On band gaps in photonic crystal fibers
Abstract
We consider the Maxwell’s system for a periodic array of dielectric ‘fibers’ embedded into a ‘matrix’, with respective electric permittivities and , which serves as a model for cladding in photonic crystal fibers (PCF). The interest is in describing admissible and forbidden (gap) pairs of frequencies and propagation constants along the fibers, for a Bloch wave solution on the crosssection. We show that, for “precritical” values of i.e. those just below (where is the magnetic permeability assumed constant for simplicity), the coupling specific to the Maxwell’s systems leads to a particular partially degenerating PDE system for the axial components of the electromagnetic field. Its asymptotic analysis allows to derive the limit spectral problem where the fields are constrained in one of the phases by CauchyRiemann type relations. We prove related spectral convergence. We finally give some examples, in particular of small size “arrow” fibers () where the existence of the gaps near appropriate “microresonances” is demonstrated by a further asymptotic analysis.
1 Problem formulation and main result
We consider the Maxwell’s system
(1.1)  
where the electric permittivity and magnetic permeability adopt two different sets of constant values in the fibers along the direction which are positioned periodically in the crosssectional plane, and in the surrounding matrix.
More precisely, let be the reference periodic cell, be an open bounded subset of , , with sufficiently smooth boundary , and . Let denote the characteristic function of the periodically extended , . Then
where , , , are positive constants.
A Bloch wave type solution to (1.1) is sought in the form
(1.2) 
The interest is in describing the pairs ( is the frequency and is the “propagation constant”) for which there exists a nontrivial solution of the form (1.2) with and quasiperiodic in .
Upon substituting (1.2) into (1.1), we find that , necessarily satisfy the following system of equations
(1.3)  
(1.4)  
(1.5) 
(1.6)  
(1.7)  
(1.8) 
Resolving then (1.4) and (1.6) for and , and (1.3) and (1.7) for and , gives the following representations of the crosssectional components and in terms of the “axial” components and
(1.9)  
(1.10)  
(1.11)  
(1.12) 
assuming
(1.13) 
is not zero. Substituting then (1.9)–(1.12) into (1.5) and (1.8) reduces (1.3)–(1.8) to the following system for and only
(1.14)  
(1.15) 
Setting , multiplying (1.14) and (1.15) by smooth test functions and respectively, adding up and integrating over gives upon integration by parts the following equivalent weak formulation:
(1.16) 
Here the bilinear form is given by
(1.17) 
where , as above,
(1.18) 
and
Notice that the form (1.17) is symmetric, i.e. , as follows by direct inspection. Noticing further that, for any ,
it is readily seen that if satisfy
(1.19) 
then is coercive on ; namely, there exists a constant such that,
(1.20) 
Henceforth, we shall consider a ‘nonmagnetic’ photonic crystal fiber, i.e. we set the magnetic permeability a constant, and we assume that in (1.18). We shall consider the set of pairs for which the problem (1.16) admits a nontrivial solution when approaches from below the critical line . (This line corresponds to the dispersion relation in the matrix  the wavenumber for a plane wave  which appears exactly where loses its coercivity.) More precisely, for a fixed small parameter , we can say that (1.16) admits a nontrivial solution for the pair if belongs to the spectrum of the selfadjoint operator generated by the bilinear form given by (1.17). This form can be conveniently represented as
(1.21) 
Here is easily found by rearranging (1.17) upon substituting ; see (3.3) below. Then the domain and the range of the selfadjoint operator consist of all and respectively, such that ()
Notice that, for a fixed , the operator and hence its spectrum depend on the spectral parameter , cf. (1.17), (1.13), hence one generally deals with an operator pencil. We will see however that, as , it asymptotically bahaves as a conventional spectral problem (although for a “partially degenerating” operator).
So, for converging to zero, we aim to characterise the (FloquetBloch) spectrum . This brings us to our main result
Theorem 1.1.
Let be the spectrum of the operator generated by the bilinear form , see (1.21). Then the set converges to the set
as in the sense of Hausdorff. Here is the set of , , such that there exists a nonzero solution to
(1.22) 
where and the unitary mapping is defined by ;
(1.23) 
where
Remark 1.1.
We say that sets converge to a set as in the sense of Hausdorff when the following two conditions hold:

For every there exists such that as .

Suppose such that converges to some as , then .
Symbolically, we represent this convergence as follows
Remark 1.2.
Corollary 1.1.
The set also admits the following equivalent representation
where is the set of , , such that there exists a nonzero solution to
(1.24) 
where
Here for a given multiindex is a “multicell” of size , and is the periodic extension of the set to the whole space .
The ‘limit spectrum’ has a gap if two adjacent bands do not overlap: i.e. if for some , one has
An immediate consequence of Theorem 1.1 is the following result
Corollary 1.2.
Proof.
Suppose has a gap, i.e. there exists an such that . Let us fix such that .
We now argue that there exists a such that the set denoted by
satisfies the statements of the corollary. For if not, there exists a sequence for which each set contains a pair such that (1.14)(1.15) admits a nontrivial solution, i.e. belongs to . Since , it has a subsequence which converges to some . By Theorem 1.1, we conclude that which contradicts the fact is a gap in the spectrum . ∎
In Section 5 we prove the existence of gaps in the limit operator for several subclasses of twodimensional photonic crystal.
2 Limit Bloch operators and some properties
In this section we shall study some properties of the family of selfadjoint operators , , associated with Theorem 1.1. In particular, we shall establish several important properties concerning the continuity of these operators, their domains and their spectra with respect to the quasiperiodicity parameter .
Suppose and let be defined by (1.23).
Remark 2.1.
is ‘equivalent’ to solving the conjugate CauchyRiemann equations in : for if we define a complex valued quasiperiodic function then solves the CauchyRiemann equations in if, and only if, .
Note that is a closed subspace of and therefore a Hilbert space when equipped with the following equivalent norm
(2.1) 
Remark 2.2.
The equivalence of the norm (2.1) follows from standard Poincaré type inequalities and the following equality, which is a simple consequence of integration by parts and the quasiperiodicity of the functions in question,
Denoting by the closure of in
(2.2) 
It is clear that is closed in due to the norm (2.1). Therefore, generates a unique selfadjoint operator whose domain is a dense subset of and whose action associates to the unique solution of
(2.3) 
As is compactly embedded into we find that the resolvent of is compact and, in particular, the spectrum of is discrete: it consists of countably many isolated eigenvalues of finite multiplicity which converge to infinity. We order the eigenvalues, allowing for multiplicity, as follows:
An important distinct feature of the operators (as opposed to the usual FloquetBloch operators concerned with periodic elliptic PDEs in the whole space) is the nontrivial dependence on of the operator form domain . The statement of continuity of is therefore not a simple consequence of the Blochwave representation of functions belonging to , and relies on establishing that the underlying space is continuous with respect to , see Lemma 2.1. We shall show that the continuity of leads to the operators being continuous with respect to in the normresolvent sense, which in turn implies the continuity of , see Lemma 2.4.
In the rest of this section we shall introduce and prove the aforementioned results.
Lemma 2.1 (Continuity of ).
Let be such that for some . Then, for each there exists such that strongly in as
To prove this lemma we shall use the following characterisation of the spaces .
Lemma 2.2.
(i). Suppose , then if, and only if,
for some with
(2.4) 
where with and contained in . Here, for a given scalar function , we denote by the vector .
(ii). Suppose . Then if, and only if,
for some constant and with
(2.5) 
where , ,
Proof.
In both cases (i) and (ii) the necessity condition is easy to demonstrate. Let us now show the sufficiency condition. First, we shall consider case (i). Suppose, , and let us fix and set , . It is clear that and that their supports are contained in . Now, let us introduce the unique solutions of
We claim that . Indeed, this can be seen by noting that belongs to with , which implies .
For case (ii), fix ; then it is clear that for some such that . Now we repeat the above argument, that is let be the unique solutions of
As above, we show that . ∎
Additionally, we shall use in the proof of Lemma 2.1 the following regularity results.
Lemma 2.3.

Suppose , and let be the unique solution of
(2.6) for some . Then, and
(2.7) 
Suppose and let , be the unique solution of
(2.8) for some such that . Then, and there exists a constant , independent of , such that
Proof.
Proof of (i): The eigenfunctions of the quasiperiodic Laplacian form an orthonormal basis of . By decomposing and in terms of this basis, we have
Now (2.6) tells us , where . Since
For we see . Hence,
Proof of (ii): The eigenfunctions of the periodic Laplacian form an orthonormal basis in . Decomposing and in terms of this basis we have
By the assumptions and , we have and respectively. Now (2.8) tells us that for , . Since , one has
i.e.
∎
Proof of Lemma (2.1)..
Consider a sequence such that as . There are two separate cases to consider, the cases and .
Case 1: . Let us assume without loss of generality that . For fixed we know, by Lemma 2.2, that for some where
for some . We shall now construct the desired as follows: Set where solve
Notice, by Lemma 2.2, that . It remains to show strongly in . To this end, it is sufficient to show that and strongly in as . Let us show that as . By defining , one notices that and uniquely solves
(2.9) 
where . In particular, with , and therefore by Lemma 2.3(i) one has
(2.10) 
Furthermore, since , by an application of Lemma 2.3(i) one notices
where is independent of (here we used the fact that as ). Therefore, one has