416

JOIRNAL OF I ICHTWAVF 1ECHNOI OLY VOL 11 NO

MARCH 1993

Loss Calculations for Antiresonant Waveguides

J-.
Fig. 1. Refractive-index profile of either a symmetrical planar or a cylindrical multilayered leaky waveguide (n.y 2 111 ).

%-I

I. INTRODUCTION
N most optical waveguides, light is confined inside a region, the core, where the refractive index is higher than in the surrounding medium, the cladding. There is, however, a class of optical waveguides - leaky waveguides - in which the core index is lower than that of the cladding. These waveguides have unique properties which have been exploited for several applications, such as building waveguides on highindex substrates [1]-[3], confining light in a gaseous medium [4], [5] and remote switching [6]. A leaky waveguide does not support any guided modes and its transmission capabilities are therefore limited by leakage loss. This loss can, however, be reduced to an acceptable level providing the refractive-index profile of the waveguide is carefully designed. The most effective way of realizing a low-loss leaky waveguide is to surround the core with highly reflective antiresonant claddings. Antiresonance is a well-known effect in Fabry-Perot interferometers: light is in antiresonance when it interferes destructively with itself inside a Fabry-Perot cavity; the reflectivity of the interferometer is then at its maximum. Antiresonant waveguides use this effect to provide a transverse confinement of the light in the core. There are in the literature a few examples of antiresonant waveguides, such as ARROW [1]-131 and hollow [4], [5] waveguides. The theory published so far only covers a limited number of possible leaky waveguides. For instance, no one has, to our knowledge, treated the case of hybrid (HE and EH) modes of cylindrical antiresonant waveguide. Our objective in this paper is to present a general study of the properties of antiresonant waveguides using modal solutions of the vector wave equation. In particular, we examine how the leakage loss of these waveguides is influenced by the geometry of their refractive-index profiles, by wavelength and by the presence of absorption. In doing so, we derive some very useful and general analytical expressions which provide,
Manuscript received June 5. 19Y2. J.-L. Archambault is with Optoelectronics Research Centre, University of Southampton, Southampton, SOY 5NH. United Kingdom. R. J. Black, S. Lacroix, and J. Bures are with Ecole Polytechnique de Montreal. DCpartement de genie physique, C.P. 6079. succ. A. Montreal, Quebec, H3C 3A7, Canada. IEEE Log Number 9206939.

in most cases of interest, a good approximation of the loss. These expressions are applicable to TE and TM modes of slab waveguides or TE, TM and hybrid modes of cylindrical fibers, with step-index profiles of up to four layers. For threelayer waveguides, we also present an expression giving the leakage loss of these various modes when the second layer is absorbing.
11. MODESOF MULTILAYERED LEAKYWAVEGUIDES

Fig. 1 illustrates the refractive-index profile of a multilayered leaky waveguide. This profile can represent either the refractive-index of a symmetrical slab waveguide or of a cylindrical fiber. I t consists N layers of different materials: a central region, the core, of half-width or radius p 1 and uniform index ra,; N - 2 internal claddings, each of thickness f , = 0, - pq-l and refractive index n q ,where 2 < q < N - 1: finally, an external cladding which extends to infinity and has a refractive-index equal to 7 1 5 . The refractive-indexes can be in general complex to account for the presence of absorption. We consider the waveguide of Fig. 1 to be leaky if n1 5 n ; ~ . We choose to study this waveguide by calculating its discrete leaky and guided modes. Each mode is characterized by its propagation constant, /j. which is real for guided modes and complex for leaky modes. The effective index of a mode is given by n,,R = / Y / k . where ,b is the real part of p and k = 2n/X is the wave number in vacuo, with X representing the wavelength. The type of transverse dependence of the modal fields can be deduced by comparing this effective index ~ nlq. to the local refractive index of the waveguide. If n , < the modal fields are oscillatory in layer q: if n,,R > n,q3 they are evanescent. A mode is therefore guided if 7 2 , ~ > 7 1 , . ~ and leaky if ne^ < n , . ~ The . imaginary part of [?,[Y. is the attenuation coefficient of the modal fields; a = 2/?2 is the power attenuation coefficient of the mode. The modes of a multilayered ( N > 2) waveguide may be separated into core and cladding modes, depending on their
0 1Y93 IEEE

0733-X724/93$03.00

ARCHAMBAULT er al.: LOSS CALCUL.ATIONS FOK ANTIRESONANT WAVEGUIDFS

417

effective indexes and the transverse distributions of their modal fields. Because of the condition n 1 5 71 \ . all core modes of a leaky waveguide are leaky modes. Some of the cladding modes may however be guided, if their effective index satisfies the condition 711;- < nee < n i . The mode numbering used throughout the paper refers only to core modes and is in particular associated with the number of zeros of modal fields in the core, regardless of the field distribution in the claddings.
111. RESONANCE AND ANTIRESONANCE A leaky mode can be in resonance or in antiresonance in each of the internal claddings of a multilayered waveguide, providing its effective index is lower than the local refractive index, i.e., if the modal fields are oscillatory in the internal cladding considered. To derive the condition for resonance or antiresonance, we move to geometrical optics, replacing each mode by a plane wave characterized by a wave vector k with (kl = kn,. The components of k along the longitudinal and transverse directions of the waveguide are respectively /I and U,. with
71,

Otherwise, the even values of 1) give rise to resonance and the odd ones to antiresonance. Equation (4) holds two interesting limiting cases

w Jm.
=

(1)

Each of the N - 2 internal claddings of the waveguide is equivalent to a transverse cavity in which the plane wave oscillates (if n,ff < n q ) . At each round trip in one of these cavities, the wave accumulates a phase

In the first case, the critical thicknesses t,, are inversely proportional to U , and are thus different for each mode; they are, however, wavelength-independent. In the second case, these thicknesses are proportional wavelength but do not depend on the mode number. This creates the possibility of having a leaky waveguide where the different low-order core modes are simultaneously in antiresonance, for a given wavelength. Equation (5) also sets the tolerances on the thicknesses of the internal claddings for the actual fabrication of antiresonant waveguides. The error in controlling the thickness of an antiresonant layer should be a fraction of the difference between two successive critical thicknesses. For an internal cladding falling in the first case of ( 5 ) , the tolerance can be fairly large, of the order of p 1 / 2 for the fundamental mode. In the second case, however, the tolerance could be typically A/10. which can be very difficult to achieve, especially in fiber geometry.

4 = 2fq71; + 4,

1v. SOLVING THE

EIGENVALUE EQUATION

where qbq(= 0. 7r or 27r) is due to the reflections in ,oq-l and p,. If this phase 4 is an even multiple of 7 r . then the wave interferes constructively with itself and is in resonance in layer q of the waveguide. The plane wave is of course in antiresonance if the phase shift is an odd multiple of 7 r . TO derive an analytical expression for 4. we consider the case where the waveguide parameter for the core, defined as V = k p l d m [ 7 ] , is large. In the limit where V tends to infinity, the transverse propagation constant of a core mode is given by 111 = U-Jpl. For the m t h core mode of a slab waveguide, Crx = ( m 1)7r/2. In a cylindrical fiber, U , = j u p .where v and 11 are respectively the azimuthal and radial mode numbers and j U , , is the pth zero of Bessel function .JU. When V is large, 111 7.z U X / p 1 and, consequently,

The propagation constants of the modes are found by solving the eigenvalue equation of the waveguide numerically in the complex plane. Alternatively, we are able to derive useful analytical expressions for the eigenvalues using a firstorder perturbation approach [8J.The first step of this derivation is to write the eigenvalue equation of the leaky waveguide in the form

For step-index profiles, obtaining (6) is straightforward as the fields of slab waveguides and cylindrical fibers can then be expressed as combinations of trigonometric and exponential or Bessel functions. Upon performing a Taylor expansion of G(P) around /j - /Y, and keeping only the zeroth- and first-order terms, we find

(3)
We can thus write the resonance and antiresonance condition from ( 2 ) as

(7)
since [j' is real by definition. This last expression is accurate when ,9' is small compared to 13', which corresponds to lowloss modes. The difficult part of this approach is to reduce (7) to a simple, useful expression, since the eigenvalue equation of multilayered waveguides becomes extremely cumbersome as the number of layers is increased. The results we have obtained using (7) are summarized jn Table I. Here, we have used common expressions for the attenuation coefficients of TE and TM modes of both slab waveguides and cylindrical fibers. For T E modes, one must use

(4)

where p is a positive integer. If n , is between n q - l and n q + 1 . then 4 , = 7r and the even values of p correspond to antiresonance while the odd ones correspond to resonance.

JOURNAL OF LIGHTWAVE TECHNOLOGY. VOL. 11. NO. 3, MARCH 1993

TE,, , antiresonance ( p = I I )
1

T I

attenuation coefficient, a f v )

minimum

maximum

;z^ 0.5

z
U

10

dh
Fig. 2. Field intensity distribution of the TEo leaky core mode of a three-layer planar waveguide. The limits of the different regions of the waveguide are shown by the thin vertical lines. The thickness of the high-index internal cladding is 2.4GX ( p z 11) so that the mode is in antiresonance. A four hundred times magnification (dotted line) shows a standing wave pattern of five and a half cycles inside the internal cladding.

the definition Cq = ii;pl and, for TM modes, Lrq = ! i r ; p 1 / n i . Table I also provides the means to approximate the attenuation coefficients of hybrid modes of a cylindrical fiber. In the first approximation, a hybrid mode is composed in equal parts of T E and T M polarized waves. Therefore we expect the attenuation coefficient of a hybrid mode to be given by the average value of the corresponding coefficients for TE and TM polarizations, i.e.,

(u(-'-)(HE)z i(o(-'-)(TE)+ a(.'-)(TiLf)).

(8)

We can confirm this assumption in two ways. In the case where N = 2. the attenuation coefficients obtained analytically from (7) satisfy this last relation exactly. For N > 2, we show in the following that (8) is in good agreement with numerical results, within the regions of validity of our analytical expressions (see Figs. 7 and 12). Due to the complexity of the analytical calculations involving hybrid modes for N > 2 >we will adopt this approximation. One can readily verify that the various expressions of Table I are consistent with each other. For example, by setting c2 7 - C3 = (n.2 = 713 = ? L A ) one finds, as expected, a(2) ( y ( 3 ) = ( ~ ( ~ 1Similarly, . setting l T 3 = U, yields
Q(3)

index. These waveguides were extensively studied in the sixties by Marcatili and Schmeltzer [9]. The results we have derived in Table I for TE, TM and hybrid modes, in the particular case where 7 ) 1 = 1. are identical to those of [9]. These analytical expressions show that the modal attenuation coefficients are inversely proportional to the third power of the core's width. By increasing the dimensions of the waveguide, one can therefore reduce these coefficients considerably, to the expense of compactness. However, as pointed out by Marcatili and Schmeltzer and later confirmed experimentally by other authors [lo], the bending loss of leaky waveguides increases with core size. B. N = 3 Waveguides The modes of a three-layer leaky waveguide can either be core ne^ < nl) or cladding ( n l < n,ff < 71.2) modes. Cladding modes are guided if n,ff > 71.3. Figs. 2 4 illustrate the field intensity distributions of the lowest order TE core and cladding modes of a slab waveguide with 71.1 = 723 = 1.0, n.2 = 1.5, p1 = 5X (V 3 5 ) . For these particular values of 711 and 7i.2, the critical thicknesses of the internal cladding are approximately given by (5-2) as t 2 p z 0.224pX. In Fig. 2, p = 11 ( t 2 z 2.46X) and the mode is therefore in antiresonance. The field distribution inside the core is very similar to that of a guided mode of a conventional waveguide. By increasing the scale four hundred times (dotted curve), we see that light forms a standing wave pattern in the highindex cladding, going through exactly p / 2 = 5.5 cycles. In the external cladding, the intensity appears to be constant, but is in fact increasing very slowly, which is characteristic of a low-loss leaky mode. Fig. 3 shows the same mode very close to resonance inside the second layer, with p = 10, t 2 z 2.24A. The modal intensity distribution inside the core is now almost flat, as the internal cladding is effectively transparent to the mode. The standing wave pattern now undergoes five cycles. The intensity grows rapidly in the third layer, indicating a

= o(Jl,

The last two columns of the table give the minimal and maximal values of when the internal claddings' thicknesses are varied. For these results, we have assumed that the refractive indexes of the successive layers alternate, i.e., 712 > 711. 763 < 712 and 7j.k > 71,3.The star (*) appearing in the last row indicates that the internal cladding of this three-layer waveguide is absorbing, i.e., 71; = 712 + i y . where 712 and y are respectively the real and imaginary part of the complex refractive index.

v.
A.
J \ A

DISCUSSION AND COMPARISON O F RESULTS

= 2 Waveguides

These are the simplest leaky waveguides, consisting of a core surrounded by an infinite cladding of higher refractive

ARCHAMBAULT ff

U/.: LOSS

CALCULATIONS FOR ANTIRESONANT WAVEGUIDES

419

TE,, , resonance (p=IO)

HE
loo E

1P

( p - l = # of zeros in core)
I

3 0.5 2
U

0,
U

10

xlh
Fig. 3. Same as in Fig. 2, but this time the internal claddings thickness is set at 2 . 2 4 X ( p = 10). putting the TEo mode close to resonance. The intensity is now very large outside the core with the standing wave pattern going through five cycles.

t,/O. 224L
Fig. 5. Numerically calculated attenuation coefficients of the first four HEl,, modes of a three-layer cylindrical waveguide, with 1 1 1 = n 3 = 1.0, 112 = 1.3. and p i = . :A. The second layers thickness t Z is varied, driving the modes through antiresonances (minima) and resonances (maxima) at critical thicknesses given by t 2 % 0.224pA. with 1 an integer.

first cladding mode

U OO 5

10

x/h
Fig. 4. Field intensity distribution of the lowest-order TE cladding mode of the same leaky waveguide as in Figs. 2 and 3, with the second layers thickness set at 2X. This mode is guided and highly confined in the high-index region.

large attenuation coefficient. Fig. 4 illustrates the intensity distribution of the lowest-order cladding mode with t 2 = 284. This mode is guided, decaying exponentially in both the core and the external cladding. It is almost totally confined to the second layer and would therefore be rapidly attenuated if that region were absorbing. The attenuation coefficients of the first four HE1, modes of a nonabsorbing three-layer cylindrical fiber are plotted in Fig. 5 against the thickness t 2 of the internal cladding. Here, we have used alternatively solid and dashed lines for the four different modes to aid the eye. The parameters used for this numerical calculation are the same as before (721 = 713 = 1.0, 712 = 1 . 5 . p l = 5X). At evenly spaced critical thicknesses, given by (4), the attenuation coefficients go through flat minima (antiresonances) and sharp maxima (resonances), reminiscent of the transmission function of a Fabry-Perot cavity. This example enters the second limiting case of (5) and, consequently, the positions of the maxima and minima appear to be independent of the mode number. Examining closely the plots of Fig. 5 around the first resonance (t2/0.22484 = 2 ) . we notice that the attenuation

coefficient of the HE11 mode suddenly drops to zero after going through a maximum; the HE12 mode then appears to take the space left vacant by the fundamental mode. Similarly, the other higher order HE1, modes take the place of the previous HEl(,L-l) mode when they go through resonance as the thickness t 2 is increased. The explanation of these results comes from Fig. 6, which plots the effective indexes of these various modes versus t 2 . As we force the HE11 mode through resonance by increasing t 2 . we see that its effective index becomes higher than the core and external cladding index, 711 = n 3 = 1.0. The mode has then become a guided cladding mode and its intensity distribution has shifted from the core to the second layer. Each resonance corresponds in fact to the cutoff of a guided cladding mode [3].This pattern repeats itself at each resonance. The same is true for modes of other symmetries. As f 2 is increased, the total number of field zeros of each mode is conserved; the zeros simply shift toward the outer layers of the waveguide and, consequently, the radial mode number, given by the number of zeros in the core, decreases. Miyagi and Nishida proposed using a three-layer waveguide in antiresonance to reduce the leakage loss in 1980 [4]. They derived an analytical expression for the attenuation coefficients of TE modes using a transverse transmission-line model, in the particular case where 711 = 723 = 1. This expression agrees with the more general one presented in Table I for N = 3 . In Fig. 7, we compare our approximate analytical results (from (8)) to exact numerical solutions by calculating the attenuation coefficient of the HE11 mode of a leaky fiber. In this example, rtl = 713 = 1 . 0 . 7 1 2 = 1.5 and we use four different values of core radius giving V-numbers evenly between 17.5 and 140. As expected, the analytical expression gives its best results around antiresonance, where the attenuation coefficients are small. The error increases rapidly close to the resonance, where the attenuation coefficients can become very large. The comparisons of Fig. 7 indicate that the average relative error (integrated over f 2 ) is inversely proportional

.IOUKNAI

or

LIGHTWAVF TFC'HF.;OI.OG-Y. VOL. I

I,

NO 3. MARCH 1493

1.02 1.oo

50
40

0.98 0.96
0.94

30
20
'; .....$'"
'

1 0.92 I.'. 0 1

10 2
'

'

'

'

'

'

n
8

1.05

1.1

t2I0.224h
Fig. 6. Effective indexes of the modes of Fig. 5. Each modc is a leaky core mode if n e R < 1 1 1 = 1.0 (as in Figs. 2 and 3) and a guided cladding mode if ne^ > r t l (as in Fig. 4).Resonance of the HE11 core mode corresponds to the cutoff ( n , = ~ 11 1 ) of a cladding mode.

n3
Fig. X. Los\ reduction factor of the TEo mode of a .\-= 3 planar waveguide = 1 0 X ) relative to the attenuation coefficient ( n l = 1.0, 112 = 1..5. and n i l ) of the TEo modc of the corresponding .\- = 2 waveguide. The external cladding's index is varicd and the thickness of the internal cladding chosen at each point to put thc mode in antiresonance.

1oo

13-2

1o

-~
I

t7/0.224h Fig. 7. Comparison of the attenuation coefficient of the HE1 1 mode calculated numerically (solid lines) and via the analytical expressions of Table I (dotted lines) for a 1 -= 3 cylindrical waveguide with n1 = 11.3 = 1.0, na = 1..5. and p i = 2 . j X , i X , 10X. and 20X. giving l.-values ranging from 17.5 to 140.

to @. As V is increased, both the perturbation approach and the approximation I L ~N, l J x / p l become more accurate and, consequently, the error diminishes. The error would have actually been smaller for TE or TM modes, since (8) introduces an additional approximation for hybrid modes. We may now use the analytical expressions of Table I to determine how much effect an antiresonant cladding can have on the leakage loss. If a mode is in antiresonance in a three-layer leaky waveguide and 712 > n 3 2 n 1 . then : sin(u5t2) = 1, C2 = cos ( u ; t 2 ) = 0 and a('3)= a:?" = a(2)U3/C72.For a TE mode, if V is large then U2 z V : if n 1 = n 3 , U 3 = U 1 N , U,. In this particular case, = c ~ ( ~ ) U , / k :With the addition of an antiresonant cladding, the modal attenuation coefficients of a two-layer waveguide are thus reduced by a factor l r / U x . Since I-values of leaky waveguides are typically large, this is a very significant reduction of the leakage loss. At resonance, under the same conditions, S2 = 0, C2 = 1 leading to
($3)
(3) ( - Qmax - Q W 2 / C T : 3 .

Since the perturbation calculation assumes small loss, this estimate of the attenuation coefficient is less reliable than at antiresonance, but it gives the right order of magnitude. The attenuation coefficient of the two-layer waveguide is in this case multiplied by the same factor that was dividing it at antiresonance, i.e., z za(2)I~~/~ inj our x example. If we choose n1 # 7?,3. however, the effect of antiresonance o r resonance can be much smaller. In Fig. 8, we plot the loss reduction factor, rr(2)/oI(XI,, = { T ~ / U S . as a function of 763. for the TEo mode of a three-layer slab waveguide in antiresonance. ~ I, 7 1 2 = 1.5 and p l = 1OX. As can be seen, We choose 7 1 = the loss reduction factor drops extremely rapidly as soon as 7 i ~ : j is increased above the value of n l . This observation is easily interpreted by refering to geometrical optics. Low-order modes can be associated with rays which travel at grazing incidence with respect to the core boundary. The resulting Fresnel reflection coefficient is very high. If 113 = 711, the reflection coefficient at the second interface (between the internal and external claddings) is the same. If 71,3 is increased, the ray angle relative to the interface increases in the external cladding and the reflection coefficient drops rapidly. In general, antiresonance is effective in reducing modal attenuation only if the low-index claddings of a multilayered leaky waveguide have the same refractive index as the core. From the results of Table I, we can also calculate an average attenuation coefficient, ( r ~ ( ' ~ by ) ) integrating with respect to t 2 over one period. We then find (a")')= n ( 2 )as ; already pointed out by Bornstein [5]. The implication of this result is that if t 2 varies randomly along the waveguide with a standard deviation of more than the difference between two critical thicknesses, the effect of the third layer is neutralized and the waveguide is expected to behave like a two-layer waveguide.

(x!2ix

C. N = 3 Waveguides
The four-leaky waveguides are of particular interest as they include the ARROW waveguide [ 11-[3], which constitutes the most successful application of antiresonance in waveguide

4RCHAMBAULT el

a/

LOSS CALCULATIONS FOR ANTIRESONANT WAVEGUIDE\

421

ARROW waveguide

equivalent waveguide

IO0
lo-'

T E m odd modes

3 8 lo-*
v
h

10.~

ts
Fig. 9. Equivalent waveguide model of an ARROW waveguide. The air-core interface of the ARROW acts as a mirror for the lowest-order modes; an equivalent symmetrical waveguide may be constructed using the actual waveguide and its mirror image across that interface.

1o

-~
5
10 15

10.~

I
t,/h

geometry. Fig. 9 describes how results derived for a four layer symmetrical slab waveguide can be applied to an ARROW waveguide. The lowest order core modes of an ARROW waveguide are guided on the air side and leaky in the substrate. Their fields almost vanish at the core-air interface because of the large index difference. This interface is in practice equivalent to a highly reflective mirror. This is why we can build the equivalent symmetrical waveguide of Fig. 9 using the actual ARROW waveguide on the right hand side of the interface and its mirror image on the left. Because modal fields must vanish at the interface, only the odd modes of the equivalent waveguide are allowed. In [2], Baba er al. have published an analytical expression for the attenuation coefficients of the modes of an ARROW waveguide when these modes are in antiresonance in both internal claddings. Setting S2 = S3 = 1 and ( 3 2 = C? = 0. we find that the expression for in Table 1 is in accordance with their result. Most of the discussion concerning the three-layer leaky waveguide can be transposed directly to the four-layer case. A four-layer waveguide has two internal claddings whose thicknesses may be set independently for resonance or antiresonance. The attenuation coefficient of a mode is of course minimum if this mode is simultaneously in antiresonance in both internal claddings. The dependance of the attenuation coefficients on the thickness of the third layer, t ~ is, given in Fig. 10, for the first odd TE, modes (711 = 1,3.5.7) of a four-layer slab waveguide. The fixed parameters are: 7b1 = 7 ~ 3 = 1.5, n2 = 714 = 2.0, p1 = 5X. t 2 = 1.7X which is approximately the condition for antiresonance with p = 9 in (5-2). There are two noticeable differences between this plot and that of Fig. 5, where we were varying the thickness of a high-index layer. The critical values of tJ are this time given by the first limiting case of (5) (since n l = 773) which gives t 3 p = p p l / ( m + 1).They now depend on the mode number m. which is why the minima and maxima of different modes do not coincide in Fig. 10. The other main difference is that the attenuation coefficient of the lowest order mode (TE1) does not fall to zero but instead decays slowly as f J is increased beyond the resonance. Similarly to what we described in Fig. 6 for the HE11 mode, the TE1 core mode becomes a mode of the second internal cladding; but since the refractive-index 7) J is lower than that of the external cladding, n 4 . the cladding mode is in this case leaky. As in Fig. 6, when the TE, mode goes through resonance, the TE,+2 mode (the next higher order

Fig. 10. Numerically calculated attenuation coefficients of the first three odd TE modes of a . Y= 4 planar waveguide ( f i r = 773 = 1..5, n 2 = n4 = 2.0, = SA. and t 2 = 1 . T X ) versus the thickness of the low-index third layer. The modes of the second internal cladding are all leaky, since n q > n3.

mode of the same symmetry) takes its place. However, because the different modes are not simultaneously in resonance, the TE,,+2 mode becomes a cladding mode between its resonance and that of TE,,,. We can derive an analytical expression for attenuation coefficients of the modes of the third layer using an equivalent waveguide model, similar to what is described in Fig. 9. The third layer is bound on one side by the external cladding and, on the other, by a high reflector: the second layer where light is in antiresonance. The second layer thus acts as a mirror and we can build an equivalent waveguide formed by the third and fourth layers and their mirror image about the interface between the second and third layers. The result is simply a two-layer leaky waveguide with a core of index n 3 and half-width t3 surrounded by an infinite cladding of index n.4. Because of the presence of the mirror, only the odd modes of this equivalent waveguide are permitted ( m = 1 , 3 , 5 . . .). The attenuation coefficients of the cladding modes are then given directly by the expression for in Table I. On the other hand, the expression for a(') in Table I approximates the attenuation coefficients of the various core modes of the four-layer waveguide. In Fig. 11, we compare the results given by both these expressions to the numerical calculations of Fig. 10. Once again, the agreement is very good, especially when the loss is low. Upon integrating ( Y ( ~ ) over f : j , we find as expected the average attenuation coefficient (a(4)) to be equal a ( 3 ) , the attenuation coefficient of the three-layer waveguide. Again, this means that an insufficient control of the internal cladding's thickness would neutralize its effect. is minimum when a mode is in antiresonance in both internal claddings, i.e., when S2 = S3 = 1 and C2 = C3 = 0 (if n1 < 722, n.2 > n3 and then reduces to n . 3 < 714); (9) From this result and from further numerical calculations for N > 3 (as in Fig. 13, for y z 0). we have deduced and

JOUR*.iAI. OF LIGHTWAVE TECHNO1 OGY. VOL I I . NO. 3, MARCH 1903

10" lo-'
h

N-layer absorbing waveguide

1o-2
h

b ,

1o
1o

-~

ts

-~
0

\ 1 2 3 4 5
t,lh

Y
Fig. 11. Comparison of analytical (dotted) and numerical (solid) results for Fig. 13. Loss reduction factor versus absorption for the TE1 mode of a planar the first part of Fig. S . The attenuation of the core modes are waveguide, with . y = ,3. 1.and 5 . The parameters are: 111 = 113 = 115 = 1..5> calculated from ,($) and those of the cladding ,,,"des from (,(2) in ~ ~ 1. b l ~ ,I; = t1: = 2 . 0 + (-,, p1 = SA, t 2 = t , = 1.7X. and t 3 = 2 . 5 X . so that the mode is everywhere in antiresonance. For = 3 . analytical results calculated from a ( ' 3 1 in Table I are shown as a dotted line with open circle, overlapping the numerical calculation. 1O0

5
ts

10.'

'-'O

t,/O .224 h
Fig. 12. Same as in Fig. 5 (.Y = 3 fiber, t i , = I , . ] = 1.0, p1 = > A ) . except that the second layer is now absorbing, with a refractive-index 1 1 ; = ~2 i ; = 1 . j + 0.01). Results from the expression for n ( 3 J - in Table 1 are shown in dotted line for comparison.

now have large attenuation coefficients ( ~ 0 . 5 dB/X), 5 much larger than that of the fundamental mode at resonance ( ~ 0 . 1 dB/X) making the connection impossible. Consequently, the HEl2 mode does not replace HE11 at resonance. The second effect is that the contrast between resonance and antiresonance diminishes as t 2 is increased. The reason is the following: as the absorbing layer becomes thicker, it effectively masks the outer cladding to the core modes; light which escapes the core is absorbed before it can be reflected back into it and therefore, the internal cladding becomes equivalent to an infinite cladding. The three-layer leaky waveguide thus tends to act as a two-layer waveguide. This interpretation is indeed the expression for rr(3)- in I. From this confirmed expression, we see that, as the product ~ t increases, 2 the hyperbolic tangent tends to 1, U; z U,* z U2 U, and o(~) z' ~ ( ~ On 1 . the other hand, if the condition

verified the following general relation:

Q ,

(.v-l)~Av-l

U\

if n-, -1

< nAv
is verified, then the effect of absorption is small. In Fig. 13, we calculate the loss reduction factors, w ( ~ ) / o ( " ) . of the TE1 modes of multilayered planar waveguides, with N = 3 , 4 or 5. We choose, when applicable, = n; = 2.0 27, p1 = 5x, 711 = I l j = 7 t 3 = 1.5, 7); f 2 = 1.7X, t 3 = 2.5X, t4 = 1.7X. with a variable absorption y in layers 2 and 4. These values of the thicknesses t, insure that the mode is everywhere in antiresonance. It is clear from this figure that the beneficial effect of the outer layers of an antiresonant waveguide vanishes rapidly as y increases. ( ~ ( ~ and 1 tu(') are very close together and For 2 > for y > all three attenuation coefficients merge with m ( 2 ) / r t ( 1 y )z 1. We have included in Fig. 13 results from the analytical expression for (dotted line with open circles). The agreement with the numerical results is in this case excellent (since the mode is always in antiresonance) and the two curves are hardly distinguishable on this scale.

'

VI. &SORBING

WAVEGUIDES

We now examine the effect of material absorption on the modes of leaky waveguides. As a first example, we look once more at the HE modes of a cylindrical fiber, setting the parameters as in Fig. 5, with the difference that the refractive-index of the second layer now has an imaginary component, y = 0.01. This value of y would correspond to an attenuation coefficient of 0.55 dB/X for light propagating in a bulk material. Plotting the attenuation coefficients of the HE11 and HE12 modes (Fig. 12) and comparing to the results of Fig. 5, we see that the absorption has had two noticeable effects. First, at resonance, the fundamental mode does not become a cladding mode. This is because cladding modes

ARCHAMBAULT el al.: LOSS CALCULATIONS FOR ANTIRESONANT WAVEGUIDES

423

VII. CONCLUSION

In this paper, we have examined many different cases of multilayered leaky waveguides, in planar and cylindrical geometry, using a leaky mode analysis. Some very useful and general analytical expressions have been derived from a perturbation approach, giving the attenuation coefficients of TE, TM, HE, and EH polarized modes for step-index waveguides of up to four layers. In the case of a threelayer waveguide, the expression takes into account material absorption. All these analytical results have been compared to numerical solutions and found to be very accurate when the modal attenuation is not too large. In the light of this study, we were able to draw some general conclusions regarding the properties of antiresonant waveguides. First, we saw that the antiresonance condition can fall into two very different limiting cases, depending on the refractive-index of the layer considered. In one of these cases, the condition is wavelengthindependent, in the other, it is mode-independent. It is clear that antiresonant claddings can reduce the modal attenuation coefficients of a leaky waveguide considerably, provided that three conditions are met: 1) the low-index regions of the waveguide must have the same refractive-index as the core; 2) the thicknesses of the antiresonant claddings must be controlled very accurately; 3) the product of the absorption coefficient and the thickness of an internal cladding must be kept small. Otherwise, the effect of antiresonance is neutralized. REFERENCES
[ l ] M. A. Duguay. Y. Kokubun, T. L. Joch, and L. Pfciffer, Antiresonant reflecting optical waveguides in SiO2-Si multilayer structures, Appl. Phys. Lett., vol. 49, no. 13, 1986.

T. Baba, Y . Kokubun, T. Sakaki, and K. Iga, Loss reduction of an ARROW waveguide in shorter wavelength and its stack configuration, 1. Lighmave Technol., vol. 6, p. 1440, 1988. T. Baba and Y . Kokubun, High efficiency light coupling from antiresonant reflecting optical waveguide to integrated photodetector using an antireflecting layer, Appl. Opt., vol. 29, p. 2781, 1990. M. Miyagi and S. Nishida, A proposal of low-loss leaky waveguide for submillimeter waves transmission, IEEE Trans. Microwave Tech., vol. 28, p. 398, 1980. A. Bornstein and N. Croitoru, Experimental evaluation of a hollow glass fiber, Appl. Opt., vol. 25, p. 355, 1986. M. Cantin er al., Remotely switched hollow-core antiresonantreflecting optical waveguide, Opt. Lett., vol. 16, p. 1738, 1991. We have adapted this definition in order to keep I . real if ng > n l . A. W. Snyder and J. D. Love, Optical Waveguide Theory. New York: Chapman and Hall, 1983. E. A. J. Marcatili and R. A. Schmeltzer, Hollow metallic and dielectric waveguides for long distance optical transmission and lasers, Bell Syst. Tech. J . , vol. 43, p. 1783, 1964. C. A. Hill, R. M. Jenkins, and R. W. J. Devereux, Transmission of linearly polarized infrared light through curved hollow dielectric waveguides, IEEE J . Quantum Electron., vol. 24, p. 618, 1988.

J.-L. Archamhauit, photograph and biography not available at the time of publication.

R.J. Black, photograph and biography not available at the time of publication.

S. Lacroix, photograph and biography not available at the time of publication.

J. Bures. photograph and biography not available at the time of publication.