1926

JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 13, NO. 9, SEPTEMBER 1995

An Efficient Numerical Approach for Determining the Dispersion Characteristics of Dual-Mode Elliptical-Core Optical Fibers
Tzong-Lin Wu, Student Member, IEEE, and Hung-chun Chang, Member, IEEE
Abstract-Based on vectorial formulations which combine the surface integral equation method and the finiteelement method, a novel numerical approach is proposed for calculating the dispersion coefficients of dual-mode elliptical-core fibers with arbitrary refractive index profiles. By differentiating the original formulations involving the propagation constant p and the guided mode fields H , and H , once and twice with respect to the normalized frequency V , the new formulations for { d p / d V , d H , / d V , d H , / d V } and for { d 2 p / d V Z d , 2 H , / d V Z ,d 2 H y / d V 2 } are obtained, respectively. Once {p,H , , H , } is solved through the eigenvalue procedure which dominates the computing time, only a few matrix manipulations are required to obtain { d p l d V , d H , / d V , d H y / d V ) and { d 2 p / d V 2 , d 2H , / d V 2 , d2H , / d V 2 } . Some numerical examples are examined to see the influence of different refractive index distributions with dips on the dispersions of the four nondegenerate LPll modes for elliptical-core fibers.

wavelength, the difference in the dispersion among the four nondegenerate LP1 modes, which include the first-higherorder modes of 2 and y polarizations (LP;';e,", LPyTt) and the second-higher-order modes of 2 and y polarizations (LPi'f;, LPyfi), is significant. Rigorous vectorial waveguide theory is thus needed in order to investigate the dispersions of the four nondegenerate LPll modes for the e-core fibers. Some related analysis results have recently been reported by Poole et al. [6]. In this paper, a novel numerical approach based on the vectorial waveguide theory is proposed to efficiently calculate the dispersions of the e-core fibers. The computation of dispersion coefficient requires the first and second derivatives of the propagation constant with respect to wavelength. A simple way to evaluate the dispersion is the direct numerical calculation, based on finite differences, of the first and second derivatives from the propagaion constant versus wavelength data. However, such simple method I. INTRODUCTION requires large computational efforts and may result in great UAL-MODE elliptical-core (e-core) optical fibers have errors [7]. Some improved numerical approaches based on been used extensively in many sensors and optical dethe scalar theory for circular optical fibers, such as that vices such as intermodal couplers, modal filters, and acoustousing the Rayleigh quotient to obtain the first derivative of optic frequency shifter [ 11-[3]. Recently, a chromatic disthe propagation constant [81, that solving three differential persion compensation technique that employs the dual-mode equations which include the first and second derivatives of fiber with small ellipticity was proposed by Poole et al. the propagation constant [9], and that based on the matrix [4]. This technique is based on the principle of converting perturbation method [lo], have been proposed. Besides, based the LPol mode from the communication fiber operated near 1.55 pm to the LPll modes in an e-core compensating on the variational finite-element formulation of the scalar fiber and using the large negative dispersions of the LPll theory, an algorithm [ 111, in which formulas for the first and modes close to their cutoff wavelengths to compensate for second derivatives of the propagation constant with respect the dispersion in the communication fiber. It has been shown to V are explicitly derived by intrinsically differentiating the that the dispersion of the LPll mode for the circular-core scalar wave equation with respect to V and by using a reaction compensator is sensitive to the variation of the refractive formula, has been presented for calculating the dispersions index profile using a scalar waveguide theory [5], but, to our of optical fibers. This algorithm which eliminates the need knowledge, the dispersion of the higher-order modes for the e- of numerical differentiation and is efficient in computational core compensator has not been studied theoretically in detail. effort can only treat the dispersion of circular-core fiber with Because the e-core compensator is operated near the cutoff circularly symmetric refractive index profile (1-D problem). In this paper, stemming from the same principle as in Manuscript received November 17, 1994; revised April 7, 1995. This work [ l l ] and based on vectorial formulations which combine was supported in part by the National Science Council of the Republic of China Grant NSC83-0417-E002-009 and in part by Telecommunications the surface integral equation method (SIEM) and the finiteLaboratories, Ministry of Transportation and Communications, Republic of element method (FEM) [12], a novel numerical approach is China, Grant TL-83-5101. proposed for calculating the waveguide dispersions of e-core T. L. Wu is with the Department of Electrical Engineering, National Taiwan fibers with arbitrary refractive index profiles (2-D problem). University, Taipei, Taiwan 106-17, Republic of China. H.-C Chang is with the Department of Electrical Engineering and the First, the propagation constant p and the guided mode fields Graduate Institute of Electro-optical Engineering,National Taiwan University, H , and HY are solved by the combined formulations of Taipei, Taiwan 106-17, Republic of China. IEEE Log No. 9413045 SIEM and FEM and the continuity requirement of E, and

0733-8724/95$04.00 0 1995 IEEE

Authorized licensed use limited to: National Taiwan University. Downloaded on February 23, 2009 at 04:10 from IEEE Xplore. Restrictions apply.

WU AND CHANG DISPERSION CHARACTERISTICSOF DUAL-MODE ELLIPTICAL-CORE OPTICAL FIBERS

1927

H , at the core-cladding boundary. Second, new formulaY tions about h,, hy }, where the dot denotes differentiation with respect to the normalized frequency V, with the solved {p, H,, H y } being substituted are derived by differentiating -. the original formulations with respect to V. Matching the continuity of fi, and fi, at the boundary, {b,hx.?hy} is obtained. Similarly, the formulations about fi,, H y } with the solved {p,H,, H y } and (8,f i x ,Hy} being substituted can be derived by differentiating the original formulations twice with respect to V and B:, By} is obtained by matching the continuity requirement of E, and fi, at the boundary. When p,p, and are obtained, the dispersion can be calculated by the definition. Most of the computational efforts are spent in solving {p,H,, Hy}. Only some matrix manipulations are required in obtaining {~,h,,hy} and once Fig. 1. Elliptical-core fiber with a and b being the major and minor axes, respectively. {p,H,, H y } is solved. The detailed mathematical formulations about {p, H,, H y}, value integral with the singularity at {b,Az,fiy},and {p,BZ,By} are presented in Section II. the Cauchy principal -+ The solution procedure described by matrix representation is the point of r' = r' being removed. How the P s integral discussed in Section 111. In Section IV, as a numerical example, is approximated will be described in next section. Through the influence of the refractive index distributions, including the integral equation. ( l ) , the transverse magnetic field at an different cy-power profiles and the index dip effect, on the arbitrary position r' in the cladding region can be described dispersions of the four nondegenerate LPll modes for the e- by the field on r and its normal derivative along A. When core compensators is investigated. The vectorial mode field F ( 3 is chosen as the field on the boundary I?, a complete patterns of these four modes are also presented. Section V relation between F and dF/dn on r for the cladding region gives the conclusions. is established in (1). In the inhomogeneous core region with the relative permittivity distribution E,(x,y ) , the magnetic fields of the guided 1 1 . FORMULATIONS modes satisfy the following source-free equation As described in [12], based on a full-wave formulation, a combined method employing the surface integral equation method and the finite-element technique can treat the propagation characteristics of inhomogeneous optical waveg- By making the dot product of the left-hand side of (2) with which is independent of uides with arbitrary cross-sections. Although similar numerical an arbitrary vector function lfc, and integrating the scalar product over the entire space, techniques have been proposed for optical waveguides with 3, arbitrary cross-sections [13], [14], they can only treat the the differential equation. (2) can be transformed into the variational-equation formulations. After some manipulations propagation characteristics of homogeneous waveguides. Fig. 1 gives the sketch of an inhomogeneous dielectric as shown in [12], the variational equations with the arbitrary waveguide (core region) embedded in a homogeneous sur- fields I f c being chosen as H:? and H;& respectively, are rounding medium (cladding region) with an elliptical crossaH, aH; aH, aH: (IC& - P ~ ) H , H ;- --- -section. In the homogeneous cladding with the core-cladding ax ax ay ay boundary I?, the transverse magnetic field F(= H, or H y ) and its normal derivative dF/dn(= dH,/dn or d H y / d n ) on I? can be related through a surface integral equation and aH, aH; aHyaHi / ( k g c c - p 2 ) H H c - --- -ax ax ay ay

{p,

t *

{p,

{p,

{fi,&,ay}

where kd is related to the wave number in free space ko and the propagation constant of the guided mode by kd = (p2- kid)1/2 with ~d being the relative permittivity of the homogeneous cladding, r' and 2 are the position vectors in the 2-D vector space with 2 on r as shown in Fig. 1, G = (112~)KO(kdlF - 21) with KO being the modified Bessel function of the second kind and order zero denotes the 2-D Green function in the homogeneous cladding, and P denotes

By making a variation of (3a) and (3b) with respect to H; and H i , one obtains the differential equation of H, and H y in (2). For guided modes, the boundary condition that E, and H , are continuous at the boundary r should be satisfied. From the Maxwell equations j w & = V x I ? and V 3 = 0, E, and H , can be expressed as

Authorized licensed use limited to: National Taiwan University. Downloaded on February 23, 2009 at 04:10 from IEEE Xplore. Restrictions apply.

1928

JOURNAL OF LIGHTWAVE TECHNOLOGY,VOL. 13, NO. 9, SEPTEMBER 1995

and

+/ ( l k ~ k ~ 2/3p)HyHGdxdy ~ , =0
and that a/aV(jwE,) and a / a V ( j p H , ) on pressed as

(6b) can be ex-

where E = E O E , or E = EoEd with EO being the permittivity of the vacuum, H , and He are the magnetic fields normal to and tangential to the boundary, respectively, and d / d n and denote the partial derivatives along the normal and the tangential directions, respectively. Explicitly, H,, = H, cos(B)+Hysin(8) and Hl = H, sin(B)-Hy cos(I3),where I3 is the angle between A and the x direction. By discretizing (1) and (3) with the boundary element method and the finite element method, respectively, the normal derivative fields (aH,/an and a H y / a n ) on r can be expressed explicitely in terms of the boundary fields ( H , and H y ) in their corresponding regions. The tangential derivative fields (dH, /&! and aHy/af!) in (4) are approximated by a three-point finite difference. Through matching the continuity of E, and H,, respectively, in (4a) and (4b) which are now in terms of H, and H y only, the propagation characteristics of the guided modes can be obtained. The integral equations. (1) and (3) and the continuity requirement (4) on r give a complete description of the electromagnetic behaviour of the guided mode (p,H,, H y ) .When we perturb (l), (3), and (4) with respect to the normalized frequency V , similar descriptions about the electromagnetic behaviour of the perturbed field (b,fix,fiy) can be constructed. Described in the following is the details for the perturbation process. To calculate the first derivative of , B with respect to the normalized frequency V ,we differentiate (l),(3), and (4) with respect to V . In doing so, we obtain that #(= fix or fiy)and on I? should satisfy the integral equation

a
and

w ( j D H * )= an + al

aHn

aHe

(7b)

where the dot denotes the derivative with respect to V and the waveguide material is assumed to be nondispersive (i.e. &/aV = 0). Note that in (6) the arbitrary fields ( H i and H:) are not changed as V varies. By differentiating G and dG/dn with respect to V, the closed forms of G and dG/dn can be expressed as
i d G = - -tK1 (kdt) 27r

(Sa)

and
'G - -k d k d A'

dn

2T

(?- '?)Ko(kdt)

(8b)

where t = I?- 2 1 , A is the unit vector along the normal direction of the cladding boundary at position 2 and K i is the modified Bessel functions of the second kind and order i. Similarly, to calculate the second derivative of p with respect to V , we differentiate (perturb) (3, (6), and (7) with respect to V . In doing so, we obtain that F and at the cladding boundary should satisfy the integral equation

dG(kd,? , 2) d; A F(~ = P S, F(r1) dn 2

+

G(kd,F, 2
+

d F ( 9 ) dr3

) r

? , 2) d2 + 2~ S,F(r'>dG(kd, dn
-2

(5) that Ikxl Hy, and in the core region can be related as

d2 + P Jr F ( r-' )dG(kd,T,J) dn
-

G(kd,F, r

- d #' ( 2 ))d r ; ~

\(k:c,

8Hx aHg 8Hx aHg - p2)HXH,C - --- -ax ax ay ay

that

d F ( 2 ) d; G(kd,F, r') 7

(9)

I?,, fiY , and

in the core region can be related as

J
aHy dHy" aHyally" / ( k : ~-~ p2)H H" - --- -a x ax ay ay

aHx aHg aHx 8Hg (k:ec - p2)I?,H: - -- ax ax a y ay

+

(g%)
-

%LH,"dxdy a y E" dxdy

(loa)

+ ($$ - x aHy ) --H,"dxdy 1

aEc

+/2(k&

b2+ k o i O ~-, pp)H,H;

pj)H,H; dxdy = 0

ax

E,

+ J4(kokoEc

Authorized licensed use limited to: National Taiwan University. Downloaded on February 23, 2009 at 04:10 from IEEE Xplore. Restrictions apply.

WU AND CHANG: DISPERSION CHARACTERISTICS OF DUAL-MODE ELLIPTICAL-CORE OPTICAL FIBERS

1929

+ Match boundary continuity of (E,, H,)

Given V

+ solve (1) and (3)

with (4)

then with V and {H,, Hy,p } Match boundary continuity of (g,, fi,) with (7) then with V,{H,,Hy,P}, and + Match boundary continuity of (I?,, g,) with (11)

{b,&,fiy}

+ solve (5) and (6) + (8,I?,, Hy} + solve (10) and (1 1) * {$,fi,, ky}
111. NUMERICAL PROCEDURES For clarity, we summarize the numerical procedure for solving as shown at the top of the page. In the following, the steps of the numerical procedures are described in detail with the matrix-form representations. Steps 1 and 2 describe the numerical procedures for solving the propagation characteristics of the guided modes (p,H,, Hy) for a given normalized frequency V . Steps 3 and 4 describe the procedures for solving the first-order perturbed propagation characteristics H,, hy) of the guided modes for the given V and the solved (p,H,, Hy) in steps 1 and 2. Steps 5 and 6 describe the procedures for solving the secondorder perturbed propagation characteristics B,, I?,) of the guided modes for the given V and the solved (p,H,, Hy) in steps 1 and 2 and By) in steps 3 and 4. Step 1: Express the normal derivatives of the fields on the boundary in terms of the boundary fields for the cladding and core regions by discretizing (1) and (3), respectively. Based on the boundary-element method with the pulse bases for (l), the relation between the boundary fields and their normal derivatives on the cladding boundary for the cladding region can be expressed as

{p, H,, Hy}

and

+/2(k&

8' + kol&c

- pp)HyH;dxdy

+ / 4 ( k o k o ~ ,- ,O,b)HyHGdxdy = 0

(lob)

(b,

and ) a 2 / a V 2 ( j p H , ) on I? can be and that a 2 / a V 2 ( j ~ E Z expressed as

(p,

@,a,,

and

The closed forms of G and dG/dn in (9) can be derived as

and

-f i =
.7=1 3#%

M j-jc;ej -

fj"Gije3
i=l

(14)

+ k;(KO(kdt) - kdtKl(kdt))]
(8,

(12b)

respectively, by differentiating (8) again with respect to V . The remaining work is to solve the propagation characteristics {p, H,, Hy}, I?,, fiy}, and (8, The integral equations (l), ( 5 ) , and (9) on the cladding boundary r and the variational equations (3), (6), and (10) in the core region are discretized by the boundary element method and the finite-element method, respectively. By enforcing the continuity requirement of (E,, H , ) , (I?,, I?,), and (8,,& ,) expressed as (4), (7), and (1 1) on the core-cladding boundary r, the propagation characteristics (0, H,, Hy}, &, hY}, and {b,&,,Hy} can be solved, respectively. Once fi is obtained, the waveguide dispersion D , can be calculated by

B,,BY}.

(8,

for i = 1,2, . . . , M , where M is the total number of divided segments on r, and ej is the arc length of the j-th segment. f j and fj", which are both unknowns, denote the transverse magnetic field and its normal derivative, respectively, on the boundary r at node j. Similarly, the matrix elements Gij and Gn. denote the values of the Green function and its normal ". denvative, respectively, on the boundary r between node i and node j. Note that the first summation in (14) excludes the point i = j due to the fact that in the Cauchy principal integral the singular value GG for i = j is removed. By some linear algebra operations on the M equations for i = 1,2, . . . ,M in (14), a matrix form representation of (14) can be written as
[ f n ]x ~ 1

= [Mdl

[fl M X 1

(15)

27rc where c is the velocity of light in free space.

--p V2

where f = [ f i fz, , . . . ,f ~ ] fn ~ = , f;, . . . ,&IT, and the matrix elements of [Md]are related to the Green function and its normal derivative as in (14). Because f represents either the x component (H,) or the y component (Hy)of the transverse magnetic field on the cladding boundary I?, we denote f as h : for the H , component on r and as hz for the Hy component. Similarly, f, is denoted as hz for the normal derivative of the H , component on and as hd,, L for that of the Hy component.

[fr,

Authorized licensed use limited to: National Taiwan University. Downloaded on February 23, 2009 at 04:10 from IEEE Xplore. Restrictions apply.

1930

JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 13, NO. 9, SEPTEMBER 1995

By employing the FEM with the triangular elements and linear bases for both the unknow fields ( H , and H y ) and the arbitrary fields ( H i and H,") and taking the variation on the arbitrary fields (see [12] for details), (3a) and (3b) are discretized as

respectively, and by the finite difference with (21) as in step 1 [h;,,,h;,,lT is expressed as

I."[

hz,n

2Mx1

= [ M , . ]hd P]

2Mx1

where N is the entire number of nodes in the core region including the nodes on I?, hi = Iz;,~,. . . , h&M]T denotes the column vector of nodal unknowns inside the core region (excluding the nodes on I?) with j polarization (j= x,y), and [Q] is a matrix determined from (3) as [Md] in (15) is derived from (1). The relation between the associated fields and their normal derivatives on the core boundary I? can be obtained by the finite difference with (16) 1121 as

where the column vecotors [Td] and [&I are determined by the solved / I , the boundary nodal fields [f,, f] in steps 1-2, and the first derivatives of the Green function and its normal derivative on the cladding boundary I' (G and d G / d n ) with respect to V, and the column vectors [-ycj and [&I are determined by p and the nodal fields h in the core region. Step 4: Solve ,b and h. By substituting (20) and (22) into (7) and enforcing the continuity requirement of & and at the boundary I?, one obtains

k,

$.I

= [ M . ]hd P ] 2Mx1
2Mx1

(17)

where hi,, is the column vector with its elements being the normal derivatives of the j-polarized fields on the core boundary r. For the sake of convinience, in the following discussion we define h = [h;, h i , h;, h:IT. Step 2: Solve for p and h. By substituting (15) and (17) in (4) and enforcing the continuity requirement of E, and H , at the core-cladding boundary I?, the relation between h i and h : can be expressed as

In (23), there are 2M 1 unknowns (hi,h: and but only 2 M equations. By differentiating (19) with respect to V, another constraint

8).

C(h&& + h&A&) = 0
i=l

(24)

For guided modes, the eigenvalue p is determined from det(R(/?)) = 0 and the eigenvector [hi,h : I T is nontrivial. The bisection method serves to locate the desired p. Because step 1 is repeated at each trial in the bisection root search, it is the most time-comsuming step. Once p and [hi,h;IT with the constraint
M
i=l

is obtained. Only one solution of h i ,h : } can be solved from (23) and (24). Step 5: Discretize (9) and (10) and obtain the relation between [fn]and [f]and the relation between [hf,,, hz,n]T and [h:, hZ]l'. By the same method with the same basis functions and adopting the same nodes as in steps 1 and 3, (9) and (10) can be transformed into

(8,

+ h:,i2) = 1

and
=
Z(N-M)xl

are obtained, [hi,,, h;,,lT and [hf,hZIT can be found from (15) and (16), respectively. The reason for the constraint on [hi,hY,ITwith normalized norm will become clear in the next few steps. Step 3: Discretize ( 5 ) and (6) and obtain the relation between [f,] and [f] and the relation between [hf,,, hZ,JT and [hi,h;IT. By the same method with the same basis functions and adopting the same nodes as in step 1, (5) and (6) can be transformed into
[ f n ] = ~ ~ ~

[Q]p]
hd

2(N-M)xl

respectively, and by the finite difference with (26) as in steps 1 and 3 [hz,n,hZ,n]T is expressed as

[Md][f]~x +b[rd] l + [ad]

(20)

and
= [Ql

2(N-M)x 1

["I

hs

2(N-M)xl

+ B[Sll + [a21

(21)

where [TA] and [SA] are determined by the first and second derivatives of the Green function and its normal derivative on the cladding boundary I? (G, d G / d n , G, d G / d n ) with respect to V, and the solved p, [fn, f]: and [fn, f] in steps 1 4 [?;I, and [SL] are determined by p, p, h, and h which have been solved in steps 1-4.

p,

Authorized licensed use limited to: National Taiwan University. Downloaded on February 23, 2009 at 04:10 from IEEE Xplore. Restrictions apply.

W U AND CHANG: DISPERSION CHARACTERISTICS OF DUAL-MODE ELLIPTICAL-CORE OPTICAL FIBERS

*
' I

1931

. .

'

'

4 } , ' I 1" I # , # # *
I :#.*--

.
*

--

_ - - _ _ < -

+
I
4
'

_ _ * . I "

.,**.:
, ,
I

I
'

#
1

'
LPIY

- TMm

TEm

Fig. 2. (a), (b), (c) and (d) show the patterns changing from T E o l to LyT:;, T M o l to as the fiber is deformed from a circle ( a / b = 1) to an ellipse (a/b = 1.05).

Lyyeg,

HEgl to

Lyti

and

H E : ,

to Ly?dz, respectively,

Step 6: Solve fi and h. By substituting (25) and (27) into refractive index profile is defined as (11) and enforcing the continuity requirement of Ez and kz - (nf - nz)R* for R 5 1 at the boundary l 7 , one obtains n2(R) = elsewhere
[R(P)12MX2M

[" I

hY d 2 ~ x 1

+&'I

+ [(?I

=I ) .

(28)

By differentiating (24) again with respect to V, the constraint becomes

i=l

Only one solution of h : , hz} can be solved from (28) and (29). Based on the full-wave vectorial formulations of the SIEM and FEM, the approach described above can efficiently calculate the first and second derivatives of the propagation constants with respect to V of the guided modes for optical waveguides with inhomogeneous core and uniform cladding. Most of the computational time is spent in the first two steps when searching for P and h. In steps 3-6, only some matrix generations, matrix multiplications, and matrix inversions are required.

(8,

Iv. RESULTS AND DISCUSSIONS

Consider a dual-mode e-core fiber with the major and minor axis radii being a and b, respectively, as shown in Fig. 1. The

y2/b2)1/2.Note that E , ( R )= n2(R) where R = (.'/a2 and Ed = ni. When a = 1, 2, and 00, the index profile is a triangular, parabolic, and step distribution, respectively. The change of the vectorial magnetic field patterns of the four higher-order LPll modes as the fiber deforms from a circle ( a / b = 1) to an ellipse ( a / b = 1.05) is shown in Fig. 2 with the parameters being A (nf - nE)/2nf = 2%, V ( 2 7 r b / X ) d m = 4 and a = 2. Fig. 2(a), (b), (c), and (d) shows the patterns changing from TEol to LP??,", T M o l to Lc;;:, HE& to LP$: and HE;, to LPyf,,d,,respectively. Each arrow represents the orientation and the relative strength of the magnetic field at the point specified by the arrow root. It can be seen that the small deformation of core geometry results in the linearly polarized field patterns with two lobe orientations. To check the correctness of this novel numerical approach, we compare the dispersions, which are calculated by our approach, of the LP:\;e" modes for the e-core fiber with the results by Poole et al. [6] which are also based on a vectorial waveguide theory. The e-core parameters used in [6] are e 2(a - b ) / ( a b) = 10% and a = 00 (step index distribution). The cutoff wavelength of the LET; mode is designed to be at 1600 nm for the e-core fiber. Fig. 3 shows the dispersions of the LyT:," and L E T : modes versus wavelength for three different index steps A = OS%, 2%, and 4%, respectively.

Authorized licensed use limited to: National Taiwan University. Downloaded on February 23, 2009 at 04:10 from IEEE Xplore. Restrictions apply.

1932

JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 13, NO. 9, SEPTEMBER 1995

E
E

-100 -200 -

0.5%

LP11,x

even LP11,y e ~ n(This work)

mol, and
-

index distribution on the waveguide dispersions of the TEol, HE21 modes for the circular fiber with A = 2%. The radius b = 2.2 pm is chosen such that the cutoff wavelength is 1.623 pm for the HE21 mode. It can be seen

I"

\-...I

Fig. 3. Dispersion versus wavelength for the step-index e-core fiber with three different index deoths (A = 0.5%. 2% and 4%).ComDarision between the results in [6] and ours is also shown.
\

for a = 2 and a = co cases. Assume that in the presence of be written as the dip the index profile

where ni(R),which can be any a-power distribution, is the refractive index profile in the absence of the dip and p and d define the fractional dip depth and the fractional dip width, respectively. A case of the profile with ni(R) being the step distribution (a = co) is shown in Fig. 5. The dispersions of -250 the four LPll modes versus wavelength from 1.53 pm to 1.56 p m for three different dip parameters (d = p = 0, d = p = 0.2, and d = p = 0.3) with a = co,a / b = 1.05, A = 2% and b = 2.137 pm are shown in Fig. 6. For each pair of curves the upper curve is for x polarization and the lower curve is for y polarization for each case with different dip parameter. Scalar Theory in Ref. [ll] It can be seen that the index dip causes the increase of the -400 153 1.535 1.54 1.545 155 1.555 156 1.565 dispersions of the LPll modes. The reason is that due to the hclrm) dip the fractional power guiding in the core decreases and Fig. 4. Dispersion versus wavelength for the circular fiber with three differ- the compensator is operated much near the cutoff wavelength. ent a-power distributions. The results calculated by the scalar theory in [ l l ] The influence of the dip on the dispersions of the compensator are also presented. with a = 2, a / b = 1.05, A = 2%, and b = 3.06 pm is shown in Fig. 7. The dip effect of increasing the dispersion values is the same as in the previous case (a = co). Note that the Our results are shown as the solid and dashed curves, while dispersion discussed in Figs. 3, 5, and 6 is the summation of those in [6] are plotted as dots and crosses. It can be seen that the waveguide dispersion calculated by the approach described the dispersions increase with the increase of the index step in Section I11 and the material dispersion obtained from the and that for A = 0.5% the dependence of the dispersion on empirical Sellmeier equation with 13% GeO2 doping in silica the polarization is indistinguishable. Our results are generally for the core and pure silica for the cladding [ 151. Compared consistent with [6], except that for fibers with larger index with the waveguide dispersion, the contribution of the material steps (A = 2% and 4%) the difference between the dispersions dispersion is relatively small for the e-core Compensator which for the two polarizations in our results is smaller than that is operated near the cut-off wavelength. predicted by [6]. This discrepancy becomes more significant for larger wavelength which is close to cutoff. It should be V. CONCLUSION noted that the cutoff wavelengths for the Ley:; and L E T : Based on the full-wave vectorial formulations of combining modes splits as the index step increases. We design the minor axis b of the e-core fiber such that the cutoff wavelength of the the surface integral equation method (SIEM) and the finiteLP:T," mode is 1600 nm. However, how the cutoff wavelength element method (FEM),a novel approach for determining 1600 nm was related to the modes was not clearly stated in [6], the dispersion characteristics of the dual-mode elliptical-core making the comparision of the exact values of the dispersion fibers with arbitrary index profile in the core is proposed without resorting to numerical differentiation.Most computing more difficult.
150
Authorized licensed use limited to: National Taiwan University. Downloaded on February 23, 2009 at 04:10 from IEEE Xplore. Restrictions apply.

----.---;=-----_____

a=1 ---------

WU AND CHANG DISPERSION CHARACTERISTICS OF DUAL-MODE ELLIPTICAL-CORE OPTICAL FIBERS

1933

2 n1

I
I I I I I

2 n2

-1

R
Fig. 5. The index profile of the step (a= co) distribution with the Gaussian index dip.

_____
-300
I

p=dI 0.3 D.dIa2 p-d-0

b 3.0Clum

1.53

1.535

1.54

1.545

1.55

1.555

1.56

a (w)
Fig. 7. The same as in Fig. 5 but for the case of parabolic index profile.

-320 -340-

_____

p = d 10.3 p Id I0.2 p-d.0

a / b 1.05 b = 2.1368

3 6 0

1.53

1.535

1.54

1.545

1.55

1.555

1.56

(Pm
Fig. 6. Dispersion versus wavelength for three different dip parameters for the case of step index profile. For each pair of curves the upper curve is for z polarization and the lower curve is for y polarization for each case with different dip parameter.

efforts are spent in solving the propagation constant , b and the guided mode fields H , and Hy. Once they are obtained, only some matrix manipulations are required to obtain the first and second derivatives of the propagation constant with respect to V. By this proposed approach, the influence of the refractive index distribution parameters, such as the different a-powers and the index dip effect, on the dispersions of the four nondegenerate LPll modes for the e-core compensators are investigated as numerical examples. The vectorial field patterns of these four modes for the e-core fibers have also been presented.

[4] C. D. Poole, J. M. Wiesenfeld, A. R. McCormick, and K. T. Nelson, Broadband dispersion compensation by using the higher-order spatial mode in a two-mode fiber, Opt. Lett., vol. 17, pp. 985-987, 1992. [5] A. M. Vengsarkar, W. A. Reed, and C. D. Poole, Effect of refractiveindex profiles on two-mode optical fiber dispersion compensators, Opt. Left., vol. 17, pp. 1503-1505, 1992. [6] C. D. Poole, J. M. Wiesenfeld, D. J. DiGiovanni, and A. M. Vengsarkar, Optical fiber-based dispersion compensation using higher order modes near cutoff, J. Lightwave Technol., vol. 12, pp. 1746-1758, 1994. [7] R. A. Sammut, Analysis of approximations for the mode dispersion in monomode fiber, Electron. Lett., vol. 15, pp. 590-591, 1979. [8] W. L. Mammel and L. G. Cohen, Numerical prediction of fiber transmission characteristics from arbitrary refractive-index profiles, Appl. Opt., vol. 21, pp. 699-703, 1982. [9] E. K. Sharma, A. Sharma, and I. C. Goyal, Propagation characteristics of single mode optical fibers with arbitrary index profiles: A simple numerical approach, IEEE J. Quantum Electron., vol. QE 18, pp. 1484-1489, 1982. [lo] A. Sharma and S. Banerjee, Chromatic dispersion in single mode fiber with arbitrary index profiles: A simple method for exact numerical evaluation, J. Lightwave Technol., vol. 7, pp. 1919-1923, 1989. [ l l ] H. Y. Lin, R.-B. Wu, and H.-C. Chang, An efficient algorithm for determining the dispersion characteristics of single-mode fibers, J. Lightwave Technol., vol. 10, pp. 705-711, 1992. [12] C.-C. Su, A combined method for dielectric waveguides using the finite-element technique and the surface integral equations method, ZEEE Trans. Microwave Theory Tech., vol. MlT-34, pp. 1140-1 146, 1986. [13] L. Eyges, P. Gianino, and P. Wintersteiner, Modes of dielectric waveguides of arbitrary cross sectional shape, J. Opt. Soc. Am., vol. 69, pp. 1226-1235, Sept. 1979. [I41 N. Morita, A method extending the boundary condition for analyzing guided modes of dielectric waveguides of arbitrary cross-sectional shape, IEEE Trans. Microwave Theory Tech., vol. MTT-30, pp. 6-12, 1982. [I51 J. W. Fleming, Material dispersion in lightguide glasses, Electron. Lett., vol. 14, pp. 326329, 1978.

REFERENCES
[ l ] J. N. Blake, B. Y. Kim, and H. J. Shaw, Fiber-optic modal coupler using periodic microbending, Opt. Lett., vol. 11, pp. 177-179, 1986. [2] W. V. Sorin, B. Y. Kim, and H. J. Shaw, Highly selective evanesent modal filter for two-mode optical fibers, Opt. Lett., vol. 11, pp. 581-583, 1986. [3] B. Y. Kim, J. N. Blake, H. E. Engan, and H. J. Shaw, All-fiber acousto-optic frequency shifter, Opt. Lett., vol. 11, pp. 389-391, 1986.

Tzong-Lin Wu (S95) was born in Hsinchu, Taiwan, Republic of China, on June 23, 1969. He received the B.S.E.E. and Ph.D. degrees from National Taiwan University, Taipei, Taiwan, in 1991 and 1995, respectively.His research interests include the theory of optical fibers and modeling of fused fiber couplers.

Authorized licensed use limited to: National Taiwan University. Downloaded on February 23, 2009 at 04:10 from IEEE Xplore. Restrictions apply.

1934

JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 13, NO. 9, SEPTEMBER 1995

Hung-chun Chang (S78-M83) was born in Taipei, Taiwan, Republic of China, on February 8, 1954. He received the B.S. degree from National Taiwan University, Taipei, R.O.C., in 1976, and the M.S. and Ph.D. degrees from Stanford University, Stanford, CA, in 1980 and 1983, respectively, all in electrical engineering From 1978 to 1984, he was with the Space, Telecommunications, and Radioscience Laboratory of Stanford University. In August 1984, he joined the faculty of the Electrical Engineering Department of National Taiwan University, where he is currently a Professor. He served as Vice-chairman of the EE Department from 1989 to 1991. Since August 1992, he has been Chairman of the newly established Graduate Institute of Electro-Optical Engineering at the same university. His current research interests include the theory, design, and application of guided-wave structures and devices for fiber optics, integrated optics, optoelectronics,and microwave and millimeter-wave circuits. Dr. Chang is a member of Sigma Xi, the Phi Tau Phi Scholastic Honor Society, the Chinese Institute of Engineers, the Optical Society of America, ChindSRS (Taipei) National Committee (a Standing Committee member since 1988) and Commission H of the US.National Committee of the International Union of Radio Science (URSI). In 1987, he was among the recipients of the Young Scientists Award at the URSI XXIInd General Assembly. In 1993, he was one of the recipients of the Distinguished Teaching Award sponsored by the Ministry of Education of the Republic of China.

Authorized licensed use limited to: National Taiwan University. Downloaded on February 23, 2009 at 04:10 from IEEE Xplore. Restrictions apply.