You are on page 1of 17

Chapter

Optical Waveguide Theory


2.1 Introduction

An optical waveguide is dened as any dielectric structure that is capable of transporting electromagnetic energy. They are characterized by a central region of refractive index (the core - nco ) which is higher than that of the surrounding region of normally uniform index (the cladding - ncl ). In general, the light guidance maybe treated as a consequence of the total internal reection of the light from the interface between the core and the cladding. In this chapter, some basic concepts of optical waveguides will be briey reviewed, in particular, that of the 3-D rectangular waveguides.

2.1.1

Waveguide Parameters

The propagation characteristics [5] of waveguides can be expressed in terms of the parameters discussed in the following subsections. Figure 2.1 shows some of the typical waveguide cross-sections that will be encountered in this thesis.

2.1 Introduction

ncl nco 2

ncl nco 2 x (a) (b) 2 y

ncl nco

(c)

Figure 2.1: Typical waveguide cross-sections: (a) step-index square-core buried channel
waveguide (BCW); (b) rectangular core with the characteristic dimension dened by the geometrical mean = x y . (Diagram adapted from Ladouceur & Love (1996) [5])

Characteristic Dimension The characteristic dimension, , of the waveguide cross-section is dened as the linear dimension that quanties the transverse variation in index of the waveguiding structure. As illustrated in Figure 2.1, corresponds to the half-side of the squarecore waveguide and the geometrical mean of the half-side lengths of the rectangular core waveguides.

Relative Index Dierence The relative index dierence or relative prole height, , is dened as:

n2 n2 co cl 2n2 co

(2.1)

where nco is the core index and ncl is the uniform cladding index of the waveguide. In the case of weakly guiding waveguides, where nco ncl and hence, << 1, the above expression can be approximated by:

2.1 Introduction

nco ncl nco

(2.2)

Waveguide Parameter The degree of guidance or waveguide parameter, V , is a dimensionless quantity, dened as: 1 2 2 nco 2 = (n2 n2 ) 2 co cl

V =

(2.3)

where refers to the characteristic dimension, is the prole height and is the operating wavelength of the source of excitation. The quantity (n2 n2 ) 2 is also co cl commonly known as the numerical aperture.
1

Normalized Index Prole The refractive index prole n(x, y) may also be expressed in a more convenient normalized form as shown below:

n2 (x, y) = n2 [1 2f (x, y)] co

(2.4)

where is the prole height and f (x, y) is the normalized index prole.

2.1.2

Fresnel Equations: Propagation of Light across Media

The propagation of light from one medium to another is governed by the laws of reection and refraction. As illustrated in Figure 2.2, the two media with dierent

2.1 Introduction refractive indices are separated at the boundary given by the plane y = 0. In this case, we assume a linearly-polarized plane wave from the half space y > 0 impinging upon the boundary. The plane of incidence is the x y plane as shown. The two dierent cases are discussed below:

y Ei Hi n1 n2 t Ht
Figure 2.2: Reection and refraction of an electromagnetic wave at a boundary between
two dielectric media. The illustrated wave is polarized with the E vector parallel to the plane of incidence. (Diagram adapted from Dun (1990) [6])

Er i r Hr x Et

Transverse Electric (TE) polarizations In the case of TE -polarized light, the electric vector is perpendicular to the plane of incidence (i.e. the E-eld is in the z-direction). The reection coecient r (the ratio Er /Ei ) and the transmission coecient t (the ratio Et /Ei ) are given as: cos i ( n2 ) cos t Er sin(i t ) n1 = r = = n2 Ei cos i + ( n1 ) cos t sin(i + t )

(2.5)

2.1 Introduction

10

t =

2 cos i cos i + ( n2 ) cos t n1

(2.6)

Transverse Magnetic (TM) polarizations In the case of TM -polarized light, the magnetic vector is perpendicular to the plane of incidence (i.e. the B-eld is in the z-direction). Similarly, the reection coecient r and the transmission coecient t are given as:

r =

n2 n1

cos i cos t n2 ( n1 ) cos i + cos t

tan(i t ) tan(i + t )

(2.7)

t =

( n2 ) cos i n1

2 cos i + cos t

(2.8)

2.1.3

Critical Angle & Total Internal Reection

As shown in Figure 2.3(a), a plane wave incident at an angle 1 on two semi-innite media with refractive indices n1 and n2 , such that n1 > n2 will result in a plane reected beam in medium 1 and a transmitted beam in medium 2. The angle of the transmitted beam is dened by Snells law :

n1 sin 1 = n2 sin 2

(2.9)

At the critical angle, 1 = c , the transmitted wave is parallel to the interface (i.e. 2 =
) 2

as shown in Figure 2.3(b). The critical angle [8] for the boundary

separating two optical media is dened as the smallest angle of incidence, in the medium of greater index, for which the light is totally reected.

2.1 Introduction

11

Incident Wave

Reflected Wave

1
n1 n2 Interface

Incident Wave n1 n2

Reflected Wave Interface Boundary Wave

Transmitted Wave

(a)
Incident Wave n1 n2

(b)
Reflected Wave Interface

(c)
Figure 2.3: Reection and refraction of a parallel beam: (a) Incident light on an interface between two media; (b) Incident light on an interface at a critical angle; (c) Total reection beyond the critical angle. (Diagram adapted from Syms & Cozens (1992) [7])

c = sin1 (

n2 ) n1

(2.10)

For an interface between a polymer (e.g. PMMA) and air, where n1 = 1.490 and n2 = 1.00, the critical angle is 42.1 . However, there will be no critical angle when the wave is incident from the lower index side (i.e. if the values of n1 and n2 are exchanged).

For, 1 > c , Total Internal Reection (TIR) occurs as shown in Figure 2.3(c). TIR is really total in the sense that no energy is lost upon reection. However,

2.1 Introduction in any real devices utilizing this property, there will always be small losses due to absorption in the medium and due to reections at the surfaces where the light enters and leaves the medium. Nonetheless, TIR presents a mechanism for conning a wave in one region of space.

12

2.1.4

Mode Propagation Constant & Eective Indices of Modes N

Consider the case of the guided mode in the ray-optics picture of a 2-D waveguide as shown in Figure 2.4(a). In this case where s < < 90 , the light is conned in the guiding layer by the total internal reections at both the upper and lower interfaces and propagates along the guide in a zig-zag path.

Cover/Cladding Guide/Core Substrate

ncl

nco ns

(a)
ko n co

kx

kz =

(b)
Figure 2.4: (a) Zig-zag wave picture of modes propagating along a 2-D waveguide in
ns the case of the guided mode where s = sin1 ( nco ) < < 90 . (b) Wave vector diagram.

(Diagrams adapted from Nishihara (1985) [9])

In the wave-optics picture, the modes are characterized by their propagation constants. As shown in Figure 2.4(b), the plane wave propagation constant in the

2.2 Weak Guidance Approximation wave-normal direction is dened as k0 nco where k0 =


2

13 and is the free space

wavelength of light. The propagation constants along the x and z directions are:

kx = k0 nco cos kz = k0 nco sin =

(2.11) (2.12)

For lossless waveguides, kz = where is equivalent to the mode propagation constant in an innite medium with an index of nco sin . Hence, the eective indices of modes, N, is dened as:

N=

= nco sin k0

(2.13)

The eective index, N, species the ratio of the wave velocity in a vacuum to that in the propagating direction (z ) of the waveguide. Hence, the guided mode propagating along the z direction sees the eective index N.

2.2

Weak Guidance Approximation

Maxwells equations (or the vector wave equations) for electro-magnetic elds are the basic starting point of any optical waveguiding analysis. However, due to the coupling of the three scalar components of both the electric and magnetic eld vectors, the equations become very complex and as a result, there are only a few exact analytical solutions [10].

In this work, our waveguides normally have a slight variation of the refractive index over their cross-sections (i.e. 0, the core and cladding indices become very similar) due to the nature of the fabrication process which will be discussed in the

2.2 Weak Guidance Approximation respective chapters. This slight variation in the index allows the vector equations for the electromagnetic elds to be replaced by a single scalar equation involving only a single component of the electric eld, with negligible loss of accuracy. This simplication is known as the weak-guidance approximation [11, 12].

14

2.2.1

Vector Wave Equation

For a source-free, dielectric media, Maxwells equations can be expressed in the following form [10]:

E = ik(
2

o
o

) H ) E
1 2

1 2

(2.14) (2.15)

H = ikn (

o
o

using rationalized MKS units, where

and o are, respectively, the free space


2

dielectric constant and permittivity; k =

is the free-space wavenumber; is

the wavelength of light in free space; and n is the refractive index. By taking the curl of Equation 2.14 and using Equation 2.15 to eliminate the magnetic eld, the vector wave equation for the electric eld is obtained:

E k 2 n2 E = 0

(2.16)

The refractive index prole is needed to determine the propagation along the waveguide. Assuming, for simplicity, that the waveguides are isotropic, non-absorbing and translationally invariant, the refractive index prole merely exhibits transverse dependence (n = n(x, y)). This allows the E and H eld vectors to be written in separable forms [5]:

2.2 Weak Guidance Approximation

15

E(x, y, z) = e(x, y) exp(iz) exp(iwt) H(x, y, z) = h(x, y) exp(iz) exp(iwt)

(2.17) (2.18)

where e and h are vector expressions which contain the transverse dependence of the electric and magnetic elds respectively; z refers to the accumulated phase change at a distance z along the waveguide in terms of the propagation constant ; and iwt refers to the monochromatic time dependence in terms of the source frequency w and time t.

Equations 2.17 and 2.18 may be further re-expressed to distinguish the longitudinal scalar components of the e and h elds (denoted as ez (x,y) and hz (x,y)) and the corresponding transverse vector components e (x, y) and h (x, y).

e(x, y) = e (x, y) + ez (x, y) z h(x, y) = h (x, y) + hz (x, y) z

(2.19) (2.20)

where the refers to the unit vector parallel to the z axis. Based on the above z denition for the elds, Equation 2.17 may be substituted into Equation 2.16 to give a more suitable form:

+ k 2 n2 (x, y) 2 )e = (

+ i)e z

ln n2 (x, y)

(2.21)

where the divergence condition vector Laplacian operator.

(n2 E) = 0 has been used and

is the transverse

Using Equation 2.19, the e eld can be decomposed into its respective transverse and longitudinal components:

2.2 Weak Guidance Approximation

16

+ k 2 n2 (x, y) 2 )e = ez = i

ln n2 (x, y)
2

(2.22) (2.23)

e + (e

) ln n

(x, y)

Despite the simpler form of the wave equation (Equation 2.22), it is impossible to obtain any analytical solution for the transverse component of the vector electric eld for the types of proles and cross-sections encountered in planar waveguides. A numerical solution (for the vector wave equations) would also be complicated due to the implicit coupling of the components of the eld in this equation. Fortunately, a simplication based on the weakly-guiding nature of these waveguides oers a solution to this dilemma.

2.2.2

Scalar Wave Equation

To perform a perturbation expansion of the transverse vector wave equation (i.e. Equation 2.22), this equation is rst re-expressed in terms of the normalized index prole f (x, y), the prole height , the degree of guidance V and the characteristic dimension as discussed earlier in 2.1.1:

(2

+ U 2 V 2 f )e =

ln n2

(2.24)

where the normalized core mode parameter U is related to the propagation constant by:

U = (k 2 n2 2 )1/2 co

(2.25)

Considering as the perturbation parameter, the transverse eld components and the normalized propagation constant is expanded in a power series:

2.2 Weak Guidance Approximation

17

e () = e + e + 2 e + ... U () = U + U (1) + 2 U (2) + ...

(1)

(2)

(2.26) (2.27)

where the superscript indicates the order of the perturbation and the tilde denotes the zeroth-order solution. The right hand side of the normalized vector wave equation (Equation 2.24) is also expanded using Equation 2.4, where the expansion of the logarithmic term for a small argument is:

ln n2 = 2

42

+ ...

(2.28)

Substituting the expansions (Equation 2.26 and Equation 2.28) into the normalized vector wave equation (Equation 2.24 and equating the terms of the same order in yields the scalar wave equation (i.e. the zeroth order eld):

+ U 2 V 2 f )e = 0

(2.29)

where the zeroth-order approximation of the propagation constant is related to U through:

U = (k 2 n2 2 )1/2 co

(2.30)

In Cartesian coordinates, the two scalar components of e decouple such that they obey the same scalar wave equation:

+ k 2 n2 (x, y) 2 )(x, y) = 0

(2.31)

where (x, y) refers to either ex (x, y) or ey (x, y).

2.3 Guided Modes in 3-D Waveguides

18

2.3

Guided Modes in 3-D Waveguides

3-D waveguides or channel waveguides are the cornerstone to light switching and modulation. In 3-D waveguides, the connement of light transversely (along the x-direction - i.e. along the depth) and laterally (along the y-direction) are essential. Unlike microwave rectangular waveguides with perfect conductor walls, optical 3D waveguides surrounded by dierent dielectric materials do not support pure TE and TM modes. Instead, two dierent types of hybrid modes are supported. These are essentially TEM modes polarized along the x and y directions, where its classication is dependent on the direction of the main electric eld component.
x The mode having the main electric eld Ex , is called the Epq , while that having y the main electric eld Ey , is called the Epq . The subscripts p and q denote the x y number of nodes of the E eld in the x and y directions, respectively. Epq and Epq

are also generally known as the TM -like mode and TE -like mode, respectively.

In general, the boundary value problem for the 3-D waveguide cannot be accurately solved without the use of a computer. For the analysis of these waveguide modes, one can make use of several approximation and numerical methods. Some of the approximation methods are: Marcatilis method [13] and the eective index method (EIM ) [14]; while some of the numerical methods include: fourier decomposition method (FDM ) [15], nite element method (FEM ) [16, 17] and the beam propagation method (BPM ) [18]. Nonetheless, there are no exact analytical solutions for the modes of the channel waveguides, even in the limit of the weak-guidance approximation. One has to resort to numerical methods to obtain accurate solutions of the scalar wave equation for waveguides with rectangular or square core cross-sections. In this academic exercise, a commercial optoelectronic CAD software BeamP ROP T M (from RSoft Inc.), which is based on the beam propagation method (BPM), was used to design the SU-8 y-branching waveguides.

2.3 Guided Modes in 3-D Waveguides

19

2.3.1

Beam Propagation Method (BPM)

Computer-aided design and modelling software has in many ways spurred the development of lightwave components and systems. With these softwares, new device concepts can be easily evaluated; designs can be optimized and the design cycle can be shortened signicantly. One such commercial optoelectronic CAD software is BeamP ROP T M from RSoft Inc.. This simulation software is based on the beam propagation method (BPM). The basic principles of BPM will be described in this section.

The beam propagation method (BPM) was rst applied to problems of integrated optics by Feit and Fleck [1823]. The BPM approach is based on the approximation of the exact wave equation for monochromatic waves and the numerical solution of the resulting equations. BPM oers many simplications to the guided wave problem, thus reducing the computational complexity and speeding up the computational process of many problems. An overview of the basic BPM theory [24, 25] is described below.

In the basic form of BPM, the electric eld is represented as a scalar value instead of a vector value. As a result, polarization eects can be neglected and the propagation is assumed to be paraxial (i.e. conned to a narrow range of angles). The scalar eld assumption allows the wave equation to be written in the form of the Helmholtz equation for monochromatic waves:

(x, y, z) + k(x, y, z)2 = 0

(2.32)

where

(x, y, z) is

2 x2

2 y 2

2 ; z 2

the scalar electric eld has been written as

E(x, y, z, t) = (x, y, z)eiwt , and k(x, y, z) = k0 n(x, y, z) is the spatially dependent wavenumber, with k0 =
2

is the wavenumber in free space. It can be seen that

2.3 Guided Modes in 3-D Waveguides the refractive index distribution n(x, y, z) denes the geometry of the problem. It should be noted that Equation 2.32 is exact.

20

For typical guided wave problems, the phase variation due to propagation along the guided axis (i.e. the z axis) gives the most rapid variation in the eld . To enhance the eciency of the technique, this rapid phase variation is factored out of the problem with the introduction of a slowly varying eld u, where:

(x, y, z) = u(x, y, z)eikz

(2.33)

where k is the reference wavenumber, a constant representing the average phase variation of the eld ; and is frequently expressed in terms of a reference refractive index n, where k = k0 n. In this approximation, the slowly varying eld can be represented numerically on a longitudinal grid that is much coarser than the wavelength for many problems, thus increasing the eciency of the technique. However, this approximation can only be applied to paraxial elds (i.e. the eld is propagating mainly along the z axis).

Substitution of Equation 2.33 into Equation 2.32 yields the following equation for the slowly varying eld:

2 2 2u u + u + u + (k 2 k 2 )u = 0 + 2ik z 2 z x2 y 2

(2.34)

Assuming that the variation u with z is suciently slow, the second derivative term in z may be neglected. This is known as the paraxial or parabolic approximation, which reduces the guided wave problem from a second-order boundary value problem requiring eigenvalue analysis, to a rst-order initial value problem that can be simply solved by integration. One direct consequence of the elimination of the

2.3 Guided Modes in 3-D Waveguides second derivative is that devices where reections are signicant (i.e. backward travelling wave solutions) cannot be accurately modelled based on the basic BPM equation. Equation 2.34 then becomes:

21

i u = z 2k

2u 2u + + (k 2 k 2 )u x2 y 2

(2.35)

Equation 2.35 is the basic BPM equation in three dimensions (3-D) which is also known as the scalar, paraxial BPM. This basic form of BPM models the continuous wave (CW ) optical elds propagating in the z-direction. By discretizing the cross-section of a waveguide structure into grid points, the cross-sectional prole is calculated one slice at a time using the above equation (Equation 2.35) in the z-direction, with each successive slice being mathematically dependent on the current slice. This process is repeated until the wave has propagated through the entire structure.

2.3.2

Further Extensions to the Basic BPM Theory

The basic BPM theory may be extended in many ways to address the above limitations. For example, vectorial BPM calculations may be performed by incorporating the polarization eects in the BPM (i.e. treat the E eld as a vector and derive the equations based on the vector wave equation). Wide-angle BPM calculations [26] may be performed to address the paraxial restrictions on the basic BPM. The paraxial limitations may be reduced by incorporating the eects of the second order derivative (that was neglected in the basic BPM) according to the dierent degrees of approximation. Lastly, bidirectional BPM [27] may also be incorporated to handle simultaneous propagation along the negative z axis based on a transfer matrix approach. More details on these can be found in [28, 29].

2.4 Summary

22

2.4

Summary

Some of the basic equations and the fundamental concepts on optical waveguide theory were discussed in this chapter. These include: the vector wave equations; the Fresnel equations; the concepts of critical angle and total internal reection; the mode propagation constant and the eective indices of modes. In the case of 3-D waveguides, there are no exact analytical solutions for the modes. Numerical methods must be employed to obtain an accurate solution for these waveguides. Softwares based on BPM are very popular in the design and modeling of 3-D waveguide components such as y-branches. More details of the design and modeling of channel waveguides will be provided in the subsequent chapters. In the next chapter, the common waveguide fabrication techniques for 3-D polymer and glass waveguide fabrication will be reviewed. In addition, proton beam writing, a new technique for 3-D waveguide fabrication, will be introduced.

You might also like