Model of point source for layered metamaterials

0 downloads 0 Views 11MB Size Report
Before introducing the Green's function, we apply the Laplace transform to the Maxwell's equations (1)-‐. (4) [8]:. ˆ. ˆ. ( , ). ( ,0). ( , ) iz ..... From (19) we can express ..... G yyz κ. + means left-‐sided and right-‐sided limits with x y. → . Then, in the.
ÓPTICA PURA Y APLICADA. www.sedoptica.es

Type:  Research  Paper  /  Tipo:  Artículo  de  Investigación   Section:  Electromagnetic  theory  /  Sección:  Teoría  Electromagnética  

Model  of  point  source  for  layered  metamaterials    

K.  V.  Pravdin,  and  I.  Yu.  Popov*     Department  of  Higher  Mathematics,  ITMO  University,  Kronverkskiy,  49,   Saint  Petersburg,  197101,  Russia    (*)  

Email:  [email protected]

Received  /  Recibido:  27/11/2014  

 Accepted  /  Aceptado:  15/02/2015  

DOI:  http://dx.doi.org/10.7149/OPA.48.1.31  

ABSTRACT:   The  multilayer  system  including  negative  index  material  (NIM)  layers  is  examined.  We  deal  with  the   NIM   system   composed   of   arbitrary   finite   number   of   parallel   alternated   layers   filled   with   isotropic   homogeneous   NIM   and   vacuum.   The   Maxwell’s   equations   for   the   point   source   are   considered.   The   NIM   layer   has   the   electric   permittivity   and   magnetic   permeability,   which   are   equal   to   –1   for   the   certain  frequency  (NIM  frequency).  We  set  the  goal  of  obtaining  expressions  for  the  electric  Green’s   function.   The   Laplace   and   Fourier   transforms   are   used.   The   differential   equations   for   the   scalar   s-­‐   and  p-­‐polarization  parts  of  the  electric  Green’s  function  are  obtained.  The  solutions  of  the  differential   equations   are   obtained   in   the   travelling   wave   form   with   unknown   coefficients.   With   the   standard   boundary   conditions   for   every   layer,   the   recurrence   relations   for   the   coefficients   are   obtained.   The   solution   is   obtained   by   the   generating   function   method.   The   expressions   for   the   scalar   s-­‐   and   p-­‐ polarization  and  vector  part  of  the  electric  Green’s  function  are  derived.    Under  some  assumptions,   we   observe   the   reflection   absence   (for   the   main   term   of   the   solution   asymptotics   near   the   NIM   frequency).   The   obtained   results   can   be   used   in   simulation   or   engineering   of   real   objects,   such   as   superlens  systems  and  multilayer  NIM  coverings.   Key  words:  metamaterials,  negative  index  material,  Maxwell’s  equations,  Green’s  function     REFERENCES  AND  LINKS  /  REFERENCIAS  Y  ENLACES   [1]     A.  E.  Dubinov,  L.  A.  Mytareva,  “Invisible  cloaking  of  material  bodies  using  the  wave  flow   method,”  Phys  Usp  53,  455–479  (2010).   http://dx.doi.org/10.3367/UFNe.0180.201005b.0475  

[2]     N.N.  Rozanov,  Priroda  6  (3),  (2008)  (in  Russian).   [3]     E.  Ozbay,  Z.  Li,  K.  Aydin,  “Super-­‐resolution  imaging  by  one-­‐dimensional,  microwave  left-­‐handed   metamaterials  with  an  effective  negative  index,”  J  Phys  Condens  Matt  20,  304216  (2008).   http://dx.doi.org/10.1088/0953-­‐8984/20/30/304216  

[4]     A.  Iyer,  G.  Eleftheriades,  “Free-­‐space  imaging  beyond  the  diffraction  limit  using  a  Veselago-­‐Pendry   transmission-­‐line  metamaterial  superlens,”  IEEE  Trans  Antennas  Propag  57,  1720-­‐1727  (2009).   http://dx.doi.org/10.1109/TAP.2009.2019890  

[5]     B.  Casse,  W.  Lu,  Y.  Huang,  E.  Gultepe,  L.  Menon,  S.  Sridhar,  “Super-­‐resolution  imaging  using  a  three-­‐ dimensional  metamaterials  nanolens,”  Appl  Phys  Lett  96,  023114  (2010).   http://dx.doi.org/10.1063/1.3291677  

[6]     M.  Lequime,  B.  Gralak,  S.  Guenneau,  M.  Zerrad,  C.  Amra,  “Optical  properties  of  multilayer  optics   including  negative  index  materials,”  arXiv:1312.6288  (2013)   [7]     S.  Burgos,  R.  de  Waelwe,  A.  Polman,  H.  Atwater,  “A  single-­‐layer  wide-­‐angle  negative-­‐index   metamaterial  at  visible  frequencies,”  Nat  Mater  9,  407-­‐412  (2010).   http://dx.doi.org/10.1038/nmat2747  

Opt. Pura Apl. 48 (1) 31-53 (2015)

31

© Sociedad Española de Óptica

ÓPTICA PURA Y APLICADA. www.sedoptica.es

[8]     B.  Gralak,  A.  Tip,  “Macroscopic  Maxwell’s  equations  and  negative  index  materials,”  J  Math  Phys  51,   052902  (2010).     http://dx.doi.org/10.1063/1.3374670  

[9]     B.  Gralak,  D.  Maystre,  “Negative  index  materials  and  time-­‐harmonic  electromagnetic  field,”  C  R   Physique  13,  786-­‐799  (2012).   http://dx.doi.org/10.1016/j.crhy.2012.04.003  

[10]   R.  Collin,  “Frequency  dispersion  limits  resolution  in  Veselago  lens,”  PIER  B  19,  233-­‐261  (2010).     http://dx.doi.org/10.2528/PIERB09120904  

[11]    K.  V.  Pravdin,  I.  Y.  Popov,  “Photonic  Crystal  with  negative  index  materials  layers,”  Nanosystems:   Phys.,  Chem.,  Math.  5,  626-­‐643  (2014).   [12]   Y.  Liu,  S.  Guenneau,  B.  Gralak,  “A  route  to  all  frequency  homogenization  of  periodic  structures,”   arXiv:1210.6171  (2012).   [13]   M.  Lequime,  B.  Gralak,  S.  Guenneau,  M.  Zerrad,  C.    Amra,  “Negative  Index  Materials:  The  Key  to   «White»  Multilayer  Fabry-­‐Perot,”  arXiv:1312.6281  (2013).   [14]   K.  Lai,  L.  Tsang,  C.  Huang,  “Spatial  domain  Green's  functions  for  planar  multilayered  structures,”   Micro  Opt  Tech  Lett  44,  86-­‐91  (2005)   http://dx.doi.org/10.1002/mop.20555  

[15]   M.  Maksimovic,  M.  Hammer,  Z.  Jaksic,  “Thermal  radiation  antennas  made  of  multilayer  structures   containing  negative  index  metamaterials”,  Proc.  SPIE  6896,  689605  (2008)     http://dx.doi.org/10.1117/12.762616  

[16]   A.  Tip,  “Linear  dispersive  dielectrics  as  limits  of  Drude-­‐Lorentz  systems”,  Phys  Rev  E  69,  016610   (2004)     http://dx.doi.org/10.1103/PhysRevE.69.016610    

1.  Introduction   Metamaterials   (negative   index   materials,   NIM)   are   artificial   materials,   the   main   feature   of   which   is   the   negative   refractive   index.   By   using   NIMs,   new   covered   surfaces,   cloaking   and   invisible   materials   can   be   created   [1,  2],   as   well   as   superlenses   with   the   resolving   power,   many   times   exceeding   the   diffraction   limit   [3-­‐5].   In   the   general   case,   NIMs   are   characterized   by   an   occurrence   of   the   ω   frequencies   wherewith   the   electric   ε(ω)   permittivity   and   μ(ω)   magnetic   permeability   (and,   consequently,   the   refractive   index)   possess   negative   values   [6,  7].   In   the   particular   case   [8]   called   NIM   situation,   with   the   ωˆ   frequency   called   NIM  frequency,  these  ε(ω)  and  μ(ω)  are  equal  to  –1  (which  is  opposed  to  the  vacuum  value  of  +1).  Systems   with  NIM  elements  are  called  NIM  systems.   Among   NIM   systems,   layered   NIM   systems   are   widely   known.   The   simplest   model   of   the   layered   NIM   system  (which  is  a  two-­‐layer  system)  was  considered  in  [8,  9].  The  three-­‐layer  NIM  system,  which  is  the   model  for  the  superlens,  was  studied  in  [3,  4,  5,  10,  11].  The  investigations  of  the  multilayer  NIM  systems   (i.e.,  systems  with  the  count  of  layers  greater  than  three)  are  presented  in  [12-­‐15].   A   number   of   NIM   studies   are   dedicated   to   obtaining   the   Green’s   function   which   describes   an   electromagnetic  field  of  a  point  source.  An  electric  field  value  can  be  determined  by  using  of  the  Green’s   function.  In  [8]  a  two-­‐layer  NIM  system  composed  of  homogeneous  isotropic  half-­‐spaces  filled  with  NIM   and  vacuum,  was  in  detail  studied.  The  expressions  for  the  scalar  s-­‐  and  p-­‐polarization  parts  of  the  electric   Green’s   function   were   obtained   in   the   NIM   situation.   In   [11]   the   three-­‐layer   isotropic   NIM   system   was   examined.     The  goal  of  our  work  is  to  obtain  expressions  for  the  electric  Green’s  function  in  the  NIM  situation  for  a   multilayer  NIM  system,  composed  of  arbitrary  finite  number  of  parallel  layers.  The  layers  are  filled  with   NIM  and  vacuum,  and  disposed  alternately.  Taking  into  account  boundary  conditions  at  layer’s  surfaces,   we   use   the   recurrence   relation   method.   The   method   is   easy   to   employ   because   of   the   solution   interdependence  in  adjoining  layers,  and  appropriate  to  analyze  and  compare  the  solutions  obtained.  

Opt. Pura Apl. 48 (1) 31-53 (2015)

32

© Sociedad Española de Óptica

ÓPTICA PURA Y APLICADA. www.sedoptica.es

 

2.  Green’s  function  approach   2.a.  Maxwell’s  equations   Maxwell’s   equations   describe   the   electromagnetic   field,   and,   for   the   case   that   permanent   polarization   and   magnetization   are   absent,   are   presented   in   a   differential   form   as   follows   (we   set   ε 0 = µ0 = 1   for   brevity   [8]):  

 

dD (x,t) = ∇ × H(x,t) ,   dt

(1)  

 

dB (x, t ) = −∇ × E(x, t ) ,   dt

(2)  

 

∇⋅ D(x, t ) = 0 ,  

(3)  

 

∇⋅ B(x, t ) = 0 ,  

(4)  

where   the   x   vector   is   located   in   the   {e1 , e2 , e3 } basis   of   the   Cartesian   coordinates,   ∇   is   the   Hamilton     operator,   ×   is   a   cross   product   symbol,   ⋅   is   an   inner   product   symbol   as   well   as   a   symbol   for   the   matrix   product.   A   medium   reaction   to   the   electromagnetic   field   is   described   with   the   following   auxiliary   field   equations:   t

 

D(x, t ) = E(x, t ) + P(x, t ) ,   P(x, t ) = ∫ χ e (x, t − s) ⋅ E(x, s) ds ,  

 

B(x, t ) = H(x, t ) + M(x, t ) ,   M(x, t ) = ∫ χ m (x, t − s) ⋅ H(x, s) ds ,  

t0

t

t0

(5)   (6)  

where   χ e (x, t ) ,   χ m (x, t )   are   the   electric   and   magnetic   susceptibility   tensors.   According   to   the   causality   condition,  one  has    

χ e (x, t ) = χ m (x, t ) = 0  for  t  <  t0.  

(7)  

Here  t0  is  the  initial  time.  It  means  that  the  polarization  and  magnetization  vanish  for  times  smaller  than   some  finite  t0,  which  can  have  any  value  [8].  In  relation  to  the  Laplace  transform    

∞ 1 e− izt fˆ (z) dz ,   fˆ ( z ) = ∫ eizt f (t )dt ,   f (t) = 0 2π ∫Γ

(8)  

where   z = ω + iα ,   α → 0 ,   α > 0 ,  and  Г  is  a  straight  line  located  parallel  at  an   α  range  to  the  real  axis,  it  is   proper  to  consider  that  t  ≥  0,  therefore  we  assume  t0  =  0.  We  also  assume  that  the  NIM  system  is  passive   [8].  Then,  the  electromagnetic  energy    

U em (t ) =

1 ⎡E2 (x, t ) + H2 (x, t ) ⎤⎦ dx   2∫⎣

(9)  

is   a   non-­‐increasing   function   of   time.   With   the   causality   and   passivity   conditions   and   AFF   (auxiliary   field   formalism),   the   NIM   system   has   a   proper   time   evolution   [8].   In   case   the   initial   fields   are   square   integrable   they  remain  so  for  all  later  times.   Before  introducing  the  Green’s  function,  we  apply  the  Laplace  transform  to  the  Maxwell’s  equations  (1)-­‐ (4)  [8]:    

ˆ (x, z) − D(x,0) = ∇× H ˆ (x, z) ,   −izD

(10)  

 

ˆ (x, z) − B(x,0) = −∇× E ˆ (x, z) ,   −izB

(11)  

 

ˆ (x, z) = ε(x, z) ⋅ E ˆ (x, z) ,   D

(12)  

 

ˆ (x, z) = µ(x, z) ⋅ H ˆ (x, z) ,   B

(13)  

Opt. Pura Apl. 48 (1) 31-53 (2015)

33

© Sociedad Española de Óptica

ÓPTICA PURA Y APLICADA. www.sedoptica.es

where    

ε(x, z) = 1 + χˆ e (x, z) ,  

(14)  

 

µ(x, z) = 1 + χˆ m (x, z) .  

(15)  

According  to  the  causality  condition  (7),  we  obtain  the  following  relations  from  the  auxiliary  equations  (5)   and  (6):    

D(x,0) = E(x,0) ,  

(16)  

 

B(x,0) = H(x,0) .  

(17)  

 

ˆ (x, z) − E(x,0) = ∇× H ˆ (x, z) ,   −izε(x, z) ⋅ E

(18)  

 

ˆ (x, z) − H(x,0) = −∇× E ˆ (x, z) .   −izµ(x, z) ⋅ H

(19)  

Then,  equations  (10)  and  (11)  acquire  a  form:  

ˆ (x, z )  value  from  equation  (19)  and  substituting  it  into  equation  (18),  we  obtain:   Expressing  the   H ˆ (x, z) = ge (x, z) ,   Le (x, z) ⋅ E

(20)  

 

Le (x, z ) ⋅ Eˆ (x, z ) = z 2ε(x, z )Eˆ (x, z ) − ∇ × ⎡⎣µ −1 (x, z ) ⋅∇ × Eˆ (x, z ) ⎤⎦ ,  

(21)  

 

g e (x, z ) = izE(x, 0) − ∇ × ⎡⎣µ −1 (x, z ) ⋅ H(x, 0) ⎤⎦ ,  

(22)  

  where  

Le ( z )   is   the   electric   Helmholtz   operator,   g e (x, z )   is   a   function   of   the   initial   electric   field   configuration.   Let   us  introduce  the   G e (x, y , z ) electric  Green’s  function  [8]  that  satisfies      

Le (x, z ) ⋅ G e (x, y, z ) = δ (x − y )U ,  

(23)  

where   δ (x − y)   is   the   Dirac   delta   function,   U   is   the   3×3   unit   matrix.   Then,   the   E(x, t ) electric   field     function    is  given  by  the  inverse  Laplace  transform  (8)  of    

Eˆ (x, z ) = ∫ G e (x, y , z ) ⋅ g e (y , z ) dy .  

(24)  

Note,   that   the   magnetic   Green’s   function   G m (x, y, z ) and   the   magnetic   field   function   H(x, t ) can   be       obtained  in  a  similar  way  [8].  Therefore,  below  we  consider  only  electric  Green’s  function.  We  drop  the  e   superscript  for  brevity,  and  use  sometimes  the  "Green’s  function"  notion  without  the  word  "electric".     2.b.  Electric  Green’s  function  for  the  layered  system   We   study   the   NIM   system   composed   of   (n+m+1)   parallel   layers,   where   n,  m  ≥  3   are   natural   odd   integers   (Fig.   1).   e1 ,   e 2   unit   vectors   set   a   plane   of   layer’s   surfaces.   An   x   axis   is   located   parallel   to   the   normal   of   layer’s  surfaces,  and  collinear  to  the   e3  unit  vector.  Let  us  denote  a  coordinate  of  the  left  surface  in  a  k-­‐th   layer  as  xk,  a  coordinate  of  the  left  surface  in  the  zero  layer  as  x0  =  0.  From  the  right  side  of  the  zero  layer,  n   layers   are   located,   from   the   left   side   there   are   m   layers.   Thus   we   have   k = −m,…0,…n .   Layers   alternate   with  each  other.  All  even  layers  (as  well  as  the  zero  layer)  are  Δ1  in  width,  and  filled  with  a  NIM.  All  odd   layers  are  Δ2  in  width,  and  filled  with  a  vacuum.  The  last-­‐to-­‐left-­‐side  (with  –m  index)  and  last-­‐to-­‐right-­‐side   (with  n  index)  layers  are  the  half  spaces  unbounded  along  the  direction  of  the  x  axis  ( x− m = −∞ ,   xn +1 = +∞ )   and  are  empty  (vacuum).  The  point  source  is  located  at   y  coordinate  in  the  zero  layer,  i.e.   x0 < y < x1  or   0 < y < Δ1 .  We  assume  a  translation  invariance  along  the  plane  of  layer’s  surfaces.    

Opt. Pura Apl. 48 (1) 31-53 (2015)

34

© Sociedad Española de Óptica

ÓPTICA PURA Y APLICADA. www.sedoptica.es

   

Fig.  1.  The  NIM  system  composed  of  (n+m+1)  parallel  layers  filled  with  NIM  and  vacuum.  The  point  source  is  located  at  y  coordinate   in  zero  NIM  layer.  

We   consider   NIM   in   layers   isotropic   homogeneous.   Therefore,   the   electric   permittivity   and   magnetic   permeability  do  not  depend  on  coordinates  of  the   x  vector,  i.e.   ε(x, z) = ε ( z)U ,   µ(x, z) = µ ( z)U .  Also,  we   assume  that  NIM  is  a  dispersive,  non-­‐absorptive  medium.  In  that  case,  the  susceptibilities  consist  of  a  sum   of   Lorentz   contributions   [16].   We   deal   with   a   single   dispersive   Lorentz   contribution   [8].   The   electric   permittivity  and  magnetic  permeability  are  

Ω2 ,   z − ω02

ε ( z) = µ ( z) = 1 −

(25)  

2

 

where   Ω ,   ω0 are   constants,   z = ω + iα ,   α → 0 ,   α > 0 ,   and   ε (±ωˆ ) = µ (±ωˆ ) = −1  for  the   ωˆ = ω02 + Ω2 2     NIM  frequency  in  layers  filled  with  NIM,  and   ε ( z) = µ ( z) = 1 in  layers  filled  with  a  vacuum.     Let   κ = κ eκ   be   a   two-­‐dimensional   wave   vector,   κ   be   a   coordinate   of   the   κ   vector,   eκ   be   a   unit   vector   parallel   to   the   plane   of   layer’s   surfaces.   Therefore,   e3 × eκ   is   a   unit   vector   parallel   to   the   plane   of   layer’s   surfaces  and  a  set  of   eκ ,   e3 × eκ ,   e3  unit  vectors  forms  the  Cartesian  basis.  Thus,  the   e′ = (eκ basis  is  given  from  the   e = (e1

e2 e3 )  basis  with  a  rotation  transform  by   γ κ  angle  about  the  x  axis:  

e′ = e ⋅ Tκ ,  

 

 

e3 × eκ e3 )  

⎛ cos γ κ ⎜ Tκ = ⎜ sin γ κ ⎜ 0 ⎝

− sin γ κ cos γ κ 0

(26)  

0⎞ ⎛ cos γ κ ⎟ ,   −1 ⎜ 0 ⎟ Tκ = ⎜ − sin γ κ ⎜ 0 1 ⎟⎠ ⎝

sin γ κ cos γ κ 0

0⎞ ⎟ 0 ⎟ .   1 ⎟⎠

(27)  

Let   ⊥  be  a  superscript  for  the  vector,  which  is  normal  to  the   e3  unit  vector,  i.e.,   x ⊥  is  the  projection  of  the   x  vector  to  the  plane  of  layer’s  surfaces.   Let   us   consider   the   Fourier   transform   for   the   G(x, y, z) Green’s   function   with   the   coordinates     corresponding  with  the   e1  and   e 2 unit  vectors  [8]:      

G(x, y, z) = (2π )−2 ∫ e−iκ⋅( x

 

Gκ ( x, y, z) = ∫ eiκ⋅( x





−y⊥ )

− y⊥ )

Gκ ( x, y, z)dκ ,  

G(x, y, z)dx⊥ ,  

(28)   (29)  

 

m(x) = (2π )−2 ∫ e−iκ ⋅ x mκ ( x)dκ ,  

(30)  

 

mκ ( x) = ∫ eiκ ⋅ x m(x)dx⊥ ,  

(31)  





   

Opt. Pura Apl. 48 (1) 31-53 (2015)

35

© Sociedad Española de Óptica

ÓPTICA PURA Y APLICADA. www.sedoptica.es

ˆ (x, z) ,   E(x, t ) ,   H ˆ (x, z ) ,   etc.   Then,   the   electric   field   function   is   expressed   in   terms   of   the   where   m   can   be   E Green’s  function  as  follows:    

ˆ z) dz = E(x,t) = (2π )−1 ∫ e− izt E(x,

 

= (2π )−1 ∫ e− izt dz ∫ G(x,y, z) ⋅g(y, z) d y =

Γ

Γ

= (2π )−3 ∫ e− izt dz ∫ e− iκ ⋅(x

 



−y ⊥ )

Γ

dκ ∫ Gκ (x, y, z) ⋅gκ ( y, z) dy ,  

(32)  

∂ 1   e3 ) × Hκ ( y,0) .   dy µ ( z)

(33)  

where  (22)  is  presented  as  follows:  

gκ ( y, z ) = izEκ ( y,0) − (iκ +

 

 

Let  

ζ 2 ( z, κ ) = z 2ε ( z ) µ ( z ) − κ 2 ,  

 

(34)  

then,  the  Helmholtz  equation  (20)  is  presented  as  follows:  

ˆ ( x, z) = g ( y, z) ,   Lκ ( z) ⋅ E κ κ

 

(35)  

and  the  Helmholtz  operator  (21)  in  a  certain  layer  has  the  form:  

Lκ ( z ) =  

⎞ ⎞ 1 ⎛ ∂2 1 ⎛ ∂2 ζ 2 ( z, κ ) 2 2 e3 e3 + ⎜ 2 + ζ ( z , κ ) ⎟ e 3 × e κ e 3 × eκ + ⎜ 2 + z ε ( z ) µ ( z ) ⎟ eκ eκ + µ ( z ) ⎝ ∂x µ ( z ) ⎝ ∂x µ ( z) ⎠ ⎠

−  

iκ ∂ (eκ e3 + e3eκ ) ,   µ ( z ) ∂x

 

(36)  

where   ε ( z ) ,   µ ( z ) ,   ζ 2 (κ , z )   correspond   to   the   certain   layer.   We   introduce   a   notation   for   a   componentwise   matrix  representation.  E.g.,  for  a  certain   f ( x)  function  and   e3eκ  term  

 

⎡ ⎢ f ( x ) e3 e κ = f ( x ) ⋅ ⎢( e κ ⎢⎣ = ( eκ

 

⎛ 0 ⎞⎤ ⎡ ⎛ eκ ⎞ ⎤ ⎜ ⎟⎥ ⎢ ⎜ ⎟⎥ e3 ) ⋅ ⎜ 0 ⎟ ⎥ ⋅ ⎢(1 0 0 ) ⋅ ⎜ e3 × eκ ⎟ ⎥ =   ⎜ 1 ⎟⎥ ⎢ ⎜ e ⎟⎥ ⎝ ⎠⎦ ⎣ ⎝ 3 ⎠⎦

e3 × e κ

e3 × eκ

0 0 ⎞ ⎛ eκ ⎞ ⎛ 0 ⎜ ⎟ ⎜ ⎟ e3 ) ⋅ ⎜ 0 0 0 ⎟ ⋅ ⎜ e3 × eκ ⎟ .   ⎜ f ( x) 0 0 ⎟ ⎜ e ⎟ ⎝ ⎠ ⎝ 3 ⎠

(37)  

Note,  that  the   e3eκ  term  in  (37)  does  not  mean  the  inner  product  of  the   e3  and   eκ  unit  vectors.  In  terms  of   bra-­‐ket  notation,  an   e3eκ  term  means   e3 eκ .  Then,  (23)  is  presented  as  follows:    

Lκ ( z) ⋅ Gκ ( x, y, z) = δ ( x − y)U .  

(38)  

Formulae   for   (33)   and   (36)   are   obtained   in   the   {eκ , e3 × eκ , e3 }   basis.   Therefore,   we   need   to   obtain   the  

Gκ ( x, y, z)  Green’s  function  from  (38)  in  the  basis,  too.  The  Green’s  function  in  the   {e1 , e2 , e3 }  basis  can  be  

obtained   by   the   rotation   transform   (27).   The   incident   plane   is   formed   by   eκ  and   e3   unit   vectors.   Thus,   all   matrices  can  be  expressed  in  terms  of  the  s-­‐  and  p-­‐polarization  parts.  According  to  [8]      

Lκ = Lsκ + Lpκ ,  

(39)  

i.e.   Lκ  is  resolved  into   Lsκ ,  that  is  the  s-­‐polarization  part  (the  term  with   e3 × eκ e3 × eκ  in  (36)),  and   Lpκ ,  that   is   the   p-­‐polarization   part   (the   rest   of   terms   in   (36)).   Then,   the   Gκ ( x, y, z) Green’s   function   in   the     decomposition  (32)  is  presented  as  follows  [8]:  

Opt. Pura Apl. 48 (1) 31-53 (2015)

36

© Sociedad Española de Óptica

ÓPTICA PURA Y APLICADA. www.sedoptica.es

 

G κ ( x, y, z ) = G s ( x, y, z, κ ) + G p ( x, y, z, κ ) ,  

(40)  

 

G s ( x, y, z, κ ) = G s ( x, y, z, κ )e3 × eκ e3 × eκ ,  

(41)

(42)  

 

⎛ iκ ∂ ⎞⎛ iκ ∂ ⎞ p G p ( x, y , z , κ ) = ⎜ e κ + 2 e3 ⎟⎜ eκ − 2 e3 ⎟ G ( x, y, z, κ ) ,   ζ ( x, z, κ ) ∂x ⎠⎝ ζ ( y, z, κ ) ∂y ⎠ ⎝

 

where   ζ 2 ( x, κ , z )  is  presented  in  a  layer  that  corresponds  with  the  x  coordinate,   ζ 2 ( y, κ , z )  is  presented   in  the  zero  layer,  and  scalar   G s  and   G p  functions  satisfy  the  following  differential  equations:   (43)  

 

⎞ 1 ⎛ ∂2 + ζ 2 ( z, κ ) ⎟ G s ( x, y, z, κ ) = δ ( x − y) ,   ⎜ µ ( z ) ⎝ ∂x 2 ⎠

(44)  

 

⎞ p z 2ε ( z ) ⎛ ∂ 2 2 ⎜ 2 + ζ ( z, κ ) ⎟ G ( x, y, z, κ ) = δ ( x − y) .   2 ζ ( z, κ ) ⎝ ∂x ⎠

  2.c.  Boundary  conditions   We  study  the  system  composed  of  layers  divided  by  plane  unbounded  surfaces.  The  boundary  conditions   for  each  one  are  presented  in  the  general  form  as  follows:    

(E1 − E2 ) × n = 0 ,  

(45)  

 

( H1 − H2 ) × n = 0 ,  

(46)  

 

( D1 − D2 ) ⋅ n = 0 ,  

(47)  

 

(B1 − B2 ) ⋅ n = 0 .  

(48)  

E j = E j ( x! ,t) ,   H j = H j ( x! ,t) ,   D j = D j ( x! ,t) ,  and   B j = B j ( x! ,t)  stand  for  the  limits  with   x → x! ,  where   x!  is  a   vector   located   parallel   to   the   plane   of   layer’s   surfaces,   j   is   an   index   distinguishing   limits   calculated   on   different  surface  sides  ( j = 1, 2 ),   n  is  an  unit  vector  of  normal  to  the  plane  of  layer’s  surfaces  ( n = e3 ).  The   conditions  (45),  (46),  (47)  after  Laplace  transform  (8)  being  applied  are  presented  as  follows:  

(Eˆ (x! , z) − Eˆ (x! , z)) × e = 0 ,   ( Hˆ (x! , z) − Hˆ (x! , z)) × e = 0 ,  

     

1

2

3

(49)  

1

2

3

(50)  

(ε (z)Eˆ (x! , z) − ε (z)Eˆ (x! , z)) ⋅ e 1

1

2

2

3

= 0 .  

(51)  

It  is  proper  to  represent  the  boundary  conditions  for  the  s-­‐  and  p-­‐  polarization  parts  in  the   {eκ , e3 × eκ , e3 }

  basis   obtained   from   the   {e1 , e2 , e3 }   basis   with   the   rotation   transform   (27).   Note,   that   equations   (49),   (50),   (51)  are   invariant  in  respect   to  the  rotation  transform   (27),   i.e.   an  equality   of   components   located   parallel   to  the  plane  of  layer’s  surfaces  or  collinear  to  the   e3  unit  vector,  does  not  depend  on  the   γ κ  rotation  angle.   Thus,          

Opt. Pura Apl. 48 (1) 31-53 (2015)

(Eˆ (x! , z) − Eˆ (x! , z))



( Hˆ (x! , z) − Hˆ (x! , z))



1

2

1

2

(Eˆ (x! , z) − Eˆ (x! , z)) 1

2

37

e3 ×eκ

= 0 ,  

(52)  

= 0 ,  

(53)

= 0 ,  

(54)  

 

© Sociedad Española de Óptica

ÓPTICA PURA Y APLICADA. www.sedoptica.es

(ε (z)Eˆ (x! , z) − ε (z)Eˆ (x! , z)) 1

 

1

2

2

e3

= 0 ,  

(55)  

where  for  a  certain   A  vector,   A e  notation  means  its  projection  on  the   e  unit  vector.  

ˆ ( x, z) electric  field  function  from   Let  us  consider  conditions  (52),  (55)  for  the  p-­‐polarization  case.  The   E κ   equation   (35)   is   the   result   of   the   Fourier   transform   (30)   with   coordinates   of   the   κ   vector   located   parallel   to  the  plane  of  layer’s  surfaces,  i.e.  with  the  coordinates  corresponding  to  the   e1  and   e 2  unit  vectors.  Thus,  

(Eˆ

 

! z) − Eˆ 2,κ ( x, ! z) ( x,

1,κ

! z) − ε 2 (z)Eˆ 2,κ ( x, ! z) ( x,

(ε (z)Eˆ 1

 

)

1,κ



= 0 ,  

)

(56)

= 0 ,  

e3

 

(57)  

where   x!   is   the   x!   vector’s   coordinate   corresponding   to   the   e3   unit   vector.   Then,   according   to   the   decomposition  (32),  the  boundary  conditions  for  the  p-­‐polarization  part  of  the  Green’s  function  are  given   as  follows:  

(

)

 

⎡ G1p ( x! , y, z,κ ) − G 2p ( x! , y, z,κ ) ⋅gκ ( y, z) ⎤ = 0 ,   ⎣ ⎦ eκ

 

⎡ ε1 (z)G1p ( x, ! y, z,κ ) − ε 2 (z)G 2p ( x, ! y, z,κ ) ⋅gκ ( y, z) ⎤ = 0 .   ⎣ ⎦ e3

(

(58)  

)

(59)  

Equations  (58),  (59)  are  valid  for  any  function  of  initial  field  configuration  (33).  Therefore,  the  first  row  of

  p the   G1 ( x! , y, z,κ )  tensor  function  corresponding  to  the   eκ  unit  vector  is  completely  equal  to  the  first  row   p

p

of   G 2 ( x! , y, z,κ ) ,  and  the  third  row  of   G1 ( x! , y, z,κ )  corresponding  to  the   e3  unit  vector  is  equal  to  the  third   row  of   G 2 ( x! , y, z,κ )  within  the  accuracy  of  the   ε 2 ( z ) / ε1 ( z)  term.  Thus,  we  obtain  the  boundary  conditions   p

for  the  p-­‐polarization  part  of  the  Green’s  function:    

G1p ( x! , y, z,κ ) = G2p ( x! , y, z,κ ) ,

 

∂G1p ε (z) ζ 12 (z,κ ) ∂G2p ( x! , y, z,κ ) = 2 ( x! , y, z,κ ) . ∂x ε1 (z) ζ 2 2 (z,κ ) ∂x

(60)

 

 

(61)    

Now,  let  us  consider  the  s-­‐polarization  case.  It  can  be  shown  that  equation  (19)  is  invariant  to  the  rotation   transform  (27).  From  (19)  we  can  express  

ˆ (x, z ) = H  

(

)

1 ˆ (x, z) − H(x,0) ,   ∇e ′ × E iz µ ( z )

where   ∇e′   is   the   Helmholtz   operator   acting   on   the   e′ = (eκ obtain   x= x!

⎛ ⎡ ⎤⎞ µ1 (z) ˆ ˆ ⎜ ∇ e′ × ⎢E1 (x, z) − µ (z) E2 (x, z) ⎥⎟ ⎝ ⎣ ⎦⎠ e 2

  where   f (x)

κ

x= x!

(62)  

e3 × eκ e3 )   basis.   From   (62)   and   (53)   we  

⎛ ⎞ µ (z) = ⎜ H1 ( x! ,0) − 1 H 2 ( x! ,0)⎟ ,   µ (z) ⎝ ⎠e 2

(63)  

κ

ˆ (x, z ) = f ( x! )   for   a   certain   f (x)   function.   For   the   s-­‐polarization   case   E

e3

= 0 .   Also,  

H1 (x,0) = H2 (x,0) = 0  for  any   x  vector  ( x ≠ y )  because  in  the  initial  time  t0  (t0  =  0)  the  electromagnetic   field  is  absent  at  any  point  of  the  system  except  the  coordinates  of  the  y  vector  (then,  with  time,  the  field   propagates  in  the  system  from  the  coordinates  of  the  y  vector).  Thus,  

 

Opt. Pura Apl. 48 (1) 31-53 (2015)

⎞ µ (z) ∂ ⎛ˆ E (x, z) − 1 Eˆ 2 (x, z)⎟ ∂x ⎜⎝ 1 µ2 (z) ⎠

38

x= x!

= 0 .  

(64)  

e3 ×eκ

© Sociedad Española de Óptica

ÓPTICA PURA Y APLICADA. www.sedoptica.es

Using  the  Fourier  transform  (30)  for  the  conditions  (53),  (54),  we  obtain  

(Eˆ

1,κ

 

! z) − Eˆ 2,κ ( x, ! z) ( x,

)

e3 ×eκ

= 0 ,  

⎞ µ (z) ∂ ⎛ˆ E1,κ (x, z) − 1 Eˆ 2,κ (x, z)⎟ ⎜ ∂x ⎝ µ2 (z) ⎠

 

(65)  

x= x!

= 0 .  

(66)  

e3 ×eκ

According   to   the   decomposition   (32),   the   boundary   conditions   for   the   s-­‐polarization   part   of   the   Green’s   function  are  given  as  follows:  

(

)

 

⎡ G1s ( x, ! y, z,κ ) − G s2 ( x, ! y, z,κ ) ⋅gκ ( y, z) ⎤ ⎣ ⎦ e3 ×eκ = 0 ,  

(67)  

(68)  

 

⎡⎛ ⎤ ⎞ ∂G s ∂G s2 ! y, z,κ ) − µ1 (z) ! y, z,κ )⎟ ⋅gκ ( y, z) ⎥ ( x, = 0 .   ⎢⎜ µ2 (z) 1 ( x, ∂x ∂x ⎠ ⎢⎣⎝ ⎥⎦ e ×e 3 κ

Equations   (67),   (68)   are   correct   for   any   function   of   the   initial   field   configuration   (33).   Therefore,   the   s second   row   of   the   G1 ( x! , y, z,κ ) tensor   function   corresponding   to   the   e3 × eκ   unit   vector   is   completely     ∂G1s s equal  to  the  second  row  of   G 2 ( x! , y, z,κ ) ,  and  the  second  row  of   ( x! , y, z,κ )  is  equal  to  the  second  row   ∂x ∂G s2 of   ( x! , y, z,κ )  within  the  accuracy  of   µ1 ( z) / µ2 ( z)  term.  Thus,  we  obtain  the  boundary  conditions  for   ∂x the  s-­‐polarization  part  of  the  Green’s  function:  

 

! y, z,κ ) = G2s ( x, ! y, z,κ ) , G1s ( x,

 

∂G1s µ (z) ∂G2s ( x! , y, z,κ ) = 1 ( x! , y, z,κ ) . ∂x µ2 (z) ∂x

(69)

 

 

(70)    

  2.d.  Scalar  electric  Green’s  function  with  boundary  conditions   Let   us   consider   equation   (44).   Note,   that   the   solutions   of   the   equation   in   every   layer   are   well-­‐known.   Therefore,  the  Green’s  function  is  obtained  through  the  fundamental  system  of  solutions  of  equation  (44)   with  the   G1  and   G2  coefficients:    

G p ( x, y, z, κ ) = G1eiζ x + G2 e −iζ x ,  

(71)  

where   ζ = ζ ( z, κ )   is   given   in   the   corresponding   layer.   The   coefficients   G1 = G1 ( y, z, κ )   and   G2 = G2 ( y, z, κ )   are  the  functions  of  y,  z,  and  κ  variables  but  we  write  them  as   G1  and   G2  for  simplicity   (this  also  holds  true  for  another  functions  of  y,  z,  and  κ  variables).  These  coefficients  satisfy  a  system  of   equations   obtained   from   the   standard   boundary   conditions   for   all   layers.   Formally,   the   solutions   of   that   system   are   obtained   in   the   general   form   (by   the   Cramer’s   theorem   or   by   another   way).   However,   it   is   difficult  to  analyze  the  solutions  in  such  a  form.  A  much  more  useful  way  is  the  construction  of  a  system  of   recurrence   relations   for   the   coefficients   of   the   Green’s   function   in   every   layer.   Physically,   it   means   the   consideration   of   the   serial   reflections.   The   method   is   evident   in   usage   and   easy   in   analysis   of   obtained   solutions.  In  our  investigation  we  choose  exactly  that  way.   The   general   solution   (71)   of   equation   (44)   is   composed   of   the   two   eiζ x   and   e − iζ x travelling   waves   with   G1   and   G2 coefficients.   We   get   three   travelling   waves   at   the   stage   of   propagation   of   travelling   wave   from   the     medium  with  conventional  index  1  to  the  medium  with  conventional  index  2  through  the  layer  surface  in   the  positive  direction  of  the  x  axis:    

Opt. Pura Apl. 48 (1) 31-53 (2015)

G1p ( x, y, z, κ ) = G0 eiζ1x + G1e−iζ1x ,  

39

(72)

 

© Sociedad Española de Óptica

ÓPTICA PURA Y APLICADA. www.sedoptica.es

G2p ( x, y, z, κ ) = G2 eiζ 2 x .  

 

(73)  

The   boundary   conditions   (60),   (61)   on   the   layer   surface   in   the   x!   coordinate   lead   to   the   following   relations:   + σ 1,2 E0 = E2 ,     2

 

E1 =

 

σ k±,l =

  where   E0 = G0e

iζ1x!

 is  the  incident  wave,   E1 = G1e

−iζ1x!

(74)  

− σ 1,2 E2 ,   2

(75)  

ε k ζ l ± ε lζ k ,   ε kζ l

(76)  

 is  the  reflective  wave,   E2 = G2e

iζ 2 x!

 is  the  passed  wave,  

ε = ε ( z)   and   ζ = ζ ( z, κ )   are   calculated   in   the   corresponding   medium,   σ = σ ( z, κ ) .   There   are   similar   ± kl

± kl

relations  between  the  incident,  reflective,  and  passed  waves  during  the  reverse  wave  propagation  through   the  surface  from  the  medium  2  to  the  medium  1:  

 

 

E0 =

+ σ 2,1 E2 ,     2

(77)  

E1 =

− σ 2,1 E2 .   2

(78)  

Thus,   the   boundary   conditions   are   considered   twice   on   every   layer’s   surface,   for   the   waves   propagated   in   the   positive   direction   of   the   x   axis,   and   the   negative.   The   scalar  p-­‐polarized   part   of   the   Green’s   function   in   the  k-­‐th  layer  (excluding  the  zero  layer  with  the  point  source)  is  given  as  follows:    

Gkp ( x, y, z, κ ) = ( Ak + Bk ) eiζ k x + (Ck + Dk ) e−iζ k x ,  

(79)  

where   Ak   and   Dk   are   the   coefficients   of   the   waves   passed   into   the   k-­‐th   layer   through   the   left   and   right   surface,   correspondingly,   Bk   and   Ck   are   the   coefficients   of   reflective   waves   inside   the   k-­‐th   layer   from   the   left  and  right  surface  correspondingly.   There  is  a  point  source  in  the  y  coordinate  in  the  zero  layer.  The  waves  progressing  to  the  left  side   I − e−iζ 0 x

  and   to   the   right   side   I + eiζ 0 x   from   the   point   source,   are   the   solutions   of   equation   (44),   too.   So,   to   the   left   side  of  the  zero  layer      

p G0,− (x, y, z,κ ) = A0 + B0 e

(

)

iζ 0 x

(

)

+ C0 + D0 + I − e

−iζ 0 x

,  

(80)  

and  to  the  right  side  of  the  zero  layer    

G0,p + ( x, y, z, κ ) = ( A0 + B0 + I + ) eiζ 0 x + (C0 + D0 ) e −iζ 0 x .  

(81)  

The   I −   and   I +   coefficients   are   given   by   the   continuity   condition   of   the   Green’s   function   and   the   condition   of  a  discontinuous  jump  of  the  Green’s  function’s  derivative  in  the  y  coordinate:  

G0,p− ( y, y, z, κ ) = G0,p+ ( y, y, z, κ ) ,  

 

∂G0,p +  

∂x

( y, y, z , κ ) −

∂G0,p − ∂x

( y, y, z , κ ) =

(82)  

z 2ε 0 ( z ) ,   ζ 02 ( z, κ )

(83)  

where   G0,p− ( y, y, z, κ )  and   G0,p+ ( y, y, z, κ )  means  left-­‐sided  and  right-­‐sided  limits  with   x → y .  Then,  in  the   zero  layer    

Opt. Pura Apl. 48 (1) 31-53 (2015)

G0p ( x, y, z, κ ) = ( A0 + B0 ) eiζ 0 x + (C0 + D0 ) e −iζ 0 x + I 0e −iζ 0 | x − y| ,  

40

(84)  

© Sociedad Española de Óptica

ÓPTICA PURA Y APLICADA. www.sedoptica.es

 

I0 =

ζ 0 ( z, κ ) .   2iz 2ε 0 ( z )

(85)  

Physically,  the  waves  progressing  only  away  from  the  point  source  can  be  solutions  of  equation  (44)  in  the   last-­‐to-­‐left  and  last-­‐to-­‐right  side  of  layers,  i.e.  in  the  layers  with  n  and  –m  indices:    

Gnp ( x, y, z, κ ) = An eiζ n x ,  

(86)  

 

G−pm ( x, y, z, κ ) = D−me−iζ − m x .  

(87)  

The  travelling  waves  passing  through  the  surfaces  in   xn  and   x− ( m −1)  coordinates  into  the  layers  adjoining   to  the  last  layers,  are  absent  in  the  solutions  (79),  i.e.  in  the  layers  with  (n–1)  and  –(m–1)  indices      

Gnp−1 ( x, y, z, κ ) = ( An−1 + Bn−1 ) eiζ n−1x + Cn−1e−iζ n−1x ,   G−p( m −1) ( x, y, z, κ ) = B− ( m −1) e

iζ − ( m−1) x

+ (C− ( m −1) + D− ( m −1) ) e

(88)  

− iζ − ( m−1) x

.  

(89)  

As  a  result,  the  scalar  p-­‐polarized  part  of  the  Green’s  function  is  expressed  as  follows:  

G p (x, y, z,κ ) = − iζ x ⎧ D− me − m x ∈(−∞, x−( m−1) ) ⎪ iζ − ( m−1) x − iζ − ( m−1) x − iζ − ( m−1) x ⎪ B−( m−1) e +C−( m−1) e +D−( m−1) e x ∈(x−( m−1) , x−( m−2) ) ⎪ iζ − ( m−2 ) x iζ − ( m−2 ) x − iζ − ( m−2 ) x − iζ − ( m−2 ) x ⎪ +B−( m−2) e +C−( m−2) e +D−( m−2) e x ∈(x−( m−2) , x−( m−3) ) ⎪ A−( m−2) e ⎪… … … … … ⎪ iζ −1x iζ −1x − iζ −1x − iζ −1x +B−1e +C−1e +D−1e x ∈(x−1 , 0) ⎪ A−1e ,   ⎪ iζ x i ζ x − i ζ x − i ζ x − i ζ |x− y| 0 0 0 0 ⎨A e 0 +B e +C e +D e +I e x ∈(0, x ) 0 0 0 0 1 ⎪ 0 iζ 1x iζ 1x − iζ 1x − iζ 1x ⎪ A1e +B1e +C1e +D1e x ∈(x1 , x2 ) ⎪ … … … … ⎪… iζ n−2 x − iζ n−2 x − iζ n−2 x ⎪ A eiζ n−2 x +Bn−2 e +Cn−2 e +Dn−2 e x ∈(xn−2 , xn−1 ) ⎪ n−2 iζ n−1x iζ n−1x − iζ n−1x ⎪ An−1e +Bn−1e +Cn−1e x ∈(xn−1 , xn ) ⎪ iζ x ⎪ An e n x ∈(xn , + ∞) ⎩

(90)  

where   ζ k = ζ k ( z, κ )  (34)  is  obtained  in  the  k-­‐th  layer  with   k = −m,…,n ,  and   I 0  is  given  in  (85).   The  reasoning  similar  to  the  above  mentioned  one  leads  to  an  expression  for  the  scalar  s-­‐polarized  part  of   the   G s ( x, y, z, κ )   Green’s   function,   which   is   the   same   as   in   (90).   But   there   are   differences.   The   first   one   consists  in  relations  (74),  (78)  for  the   E0 incident  wave,  the   E1 reflective  wave,  and  the   E2  passed  wave:      

 

   

E0 =

+ τ 1,2 E2 ,     2

− τ 1,2 E1 = E2 ,   2

τ k±,l =

µl ζ k ± µk ζ l ,   µl ζ k

(91)  

(92)   (93)  

where   µ = µ ( z )  and   ζ = ζ ( z, κ )  are  obtained  in  the  corresponding  medium,   τ kl± = τ kl± ( z, κ ) .  Equation  (25)   implies    

Opt. Pura Apl. 48 (1) 31-53 (2015)

τ kl± = σ lk± .  

41

(94)  

© Sociedad Española de Óptica

ÓPTICA PURA Y APLICADA. www.sedoptica.es

Thus,  for  the  s-­‐polarization  case,  relations  (74),  (75)  and  also  (77),  (78)  are  valid  with  changing  subscripts   for   σ kl± .     The  last  term  in  expression  (85)  for  the  s-­‐polarization  case  has  the  form  

I0 =

 

µ0 ( z) .   2iζ 0 ( z, κ )

(95)  

 

3.  Recurrence  relations   3.a.  Green’s  function  coefficients   Let  us  consider  the  scalar  p-­‐polarized  part  of  the  Green’s  function  in  the  k-­‐th  layer  (79).  It  is  related  with   the  Green’s  functions  in  the  (k–1)-­‐th  and  (k+1)-­‐th  layers  correspondingly  by  the  boundary  conditions  (74)   and   (78)   on   the   left   surface   and   (75)   and   (77)   on   the   right   surface.   By   considering   these   conditions   for   each   layer   of   the   system,   we   obtain   expressions   for   the   coefficients   of   the   scalar   p-­‐polarized   part   of   the   Green’s  function  [11].  Thus,  for   k = 1,…(n − 1)      

Ak = βk An ,   Bk =

dk h h γ k An ,   Ck = k βk +1 An ,   Dk = γ k +1 An ,   Cn −1 = n −1 An ,   Dn −1 = 0 ,   ck gk g n −1

(96)  

for   k = −(m − 1),…0    

 

Ak = βk An + ξk ,   Bk = C0 =

dk h (γ k An + ηk ) ,   Ck = k (β k +1 An + ξk +1 ) ,   Dk = γ k +1 An + ηk +1 ,   gk ck

h0 β1 An ,   D0 = γ 1 An ,   A− ( m −1) = 0 ,   D− m = γ − ( m −1) An + η− ( m −1) .   g0

(97)  

The   β k ,   γ k ,   ξ k ,  and   η k    values  in  (96)  and  (97)  satisfy  the  following  recurrence  relations  [11]:  

 

⎛ ad ⎞ fk ⎛ ⎛ a d e h ⎞ ⎞ ⎟ γ k +1 ,   βn−1 = J n−1 ,   J k = ⎜⎜1 − ⎜ ⎟ ⎟ ,   ek ⎝ ⎝ b c f g ⎠k ⎟⎠ ⎝ b c ⎠k

β k = J k β k +1 − ⎜

γk =  

⎞ ⎛ ah ⎞ ak ⎛ hk ⎜ β k +1 + γ k +1 ⎟ ,   γ n −1 = ⎜ ⎟ ,   bk ⎝ g k ⎝ b g ⎠ n −1 ⎠

(99)  

⎛ −iζ y ⎛ a d ⎞ iζ 0 y ⎞ ⎛ ad ⎞ ⎟ ηk +1 ,   ξ0 = I 0 ⎜⎜ −e 0 − ⎜ ⎟ e ⎟⎟ ,   ⎝ b c ⎠k ⎝ b c ⎠0 ⎝ ⎠

(100)  

⎞ ak ⎛ hk a0 iζ y ⎜ ξ k +1 + ηk +1 ⎟ ,   η0 = I 0 e 0 ,   bk ⎝ g k b0 ⎠

(101)  

ξ k = J k ξ k +1 − ⎜

 

ηk =  

(98)  

where      

ak = 2e−iζ k xk ,   bk = σ k+, k −1e−iζ k −1xk ,   ck = 2eiζ k xk ,   d k = σ k−, k −1e−iζ k −1xk ,  

 

ek = 2eiζ k xk +1 ,   f k = σ k+, k +1eiζ k +1xk +1 ,   gk = 2e−iζ k xk +1 ,   hk = σ k−, k +1eiζ k +1xk +1 ,  

(102)  

⎛ ad ⎞ ak d k and   σ kl±  is  defined  in  (76),   I 0  is  defined  in  (85).  We  introduce  the   ⎜  notation.  The  relations   ⎟ = ⎝ b c ⎠k bk ck for   βn −1   in   (98),   γ n −1   in   (99),   ξ 0   in   (100),   and   η0   in   (101)   are   called   the   initial   conditions   for   the   recurrence  relations  (98)-­‐(101).   All  the  coefficients  (96)  and  (97)  are  functions  of   An ,  which  satisfies  the  following  relation  [11]:  

Opt. Pura Apl. 48 (1) 31-53 (2015)

42

© Sociedad Española de Óptica

ÓPTICA PURA Y APLICADA. www.sedoptica.es

An = −  

ξ − ( m −1) .   β − ( m −1)

 (103)  

Thus,  the  problem  of  obtaining  the  scalar  p-­‐polarized  part  of  the  Green’s  function  for  the  considered  NIM   system   is   reduced   to   the   construction   of   the   solution   of   the   recurrence   relations   (98)-­‐(101)   and   obtaining   the   An coefficient.       3.b.  General  solutions  of  recurrence  relations   Note,   that   relations   (98)   and   (100)   have   similar   structure   with   different   initial   conditions.   The   same   takes   place   for   relations   (99)   and   (101).   Thus,   it   is   enough   to   obtain   general   solutions   for   relations   (98)-­‐(99),   and  to  use  then  the  corresponding  initial  conditions.  The  relation  for   β k  is  obtained  from  (98)-­‐(99)  and   has  the  form:  

βk = Kk βk + 2 − Lk βk + 4 ,  

 

(104)  

with  the  initial  conditions  

βn−4 = Pn−4 βn−2 − Qn−4γ n−2 ,   βn −3 = Pn −3 βn −1 − Qn −3γ n −1,  

 

⎛ ad ⎞ ⎟ γ n −1 ,   βn−1 = J n−1 .   ⎝ b c ⎠n − 2

β n − 2 = J n − 2 β n −1 − ⎜

 

(105)  

Correspondingly,  for   γ k  we  have  

γ k = M k γ k + 2 − Nk γ k + 4 ,  

 

(106)  

with  the  initial  conditions  

γ n−4 = Rn−4 βn−2 + Sn−4γ n−2 ,   γ n−3 = Rn−3 βn−1 + Sn−3γ n−1 ,  

 

γ n−2 =

 

⎞ ⎛ ah ⎞ an − 2 ⎛ hn − 2 β n −1 + γ n −1 ⎟ ,   γ n −1 = ⎜ ⎜ ⎟ .   bn − 2 ⎝ g n − 2 ⎝ b g ⎠ n −1 ⎠

(107)  

We  use  the  following  notations:  

 

⎛ ad ⎞ ⎛ ah ⎞ ⎛ ad ⎞ ⎛ ad ⎞ ⎛ a ⎞ Pk = J k J k +1 − ⎜ ⎟ ⎜ ⎟ ,   Qk = J k ⎜ ⎟ +⎜ ⎟ ⎜ ⎟ ,   ⎝ b c ⎠ k ⎝ b g ⎠ k +1 ⎝ b c ⎠ k +1 ⎝ b c ⎠ k ⎝ b ⎠ k +1

 

⎛ ah ⎞ ⎛ a ⎞ ⎛ ah ⎞ ⎛a ⎞ ⎛a ⎞ ⎛ ah⎞ ⎛ ad ⎞ Rk = ⎜ ⎟ J k +1 + ⎜ ⎟ ⎜ ⎟ ,   S k = ⎜ ⎟ ⎜ ⎟ − ⎜ ⎟ ,   ⎟ ⎜ ⎝ b g ⎠k ⎝ b ⎠ k ⎝ b g ⎠ k +1 ⎝ b ⎠ k ⎝ b ⎠ k +1 ⎝ bg ⎠ k ⎝ b c ⎠ k +1

   

Kk = Pk +

Qk Sk + 2 P QS ,   Lk = Qk Rk + 2 + k + 2 k k + 2 ,   Qk + 2 Qk + 2

M k = Sk +

Rk Pk + 2 P RS ,   N k = Rk Qk + 2 + k + 2 k k + 2 .   Rk + 2 Rk + 2

(108)  

(109)  

Note,   that   all   the   layers   with   the   same   parity   are   identical.   Therefore,   ζ k = ζ k + 2   for   −m ≤ k ≤ (n − 2) ,  

σ k±+1,l +1 = σ k±−1,l −1   and   σ k±,k +1 = σ k±,k −1   for   −(m − 1) ≤ k , l ≤ (n −1) .   For   the   Δ k   width   of   the   k-­‐th   layer,   we   have   Δk = Δk + 2   for   −m ≤ k ≤ (n − 2) .   For   the   considered   system   let   Δk  =  Δ1   for   even   k   (and   k  =  0),   and   Δk  =  Δ2   for  

odd  k.  Then,  the  expressions  (108)  are  as  follows:  

 

⎡σ + σ + ⎤ σ− σ− Pk = ⎢ k , k +1 k +1, k + k , k +1 k +1, k eiζ k 2 Δk + eiζ k +1 2 Δk +1 − eiζ k 2 Δk +iζ k +1 2 Δk +1 ⎥ ei (ζ k −ζ k +1 ) Δk +1 ,   4 4 ⎥⎦ ⎣⎢

 

Opt. Pura Apl. 48 (1) 31-53 (2015)

(

Qk = −

σ k−, k +1 2

)

(1 − e

43

iζ k 2 Δ k

)e

− i (ζ k + ζ k +1 ) xk +1

,  

(110)  

(111)  

© Sociedad Española de Óptica

ÓPTICA PURA Y APLICADA. www.sedoptica.es

Rk = −

 

σ k−+1, k 2

(1 − e

iζ k +1 2 Δ k +1

)e

iζ k ( xk + 2 Δ k +Δ k +1 ) + iζ k +1 ( xk −Δ k +1 )

,  

(112)  

Sk = ei (ζ k −ζ k+1 )Δk ,  

(113)  

Pk − 2 = Pk ,   Sk − 2 = Sk ,   Qk −2 = Qk ei (ζ k +ζ k+1 )(Δk +Δk+1 ) ,   Rk −2 = Rk e−i (ζ k +ζ k+1 )( Δk +Δk+1 ) .  

(114)  

  where    

The  expressions  (109)  are  as  follows:   (115)  

 

⎡σ + σ + ⎤ σ− σ− Kk = ⎢ k , k +1 k +1, k 1 + eiζ k 2Δk +iζ k +1 2Δk +1 + k , k +1 k +1, k eiζ k 2Δk + eiζ k +1 2Δk +1 ⎥ ei (ζ k −ζ k +1 ) Δk +1 ,   4 4 ⎥⎦ ⎣⎢

 

Lk = eiζ k 2( Δk +Δk +1 ) ,  

(116)  

 

(

)

(

)

⎡σ + σ + ⎤ σ− σ− M k = ⎢ k , k +1 k +1, k 1 + eiζ k 2Δk +iζ k +1 2Δk +1 + k ,k +1 k +1,k eiζ k 2Δk + eiζ k +1 2Δk +1 ⎥ e−iζ k Δk −iζ k +1 ( Δk + 2Δk +1 ) ,   (117)   4 4 ⎥⎦ ⎣⎢

(

)

(

)

Nk = e−iζ k +1 2( Δk +Δk+1 ) .  

 

(118)  

We  obtain  the  relations  of   Kk = Kk + 2 ,   Lk = Lk + 2 ,   M k = M k + 2 ,  and   Nk = Nk + 2  for   −m ≤ k ≤ (n − 2) .  Thus,  for   each   K k ,   Lk ,   M k ,   N k ,   ζ k ,   ε k ,  and   µ k  value  there  are  only  two  values  possessed  according  to  parity  of   the  k  index.  The  same  is  correct  for  the   σ k±, k +1  values.  From  now  on,  we  use  these  values  with  the  subscript   − − “1”   (e.g.,   K1 ,   σ 1,2 )   if   k   is   even   and   with   the   subscript   “2”   (e.g.,   K 2 ,   σ 2,1 )   if   k   is   odd.   Therefore,   equations  

(104)  and  (106)  for  even  k  (and  k  =  0)  or  for  odd  k  are  separately  considered  as  the  recurrence  relations   with   constant   coefficients.   The   solutions   of   these   relations   are   obtained   by   the   generating   function   method.   Let   us   consider   equation   (104)   with   the   boundary   conditions   (105)   for   even   k   with   the   following   substitution:  

u j = β k ,  

 

(119)  

where   j = (n − 1 − k ) / 2  and  k  is  even.   Thus,  for   k = (n − 1), (n − 3),…,   j = 0, 1, … ,   u0 = β n −1 ,   u1 = βn −3 etc  equation  (104)  becomes    

u j = K1u j −1 − L1u j − 2 ,  

 

(120)  

where   (121)  

 

⎡σ + σ + ⎤ σ− σ− K1 = ⎢ 1,2 2,1 1 + eiζ1 2 Δ1 +iζ 2 2 Δ2 + 1,2 2,1 eiζ1 2 Δ1 + eiζ 2 2 Δ2 ⎥ ei (ζ1 −ζ 2 ) Δ2 ,   4 ⎥⎦ ⎣⎢ 4

 

L1 = eiζ1 2( Δ1 +Δ2 ) ,  

(122)  

(

u0 =

)

+ ⎛ σ1,2

 

u1 =  

+ σ 1,2 2

− − ⎧⎪ ⎡ σ 1,2 σ 2,1 1 − eiζ1 2 Δ1 ⎨ ⎢1 − 4 ⎪⎩ ⎢⎣

(

(

− − σ1,2 σ 2,1

⎜1 + + + e 2 ⎜⎝ σ1,2 σ 2,1

) (1 − e 2

iζ 2 2 Δ 2

iζ 1 2 Δ1

)

⎞ −i (ζ1 −ζ 2 ) xn ,   ⎟⎟ e ⎠

⎫ iζ 1 4 Δ1 ⎪ − iζ 1 ( xn−1 +Δ1 −Δ 2 ) + iζ 2 ( xn−1 +Δ1 −Δ 2 ) e .   ⎬e + + 1,2σ 2,1 ⎪⎭

− − ⎤ σ 1,2 σ 2,1

)⎥ + σ ⎥⎦

(123)

  (124)  

The  generating  function   ∞

 

Opt. Pura Apl. 48 (1) 31-53 (2015)

U (t ) = ∑ u j t j  

(125)  

j =0

44

© Sociedad Española de Óptica

ÓPTICA PURA Y APLICADA. www.sedoptica.es

for  equation  (120)  is  defined  as  follows:  

U (t ) =

u0 − ( K1u0 − u1 ) t L1t 2 − K1t + 1

 

.  

(126)  

The  rational  function  from  (126)  is  expanded  into  the  partial  fractions  

U (t ) =  

U1 U − 2 ,   t − t1 t − t2

(127)  

where  

K1 + ( K12 − 4 L1 )

K1 − ( K12 − 4 L1 )

1/ 2

t1 =

 

2 L1

Ul =  

1/ 2

,   t2 =

2 L1

1 ⎛ u0 − ( K1u0 − u1 ) tl ⎜ L1 ⎝ t1 − t2

,  

(128)  

⎞ ⎟ ,   l = 1, 2 .   ⎠

(129)  

The  expansion  of  each  partial  fraction  (127)  into  a  series  is  as  follows:   j

1 1 ∞ ⎛t⎞ = − ∑ ⎜ ⎟ ,   l = 1, 2 .   t − tl tl j =0 ⎝ tl ⎠

 

(130)  

Comparing   the   u j unknown   in   (125)   to   the   coefficients   of   corresponding   t j   in   (126),   we   obtain   the     general  formula  for  the  solutions  of  equation  (120)  

uj = −

1 2

1 L1 ( K12 − 4 L1 )

j+2

1/ 2

(

(

)) ( )) (

)

1/ 2 1/ 2 j +1 ⎤ ⎡ 2 2 2 L u − K u − u K + K − 4 L K − K − 4 L ( ) ( ) ( ) 1 0 1 1 1 1 1 1 1 ⎢ 1 0 ⎥ ⎢ ⎥ ,   j + 1 ⎢ − 2 L u − K u − u K − K 2 − 4 L 1/ 2 K + K 2 − 4 L 1/ 2 ( 1 0 1) 1 ( 1 ( 1 1 ) ⎥⎥⎦ 1 0 1) 1 ⎢⎣

(

(

)

(131)  

where   j = 0, 1, … ,   K1   and   L1     are   obtained   from   (121)   and   (122)   correspondingly,   u0 ,   u1   are   obtained   from  (119)  and  (105)  correspondingly.  The  expression  (131)  contains  the  term   1/ 2

(K

2 1

− 4 L1 )

1/ 2

2 ⎡⎛ σ + σ + ⎤ ⎞ σ− σ− = ⎢⎜ 1,2 2,1 (1 + eiζ1 2 Δ1 + iζ 2 2 Δ2 ) + 1,2 2,1 ( eiζ1 2 Δ1 + eiζ 2 2 Δ2 ) ⎟ − 4eiζ1 2 Δ1 + iζ 2 2 Δ2 ⎥ ⎟ 4 ⎢⎜⎝ 4 ⎥ ⎠ ⎣ ⎦

ei (ζ1 −ζ 2 ) Δ2 .  

(132)  

The  general  formula  for  the  solutions  of  equation  (104)  with  the  boundary  conditions  (105)  for  odd  k  is   obtained  by  using  of  the  following  substitution:  

v j = β k ,  

 

(133)  

where   j = (n − 2 − k ) / 2 ,  k  is  odd.  Equation  (106)  with  the  boundary  conditions  (107)  is  solved  in  the   similar  way  as  equation  (104)  with  the  boundary  conditions  (105).     Equations   (98)   and   (100),   and   also   (99)   and   (101),   have   the   identical   structure   but   different   initial   conditions.   Therefore,   the   solutions   of   equations   (100)   and   (101)   are   obtained   using   the   corresponding   initial   conditions   in   the   way   similar   to   the   introduced   one.   Thus,   the   solutions   of   equations   (98)-­‐(101)   are   obtained  in  the  general  form  by  using  relations  (131),  (119),  and  similar  to  them.   3.c.  NIM  situation   We  are  interested  in  the  NIM  situation.  For  fixed   κ ,   Ω  and   ω0 ,  we  denote     2

⎛ Ω2 ⎞ ρ (ω ) = ω ⎜1 − 2 −κ 2 2 ⎟ ω − ω 0 ⎠ ⎝ 2

 

Opt. Pura Apl. 48 (1) 31-53 (2015)

45

12

.  

(134)  

© Sociedad Española de Óptica

ÓPTICA PURA Y APLICADA. www.sedoptica.es

According   to   the   results   of   [8],   there   are   two   cases   for   the   ζ k ( z) = ζ k ( z, κ )   value   (34)   obtained   in   the   k-­‐th   layer  with   z → ±ωˆ :   1. The   ωˆ > κ  case  is  called  "radiative  regime".  We  have   ζ k (±ωˆ ) = ∓ ρ (ωˆ )  if  k  is  even  (or  k  =  0),   − − ζ k (±ωˆ ) = ± ρ (ωˆ )  if  k  is  odd,  and   σ1,2 = σ1,2 ( z, κ ) → 0  with   z → ±ωˆ .  

  2. The   ωˆ < κ  case  is  called  "evanescent  regime".  We  have   ζ k (±ωˆ ) = i ρ (ωˆ )  for  any  index  k,  and   + + σ1,2 = σ1,2 ( z, κ ) → 0  with   z → ±ωˆ .  

( )

± We   introduce   the   conventional   notation   O σ 1,2 ,   which   implies   a   certain   value   that   has   the   same  

± ± infinitesimal   order   as   σ .   We   noted   above   (see   subsection   3.2)   that   σ k±, k +1   is   equal   to   σ 1,2   or   σ 2,1   ± 1,2

± ± ± ± according   to   parity   of   the   k   index.   The   σ1,2 = σ1,2 ( z, κ )   and   σ 2,1 = σ 2,1 ( z, κ )   values   tend   to   zero  

( )

± ± simultaneously  with   z → ±ωˆ   and  have  the  same  infinitesimal  order,  i.e.   σ 2,1 .  Thus,  the  general   = O σ 1,2

solutions   of   equations   (104)   and   (106)   in   the   NIM   situation   are   expressed   in   terms   of   the   asymptotic   ± approximations  with   σ1,2 → 0 .   Let   us   consider   the   solutions   (131)   and   (119)   of   relations   (104)   with   the   initial   conditions   (105)   in   the   NIM  situation.  With   z → ±ωˆ ,  the  expression  (132)  simplifies    

(K

− 4 L1 )

 

(K

− 4 L1 )

2 1 2 1

1/ 2

1/ 2

− = ei (ζ1 −ζ 2 ) Δ2 (1 − eiζ1 2 Δ1 + iζ 2 2 Δ2 )  for   ωˆ > κ ,   σ1,2 → 0  with   z → ±ωˆ ,  

(135)

+ = ei (ζ1 −ζ 2 ) Δ2 ( eiζ1 2 Δ1 − eiζ 2 2 Δ2 )  for   ωˆ < κ ,   σ1,2 → 0  with   z → ±ωˆ .  

(136)  

 

For  the   ωˆ < κ  case  with   z → ±ωˆ ,  the  expression  (131)  contains  the  first  order  pole:  

W=  

1

⎛ ρ (ωˆ ) Ω ⎞ 1 .   ~ ⎜ ⎟ ⎝ 2κ ⎠ ( z − ωˆ )( z + ωˆ ) 2

z →±ωˆ

+ σ1,2

(137)  

+ Therefore,   we   obtain   the   expression   (131)   as   asymptotic   approximation   with   σ1,2 → 0   ( z → ±ωˆ ).   From  

the   relation   for   σ kl+ (76)   we   have   the   following   expressions,   which   are   required   for   obtaining   the   below   presented  results:   − + ,   σ1,2 = 2 − σ1,2

 

(138)  

 

+ + + σ 2,1 = −σ 1,2 + O (σ 1,2 ) ,  

(139)  

 

− + + σ 2,1 = 2 + σ 1,2 + O (σ 1,2 ) ,  

(140)  

2

2

+ + σ1,2 σ 2,1

4

 

(σ ) =−

+ 2 1,2

4

+ + O (σ 1,2 ) .   3

(141)  

The  asymptotic  approximation  for  the  expression  (132)  is  obtained  by  the  expansion  in  the  Taylor’s  series   + + in  terms  of  powers  of   σ 1,2 ,  where   σ1,2 → 0  with   z → ±ωˆ .    

(K

2 1

− 4 L1 )

1/ 2

+ = ei (ζ1 −ζ 2 ) Δ2 ( eiζ1 2 Δ1 − eiζ 2 2 Δ2 ) + O (σ 1,2 ) .   2

(142)  

The  same  is  correct  for  other  values  that  are  contained  in  the  expression  (132):   1/ 2

 

K1 + ( K12 − 4 L1 )

+ = 2eiζ1 2 Δ1 ei (ζ1 −ζ 2 ) Δ2 + O (σ1,2 ) ,  

(143)  

1/ 2

 

K1 − ( K12 − 4 L1 )

+ = 2eiζ 2 2 Δ2 ei (ζ1 −ζ 2 ) Δ2 + O (σ1,2 ) ,  

(144)  

Opt. Pura Apl. 48 (1) 31-53 (2015)

2

2

46

© Sociedad Española de Óptica

ÓPTICA PURA Y APLICADA. www.sedoptica.es

 

+ u0 = −2Weiζ 1 2 Δ1 e − i (ζ 1 −ζ 2 ) xn + O (σ 1,2 ) ,  

(145)  

 

+ u1 = −2We− i (ζ1 −ζ 2 )( xn −Δ2 ) eiζ1 4 Δ1 + O (σ 1,2 ) ,  

(146)  

 

+ 2 L1u0 = −4Weiζ1 2(2 Δ1 +Δ2 ) e −i (ζ 1 −ζ 2 ) xn + O (σ 1,2 ) ,  

(147)  

 

+ K1u0 − u1 = −2Weiζ 1 2 Δ1 + iζ 2 2 Δ2 ei (ζ 1 −ζ 2 ) Δ2 e − i (ζ 1 −ζ 2 ) xn + O (σ 1,2 ) .  

(148)  

The   general   solution   (131)   of   the   recurrence   relation   (120)   in   the   NIM   situation   for   ωˆ < κ   is   presented   as   follows:   + u j = −2We −i (ζ1 −ζ 2 )( xn − jΔ2 ) eiζ 1 ( j +1)2 Δ1 + O (σ 1,2 )  with   j = 0,1,… .  

 

(149)  

This  expression  contains  the  first  order  pole  (137)  and  grows  to  infinity  with   z → ±ωˆ .     3.d.  Recurrence  relation  solutions  in  the  NIM  situation   Using  the  presented  above  way  (see  subsection  3.2,  3.3),  we  obtain  asymptotic  approximations  in  the  NIM   situation   for   the   β k ,   γ k ,   ξ k ,   and   η k   unknowns   that   satisfy   the   recurrence   relations   (98)-­‐(101).   For   the   ωˆ > κ  case,  they  are  singularity-­‐free,  and  for  even  k  are  represented  as  follows:   − β k = e± i ρ (( n +1) Δ + k Δ ) + O (σ 1,2 ) ,   γ k = O (σ 1,2− ) ,   1

 

ξk = − I 0 e±iρ ( k Δ

 

2 + y)

2

2

− + O (σ 1,2 ) ,   ηk = I 0 e±iρ (k Δ1 − y ) + O (σ1,2− ) ,  

(150)  

for  odd  k:   − β k = e± iρ ( n − k ) Δ + O (σ1,2 ) ,   γ k = O (σ 1,2− ) ,  

   

2

1

− ξ k = − I 0 e ± i ρ ( − ( k +1) Δ + y ) + O (σ 1,2 ) ,   ηk = I 0 e±iρ (−(k −1)Δ 1

2 − y)

− + O (σ 1,2 ) ,  

(151)  

where   ρ = ρ (ωˆ )  (134),  the  value  (85)  is  

I0 = ±

 

ρ (ωˆ ) .   2iωˆ 2

(152)  

Relation  (103)  in  that  case  is   − An = I 0 e ± i ρ (ωˆ )( − ( n +1) Δ1 + y ) + O (σ 1,2 ) .  

 

(153)  

For  the   ωˆ < κ   case,  we  obtain  asymptotic  approximations,  which  contains  the  first  order  pole  (137).  For   even  k    

+ β k = −2 (W − 1) e− ρ ( n +1− k ) Δ + O (σ 1,2 ) ,  

(154)  

 

+ γ k = ( 2W − 1) e− ρ (( n +1) Δ + k Δ ) + O (σ 1,2 ) ,  

(155)  

(156)  

 

⎡ − ( 2W − 1) e ρ ( k Δ1 − y ) ⎤ ⎢ ⎥ + ξ k = I 0 ⎢ 1 − e− ρ 2 Δ1 e − ρ ( k − 2) Δ1 − 1 − e − ρ 2 Δ2 e − ρ ( k − 2) Δ2 ρ ( k − 2)( Δ +Δ ) ρ y ⎥ + O (σ 1,2 ) ,   1 2 e e ⎢+ ⎥ − ρ 2 Δ1 − ρ 2 Δ2 e −e ⎢⎣ ⎥⎦

 

⎡ 1 − e ρ 2 Δ2 e− ρ k Δ1 − e− ρ k Δ2 − ρ 2( Δ +Δ ) ρ y ⎤ − ρ ( k Δ2 + y ) + 1 2 ⎢ ηk = I 0 2We + e e ⎥ + O (σ 1,2 ) ,   − ρ 2 Δ1 − ρ 2 Δ2 ⎢ ⎥ e − e ⎣ ⎦

1

1

(

)

(

(

(

(

)

)(

2

)

)

)

(157)  

for  odd  k  

Opt. Pura Apl. 48 (1) 31-53 (2015)

47

© Sociedad Española de Óptica

ÓPTICA PURA Y APLICADA. www.sedoptica.es

βk

(e =

− ρ ( n + 2 − k ) Δ1

(e

 

γk = −

(1 − e

− ρ 2 Δ1

ξ k = − I 0 ⎢ 2We ρ ( k −1) Δ e ρ y + 2

)( e

(e

 



)

( )

− e − ρ ( n + 2 − k ) Δ2 − e − ρ 2( Δ1 +Δ2 ) e− ρ ( n − k ) Δ1 − e− ρ ( n − k ) Δ 2 −e

− ρ ( n − k ) Δ1

− ρ 2 Δ1

(1 − e

− ρ 2 Δ1

−e

− ρ 2 Δ2

(e

− e− ρ ( n − k ) Δ2

− ρ 2 Δ2

)(e

− ρ 2 Δ2

)

− ρ ( k −1) Δ1

− ρ 2 Δ1

)e

− ρ ( k +1)( Δ1 +Δ 2 )

− e − ρ ( k −1) Δ2

− ρ 2 Δ2

)

)e

) + O σ ,   ( ) + 2 1,2

+ + O (σ 1,2 ) ,  

ρ ( k −1)( Δ1 +Δ 2 ) − ρ y

e

⎤ + ⎥ + O (σ 1,2 ) ,   ⎥ ⎦

 

⎢ ⎣

 

⎡( 2W − 1) e − ρ ( k +1) Δ1 e ρ y ⎤ ⎢ ⎥ + ηk = I 0 ⎢ 1 − e− ρ 2 Δ2 e− ρ ( k +1) Δ1 − 1 − e− ρ 2 Δ1 e− ρ ( k +1) Δ2 − ρ y ⎥ + O (σ 1,2 ) ,   e ⎢+ ⎥ − ρ 2 Δ1 − ρ 2 Δ2 e −e ⎣⎢ ⎦⎥

(

)

(

−e

(

)

)

(158)  

(159)  

(160)  

(161)  

where   ρ = ρ (ωˆ ) (134),  the  value  (85)  is    

I0 = −

 

ρ (ωˆ ) .   2ωˆ 2

(162)  

Relation  (103),  in  that  case,  is   + An = − I 0 e ρ (ωˆ )(( n +1) Δ1 − y ) + O (σ 1,2 ) .  

 

(163)  

 

4.  Results   4.a.  p-­‐polarization  case   The  coefficients  (96)  and  (97)  are  obtained  in  the  general  case  for  any  z  with  relations  (131),  (119)  and   other   ones   (see   subsection   3.2).   We   consider   the   particular   case,   i.e.   the   NIM   situation,   which   implies   that   the   coefficients   (96)   and   (97)   are   expressed   with   relations   (150),   (152),   (153),   (154)-­‐(161),   and   (163).   Thus,  we  obtain  the  expressions  for  the  scalar  p-­‐polarization  part  of  the  electric  Green’s  function  (90)  for   every  layer.  In  the  k-­‐th  layer  for   ωˆ > κ  with  fixed   κ  

Gkp ( x, y, ±ωˆ , κ ) = ±

 

ρ (ωˆ ) ±i ρ (ωˆ ) X ( x , y , k ) ,   e 2iωˆ 2

(164)  

ρ (ωˆ ) − ρ (ωˆ ) X ( x, y , k ) ,   e 2ωˆ 2

(165)  

for   ωˆ < κ  

Gkp ( x, y, ±ωˆ , κ ) =

  where      

X ( x, y, k ) = k Δ2 − x − y  with   k = −(m − 1),…− 2,0,2,…(n − 1) ,  

(166)  

 

X ( x, y, k ) = − k Δ1 + x − Δ1 + y  with   k = −m,…− 3,−1,1,3,…n .  

(167)  

The   expressions   (164)   and   (165)   are   continuous   on   the   layer’s   surfaces   at   the   xk   coordinates,  

k = −(m − 1),…,n ,  and  at  the   y  coordinate   of   the   point   source.   The   second   one   (165)   has   no   singularities   despite   the   fact   that   the   asymptotic   approximations   (154)-­‐(155),   (160),   and   (161)   for   the   ωˆ < κ   case   contain   the   first   order   poles   (137).   During   solving   relations   (98)-­‐(101),   all   approximations   have   the   + + + accuracy  of  the  same  infinitesimal  order  with   σ 1,2 ,  where   σ1,2 O σ 1,2 → 0  with   z → ±ωˆ .  Thus,  we  can  be     sure  of  the  accuracy  of  the  expression  obtained,  and  use  equal  sign  instead  of  asymptotic  equivalence  sign   with   z → ±ωˆ .  

( )

Opt. Pura Apl. 48 (1) 31-53 (2015)

48

© Sociedad Española de Óptica

ÓPTICA PURA Y APLICADA. www.sedoptica.es

One  can  stress  that  relations  (164)  and  (165)  are  composed  of  one  term.  Therefore,  the  reflection  terms   are  absent,  and  in  the  NIM  situation  the  reflection  of  the  p-­‐polarization  part  of  the  electric  field  is  absent,   too.     4.b.  s-­‐polarization  case   Let  us  consider  the  s-­‐polarization  case.  It  can  be  shown  that  relations  (96),  (97),  (98)-­‐(102),  and  (103)  are   correct  for  the  s-­‐polarization  case  with  the  substitution  of   σ kl±  (76)  to   τ kl±  (93),  taking  into  account  relation  

( )

− − − − (95).   In   the   NIM   situation   for   ωˆ > κ ,   the   value   τ1,2   = O σ 1,2 = τ1,2 ( z, κ ) → 0   with   z → ±ωˆ   and   τ 1,2

according   to   (94).   For   ωˆ < κ ,   the   value   τ

+ 1,2

= τ ( z, κ ) → 0   with   z → ±ωˆ   and   τ + 1,2

+ 1,2

= O (σ

+ 1,2

)   according   to  

(94).  Then,  the  solutions  of  equations  (98)-­‐(101)  and  the  value  (103)  are  obtained  in  the  NIM  situation  for   ωˆ > κ  by  relations  (150),  (151),  and  (153),  where  (152)  is  

1 ,   2i ρ (ωˆ )

I0 = ±

 

(168)  

for   ωˆ < κ  by  relations  (154)-­‐(161),  and  (163),  where  (162)  is  

I0 =

 

1 ,   2ρ (ωˆ )

(169)  

and  

W=  

1 + τ1,2

=−

1 + σ1,2

⎛ ρ (ωˆ ) Ω ⎞ 1 .   ~ −⎜ ⎟ ⎝ 2κ ⎠ ( z − ωˆ )( z + ωˆ ) 2

z →±ωˆ

(170)  

Analogously  to  the  p-­‐polarization  case  we  obtain  the  expressions  for  the  scalar  s-­‐polarization  part  of  the   electric  Green’s  function  (90)  for  every  layer.  In  the  k-­‐th  layer  for   ωˆ > κ  with  fixed   κ ,  

Gks ( x, y, ±ωˆ , κ ) = ±

 

1 e± i ρ (ωˆ ) X ( x, y , k ) ,   2i ρ (ωˆ )

(171)  

1 e− ρ (ωˆ ) X ( x, y , k ) ,   2ρ (ωˆ )

(172)  

for   ωˆ < κ ,  

Gks ( x, y, ±ωˆ , κ ) =

 

where   X ( x, y, k )  is  defined  in  (166)  and  (167).   The   expressions   (171)   and   (172)   differ   from   (164)   and   (165)   only   in   the   coefficients   of   exponents.   Therefore,  they  are  continuous  on  the  layer’s  surfaces  at  the   xk  coordinates,   k = −(m − 1),K , n ,  and  at  the  y   coordinate  of  the  point  source.  The  reflection  of  the  s-­‐polarization  part  of  the  electric  field,  as  well  as  of  the   p-­‐polarization  part,  is  absent.   4.c.  Electric  Green’s  function   Now  we  return  to  the  scalar  representation  of  the  vector   s-­‐  and  p-­‐polarization  parts  of  the  electric  Green’s   function  (41)  and  (42).  For   ωˆ > κ  and  even  k,  where   k = −(m − 1),…− 2,0,2,…(n − 1) ,  

 

∂ p ∂ Gk (x, y,±ωˆ ,κ ) = − Gkp (x, y,±ωˆ ,κ ) = ∓iρ (ωˆ )sign(x − y)Gkp (x, y,±ωˆ ,κ ) ,   ∂x ∂y

(173)  

for  odd  k,  where   k = −m,…− 3,−1,1,3,…n ,  

 

∂ p ∂ p Gk (x, y,±ωˆ ,κ ) = G (x, y,±ωˆ ,κ ) = ±iρ (ωˆ )sign(x − Δ1 + y)Gkp (x, y,±ωˆ ,κ ) ,   ∂x ∂y k

(174)  

for   ωˆ < κ  and  even  k,  where k = −(m − 1),…− 2,0,2,…(n − 1) ,    

Opt. Pura Apl. 48 (1) 31-53 (2015)

49

© Sociedad Española de Óptica

ÓPTICA PURA Y APLICADA. www.sedoptica.es

 

∂ p ∂ Gk ( x, y, ±ωˆ , κ ) = − Gkp ( x, y, ±ωˆ , κ ) = ρ (ωˆ )sign( x − y) Gkp ( x, y, ±ωˆ , κ ) ,   ∂x ∂y

(175)  

for  odd  k,  where   k = −m,…− 3,−1,1,3,…n ,  

 

∂ p ∂ Gk ( x, y, ±ωˆ , κ ) = Gkp ( x, y, ±ωˆ , κ ) = − ρ (ωˆ )sign( x − Δ1 + y) Gkp ( x, y, ±ωˆ , κ ) ,   ∂x ∂y

(176)  

where  the  sign  function  is  

⎧+1 x ≥ 0 .   sign( x) = ⎨ ⎩−1 x < 0

 

(177)  

Then   the   electric   Green’s   function   (40)   from   the   decomposition   (32)   in   the   NIM   situation   is   expressed   with   the   componentwise   representation   of   matrices   (37)   in   the   following   way:   for   ωˆ > κ   in   the   layers   filled  with  NIM,  i.e.  with  the  even  index  k,  where   k = −(m − 1),…− 2,0,2,…(n − 1) ,  

Gκ ( x, y, ±ωˆ ) = ±

 

 

1 ± i ρ (ωˆ ) ( k Δ2 − x − y ) e ⋅ 2i ρ (ωˆ )

  ⎧⎪ ⎤ ⎫⎪ ρ 2 (ωˆ ) ⎡ κ κ2 ⋅ ⎨e3 × eκ e3 × eκ + e e ± sign( x − y ) e e + e e + e3e3 ⎥ ⎬ ,   ( ) κ κ κ 3 3 κ ⎢ 2 2 ρ (ωˆ ) ωˆ ⎣ ρ (ωˆ ) ⎦ ⎭⎪ ⎩⎪

(178)

 

in  the  layers  filled  with  a  vacuum,  i.e.  with  the  odd  index  k,  where   k = −m,…− 3,−1,1,3,…n ,  

Gκ ( x, y, ±ωˆ ) = ±

 

 

1 ± i ρ (ωˆ ) ( − k Δ1 + x −Δ1 + y ) e ⋅ ˆ 2i ρ (ω )

  ⎤ ⎪⎫ ρ 2 (ωˆ ) ⎡ κ κ2 ⎪⎧ ⋅ ⎨e3 × eκ e3 × eκ + e e ± sign( x − Δ + y ) e e − e e − e3e3 ⎥ ⎬ ,   ( ) κ κ 1 κ 3 3 κ ⎢ 2 2 ρ (ωˆ ) ωˆ ⎣ ρ (ωˆ ) ⎪⎩ ⎦ ⎪⎭

(179)  

for   ωˆ < κ  in  the  layers  filled  with  NIM,  i.e.  with  the  even  index  k,  where   k = −(m − 1),…− 2,0,2,…(n − 1) ,  

G κ ( x, y, ±ωˆ ) =

 

 

1 − ρ (ωˆ ) ( k Δ2 − x − y ) e ⋅ ˆ 2 ρ (ω )

  ⎤ ⎪⎫ ρ 2 (ωˆ ) ⎡ iκ κ2 ⎪⎧ ⋅ ⎨e3 × eκ e3 × eκ + e e − sign( x − y ) e e + e e − e3e3 ⎥ ⎬ ,   ( ) κ κ κ 3 3 κ ⎢ 2 2 ρ (ωˆ ) ωˆ ⎣ ρ (ωˆ ) ⎪⎩ ⎦ ⎪⎭

(180)  

in  the  layers  filled  with  a  vacuum,  i.e.  with  the  odd  index  k,  where   k = −m,…− 3,−1,1,3,…n ,  

G κ ( x, y, ±ωˆ ) =

 

 

1 − ρ (ωˆ ) ( − k Δ1 + x −Δ1 + y ) e ⋅ 2 ρ (ωˆ )

  ⎤ ⎪⎫ ρ 2 (ωˆ ) ⎡ iκ κ2 ⎪⎧ ⋅ ⎨e3 × eκ e3 × eκ + e e − sign( x − Δ + y ) e e − e e + e3e3 ⎥ ⎬ ,   ( ) κ κ 1 κ 3 3 κ ⎢ 2 2 ρ (ωˆ ) ωˆ ⎣ ρ (ωˆ ) ⎪⎩ ⎦ ⎪⎭

(181)  

where   the   sign   function   sign( x)   and   the   function   ρ (ω)   is   defined   in   (177)   and   (134),   correspondingly.   Thus,  with  the  decompositions  (32)  and  (33),  we  can  obtain  the  expression  of  the  electric  field   E(x, t )  for   t  ≥  t0,   where   t0   =   0.   The   electric   Green’s   function   can   be   obtained   by   the   inverse   Laplace   transform   (8)   and   the  Fourier  transform  (28).    

5.  Particular  case  of  the  layered  NIM  system   Now,  let  us  consider  the  particular  case  of  the  NIM  system  studied  in  the  paper  (see  subsection  2.2).  We   set   m  =  1,   where   m   is   the   number   of   layers   located   to   the   left   side   of   the   zero   layer.   Then,   the   system   composed   of   (n+2)   parallel   alternated   layers   with   numbers   k = −1, 0,1,...n   (Fig.   2).   The   (–1)-­‐th   layer   is   the  

Opt. Pura Apl. 48 (1) 31-53 (2015)

50

© Sociedad Española de Óptica

ÓPTICA PURA Y APLICADA. www.sedoptica.es

half   space   unbounded   along   the   direction   of   the   x   axis   ( x−1 = −∞ )   and   is   empty   (vacuum).   The   point   source  is  located  at  y  coordinate  in  the  (–1)-­‐th  layer,  i.e.   x−1 < y < x0  or   −∞ < y < 0 .    

   

Fig.  2.  The  NIM  system  composed  of  (n+2)  parallel  layers  filled  with  NIM  and  vacuum.  The  point  source  is  located  at  y  coordinate  in   (–1)-­‐th  empty  (vacuum)  layer  that  is  the  half  space  unbounded  along  the  negative  direction  of  the  x  axis.  

Analogously  to  the  way  proposed  in  the  paper  (see  sections  2-­‐4)  we  obtain  the  expressions  for  the  scalar   s-­‐   and   p-­‐polarization  parts  of  the  electric  Green’s  function  (90)  for  every  layer.  In  the  k-­‐th   layer   for   ωˆ > κ   with  fixed   κ ,  

   

ρ (ωˆ ) ±i ρ (ωˆ ) X ( x , y , k ) ,   e 2iωˆ 2

(182)  

1 e± i ρ (ωˆ ) X ( x, y , k ) ,   2i ρ (ωˆ )

(183)  

ρ (ωˆ ) − ρ (ωˆ ) X ( x, y , k ) ,   e 2ωˆ 2

(184)  

1 e− ρ (ωˆ ) X ( x, y , k ) ,   2ρ (ωˆ )

(185)  

Gkp ( x, y, ±ωˆ , κ ) = ± Gks ( x, y, ±ωˆ , κ ) = ±

for   ωˆ < κ ,  

   

Gkp ( x, y, ±ωˆ , κ ) = Gks ( x, y, ±ωˆ , κ ) =

where    

X ( x, y, k ) = k Δ2 − x − y  with   k = 0,2,…(n − 1) ,  

(186)  

 

X ( x, y, k ) = −(k + 1)Δ1 + x − y  with   k = −1,1,3,…n .  

(187)  

The   expressions   in   (182),   (183),   (184),   and   (185)   are   the   same   as   in   (164),   (171),   (165),   and   (172)   correspondingly.   However,   relations   (186)   and   (187)   differ   from   (166)   and   (167).   Nevertheless,   the   expressions   (182)-­‐(185)   are   continuous   on   the   layer’s   surfaces   at   the   xk   coordinates,   k = 0,…,(n − 1) ,   and   at   the   y   coordinate   of   the   point   source.   The   reflection   of   the   s-­‐   and   p-­‐polarization   parts   of   the   electric   field   are  absent.   Then   the   electric   Green’s   function   (40)   from   the   decomposition   (32)   in   the   NIM   situation   is   expressed   with   the   componentwise   representation   of   matrices   (37)   in   the   following   way:   for   ωˆ > κ   in   the   layers   filled  with  NIM,  i.e.  with  the  even  index  k,  where   k = 0,2,…(n − 1) ,  

 

 

Gκ ( x, y, ±ωˆ ) = ±

1 e±i ρ (ωˆ )( k Δ2 − x − y ) ⋅ 2i ρ (ωˆ )  

⎡ ⎞⎛ ⎞⎤ ρ 2 (ωˆ ) ⎛ κ κ ⋅ ⎢ e3 × eκ e3 × eκ + eκ ± e3 ⎟ ⎜ eκ ∓ e3 ⎟ ⎥ ,   2 ⎜ ρ (ωˆ ) ⎠ ⎝ ρ (ωˆ ) ⎠ ⎦ ωˆ ⎝ ⎣

(188)  

Opt. Pura Apl. 48 (1) 31-53 (2015)

51

© Sociedad Española de Óptica

ÓPTICA PURA Y APLICADA. www.sedoptica.es

in  the  layers  filled  with  a  vacuum,  i.e.  with  the  odd  index  k,  where   k = −1,1,3,…n ,  

Gκ ( x, y, ±ωˆ ) = ±

 

 

1 ± i ρ (ωˆ )( − ( k +1) Δ1 + x − y ) e ⋅ 2i ρ (ωˆ )  

⎡ ⎞⎛ ⎞⎤ ρ 2 (ωˆ ) ⎛ κ κ ⋅ ⎢ e3 × eκ e3 × eκ + eκ ∓ sign(x − y) e3 ⎟ ⎜ eκ ∓ sign(x − y) e3 ⎟ ⎥ ,   2 ⎜ ρ (ωˆ ) ⎠ ⎝ ρ (ωˆ ) ⎠ ⎦ ωˆ ⎝ ⎣

(189)  

for   ωˆ < κ  in  the  layers  filled  with  NIM,  i.e.  with  the  even  index  k,  where   k = 0,2,…(n − 1) ,  

G κ ( x, y, ±ωˆ ) =

 

1 e− ρ (ωˆ )( k Δ2 − x − y ) ⋅ 2 ρ (ωˆ )

 iκ ⎡ ⎞⎛ ⎞⎤ ρ 2 (ωˆ ) ⎛ iκ ⋅ ⎢ e 3 × eκ e 3 × e κ + eκ − e3 ⎟⎜ eκ + e3 ⎟ ⎥ ,   ⎜ 2 ρ (ωˆ ) ⎠⎝ ρ (ωˆ ) ⎠ ⎦ ωˆ ⎝ ⎣

 

(190)  

in  the  layers  filled  with  a  vacuum,  i.e.  with  the  odd  index  k,  where   k = −1,1,3,…n ,    

Gκ ( x, y, ±ωˆ ) =

 

 

1 − ρ (ωˆ )( − ( k +1) Δ1 + x − y ) e ⋅ 2ρ (ωˆ )  

⎡ ⎞⎛ ⎞⎤ ρ 2 (ωˆ ) ⎛ iκ iκ ⋅ ⎢ e 3 × eκ e 3 × e κ + eκ + sign( x − y) e3 ⎟⎜ eκ + sign( x − y) e3 ⎟ ⎥ ,   ⎜ 2 ρ (ωˆ ) ⎠⎝ ρ (ωˆ ) ⎠ ⎦ ωˆ ⎝ ⎣

(191)  

where   the   sign   function   sign( x)   and   the   function   ρ (ω)   is   defined   in   (177)   and   (134),   correspondingly.   Thus,  with  the  decompositions  (32)  and  (33),  we  can  obtain  the  expression  of  the  electric  field   E(x, t )  for   t  ≥  t0,   where   t0   =   0.   The   electric   Green’s   function   can   be   obtained   by   the   inverse   Laplace   transform   (8)   and   the  Fourier  transform  (28).      

6.  Conclusions   In  this  paper,  we  solved  the  problem  of  obtaining  the  electric  Green’s  function  for  the  layered  NIM  system.   The   point   source   is   located   inside   a   certain   NIM   layer   of   the   system.   We   proposed   the   way   of   its   obtaining   for   any   frequency   ω.   In   the   NIM   situation,   we   obtained   relations   for   the   Fourier   transformed   electric   Green’s   function,   which   is   the   kernel   of   the   decomposition   (32).   Then,   with   the   decomposition   (32)   and   the   function   of   the   initial   field’s   configuration,   the   electric   field’s   values   at   any   point   of   the   system   can   easily   be   evaluated.   The   magnetic   field   can   as   well   be   obtained   with   reasoning   similar   to   the   presented   above.  The  obtained  formulae  are  symmetric,  relative  to  the  position  of  the  point  source.  This  fact  shows   the   correlation   with   the   physical   conception   of   the   electromagnetic   field   propagated   into   the   system   composed  of  isotropic  homogeneous  layers.  Besides,  the  reflection  in  the  NIM  situation  is  absent.   For   the   particular   case   of   considered   NIM   system,   where   the   point   source   is   located   in   the   last-­‐to-­‐left-­‐side   unbounded  empty  (vacuum)  layer,  we  obtained  the  similar  formulae  for  the  electric  Green’s  function  and   observed  that  the  reflection  in  the  NIM  situation  also  is  absent.   We  did  not  consider  absorption  in  the  NIM  case.  This  is  straightforward  to  do  along  the  same  lines.  The   two  frequencies   ±ωˆ  now  acquire  a  negative  imaginary  part,  so  the  reflection  term  in  Green’s  function  no   longer  vanishes.   We   present   formulae   for   the   NIM   system   under   some   particular   non-­‐essential   limitations   (The   number   of   layers   is   odd   and   can   be   7,   11,   15,   etc;   the   point   source   is   located   in   the   layer   filled   with   NIM;   the   number   of   layers   located   to   the   left   and   to   the   right   from   the   point   source,   is   odd   and   greater   than   or   equal   to   three,   so   at   least   seven   layers   can   form   the   system).   However,   the   corresponding   formulae   for   other   cases   could  be  evaluated  by  the  same  way.   The  system  can  be  composed  of  arbitrary  finite  number  of  layers.  This  fact  allows  us  to  use  the  considered   system   as   a   model   for   simulation   or   engineering   of   the   real   objects,   such   as   superlens   systems   and  

Opt. Pura Apl. 48 (1) 31-53 (2015)

52

© Sociedad Española de Óptica

ÓPTICA PURA Y APLICADA. www.sedoptica.es

multilayer  NIM  coverings.  Obtained  absence  of  reflection  (for  the  leading  asymptotic  term  near  the  NIM   frequency)  opens  an  intriguing  prospect.     Acknowledgements  

 

We  thank  Professor  G.P.  Miroshnichenko  for  many  helpful  discussions.  This  work  was  partially  financially   supported  by  the  Government  of  the  Russian  Federation  (grant  074-­‐U01),  by  the  Ministry  of  Science  and   Education  of  the  Russian  Federation  (GOSZADANIE  2014/190,  Project  14.Z50.31.0031),  by  grant  MK-­‐ 5001.2015.1  of  the  President  of  the  Russian  Federation.    

 

Opt. Pura Apl. 48 (1) 31-53 (2015)

53

© Sociedad Española de Óptica

Suggest Documents