On the Van Cittert - Zernike Theorem for Intensity Correlations and Its Applications
Total Page:16
File Type:pdf, Size:1020Kb
On the van Cittert - Zernike theorem for intensity correlations and its applications TIMUR E. GUREYEV,1,2,3,4,* ALEXANDER KOZLOV,1 DAVID M. PAGANIN,2 YAKOV I. NESTERETS,4,3 FRANK DE HOOG,5 HARRY M. QUINEY1 1ARC Centre of Excellence in Advanced Molecular Imaging, The University of Melbourne, Parkville, VIC 3010, Australia 2School of Physics and Astronomy, Monash University, Clayton, VIC 3800, Australia 3School of Science and Technology, University of New England, Armidale, NSW 2351, Australia 4Manufacturing, Commonwealth Scientific and Industrial Research Organisation, Clayton, VIC 3168, Australia 5Data61, Commonwealth Scientific and Industrial Research Organisation, Canberra, ACT 2601, Australia *Corresponding author: [email protected] Received A reciprocal relationship between the autocovariance of the light intensity in the source plane and in the far-field detector plane is presented in a form analogous to the classical van Cittert - Zernike theorem, but involving intensity correlation functions. A "classical" version of the reciprocity relationship is considered first, based on the assumption of circular Gaussian statistics of the complex amplitudes in the source plane. The result is consistent with the theory of Hanbury Brown - Twiss interferometry, but it is shown to be also applicable to estimation of the source size or the spatial resolution of the detector from the noise power spectrum of flat-field images. An alternative version of the van Cittert - Zernike theorem for intensity correlations is then derived for a quantized electromagnetic beam in a coherent state, which leads to Poisson statistics for the intrinsic intensity of the beam. OCIS codes: (110.0110) Imaging systems; (110.4280) Noise in imaging systems; (110.2990) Image formation theory; (110.4980) Partial coherence in imaging; (340.7440) X-ray imaging. http://dx.doi.org/ estimating the source size or the width of the detector point-spread 1. INTRODUCTION function (PSF) from noise power spectra in so-called flat-field images (i.e. images obtained without any objects between the source and the detector) collected in a plane orthogonal to the optic axis. Such an The van Cittert - Zernike (vCZ) theorem is a well-known result of approach can be useful for simple measurements of spatial resolution classical statistical optics [1]. A generalized form of this theorem is of optical imaging setups and instruments, eliminating the need for sometimes referred to as the reciprocity relationship between the special resolution masks, sharp ("knife") edges or interferometric intensity distribution and the degree of (spatial) coherence of a quasi- measurements in some cases. It could be useful, for example, for monochromatic beam in the source and the detector planes [2]. This measuring coherence properties of synchrotron and X-ray free- result involves spatial correlation functions of the second order [2, 3], electron laser (XFEL) sources, which was recently studied e.g. in [5, 6] i.e. statistical averages of the product of complex amplitudes of an using the HBT approach. electromagnetic beam at two different points on a plane orthogonal to We also consider how the classical theory of HBT interferometry the optic axis. Correlation functions of the fourth order with respect to can be modified if one assumes a Poisson illumination model instead of complex amplitudes, such as statistical averages of the product of Gaussian sources. As an analogue of the Gaussian moments theorem optical intensities at two different points, are used in the theory of the for multivariate Poisson statistics does not seem to exist, we abandon Hanbury Brown - Twiss (HBT) interferometer [2, 3, 4]. In the present classical statistical optics in favour of a quantum electrodynamics paper we consider an extension of the vCZ theorem to such intensity treatment in this case. It turns out that the intensity autocovariance of a correlation functions. Much of the related material can be found at quantized electromagnetic beam in a coherent state does have Poisson different places in well-known textbooks [1, 2]. However, it appears statistics, which is intrinsic to the field's intensity variations, rather that an explicit form of this vCZ theorem for intensity correlations (we than being a feature of the probabilistic photon detection. This allows will call it the vCZ2 theorem) may not be widely known, particularly in us to derive an alternative form of the vCZ2 theorem for intensity the general polychromatic partially coherent case that we consider autocovariance with spatially correlated Poisson statistics. The here. We discuss how this result can be applied not only to source size measurements with the help of the HBT interferometer, but also for variance due to the photon shot noise in the detector is still present, of the complex amplitude in the source plane having circular Gaussian course, and is added to the intrinsic variance of the field's intensity. statistics. Let us now consider a general ergodic polychromatic beam and a 2. INTENSITY AUTOCOVARIANCE IN THE CASE OF model detector measuring average intensity over a time duration T in a T /2 single exposure, I( x )= T−12 | U ( x , t )| dt . Usually a light GAUSSIAN STATISTICS TR, ∫−T /2 R detector would be expected to integrate the instantaneous intensity Let a quasi-monochromatic partially-coherent extended source be over the exposure time, rather than average it. However, it is trivial to located immediately upstream from the "source plane" z = 0 implement the division by the exposure time after the exposure. We orthogonal to the optic axis z. We want to calculate first the will see that this normalization leads to results that have a convenient and natural mathematical form. It can be shown (see Appendix) that in autocovariance of light intensity IxR () at two points, x1 and x2 , in a distant "detector plane" zR= downstream from the source: this case eq.(5) becomes: CTR,(, xx 12 ) CxxR(,12 )=< IxIxRR ()1 () 2>−< Ix R ()1 >< Ix R ()2 >, (1) ≡< IxIx() ()>−< Ix () >< Ix () > (6) where the angular brackets denote ensemble average. In order to TR,1,2 TR TR ,1TR ,2 simplify the notation, we will work with the one-dimensional case, i.e. ∗ 2 = WRR(, xx121 ,)ν W (, xx 122 , ν )sinc[( πν1− ν 2 )]Tdd ν 1 ν 2 . the transverse coordinate x will be a scalar. Extension of the main ∫∫ results to the two-dimensional case is straightforward. where νλ= c / is the (temporal) frequency and W( xx , ,)ν is We assume that the complex amplitudes in the source plane have R 12 circular Gaussian statistics. This assumption allows one to use the the cross-spectral density of the beam in the detector plane. Gaussian moments theorem [1] for the fourth order correlations of the Let us consider two limit cases. ′ 1. When the exposure time T is so short and/or the radiation complex amplitudes Ux0 () of light in the source plane: 2 spectrum width ∆ν is so narrow that sinc [πν (12−≅ ν )T ] 1 over < **′ ′ ′ ′>= ′′ ′′ UxUxUxUx01() 02 () 03 () 04 () JxxJxx 013024 (,)(,) the support of WR ( xx12 , ,)ν , we have (2) +JxxJxx023(′′ , ) 014 ( ′′ , ), ≅ νν2 . (7) CTR,(, xx 12 )|∫ WR (, xx12 ,) d | where J(, x′′ x )=<> U* () xU ′ ( x ′ ) is the mutual intensity of 0ij 00i j This can be seen as the "temporal coherent limit", corresponding to ′ ′ light at two points, xi and x j , in the source plane. Equation (2) the case when the exposure time T is much shorter than the coherence applies, for example, to sources consisting of multiple independent time Tc of the beam incident on the detector (the coherence time is radiators, such as thermal light sources [1,2]. reciprocal to the width of the temporal spectrum, see the exact Under the conditions of paraxial illumination, we can express each definition below). This case may be applicable only to some complex amplitude in the detector plane via the Fresnel diffraction experiments involving imaging with high-quality single-mode laser integrals with the integration over the source plane: light or XFEL sources [1, 5, 6]. For cross-spectrally pure light, i.e. when W(, xx ,)νν= J (, xxs )(), with s()ν ≥ 0 and sd()νν= 1, exp(ikR ) iπ RR12 12 ∫ = − ′2 ′′, (3) UR () x ∫ exp (x x ) U0 ( x ) dx it is straightforward to verify that eq.(7) transforms into eq.(5). iRλ λR 2. When the exposure time is so long and/or the spectrum width is where λ is the radiation wavelength and k = 2/πλ. Evaluating 2 so broad that sinc[()]()/πνν12−TT ≅− δνν 12 (see Appendix the ensemble average for details regarding this approximation) over the support of < IxIx() ()>=< UxUxUxUx** () () () ()> with the RR12 RR 1212 RR WR ( xx12 , ,)ν , we have help of eqs. (2) and (3), one obtains (see e.g. [1, 2]): C(, xx )≅ (1/)| T W (, xx ,)|νν2 d . (8) < IxIx() ()>=< Ix () >< Ix () >+ | Jxx (, )|2 TR, 12 ∫ R 12 RR1 2 R1 R2 R12 (4) =<Ix( ) >< Ix ( ) > [1 + | gxx ( , ) |2 ], Note that the basic fact that the covariance should tend to zero when RR1 2 R12 the exposure time tends to infinity can be understood directly from the where Ix()= | Ux ()|2 , Jxx(, )=<> UxUx* () () and first line of eq.(6): the right-hand side tends to zero in this case because RRR12 RR1 2 for ergodic processes 1/2 T /2 g(,) xx= J (,)/ xx < I () x >< I () x > is the −12 2 R12 R 12 ( RR1 2 ) T| U ( xt , )| dt → <| U ( xt , )| > = I ( x ) for any t. ∫− R T →∞ RR corresponding complex coherence factor. Equations (1) and (4) imply T /2 that the autocovariance of light intensity in the detector plane is equal The case described by eq.(8) corresponds to the "temporal incoherent to limit", i.e. the case when the exposure time T is much longer than the coherence time Tc.