Electromagnetic density of modes for a finite-size three-dimensional structure Giuseppe D’Aguanno, 1,2,3, * Nadia Mattiucci, 1,2,4 Marco Centini, 2,3 Michael Scalora, 2 and Mark J. Bloemer 2 1 Time Domain Corporation, Cummings Research Park, 7057 Old Madison Pike, Huntsville, Alabama 35806, USA 2 Weapons Sciences Directorate, Research Development and Engineering Center, U.S. Army Aviation & Missile Command, Building 7804, Redstone Arsenal, Alabama 35898-5000, USA 3 INFM at Dipartimento di Energetica, Università di Roma “La Sapienza,” Via A. Scarpa 16, I-00161 Rome, Italy 4 Università “RomaTre,” Dipartimento di Fisica “E. Amaldi,” Via Della Vasca Navale 84, I-00146 Rome, Italy (Received 29 September 2003; published 24 May 2004) The concept of the density of modes has been lacking a precise mathematical definition for a finite-size structure. With the explosive growth in the fabrication of photonic crystals and nanostructures, which are inherently finite in size, a workable definition is imperative. We give a simple and physically intuitive defini- tion of the electromagnetic density of modes based on the Green’s function for a generic three-dimensional open cavity filled with a linear, isotropic, dielectric material. DOI: 10.1103/PhysRevE.69.057601 PACS number(s): 42.70.Qs, 41.20.-q Several attempts have been made to generalize the notion of the local density of modes (LDOM)—or local density of states—and the density of modes (DOM)—or density of states—to the case of open cavities—i.e., structure of finite size where electromagnetic energy can flow in and out of the volume bounded by the surface S of the cavity [1–6]. Quite surprisingly, the concept of the density of states for a finite- size structure still lacks a simple, concise definition. In the words of Felbacq and Smaali, “Such notions as that of the density of states or local density of states, which are crucial in the description of the coupling between field and matter, cannot be straightforwardly defined for finite structures” [5]. In our estimation, the most likely reason why the concepts of the LDOM and DOM have not yet found straightforward extensions to the case of three-dimensional (3D), finite struc- tures is probably due to the fact that most approaches have focused on the mathematical rather than the basic physical aspects of the problem. Furthermore, due to their basic sim- plicity, 1D and 2D open cavities filled with nonabsorbing materials remain the subject of choice of most researchers. If we have a closed cavity such that the field vanishes at the edges or a cavity where periodic boundary conditions can be applied (such as a multilayer stack of infinite length), filled with a nonabsorbing medium, we can expect that the electromagnetic energy will be conserved inside the cavity, and the problem in Hermitian. In this case, the LDOM stands for the number of eigenmodes per unit volume and unit fre- quency at a point r inside the cavity. If the electromagnetic field can be specified by a single field component (TE or TM polarization), then the scalar Green’s function can be ex- panded in terms of the eigenmodes of the cavity and the LDOM can be calculated through the imaginary part of the scalar Green’s function: r -ImG r , r  [4,7,8]. The DOM is then defined as the average LDOM inside the vol- ume V of the cavity. So we ask (i) what happens if the cavity is open, such that electromagnetic energy can flow in and out of the volume bounded by the surface S of the cavity? (ii) What happens if the cavity is filled with an absorbing mate- rial? These questions have no easy answers fundamentally because the electromagnetic problem is no longer Hermitian and the cavity does not admit eigenmodes in the usual sense of the word. As a consequence, the Green’s function does not admit a straightforward expansion in terms the cavity modes [7,9] and the very notion of the LDOM would seem to lose its validity. In other words, can a LDOM still be defined when the problem is not Hermitian (the case of open cavity or material absorption), and what is its physical meaning in this case [10]? To answer these deceptively simple but crucial questions we go back to the usual starting place—that is, Maxwell’s equations, which we write in MKSA units, in the frequency domain, assuming a harmonic time dependence of the type e -it and nonmagnetic materials r 1: E = iB , 1a B = 0 J r - i c 2 r E , 1b where J r is a complex current density, r is the spa- tially dependent, relative, complex dielectric function of the material, and r = R r + i I r . Note that we use a ge- neric, linear, and isotropic dielectric material. Equations (1) describe the steady-state case; i.e., we assume that both the electromagnetic field and the source J r oscillate with a harmonic time dependence. Taking the curl of Eq. (1a) and using Eq. (1b), we arrive at the following equation for the electric field: - E + 2 c 2 r E =- i 0 J . 2 Let us now consider a cavity of volume V and surface S filled with a dielectric material of dielectric function r , where a known current density J r is present. Multiplying Eq. (2) by E * , integrating over the volume V, using the vector ana- log of Green’s first identity [11], and using Eq. (1a), we arrive at the equation *Electronic address: giuseppe.daguanno@timedomain.com PHYSICAL REVIEW E 69, 057601 (2004) 1539-3755/2004/69(5)/057601(4)/$22.50 ©2004 The American Physical Society 69 057601-1