You are on page 1of 7

Eur. J. Phya. 11 US%) 121-127. Primlea In tne I .

121

Examples of Fresnel diffraction using Gaussian modes


J A Murphy and A Egan
Experimental Physics Department, St. Patricks College, Maynooth, Co. Kildare, Republic of Ireland Received 12 August 1992, in final form 2 December 1992 Abstract. Gaussian beam modes, which are normally associated with laser cavities and quasi-optical propagation, can be used to solve some classic problems in Fresnel diKraction. A particular advantage of the technique is that, for the cases of the uniformly illuminated circular aperture, narrow slit and straight edge, it is possible to derive analytical recursive relationships for the mode amplitude coefficients, which allow for speedy computation of the diffraction pattems.
RCsumC. Les modes de faisceaux gaussiens, qui sont normalement associ6s avec les cavit& lasers et la propagation quasi-optique, peuvent etre utilisis pour koudre quelques probldmes classiques dans la diffraction de Fresnel. Un avantage particulier de la techniques dans les cas dune ouverture circulaire uniformbment klaide, une fente &troite,et un demi plan, est quil est possible dhoncer des relations recursives analytiques pour les coefficients damplitude du made. ce qui permet un calcul numCrique rapide des figures de diffraction.

1. Introduction

The usual approach followed in physical optics when solving Fresnel diffraction probiems is to apply the Fresnel integral transform (Goodman 1968, Klein and Furtak 1986). Even in the case of simple situations, such as the uniformly illuminated circular or rectangular aperture, it is not possible to obtain an analytical description of the diffraction pattems resulting, except in the Fraunhofer limit, and one is forced to rely on quasiquantitative constructs such as the Cornu spiral and Fresnel zones (Klein and Furtak 1986, Born and Wolf 1980). An alternative approach presented here is based on what are known as Gaussian beam modes (Goubau 1969, Siegman 1986, Hauss 1984). In this view a monochromatic spatially coherent beam, represented by the complex scalar field E(x,y , z) is considered to he made up of a linear sum of complex independently propagatingmodes $,(x,y,z), these being solutions of the wave equation (assuming paraxial propagation) that maintain the same amplitude profile as they diffract. These modes can be regarded as the free space analogue of waveguide modes. At any point the field E(x,y, z) can be expressed simply hy E(x,y, z ) = &4j$j(x,y,z), where the Ai are constants. If the E field is known on some transverse plane (e.g. a uniformly illuminated aperture in an opaque screen), the mode coefficients,A,, can be determined by perfonning the appropriate overlap integral between E and $ j over that plane; describing the E field elsewhere is then just a question of propagating the individual

modes and re-summing (analogous to Fourier decomposition and synthesis.) Gaussian beam modes were first introduced in the field of laser optics (Kogelnik and Li 1966). More recently, Gaussian beam mode analysis of beam propagation has been widely and successfully applied in coherent millimetre-wave (and submillimetrewave) optical systems (Wylde 1984, Lesud 1990, Murphy 1988, Withington and Murphy 1992). The approach has the advantage that often in such systems the beam has approximately a simple Gaussian profile, and thus can be considered as consisting of a single mode (the fundamental), the simplest possible scenario. However, if the beam is only poorly approximated by a simple Gaussian then many modes have to be considered for an accurate description of the beam. Nevertheless, it has been realized that in such cases the modal approach still offers a powerful picture of beam propagation. For example, (thin) lenses are trivial to deal with, as they cause the same phase curvature transformation for all modes and do not disturb the mode amplitude profiles; the mode coefficients, A , , thus remain unchanged. This is a much simpler case than for aperture integration techniques, where at each lens in an optical system one would have to re-evaluate the diffraction integrals. Also, the effect of any truncating aperture or stop can be thought of as scattering power between the component modes of the beam. In fact a scattering matrix formalism analogous to quantum mechanics can be used with the beam represented by a vector of mode amplitudes (Padman-aEd, Murphy 1991). The

122

J A Murphy and A Egan

modal approach has thus been found to be both conceptually and computationally superior to diffraction integral techniques in the analysis of complex millimetre-wave optical systems. It is therefore pertinent to ask whether the modal approach to diffraction can be applied to the classic problems of Fresnel diffraction encountered in undergraduate courses, and this is what we aim to demonstrate in this paper. Furthermore, we show that a particular advantage of the modal technique is that for the three fundamental examples in Fresnel diffraction, the uniformly illuminated circular aperture, narrow slit and straight edge, it is possible to derive analytical recursive relationships for the mode amplitude coefficients. This enables diffraction patterns to be determined computationally in a very efficient manner without the need for any numerical integrations. The approach is therefore interesting for those teaching optics using the methodology of computational physics. Since beam mode amplitude profiles maintain the same mathematical form with propagation (apart from some width scaline,factor). we can determine the beam modes approprise to the Fresnel approximanon bv lookina for eieenfunction tvne solutions of the Fresnel tran;fonnagon. Altem&ely, the beam modes can also be derived from the wave equation by assuming paraxial propagation, since it can be shown that this is equivalent to the Fresnel description of diffraction (Hauss 1984, Siegman 1986). These points are discussed in section 2. In section 3, as referred to above, the modal approach is applied to Fresnel diffraction at a circular aperture, slit and straight edge, and analytical recursion relationships for the beam mode coefficients are derived. As will be discussed, describing the Fresnel diffraction pattem to a high precision is then a question of summing a sufficientlylarge number of modes.
2. Derivation 01 the Fresnel transform beam modes

r:

, for beam propagation between the xo,yoplane at z and the x.y plane at z = zo + L. To generate a modal description we rewrite the mode in a functional form @(x, y, z ) = u(x,y; W(z), R(x))exp( -jkz), which is appropriate for all z as the beam propagates, Using the fact that (Gradshteyn and Ryzhik 1980):
&'-YP

'q(M)dr= z y l

-$ ' )"I2

(where H.(x) is the Hermite polynomial of order n), one can show that beams which are described by the functional form:

x exp( jkz)

(3)

are one complete set of such modes. Wand R are the beam width parameter and phase front radius ofcurvature, respectively, both functions of z. If we choose the origin z = 0 to be where the modes are of narrowest extent in the x,y plane (a position known as the mode waist position), then Wand Rare given by:

(4)

We generate a set of modes which will act as a basis set for a spatially coherent monochromatic beam by searching for the eigenfunctions of the Fresnel integral transform. This approach is taken by Siegman (1986) and Klein and Furtak (1986), for example, in deriving the simplest such mode, the fundamental, which has a simple Gaussian profile. For beam propagation along the z axis, the Fresnel transformation can be written as (e.g. Siegman 1986):

W Ois called the waist radius and, clearly, R = to there. The term & , , known as the phase slippage (with respect to a plane wave propagating in the z direction), is given by:

These Hermite-Gaussian beam modes are orthonor*,&. For mal in the sense that: JJ@& m.nl dxdy = a, convenience we define the normalized Hermite-Gaus' ) by: sian function h,(x; U h,(x; W )=

J W

a( 4 ; )

eXP( -

5)

(I)

and this is the form that we shall use hereafter.

Examples Ot Fresnel diffraction using

Gaussian

modes

123

The phase slippage term turns out to be very important in the understanding of how an arbitrary beam amplitude profile changes shape as it propagates. Effectively different modes have different phase velocities. If a field is not a pure mode, but rather consists of sum of modes, the relative phase between different component modes varies along the axis of propagation, since the phase slippage terms are both functions of z and also mode number dependent. This results in the amplitude distribution of the composite field altering shape with z, which is expected from diffraction theory. Beam modes can also be derived by searching for mode-like solutions to the time independent scalarwave equation: V2$ ! 2 $ = 0. In a well defined beam with a narrow opening angle in the far field (the paraxial approximation) the diffraction process will not be noticeable over distances o f a few 2. Thus, the beam will behave approximately like a plane wave, exp(-jkz), with some slowly varying amplitude u(x,y,z). For such beams we thus write $ = u(x, y, z) exp( -&), and neglect the second partial derivative of U with respect to z. The scalar wave equation then reduces to the following second order partial differential equation for U:

xexp[-jk(z+&)+j(2n+
x tan-

1)

I)(

n W

where Ln(x)is the Laguerre polynomial of order n (as defined in Gradsbteyn and Ryzhik 1980). These modes form the complete orthonormal LaguerreGaussian beam-mode set. Ifwe wish to choose the set of modes with which to describe the propagation of some arbitrary field, we are free to choose Wand R for the component-mode set. If we fix on some Wand R, then the distance Az to the waist position, and the waist radius WO = W(0). are determined by:

a% 7 a% , aU -+ 24- = 0 . ax2 ay az

(7)

Ar =

This approach is the one normally taken in deriving beam modes for laser cavities and beam waveguide a r t i n quasi-optical systems (Kogelnik and Li 1966, M and Lesurf 1978). Since the paraxial approximation is equivalent to the Fresnel approximation (Hauss 1984, Siegman 1986), the set of Hermite-Gaussian beam modes is the same as the set of eigenfunction solutions of the Fresnel transformation. If one deals with a problem which has a rotational symmetry then a beam can be expanded in terms of the appropriate mode set for cylindrical polar coordinates (r, e). The two-dimensional Fresnel transform in terms of x,,y, reduces to the Hankel-like onedimensional integral transform:

R 1 + (AR/nW2)

It is best to choose Wand R so as to describe the be& to a given accuracy with as few modes as possible. We can quantify the goodness of fit by calculating the total power associated with the finite modal sum and comparing it with that of the real beam being approximated (this will be discussed in the next section). If to the desired accuracy the beam is described by the minimum number of modes, then clearly the diffraction patterns can be calculated with minimum number of computational operations. In general the expansion for the beam a1 any point (beyond a diffracting aperture) can be written in the form:

for the HermiteGaussian modes, or


x 2nJ,(kr0r/L)rodr0.

(8)

Then using that (Gradshteyn and Ryzhik 1980):

W , x )=

=O

1 A $ ! , ( r , z ; W(-%R(zD.

(14)

Iom =-

B ~ ~ ~ , ( = ~ (P ~ e -~ u 2~ i 4 s( ~ , . ) ~

2$+

(9) one can show that the normalized eigenfunctions of the Fresnel transformation for cylindrically symmetric fields turn out to be:

for the Laguerre-Gaussian beams, in the case of cylindrically symmetric beams. Since the two beammode sets are orthogonal and complete, the amplitude coefficients Am,or A,, can be simply determined by performing the integrals:

A, =

il E ( X , ~ , Z ~ ) [ $ ~ ( ~Wz.), ,Y,~ R(Z,)ll* ~;


x dxdy

(15)

124

J A Murphy

and A Egan determined the amplitude coefficients for n up to 2000. The value of WO= W(0)can he chosen arbitrarily; thus, the A, calculated are actually functions of W?/a, the ratio of the waist radius to the aperture radius. The intesity attern is given by f(r,z) cc IE(r, z)I2 = &4&&,z)I . Thus, if one wishes to use just a few modes to approximate the diffracted beam, then one needs to choose a value for Wo/a that optimizes Z:lA,1', since the power associated with each mode is proportional to IAJ'. If a large number of modes are present in the summation then the cboise of Wo/ais less critical. For example, for a total number N = 500, approximately 99.9% of the power in the beam is included when Wo/a = 0.11 (these were the values used for . the examples to be discussed below.) We have chosen to look at exaipies which have special significance in terms of the traditional Fresnel zone approach to diffraction. A maximum (or minimum) in intensity is predicted, as one moves along the axis of propagation of the beam, at those points where there are exactly an odd (or even) integer number of Fresnel zones contained within the aperture boundary. The position of an extremum (maximum or minimum) corresponds to values of z = 2 / N r L , where Nr is the total number of Fresnel zones filling the aperture. We notice that these points are related to the fundamental mode phase slippage factor 6 ,= tad(nW'/%R), through the following relationships:

or
A, =

JJ E(~',Z~)I$!,(~'.~~; Wz.),R(z0))l*2nr'dr'

(16)

respectively, assuming that E i s known over the transverse plane at z = 2 . . The power of the modal discription is that, once the beam mode amplitudes have been determined, following the evolution of the diffracting beam with z is just a question of evaluating the modal sum, which is fast computationally. Thus, ope does not have to perform a whole series of computationally intensive numerical integrations, as would be the case if one were to use the Fresnel transfonnation directly.

3. Examples
3.1. Fresnel diffraction at a circular aperture

Consider a circular aperture in a screen placed in the path of a plane wave E ( x , y , z ) = E,exp(-jkz). We determine the mode coefficients at the circular aperture, where the E field is well defined. The diffracted field E, will possess cylindrical symmetry, and thus is expanded as a sum of Laguerre-Gaussian modes, making the usual approximation that the aperture size is large enough so as not to affect the field in the aperture appreciably. Since the incident plane wavefront radius of curvature is infinity, we choose R for the modes to be infinity, implying that the common mode waist position (where we choose z = 0) coincides with the aperture in the screen. If the aperture has radius U , the A , are then determined by:

x exp[-(')]~nrdr

where we have assumed the amplitude of the incident plane wave E, is unity. Using the recursion relationships for Lagueme polynomials:

d Z;[L"(X) - L"+l(x)l=

U)

(19)

the following recursion relationship between the A , is obtained

Thus, to determine the A. we need to just calculate A,, which is easy analvticallv, and all other values of A. are determined r&ursiv;ly This approach we have round to be robust compuiationally and we have

Since it is the phase slippage term that essentially determines the amplitude (or intensity) profile changes as the beam diffracts, it is not surprising that there should be this relationship between Fresnel zones and phase slippage. The Fresnel diffraction pattems shown in figures l(a) to I(d) are for values of z = 0, 2/6?.,2/31 and 2 d / & respectively. For each pattern the intensity (which for convenience we define to be IE(r,z)l*) is plotted as a function of the parameter r/a (the edge of the geometrical shadow of the aperture is at r/a = 1). As predicted an on-axis zero in intensity occurs when (d/Lz) is an even integer (e.g. figure l(b)). Similarly, an on-axis maximum occurs when (a'/%z)is an odd integer (e.g. figure I(c)), with an intensity four times that which would be present in the absence of the diffracting screen (consistent with the Fresnel zone construction). Figure l(a) shows the reconstructed fields at the uniformly illuminated aperture itself, z = 0, while figure I(d) corresponds to z/(d/l) = 2.0, and we see that the field distribution begins to resemble the appropriate farfield A ~ N Dattem. Some hieh freauencv .rineing can be seen in ;he apenure recon;truction figure I(a); this IS due to the finite number of modes used (500 in this
I

Examples of Fresnel dinraction using Gaussian modes


I.Zi

125

. .

. .

4.0,

Flnure 1. Fresnel diffraction Dattern at a circular aoerture for cases where z is equal to (a) 0, (b) a2/6 (c) a2/s
( d j2a'I.l.

case) and is related to high frequency ringing and the Gihbs phenomenon observed in Fourier series reconstruction.
3.2. Fresnel diffraction at a slit and straight edge The diffraction pattern produced by a slit in a screen, or a straight edge, placed in the path of a plane wave E(x,y,z) = Eoexp(-jkz) can he regarded essentially as a one-dimensional problem. The most convenient modal set to use is the Hermite-Gaussian set, and thus we just consider an expansion of the form:

the following recursion relationship for the A, is obtained (form odd):

We determine the mode amplitudes at the plane of the aperture, since the incident fields are well defined there. If the slit has width d = 2a, and we set the amplitude of the incident plane wave equal to unity, the A, are determined by:
A, =

s" h,(x;
-0

(26) Only even numbered modes have non-zero amplitudes because of the symmetry of the problem. Figure 2 shows a series of Fresnel diffraction patterns produced by a uniformly illuminated slit. In the examples shown here 500 modes were used with Wo/a= 0.13, ensuring 99.9% of the power is included in the description. The plots are parametrized in terms of the usual dimensionless coordinates associated with the Cornu spira! (e.g. Goodman 1968) q = x m and AV = For comparison, the examples correspond to some of the values of AV in figure 7.17 of Klein and Furtak (1986). In the case of the straight edge, we assume that the edge lies at the axis of propagation, such that the Screen blocks the beam for all values of x < 0. In that case the mode amplitudes for the one-dimensional Hermite-Gaussian set, A,, are determined by:

m.

W,)dx

(23)
A,

(on setting R = so for modal set). Using the recursion relationships for Hermite-Gaussian functions:

=Io

hm(x;",)dx.

(27)

The followingrelationship between the A, is obtained


Am+I =&.(O:W,)+&A,-l.

(28)

Figure 3 shows a plot of the Fresnel diffraction pattern due to a straight-edge calculated in this way. The off-axis distance is parametrized in terms of q =x m (as for the slit discussed above). The pattern is appropriate for all values of z , as long as z is much smaller than d2/L,where d is the extent of the

126

J A Murphy and A Egan


2.0,

. .

. . , . . . . .

2.0r,

. .

. . .

. 7 and (d) 2.4, Figure 2. Fresnel diffraction panern at a slit for cases when? A V is equal to (a) 10.1, (b)6.2,(c) 3

approximation to the infinite plane wave provided by the finite total number of modes in the sum (of order fi6). Some high frequency ringing can be seen in the pattern; again this is due to the finite number of modes used (500 also in this case.)
4. Conclusions

slippage occurs between modes as they propagate. In the case of a uniformly illuminated circular aperture, slit and straight edge, it was shown that the mode coefficients could be related through analytical recursive relationships. Reconstruction of the beam by modal summation is achieved by simply propagating the modes to the appropriate plane, and a number of examples were presented.

In this paper the analysis of Fresnel diffraction problems by the use of Gaussian beam modes was discussed. These modes, which are normally associated with quasi-optical systems, were shown to be eigenfunctions ofthe Fresnel transformation, their amplitude profiles remaining unchanged during diffraction, apart from a width scaling factor. Diffraction of an arhitary beam profile, which can be considered as a modal sum with constant mode coefficients,was explained by the fact that the modes have different phase velocities, so that relative phase
Figure 3 . Fresnel diffraction pattern at a straight edge.

Acknowledgments
The authors are indebted to Stafford Withington and Rachael Padman for many useful conversations, and to Francisco Perales for translation of the abstract, We would also like to thank EOLAS and St. Patrick's College, Maynooth, for financial assistance.

References
Born M and Wolf E 1980 Principles of Optics 6th ed (Oxford Pergamon) Goodman J W 1968 Introduction I O Fourier Oplics (New York McGraw Hill) Goubau G 1969 Millimeter and Submillimeter Waves ed F A Benson (London: Iliffe) Ch 19 Gradshteyn I S and Ryzhik I M 1980 Table o/lnregmlr. Series and ProducIs (New York Academic) Hauss H A 1984 Wavei and Fie/& in Optoelectronics (Englewood Cliffs, NJ: frentice-Hall) Klein M V and Furtak T E 1986 Optics 2nd edn (New York: Wiley) Kogelnik H and Li T 1966 Proc. IEEE 54 1312-29 k u r f J C G 1990 Millimetre-wave Oplics. Devices and System (Bristol: Adam Hilger) Martin D H and Lesuri 1 C G 1978 Infared Phys. 18
~ ~

405-12 . .

Murphy J A 1988 IEEE Transactions Anrennas Propagat.


AP-36 510-5

Examples 01 Fresnel diffraction using Gaussian modes Padman R and Murphy 1 A 1991 Proc. IEEIURSI 7th ConJ on Antennas and Propagation (York) (London: IEE) pp 201-4 Siegman A E 1986 h e r s (Mill Valley, C A University

127

Science Books) Withington S and Murphy J A 1992 IEEE Transacrions


Antemar plopogn:. M - 4 0 198-206

Wylde R J 1984 Proc. IEE

H 131 258-62

You might also like