* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Download BASIC THEORY OF PARTIAL COHERENCE
Retroreflector wikipedia , lookup
Preclinical imaging wikipedia , lookup
Image intensifier wikipedia , lookup
Super-resolution microscopy wikipedia , lookup
Confocal microscopy wikipedia , lookup
Thomas Young (scientist) wikipedia , lookup
Chemical imaging wikipedia , lookup
Night vision device wikipedia , lookup
Johan Sebastiaan Ploem wikipedia , lookup
Phase-contrast X-ray imaging wikipedia , lookup
Image stabilization wikipedia , lookup
Fourier optics wikipedia , lookup
Optical aberration wikipedia , lookup
BASIC THEORY OF PARTIAL COHERENCE George B. Parrent, Jr. Technical Operations Research Burlington, Massachusetts INTRODUCTION In (1) V is the analytic signal associated with the optical disturbance, which we assume to be a single Cartesian component of the electric field vector. In terms of the mutual coherence function, the complex degree of coherence, 1'12 (T) is defined as The structure for a fundamental treatment of image formation problems already exists in the formalism of modern coherence theory as introduced by Wolf.l An adequate introduction to the subject is provided by Born and Wolf,2 (Chap. 10), and a detailed description of most of the results of the theory to date may be found in Beran and Parrent. 3 Therefore it will not be necessary to review the subject extensively here. Rather, we shall limit ourselves to a statement of the pertinent definitions and a summary of the treatment of the imaging problem in coherence theory. (3) It should be noted that the complex degree of coherence, like the mutual coherence function, is a function of seven variables, six position coordinates, and the time-delay coordinate T. The physical significance of these parameters is illustrated by the example discussed below. The treatment of problems involving partially coherent light involves the solution of the two wave equations: BASIC DEFINITIONS Mutual Coherence Function The basic entity in the theory of partial coherence is the mutual coherence function, r I2 (T), which may be defined by r I2 (T) == (~h~2,T) = <V(~ht) V*(~2,t + T» (s = 1,2) where V; denotes the Laplacian operator in the coordinates of the point Xs' A typical problem involves determining the mutual coherence in the source or object plane, solving (4) to obtain the mutual coherence on a later surface, such as the image plane, and then recovering the intensity, I, in the plane of interest from the relation (1) Here the underscore denotes position vector, the asterisk a complex conjugate, and the sharp brackets indicate a long time average, * i.e., 1 <J> == lim -2 T-oo T IT Jdt (4) (2) -T (5) * Equation (2) is equivalent to the definition introduced by Wolf, though in a slightly different form. Equation (5) follows directly from the definition of 17 From the collection of the Computer History Museum (www.computerhistory.org) !8 PROCEEDINGS-SPRING JOINT COMPUTER CONFERENCE, 1966 the mutual coherence function and the properties of the analytic signal. For a large class of problems the theory outlined in the preceding paragraph may be greatly simplified. These problems are characterized by the quasimonochromatic approximations, which are stated as {~;L\: :< I I} T where D.v is the spectral width. Of these two constraints, the second is obviously the more significant. White light may often be treated as quasimonochromatic if the path differences, CiT I, involved in the experiment are suitably small. In those circumstances for which the approximations above are applicable, the mutual coherence function may be replaced by the mutual intensity function, r(~I'-~2)' r (x 1 , ~ 2) == r 12 = r (KI , ~2 , 0) (6) The complex degree of coherence reduces to 1"12(0) == 1"12 and the wave equations (4) reduce to the two Helmholtz equations (s 1,2) (7) of these theorems. We shall reserve a discussion of the significance of the incoherent limit for a later point (a comprehensive treatment may be found in Beran and Parrent,3 Chaps. 2 and 3). The van Cittert-Zernike Theorem An additional theorem is required before attacking the treatment of the image formation probl{~m. The van Cittert-Zernike theorem may be stated as follows: The mutual intensity of the illumination derived from a distant incoherent source may be expressed in the form 2 ... i r(XI, X2) = J 1(~)eXR f'(~I-~2) d~ (9) Here I is the intensity distribution across the source, and R is the distance from the source plane to the observation plane. If the source is pl.aced- in the focal plane of a lens and the coherence of the emergent beam examined, it is found to follow the same law with the R replaced by the focallengthf THE IMAGING PROBLEM where Is:. is the wave number. Coherent and Incoherent Fields Equations (1) through (7) provide the basis of the theory of partial coherence as introduced by Wolf. To apply this theory to the imaging problem arid recover the familiar limiting forms, several the~rems due to Parrent are required. Principal among these are: 1. A field is coherent if and only if the mutual intensity function describing it can be factored in the form Review of Image Theory where V2UC~1) We may now direct our attention to the formulation of the general imaging problem. As will become clear in the following discussion, a basic description of image formation (at least as far as the lenses are concerned) already exists in coherence theory and, in fact, may be found in Refs. 2 and 3. This theory has not however been appli<;:d to the significant problems of image evaluation. Indt!ed, the theory has been applied to very few problems. In the next section the basic theory is outlined and those pertinent problems that have been solved are reviewed and discussed. + 2 k U(Kd = 0 (8) 2. An incoherent field ~annot exist in free space; however, an incoherent source consistent with this· result may be defined. (For the proof of these theorems and their extensions to polychromatic fields the reader is referred to Beran and Parrent. J ) Of particular significance for the problem of image evaluation is the second In coherence theory an object is described by its mutual intensity* (or mutual coherence) distribution rather than its intensity distribution. Thus the object described by rO(~l' ~2) and the relationship between object and image,-ri(~h.x2)' is developed by solving the two Helmholtz equations (7) subject to the appropriate boundary conditions. The gen*Our discussion in this section will be limited to quasi· monochromatic radiation. This serves to introduce the con(:epts, and at the same time keeps the development tractable:. From the collection of the Computer History Museum (www.computerhistory.org) 19 BASIC THEORY OF PARTIAL COHERENCE eral solution is (see Ref. 3, Chaps. 7 and 8): C(XI,-X2) = J J rO(~I' ~2)K(~1 - ~I) ·K*(~2 - ~2)d~ld~2 (10) Here K denotes the amplitude impulse response of the lens; i.e., denoting the complex transmission of the aperture by A (g), we may write K(~) = K ( f3 ) 2-.:i AI = J A(g)e a·fJ >./-- det (II) The two familiar limits may be recovered from (10) by using the theorems of the previous section. Thus, in the coherent limit, r l2 = VI Vf, and (10) reduces to CC~I' K2) = J VO(~I )K(~I - ~dd~1 .J V6' (~2 )K* C~2 - ~2 )d~2 (12) From (12) and theorem I ("Coherent and Incoherent Fields," above), it is clear that the image of a coherently illuminated object is coherent. A somewhat more surprising result (and certainly more interesting in the image evaluation problem) is obtained in the incoherent limit. Thus, we may take* r l2 = 1(~1 )O(~I - 6) to describe the object. The general image~Eq. (10), then reduces to C(~I'~2) = J 1(~)K(~1 - ~)K*(:K2 - ~)d~ (13) From (13) it is clear that the image mutual intensity is no longer of the same form as the object mutual intensity; i.e., the image of an incoherent object is not incoherent but is partially coherent. This result will be seen to have rather far-reaching implications in the problems of image formation. For most applications, the primary exposing radiation may be safely taken as incoherent. For example, sunlight is coherent only over a distance of approximately 1/20 mm. Thus, even a reconnaissance system which resolved an inch on the ground could probably be safely described by the incoherent limit of Eq. (10). In this case, the intensity in the image can be obtained by setting ~I = ~2 In (13); thus !i(~) = J 10 (f) 1 K(~ - ~) 12 d~ (14) Equation (14) will be recognized as the familiar incoherent imaging equation. The difficulty arises, of course, when the scale of the mutual coherence function becomes comparable with the resolution of the optical instrument. (This point will be discussed * Actually this form for the incoherent limit is only an approximation and must be used with care. However, it is sumciently precise to illustrate the present problem. at length in a later section.) While this condition is not likely to arise in the original taking system in the near future, it becomes serious in viewing and analyzing equipment such as microscopes, enlargers, and microdensitometers at the present state of the art. If one envisions improvements in taking equipment of a factor of two or more, it will become even more serious. This point will become clear as we analyze transilluminated objects. While (10) represents the general solution to the partially coherent imaging problem, a more useful form for application to spatial filtering is obtained by considering the object to be a transparency that is transilluminated. This is, of course, the case in almost all viewing of reconnaissance imagery, ~nd certainly in all uses of microscopes and microdensitometers in image evaluation. To describe this class of problems, it is necessary to describe the object in terms of its complex transmission t(~). For transilluminated objects Eq. (10) may be expressed as ri(~I'~2) = J JrO(~I'~2)t(~I)t*(~2) ·K(~I - ~1)K*(~2 - ~2)d~ld~2 (15) In most cases, one is interested in the intensity of the image, which may De obtained from (15) by setting ~I = ,!2. Thus, li(~) = J J ro(~], ~2)t(~dt*(~2)K(~ - {d ·K*(~ - ~2)d~ld~2 (16) In (15) and (16) rO(~I' 6) must be interpreted as the coherence of the illumInation incident on the transparency. The illumination in most cases of practical interest will be derived from a primary incoherent source. In this case rO(~I' ~2) takes a special form(because of the van Cittert-Zernike theorem): (17) That is, it becomes a function of coordinate differences only. Under these circumstances (16) becomes lie,!) = J J rO(~1 - ~2)t(~1 )t*(~2)K(~ - ~) ·K*(~ - ~2)dgldg2 (18) From (18) it is clear that for transilluminated objects the transition from object intensity 1 t(~) 12 to image intensity is nonlinear. The significa-nce of this conclusion is that the customary image evaluation techniques and criteria are not, in general, applicable to such systems. For example, knowing how such a system images sine waves or edges does not permit us to describe how it images other objects. Furthermore, the same optical system could From the collection of the Computer History Museum (www.computerhistory.org) 20 PROCEEDINGS-SPRING JOINT COMPUTER CONFERENCE, 1966 be expected to yield different results if the coherence of the illumination varied. At high resolutions a small variation in the scale of the coherence function can produce dramatic; changes in the image. This may account, in part, for the difficulty encountered in intercalibrating instruments in different laboratories, or in the cross-checking of microdensitometers that have essentially equivalent optical components but produce different results in edge trace analysis. Since systems of this type are inherently nonlinear, it is impossible to characterize them by a transfer function. This point is easily established by taking the Fourier transform of both sides of (18). Thus, i~) = J i({})7*(~_ - Q){J r[~ K(~ - ~)K*(~)dqJdf1 - (~ + ~)] v v /1 z, 0(. Figure 1. Coordinate system. The geometry is illustrated in Fig. 1, in which ~ is the coordinate in object space, a is the coordinate in the aperture plane, and x is the coordinate image space. Assuming paraxial optics and ignoring obliquity factors, we may express the rdation between object and "image" as follows (Beran and Parrent,3 Chaps. 3 and 7): fi(XI, X2) = J JJJ rO(~I' ~2)R(al )R*(a2) (19) In (19) the inner integral is characteristic of the instrumentonly, while the factors t({3) and t* (Il - (3) are determined solely from the object spectrum. However, (19) is not in the form of "object spectrum times transfer function equals image spectrum." The inner integral has been referred to as a generalized transfer function, but that nomenclature is rather misleading since the function is not used as a transfer function at all. A better terminology is the more cumbersome one introduced by Wolf, the "transmission cross coefficient," which emphasizes that it is a function of two frequencies. With these general reservations in mind, we may direct our attention to the development of the system analysis for spatial filtering systems. Here R (a) describes the transmission of the aperture and all integrals are infinite. The term a~2 /2/ is, of course, the saggital approximation and the r's remain from the Green's function. Assuming coherent quasi-monochromatic radiation, we find that rO(~1,6) UO(~I)U~(~2) (21) = U i (XI)U1(X2) (22) = and the image becomes r i (XI,X2) where Ui(x) = JJ Uo(~)R(a)e ik ~U.a)- ~ +r(a.x)] 2/ SYSTEM ANALYSIS In this section the relationships between "object" and "image" for three cases of imaging with coherent radiation are derived. Denoting by ZI and Z2, respectively, the object and image distances, we define these cases as follows: d~da (23) f d K (~ - a)4 . · we may wnte I gnonng terms 0 or er Z3 r(~, a) = ZI + (~ _ a)2 (24) 2z1 and l.~+~=~ ZI 2. 3. Z2 ZI = Z2 ZI = 0, (25) / =/ Z2 = f Condition (1) produces an image in the ordinary sense only if the object is: in the near field of the lens. Condition (2) yields an "image" which is the Fourier transform of the object, and condition (3) yields a Fourier transform multiplied by a quadratic phase term. Hence, omitting constant phase terms we may rewrite (23) as . J R(a)e L+ ~ a(X ~ ~ )] [~ (~+ z.~ _/~) +2z, 2z. - ;;- z: ik 2 z, From the collection of the Computer History Museum (www.computerhistory.org) ·dad~ (26) 21 BASIC THEORY OF PARTIAL COHERENCE I Case (1): Here an ideal lens is assumed again. Provided a, x, and ~ and the condition / Under these conditions (26) becomes U;(X) = J Uo(~) J 'k [r aE x' R(a)e 2z,- z,-+ 2z. o a2 » ax] z.- da J Uo(~)e 2aEl 2z, k d~ (27) is met, the inner integral yields a constant C and (37) becomes' Consider first the integral ;k[r - 2/ ;kEx ;ka' d~ e -zz,- U(a) = (28) Here U;(x) = C J Uo(~)e -V d~ (39) ;k(E-a)' U(a) = J Uo(~)e-2Z'-d~ (29) Equations (28) and (29) are obtained by simply completing the square in the exponent. We may now write (27) as Uj(x) = J R(a) U(a)e;kG:.- ::- ;:,J da i.e., a Fourier transform with no quadratic phase term. Z1 = 0 Z2 = / Under these conditions we have immediately from Eq. (20) Case (3): (30) Or completing the square again we have ikx' Ok .[Z' Z~] ~+~ U;(x) I X J R(a) U(a)e 2z, e U;(x) ;kx'z, e 2Jz. da ;k(a+ J R(a) U(a)e (31) (32) ;k(a+z;x)" ikx'z, f a de¥. 2z1 -a U(a)e (33) Under the condition (34) i.e., object in near field of lens, the limits -a to a (the aperture size) may be regarded as infinite and (33) may be evaluated by the inversion theorem for Fresnel transforms, giving V,(x) - elk"g«,,<,) Vo (:: x) (35) that is, an image mUltiplied by a quadratic phase term. Case (2): Zl = Z2 (40) or (41) da If the lens is unapodized and unaberrated, (32) becomes U;(x) = e 2Jz. ikax = e 2T J Uo(a)R(a)e -21 da ;;-xr 2z, (38) or Spatial filtering systems are properly constructed around the configuration characteristics of case 2. That is, this setup is used to display the Fourier transform of the input transparency. A second transparency is placed in the Fourier transform plane to modify the spectrum. Then the process is repeated and the "filtered" image is displayed in the final Fourier transform plane. In each spatial filtering experiment the coherence conditions should be adjusted in accordance with the foregoing analysis in order to be sure that the approximations of the theory have been satisfied and at the same time to minimize the degrading effects associated with coherent imaging. These effects arise primarily from the fact that such systems are nonlinear in intensity. Some examples of these effects are shown below. EXPERIMENTAL EXAMPLE OF COHERENCE EFFECTS =/ Under these conditions, (26) reduces to Ui(x) = J J Uo(~)R(a)e ;k(a' + E' +x' - 2a(x + E)l 2J da d~ (36) Completing the square on the exponent in (36) gives U;(x) = f _lkEx I.a Uo(~)e E-J Ik(a-(x+E»)' e -a 2J da d~ (37) A simple experiment that illustrates the effects of the spatial and temporal coherence is to form twobeam interference fringes by division of a wavefront. Figure 2 shows the result of such an experiment. Figure 2a shows high-contrast fringes formed with a He-Ne gas laser illuminating a pair of small circular From the collection of the Computer History Museum (www.computerhistory.org) 22 PROCEEDINGS-SPRING JOINT COMPUTER CONFERENCE, 1966 Figure 2. a b c d Effect of coherence length. A and B-with gas lasers; C and D-with mercury arc. apertures~ The envelope function is the diffraction pattern of the single aperture. In Fig. 2b, a piece of plane optical quality glass 0.5 mm thick was introduced in front of one of the apertures only, to add an extra optical path. Again illuminating with the He-Ne gas laser, we observe no difference in the fringe contrast. However, when the experiment is repeated with a coherent field produced by a mercury arc lamp without the glass plate, highcontrast fringes are again seen (Fig. 2c), but with the glass in place, the fringes disappear (Fig. 2d). The slight scale change between the two pairs of illustrations results from the different wavelengths (6328 A for He-Ne and 5461 A for the Hg green line). This illustration shows that the coherence length of the mercury arc radiation is quite small. Both fields were spatially coherent but the coherence lengths were quite different. Introducing fine ground glass across the pair of pinholes resullts in the intensity distribution of Fig. 3. The c!xtra paths introduced by the ground glass did not exceed the coherence length; hence, high-contrast fringes are seen over the whole field. A discussion of these types of speckle patterns in terms of their autocorrelation function and their power-sp,ectral density are to be found in a paper by Goldfischer. 4 An attempt to build a coherent projection printer is reported by Milinowski,5 in which a rotating pi(~ce of From the collection of the Computer History Museum (www.computerhistory.org) BASIC THEORY OF PARTIAL COHERENCE Figure 3. Two-beam interference with diffusing plate. ground glass is used to remove some of the coherence effects. REFLECTED LIGHT The different speckle patterns formed when coherent light is reflected from a rough surface have been commented upon a number of times and perhaps form the most objectionable feature of coherent imaging by reflected light (as opposed to transmitted light discussed in the last section). Figure 4 shows a standard bar target that has been printed on a matte photographic paper and then Figure 5. Figure4. 23 Photograph in reflected coherent light. rephotographed in reflected coherent light. The edge-ringing effects are masked by the speckle patterns. The speckling is produced by the interference between the scattered light and is determined by the coherence length of the incident radiation. Figure 5a is a photograph of a portion of a cementblock wall illuminated by a mercury arc so that the light is spatially coherent. In Figure 5a the same portion of the wall is illuminated by a gas laser; the speckling completely obscures any structure of the wall. Both beams had approximately the same spatial coherence, but the gas laser has a considerable longer coherence length. a b Effect of coherence length. A-spatially coherent only; B--spatially and temporally coherent. From the collection of the Computer History Museum (www.computerhistory.org) 24 PROCEEDINGS-SPRING JOINT COMPUTER CONFERENCE, 1966 REFERENCES 1. E. Wolf, Proc. Roy. Soc., vol. (A) 230, p. 246 (1954). 2. M. Born and E. Wolf, Principles of Optics, 2nd ed., Pergamon Press, New York, 1964. 3. M. Beran and G. B. Parrent, Theory of Partial Coherence, Prentice-Hall, Englewood Cliffs, N.J., 1963. 4. L. I. Goldfischer, J. Opt. Soc. Am". vol. 55,· p. 247 (1965). 5. A. S. Milinowski, ibid, vol. 54, p. 1406 (1964). From the collection of the Computer History Museum (www.computerhistory.org)