Optics Communications 283 (2010) 3367–3370

Contents lists available at ScienceDirect

Optics Communications
j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / o p t c o m

Polarization-dependent Goos–Hänchen shift at a graded dielectric interface

W. Löffler a,⁎, M.P. van Exter a, G.W. 't Hooft a, E.R. Eliel a, K. Hermans b, D.J. Broer b, J.P. Woerdman a
a
Huygens Laboratory, Leiden University, P.O. Box 9504, 2300 RA Leiden, The Netherlands
b
University of Technology Eindhoven, PO Box 513, 5600 MB Eindhoven, The Netherlands

a r t i c l e i n f o a b s t r a c t

Article history: We examine the polarization differential Goos–Hänchen beam shift upon total internal reflection, for a graded-
Received 11 February 2010 index dielectric interface. We find a generic scaling law where the magnitude of this shift depends solely on the
Received in revised form 21 April 2010 product of wavelength and gradient steepness. The analytic results are extended using transmission matrix
Accepted 22 April 2010
calculations in cases where the assumptions made to allow analytical treatment might become questionable. Two
important cases in this category are: (i) incident angle close to the critical angle and (ii) gradients with an overall
Keywords:
thickness of the order of a wavelength. We demonstrate this effect experimentally using a polymer-blend sample
Graded-index (GRIN) optics
Goos–Hänchen (GH) shift
with a gradual refractive-index transition induced by diffusion.
Polarization © 2010 Elsevier B.V. All rights reserved.
Inhomogeneous media

1. Introduction differential GH beam displacement in GRIN media have been reported.

The GH shift in 1D GRIN media has only been studied theoretically, for
When a light beam undergoes total internal reflection 100% of its perpendicular s (TE) polarization [9]. A related effect, the transverse shift
energy is reflected. Because evanescent fields penetrate into the lower- of a circularly polarized beam in a GRIN medium can be described within
index medium, the reflected beam does not follow the geometrically the Berry-phase framework [10]. Recently, Bliokh et al. [11] developed
expected path, but is translated a little, the amount depending on the an analytic theory describing ray paths in general inhomogeneous media
polarization of the incident light. This translation, known as the Goos– emphasizing geometric phases. In principle the analytic part of our
Hänchen (GH) shift [1], is a correction to geometric optics. It arises investigations is contained in this theory, we however find direct
because the plane-wave components of the incident beam pick up a analytic and numeric results for the special, experimentally attractive
phase jump upon reflection that depends on the incident angle. case of a planar (1D) structure, which was not discussed before [10,11].
The GH effect has been studied for interfaces separating homoge- Here we discuss only the case of total reflection, tunneling of light in
neous dielectric media [1] and metamaterials [2], as well as for GRIN media has been studied in [12,13]. The 1D GRIN medium is also the
metallic [3] and absorbing [4,5] layers. In inhomogeneous media, the simplest case in the very young field of transformation optics [14,15]:
polarization-dependent GH effect is unexplored yet; since abrupt there, the full potential of inhomogeneous media is demonstrated,
interfaces are a special case of graded transitions, new insight can be concealing objects and allowing omnidirectional retro-reflection to
obtained by comparing these two cases (see, e.g., [6,7]). The charac- name a few examples. The GH shift that we discuss here marks a general
teristic geometric parameter of total internal reflection, i.e. the critical limit of such devices.
angle, is the same for gradual and abrupt transitions but the beam is In this paper we first give an analytical theory of the polarization-
not abruptly but gradually steered into the reflected direction. This dependent GH beam shift for a graded interface, subsequently we
brings us to the question that we address here: how does the extend the results using numerical simulation for cases where the
(polarization-dependent) GH beam shift depend on the refractive assumptions made for analytical treatment are questionable. Finally,
index profile? we give a first experimental demonstration of the effect.
Since graded-index (GRIN) dielectrics are very important for wave-
guide optics, the influence of polarization on the phase of a propagating 2. Theory and numerical simulations
mode has been investigated theoretically in a cylindrical geometry [8].
However no analytical results or experimental tests for the polarization As a starting point of our analytic description, we take a linearly
polarized Gaussian light beam that undergoes total internal reflection in
a 1D graded dielectric structure (see Fig. 1). Because we are interested in
beam displacements in the plane of incidence, it is sufficient to analyze
Maxwell's equation in two dimensions.
⁎ Corresponding author. For simplicity we assume an infinitely extended refractive-index
E-mail address: loeffl[email protected] (W. Löffler). gradient. The gradient is along the x axis and n(x) is decreasing with

0030-4018/$ – see front matter © 2010 Elsevier B.V. All rights reserved.
doi:10.1016/j.optcom.2010.04.055
3368 W. Löffler et al. / Optics Communications 283 (2010) 3367–3370

differences of s and p polarized beams manifest themselves in the

relative beam displacement; this is evident by comparing Eqs. (6) and
(7). Therefore we focus here on the reflection phase difference
between s and p polarization:


xtp
δV x′ ′
ϕp −ϕs = ∫   dx ð8Þ
0 kx x′

For easy interpretation we introduce a scaling parameter x0, being a

measure of the width of the gradient, and rewrite the refractive index
Fig. 1. Scheme of the experiment: a light beam impinges from the high-refractive index profile as n(x) = n0f(x/x0) where n0 is a constant. We introduce the
(n1) material upon the graded-index (GRIN) transition layer where the refractive index incident angle as θinc, β ≡ λ− 1
̄ n0 sinθinc with λ̄n0 ≡ λ/[2πn(0)], and x̃ ≡ x′/x0.
gradually reaches n2. Right: exemplary refractive index profile. The turning point of the
Then the differential phase shift is
beam is at x = xtp.
h i2
λ̄n0 x̃tp f ðx̃Þf ″ ðx̃Þ−2 f ′ ðx̃Þ
ϕp −ϕs = ∫ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi dx̃ ð9Þ
positive x; the whole system is translationally invariant in the y-direction. x0 0
½f ðx̃Þ2 ½f ðx̃Þ2 −sin2 θinc
We assume n to be real, i.e. neglect absorption. The fields can be separated
like E(r)=E(x)exp(iβz) for the electric field and H(r)=H(x)exp(iβz) for
the magnetic field (β≡kz is the propagation constant parallel to the This differential phase is dependent on λ̄n0 and therefore essentially
interface). We only need to consider one component of the fields E and H, different from the conventional phase shift at an abrupt interface, which
Ey and Hy. For s polarization (TE, Ey(x) component) the wave equation is is independent of wavelength (i.e. zero-order in λ̄n0). The latter disap-
(see, e.g., Ref. [8]): pears in GRIN media and Eq. (9) gives the first-order correction. This was
also found by Liberman and Zeldovich [19] as well as by Bliokh and


2 2 2 2
∂x Ey + n k0 −β Ey = 0 ð1Þ Stepanovskii [20]; they retrieved equivalent results for the phase differ-
ence. Our expression is more straightforward owing to the introduction
of the scaling factor x0.
and for p polarization (TM, Hy(x) component)
The integral factor in Eq. (9) depends on how smooth the index



2 2 2 2 2 transition from f(0) = 1 to f(x̃tp) = sin θinc is; in general the integral is
∂x Hy −∂x ln n ∂x Hy + n k0 −β Hy = 0; ð2Þ
negative and of order O(1). For quadratic profiles of the form n2(x) =
n2o(1 − x2/x20) the integral is approximately −π.
where ∂x is the partial derivative with respect to the x-coordinate; the
In contrast to [19,20] we are interested in the differential parallel
vacuum wave number is k0 = 2π/λ. We note that Eq. (2) is not
beam displacement (Dps = Dp − Ds). This is calculated from Artmann's
analytically solvable for a general refractive index profile n(x) (see
formula [21] and depends only on the derivative of the reflection
[16]).
phases ∂ϕp,s/∂θinc:
The TM (p) case can be rewritten into a form that resembles the TE (s)
case by making the substitution [17] Hy(x) = n(x)g(x), which results in ∂

Dps = −λ̄n0 ϕ −ϕs ð10Þ
h
 i ∂θinc p
2 2 2 2
∂x g + n k0 −β + δV ðxÞ g = 0 with ð3Þ
Substituting Eq. (9) into Eq. (10) leads to the main result
1 2 2 2
δV ðxÞ = ∂ n− 2 ð∂x nÞ : ð4Þ
n x n λ̄2n0
Dps ∼ : ð11Þ
x0
For sufficiently smooth GRIN media we can treat δV as a small
perturbation. This allows us to approximate the propagation constant
This highlights the scaling of the differential Goos–Hänchen shift
in the x-direction, kx,p, as
at a GRIN interface; it vanishes if the refractive-index transition region
  x0 is much larger than the (reduced) wavelength λ̄n0. The conven-
2 2 δV ðxÞ 2
kx;p = kx + δV ðxÞ≈ kx + : ð5Þ tional first-order (in λ̄n0) differential GH shift disappears in a graded-
2kx
index medium; from Eq. (11) we learn that only the second-order
differential GH shift survives.
Our problem is ideally suited to be solved using the WKB method
The analytic solution (Eqs. (9) and (10)) has three limitations: (i)
(see, e.g., [18]): we consider “smooth” index variations where the
close to the critical angle the finite spatial extent of the index gradient
refractive index changes only little over a wavelength. The path of the
becomes apparent, (ii) steep wavelength-scale gradients violate the
beam is treated ray-optically, but we keep track of the phase. The
slowly-varying potential requirement for the WKB method, and (iii)
acquired phase starting at x = 0 up to the turning point x = xtp and
the role of the exact shape of the refractive-index profile remains to be
back (see Fig. 1) is for s polarization
quantified. These cases require numerical calculation.

 The transmission matrix method [22] is ideally suited for numerical
x ′ ′
ϕs = 2∫ tp kx x dx ð6Þ calculation of the reflection phase for a 1D GRIN structure. The GRIN
0
medium is approximated as a multilayer structure with 10 nm/layer (at
while for p polarization: λ0 = 675 nm). We verified that this is sufficiently fine-grained. We obtain
the reflection phase depending on incident angle, and using Artmann's



 ′ formula (Eq. 10) the differential GH displacement Dps of p and s polarized
′ ′
δV x
ϕp = 2∫
xtp
kx;p x dx ≈ϕs + ∫
xtp
 ′  dx′ ð7Þ beams upon passage through the GRIN structure is determined. All
0 0 kx x calculations are for the glass (n =1.5) — air (n =1.0) transition.
The absolute GH shift at a GRIN interface has little meaning since The first index profile we discuss is a linear refractive index gradient
the reflection plane is undefined. However, the (interesting) physical n(x̃) = n0(1 − x̃) in the configuration shown in Fig. 1. The resulting
 W. Löffler et al. / Optics Communications 283 (2010) 3367–3370 3369

differential GH shift is plotted in Fig. 2 for different gradient thicknesses.

The largest shift (black curve) depicts the well-known case of an abrupt
interface. We focus first on angles of incidence away from the critical
angle. By introducing a gradual index change the differential GH shift is
strongly reduced, already for a 1 μm thick gradient, and more so for 2, 3
and 4 μm thick gradients. This reflects the scaling law of Eq. (11): by
increasing the gradient width the scaling factor λ̄n0/x0 → 0. At the critical
angle, the conventional differential GH shift (at a hard interface)
diverges, this remains true for the GRIN case. However, for θ N θcrit, the
shift is reduced more quickly for smoother interfaces than for abrupt
ones (Fig. 2).
Now we focus on the influence of the shape of the refractive-index
transition on the differential GH beam shift. To this end we calculated
the GH shift for different profiles: linear, complementary error
function (profile achieved by many diffusion-driven processes [23]),
and hyperbolic secant (ideal profile for graded-index fibers [8]). The
refractive-index difference is kept constant (0.5) and the profiles have
Fig. 3. Transmission matrix calculations of the differential beam shift for different
a comparable width of the index gradient (see inset Fig. 3).
refractive-index profiles: linear, complementary error function, and hyperbolic secant
From an experimental point of view, the differential GH shift (the black curve shows the case of an abrupt interface for comparison). The inset shows
curves for different profiles show a similar reduction (compared to the GRIN profiles.
the abrupt interface case, black curve in Fig. 3); this is well explained
by the overall scaling of the GHshift in GRIN media (Eq. (11)). More
precisely, the GH shift follows this scaling only on average; the details undisturbed to allow diffusion of the monomer TFPMA into the
of the GH shift Dps(θinc) are indeed highly sensitive to the GRIN profile partially cross-linked matrix of EBADA. Finally, the whole sample was
shape. This is because WKB scattering amplitudes are very sensitive to fully polymerized by UV exposure. The refractive indices of the single
the potential profile [24,25]. polymer layers have been determined to be 1.54 (EBADA) and 1.42
(TFPMA). This corresponds to a critical angle of 67.23°, which also
3. Experimental demonstration applies for the graded interface.
Collimated light from a superluminescent diode (λ = 677 nm,
For a first experimental demonstration we require samples with a Δλ= 9 nm) was incident on the sample in a θ− 2θ setup. This light
large refractive-index difference to (i) maximize the GH shift and to was modulated continuously between s and p polarization using a photo-
(ii) achieve an experimentally accessible angle of total internal reflec- elastic modulator. The differential positional shift of the reflected beams
tion. The system we have chosen is a graded polymer-blend made out for s and p polarization was detected using a silicon quadrant photodiode
of ethoxylated Bisphenol-A diacrylate (EBADA) and tetrafluoropro- as detector, together with a lock-in amplifier [26]; correction for Fresnel
pylmethacrylate (TFPMA) with refractive indices of 1.54 and 1.42, refraction at all interfaces has been done. A residual birefringence is
respectively and negligible absorption in the relevant spectral range. present in the high-index layer, this results in an offset in the differential
To induce polymerization by UV light the monomers were mixed with displacement signal. For clarity, we took care of this by vertically shifting
the photo-initiator IRGACURE 819 (EBADA with 1%, TFPMA with 2% by the experimental curve.
weight). The resulting differential GH shift close to the critical angle is
The demonstration sample was prepared directly on a face of an shown in Fig. 4: we see clear indication that the relative displacement
equilateral BK7 prism. Below a cover slip, an approximately 130 μm thick is strongly reduced for a GRIN interface compared to the hard inter-
film of the EBADA was deposited and (partially) polymerized initiated face case. The theoretically expected GH shift for a 2 μm GRIN is plot-
by a short UV exposure, with a dose below that for full cross-linking. ted for comparison.
Using the same technique, a layer of TFPMA was deposited on top
of this layer. During the diffusion time TD the system was kept

Fig. 4. Experimental GH shift for the polymer-blend GRIN sample in comparison with
Fig. 2. Transmission matrix calculation of the differential GH shift Dps as a function of the theoretical curve for a 2 μm thick gradient (n = 1.54...1.42). The experimental curve
incident angle. Shown are calculations for different gradient thicknesses, the inset is displaced vertically as explained in the text. The inset sketches the experimental
shows the profiles. configuration.
3370 W. Löffler et al. / Optics Communications 283 (2010) 3367–3370

4. Conclusions [4] W.J. Wild, C.L. Giles, Phys. Rev. A 25 (1982) 2099.
[5] J.B. Götte, A. Aiello, J.P. Woerdman, Opt. Express 16 (2008) 3961.
[6] H. Wolter, J. Opt. A-Pure Appl. Op 11 (2009) 090401 Translation of H. Wolter, Z.
In conclusion, smoothening of a refractive index gradient results in Naturf. A 5, 27683 (1950).
a reduction of the polarization differential Goos–Hänchen effect if the [7] L.A. Whitehead, M.A. Mossman, Opt. Photon. News 20 (2009) 28.
[8] A.W. Snyder, J.D. Love, Optical waveguide theory, Springer, 1983.
GRIN thickness is of the order of or larger than a wavelength. This can [9] S. Kozaki, Y. Mushiake, J. Appl. Phys. 46 (1975) 4098.
be seen as a manifestation of decoupling of polarization and spatial [10] V.S. Liberman, B.Y. Zeldovich, Phys. Rev. A 46 (1992) 5199.
degrees of freedom in inhomogeneous media, provided that the [11] K.Y. Bliokh, A. Niv, V. Kleiner, E. Hasman, Nat. Phot. 2 (2008) 748.
[12] A. Shvartsburg, V. Kuzmiak, G. Petite, Phys. Rep. 452 (2007) 33.
inhomogeneity λ̄ n0/x0 is small. The differential Goos–Hänchen effect [13] A. Shvartsburg, V. Kuzmiak, G. Petite, Phys. Rev. E 76 (2007) 016603.
in GRIN media is a second-order correction to geometric optics; the [14] U. Leonhardt, Science 312 (2006) 1777.
well-known first-order differential Goos–Hänchen shift disappears [15] J.B. Pendry, D. Schurig, D.R. Smith, Science 312 (2006) 1780.
[16] L.M. Brekhovskikh, Waves in Layered Media, Academic Press, 1965.
here. This generalizes the conventional Goos–Hänchen effect.
[17] A. Ghatak, L. Kraus, IEEE J. Quant. Elect. 10 (1974) 465.
This work was supported by the Netherlands Foundation for [18] D.J. Griffith, Introduction to Quantum Mechanics, Prentice Hall, 1994.
Fundamental Research of Matter (FOM) and by the Seventh Framework [19] V.S. Liberman, B.Y. Zeldovich, Phys. Rev. E 49 (1994) 2389.
Programme for Research of the European Commission, under the FET- [20] K. Bliokh, Y. Stepanovskii, J. Exp. Theor. Phys. 97 (2003) 479.
[21] K. Artmann, Ann. Phys. 437 (1948) 87.
Open grant agreement HIDEAS, no. FP7-ICT-221906. [22] R.M. Azzam, N.M. Bashara, Ellipsometry and Polarized Light, North-Holland, 1977,
1977.
References [23] R.A.L. Jones, R.W. Richards, Polymers at surfaces and interfaces, Cambridge
University Press, 1999.
[1] F. Goos, H. Hänchen, Ann. Phys. 436 (1947) 333. [24] M.V. Berry, K.E. Mount, Rep. Prog. Phys. 35 (1972) 315.
[2] D.-K. Qing, G. Chen, Opt. Lett. 29 (2004) 872. [25] M.V. Berry, J. Phys. A: Math. Gen. 15 (1982) 3693.
[3] M. Merano, A. Aiello, G.W. 't Hooft, M.P. van Exter, E.R. Eliel, J.P. Woerdman, Opt. [26] H. Gilles, S. Girard, J. Hamel, Opt. Lett. 27 (2002) 1421.
Express 15 (2007) 15928.