ k U I k = ----------------2 2 T k + Lk 2 T k + Lk ( 11 ) and reduces to ′ Ik = T k – Lk T k + Lk ( Q cos 2φ k + U sin 2φ k ) ----------------- ⋅ I + ----------------2 2 ( 12 ) Therefore, the transmission expressions in equation (9) become T k – Lk T k + Lk cos 2φ, C k = A k = -----------------, B k = ----------------2 2 T k – Lk ----------------- sin 2 φ 2 ( 13 ) Another generic approach is to simply let the transmittances be defined as zk with a polarizing efficiency of εk. The expressions for Ak, Bk, and Ck, including the polarizing efficiency, then become 1 1 1 A k = --- z k , B k = --- ε k z k cos 2φ, C k = --- ε k z k sin 2 φ 2 2 2 4 ( 14 ) with Equation (1) transforming to I ′ k 1 = --- z k ( I + ε k Q cos 2φ k + ε k U sin 2φ k ) 2 ( 15 ) 3. Conversions Equating the transmission coefficients from equations (8), (13), and (14) will yield general transformations between the polarization conventions stated. That is, by setting any two expressions for Ak, Bk, and Ck (accounting for the polarizing efficiency) equal to each other and then solving the system of equations for the various transmission variables, a straightforward relation can be made. The relations are as follows: tk in terms of Tk and Lk: tk = T k + Lk T k – Lk ------------------ + ----------------2 2 ( 16 ) lk in terms of Tk and Lk: T k + Lk – T k – Lk l k = ------------------------------------------------------------T k + Lk + T k – Lk ( 17 ) Tk in terms of tk and lk: 2 2 T k = 0.5t k ( 1 + l k ) ( 18 ) Lk in terms of tk and lk: 2 ( 19 ) Lk = t k lk zk in terms of Tk or Lk: Tk z k = 2 ------------------- 1 + ε 2 k or Lk z k = 2 ------------------- 1 – ε 2 k ( 20 ) zk in terms of tk and/or lk: t k z k = 2 ------------------( 1 + εk ) or t k lk z k = 2 ------------------( 1 – εk ) ( 21 ) 5 tk in terms of zk: 1 t k = --- z k ( 1 + ε k ) 2 ( 22 ) lk in terms of zk: 1 l k = ------- z k ( 1 – ε k ) 2t k ( 23 ) Tk in terms of zk: 1 2 2 T k = --- z k ( 1 + ε k ) 4 ( 24 ) Lk in terms of zk: 1 2 2 L k = --- z k ( 1 – ε k ) 4 ( 25 ) 4. Solutions for 3-polarizer case All three conventions used to express the polarization reduction algorithm reduce to a set of three equations with three unknowns. The solution results in the Stokes parameters I, Q, and U for the incoming light. Mathematically, the solution is derived from a matrix product of the inverse of the transmission coefficients matrix with the vector of intensities observed through the three polarizers. Let x be the incoming Stokes vector (I, Q, U), and let the vector of observed intensities be b = (I1, I2, I3). Then analytically, the linear equation [A]x = b can be solved for the incoming Stokes vector x, which yields the equation x=[A-1]b. The matrix [A] contains the transmission coefficients for each Stokes parameter I, Q, and U. In general the matrix [A] corresponding to equation (1) is A1 ε1 B1 ε1 C 1 A = A2 ε2 B2 ε2 C 2 A3 ( 26 ) ε3 B3 ε3 C 3 6 More explicitly, for the single polarizer convention, the matrix [A] is t1 ---- ( 1 + l 1 ) 2 1–l t ------------1⋅- ---1-( 1 + l 1 ) cos 2φ 1 1 + l1 2 1–l t ------------1⋅- ---1- ( 1 + l 1 ) sin ( 2φ 1 ) 1 + l1 2 t A = ---2- ( 1 + l ) 2 2 1–l t ------------2⋅- ---2-( 1 + l 2 )cos 2φ 2 1 + l2 2 1–l t ------------2⋅- ---2- ( 1 + l 2 ) sin ( 2φ 2 ) 1 + l2 2 t3 ---- ( 1 + l 3 ) 2 1–l t ------------⋅3- ---3- ( 1 + l 3 )cos 2φ 3 1 + l3 2 1–l t ------------3⋅- ---3- ( 1 + l 3 ) sin ( 2φ 3 ) 1 + l3 2 ( 27 ) For the second convention of dual polarizers [A] is A = T 1 + L1 -----------------2 T 1 – L1 ----------------⋅- cos 2φ 1 2 T 1 – L1 ----------------⋅- sin 2 φ 1 2 T 2 + L2 -----------------2 T 2 – L2 ----------------⋅- cos 2φ 2 2 T 2 – L2 ----------------⋅- sin 2 φ 2 2 T 3 + L3 -----------------2 T 3 – L3 ----------------⋅- cos 2φ 3 2 T 3 – L3 -----------------⋅- sin 2 φ 3 2 ( 28 ) and for the third convention [A] is z ----1 2 z2 A = ---2 z ----3 2 z 1 ε 1cos 2φ 1 --------------------------2 z 2 ε 2cos 2φ 2 --------------------------2 z 3 ε 3cos 2φ 3 --------------------------2 z 1 ε 1 sin 2φ 1 -------------------------2 z 2 ε 2 sin 2φ 2 -------------------------2 z 3 ε 3 sin 2φ 3 -------------------------2 ( 29 ) All three expressions for the matrix [A]can then be inverted to find the same solution to the original reduction equation. For simplicity, we present the analytical solution to the third convention using the matrix from equation (29). Using the reduction algorithm according to the convention in equation (15), let the following conditions exist: 2 ′ - I be the set of three observed intensities, and let I k∗ = ---z k k define an incoming Stokes vector to be { S1 } { S2 } { S3 } ≡ I Q U 7 . The matrix [A] from equation (29) becomes 1 ε 1cos 2φ 1 ε 1sin 2 φ 1 ( 30 ) A = 1 ε 2cos 2φ 2 ε 2sin 2 φ 2 1 ε 3cos 2φ 3 ε 3sin 2 φ 3 Solving for the incoming Stokes vector yields 3 Sk = ∑ akj I j∗ ( 31 ) j=1 where akj represent the elements of the inverted matrix a kj ε 2 ε 3sin( – 2φ 2 + 2φ 3 ) ε 3ε 1sin( – 2φ 3 + 2φ 1 ) 1 = ---- ε 2 sin 2φ 2 – ε 3 sin 2φ 3 ε 3 sin 2φ 3 – ε 1 sin 2φ 1 Ω ε 3 cos 2φ 3 – ε 2 cos 2φ 2 ε 1 cos 2φ 1– ε 3 cos 2φ 3 ε 1ε 2sin ( – 2φ 1 + 2φ 2 ) ε 1 sin 2φ 1 – ε 2 sin 2φ 2 ( 32 ) ε 2 cos 2φ 2– ε 1 cos 2φ 1 where Ω = ε 1 ε 2 sin ( – 2φ 1 + 2φ 2 ) + ε 2 ε 3 sin ( – 2φ 2 + 2φ 3 ) + ε 3 ε 1 sin ( – 2φ 3 + 2φ 1 ) ( 33 ) The solution is a vector consisting of the Stokes parameters for the incoming beam. Note that for the ideal configuration when ε 1 = ε 2 = ε 3 = 1 and ° ° ° ( φ 1 = 180 ) ,( φ 2 = 60 ) ,( φ 3 = 120 ) this reduces to the equation given in the NICMOS handbook4 with the exception of a sign convention difference I ∗ 1 I 1 1 1 1 Q = --3- 2 – 1 – 1 • I 2∗ 0 3 – 3 U I ∗ ( 34 ) 3 8 5. Numerical Example: NICMOS Polarizers In the case of NICMOS, the three polarizers have been characterized by Hines (Hines, et al,1997). Referring to the Hines’ paper previously cited, let the characteristics of the polarizers for POL0S, POL120S, and POL240S be the following: FILTER ϕk εk tk lk POL0S 1.42 0.9717 0.7760 0.0144 POL120S 116.30 0.4771 0.5946 0.3540 POL240S 258.72 0.7682 0.7169 0.1311 Table 1. Filter Characteristics for Convention One The resulting coefficient matrix (as computed from equation 27) is 0.3936 0.3819 0.0189 A = 0.4025 – 0.1166 – 0.1526 0.4054 – 0.2876 0.1195 ( 35 ) which can be used to compute the expected (I1, I2, I3) for a given set (I, Q, U). The inverted matrix is numerically calculated to be 0.8754 0.7735 0.8488 –1 A = 1.6647 – 0.5957 – 1.0246 1.0368 – 4.0582 3.0227 ( 36 ) which can be used to compute the Stokes parameters (I, Q, U) from a set of observations (I1, I2, I3). To complete this example, let the incoming Stokes vector be a beam fully polarized at a position angle of 45 degrees x = 1 1 = cos ( 2 ⋅ 45° ) 0 sin ( 2 ⋅ 45° ) 1 ( 37 ) Then the observed intensity vector, b, is 0.3936 0.3819 0.0189 1 0.4125 b = A x = 0.4025 – 0.1166 – 0.1526 ⋅ 0 = 0.2500 0.4054 – 0.2876 0.1195 1 0.5249 9 ( 38 ) Conversely, let the observed intensity vector be 0.4125 b = 0.2500 0.5249 ( 39 ) Then x = A –1 0.8867 0.7735 0.8487 0.4125 1 b = 1.6864 – 0.5957 – 1.0246 ⋅ 0.2500 = 0 1.0504 – 4.0582 3.0227 0.5249 1 ( 40 ) thus recovering the incoming Stokes vector (I, Q, U). Of course with the mathematics now presented for two other conventions, the same matrix can be found using the same initial coefficients, but with the expressions from the respective conventions. From equations (18) through (21), Tk,Lk, and zk are computed from the original values of tk and lk to be: FILTER Tk Lk zk POL0S 0.3012 0.0087 0.7871 POL120S 0.1989 0.1252 0.8051 POL240S 0.2614 0.0674 0.8109 Table 2. Filter Characteristics for Conventions Two and Three The matrix with respect to the second convention stated uses the coefficient matrix from equation (28), and is calculated to be 0.3936 0.3819 0.0189 A = 0.4025 – 0.1166 – 0.1526 0.4054 – 0.2876 0.1195 ( 41 ) which is the same as with the first convention, as shown in equation (35). 10 Using the third convention stated, the coefficient matrix from equation (29) is used, and is also calculated to be 0.3936 0.3819 0.0189 = A 0.4025 – 0.1166 – 0.1526 0.4054 – 0.2876 0.1195 ( 42 ) which again corresponds to the inverted matrices from the first two conventions from equations (35) and (41). 6. Summary Several conventions exist that can equally be used to solve the reduction equation for polarimetry. The solution produces the Stokes parameters (I,Q,U) for the incoming beam, which once known, can be used to calculate the degree and percentage of polarization and the position angle. 7. References 1N.Carleton,ed. Methods of Experimental Physics, Polarization Techniques. Academic Press, New York, 1974, p. 364. 2D.Hines., et al. “The Polarimetric Capabilities of NICMOS”. (Steward Observatory, University of Arizona), 1997. 3Z.Kopal, ed. Advances in Astronomy and Astrophysics, Polarization of Starlight. Academic Press, New York, 1962, p295. 4D. Axon, et al. NICMOS Instrument Handbook. Version 2.0 (Baltimore: STScI), 1997. 11