Arxiv:1912.01480V2 [Physics.Optics]

arXiv:1912.01480v2 [physics.optics] 26 Jan 2020 h oainagei oendb h lt hcns and thickness plate the and by governed reciprocal is is angle effect rotation com- the the these elements For available birefringence). mercially natural a (circular exhibiting activity quartz, optical as re- such Faraday be crystals, may the rotators using to polarization alized alternative, connected an As constant Verdet effect. the depen- temperature of They and dence dispersion broadband. strong not the are from and suffer expensive and generally bulky are rotators quite Faraday polarizer However, a analyzer. with of- an connection and in therefore isolators optical are as they used ten re- direction, to propagation respect with the reciprocal versing non the being have of induc- rotators property Faraday magnetic useful [19]. a effect) (Faraday by field on induced tion rely birefringence po- which circular input used rotators the Faraday the commonly are of most rotators the independent polarization Probably is wave which direction. input angle larization polarized fixed linearly a a by of polarization the tates below. discuss will we stack that a [16–18] recent wave-plates using the several rotators also of polarization belong achromatic ideas of of polar- line studies to this (NMR) in To fault- resonance used optics. magnetic the ization widely nuclear apply The of [15] field to approach the [10]. pulses allows Ardavan composite concepts by tolerant of first transfer out related pointed two- systems, coupled of level dynamics quantum-state Schrödinger the the of for and equation change plate birefringent the the formal describing in polarization the equations broad- No- of the improved advantage between take analogy to [5–14]. [10–14] leading decades performance advances several band recent for the interest compact therefore tably, of realize been matter to have the Methods retarders (broadband) dependence retardation. achromatic wavelength phase strong the a of disper- with highly proper- intrinsically components are birefringence sive and the crystals specific on of wave-plate based ties consider are may crucial which we of example retarders, band- an becomes wide As components and such importance. robust or of a broadband operation used, where width are case devices sources the optical light In several tunable for [1–4]. elements applications key and are wave light a notclplrzto oao sa lmn htro- that element an is rotator polarization optical An of state polarization the convert to able Components rabn oaiainrttrwt ual oainangl rotation tunable with rotator polarization Broadband ohmdAl-Mahmoud, Mouhamad lt.Terbs prahisie ytecmoiepulses composite the by extent. large inspired in approach three dispersion robust wave-plate only The of consists rotator plate. The experimentally. verified ipeshm o rabn oaiainrttrwt tu with rotator polarization broadband a for scheme simple A .INTRODUCTION I. 1 eateto hsc,Sfi nvriy ae orhe b 5 Bourchier James University, Sofia Physics, of Department 2 nvri´ eLran,CnrlSpee,LOS F-5700 CentraleSupélec, LMOPS, Lorraine, Université de 1 igneCoda, Virginie Dtd aur 8 2020) 28, January (Dated: wave-plates 2 no Rangelov, Andon Pi rabn.Aogti anie eea stud- several idea main this Along single every if broadband. broadband is become WP combi- should not a these WPs was by wavelength of composed of rotator of nation any robustness change principle the In a to discussed. cases respect both modu- with in light devices However, spatial liquid-crystal lator. parallel-aligned a retarder by homogeneous spatially demon- the replacing to by [25] encoding al. polarization addressed et electrically Davis two-dimensional by strate used also was concept This etlast ual oaiainrttrfrwihthe which for angle rotator polarization rotation arrange- tunable this a intermediate to QWPs the leads the ment of of axis those bisects fast quarter- retarder the variable variable If crossed a two [24] (QWPs). plates Ye WPs, between wave context three sandwiched this of retarder combination In WP wave- a birefringent choice. considered of best has stacks the of remains and use compact lim- plates for the which Therefore rotators incidence, [23]. tunable of aperture their angular angle is the the rhombs its Fresnel besides to of Nevertheless, sensitivity drawback strong broadband. another be size, wavelength can their with device little a only vary such jumps phase these Since t eadto sdet h ieetpaejm suffered jump phase birefringence, different on the the by to rely due [22]. not is rotator does retardation its bulky rhomb very Fresnel a but achromatic Indeed, can nearly one a retarders, obtain birefringent angle of an instead by rhombs polarization Fresnel the turns angle that α an rotator making a half-wave axes to two fast lead that their known with birefringent well (HWPs) several plates is of It combination (WPs). the wave-plates on rely rotators high for modified performance the easily affect be powers. may effects cannot thermal angle and rotation obtained. be spe- the can wavelength However, components not achromatic is nearly effect and liq- This cific polarization the the molecules. of crystal reorientation thick local uid sufficiently the is adiabatically Pro- follow cell can technology. the display is crystal that which liquid principle vided of liquid same base the the nematic much at by twisted also Yet 21] of [20, bandwidth. cells use useful crystal makes the dispersion approach limits strong activity another rather optical the here the Also of tuned. be cannot ti ot oiigta,i o hsproeoeuses one purpose this for if that, noticing worth is It . h otvraiewyt elz ual polarization tunable realize to way versatile most The aepae,oeo hc safull-wave a is which of one wave-plates, p n the and - al oainagei rpsdand proposed is angle rotation nable nlg lost opnaethe compensate to allows analogy 1 n emn Montemezzani Germano and α v,16 oa Bulgaria Sofia, 1164 lvd, shl h eadto fteretarder. the of retardation the half is s ez France Metz, 0 plrzdwv pnttlreflection. total upon wave -polarized opsdo three of composed e 2 α/ 2 2 ies of composite rotators based on the quantum-optical (a) HWP FWP HWP analogies mentioned above were performed. Rangelov !" !& ! and Kyoseva [16] have proposed a broadband compos- #" %" #& % $&" # %" ite polarization rotator based on the combination of two effective broadband HWPs, each of which is composed by a number of HWPs by the composite approach. The expected device performance was analyzed theoretically α in terms of the so called fidelity (see below) for a total number of HWPs between 6 and 18. This concept was demonstrated experimentally in Ref. [17] for 6 and 10 Input polarization Output polarization HWPs. Besides for the broadband configuration, also a (rotated by an angle α) narrowband configuration was implemented by another choice of the orientations of the individual HWP fast (b) axes. This concept was developed further in a recent Polarizer Light Analyzer Removable paper [18], where an even number of HWPs (up to 10) Source mirror oriented at predetermined angles was used and the broadband behavior was tested through the transmission of a Spectral Photodiode white light source through an analyzer placed after the Filter HWP stack. Spectrometer Rotator sequence

FIG. 1. (a) Principle of the proposed composite polariza- While the above approaches generally require a rather tion rotator composed of three wave-plates, a half-wave plate large number of WPs to achieve a sufficiently broadband (HWP) followed by a full-wave plate (FWP) and another operation ( 6), in the present work we consider a simpli- HWP. The angles θi are the orientation angles of each wave- fied arrangement≥ involving only three WPs. With respect plate and ϕi are the corresponding retardations. (b) Experi- mental set-up for the characterization of the composite rota- to the approach of Ye [24], which is also based on three tor. WPs, here the middle retarder is replaced by a full-wave plate (FWP) and the two outer QWPs are replaced by HWPs. The concept exploits additionally the freedom II. THEORY of rotation of all three elements in the row. It is shown that such a simple stack leads to a broadband polariza- The broadband polarisation rotator proposed in this tion rotator provided that the intermediate wave-plate is work is composed of three wave-plate (WP) retarders as placed in such a way as to counteract the dispersion of shown in Fig. 1(a). The first and the third WPs are the HWPs. The rotator is robust against the initial po- half-wave plates for the central target wavelength λ0 of larization direction and the rotation angle can be tuned the device, while the intermediate wave-plate is a full- by rotating only one of the WPs. In Section II we give wave plate at the same wavelength. Even though this the theoretical background for the taken approach and element leaves the wave unchanged and acts as a neu- compare it with the cases where only two HWPs would tral element at the wavelength λ0, the importance of this be used or the case where the intermediate full-wave plate crucial element for the broadband behavior will become would be placed under a ”wrong” direction (phase shift clear below. of +2π rather than 2π). Notably, our sequence HWP As it is well known, a wave-plate retarder is a birefrin- (phase-shift of +π) -− FWP ( 2π) - HWP (+π) is remi- − gent element which adds different phases ϕ/2 and ϕ/2 niscent of schemes used to address fault-tolerant coherent to two perpendicular linear polarization components− of population transfer or phase gates in two-state quantum the light propagating through it. In the framework of systems using effective zero-area pulses or combination Jones calculus [1, 2] and in the HV-basis formed by the of pulses with a vanishing total area [26, 27]. An impor- Jones vectors for horizontal and vertical linear polariza- tant feature consists in the fact that, unlike for the case tions, the Jones matrix for a retarder whose slow and fast of a two HWPs rotator, the three WPs approach leads to axes are aligned along the HV-axes is given as a near unity degree of linear polarization of the output wave even at wavelengths very far from the central de- eiϕ/2 0 (ϕ)= . (1) sign wavelength. In Section III we verify successfully the J0 0 e−iϕ/2 concept experimentally either by using a monochromatic wave at the central design wavelength or a broadband Here source covering a bandwidth of approximately 400 nm. ϕ =2πL(ns nf )/λ (2) Theoretical expectation and experimental results agree − very well and confirm that such a simple arrangement is is the wave-plate retardation, i.e. the phase shift be- suitable for broadband operation. tween the two orthogonal polarization components upon 3 passing the element. The quantities nf and ns are the which gives refractive indices along the fast and slow axes, respectively, λ is the vacuum wavelength of the light and L is e−2i(θ3−θ1) 0 e−iα 0 J = = . (13) the thickness of the retarder plate. The most commonly 0 e2i(θ3−θ1) 0 eiα used retarders are the HWPs (ϕ = π) and the QWPs (ϕ = π/2). A full-wave plate (FWP)± has a retardation Obviously the last equality shows that J corresponds to of ϕ =± 2π. the rotator matrix in the LR basis in Eq. (10), therefore If the± retarder plate is turned by an angle θ around the this sequence acts as a rotator with a rotation angle light propagation axis (perpendicular to the plate), then its Jones matrix (ϕ) in the HV basis can be found α = 2(θ3 θ1) . (14) Jθ − from 0 (ϕ) as J This equivalence is exact at the central wavelength, for θ (ϕ)= ( θ) 0 (ϕ) (θ) , (3) which the retardations ϕ , ϕ and ϕ correspond exactly J R − J R 1 2 3 to those given above. However, we are principally in- where (θ) is an axes rotation matrix given as R terested in the behavior found when these retardations depart from the values π, 2π and π, as a result of us- cos θ sin θ (θ)= . (4) ing a different wavelength± (see Eq. (2)). Therefore, in R sin θ cos θ − order to explore the behavior in the ϕ space, we define Explicitly the form of (ϕ) is the fidelity factor according to Jθ 1 (ϕ) = eiϕ/2 cos2 (θ)+ e−iϕ/2 sin2 (θ), (5) Tr J −1 (α) J , (15) Jθ 11 F ≡ 2 R θ (ϕ) = i sin (2θ) sin (ϕ/2), (6) J 12 −1 where we note that J (α) = JR ( α). The fidelity θ (ϕ) = i sin (2θ) sin (ϕ/2), (7) R J 21 is therefore a kind of measure on− how close the com-F (ϕ) = e−iϕ/2 cos2 (θ)+ eiϕ/2 sin2 (θ). (8) Jθ 22 posite matrix J approaches the target matrix JR(α). In It is worth noticing that (up to an unimportant minus the case where the output light maintains a linear polar- sign) the rotation matrix (4) corresponds to the Jones ization state, the fidelity finds a more direct physical interpretation, as will beF mentioned later in the experi- matrix R (α) of an optical rotator in the HV basis for a polarizationJ rotation by +α (with the positive angles mental section. defined in counterclockwise direction), indeed Let us consider the relative deviation δ from the retardation values ϕi = ϕ(λ0) of the three wave-plates defined cos α sin α as R (α)= ( α)= − . (9) J R − sin α cos α ϕ(λ) ∆n(λ) λ δ 1= 0 1 , (16) ≡ ϕ(λ0) − ∆n(λ0) λ − To simplify the above matrices θ (ϕ) and R (α), it is convenient to express them inJ the left-rightJ circu- where λ is the central wavelength for which the compos- lar polarization basis (LR), where the basis Jones vec- 0 ite structure is designed and ∆n = n n in Eq. (2). tors are L = 1/√2(1,i) (left circular polarized) and s − f While the orientation angles θ1 and θ3 are related by R = 1/√2(1, i) (right circular polarized). We call the − Eq. (14), the optimum angle θ2 can be found by max- corresponding Jones matrices in this basis Jθ (ϕ) and imizing the integral of the fidelity over the range JR (α) (note the different symbol with respect to the HV 1 δ 1, i.e. F basis), a straightforward calculation leads to − ≤ ≤ 1 +1 e−iα 0 Q = (δ) dδ , (17) JR (α)= , (10) 2 Z |F | 0 eiα −1 and which means searching for the broadest fidelity curves (δ). The above integral can be considered as a quality ϕ ϕ F cos ie−2ıθ sin measure (figure-of-merit) for the robustness of the rotator 2 2 Jθ (ϕ)=  ϕ ϕ  . (11) and its maximum possible value equals 1. It can be shown ie2ıθ sin cos analytically that this integral is maximized if  2 2  For our sequence of HWP-FWP-HWP shown in θ2 = θ1 + α/4 π/2= θ3 α/4 π/2 . (18) − − − Fig. 1(a) we have ϕ1 = π, ϕ2 = 2π and ϕ3 = π, the corresponding orientations for the± three wave-plates are The additional angle π/2 appearing above is important. θ1, θ2 and θ3, respectively. The overall Jones matrix de- In fact, the analysis was performed by assuming ϕ2(λ0)= scribing this composite sequence is therefore +2π, however, the additional angle of π/2 implies that the role of the slow and fast axis should be permuted J = J (π) J ( 2π) J (π) , (12) for the intermediate FWP, meaning that the optimum θ3 θ2 ± θ1 4

(a) 2 WPs ( π, π) (b) 2 WPs ( π, -π) 90 1.0 1.0 (a) 2 WPs 75 | ) δ ( 0.5 0.5 60 |F

0.0 0.0 45 –1 –0.5 0 0.5 1 1 –0.5 0 0.5 1 30 (c) 3 WPs ( π, +2π, π) (d) 3 WPs ( π, -2π, π) 1.0 1.0 15 | )

δ α=15° Rotation angle (deg) angle Rotation ( 0.5 0.5 α=30° 0 |F α=60° –1 –0.5 0 0.5 1 α=90° 0.0 0.0 90 –1 –0.5 0 0.5 1 –1 –0.5 0 0.5 1 δ δ (b) 3 WPs 75 FIG. 2. Absolute value of the fidelity |F| as a function of 60 α=15° the relative retardation deviation δ for four different optical α rotator configurations. (a) Two HWP with their fast axes 45 =30° making an angle of α/2, equivalent to ϕ = π for both wave- α=60° plates at the central wavelength. (b) Two HWP with the fast 30 axis of the first making an angle α/2 with the slow axis of the α=90° second, equivalent to ϕ1 = π = −ϕ2. (c) Three wave-plates, 15 a FWP sandwiched between two HWPs. The fast axis of the FWP makes an angle of +α/4 (−α/4) with respect to the (deg) angle Rotation 0 fast axis of the first (third) waveplate. The retardations are –1 –0.5 0 0.5 1 ϕ1 = π, ϕ2 = +2π and ϕ3 = π. (d) Same as (c) but the fast and slow axes of the FWP are switched (see Eq. (18)), here δ ϕ1 = π, ϕ2 = −2π and ϕ3 = π. The curves are for following target rotation angles: α = 15 deg (dotted red line), α = 30 FIG. 3. Output polarization rotation angle as a function of deg (solid black line), α = 60 deg (dashed blue line) and the relative retardation deviation δ. (a) Sequence of two HWP α = 90 deg (dashed-dotted green line). (ϕ1 = π, ϕ2 = −π) as in Fig. 2(b); (b) Sequence of three wave-plates (ϕ1 = π, ϕ2 = −2π and ϕ3 = π) as in Fig. 2(d). The curves for four different target rotation angles α have the is found for a negative retardation ϕ2(λ0) = 2π (by same styles as in Fig. 2. simultaneously dropping the π/2 term in Eq.− (18)). To illustrate the expected robustness− of the composite rotator we depict in Fig. 2 the expected fidelity as a function of the relative retardation deviation δ. The quantity quite quickly as δ departs from zero. This means that being represented is such a structure is not spectrally robust. Interestingly, the robustness improves already significantly by means 1 −1 of a small modification, still using only two wave-plates. (δ) = Tr J (α) J(δ) , (19) |F | 2| R | If the second HWP is turned by an additional 90 degrees, with its retardation becomes negative (= π) and, as seen in Fig. 2(b), the function (δ) remains− large over a much F J(δ) Jθ3 (π(1 + δ)) Jθ2 ( 2π(1 + δ)) Jθ1 (π(1 + δ)) . wider range of the parameter δ. Figure 2(c) show the ≡ − (20) case where the sequence of Fig. 1(a) is implemented with Note that here a unique value of the parameter δ can be ϕ1 = π, ϕ2 = +2π and ϕ3 = π. This corresponds to considered for the three wave-plates provided that their the situation where the fast axes of the two HWP and dispersion is the same, which is the case for the system of the central FWP are all oriented within an angle α/2. used in our experimental study. Note also that it is suffi- In this case the retardation dispersion associated to the cient to consider target rotation angles α 90 degrees FWP reinforces the dispersion of the HWPs and the fi- because larger angles are redundant, also| |≤ the situation delity drops even faster than in the case of Fig. 2(a). Fi- for negative angles α is symmetric to the one for posi- nally, Fig. 2(d) show our chosen configuration for which tive ones. First we show in Fig. 2(a) the standard case the orientation angles follow Eq. (18) meaning that the where two HWPs under a relative angle α/2 are used to slow axis of the FWP is aligned in between the fast axes create a rotator by an angle α. While such a configura- of the external HWPs. It is evident that in this case the tion acts as a perfect rotator at the central wavelength function (δ) gets flatter on the top and is wider than ( = 1 for δ = 0), for all four considered angles α be- in any otherF case in Fig. 2. As we will discuss later, the tweenF 15 and 90 degrees the fidelity is found to drop experimentally most relevant range for the parameter δ 5 is roughly 0.5 δ 0.5, for which this three wave- 1.0 plate configuration− ≤ of≤ Fig. 2(d) is clearly outperforming any other case in Fig. 2. It is also worth noting that 2 WPs the curve (δ) is found to be broader for small target 0.5 rotation anglesF α than for larger ones. This situation is 3 WPs distinct from the one treated earlier by Rangelov et al. [16] using composite structures with a larger number of 3 0.0 HWP ( 6), where the fidelity curve was found to be S slightly≥ wider for the the largest target rotation angles. Even though the fidelity is high, when the wavelength –0.5 departs from the central design wavelength λ0, the wave- plate sequence will not act exactly as the nominally designed rotator. However, the function remains in very –1.0 good approximation the one of a rotator. On the one –1 –0.5 0 0.5 1 hand it is impossible to avoid that the output wavelength δ would get a (small) elliptical polarization component for λ = λ . On the other hand the polarization rotation 6 0 FIG. 4. S3-component of the output polarization Stokes vec- angle, characterized by the angle leading to maximum tor for the two cases shown in Fig. 3. The blue curves are for transmission through a properly oriented analyzer, will the two HWPs case of Fig. 3(a), the red curves are for the be a smooth function of δ (or λ). We discuss these issues three WP sequence of Fig. 3(b). Solid curves are for a target for the cases in the right column of Fig. 2 for which the rotation angle α = 30 deg, while dashed curves are for α = 60 fidelity curves are broader. It is worth noticing that the deg. configurations of the left column of Fig. 2 perform much worst also in this matter. Figure 3 shows the theoretical expected rotation angle sequence treated in the present work, while significant as a function of δ for the sequences used for Fig. 2(b) values of S3 are found for the sequence composed of two and Fig. 2(d). The depicted rotation angle is obtained HWPs. By defining the degree of linear polarization as by applying the matrix J(δ) in (20) to an initial horizon- ξ 1 (S )2 (22) tal linear polarization and projecting the obtained Jones ≡ − 3 vector into an analyzer oriented at a variable angle γ. p The angle γ for which the transmission would be maxi- we find that ξ is always very close to one for the three WP sequence (ξ > 99.9% for α = 30 deg, ξ > 99.4% mum (maximum squared projection) corresponds to the expected rotation angle. It is seen that in both cases the for α = 60 deg, and still ξ > 95% for the worst case of α = 90 deg). In contrast, for the two HWP case, since rotation angle corresponds to the target one at δ = 0 and decreases symmetrically for positive and negative values the circular polarized component is much stronger, the degree of linear polarization can decrease to ξ 87.5% of δ. In the most important case of three wave-plates for α = 30 deg, to ξ 61.7% for α = 60 deg≈ and to the behavior is very well approximated (but not exactly ≈ 2 ξ 40% for α = 90 deg. given) by the function α(δ)= α(δ = 0)cos (δπ/2). ≈ For the two HWP configuration the decrease of the It is worth to note that the curves for the rotation angle in Fig. 3 give the orientation of the longest main axis rotation angle with increasing δ is initially less steep. of the output polarization ellipse. Only in the case where However, for this configuration the| | output light field polarization exhibits a rather strong ellipticity. In contrast, the output light is still linear polarized, which is essen- tially true for Fig. 3(b) as discussed above, this angle cor- for the three wave-plates sequence the output light remains always very close to a linear polarization state, responds to the rotation angle of the linear polarization. In the specific case of Fig. 3(a) the apparently broader what represent the major advantage of this configuration. To prove this we represent for both cases in Fig. 4 curves with respect to Fig. 3(b) are related to the strong ellipticity of the output wave for values of δ close to 0.5 the component S3 of the Stokes vector [1] as a function ± of δ. The S component is defined as seen in Fig. 4. This pushes up artificially the apparent 3 rotation angle around these values. This kind of artifact

S3 IR IL , (21) does not mean that the two WPs configuration would be ≡ − better than the three WPs configuration. In fact, the where IR and IL are the normalized intensity transmis- fidelity, which is broader in the three WPs configuration, sions of the output light field through a right-circular gives a much better measure of the usefulness of the de- and a left-circular analyzer, respectively. Right and left vice, as discussed later. circular polarized light have S = +1 and S = 1, re- The above discussion clearly shows that, unlike the 3 3 − spectively, while linear polarized light has S3 = 0 and other configurations, the three wave-plate sequence with lies on the equator of the polarization Poincarésphere. nominal retardations (ϕ , ϕ , ϕ ) = (π, 2π, π) offers a 1 2 3 − A scrutiny of Fig. 4 clearly shows that the S3 component combination of a robust behavior under variation of δ and remains always very small in the case of the three WP of the wavelength, as well as of an always nearly linearly 6 polarized output light field. In the next section we will investigate this configuration experimentally. 0.6

III. EXPERIMENTS 0.4 0.2 The experimental set-up for the characterization of our optical rotator is shown in Fig. 1(b). The heart of the δ system is composed of two crystal polarizers (acting as 0.0 polarizer and analyzer) surrounding the three wave-plate rotator sequence. The light source is either a monochro- –0.2 matic laser source (He-Ne laser at λ = 632.8 nm) or a broadband white light source (Thorlabs SLS201L/M). –0.4 Three distinct operation modes are used. In the case of 400 500 600 700 800 the monochromatic He-Ne laser source the spectral filter and the mirror behind the analyzer are removed and λ (nm) the detection of the light transmitted through the analyzer occurs by means of a Si-photodiode. The same FIG. 5. Experimental wave-plate relative retardation devia- detection mode is used in the case where the broadband tion δ as a function of λ referred to the retardation at a central light source is combined with a spectral filter at different wavelength λ0 = 633 nm (blue circles). The red fitted curve wavelengths to leave a quasi-monochromatic 10 nm wide is according to (23). (FWHM) spectrum through the system. Finally, by inserting the switchable mirror before the detector, the use of the broadband source without the spectral filter al- the He-Ne laser source at λ0 = 632.8 nm is used and the lows to detect the whole transmitted spectrum through LC wave-plates are adjusted to be half- or full wave-plates the analyzer with the help of a spectrometer connected at this wavelength. Figure 6 compares the target rotation to a computer (OceanOptics USB4000-VIS-NIR). angles α with the experimentally measured polarization The three wave-plates composing the rotator sequence rotation angles and confirms an excellent agreement. The are realized using three tunable liquid crystal (LC) re- target rotation angles are adjusted by changing the orien- tarders (Thorlabs LCC1413-A), whose retardations are tations θ1 and θ3 according to (18) and the experimental adjusted with an external applied voltage. The voltage- rotation angles αexp are obtained by finding the mini- retardation curve of each of the LC retarder has been pri- mum and maximum transmitted intensity upon rotation orly calibrated over the spectral range of interest using a of the analyzer behind the rotator sequence in Fig. 1(b). Soleil-Babinet optical compensator (Thorlabs SBC-VIS) The insets in Fig. 6 show examples of such measurements put in series with the LC retarder and whose mechani- for α = 30◦ and α = 60◦. The dependence of the trans- cally adjustable retardation is known. mitted intensity I(β) on the analyzer orientation β (with The above calibration allows to establish the LC WP β = 0 being the extinction position in absence of the ro- retardation for each applied voltage at each of the filtered tator sequence) is I(β)= I sin2 (β α )+I , where 0 − exp min wavelengths (450, 500, 550, 600, 633, 700, 750 and 800 nm) I0 is the modulation amplitude and Imin is the minimum and to determine the spectral dependence of the relative transmission. The full contrast seen in the insets in Fig. 6 deviation δ for a given central wavelength λ0. The latter confirms that the output light maintains its linear polar- is determined by considering the wavelength dependence ization. Indeed, the degree of linear polarization ξ in (22) of the maximum retardation (maximum ∆n) found for corresponds to the fringe visibility of such measurements zero applied voltage. The obtained nonlinear relationship given by (I I )/(I + I )= I /(I +2I ). max − min max min 0 0 min δ(λ) is shown in Fig. 5 for the case where λ0 = 633 nm The rotation angle of the above WP sequence can be is the wavelength at which δ(λ0) = 0. Clearly in this used as a tunable rotator by reorienting only one of the case the spectral range between 450 and 800 nm spans a three wave-plates, the first or the third. Let us rotate the δ-range between 0.65 and 0.3. The fitted dispersion ≈ − first wave plate, initially oriented under the angle θ1 by curve in Fig. 5 follows the relationship ′ a supplementary angle ∆θ1, so that θ1 = θ1 + ∆θ1. With ′ ′ λ2 λ¯2 λ α = 2(θ3 θ1) and α = 2(θ3 θ1) the new target rotation δ(λ)= 0 1 , (23) angle α′ shall− vary double as− fast as θ′ , i.e. α′ = α 2∆θ . 2 − 2 1 1 λ λ¯ λ0 − − − This is confirmed in Fig. 7 that shows the tuned rotation ′ ◦ ◦ with an effective oscillator wavelength λ¯ = 221.2 nm. angle α for a variation of ∆θ1 between 90 and +90 ◦ − The above expression is obtained by assuming a simpli- and an initial target angle α = 60 . fied Sellmeier-like function for the dispersion of the LC Unlike the rotation of polarization by a single half-wave birefringence ∆n = ns nf entering Eq. (2). plate, which depends on the input polarization direction We first analyze the− behavior of the three-WP rotator γ, a rotator should rotate the linear polarization by the at the wavelength λ0 for which it is designed. In this case same amount independently of γ. The inset in Fig. 7 7

1 1 1 90 α=15° α=30° I 0.5 0.5 0 60 0 90 180 β 0 0 1 1 1 (deg) α=45° α=60° I exp 30 α 0.5 0.5 (arb. units) (arb.

0 ⊥

0 90 180 I ,

β ||

I 0 0 0 1 α 1 α 0 30 60 90 =75° =90° Target rotation angle α (deg) 0.5 0.5

FIG. 6. Experimentally measured rotation angle αexp (blue dots) vs. the target rotation angle α for the nominal wave- 0 0 length λ = λ0 = 632.8 nm. The two insets show the measured 400 600 800 400 600 800 transmitted intensity I through the analyzer oriented at an λ (nm) λ (nm) angle β for α = 30◦ and 60◦, as well as the function I(β) given in the text. The orientation angles of the WPs are arranged FIG. 8. Spectra of the normalized intensities Ik (red curves) according to Eq. (18). and I⊥ (blue curves) transmitted through the analyzer in transmission, resp. extinction mode. For Ik the transmis- 90 sion direction of the analyzer corresponds to the target out- 360 put polarization at the nominal wavelength λ0 = 550 nm of 60 the composite rotator.

(deg) 30

240 exp

α 0 –90–45 0 45 90 ter center the available spectrum into the corresponding γ (deg) range of the δ-parameter, we choose λ0 = 550 nm. First (deg) 120 we send the whole spectrum of the broadband source

exp through the composite rotator and detect the correspond- '

α 0 ing spectrum after passing the analyzer with the optical spectrometer (see Fig. 1(b)). The analyzer is put either in transmission mode (transmitted intensity = I ) or in –120 k extinction mode (transmitted intensity = I⊥). In trans- –90 –45 0 45 90 mission mode the analyzer transmission axis is put par- θ allel to the expected output polarization direction under ∆ 1 (deg) the target rotation angle α for the nominal wavelength, while in extinction mode it is put perpendicular to this ′ FIG. 7. Tuning of the experimental rotation angle αexp upon direction. Figure 8 shows the intensities Ik (red) and I⊥ a variation ∆θ1 of the orientation of the first HWP in the (blue) for six values of α. While for small α the extinc- ◦ sequence and α = 60 for ∆θ1 = 0. The solid red curve tion in extinction mode is nearly perfect over the whole ′ corresponds to the relation α = α − 2∆θ1. The inset shows spectrum, for larger rotation angles a weak transmitted the experimental rotation angle αexp as a function of the input ◦ intensity persists at the border of the spectrum. This polarization angle γ for the target rotation of α = 60 . γ = 0 is connected to the narrower fidelity function (δ) for corresponds to horizontal input polarization (the bisectrix of large than for small angles α seen in Fig. 2(d).|F | the HWP orientation angles), λ0 = λ = 632.8 nm. The red curves for Ik in Fig. 8 do not permit a direct comparison with the theoretically expected intensity in transmission mode because they are influenced shows the constancy of αexp upon variation of γ for the same rotator sequence used for Fig. 6 and α = 60◦. The by the source spectrum and detector sensitivity, as well expected independence on the input polarization angle is as the transmission spectrum of all optical elements in therefore well verified experimentally. the set-up. To take away all these effects and permit Next we test the broadband behavior and the robust- this comparison we normalize therefore the intensity Ik ˜ ness of the rotator if the used wavelength differs from as Ik Ik/(Ik + I⊥). Figure 9 shows the corresponding ≡ ˜ the nominal wavelength λ0. In this case, in order to bet- spectra for Ik(λ) together with the theoretically expected 8 δ δ δ 0.2 0 -0.2 -0.4 0.2 0 -0.2 -0.4 0.6 0.3 0 -0.3 1.0 1 )

λ 90 ( α α

=15° =30° ˜ I 75 0.5 0.5 Theory 60 α=90° Experiment α=75°

(deg) 45 0.0 0 α 1 1 =60° exp 30 α=45° α α α α =45° =60° 15 =30° 0.5 0.5 α=15° 0 400 500 600 700 800 λ 0 0 (nm) 1 1

FIG. 10. Variation of the polarization rotation angle αexp α=75° α=90° with wavelength. The points are measured values at seven 0.5 0.5 distinct wavelengths and the curves give the theoretically expected dependence. The α-values characterizing each curve are the target rotation angles for the nominal central wave- Normalized transmitted intensity 0 0 length λ0 = 550 nm. The upper abscissa give the values for δ 500 600 700 800 500 600 700 800 corresponding to the wavelengths. λ (nm) λ (nm)

˜ FIG. 9. Normalized transmitted intensity spectra Ik(λ), the dependence of the rotation angle on the parameter δ dis- dotted blue curves give the experimental spectra while the cussed in Fig. 3(b). To check this we have performed underlying solid red lines are the theoretically expected spec- measurements of the experimental rotation angle at seven tra. The nominal wavelength for the three WPs rotator is filtered wavelengths (λ = 450, 500, 550, 600, 650, 700 and λ0 = 550 nm. The upper scale give the corresponding values for δ. 750 nm) by inserting the corresponding spectral filter after the broadband light source in the set-up. The experimentally measured and theoretically expected rotation ones. The latter are obtained by applying the resulting angles (see also Fig. 3(b)) are shown in Fig. 10 and show Jones matrix (19) to the input polarization Jones vector a very good agreement. and projecting the resulting expected output polarization onto the analyzer to obtain the expected transmitted intensity as the square of the module of the projected IV. DISCUSSION AND CONCLUSION polarization vector. The conversion between the wavelength λ and the relative retardation deviation δ is made We have proposed theoretically and verified experi- by means of Eq. (23) as discussed in connection to Fig. 5. mentally a new design for a composite broadband polar- As can be seen in Fig. 9, the agreement between the mea- ization rotator composed of only three wave-plates, two sured and the expected normalized transmission spectra half-wave plates and one full-wave plate for the central ◦ is excellent. Even in the worst case scenario (α = 90 nominal wavelength. This design is simpler with respect and a wavelength exceeding the nominal wavelength by of earlier broadband composite rotators composed of a ˜ 300 nm) the normalized transmitted intensity Ik is still larger number of wave-plates [16–18]. We have shown 70%. A careful analysis shows that, in the case where ≈ that the output polarization state remains nearly linear the output light is still linearly polarized, the normalized even for strong departure from the nominal central wave- transmitted intensity I˜ corresponds to the square of the k length λ0 and that the polarization rotation angle has fidelity (δ) 2, what justifies the choice of the fidelity its maximum at λ and diminishes smoothly away from |F | 0 as the quantity to optimize. In our case, by measuring this wavelength. The rotation angle can be tuned by ro- the fringe visibility for wavelength filtered radiation as tating only one of the wave-plates and is robust against in the insets of Fig. 6, the degree of linear polarization the initial polarization direction. It is useful to compare ξ is found to be close to 100% throughout, as expected the values of the figure-of-merit Q defined in Eq. (17) theoretically (see Fig. 4 and related discussion). Thus for our selected configuration, the three other configu- the above condition is satisfied and the measurements of rations of Fig. 2, and the case of Ref. [18], which gave Fig. 9 can be considered as direct measurements of the the wider fidelity curves to date. By taking the example fidelity squared. of a target rotation angle α = 60◦, the standard rotator ˜ Therefore the decrease of Ik far from the nominal wave- composed of two HWPs of Fig. 2(a) has a figure-of-merit length is not due to a decrease of ξ but it is due to the Q = 0.504, the addition of a FWP with the ”wrong” 9 sign of the phase shift as in Fig. 2(c) leads to a similar N = 8 is associated to a significantly better figure-of- value of Q = 0.491. In contrast, the zero-pulse-area-like merit than for the present much simpler configuration. two-HWPs configuration of Fig. 2(b), that we have dis- In conclusion, we believe that the three wave-plates ro- cussed in larger extent, leads to an improvement since tator discussed in this work will be useful for light polar- Q = 0.750. However, as discussed above, this configu- ization management of broadband, short-pulse or tunable ration is related to strongly elliptical polarized output light sources. light for a certain range of the detuning parameter δ. Finally, the three-WPs design studied in this work has Q =0.808. This value compares favorably with the cor- ACKNOWLEDGMENTS responding values for the schemes of Ref. [18], which are 0.812; 0.720; 0.863 and 0.727 for a total number of WPs This work is supported by EU Horizon-2020 ITN N equal to 4; 6; 8 and 10, respectively. Only the case project LIMQUET (contract number 765075) and by the Bulgarian Science Fund Grant No. DN 18/14.

[1] M. A. Azzam, N. M. Bashara, Ellipsometry and Polarized [14] E. Dimova, W. Huang, G. Popkirov, A. A. Rangelov, Light, North Holland, Amsterdam, 1977. and E. Kyoseva, Broadband and ultra-broadband modu- [2] D. H. Goldstein, Polarized Light, 3rd Ed., (CRC Press, lar half-wave plates, Opt. Commun. 366, 382-385 (2016). Boca Raton, 2011). [15] M. H. Levitt, Composite pulses, Progr. Nucl. Magn. Re- [3] F. J. Duarte, Tunable Laser Optics, 2nd Ed., (CRC son. Spectrosc. 18, 61-122 (1986). Press, Boca Raton, 2015). [16] A. A. Rangelov and E. Kyoseva, Broadband composite [4] D. S. Kliger, J. W. Lewis, C. E. Randall, Polarized Light polarization rotator, Opt. Commun. 338, 574-577 (2015). in Optics and Spectroscopy,(Academic Press, Boston, [17] E. Dimova, A. A. Rangelov, and E. Kyoseva, Tun- 1990). able bandwidth optical rotator, Photon. Res. 3, 177-179 [5] C. D. West, and A. S. Makas, The spectral dispersion of (2015). birefringence, especially of birefringent plastic sheets, J. [18] E. Stojanova, M. Al-Mahmoud, H. Hristova, A. A. Opt. Soc. Am. 39, 791-794 (1949). Rangelov, E. Dimova, and N. V. Vitanov, Achromatic [6] M. G. Destriau, and J. Prouteau, Réalisation d’un quart polarization rotator with tunable rotation angle, J. Opt. d’onde quasi achromatique par juxtaposition de deux 21, 105403 (2019). lames cristallines de même nature, J. Phys. Radium 10, [19] K. D. Möller, Optics, 3rd Ed., (Univ. Science Books, Mill 53-55 (1949). Valley, 1988). [7] S. Pancharatnam, Achromatic combinations of birefrin- [20] Z. Zhuang, Y. J. Kim, and J. S. Patel, Achromatic linear gent plates. Part I. An achromatic circular polarizer, polarization rotator using twisted nematic liquid crystals, Proc. Indian Acad. Sci. A 41, 130-136 (1955). Appl. Phys. Lett. 76, 3995-3997 (2000). [8] S. Pancharatnam, Achromatic combinations of birefrin- [21] T. Y. Chung, M. C. Tsai, C. K. Liu, J. H. Li, and K. T. gent plates. Part II. An achromatic quarter-wave plate, Cheng, Achromatic linear polarization rotators by tan- Proc. Indian Acad. Sci. A 41, 137-144 (1955). dem twisted nematic liquid crystal cells, Sci. Rep. 8, [9] C. M. McIntyre, and S. E. Harris, Achromatic wave plates 13691 (2018). for the visible spectrum, J. Opt. Soc. Am. 58, 1575-1580 [22] A. Messaadi, M. M. Sanchez-Lopez, A. Vargas, P. Garcia- (1968). Martinez, and I. Moreno, Achromatic linear retarder with [10] A. Ardavan, Exploiting the Poincaré–Bloch symmetry tunable retardance, Opt. Lett. 43, 3277-3280 (2018). to design high-fidelity broadband composite linear re- [23] H. Fabricius, Achromatic prism retarder for use in po- tarders, New J. Phys. 9, 24 (2007). larimetric sensors, Appl. Opt. 30, 426-429 (1991). [11] S. S. Ivanov, A. A. Rangelov, N. V. Vitanov, T. Peters, [24] C. Ye, Construction of an optical rotator using quarter- and T. Halfmann, Highly efficient broadband conversion wave plates and an optical retarder, Opt. Eng. 34, 3031- of light polarization by composite retarders J. Opt. Soc. 3035 (1995). Am. A 29, 265-269 (2012). [25] J. A. Davis, D. E. McNamara, D. M. Cottrell, and T. [12] T. Peters, S. S. Ivanov, D. English, A. A. Rangelov, N. Sonehara, Two-dimensional polarization encoding with a V. Vitanov and T. Halfmann, Variable ultrabroadband phase-only liquid-crystal spatial light modulator, Appl. and narrowband composite polarization retarders, Appl. Opt. 39, 1549-1554 (2000). Opt. 51, 7466-7474 (2012). [26] G. S. Vasilev, and N. V. Vitanov, Complete population [13] E. Dimova, S. S. Ivanov, G. Popkirov, and N. V. Vi- transfer by a zero-area pulse, Phys. Rev. A 73, 023416 tanov, Highly efficient broadband polarization retarders (2006). and tunable polarization filters made of composite stacks [27] B. T. Torosov, and N. V. Vitanov, High-fidelity error- of ordinary wave plates, J. Opt. Soc. Am. A 31, 952-956 resilient composite phase gates, Phys. Rev. A 90, 012341 (2014). (2014).