1 Optical Fibers — supplementary notes

Optical fibers are cylindrical dielectric waveguides. Their operation, in analogy with dielectric slab waveg-
uides, depends on total internal reflections from the boundary of a high refractive index cylindrical core
embedded in a lower refractive index cladding. They support TE and TM modes just like slab waveguides,
but these modes are not “fundamental” within the optical fiber — that is, they have finite cutoff frequencies
greater than a cutoff frequency of yet another mode, the fundamental one, that happens to be, in step-
index optical fibers (as opposed to graded-index), one particular one of an infinitely many hybrid modes
having non-zero Ez and Hz variations. To understand the total collection of TE, TM, and hybrid HE and
EH modes that populate our step-index optical fibers we need to look at the solutions of Maxwell’s wave
equation in a cylindrical geometry.

1.1 Wave equation in cylindrical coordinates

Maxwell’s curl equations furnish us with vector wave equations having the z-components

r 2 Ez + k 2 Ez = 0
r 2 Hz + k 2 Hz = 0 (1)

where
!2 2
k2 =
n = ko2 n2 (2)
c2
p
with n = µr ✏r denoting the refractive index in a homogeneous medium1 . In a cylindrical geometry of a
step-index fiber with (r, , z) coordinates we seek propagating solutions of a separable form
j z
Ez , Hz / R(r)P ( )e , (3)

in which case the equations above reduce to

1 @ @P R 1 @2
r2? RP + (k 2 2
)RP = (r ) + 2 2 (P R) + (k 2 2
)RP = 0
r @r @r r @
that further simplifies as
r @ @R 1 @2P
(r )+ + r2 (k 2 2
) = 0.
R @r @r P @ 2
Let
1 @2P @2P
= m2 ) + m2 P = 0 ) P ( ) / e±jm (4)
P @ 2 @ 2
and consequently
q q
r @ @R
(r ) + r2 (k 2 2
) = m2 ) R(r) / Jm ( k 2 2 r), Km ( 2 k 2 r). (5)
R @r @r
1
Also valid in inhomogeneous graded-index fibers with an index profile n(r) = n1 f (r) n, with 0 = f (0) < f (r) < 1 = f (a),
so long as “weakly guiding” condition n ⌧ n1 holds. Since the Helmholtz equation (1) with inhomogeneous k(r) is also
the Klein-Gordon equation describing the quantum wavefunctions (r) of massless spin-zero particles confined in potential
wells / f (r), graded-index fibers are found to support quantized Ez (r) variations (modes) matching the (r) eigenfunctions
of quantized energy states of such particles in cylindrical potential wells. See Gloge [Rep. Prog. Phys., 42, 1979] for a
discussion of Ez (r) “eigenfunctions” resulting from a parabolic f (r) = r2 /a2 profile, analogous to the quantum mechanical
harmonic oscillator and leading to a ladder of 2 /ko2 “eigenvalues” varying with a pair of “quantum numbers” m and l
(instead of 3 as in a spherical hydrogen atom) for a total of V 2 /4 modes expressed in terms of the V-number (11) of
the fiber. Multimode graded-index fibers exhibit weaker “modal dispersion” — due to different group velocities — than
multimode step-index fibers because of a compensation effect involving path lengths and propagation speeds.

1
 Single valued P ( ) demands integer values of m above, while Jm (X) and Km (Y ) denote mth order Bessel
functions of the 1st kind and modified 2nd kind, respectively, fitting the linear differential equation in the
left known as Bessel’s differential equation. We disregard other solutions — 2nd kind and modified 1st kind
— of Bessel’s differential equation leading to unbounded fields in core and cladding regions for reasons of
realizability.
With the above 1D solutions we can construct bounded 3D phasor solutions of the wave equation as

Ez = Jm (kt r)(Aejm + cc)e j z

Hz = Jm (kt r)(Bejm + cc)e j z

(6)

in the core region r < a where n = n1 and

kt2 ⌘ k 2 2
= ko2 n21 2
. (7)

Likewise, for the cladding region where n = n2 < n1 we use the bounded options

Ez = Km (↵r)(Cejm + cc)e j z

Hz = Km (↵r)(Dejm + cc)e j z
(8)

where
↵2 ⌘ 2
k2 = 2
ko2 n22 . (9)

(a) (b)

Figure 1: (a) The red curve depicts J1 (r) in the range 0 < r < 3. The green curve is K1 (r) normalized by
K1 (3)/J1 (3) plotted over 2 < r < 10. Their combination illustrates how Jm and Km functions
can be patched together at some r = a boundary to construct bounded field profiles for optical
fiber modes. (b) Plots of Jm (r) and Km (r) for m = 0, 1, 2.

Note that
V2
↵2 + kt2 = ko2 (n21 n22 ) ⌘ . (10)
a2
This leads to q
!
V = ko a n21 n22 =
aN.A. (11)
c
which is a normalized operation frequency of the fiber known as its V-number, while
q
N.A. ⌘ n21 n22 = sin ✓a (12)

is the so-called numerical aperture of the fiber corresponding to the sine of the acceptance angle — rays
making incidence angle ✓ < ✓a with the normal to the external cross-section of the fiber can be guided by

2
the fiber since the internal incidence angle at the core-cladding boundary will then exceed the the critical
angle ✓c = sin 2 n2 /n1 as required for total internal reflection (HW problem).
Given Ez , Hz , the remaining field components in the transverse plane can be obtained from Maxwell’s
curl equations as follows (HW problem):
j @Ez @Hz
E = ( !µ )
k2 2 r @ @r
j !µ @Hz @Ez
Er = ( + )
k2 2 r @ @r
j @Ez @Hz
H = (!✏ + )
k2 2 @r r @
j @Hz !✏ @Ez
Hr = ( ). (13)
k2 2 @r r @
To determine the amplitude coefficients above, and the propagation constant , we need to enforce the
boundary conditions at r = a —the core radius of the fiber — of the continuity of tangential Ez , E as
well as Hz , H .
Before doing that, we note that one of the amplitude coefficients above, say A, can be selected at will.
Then the boundary conditions will identify the remaining coefficients in terms of A. Let us briefly see the
impact of choosing A = 12 and A = 2j
1
as two plausible examples.

Illustrative example: With A = 1

2 we get

Ez = Jm (kt r)(Aejm + cc)e j z

= Jm (kt r) cos(m )e j z
(14)

while with A = 1
2j

Ez = Jm (kt r)(Aejm + cc)e j z

= Jm (kt r) sin(m )e j z
(15)

• Now, by convention, = 0 is the direction of the x-axis while = 90 is the direction of y-axis.
– Accordingly, these two solutions above represent having the Ez field intensity peaking along
the x- and y-axes, respectively (for m = 1 case, at least). When we want to have the field
intensity peaking along some other direction we will need to use a linear combination of these
two solutions.

• Hence it is natural to think of these two solutions as a complete set of orthogonal basis functions
or eigenfunctions for Ez . In general, for any m 1, these will have a 90 /m rotational symmetry
property — rotating the intensity pattern of any one of these eigenfunctions by 90 /m in the xy-plane
will produce the intensity pattern of the complementary function.

• The case m = 0 is special — in that case there is no variation of field intensity and sin(m ) basis
function does not exist!

• We will refer to the the orthogonal eigensolution pairs described here as “normal modes” in the
following discussion.

By applying the boundary conditions above it can be shown that B = j⇠A, where ⇠ is a real number. This
leads to ( ) ( )
sin(m ) cos(m )
Hz / when Ez / .
cos(m ) sin(m )

3
Accordingly it is more convenient to carry out the boundary condition matching exercise with a modified
set of candidate solutions — normal modes — written as
j z
Ez = AJm (kt r) cos(m )e
j z
Hz = BJm (kt r) sin(m )e (16)

and
j z
Ez = CKm (↵r) cos(m )e
j z
Hz = DKm (↵r) sin(m )e (17)

and obtain the complementary partner after applying a 90 /m rotation in the solution obtained for the
first basis function.

1.2 Boundary condition matching

Using the candidate solutions just proposed, and the chain rule of differentiation, the transverse field
components are found to be in the core region
j @Ez @Hz j m 0
E = ( !µ )= ( AJm (kt r) sin(m ) !µ1 Bkt Jm (kt r) sin(m ))e j z
k2 2 r @ @r kt2 r
j !µ @Hz @Ez j m!µ1 0
Er = ( + )= 2 ( BJm (kt r) cos(m ) + kt AJm (kt r) cos(m ))e j z
k2 2 r @ @r kt r
j @Ez @Hz j 0 m
H = (!✏ + )= 2 (!✏1 kt AJm (kt r) cos(m ) + BJm (kt r) cos(m ))e j z
k2 2 @r r @ kt r
j @Hz !✏ @Ez j 0 m!✏1
Hr = ( )= ( kt BJm (kt r) sin(m ) + AJm (kt r) sin(m ))e j z (18)
k2 2 @r r @ kt2 r
and in the cladding region
j @Ez @Hz j m 0 j z
E = ( !µ ) = 2( CKm (↵r) sin(m ) !µ2 D↵Km (↵r) sin(m ))e
k2 2 r @ @r ↵ r
j !µ @Hz @Ez
Er = ( + ) = HW problem
k2 2 r @ @r
j @Ez @Hz j 0 m
H = (!✏ + ) = 2 (!✏2 ↵CKm (↵r) cos(m ) + DKm (↵r) cos(m ))e j z
k2 2 @r r @ ↵ r
j @Hz !✏ @Ez
Hr = ( ) = HW problem. (19)
k2 2 @r r @
Our four boundary conditions subsequently yield, with
p
X ⌘ kt a and Y ⌘ ↵a and V = X 2 + Y 2, (20)

the following four relations:

AJm (X) = CKm (Y ) —Ez matching

BJm (X) = DKm (Y ) —Hz matching
1 0 1 0
2
(m AJm (X) + !µ1 BXJm (X)) = (m CKm (Y ) + !µ2 DY Km (Y )) —E matching
X Y2
1 0 1 0
2
(!✏1 XAJm (X) + m BJm (X)) = (!✏2 Y CKm (Y ) + m DKm (Y )) —H matching (21)
X Y2

4
where Jm0 (X) and K 0 (Y ) indicate Bessel function derivatives — see below. Note that the cancellation of
m
dependent terms from these equations depended crucially on having normal mode Ez / cos(m ) paired
with Hz / sin(m ) — this result can therefore be taken as a posteriori justification of the form assumed
for our normal modes.
Now, using the first two equations to get rid of C and D in favor of A and B, we have
1 0 1 Jm (X) 0
(m AJm (X) + !µ1 BXJm (X)) = (m AJm (X) + !µ2 B Y Km (Y ))
X2 Y 2 Km (Y )
1 0 1 Jm (X) 0
(!✏1 XAJm (X) + m BJm (X)) = (!✏2 Y A K (Y ) + m BJm (X)),
X2 Y 2 Km (Y ) m
yielding, after some more work,
µ1r Jm 0 (X) 0 (Y )
µ2r Km 1 1 m
B( + ) = A( 2
+ 2) (22)
XJm (X) Y Km (Y ) X Y !µo
✏1r Jm0 (X) 0 (Y )
✏2r Km 1 1 m
A( + ) = B( 2 + 2 ) . (23)
XJm (X) Y Km (Y ) X Y !✏o
Now, multiply the equations and simplify using µ2r = µ1r = 1 to obtain the dispersion relation
0 (X)
Jm 0 (Y )
Km n2 J 0 (X) n22 Km
0 (Y ) m2 2 1 1
( + )( 1 m + )= 2
( 2 + 2 )2 . (24)
XJm (X) Y Km (Y ) XJm (X) Y Km (Y ) ko X Y

We call (24) the dispersion relation because given the frequency ! = ko c, it can be solved for the
propagation constant , onceq the fiber parameters a, n1 ,and n2 are specified.
q
p
Recall that X = akt = a ko n12 2 2 and Y = V 2 X 2 where V = ko a n21 n22 .
In computing the dispersion relation (24) we will typically plot its RHS and LHS as a function of X for
a given V and compute kt on the way to from the X-numbers at the intersections of the RHS and LHS
curves. In evaluating dispersion relation (24) use the Bessel identities
0 m
Jm (X) = ⌥Jm±1 (X) ± Jm (X) (25)
X
and
0 m
( 1)m Km (Y ) = ( 1)m±1 Km±1 (Y ) ± ( 1)m Km (Y ). (26)
Y
Finally, we note that dispersion relation (24) is equally valid for Ez / sin(m ) modes, m 0, although
the plots of field intensities that can be constructed using the equations derived above will need to be
rotated by 90 /m to be correct. More quantitatively, we can make the

sin(m ) ! sin(m( 90 /m)) = sin(m 90 ) = cos(m ) (27)

and
cos(m ) ! cos(m( 90 /m)) = cos(m 90 ) = sin(m ) (28)
replacements in above equations for plotting the Ez / sin(m ) modes properly.
Now, how are these normal mode fields polarized? The answer is, a variety of polarizations will be found,
depending on each mode. Later on we will find out that it is possible to organize some groups of the allowed
modes in certain linear combinations that yield linearly polarized field distributions in x or y directions if
the condition n = n1 n2 ⌧ n1 holds. But before we study those “quasi LP modes” of “weakly guided
fibers” we will take a look at the exact normal modes corresponding to the direct solutions we have derived
above — they are called TE and TM modes when m = 0, and HE and EH modes when m 1.

5
1.3 TE and TM modes
Let m = 0. Then our dispersion relation simplifies as

J00 (X) K00 (Y ) n21 J00 (X) n22 K00 (Y )

( + )( + )=0 (29)
XJ0 (X) Y K0 (Y ) XJ0 (X) Y K0 (Y )

while the fields in the core region are given, when Ez / cos(m ) = 1 and Hz = 0, by TM mode equations
j m 0
E = ( AJm (kt r) sin(m ) !µ1 Bkt Jm (kt r) sin(m ))e j z = 0
kt2 r
j m!µ 0 j 0
Er = 2 ( BJm (kt r) cos(m ) + kt AJm (kt r) cos(m ))e j z = 2 kt AJm (kt r)e j z
kt r kt
j 0 m j 0
H = (!✏1 kt AJm (kt r) cos(m ) + BJm (kt r) cos(m ))e j z = 2 !✏1 kt AJm (kt r)e j z
kt2 r kt
j 0 m!✏
Hr = 2 ( kt BJm (kt r) sin(m ) + AJm (kt r) sin(m ))e j z = 0. (30)
kt r

With Ez / sin(m ) = 0 and Hz / cos(m ) = 1, we have the TE mode equations

j m 0 j 0
E = ( AJm (kt r) cos(m ) + !µ1 Bkt Jm (kt r) cos(m ))e j z = 2 !µ1 Bkt Jm (kt r)e j z
kt2 r kt
j m!µ 0
Er = 2 ( BJm (kt r) sin(m ) + kt AJm (kt r) sin(m ))e j z = 0
kt r
j 0 m
H = 2 (!✏1 kt AJm (kt r) sin(m ) + BJm (kt r) sin(m ))e j z = 0
kt r
j 0 m!✏ j 0
Hr = 2 ( kt BJm (kt r) cos(m ) + AJm (kt r) cos(m ))e j z = 2 kt BJm (kt r)e j z . (31)
kt r kt
Note that TE mode is characterized by an azimuthal polarized electric field while the TM mode is carried
by a radially polarized field. One of them, TE mode, has the dispersion relation

J00 (X) K00 (Y )

= (32)
XJ0 (X) Y K0 (Y )
while the other, TM mode,
n21 J00 (X) K00 (Y )
= ). (33)
n22 XJ0 (X) Y K0 (Y )
This identification of the two separate roots of (24) for m = 0 case with TE and TM modes comes from
comparisons of (32) and (33) with (22) and (23), respectively. (22) came from E matching when Hz 6= 0,
meaning TE when m = 0. Likewise, (23) came from H matching when Ez 6= 0, meaning TM when m = 0.
The distinctions between TE mode dispersion relation (32) and TM wave relation (33) are negligible
when n1 n2 ⌧ n1 . In that case (32) applies equally well for both TE and TM modes.
Recall that q
2⇡a q 2 ! q
V = ko a n21 n22 = n1 n22 = a n21 n22 .
o c
Using this expression with o = 1.5 µm, a = 5 o , n1 = 1.5, n = n1 n2 = 0.02 we find that V ⇡ 7.7. If
the fiber radius is reduced to a = o then V-number is reduced to about 1.53.
In Figure 2 we depict a plot of both sides of (32) versus X when V = 2 and 8. The RHS curves in such
figures occupy the range 0 < X < V as explained below:

6
 Figure 2: TE0l and TM0l modes obtained with m = 0.

Notice that with V = 8, the intersections of the RHS and LHS curves in the range X < V = 8 imply
the propagation of TE01, TM01, TE02, and TM02 modes in a fiber with a V-number of 8. But when the
V-number is reduced below 2.405 all of those modes would be put into “cutoff” — this is clear from the
V = 2 curve included in the plot. When we examine the HE and EH modes in the next section, we will
find out the existence of a “fundamental” HE11 mode that enjoys the fiber all by itself if V < 2.405. This
critical number, Vc = 2.405, is the first zero of J0 (X) in X associated with the “first infinities” of LHS
curve / 1/J0 (X)!! In fact all the “infinities” of the LHS curve shown (in blue) in Figure 2 correspond to
successive zeroes of J0 (X), namely, 2.405, 5.520, 8.655, 11.792, etc. (verify in HW).
In examining Figure 2, and the subsequent dispersion plots to be shown, take a notice that the oscillatory
and blue colored LHS curves depend only on X — they are, most importantly, V -invariant! The RHS curves,
on the other hand, are very sensitive to V , and in fact only exist as real-valued functions for X < V . Thus
“guided solutions” on optical fibers are only possible for X < V (otherwise Y is imaginary and no good),
and the existence of propagating modes in this permissible range of X depends on whether the range is
populated by intersecting LHS curves for possible modes — there is no such LHS curve in Figure 2 for
X < 2.405.

1.4 HE and EH modes

Let’s take a look at the dispersion relation (24) once more which is repeated here:
0 (X)
Jm 0 (Y )
Km n2 J 0 (X) n22 Km
0 (Y ) m2 2 1 1 2
( + )( 1 m + )= ( + ) .
XJm (X) Y Km (Y ) XJm (X) Y Km (Y ) ko2 X 2 Y 2

7
We will solve this dispersion relation when m 1 after making some simplifying approximations. Let
n21 ⇡ n22 ⇡ n2 in which case we have
0 (X)
Jm 0 (Y )
Km m2 2 1 1 1 1
( + )2 ⇡ 2 2 ( 2 + 2 )2 ⇡ m2 ( 2 + 2 )2
XJm (X) Y Km (Y ) n ko X Y X Y
since if n1 ⇡ n2 ⇡ n then ⇡ nko — essentially the case of “weakly guided” waves with negligible kt and
↵. It then follows that 0 (X) 0 (Y )
Jm Km 1 1
+ ⇡ ±m( 2 + 2 )
XJm (X) Y Km (Y ) X Y
leading to
0 (X)
Jm Km0 (Y ) 1 1
⇡ ± m( 2 + 2 ). (34)
XJm (X) Y Km (Y ) X Y
These two approximate roots of our exact dispersion relation (24) for m 1 are the weakly guided HE (use
m; Ez dominates Hz in this mode) and EH (use +m; Hz dominates Ez in this mode) modes as we will
find out. Also, with m = 0 this single dispersion relation covers the weakly guided TE and TM modes that
we have already looked at2 .
Obtaining the exact roots of (24) for m 1, in order to study the “strongly guided” EH and HE modes,
is quite cumbersome and will not be attempted here — if interested in doing that, use 2 = ko2 n21 X 2 /a2
in the dispersion equation (24) and use a numerical root finder. Fortunately doing that is not necessary
because in practical optical fibers utilized in the low-loss o ⇡ 1.5 µm band (attenuation dips down to about
0.15 dB/km around here — see Figure 3) we have n1 ⇡ n2 ⇡ 1.5 (fine tuned by doping fused silica SiO2
with, e.g., GeO2 to increase its n within the core) and the weak guiding approximation works very well.

Figure 3: The attenuation constant of fused silica within the 1 µm wavelength band — from Saleh and
Teisch (2007).
2
Notice that TM0l and TE0l modes correspond to cos(m ) and sin(m ) normal modes of Ez , respectively, when m = 0.
With m 1, each EHml and HEml mode (selected by ±m choice in (34)) will have each of these cos(m ) and sin(m )
normal modes of Ez . Because of this we say that EH and HE modes have a “2-fold degeneracy” while TE and TM modes
are not degenerate.

8
(a)

(b)

Figure 4: (a) HE1l modes, (b)EH1l modes obtained with m = 1.

Figures 4a and b show the HE and EH mode hybrid solutions for the same weak-guiding optical fiber

9
examined in Figure 2 — i.e., V = 2 and 8. Also Figure 5 shows the case m = 2 for HE modes.
Clearly, an examination of Figures 4a and b and comparisons with Figure 2 show that the fundamental
mode in a weakly guided fiber is HE11 — it has no cutoff frequency, it propagates at all frequencies !, or
with all normalized frequencies V !
Furthermore, designing a fiber with a V-number of 2.405 would put all the other modes into cutoff and
ensure a “single mode operation”. This can be confirmed by an examination of Figure 5 where for V = 2
the HE21 mode is cut-off, but it is propagating for V = 8 > 2.405 ⌘ VcLP 11 .

Figure 5: HE2l modes obtained with m = 2.

To summarize, so-called “single-mode” fibers have V-numbers of 2.405 or less — this critical V-value is
independent of fiber constitution and it comes from boundary condition math in cylindrical settings, namely
the first zero of J0 (X), much like ⇡/2 is the first zero of cos(X). To maintain single mode operation we
select a/ o , n1 , and n such that V  2.405.

1.5 Field structure of HE11 mode

There are two flavors of the dominant HE11 mode — cos( ) and sin( ) flavors — each having its own
preferred polarization direction that is linear. It turns out that cos( ) flavor is x-polarized and sin( ) flavor
is y-polarized. This makes sense because as x-polarized transverse field loops back into itself via Ez , the
latter will need to switch sign as it crosses the x = 0 plane just like the cos( ) function. In Figure 6 we show
the Mathematica snippet to compute and display vector plots of both flavors of the HE11-mode electric
fields in the transverse xy-plane along with the plots themselves — V = 3 situation is depicted in the figure
using the weak guiding solutions.

10
Figure 6: Mathematica code and plots of transverse electric fields for both flavors of the fundamental or
dominant HE11 mode.

11
Figure 7: Mathematica code and plots of transverse electric fields for radial TM01 and azimuthal TE01
modes.

1.6 LP quasi-modes
Neither TE and TM nor HEml and EHml modes for m > 1 are linearly polarized — Figures 7 and 8 show
the transverse field distributions of TM01/TE01 and HE21 modes as useful examples.
However, it does turn out that for weakly guided fibers one could construct linearly polarized superposi-
tions of TM01, TE01, and HE21 modes — these are classified as a four-fold degenerate “LP11 mode”. This
is possible because under the weak guiding condition propagation constants and cutoff frequencies Vc of
these three modes are nearly identical.

12
 Figure 8: Mathematica code and plots of transverse electric fields for both flavors of HE21 mode.

Figures 9-10 illustrate how various “weighted” superpositions of TM01, TE01, and HE21 modes produce
all four flavors of the linearly polarized LP11 mode.

13
 Figure 9: LP11xx

(a)

(b) (c)

Figure 10: (a) LP11xy, (b) LP11yx, (c) LP11yy

Furthermore similar groupings of higher order TE, TM, HE, and EH modes into higher-order quasi-linear-
polarized modes labelled as “LPml modes” turns out to be possible. Some of such groupings of orders m = 0
and 1 and their relationships to the zeroes of J0 (X) and J1 (X) functions are illustrated in Figure 11 taken
from Gloge [1971].

14
Figure 11: The regions of parameter X for modes of order m = 0, 1 — from Gloge, Applied Optics, 10,
1971.

We next furnish a simple derivation of linearly polarized LP-mode fields and their dispersion relations
following Gloge [1971].
Postulate that transverse components of the vector wave equation describing a step-index fiber can be
satisfied by the linear polarized fields
( )
Hy j z cos(m ) Jm ( Xr
a )
Ex = =e (35)
⌘1 sin(m ) Jm (X)

and ( )
Hy j z cos(m ) Km ( Yar )
Ex = =e (36)
⌘2 sin(m ) Km (Y )
within core and cladding regions, respectively, while Ey = Hx = 0 everywhere. With these assumptions,
and focusing only on cos(m ) mode for the moment, we can compute the Ez and Hz from Maxwell’s curl
equations as
1 @Hy 1 @Ex
Ez = and Hz = (37)
j!✏ @x j!µ @y
in the absence of Hx and Ey (HW problem). We start with
8 0 ( Xr ) 9
< ⌘1 XJm J m ( Xr ) @
1 @Hy j z j!✏1 [cos(m ) a
aJm (X) @x
@r
+ Jm (X) @x cos(m
a
)] =
Ez = =e Y K 0 ( Y r ) @r K m ( Yar ) @
(38)
j!✏ @x : ⌘2
) aKmm (Ya) @x )] ;
j!✏2 [cos(m + Km (Y ) @x cos(m

inside and outside the core, and, likewise

8 0 ( Xr ) 9
< XJm J m ( Xr ) @
1 @Ex j z
1
j!µ1 [cos(m ) a
aJm (X) @y
@r
+ Jm (X) @y cos(m
a
)] =
Hz = =e Y K 0 ( Y r ) @r K m ( Yar ) @
. (39)
j!µ @y : 1
) aKmm (Ya) @y )] ;
j!µ2 [cos(m + Km (Y ) @y cos(m

We can test the validity of these results by trying to recover (35)-(36) from (38)-(39) by using Maxwell’s
curl equations once more and finishing the entire algebra started above. When that is done one finds Ex

15
and Hy expressions a little different from (35)-(36) that we started with, but the differences are negligible
when “weak guiding” condition is valid, that is when ⌘1 ⇡ ⌘2 , µ1 ⇡ µ2 , ✏1 ⇡ ✏2 .
Accordingly, we can accept our postulated and linear polarized solutions (35)-(36) as valid solutions (or
good approximations) of the step-index optical fiber problem under weak guiding conditions so long as
we identify the “quantized” and permissible X-numbers of these LPml modes — that requires matching
the tangential field components (38) and (39) at the r = a interface. With ⌘1 ⇡ ⌘2 , µ1 ⇡ µ2 , ✏1 ⇡ ✏2
simplifications we obtain from each one of those expressions similar constraints like
0 (X) @r
XJm J m (X) @ 0 (Y ) @r
Y Km K m (Y ) @
cos(m ) + cos(m ) = cos(m ) + cos(m )
aJm (X) @x Jm (X) @x aKm (Y ) @x Km (Y ) @x
both of which simplify identically as our LPml-mode dispersion relation
XJm0 (X) 0 (Y )
Y Km
= (40)
Jm (X) Km (Y )
after dropping identical terms from both sides of the equality.
We can now obtain the ’s of LPml modes from this simple dispersion relation using straightforward
graphical procedures that we are very familiar with. Using Bessel identities, this dispersion relation can
also be expressed as (HW problem)

XJ m±1 (X) Y K m±1 (Y )

=± . (41)
Jm (X) Km (Y )

Note that when using (40) or (41) to compute X-numbers of LPml modes from LHS and RHS intersec-
tions, remember to start with m = 0 in order to identify the fundamental mode LP01 that is the same as
our good old HE11-mode that was obtained from HEml-mode dispersion relation (34) by using m = 1! —
so, m0 s are a little different in these two contexts, be careful!
You can also use (40) or (41) with m = 1 to compute the X-number of LP11-mode approximating the
almost identical X-numbers of the TM01, TE01, and HE21 modes that we encountered and studied earlier.
Overall, dispersion relation (41) describes 4-fold degenerate and weakly-guided LPml modes with trans-
verse core electric fields ( )
Xr
j z Jm ( a ) cos(m )
E? = e {x̂, ŷ} (42)
Jm (X) sin(m )
and transverse core magnetic fields
( )
j z Jm ( Xr
a ) cos(m )
H? = e {ŷ, x̂} (43)
⌘1 Jm (X) sin(m )

accompanying some weak Ez and Hz fields in the propagation direction. These modes for m 0 and l 1
provide a complete description of the available guided modes in practical step-index fibers to an excellent
approximation. The dispersion curves for these modes are shown in Figure 12.

Well, this is THE END.

If you liked all this and want to learn more and experiment with step-index and graded-index fibers in
the lab, then take ECE 465 next fall — it will be a new 4 hr course with only ECE 350 prerequisite that
will count as an ECE lab.

Erhan Kudeki, Nov 24, 2012

16
Figure 12: X solutions of LPml-mode dispersion relation as a function of V for LP01 (red), LP11 (blue),
LP21 (green), LP02 (magenta), LP31 (orange), LP12 (yellow) modes. Dispersion curves populate
only the X < V region (the bottom triangle underneath the dashed X = V line). Each LPml-
mode propagates only when V > Vc,ml , where Vc,ml is lth zero of Jm 1 (X) in X; also, permissible
X for each LPml-mode is Vc,ml < X < Xm,ml , where Xm,ml is the zero of Jm (X) in X just
above Vc,ml — see Figure 11 for a check of this rule for 0l and 1l modes.

17