PLANAR POLARIZATION TRANSFORMER

Abstract

The planar polarization transformer for a normally incident optical wave or beam having a given vacuum wavelength λ comprises an optical planar grating between a cover medium of refractive index nc and an optical substrate of refractive index ns, this planar grating defining a binary index modulation or corrugation of substantially rectangular profile with periodic ridges. This polarization transformer is characterized in that the ridge refractive index nr is larger than the substrate refractive index ns, the grating period Λ is larger than 0.4·λ/nc, the substrate refractive index ns is smaller than 2.7·nc, and the index modulation or corrugation is designed such that, according to the grating mode dispersion equation, the effective index of the mode TE0 is larger than the substrate index ns and the effective index of the mode TM0 is larger than the cover refractive index and smaller than the substrate index.

Description

FIELD OF THE INVENTION

The invention concerns the transformation of the polarization distribution at the cross-section of a light beam into another polarization distribution by means of a planar zeroth order (0^th order) diffractive element, i.e. an optical element that diffracts the light only in the 0^th order (all higher order diffraction orders are below their cutoff and are evanescent in the incidence and transmission media).

The polarization state of laser beams is in most laser applications linear or circular. There is a growing interest in laser-matter interactions, in optical trapping of particles, in charged particle acceleration, in the generation of cylindrical waves, of axicons, etc. to have beams with a different distribution of the polarization over their cross-section, for instance the radial or the azimuthal polarization. Such polarized beams can be generated from the laser cavity directly by using a grating mirror that dictates the emission of the desired polarization distribution. This is however not always possible or desired. For instance, the beam emitted by edge emitting semiconductor lasers has a linear polarization, so have the femtosecond lasers beams resulting from an intra-cavity CPA amplification scheme. There is therefore a need for a compact, monolithic optical element capable of transforming the polarization distribution over the cross-section of an input beam into a desired polarization distribution.

BACKGROUND OF THE INVENTION

The modern solution for the above mentioned planar zeroth order diffractive element is an element that can locally rotate the local input linear polarization by the desired angle and does so by means of a spatially resolved half-wave birefringent element whose birefringence axis is oriented along the bisector of the angle between the input and output linear polarizations. This solution generally rests on the birefringence property of a 0^th order subwavelength binary corrugation grating structure defined at the surface of a high refractive index material.

More particularly, the structure of the state of the art consists of a binary, periodic grating of period Λ with essentially vertical walls wherein the grooves substantially have the refractive index of the cover material n_c (usually air) and the ridges are made of a high index material having the refractive index n_s of the substrate in which the grating is microstructured as a pure surface relief (or corrugation) grating. The period of the grating has to be small enough to avoid the generation of diffraction orders higher than the 0^th order in the cover and in the substrate. This imposes the condition (given that n_s is larger than n_c):

$\begin{array}{cc}\Lambda <\frac{\lambda }{n_s}& \left(1\right)\end{array}$

when the beam incidence is normal to the grating plane and where λ is the wavelength of the light in vacuum. The condition of equality between Λ and λ/n_s is named the cutoff condition for the +/−1^st orders under normal incidence, and condition (1) under which all diffraction orders except the 0^th order do not propagate in the substrate (they are “evanescent”) corresponds to these orders being “below cutoff”, which is equivalent to the grating being called “subwavelength”.

The polarization component of the incident beam having its electric field parallel to the local grating lines is called the TE polarization and the orthogonal polarization component having the electric field normal to the grating lines is the TM polarization, as shown in FIG. 1. The aim for a polarization transformer is to achieve a phase shift of π between the TE and TM polarizations and to have as low as possible losses due to reflection. The grating then acts as a spatially distributed half-wave element, locally mirroring the input electric field with regard to the local orientation of the grating lines. This mirroring can be seen as a rotation of the polarization, where the rotation angle α for a y-axis oriented polarization, after passing through a grating oriented at angle β with regard to the y-axis, is given by:

α=2β  (2)

In order to transform a linear polarization into a spatially non uniform polarization, the E-Field has to be turned by a specific angle β at each point within the beam cross section. As represented in FIG. 2, for the example of transforming a linear to a radial polarization, the required direction of the E-Field after the polarizer is equal to the polar angle Θ of the corresponding position in the beam in a polar coordinate system with its centre at the beam centre. The corresponding grating pattern is sketched in FIG. 2. Note that by turning the element by 90°, one can produce azimuthally polarized light from the same incident beam as also represented in this FIG. 2.

The grating has to fulfill the condition β=Θ/2 everywhere over the beam cross-section. This is not possible to achieve with continuous grating lines without accepting a change in the period (as shown in FIG. 2). If this period change is too large regarding the period tolerance of the grating with respect to its phaseshifting property, a suitable segmentation of the element must be made as represented in FIG. 3. The grating design of FIG. 3 geometrically corresponds to the one of FIG. 2 so that this segmented grating transforms a linear polarization into a radial polarization or into an azimuthal polarization as shown in FIG. 2. The lateral arrangement of such grating lines and the phase functions created by such an element, are known in the scientific literature (e.g. Bomzon, Z.; Biener, G.; Kleiner, V. & Hasman, E., “Radially and azimuthally polarized beams generated by space-variant dielectric subwavelength gratings,” Opt. Lett. 27, 285-287 (2002)).

In the solution of the state of the art for the grating in a polarization rotating element (described by Lerman G., Levy U., Optics Letters 33, 2782-2784 (2008)), the grating is a pure surface relief grating, i.e. the ridge material is the same as the grating material. It can be shown that in this described solution the 0^th order grating mode only has a propagating character, while higher order grating modes are evanescent (chapter 5 in Tishchenko, A., “Phenomenological representation of deep and high contrast lamellar gratings by means of the modal method,” Opt. Quantum. Electron., Springer 37, 309-330 (2005)). In such pure relief grating the fundamental TE₀ and TM₀ modes always have an effective index which is located between the low index n_c of the cover and the high index n_s of the substrate which implies that either the TE or the TM polarization exhibits notably less than 100% transmission. This will be clearly established later on in the description of the invention.

The requirement of zero diffraction loss in the substrate of high refractive index imposes a corrugation period Λ smaller than λ/n_s (see Equation (1)) which is necessarily very small since the condition for a high TE/TM birefringence calls for a large index n_s so as to require grooves that are not too deep and too difficult to fabricate. This can be understood from the effective medium theory used in the state of the art: it is known that with so small a period (much smaller than the wavelength in vacuum) the hypothesis of effective medium for the corrugated layer holds; accordingly, the grating modes of order higher than 0^th, which are evanescent, are not taken into consideration and the grating layer is treated as a homogeneous layer characterized by an “effective index”. This homogenized grating layer between the cover medium and the substrate medium thus defines a Fabry-Perot filter for each linear polarization component of the incident light beam. In the effective medium theory, the effective index of the homogenized grating layer “seen” by the linear polarization components of the light beam is given by:

$\begin{array}{cc}{n}_{\mathrm{eff}}^{\mathrm{TE}}=\sqrt{{\mathrm{Dn}}_s^2+\left(1-D\right)n_c^2},n_{\mathrm{eff}}^{\mathrm{TM}}=1/\sqrt{\left(\frac{D}{n_s^2}+\frac{\left(1-D\right)}{n_c^2}\right)}& \left(3\right)\end{array}$

where D is the duty cycle of the grating, calculated as D=W_r/Λ with W_r being the width of the ridges of the grating as shown in FIG. 4A. It is worth noting here that these expressions coincide with the effective index of the 0^th order TE₀ and TM₀ grating modes as calculated by the exact modal theory for an infinitely small period. In FIG. 4B, the curves of the effective indexes of the 0^th order grating modes n_eff^TE and n_eff^TM are given in function of D for three cases: n_s=1.5, n_s=2.5 and n_s=3.5. It can be seen that the birefringence effect strongly decreases with a decrease of the substrate refractive index n_s; this in turn implies an increase of the grating height needed to accumulate a phase shift of π between the TE₀ and TM₀ grating modes upon their propagation through the grating corrugation. This is why a high refractive index is selected for the substrate in which the grating is microstructured.

One of the consequences of the period having to be much smaller than the wavelength as required in the solution of the state of the art (ref. to Equation (1)) is that the field transmission and reflection coefficients of the modes at the top and bottom interfaces are essentially real, i.e., their phase is either close to 0 or close to π as implicitly assumed in the effective medium theory. This has the adverse consequence that achieving in a shallow (i.e., fabricable) grating corrugation the three needed conditions of TE and TM Fabry-Perot filters being set at the maximum resonant transmission peak with minimal grating height (i.e. minimal amount of 2π multiples of the phase for one roundtrip of the light in the Fabry Perot) and the transmitted TE and TM fields being π out of phase is in principle impossible for the refractive indexes available in nature at optical frequencies (see the explanation of this in the description of the invention). The solution of the state of the art is therefore a compromise which has to be searched for numerically, for instance with an air cover and a GaAs substrate (n_s=3.48) a TE and TM transmission of 86% and 96% respectively at λ=1064 nm with a period as small as 240 nm, i.e., smaller than a quarter of the wavelength, a ridge width of 144 nm and a groove depth of 470 nm (Lerman G., Levy U., Optics Letters 33, 2782-2784 (2008)), i.e., an aspect ratio of 3.2 for the ridges and 4.9 for the grooves, which is extremely difficult to fabricate.

It is to be noted that the reduced overall transmission, caused by the in-principle impossibility to fulfill the conditions for the π-phaseshifted 100% transmissions of the TE and TM Fabry-Perots, can be partly circumvented by depositing an antireflection layer at the top of the grating lines and at the bottom of the grooves as disclosed in Dimitri Mawet, Pierre Riaud, Jean Surdej, and Jacques Baudrand, “Subwavelength surface-relief gratings for stellar coronagraphy,” Appl. Opt. 44, pp. 7313-7321 (2005). This however adds to the fabrication difficulty, and it does not change the requirement of the critical depth and width of the corrugation.

Thus, the grating elements of the prior art used as polarization transformers face at least two problems. First, they are difficult to manufacture because of the small grating period and of the grating profile with a large groove aspect ratio (groove height divided by groove width). Secondly, their power transmission is notably smaller than 100%. This will be better understood in the following description of the invention.

In order to suppress the reflection from a pure relief grating of the prior art represented by the Optics Letter article by G. Lerman et al., document US 2008/130110 A1 discloses an additional 2D subwavelength grating on top of the 1D birefringent polarizing grating, made of a different material as the etched substrate whose function is to act as an O^th order antireflection micro-structured layer at the air side of the form-birefringent grating where the reflection is the larger. This measure does not change the requirement on the critical depth and width of the grooves and requires a second set of layer deposition, lithography and etching processes with critical alignment between 1D and 2D corrugations.

SUMMARY OF THE INVENTION

For a subwavelength grating (i.e. in a grating fulfilling the cutoff condition Λ<λ/n_s), the 0^th diffraction order only has a propagating character in the substrate and the cover media, whereas all higher orders are evanescent. This condition closely (but not exactly) corresponds to the fundamental grating modes TE₀ and TM₀ having an up and down propagating character whereas the grating modes of higher order have an evanescent character along the walls and grooves of the corrugation. Further, in the case of a relatively small period as in the prior art, i.e. Λ well below the wavelength, e.g. Λ<λ/4, those higher order grating modes can then be neglected in the analysis, and do not contribute notably in the grating transmission, given that for usual grating heights needed for a π phase shift between TE and TM polarizations their amplitude has decreased sufficiently at the output interface of the grating corrugation because of their strong exponential decay. Thus, the process of light interaction with such a grating of relatively small period can be considered as analogous to light interaction with a homogeneous layer forming a Fabry-Perot cavity. Therefore, the known Fabry-Perot resonance properties can be used to understand and design the transmission and reflection spectra of subwavelength gratings for both polarizations separately. The analogy with a homogeneous medium also approximately holds for the field reflection and transmission coefficients r and t respectively at the interfaces which are given by the well known Fresnel formulae at the interface between two homogeneous media of refractive index n₁ and n₂:

$\begin{array}{cc}r=\frac{n_1-n_2}{n_1+n_2};t=\frac{2n_1}{n_1+n_2}& \left(4\right)\end{array}$

where, in the case of a grating, one refractive index has to be replaced by the effective index of the grating mode of each given polarization. The phase of the coefficients is also close to that of the homogenous media case (i.e. 0 for transmission and 0 or π for reflection from a lower or higher index material, respectively). It is worth pointing out here that the present invention places itself in conditions where the above hypothesis on the phase of the field reflection and transmission coefficients does not hold, and it exploits the very deviation from this hypothesis to break the in-principle impossibility of the solution of the state of the art, to render the set of three conditions non-contradictory, and to satisfy the latter in a shallow, easily fabricable grating corrugation.

The Fabry-Perot concept can however still be used to approximately give the maximal achievable resonant power transmission T_max, in particular for the polarization transformers of the prior art for which this concept is a close to exact approximation. This maximal power transmission T_max of a Fabry Perot with field reflectivities r_c and r_s at the cover and substrate interface, respectively, is given by:

$\begin{array}{cc}{T}_{\max }=\frac{\left(1-r_c\right)\left(1-r_s\right)}{{\left(1-\sqrt{r_cr_s}\right)}^2}& \left(5\right)\end{array}$

which only reaches 100% in the case of r_c=r_s (called “symmetric” Fabry-Perot). Combining expressions (4) and equation (5) results in the maximal transmission under resonance which can be achieved in a Fabry-Perot made up of three different dielectric materials:

$\begin{array}{cc}{T}_{\max }=\{\begin{array}{cc}4\frac{n_cn_s}{{\left(n_{\mathrm{eff}}+n_cn_s/n_{\mathrm{eff}}\right)}^2}& \mathrm{for}n_{\mathrm{eff}}\mathrm{between}n_c\mathrm{and}n_s\\ 4\frac{n_cn_s}{{\left(n_c+n_s\right)}^2}& \mathrm{for}n_{\mathrm{eff}}\mathrm{larger}\mathrm{or}\mathrm{smaller}\mathrm{than}n_c\mathrm{and}n_s\end{array}& \left(6\right)\end{array}$

The mathematical expressions (6) reveal that there is a fundamental difference between two types of structures. On the one hand, if the effective index is between the cover and substrate index, a maximal power transmission of 100% can be obtained only for n_eff=(n_c·n_s)^1/2. This implies that, for the devices of the state of the art for which expressions (6) are essentially valid, either the TE₀ or the TM₀ mode represents an asymmetric Fabry-Perot exhibiting notably less than 100% power transmission: only one of the polarizations can fulfill the condition of equal reflectivities and thus exhibit maximum transmission of 100% by having an effective index equal to (n_cn_s)^1/2, and thus representing a symmetrical Fabry-Perot; the other polarization necessarily forms a Fabry-Perot with different reflectivities at the two interfaces exhibiting a notably lower maximal transmission according to the first expression of (6). This is thus an inherent problem of the polarization transformers of the prior art since n_eff^TE0 and n_eff^TM0 for the zeroth order TE₀ and TM₀ modes must be different to give rise to the needed polarization phase difference. Thus, it is not possible for both polarizations to have a maximum transmission. This affects the quality and efficiency of the polarization transformation. On the other hand, if the effective index is larger or smaller than both substrate and cover index, the maximal transmission, as given by the second expression of (6), does not depend on the effective index in the Fabry-Perot model. For usual substrate materials it is rather large and close to 100% (e.g. 96% for n_s=1.5 and an air cover).

The present invention places itself in the condition of the second expression of (6) for the TE-polarization, with a low index substrate material permitting the TE-polarization to “see” a non-symmetrical Fabry-Perot having a larger effective index of the grating region and thus exhibiting a maximum power transmission close to 100%. Furthermore, the present invention places itself at the same time in a condition where in the TM-polarization the difference between the TM₀ effective index and the substrate refractive index as well as the cover refractive index is very low, resulting also in a Fabry-Perot power transmission close to 100% whatever the Fabry-Perot height.

The present invention, after having identified the inherent problems of the devices of the prior art and the theoretical limitation of such devices by analyzing the physical behavior of a grating as a Fabry-Perot for the zero^th order grating modes, discloses a device that overcomes these limitations, is easier to fabricate and performs as a high efficiency, high quality polarization transformer transforming a beam of linear polarization into a beam of radial and azimuthal polarization.

Thus, the present invention concerns a planar polarization transformer for a normally incident optical wave or beam having a given vacuum wavelength λ, comprising an optical planar grating between a cover medium of refractive index n_c and an optical substrate of refractive index n_s larger than the cover refractive index, this planar grating defining a binary index modulation or corrugation of substantially rectangular profile with periodic ridges, having a ridge refractive index n_r and a ridge width W_r, and interleaved grooves having a groove width W_g and groove refractive index n_g equal to the cover refractive index n_c. The planar grating locally has a grating period Λ which is smaller than the wavelength of the incident optical wave divided by the substrate refractive index (Λ<λ/n_s). The polarization transformer is characterized in that the ridge refractive index n_r is larger than the substrate refractive index n_s, in that the grating period Λ is larger than 0.4·λ/n_c, in that the substrate refractive index n_s is smaller than 2.7n_c, and in that the index modulation or corrugation is designed such that, according to the grating mode dispersion equation, the effective index of the TE fundamental grating mode TE₀ is larger than the substrate index n_s and the effective index of the TM fundamental grating mode TM₀ is larger than the cover refractive index n_c and smaller than the substrate index n_s.

In a particular embodiment, the substrate is composed of high thermal conductivity MgF₂ or sapphire or YAG. In a first variant of the invention, the cover and the grooves are filled with air or vacuum. In another variant, the cover and the grooves are filled with a transparent polymer or with a transparent liquid.

The invention also concerned a design method for a planar polarization transformer comprising an optical planar grating between a cover medium of refractive index n_c and an optical substrate of refractive index n_s larger than the cover refractive index, this planar grating defining a binary index modulation or corrugation of substantially rectangular profile with periodic ridges, having a ridge refractive index n_r and a ridge width W and interleaved grooves having a groove width W_g and groove refractive index n_g equal to the cover refractive index n_c, this planar grating locally having a grating period Λ which is smaller than the wavelength λ of an incident optical wave or beam divided by the substrate refractive index, said ridge refractive index n_r being larger than the substrate refractive index n_s, this design method comprising the following step:

• • Choosing a wavelength λ;
  • Choosing a period Λ in the range 0.4·λ/n_c<Λ<λ/n_s;
  • Determining the duty cycle D=W_r/Λ from the TM₀ dispersion equation of the modal theory to substantially realize n_eff^TM0=(n_c·n_s)^1/2 n_eff^TM0 being the effective index of the fundamental TM mode (TM₀) of the planar grating;
  • Determining n_eff^TE0 from the TE dispersion equation of the modal theory for the chosen period and wavelength and the determined duty cycle D, n_eff^TE0 being the effective index of the fundamental TE mode (TE₀) of the planar grating;
  • Calculation of the field reflection and transmission phases φ_r^TE and φ_t^TE of the TE₀ mode by means of the modal theory;
  • Calculation of a first height H^FP from the expression

$H^{\mathrm{FP}}=\frac{\lambda }{2\pi n_{\mathrm{eff}}^{\mathrm{TE}}}\left(\pi -1/2{\varphi }_r^{\mathrm{TE}}\right)$

• • Calculation of a second height H^dif from the expression

$H^{\mathrm{dif}}=\frac{\lambda }{2\pi \left(n_{\mathrm{eff}}^{\mathrm{TE}}-n_{\mathrm{eff}}^{\mathrm{TM}}\right)}\left(\pi -{\varphi }_t^{\mathrm{TE}}\right)$

• • Calculation of the height average H=(H^FP+H^dif)/2 defining the grating height;

According to an improvement of the design method given above, this method further comprises an optimization step in which a fine tuning of geometrical parameters (Λ, D and H) of the planar grating with an appropriate numerical optimization tool by taking as departure point the corresponding geometrical parameters obtained in the design method given above. In a particular variant, the numerical optimization tool defines a gradient search method using a merit function corresponding to the transmission of the incident optical wave or beam through two crossed polarizers with the grating arranged between them and oriented at 45° relative to each polarizer direction.

BRIEF DESCRIPTION OF THE FIGURES

The present invention will be described subsequently in more detail with reference to the attached drawing, given by way of examples, but in no way limited thereto, in which:

FIG. 1 illustrates the principle of linear polarization rotation by means of a subwavelength grating;

FIG. 2 is a top view illustrating the direction of the lines of a grating transforming an input linear polarization into a radial or azimuthal polarization;

FIG. 3 is a top view of a segmented grating exhibiting a substantially constant period while substantially achieving the polarization transformation of the grating represented in FIG. 2;

FIG. 4A partially shows a polarization transforming grating of the state of the art;

FIG. 4B is a graph showing the TM₀ and TE₀ grating mode effective index dependence versus the duty cycle of a polarization transforming grating of the state of the art for three different materials forming this grating;

FIG. 5A is the cross-section of a first embodiment of a polarization transforming grating according to the invention;

FIG. 5B is the cross-section of a second embodiment of a polarization transforming grating according to the invention with a thick layer optical substrate on mechanical support;

FIG. 6 is a graph showing the TE and TM power transmission and the TE and TM field transmission phase of the Fabry-Perot for the TE₀ and TM₀ mode (using n_s=1.44, n_c=1.0, n_eff^TM0=1.2 and n_eff^TE0=2.0) versus the height H of the grating;

FIG. 7 is an explanatory sketch defining the amplitudes and phases of the field transmission and reflection coefficients considered for defining the geometrical parameters of the polarization transforming grating of the invention;

FIG. 8 is a graph showing the TM₀ and TE₀ grating mode effective index dependence versus the duty cycle D=W_r/Λ of a polarization transforming grating of the invention;

FIG. 9 is a graph showing the transmission through crossed linear polarizers versus the ridge index n_r and substrate index n_s for the optimized polarization transformer gratings of the invention, and a selected area of validity for the polarization transforming gratings of the invention in the n_r-n_s space;

FIG. 10 shows the relative period P_rel=Λ/λ of the optimized polarization transforming gratings versus the refractive indexes of ridge n_r and substrate n_s in the selected area of FIG. 9;

FIG. 11 shows the duty cycle D=W_r/Λ of the optimized polarization transforming gratings versus the refractive indexes of ridge n_r and substrate n_s in the selected area of FIG. 9;

FIG. 12 shows the relative height H_rel=H/λ of the optimized polarization transforming gratings versus the refractive indexes of ridge n_r and substrate n_s in the selected area of FIG. 9;

FIG. 13 shows the period normalized to the diffraction order cutoff period P_cop=Λ/(λ/n_s) of the optimized polarization transforming gratings versus the refractive indexes of ridge n_r and substrate n_s in the selected area of FIG. 9;

FIG. 14 shows the groove aspect ratio A_g=H/(Λ−W_r) of the optimized polarization transforming gratings versus the refractive indexes of ridge n_r and substrate n_s in the selected area of FIG. 9.

DETAILED DESCRIPTION OF THE INVENTION

A first embodiment of the invention is partially represented in cross-section in FIG. 5A. The polarization transformer 2 is intended for a light beam or light wave of a given wavelength λ in the optical domain arbitrarily defined as extending from 193 nm up to 20 micrometer vacuum wavelength. This polarization transformer is formed by a planar structure comprising a substrate 4 having refractive index n_s and a grating 6 arranged on this substrate. The incidence of the light wave is substantially normal to the planar structure. It can however be off normal (typically up to 5 degrees), as long as no orders higher than the 0^th order appear in the substrate. The optical planar grating 6 defines a grating region (also named grating layer or grating medium or grating zone) and is formed by ridges 8 of width W_r and grooves 10 of width W_g forming a corrugation of substantially rectangular profile. This planar grating is located between a cover medium, having a refractive index n_c smaller (or lower) than or equal to the refractive index of the substrate, and the substrate. According to the invention, the ridges 8 have a refractive index n_r which is larger (or higher) than the refractive index n_s of the substrate. The grooves have a refractive index n_g smaller than the index of the ridges. In a preferred variant, the grooves are filled with the same medium as the cover medium, as air for instance. The planar grating has locally a grating period Λ which is smaller than the cutoff period, i.e. smaller than the wavelength of the incident light beam or wave divided by the refractive index n_s of the substrate. The cutoff condition limits the diffraction of the light beam or wave by the grating into the substrate and the cover media to the sole zeroth (0^th) reflected and transmitted diffraction orders. Further, the refractive index n_r of the ridges, that of the grooves n_g, the period/wavelength ratio Λ/λ and the duty cycle D=W_r/Λ are selected, and the grating is designed so that the effective index of the fundamental TE₀ modes of the grating region, given by the dispersion equation of the modal theory [e.g., ref Tishchenko, A., “Phenomenological representation of deep and high contrast lamellar gratings by means of the modal method,” Opt. Quantum. Electron., Springer 37, 309-330 (2005)), Equations (17) and (20)], is larger than the refractive index n_s of the substrate.

FIG. 5B shows a second embodiment of a polarization transformer 12 according to the invention wherein the transparent substrate is formed by a thick layer 4A of low index material like ThF₄ disposed on top of a mechanical substrate (support) of any transparent material, for instance ZnSe, having a larger index n_s*(in the case of ZnSe, the index value n_s*=2.403 at λ=10.6 μm). The height of the thick layer is chosen so that the optical substrate actually “seen” by the light is essentially ThF₄ with n_s=1.32. Thus, the low index optical substrate is thick enough to preclude the tunneling of the +/−1^st transmitted diffraction orders into the high index support through the low index optically thick substrate layer.

The evanescent +1^st and −1^st diffraction orders in the low index optical substrate of index n_s have an exponential decay in the direction normal to the plane of the element. The factor in the exponential is −z((λ/Λ)²−n_s²)^1/2)2π/λ where z is the coordinate along the normal with origin at the basis of the grating corrugation. Preventing these diffraction orders from tunneling into the high index substrate through the low index optical substrate amounts to ensuring a field amplitude at the interface between the high index substrate and the low index optical substrate that is ∈ times the field amplitude at the basis of the grating corrugation (at z=0), where ∈ is a factor much smaller than 1; for instance equal to 0.1. Consequently, the thickness t_ob of the low index optical substrate is given by: t_ob>−ln(∈)λ/(2π((λ/Λ)²−n_s²)^1/2). A concrete value for t_os will be given in the 3^rd example given hereunder.

The use of a low index substrates softens the condition on the grating period for having no propagating orders in the substrate (Λ<λ/n_s) and thus prevents diffraction losses in the substrate even for larger periods, compared to those used in the state of the art. The low index is achieved by using materials like CaF₂ (n_s=1.51 at 193 nm) or MgF₂, SiO₂ (about 1.46 in the visible and near-IR range), ThF₄ (n_s=1.32 in the far IR, in particular at 10.6 μm wavelength) for instance. The choice of the substrate can be imposed by thermal conditions if the incident beam or wave has a large power. In this case it is preferable to use a MgF₂ or sapphire or YAG substrate instead of a fused quartz substrate because of their large thermal conductivity, and a heat sink of adequate geometry. The corrugated grating layer is of high index against the cover and groove index to give rise to a large birefringence between the TE₀ and TM₀ grating modes, and thus to limit the needed groove depth to accumulate the transmission phaseshift of π between the TE and TM Fabry-Perot field transmissions. This high index material can e.g. be amorphous silicon as from 0.7 μm wavelength (where it becomes transparent) upwards, single crystal silicon of the SOI technology, germanium, GaAs, InP, or GaN at wavelengths larger than that corresponding to their gap for instance.

According to the invention, for fixed refractive indexes and period/wavelength ratio, the duty cycle D of the corrugation is first selected so as to define an effective index of the TM₀ mode equal to (n_c·n_s)^1/2 in order to have the maximal power transmission through the polarization transformer according to the result given by the first expression of (6) in the summary of the invention. This is made by solving the dispersion equation (equation (20) of Tishchenko's reference given above) to create a plot of the effective index of TM₀ over the duty cycle D, which is presented in FIG. 8 for an example with n_s=1.44, n_r=3.7 and Λ=0.8λ/n_s, and subsequently choosing a suitable duty cycle to fulfill the first of expressions (6). So small a TM₀ effective index is obtained with a duty cycle smaller than 0.25, i.e. the grooves are wider than the ridges. This is an important advantage over the state of the art because it is technologically much easier to photolithographically transfer and to dry-etch grooves that are wider than the ridges than the opposite. The assumption that the 0^th mode only is responsible for the grating transmission is not as accurate as at very small periods since the decay of the next higher order TE grating mode field, which is evanescent, can penetrate deep enough into the grating layer, and its energy tunnels through it into the substrate. To avoid this situation where the grating transmission becomes unpredictable the condition is stated that the field of the TE₂ mode at the bottom of the corrugation is less than 20% of its amplitude at the air-corrugation interface. The model of considering the 0^th order only holds well enough to serve as an explanatory model of the device of the invention as well as to approximately define its operation conditions.

Having assigned to D the value giving the TM₀ mode the effective index n_eff^TM0=(n_cn_s)^1/2, the TM Fabry-Perot is symmetric and consequently has a maximum power transmission of 100%, but contrary to the devices of the state of the art, this effective index is very low (typically 1.22 with an air cover and a SiO₂ substrate), and close to the substrate as well as the cover index, so that the reflectivity for both interfaces of the respective Fabry Perot is very small (see Expressions (4)), which implies that close to 100% power transmission is achieved whatever the depth H (height) of the corrugation as shown in FIG. 6 where the typical behavior of a Fabry-Perot is represented versus H as an example wherein the effective index of the TM grating mode of zero^th order n_eff^TM0=1.22 and the effective index of the TE mode of zero^th order n_eff^TE0=2.0. Thus, in the device of the invention the requirement of close to 100% TM power transmission does not place any condition on the height H of the grating corrugation. The latter can consequently be freely chosen to adjust the resonance of the TE Fabry-Perot and the π-phase difference between the TE and TM field transmissions. It is important to point out here that n_eff^TM0 is so low that the next order TM₂ grating mode is strongly evanescent and practically does not affect the phase and amplitude of the TM field transmission coefficients across the two interfaces of the corrugated region with the cover and the substrate. On the contrary, as FIG. 6 shows, the power transmission of a Fabry-Perot with a layer having a refractive index value=2.0, corresponding to the TE polarization, is significantly dependent on the grating height H. In this last case, the power transmission varies approximately between 80% and 96%. Thus, the grating height is a critical parameter for the power transmission of the TE polarization component and has to be selected so as to be in the region of a maximum of the TE power transmission curve. FIG. 6 also shows the phase of the field transmission of the Fabry-Perot for both refractive indexes, i.e., for both polarizations. The phase dependence of both polarizations on the height H is an important feature of the polarization transformer since it is their phase difference (which must here be π) as much as their amplitude difference that determines the transmitted polarization state. The phase of the TM field transmission across the TM Fabry-Perot is quasi-linearly dependent on the height H of the grating and the TE transmission phase is only weakly modulated around a straight line. It will be shown hereunder that this property remains substantially true in the device of the invention where the grating period is close to the wavelength, with the consequence that both TM and TE transmission phases can be approximated by a straight through transmission (linear dependency on grating height).

In the device of the invention, the TE₀ effective index is larger than the substrate index n_s, thus according to the second of expressions (6) the maximal transmission does not depend on the absolute value of the refractive index n_r of the high index grating layer and of the effective index of this grating layer, but only on n_s and n_c, which in the example of a fused silica substrate and air cover gives a maximum power transmission of approx. 96% for the TE Fabry-Perot.

The quasi-100% TM Fabry-Perot transmission condition being fulfilled whatever H, two conditions remain to be fulfilled: first, one roundtrip of the TE₀ grating mode must result in a phaseshift of m_FP·2π to give constructive interference in the TE Fabry-Perot field transmission and, secondly, the phase difference between the transmission for TE and TM polarizations must amount to (2m_dif−1)·π to give the desired phase shift between TE and TM field transmissions (m_FP and m_dif are integer numbers). Since the grating should be as shallow as possible to ease the fabrication, and since the phaseshift is substantially proportional to the height of the grating, the integer numbers m_FP and m_dif should be as small as possible, i.e. equal to 1.

FIG. 7 defines the field transmission and reflection coefficients, as well as the phase values for the transmission through the grating layer (φ_eff), through the whole grating (φ_straight) and for a Fabry-Perot roundtrip (φ_FP).

At this point, the very fact of using a period larger than in the state of the art permits to make use of a not yet explicitly disclosed property of a relatively large period grating for obtaining a solution with a Fabry-Perot roundtrip of 2π and a TE-TM phase difference of π (i.e., m_FP and m_dif having the smallest value which their definition permits, i.e., 1). Whereas a relatively large period might dangerously lead to the excitation of a spurious higher order propagating grating mode or, in the present case of a weakly evanescent mode, there is a modal operation range where there is still only one propagating mode, the fundamental TE₀ mode (the TM₀ case being already settled), but where the excitation of the weakly evanescent higher order mode(s) by the incident wave affects the phase of the field reflection and transmission coefficients of the fundamental TE₀ mode at this interface. The larger the decay length of the higher order evanescent mode, the longer the life time of the photons coupled into the latter, the stronger the effect of the latter on the TE₀ reflection and transmission coefficients, mainly on their phase. Under such conditions of relatively large period the phase of these coefficients can be widely different from 0 and π and this difference can be explicitly used to intervene in the transmission phaseshift condition between TE and TM polarization components, and in the TE Fabry-Perot roundtrip condition, resulting in the possibility of satisfying them both with the smallest possible combination of π-phaseshift and 2π-roundtrip phase multiples. This will be explained hereunder.

To this end let us define the total phase of the TE₀ field reflection coefficients (φ_r) resulting from a round trip in the Fabry-Perot, and of the transmission coefficients (φ_t) resulting from crossing the Fabry-Perot from the cover to the substrate:

φ_r^TE=arg(r_00,c^TE)+arg(r_00,s^TE);φ_t^TE:=arg(t_c,0^TE)+arg(t_0,s^TE)  (8)

and for TM₀:

φ_TM=arg(r_00,c^TM)+arg(r_00,s^TM);φ_t^TM:=arg(t_c,0^TM)+arg(t_0,s^TM)  (9)

where r_00,c, r_00,s, t_c,0 and t_0,s are the field reflection and transmission coefficients at the cover-grating and grating-substrate interfaces (see FIG. 7 for their definition). These coefficients are calculated by using the modal theory (Appendixes A and B of Tishchenko, A., “Phenomenological representation of deep and high contrast lamellar gratings by means of the modal method,” Opt. Quantum. Electron., Springer 37, 309-330 (2005)).

Furthermore, the phase-shift φ_eff resulting from the 0^th order mode propagation through the grating region with its effective index n_eff is given by:

$\begin{array}{cc}{\varphi }_{\mathrm{eff}}^{\mathrm{TE}}=\frac{2\pi }{\lambda }n_{\mathrm{eff}}^{{\mathrm{TE}}_0}H;{\varphi }_{\mathrm{eff}}^{\mathrm{TM}}=\frac{2\pi }{\lambda }n_{\mathrm{eff}}^{{\mathrm{TM}}_0}H& \left(10\right)\end{array}$

which leads to the phase of the complete transmission, called “straight through” phase (FIG. 7), as:

φ_straight^TE=φ_t^TE+φ_eff^TM;φ_straight^TM=φ_t^TM+φ_eff^TM  (11)

The phase of the transmission through the grating should include the contribution of the multiple Fabry-Perot round trip reflections. However, the finesse of such Fabry-Perot is so small that considering just one round trip is a sufficient approximation.

Following the rationale described above, the two conditions to be fulfilled are the Fabry-Perot roundtrip for the TE polarization:

φ_FP^TE=m_FP2π  (12)

and the phase difference between TE and TM upon a single path through the Fabry-Perot:

Δφ=φ_straight^TM−φ_straight^TM=(2m_dif−1)π  (13)

This in itself already represents a first major step ahead of the state of the art solution since the close to 100% condition for the TM power transmission is satisfied regardless of the above two remaining conditions. At this point it is interesting to write conditions (12) and (13) in the case of a period/wavelength ratio notably smaller than 1 (state of the art situation) and with an effective index of TE₀ and TM₀ located between substrate and cover index (as is the case of the state of the art), which means that the phase of the TE field transmission coefficients are close to 0, whereas the field reflection coefficients are 0 and π at the cover and substrate interface, respectively: it would in this case be impossible to satisfy conditions (12) and (13) with the lowest possible combination of m_FP=1 and m_dif=1 since this would imply that the phase of the TM single path phase φ_straight^TM is zero which is impossible because the grating region has a non-zero thickness. With m_FP>1 the depth H of the grating corrugation becomes large and difficult to fabricate. This is precisely where the choice of a larger period characteristic of the present invention reveals to permit a breakthrough: not only can a large period grating be more easily fabricated than a short period grating, but, in addition and most importantly, a larger period leads to reflection and transmission phase shifts that deviate notably from 0 and π and enables the satisfaction of conditions (12) and (13) with m_FP=1 and m_dif=1 which ensures the operation of the polarization transformer with the shallowest possible grating as described hereafter.

Let us first consider the reflection coefficients of the 0^th order grating mode at the grating-cover and at the grating-substrate interfaces within binary gratings with period close to the cutoff of the grating mode next to the fundamental mode the decay of the evanescent field of the higher order evanescent grating modes and higher order evanescent diffraction orders is slower than in short period gratings. These evanescent modes and diffraction orders do not carry energy, but they act as a temporary energy reservoir. The 0^th grating mode, impinging on one of the interfaces, couples to these evanescent modes and diffraction orders and transfers some energy into them. This energy is returned back to the 0^th grating mode with a phase retardation, i.e. a negative phase shift. This is analogous to the known Goos-Hänchen shift occurring under total reflection at the interface between two dielectric media. Thus, the reflection coefficients always have a negative phase instead of close to zero as in short period gratings.

No general prediction about the sign of the phase of the field transmission coefficients is possible. However, repeated rigorous calculations show that in the TE case they are usually positive and only in some extreme cases (extremely small duty cycle, very large substrate index) become slightly negative. In the TM case, as already underlined previously, the field transmission coefficient phases are extremely small in the area of interest and can be set to zero, since the effective index is very small.

This general behavior of the phases of the field reflection and transmission coefficients does permit the satisfaction of conditions (12) and (13) with the smallest possible integer numbers m_FP=1 and m_dif=1: since the sign of the TE field reflection coefficients is negative, condition (12) is satisfied with m_FP=1 provided the TE Fabry-Perot thickness is larger than a half-wave thickness. Condition (13) can therefore in turn be in principle satisfied with m_dif=1 since the first term is larger than π (all the more so as the phase of the TE field transmission coefficients is positive) and since the straight through TM phase φ_straight^TM is smaller than φ_straight^TE (because the phase of the TM transmission coefficients is close to zero and because the TM₀ mode effective index is notably smaller than the TE₀ mode effective index). Thus, given that the straight through transmission for TE has a phase of π+x with x being positive, a straight through phase for TM of x results in a TE/TM phase shift of π.

Hereafter, an explicit design method for the novel polarization transforming grating is disclosed. It is assumed that n_s, n_c and n_r are given. The wavelength needs not be given because all three geometrical parameters Λ, W_r and H are scalable relative to the wavelength λ as the material dispersion is neglected.

First, the TM-effective index is chosen close to the value realizing a symmetric Fabry-Perot for the TM polarization:

n _eff ^TM=√{square root over (n_cn_s)}  (14)

Expression (14) imposes through the TM grating mode dispersion equation a condition on the period Λ and the duty cycle D. The period is preliminarily fixed at a value according to the rationale of the invention, e.g. Λ=0.8 λ/n_s. Substituting this period into the TM dispersion equation straightforwardly delivers the related duty cycle D. Introducing then these Λ and D values into the TE grating mode dispersion equation straightforwardly delivers the corresponding TE₀ effective index (FIG. 8). The modal method is now used to calculate the reflection and transmission coefficients required in Equation (8) to calculate φ_r^TE and φ_t^TE.

Now, setting m_FP=1 in Equation (12) and m_dif=1 in Equation (13), their combination with Equations (8) to (11) leads to the following two equations for two heights H^FP and H^dif of the grating, which fulfill the Fabry Perot condition (12) and the phase difference condition (13), respectively:

$\begin{array}{cc}{H}^{\mathrm{FP}}=\frac{\lambda }{2\pi n_{\mathrm{eff}}^{\mathrm{TE}}}\left(\pi -1/2{\varphi }_r^{\mathrm{TE}}\right);H^{\mathrm{dif}}=\frac{\lambda }{2\pi \left(n_{\mathrm{eff}}^{\mathrm{TE}}-n_{\mathrm{eff}}^{\mathrm{TM}}\right)}\left(\pi -{\varphi }_t^{\mathrm{TE}}\right)& \left(15\right)\end{array}$

Those two heights which separately satisfy two different conditions (12) and (13) are not necessarily equal, as the period was assigned some value that might not be optimal for the desired function of the polarization transforming grating. However, for the chosen value of Λ=0.8 λ/n_s the two values of the height are close to each other since the chosen period is located in the period domain where the phases of the TE₀ reflection and transmission coefficients enable the possible fulfillment of both Fabry-Perot and phase difference conditions with the least values of M_FT and m_dif. Thus, defining the height of the polarization transformer grating as the arithmetic mean of the two heights H^FP and H^dif of the expressions in (15) defines with the chosen period and related duty cycle a structure which is in the neighborhood of the optimum structure. The difference between H^FP and H^dif is a measure of how close to the optimal geometrical structure the initial guess on the period was. In case this difference is obviously too large, the design is resumed with another period (preferably between Λ=0.7 λ/n_s and Λ=0.95 λ/n_s) or/and a slightly different TM₀ effective index.

Three examples of polarization transformer according to the invention are given hereafter, in a non limiting way, in the case of air or vacuum in the cover medium and grooves.

In the first example the substrate is made of fused silica (n_s=1.44) and the grating ridges made of amorphous silicon (n_r=3.7 for a wavelength of λ=1030 nm). According to the design method laid out above, the following steps are applied to obtain the required grating parameters:

• • Equation (14) gives the effective index of the 0^th order grating mode for the TM polarization:

n _eff ^TM0=1.2

• • The grating period is set to a typical value below the largest possible period according to the cutoff condition (see Equation (1)):

Λ=0.8λ/n_s=591 nm

• • The resolution of the TM₀ mode analytical dispersion equation gives the duty cycle providing n_eff^TM0=1.2:

D=0.18

• • corresponding to a ridge width of W_r=D·Λ=106 nm.
  • The resolution of the TE₀ dispersion equation with D=0.18 and Λ=591 nm gives

n _eff ^TE0=2.72

• • With the grating parameters D=0.18 and Λ=591 nm, the phase of the TE₀ reflection and transmission coefficients at the cover and substrate interfaces are calculated via the modal method (e.g. MCGrating by Nikolay Lyndin; http:/www.mcgrating.com):

arg(r_00,c^TE)=0.19π,arg(r_00,s^TE)=0.28π,arg(t_c,0^TE)=0.04π,arg(t_0,s^TE)=0.07π,

• • Which leads, using Equation (8), to

φ_r^TE=0.4π,φ_t^TE=0.11π

• • (This confirms for this example the assumptions about the required signs of the phase shifts).
  • Equation (15) is now used to calculate the two necessary heights of the grating to give maximal transmission of the TE₀ Fabry Perot (H^FP) and to give a phase shift of π between TE and TM (H^dif):

H ^FP=241 nm,H^dif=311 nm

• • of which the average gives an approximate grating height:

H=276 nm

The difference of the optimal values of H for TE Fabry Perot transmission and TE-TM phase shift from the average is approx. 10%. If this deviation was notably larger, for instance 25%, the design process would be resumed with another period, e.g. 0.7 λ/n_s and 0.9 λ/n_s until the difference between the resulting H^FP and H^dif is smaller than e.g. 20% of their average.

The derived grating has the geometrical parameter

Λ=591 nm,W_r=106 nm,H=276 nm

This grating exhibits 85.7% TE and 97.9% TM power transmissions, while giving a phaseshift between TE and TM field transmissions of 0.999·π. Thus, it shows the desired effect, its performance is better than the state of the art, and the grating parameters are much more relaxed due to the larger grating period and the greatly reduced groove aspect ratio.

In a second example at 1064 nm wavelength the substrate material is MgF₂=1.38), and the grating is made of TiO₂ (n_r=2.74). The same method gives in brief:

• • the required effective index of TM₀ according to Eq. (14) is n_eff^TM0=1.17
  • with Λ=0.8 λ/n_s=617 nm, this n_eff^TM0 is obtained with D=0.20 (W_r=132 nm)
  • with these duty cycle and period, the effective index of the TE₀ mode is n_eff^TE0=1.94
  • the required phases from Eq. (9) amount to φ_r^TE=−0.47π and φ_t^TE=0.05π
  • using Equation (15) the required grating heights are H^FP=339 nm, H^dif=902 nm.

In this case the difference between the two heights required for TE₀ Fabry Perot transmittance and the TE-TM phase difference is obviously too large. The design procedure is resumed with a period closer to λ/n_s. Λ=0.95 λ/n_s which leads to H^FP=347 nm and H^dif=458 nm.

Consequently, the approximate geometrical parameters of the polarizing device for this choice of materials are:

Λ=694 nm,W_r=132 nm,H=402 nm

This grating transmits 86.1% in TE and 97.3% in TM, while giving a phaseshift between TE and TM of 1.087π.

The third example is a polarization transformer for the infrared region (λ=10.6 μm), made of a ThF₄ optical substrate (n_s=1.32) with a Germanium grating on top (n_r=4.01). This optical substrate is a ThF₄ layer of thickness t_ob larger than 3.3 μm between the Germanium corrugation and a larger index substrate, ZnSe in general. Following again the design method above a period of 0.75 λ/n_s gives the approximate structure of the polarization transformer for a CO₂ laser beam:

Λ=6023 nm,W_r=903 nm,H=2526 nm

This grating transmits 95.0% in TE and 99.2% in TM, while giving a phaseshift between TE and TM of 0.96π.

In the infrared, at 10.6 μm wavelength, the device of the invention cannot have a substrate all made of ThF₄. A hard transparent substrate (support) is needed which is usually ZnSe. It is consequently advantageous to insert a thick layer of low index ThF₄ between the high index layer of germanium and the rather large index mechanical ZnSe substrate. The ThF₄ optical substrate must be thick enough to prevent the tunneling of the +/−1st diffraction orders into the ZnSe substrate through the ThF₄ optical substrate. Considering the just determined device of the invention with Λ=6032 nm, n_s=1.32, and λ=10.6 μm, and assuming a maximum acceptable ratio ∈ between the field at the ThF₄/ZnSe interface and that at the bottom of the Ge-corrugation, the optical substrate thickness t_ob must be t_ob>−1.45 ln(∈). Assuming ∈<0.1, the optical substrate thickness must be larger than 3.33 μm, which is easily fabricable.

The presented examples show that the disclosed method based on the simple model of 0^th-order Fabry Perot transmission with m_FP=1 and m_dif=1 permits to identify the grating parameter space where the solutions of the invention lie. During the invention synthesis process some approximations were made (e.g. neglecting the contribution of higher order grating modes in the transmission, neglecting the Fabry Perot effect on the transmission phase) in order to provide an intelligible physical vision of the device operation. This led to the semi-analytic results of Equations (14) and (15), which determine the grating parameters. The two heights in Equation (15) fulfill separately the Fabry Perot condition for TE and the phase difference between TE and TM polarizations; their difference is the criterion for how far the approximate device structure is from the optimal structure; it therefore permits to tune the structure for its operation closer to the specified polarization function.

Now as the design method, based on the identification of the parameter space where solutions exist with the least values of m_FP and m_dif, is set up its results can next be considered as the input data for a numerical search for the Λ, D and H set of parameters leading to the best performance of the polarization function. This search amounts to optimizing a merit function constructed on the basis of the operation of the device of the invention. One possible and very well adapted merit function is the transmission through crossed polarizers, one positioned in front of the polarization transforming grating and one behind it, where the grating is oriented 45° relative to the input polarizer. Only a perfect π-phase-shift and a power transmission of 100% for both TE and TM polarization components results in a transmission through the crossed polarizers equal to 100% while a different phaseshift and/or a reduction of the transmission will reduce this value. Such a merit function is for example available as an output format option of the program MCGrating by Nikolay Lyndin (http://www.mcgrating.com) based on the well known and widely used RCWA or on the true-mode method.

Applying this numerical optimization to the three above examples gives the best performance achievable with each set of refractive indexes (Power transmission T_TE of the TE and T_TM of the TM polarizations as well as the TE-TM phase shift Δφ), and the related geometrical parameters:

Example 1 (amorphous silicon corrugation on a fused silica substrate at 1030 nm wavelength):

• • Λ: 591 nm→597 nm, W_r: 106 nm→105 nm, H: 276 nm→245 nm T_TE: 85.7%→98.1%, T_TM: 97.9.0%→98.3%, Δφ: 0.99π→1.03π

Example 2 (TiO₂ grating on a MgF₂ substrate at 1064 nm wavelength):

• • Λ: 694 nm→733 nm, W_r: 132 nm→146 nm, H: 402 nm→367 nm T_TE: 86.1%→88.9%, T_TM: 97.3%→97.3%, Δφ: 1.08π→1.03π

Example 3 (Ge grating on a ThF₄ substrate at 10.6 μm wavelength):

• • Λ: 6023 nm→6012 nm, W_r: 903 nm→971 nm, H: 2526 nm→2340 nm T_TE: 95.0%→99.5%, T_TM: 99.2%→99.3%, Δφ: 0.96π→1.01π

Beyond tuning the three geometrical parameters Λ, W_r and H of the specific example given above, the same code was used to give all possible devices according to the invention under the hypotheses of air- or vacuum-filled cover and grooves, under normal incidence by proceeding to the numerical fine tuning step by step in the normalized Λ/λ, D and H/λ space, each optimized triplet being the departure point for the fine tuning of the neighboring triplet. This gives rise to the exhaustive determination of all possible triplets belonging to the device of the invention with n_c=n_g=1 versus both substrate and ridge index within ranges that can be covered with existing transparent materials at the desired optical wavelengths and that fulfill n_s>n_c and n_r>n_s. Each parameter of the solution triplets is given in FIG. 10-12, where the labeled curves are the level curves on the surfaces Λ(n_s,n_r)/λ, D(n_s,n_r) and H(n_s,n_r)/λ. As a measure of the quality of the polarization transformer, FIG. 9 depicts the transmission through the mentioned setup with crossed polarizers which was used for the numerical optimization. FIG. 13 shows the ratio P_cop, between the period of the grating and the critical period according to Equation (1), showing that this ratio approaches P_cop=1 for combinations with small ridge index and large substrate index (i.e. the upper left corner in this diagram). FIG. 13 illustrates, that there is a forbidden zone in the direction of large substrate indexes and small ridge indexes (left upper corner in FIG. 13) where the gratings designed by the described algorithm turn out to have too large a period and do not fulfill the subwavelength condition, i.e. they would give rise to propagating diffraction orders in the substrate. The area of the (n_s,n_r) plane containing the devices of the invention not violating this subwavelength condition is indicated by curve 20 in FIG. 9. This area is in the case of the numerical analysis of FIG. 9 (restricted to 1.5<n_r<4.5 and 1.2<n_s<2.3) approximately defined by:

1.5<n_r<4.5

n _s<2.3

n _s<0.381n_r⁴−3.638n_r³+13.102n_r²−20.653n_r+13.206  (16)

For very large ridge indexes (n_r>4) and/or if a slightly lower efficiency of the polarization transformer is acceptable (between 80% and 90%), the condition on the substrate index is less stringent. A usable polarization transformer can in this case be designed up to n_s=2.7. In a preferred variant achieving high efficiencies, the last condition is replaced by a easier relationship corresponding to the linear curve 22 in FIG. 9, i.e. n_r>2n_s. In this limited region, the efficiency is above 90% and FIG. 9 shows that it is possible to have more than 98% in a particular validity area. It can be demonstrated that, in the state of the art, the maximum efficiency remains under 93% for n_s=n_r between 2.6 and 4.0. The present invention gives exceptional results. Not only the grating period is larger and the groove aspect ratio is smaller than in the state of the art, but also the efficiency of the polarization transformers according to the invention is much higher than the maximum efficiency of corresponding polarization transformers of the state of the art. Within this area, the diagrams of FIG. 10-12 provides practical limits of the geometrical parameters of the devices of the invention in the case of air/vacuum cover and grooves as well as the range in which they are defined. The relative and 0.9 (0.4<P_rel<0.9). The lower limit of P_rel is chosen slightly smaller than the smallest value of 0.45 appearing within the scanning range of FIG. 10, since values of P_rel<0.45 may be needed in the case of larger substrate indices (n_s>2.3 at the upper right corner of FIG. 10). According to the invention, the case of P_rel=0.4 is considered as an approximate lower limit for the relative period. In the case of P_rel<0.4 the fabricability of the polarization transformer would no longer be superior to that of the state of the art solution because of the required very high ridge aspect ratio combined with a small period. With n_c greater than 1.0, this condition according to the invention is P_rel>0.4/n_c.

The practical limits of the substrate refractive index of the polarization transformer of the invention are 1.2<n_s<2.7. The upper limit is determined from the fact that the required refractive index for the ridge material would in the case of n_s>2.7 be larger than approximately 5.4 in order to have a high efficiency, above 90% (as can be seen by considering in particular line 22 in FIG. 9, which indicates the minimum required ridge refractive index as a function of the substrate refractive index for an efficiency above 90%). This is approximately the limit for optical materials in the considered wavelength range. Furthermore, n_s>2.7 would also require a polarization transformer with P_rel in the range or below 0.4, which is considered as the practical lower limit according to the present invention.

The duty cycle D=W_r/Λ of the grating (FIG. 11) has substantially a value between 0.15 and 0.25 (0.15<D<0.25). The relative height H_rel=H/λ of the grating (FIG. 12) has substantially a value between 0.16 and 1.0 (0.16<H_rel<1.0) and preferably between 0.2 and 0.5 (0.2<H_rel<0.5). With n_c greater than 1.0, those conditions according to the invention are 0.15/n_c<D<0.25/n_c and 0.2/n_c<H_rel<0.5/n_c.

Furthermore, FIG. 14 shows the groove aspect ratio of all solution triplets, which stays below 1.0 for the major part of the area of the device of the invention, confirming the strongly relaxed requirements on the fabrication compared to the state of the art.

From the set of exact results a two-dimensional polynomial expression was derived for each optimized geometrical parameter versus n_s and n_r enabling the straightforward determination of each geometrical parameter for any couple of n_s and n_r index value. These polynomials are of the following form:

$\begin{array}{cc}f\left(n_r,n_s\right)\approx c_{00}+c_{10}n_r+c_{01}n_s+c_{20}n_r^2+c_{11}n_rn_S+c_{12}n_S^2+c_{30}n_r^3+c_{21}n_r^2n_S+c_{12}n_rn_S^2+c_{03}n_S^3+\dots & \left(17\right)\end{array}$

On the basis of expression (17) the values of the generalized grating parameters and the achievable transmission through crossed polarizers are given in Table 1 hereafter. This approximate analytical expression provides the solution for the disclosed linear polarization rotating grating for any pair of refractive index (n_s,n_r) satisfying the required inequality relationships between indexes.

TABLE 1
Coefficients of the polynomial approximation (see Expression
(17)) of the normalized grating parameters from FIG. 9 to 12.
	Merit value
	(Transmission for	relative		relative
	crossed	period	duty cycle	height
coefficient	Polarizers)	Prel = Λ/λ	D = Wr/Λ	Hrel = H/λ
C00	224.59	3.2264	−1.6212	71.6608
C01	−41.35	−2.4227	1.8312	−67.5264
C10	−145.63	0.3277	0.6719	−69.4952
C21	−5.23	0.8485	−0.6745	25.0883
C22	92.19	0.556	−0.8187	55.4527
C02	−35.2	−1.3733	0.5295	22.5732
C30	1.75	−0.1614	0.1093	−4.7556
C21	−12.3	−0.0549	0.2351	−15.3462
C12	4.82	−0.0441	−0.1221	−15.6794
C03	2.23	0.5398	−0.1122	−2.2388
C40	0	0.0125	−0.0089	0.4696
C31	0	−0.0026	0.0002	1.8858
C22	0	0.0144	−0.0635	2.7572
C13	0	−0.0177	0.1203	2.2824
C04	0	−0.0624	−0.0508	−0.5852
C50	0	0	0	−0.0203
C41	0	0	0	−0.0817
C32	0	0	0	−0.1832
C23	0	0	0	−0.1503
C14	0	0	0	−0.1731
C05	0	0	0	0.1439

All numerical examples up to this point assumed a cover and groove index of n_c=1.0, i.e. vacuum or air. However, the polynomial approximation using expression (17) and Table 1 can also be used in the case of a cover and groove index different from n_c=1.0, as will be explained hereafter.

Two binary corrugation gratings (subscript 1 and 2 in the following relationships) are identical regarding their optical diffraction spectra if their parameters are converted via an arbitrary real factor α according to the relationships:

n _r,2 =n _r,1 α,n _s,2 =n _s,1 α,n _c,2 =n _c,1α  (18)

Λ₂=Λ₁|α,W_r,2=W_r,1|α,H₂=H₁|α  (19)

To calculate a device according to the invention having a cover and groove index n_c,2>1.0, and substrate and ridge indexes n_s,2 and n_r,2, a grating with a cover and groove index of n_c,1=1.0 (as described so far in the invention) and scaled substrate and ridge indexes n_s,1=n_s,2/α and n_r,1=n_r,2/α where α=n_c,2 is first designed, according to the described method or by using FIG. 10-12 or expressions (17). Then the resulting geometrical parameters Λ₁, W_r,1, H₁ of the device with cover and groove index of n_c,1=1.0 are transformed via Equation (19) to give the geometrical parameters Λ₂, W_r,2, H₂ of the grating with the desired cover and groove index n_c,2>1.

This generalizing procedure will be illustrated on the basis of the first example (amorphous silicon ridges on fused silica substrate at 1030 nm wavelength) where the aim now is to find out the device of the invention at the same wavelength in the presence of a cover and groove medium of refractive index 1.1 (e.g. a porous sol-gel). The grating whose geometrical parameters must be found out (subscript 2) is characterized by the set of refractive indexes: n_c,2=1.1, n_s,2=1.44, n_r,2=3.7. The factor α to be used in Eq. (18) is thus 1.1. The refractive indexes of the corresponding grating with a cover index of 1.0 (subscript 1) are calculated by using Equation (18), and the polynomial expressions given by Equation (17) and Table 1 are used to determine its geometrical parameters. It is given by:

n _c,1 =n _c,2/α=1.0,n_r,1=n_r,2/α=3.36,n_s,1=n_s,2/α=1.31

Λ₁=658 nm,W_r,1=112 nm,H₁=277 nm

Relationships (19) are now used to retrieve the geometrical parameters of the grating device of the invention with cover and groove index of 1.1. These are:

n _c,2=1.1,n_r,2=3.7,n_s,2=1.44,

Λ₂=598 nm,W_r,2=102 nm,H₂=252 nm

This polarization transformer transmits 97.6% of the TE and 99.0% of the TM polarizations while giving a phaseshift between them of 1.03π.

Claims

1-20. (canceled)

21. A planar polarization transformer for a normally incident optical wave or beam having a given vacuum wavelength λ, comprising an optical planar grating between a cover medium of refractive index nc and an optical substrate of refractive index ns larger than the cover refractive index nc, this planar grating defining a binary index modulation or corrugation of substantially rectangular profile with periodic ridges, having a ridge refractive index nr and a ridge width Wr, and interleaved grooves having a groove width Wg and groove refractive index ng equal to the cover refractive index nc (ng=nc), this planar grating locally having a grating period Λ which is smaller than the wavelength of the incident optical wave divided by the substrate refractive index (π<λ/ns), wherein said ridge refractive index nr is larger than the substrate refractive index ns, wherein said grating period Λ is larger than 0.4·λ/nc, wherein said substrate refractive index ns is smaller than 2.7·nc, and wherein the index modulation or corrugation is designed such that, according to the grating mode dispersion equation, the effective index of the TE fundamental grating mode TE0 is larger than the substrate index ns and the effective index of the TM fundamental grating mode TM0 is larger than the cover refractive index nc and smaller than the substrate index ns.

22. The polarization transformer according to claim 21, wherein said substrate refractive index ns is smaller than 2.3·nc.

23. The polarization transformer according to claim 21, wherein the optical substrate is a thick layer of thickness tos on a higher index transparent mechanical support, the thickness tos being larger than −ln(0.1)·λ/(2π((λ/Λ)2−ns2)1/2).

24. The polarization transformer according to claim 21, wherein the cover and the grooves are filled with air or vacuum.

25. The polarization transformer according to claim 24, wherein said substrate is composed of high thermal conductivity MgF2 or sapphire or YAG.

26. The polarization transformer according to claim 21, wherein the cover and the grooves are filled with a transparent polymer.

27. The polarization transformer according to claim 21, wherein the cover and the grooves are filled with a transparent liquid.

28. The polarization transformer according to claim 21, wherein said grating has a groove aspect ratio, defined by the height H of said grooves divided by the width Wg of these grooves, which is smaller than 1.5.

29. The polarization transformer according to claim 21, wherein the substrate refractive index ns has a value between 1.0·nc and 2.3·nc (1.0·nc<ns<2.3·nc) and the ridge refractive index nr has a value between 1.5·nc and 4.5·nc (1.5·nc<nr<4.5·nc), and wherein for a given ridge refractive index nr the substrate refractive index ns is lower than the following function F of the ridge index nr: F(nr)=(0.381·(nr/nc)4−3.638·(nr/nc)3+13.102·(nr/nc)2)·nc−20.653·nr+13.206·nc(ns<F(nr)).

30. The polarization transformer according to claim 21, wherein the substrate refractive index ns has a value between 1.0·nc and 2.3·nc(1.0·nc<ns<2.3·nc) and the ridge refractive index nr has a value between 1.5·nc and 4.5·nc(1.5·nc<nr<4.5·nc), and wherein the value of the ridge refractive index is at least twice the one of the substrate refractive index (nr>2ns).

31. The polarization transformer according to claim 29, wherein the relative period Prel=Λ/λ has substantially a value between 0.4/nc and 0.9/nc(0.4/nc<Prel<0.9/nc).

32. The polarization transformer according to claim 29, wherein, in the case nc=1.0, the relative period Prel=Λ/λ has substantially a value, in function of a given pair nr and ns, which is determined by a polynomial expression given in expression (17) of the description with the coefficients given in Table 1 in column Prel.

33. The polarization transformer according to claim 29, wherein the relative height Hrel=H/λ of the grating has substantially a value between 0.16/nc and 1.0/nc (0.16/nc<Hrel<1.0/nc).

34. The polarization transformer according to claim 29, wherein, in the case nc=1.0, the relative height Hrel=H/λ of the grating has substantially a value, in function of a given pair nr and ns, which is determined by a polynomial expression given in expression (17) of the description with the coefficients given in Table 1 in column Hrel.

35. The polarization transformer according to claim 33, wherein the duty cycle D=Wr/Λ has substantially a value between 0.15/nc and 0.25/nc (0.15/nc<D<0.25/nc).

36. The polarization transformer according to claim 29, wherein, in the case nc=1.0, the duty cycle D=Wr/Λ has substantially a value, in function of a given pair nr and ns, which is determined by a polynomial expression given in expression (17) of the description with the coefficients given in Table 1 in column D.

37. A design method for a planar polarization transformer comprising an optical planar grating between a cover medium of refractive index nc and an optical substrate of refractive index ns larger than the cover refractive index nc, this planar grating defining a binary index modulation or corrugation of substantially rectangular profile with periodic ridges, having a ridge refractive index nr and a ridge width Wr, and interleaved grooves having a groove width Wg and groove refractive index ng equal to the cover refractive index nc, this planar grating locally having a grating period Λ which is smaller than the wavelength λ of an incident optical wave or beam divided by the substrate refractive index, said ridge refractive index nr being larger than the substrate refractive index ns, this design method comprising the following step: Choosing a wavelength λ;Choosing a period Λ in the range 0.4·λ/nc<Λ<λ/ns;Determining the duty cycle D=Wr/Λ from the TM0 dispersion equation of the modal theory to substantially realize neffTM0=(nc·ns)1/2, neffTM0 being the effective index of the fundamental TM mode (TM0) of the planar grating;Determining neffTE0 from the TE dispersion equation of the modal theory for the chosen period and wavelength and the determined duty cycle D, neffTE0 being the effective index of the fundamental TE mode (TE0) of the planar grating;Calculation of the reflection and transmission phases φrTE and φtTE of the TE0 mode by means of the modal theory;Calculation of a first height HFP from the expression

38. The design method according to claim 37, wherein it further comprises an optimization step in which a fine tuning of geometrical parameters (Λ, D and H) of the planar grating with an appropriate numerical optimization tool by taking as departure point the corresponding geometrical parameters obtained in the design method of claim 37.

39. The design method according to claim 38, wherein said numerical optimization tool defines a gradient search method using a merit function corresponding to the transmission of the incident optical wave or beam through two crossed polarizers with the grating arranged between them and oriented at 45° relative to each polarizer direction.

40. A design method for a planar polarization transformer, wherein the cover refractive index nc is greater than 1.0, wherein the design method according to claim 37 is used with nc* being equal to 1.0, a scaled substrate refractive index ns*=ns/nc, ns being the substrate refractive index of the planar polarization transformer, and a scaled ridge refractive index nr*=nr/nc, nr being the ridge refractive index of the grating forming the planar polarization transformer, and wherein the period Λ*, the ridge width Wr* and the groove height H* resulting from the design method are divided by the value of nc in order to obtain the grating period Λ, the ridge width Wr of the grating and the groove height H of the grating.