The Three-Dimensional (3-D) Transmission Cross-Coefficient For TCC Imaging

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 4

100, No.

4 (1995) 155-158 © Wissenschaftliche Verlagsgesellschaft mbH, Stnttgart 155

The three-dimensional (3-D) transmission cross-coefficient


for transmission imaging
e. 1. R. Sheppard, M. Gu
Department of Physical Optics, School of Physics, University of Sydney, Australia

diffracted light with diffracted light, and the problem be-


'I1le~()-D)~i~oefI"lCient fer
tndIsRIissioo ~ Thettam;mission~s-coefiicient for comes much easier to consider. The TCC reduces to the
three-dinI~m;i01lIll imaging in a <lonventional transmission- weak-object transfer function (WOTF), which is a func-
mOde micrOscope is examined.· This is also compared with tion of three-variables, being the spatial frequencies in the
the corresponding transmission cmils-coefiicient for a canfo- three dimensions. This case was analyzed by Streibl [3].
cal transmission system. However, if we retain the bilinear terms from interference
Der~~(3D)~~td·· of diffracted with diffracted light, we must use the full
IItr~~ DerTransmiilSIonme_oeffizient fiit TCC treatment, which involves six variables, two spatial
die dreid:bnenst~nale Bi141ibertrapng in ei:nem konventio- frequencies for each of the three dimensions. Although
nellen' nansmission~kop wir4 untersooht.& wi(d the form of the 3-D TCC has been given previously [4],
aoohmit dem entspreChenden. Transmissionsbeuzkoeffi-
zienten ei:nes konfoklllen Transmissions.Systems verglicben. it is difficult to interpret the results. Sheppard and Mao
[4] reported some interesting properties: one is that 3-D
imaging in a microscope with infinitely large effective
source does not behave incoherently.
1 Introduction An aim of our present investigation is to compare the
3-D imaging properties of confocal and conventional mi-
In two-dimensional (2-D) imaging in a microscope, the croscopes. For weakly scattering objects, such that the
degree of partial coherence depends on the aperture of image is predicted by the WOTF, a true confocal micro-
the condenser (i.e. the effective source), relative to that of scope (i.e. one in which the confocal pinhole is infinitesi-
the objective. For a small condenser aperture imaging is mally small) behaves identically to a conventional micro-
coherent, whilst for a very large condenser aperture imag- scope. It also behaves identically to a scanning micro-
ing is incoherent. Between these two limiting cases imag- scope with a very large detector. However as we shall
ing is essentially partially coherent [1]. This is so in gen- show, the TCC is different for conventional and confocal
eral even for the case of equal condenser and objective systems. Interestingly, the WOTF for a confocal micro-
apertures, when for the particular case of an object con- scope with a finite-sized detector behaves differently from
sisting of two points separated by the Rayleigh resolution either a true-confocal <;>r a conventional microscope [5].
distance the image is identical to the incoherent image.
Calculation of images in partially coherent light is con-
siderable more complicated than calculations for either
coherent or incoherent imaging. A variety of methods are 2 The TCC for a conventional transmission microscope
available, includihg direct integration of the contribu-
tions from points within the object, with inclusion of the
appropriate partially-coherent cross-product terms, inte- The TCC is a function of six spatial frequencies: m1 , m2
gration over the effective source, and integration over in the x-direction; nI' n 2 in the y-direction and SI' S2 in
pairs of spatial frequencies using the transmission cross- the z-direction. We denote these vectors by m 1 , m 2
coefficient (TCC).
When it comes to consider three-dimensional (3-D) m = mi + nj + sk . (1 )
imaging, the problem becomes more difficult. If the first Consider a microscope system in the scalar, paraxial ap-
Born approximation is satisfied, i.e. we neglect multiple proximation. Then the image intensity at point position
scattering and depletion of the incident beam, the object vector v can be written [4]
can be considered in terms of its scattering potential [2],
related to the complex refractive index variations. The J(v) = Sf C(m 1 ;m 2 ) T(m 1 ) T*(m2)
3-D object spectrum is then the 3-D Fourier transform of . exp [i(ml - m 2) . v] dml dm2 (2)
the scattering potential. If the strength of the scattered where T(m) is the 3-D object spectrum, and C(m 1 ;m 2 ) is
light is very weak, then we can neglect interference of the 3-D TCe. For simplicity we restrict our attention to
objects which are unchanging in the y-direction, so that
n 1 = n 2 = O. Then the TCC is [4]
Received February 27, 1995.
C(m 1 ;m 2 ) = Sf 1~(~,'7W Pz(~ - m 1 ,'7)
C. 1. R. Sheppard, Min Gu, Department of Physical Optics,
School of Physics, The University of Sydney, NSW 2006, Aus- . Pz*(~ - m2, '7) b(sl - m 1 ~ + tmi)
tralia. 'b(s2-m2~+tm~)d~d'7 (3)
156 C. 1. R. Sheppard, M. Gu, The three-dimensional (3-D) transmission cross-coefficient for transmission imaging

By considering the behaviour in the third quadrant, we


can thus write for both the first and third quadrants

C("2l

In the fourth quadrant, m 1 is positive and m2 is negative.


C("2l
The Tee can be written

Fig. 1. The cut-off and symmetry ofthe 2-D TCC C(m) ;m z) for
a partially coherent system with equal condenser and objective
apertures.

where PI' Pz are the pupil function of the condenser and


objective respectively, which can be written

C(m 1;m 2) = -l_SI IP1(';,'1W Pz('; - m 1,'1)


m1 m2
. Pz*(.; - m 2 , '1)
1
J ( .; - (~11 + ~ ))
for
m1 82< 1 + -m 2
- <- (10)
2 m2 2
.J(';_(~22 +~2))d';d'1' (4)
i.e.
The delta functions represent straight lines though the
area of overlap of the pupil functions, so that the Tee
only has appreciable magnitude when the two delta func-
tions coincide. This implies the condition
and
(5)

The area of overlap of the pupil functions is well-known


in the theory of partially-coherent imaging in two-dimen-
sions [1]. If the spatial frequencies are normalized appro-
priately, the cut-off in m 1 , m 2 space is shown in fig. 1 for
the case of equal condenser and objective apertures. The
Tee is non-zero inside the hexagonal shape shown. In
the first quadrant of fig. 1 the integral in eq. (4) depends
on the larger of the quantities m 1 and m 2 • Thus

C(m 1 ;m 2) = -2- J ((- +m2)


m 1m2
- 82
m2 2
- (8-1+ml))
m1
-
2 for

(14)

I.e.

1821 m2)2} ' m2 > m1· m2 8 m1


- 1 + m1 < -2 < -
. Re { 1- ( m2 + 2 (6) - -
2 m2 2
(15)
C. J. R. Sheppard, M. Gu, The three-dimensional (3-D) transmission cross-coefficient for transmission imaging 157

For the first condition 8 2 1m 2 is positive, whilst for the is also known as the weak object transfer function
second condition 8dml is negative, so that the Tee can (WOTF), and has been described elsewhere [3, 5].
be written
2 l<5
C(m l ;m 2)=lm 1Im
1 ml) - (8m2+2m2))
((8~+2 2) 81 = 82 = 0
2 l 2 For this case, corresponding to a thick object with no
variations in the axial direction, one solution follows
.Re{ 1_(~+~)2},
Im21 2
from putting 81 = 0 in eq. (21) to give

C(ml' 0; 0) = _2_ Re
Imll
{Vi - mi}.
2
(22)

There is also a corresponding solution C(0;m2' 0). A fur-


ther solution is obtained from eq. (7)

C(ml ,O; m2,0) = Iml~m21 <5(m2 - ml ) Re {Vi - ~i}.


(23)
Thus components occur only for ml = m2, m l = 0, or
m2 =0.
Similarly for the second quadrant

C(m l · m 2) =
,
2
Iml11m21
<5 ((-
m2
82+ -m2)
2
- (8- 1+ -ml))
m 2 l
The condition m l = m 2 implies also 81 = 82 from the
. Re{ 1_ (~+
Imll
!5l)2} ,
2
delta function. We obtain from eq. (4)
2
m2 8 m C(m,8 1;m,8 2)=-I-<5(8 2 -8 1)
- < ~ < 1 + -2. (18) mI ~---y---:-_--=..,.
2 ml 2
18 I m2)2}
. Re { 1 - ( I~ I + 2 . (24)
_ 2
Iml11m21
<5((~
m l
+ ml) _
2
(~+
m
2
m2))
2 This describes the strength of the constant component in

1_(~+~)2},
the image arising as a difference signal from a single
.Re{ spatial frequency.
Im21 2
m 8 m 4) m l = m2 = 0
- 1 + - 2 < - 2 < -l. (19)
2 m2 2
For m l = m 2 = 0 we need to go back to eq. (3) to give
The value of the Tee for various special cases can now
be examined. C(O, 81; 0,8 2) = SS Il't (~, 1JW .1 ~(~, 1JW <5(8 1) <5(82) d~ d1J
= n<5(8 1) <5(8 2) (25)
1) m 2 =0
The behaviour along the axis m 2 = 0 can be determined 5) m l = m, m2 = - m
from the limiting cases of the Tee in quadrants 1 and 3
or 2 and 4. From quadrants 1 and 3, For equal and opposite-signed values ofml' m 2 we obtain
from eqs. (16-19)
2 ( m~ 81m2 mlm2)
C(m l ;m2) = Imll <5 82 + 2 - --;;;-;- - -2- 2
C(m,8 1; - m,8 2) = -<5(8 1 + 82 + Iml)
Iml
(~+ ImlI2)2}
C~: + 1;ly},
. Re { 1_ (20)
Imll 2 . Re{ 1-
and thus from the delta function we have that 82 must
also equal zero. Thus Iml 8 Iml
- 1+- < - 2 < - - (26)
1~11 Re{ G~llll + Im~12y}.
2 Iml 2
C(ml;O) = 1- (21)
2
= -<5(8 1 + 82 + Iml)
In order to examine the behaviour from quadrant 4, the Iml
C~: + 1;ly},
behaviour for 81 negative follows directly from eq. (15),
but for positive 81 we must go back to eq. (6) as 8 2 1m 2 is . Re{ 1-
indeterminate. The final result agrees with eq. (21), so
Iml 8 Iml
that there is no discontinuity in the Tee. The Tee - 1 + - < - 1 < - -. (27)
C(m l ;0) describes the imaging of a weak object so that it 2 Iml 2
158 C. 1. R. Sheppard, M. Gu, The three-dimensional (3-D) transmission cross-coefficient for transmission imaging

This expression cuts off when Im I = 1, for which value Sl = S2 = 0 the Tee is non-zero only if m 1 = m2 or
Sl = S2 = - t· m 1 = 0 or m 2 = 0, whereas for confocal imaging all possi-
Once the 3-D Tee has been developed, the 2-D Tee ble pairs of spatial frequency within the hexagon on fig. 1
for transverse imaging of thin objects and the l-D Tee are imaged. In addition we note that for m 1 = m 2 ,
for an axially-varying line object can be calculated by s 1 = S2' corresponding to high transverse spatial frequen-
integration [6]. For the 2-D Tee the behaviour in the cies, the transfer function for the confocal system cuts off
first and third quadrants is straightforward. The 3-D at two, whereas for the conventional system it cuts off at
Tee is either a function of m 1 , Sl or m 2 , S2 and hence the one, so that these components are enhanced. A combina-
2-D Tee is also a function ofm1 or m 2 • In the second and tion of these effects accounts for the improved image of,
third quadrants the behaviour is more complicated. We for example, a strong point object in a confocal micro-
must evaluate scope.
As we claimed in the introduction, the Tee can be
(28)
used to model images, based on the first Born approxima-
over the Tee. The Tee for the fourth quadrant (and tion. In this case we include in the Tee treatment the
similarly for the second quadrant) can be broken up into bilinear terms, but neglect second and higher order terms
two regions, given by eqs. (9) and (12). We have in the treatment of the interaction with the object. This
approach will not then be v~lid to the second order, but
nevertheless we believe that consideration of the Tee is
a useful method for appreciating the imaging properties
of an optical system. Further, we have claimed [4] that the
Tee treatment does not strictly rely on the assumption
of the first Born approximation. It is applicable provided
+ const (29) that the scattering with the object can be described by a
function of the three spatial frequency coordinates [4].
so that This is rigorously true for some particular geometries of
C(m 1;m 2) object or optical system [7]. There is also the possibility

= 2{COS-1 (m2 ~ m1) _1m2 ~ m11 V 1 _ (m2 ~ m1)2}


for improvement of the scattering model based on, for
example, assumption of the Rytov [8, 9] approximation,
or the second Born or Rytov approximations.
(30)
which is a function of m 2 - m 1 only, as is well known
from the theory of partially coherent imaging. Acknowledgment
The authors acknowledge support from the Australian Re-
3 The TCC for a confocal transmission microscope search Council, and the Science Foundation for Physics within
the University of Sydney.
The behaviour of the confocal microscope is much sim-
pler. Because it behaves coherently, the Tee separates
C(m 1,m 2) = c(m 1) c*(m2) (31) References

where c(m) is the 3-D coherent transfer function (eTF) [1] H. H. Hopkins: On the diffraction theory of optical images. Proc.
given by Roy. Soc. Lond. A217 (1953) 408-432.
[2] E. Wolf: Three-dimensional structure determination of semi-trans-
c(m) = IS Fi (~, 1]) ~(~ - m,I]) o(s - m~ + tm2) d~ dl]. parent objects from holographic data. Opt. Commun. 1 (1969) 153-
(32) 156.
[3] N. Streibl: Three-dimensional imaging in a microscope. 1. Opt. Soc.
For a microscope with equal circular lenses, we thus have Amer. A2 (1985) 121-127.
[4] C. 1. R. Sheppard, X. Q. Mao: Three-dimensional imaging in a micro-

I~I Re { G~II + 1~12y},


scope. 1. Opt. Soc. Amer. A6 (1989) 1260-1269.
c(m) = 1- m =F 0 [5] M. Gu, C. 1. R. Sheppard: Three-dimensional partially-coherent im-
age formation in confocal microscopes with a finite-sized detector. 1.
Mod. Opt. 41 (1994) 1701-1715.
= noes), m = O. (33) [6] C. 1. R. Sheppard, M. Gu: The significance of 3-D transfer functions
in confocal scanning microscopy. 1. Microsc. 164 (1991) 377-390.
[7] J. T. Sheridan, C. 1. R. Sheppard: Modelling of images of square-wave
4 Discussion gratings. Opt. Commun. 105 (1994) 367-378.
[8] V. I. Tatarski: Wave propagation in turbulent media. McGraw-Hill,
We see that the Tee for confocal imaging covers a range New York 1961.
of spatial frequencies which are not imaged in a conven- [9] C. 1. R. Sheppard: General considerations of diffraction theory of
tional system. For example, for conventional imaging if 3-D imaging. Eur. 1. Cell. Biology, Suppl. 25, 48 (1989) 29-32.

You might also like