1. Introduction

A polarizing beam splitter (PBS) is an optical device that can separate an incident beam into two diverging beams carrying orthogonal polarization components. These devices are most often fabricated into cube structures to be used at $45^\circ$ incidence, so as to induce a $90^\circ$ angle between the directions of propagation of the two output beams. However, there are applications where a plate PBS is preferred over a cube, particularly when the PBS must perform over a wide range of incidence angles, including at normal incidence. Examples include optics for efficient liquid crystal display backlights [1] and "pancake"-style head-mounted displays [2]. We are investigating an application for a plate PBS as an element of an infrared (IR) beam steering device, where high-efficiency of both the transmitted and reflected polarization modes over a range of incidence angles and azimuth orientations is important. Therefore, we are interested in characterizing the performance of a plate PBS when used over a wide range of beam incidence angles, including normal incidence.

Common plate PBS’s used to separate linear polarization states include multi-layer thin films made with birefringent materials [3] and wire grid polarizers (WGPs) [4,5]. WGPs are first-surface reflectors, exhibiting broad wavelength response and low chromatic dispersion compared to multilayer birefringent polymers [6]. Furthermore, they can perform well over a broad range of incidence angles [4,7], as required for our application. Because the polarization state for transmission and reflection is determined by the orientation of the wires in the grid, the wire grid PBS can be oriented to pass transverse electric (TE), transverse magnetic (TM), or indeed any other orientation of linearly polarized input beam. This is an important difference between wire grid polarizers and MacNeille-type PBS’s that will only pass TM polarization and only reflect TE polarization [8].

In many cases, however, specifications for commercially-available WGPs lack comprehensive information describing their performance as PBS’s. For use as a beam splitter we need data on efficiency and polarization purity of both the transmitted and reflected beams, over a range of incidence angles and for variable pass axis alignment and input polarization orientation. For example, Moxtek provides characterization data for several WGP devices they manufacture [9]. While this tool is very helpful, specifying performance for a classical mount that passes TM and reflects TE polarization, it does not provide information for other polarizer orientations to be considered in the following.

In this paper we report the experimental characterization of IR VersaLight wire grid polarizers (manufactured by Moxtek for Meadowlark Optics, Inc.) to assess their performance as PBS’s at wavelengths near 1550 nm [10]. In this wavelength band the transmission efficiency for normal incidence, as specified by the supplier, is better than $95\%$ with contrast ratio near 5000:1. The reflected beam contrast ratio is specified to be better than 80:1 [10]. It appears that these devices could perform well as PBS’s at normal incidence. The supplier provides representative transmission efficiency and transmission contrast ratio data as a function of wavelength from 1000 nm to 2000 nm, which is reproduced in Fig. 1. However, spectral data for the reflection efficiency is not provided, nor is there data on the dependence of transmittance or reflectance on the incidence angle of the input beam. We therefore characterized these devices for both transmission and reflection efficiency versus incidence angle (polar angle $\theta$), where $-50^\circ \leq \theta \leq 50^\circ$. We performed this measurement for three orientations of the polarizer such that the azimuthal angle $\varphi$ was rotated to $\varphi =0^{\circ}$, $\varphi =45^\circ$ and $\varphi =90^\circ$, where $\varphi$ is the angle between the pass (high transmission) axis of the polarizer and the rotation axis for the angle of incidence ($\theta$). For a WGP, the angle $\varphi$ is also the angle between the grating vector ($\vec {K}$) and the normal to the incidence plane ($\hat {y}$). For each configuration of the polarizer under test, the input linear polarization state is configured to both the pass (high transmission) state and block (high reflection) state. The reflectance and transmittance are then measured for these cases.

 

Fig. 1. Spectral data provided by Meadowlark Optics, Inc. for the IR VersaLight polarizer, showing the transmittance value and contrast ratio at normal incidence. For a wavelength of 1550 nm, these are shown to be approximately 96% and 5000, respectively. [10]

Download Full Size | PDF

2. Method

A polarization-resolved reflection/transmission measurement system was built to characterize the PBS performance of the commercial WGPs, as shown in Fig. 2. We employed a 3 mW fiber-coupled tunable laser source centered at 1550 nm wavelength. This laser output was coupled into a single mode fiber and collimated into a $\sim 1$ mm diameter beam. The full-angular width (as defined in [11]) of this beam was determined by fitting the measured beam width as a function of axial position to the typical Gaussian beam hyperbolic profile, and was found to be $<0.15^\circ$. The laser output polarization angle is maintained by a precision linear polarizer with a nominal contrast ratio of 10,000:1 (Newport 10LP-NIR) and a polarizing orientation along $\hat {y}$, normal to the optical bench surface. The angle of linear polarization is then rotated using a half-wave plate (ThorLabs WPH10M-1550), in which the angle $\delta$ describes the electric field alignment relative to $\hat {y}$ after the half-wave plate. Following the half wave plate, a pick-off wedged plate beam splitter (ThorLabs BSF2550) was used to redirect a portion of the beam into a reference power meter. Note that the use of the half-wave plate and beam pick-off do slightly decrease the degree of linear polarization (DoLP) of the beam used to measure the device under test, limiting the reliability of small cross-state measurements (e.g. transmittance in the high reflectivity configuration). This degradation is discussed in the next section and is shown to be insignificant for this current characterization effort, which focuses on PBS efficiency. For measurements such as extinction ratio that require high resolution of the cross-state transmittance, the DoLP could be maintained by adding a high-contrast clean-up polarizer between the above components and the device under test.

 

Fig. 2. Characterization set-up for the WGP used to measure transmittance and reflectance with respect to incidence angle, $\theta$. System allows for control of the pass axis orientation of the WGP ($\varphi$) and the angle of the incident linear polarization state ($\delta$).

Download Full Size | PDF

For this device characterization set-up, a WGP was centered in the beam path with rotational control over the horizontal-plane incidence angle, $\theta$, as well as the polarizing orientation angle, $\varphi$, for which $\varphi =0^{\circ}$ denotes a pass polarization axis aligned with $\hat {y}$. The light passed through the WGP was monitored using the transmittance power meter while the reflected beam power was monitored using a separate reflectance power meter. For ease of describing the system, the optical bench surface is defined to be the horizontal plane with normal $\hat {y}$. For the remainder of this discussion, the terms horizontal and vertical will be used to indicate $\hat {x}$ and $\hat {y}$ directed polarization components, respectively. Additionally, the term diagonal will be used to indicate a linear polarization orientated at $\delta = 45^\circ$ (having equal projection onto $\hat {x}$ and $\hat {y}$). Finally, $\hat {z}$ will be used to represent the forward propagating beam path direction (left to right in Fig. 2).

A linear polarizer will exhibit either high transmission or high extinction/reflection depending on the pass state polarization axis orientation, $\varphi$, relative to the incident polarization orientation, $\delta$. If these two angles are equal in our system, then the WGP pass state aligns to the incident polarization state and maximum transmission occurs. If these two angles differ by $90^{\circ}$, then the WGP assumes a block state in which maximum reflection occurs. To achieve either of these states, the WGP was first set to a desired $\varphi$. The incident polarization angle $\delta$ was adjusted using the waveplate to achieve maximum extinction. For block-state testing, the system remained in this orientation, while in pass-state testing the waveplate was rotated to achieve maximum transmission through the polarizer.

In a preliminary test with a fixed source wavelength of 1550 nm, the incidence angle, $\theta$, was swept from $-9^\circ$ to $9^\circ$ with a resolution of $0.1^\circ$. For this measurement, both $\delta$ and $\varphi$ were oriented to $0^\circ$, achieving a pass state (for TE incident polarization). Transmission and reflection measurements were normalized to incidence power values (no WGP) collected before each test. Rather than relying on the cross-calibration between detectors, the normalization process for transmittance (T) was performed solely on the transmittance power meter head, while reflectance (R) was performed solely on the reflectance power meter head. To account for any fluctuation in the source power, a pick-off was used to obtain a reference value during the preliminary incidence power measurement as well as at each angular step for a given test. This ensured that source power variation did not affect the computed transmittance and reflectance values.

In this preliminary angular sweep we observed significant Fabry-Perot fringes as shown in Fig. 3(a). Using the source wavelength of 1550 nm, we find that the dominant fringes correlate to a 1mm thick glass etalon which we attribute to the Corning Eagle XG glass substrate supporting the WGP [12]. To suppress the narrowband Fabry-Perot fringes, we explored two methods of averaging: averaging data as a function of wavelength (wavelength averaging), and averaging data as a function of incidence angle (angle averaging). Note that an antireflection coating could also be applied to the back surface of the device to minimize the effect of these fringes, however this possibility was not explored in this work.

 

Fig. 3. (a) Transmittance versus incident angle, $\theta$, of WGP in a pass state at 1550 nm wavelength for $\delta = 0^\circ$ and angular resolution $0.1^\circ$. (b) Transmittance and reflectance versus wavelength for WGP with diagonal pass polarization axis. WGP placed at slight incident angle ($5^{\circ }$) for wavelength sweep from 1545nm-1555nm (solid lines). Computed average values for five trials (dashed lines) across optimized wavelength band 1546 nm -1554 nm (shaded region).

Download Full Size | PDF

To establish the window for wavelength averaging, we began by sweeping the laser from 1545 nm to 1555 nm in 0.01 nm steps while measuring the transmittance and reflectance, as seen in Fig. 3(b). For accurate reflectance measurements, the system was limited to a minimum incidence angle of $|\theta |=5^\circ$ to ensure the reflectance power meter did not obstruct the incident beam path. For the data in this figure, the system was configured with $\delta =45^\circ$ and $\varphi =0^\circ$. This allowed us to monitor nearly balanced transmitted and reflected power. Figure 3(b) presents a wavelength-dependant oscillating pattern composed of two dominant frequencies. The higher of these two frequencies (0.8 nm period) is attributed to the previously identified 1mm etalon, while the lower frequency component (4 nm period) is attributed to a thin window in the nominally identical transmission and reflection power sensors (Newport 918D-IG-OD1R) as well as the reference power sensor (Newport 918D-IR-OD1R) [13]. Following this initial test, four additional sweeps were completed with slightly perturbed incidence angles, shifting the position of the fringe pattern so that an optimized wavelength band with minimal mean biasing could be chosen. It was found that averaging across 8 nm (from 1546 nm - 1554 nm) with 0.2 nm steps reduced the maximum mean bias to approximately $\pm 0.2\%$, illustrated by the dashed horizontal lines representing each of the five wavelength averaged trials in Fig. 3(b). Hence, averaging over an 8 nm wavelength window suppressed etalon effects while also reducing the sensitivity of the system to the angular alignment of each detector.

For angular averaging, the incidence angle was adjusted by hand to locate a nearby maximum and minimum pair around each desired angle $\theta$. At normal incidence this required adjustments of up to $2^\circ$ to reach the neighboring extrema. As shown in Fig. 3(a), the angular frequency of the etalon fringes increase as a function of incidence angle, requiring only fine adjustments ($<0.5^\circ$) for angles $|\theta | > 10^\circ$. The average reflectance or transmittance was determined using the mean of the collected maximum and minimum measurement for each angle $\theta$. We confirmed that angle averaging and wavelength averaging were each effective at suppressing the etalon fringe effects, and produced results that were equivalent within measurement uncertainties of approximately 1%. Wavelength averaging had the advantage of allowing our measurements to be made at a selected fixed angle of incidence, while angle averaging relied on transmittance or reflectance changing slowly over the small angular range needed to find a nearby etalon maximum and minimum.

Following these tests, the WGP used for the preliminary measurements (polarizer 1) was examined to determine its performance as a PBS at varying incidence angles. Both transmitted power and reflected power were recorded in a system state $(\varphi,\delta )$ where $\varphi \in [0^{\circ }, 45^{\circ }, 90 ^{\circ }]$ and $\delta \in [ \varphi, \varphi + 90^{\circ }]$. For clarity in describing various polarizer orientations, we will use "pass state" to refer to the configuration with $\delta = \varphi$ and "block state" to refer to the configuration with $\delta = \varphi +90^{\circ }$. Representative results for polarizer 1 with a vertical (V) orientation $(\varphi = 0^{\circ },\delta = [0^{\circ }, 90^{\circ }])$ are shown in Fig. 4. This configuration correlates to pass state maximum transmission of transverse electric (TE) incident polarization and block state maximum reflection for transverse magnetic (TM) incident polarization. Figure 4(a) presents the pass state transmission and block state reflection as a function of incidence angle. These modes could also be described as the "desired" beam efficiencies. Conversely, Fig. 4(b) presents the block state transmission and pass state reflection as a function of incidence angle, which similarly could be referred to as the "undesired" beam efficiencies.

 

Fig. 4. Reflectance and transmittance of polarizer 1 as a function of incidence angle $\theta$ – (a) desired beam efficiencies: pass state (TE) transmittance and block state (TM) reflectance; and (b) undesired beam efficiencies: pass state (TE) reflectance and block state (TM) transmittance. Polarizer orientation is set vertical $\varphi = 0^\circ$.

Download Full Size | PDF

This data shows a pass state normal-incidence transmittance of around 92%, approximately 4% below the value given in the data sheet and reproduced in Fig. 1 [10]. The transmittance decreases as the incidence angle increases, staying within 1% of the normal incidence value until angles greater than $\pm 30^\circ$; it reached approximately 85% at an incidence angle of $\pm 50^{\circ }$. The block state reflectance shows little sensitivity to incidence angle over the entire range as seen in Fig. 4(a), showing a near normal incidence reflectance value of 92% that gradually decreases to slightly more than 90% as the incident angle approaches $\pm 50^{\circ }$. Comparing Figs. 4(a) and 4(b), we find that the transmittance and reflectance for the pass state are coupled; the transmittance decreases as the reflectance increases at angles above approximately $\pm 30^\circ$. For broad angular performance, we hope to maintain high extinction of the "undesired" beam as well as low loss of the "desired" beam across a wide range of angles. In this orientation (the TE pass state) this polarizer demonstrates the expected characteristics of a PBS, albeit with slightly decreased performance at incidence angles above $\pm 30^\circ$.

3. Results and discussion

We applied the method described above to characterize five Meadowlark IR Versalight WGP devices, which will be referred to as polarizers 1-5. Polarizers 1-3 are 0.5" (12.7 mm) in diameter (VLR-050-IR) and polarizers 4 and 5 are 1" (25.4 mm) in diameter (VLR-100-IR). The diameter had no discernible effect on the performance of the devices in this set-up, and therefore will not be distinguished in the discussion that follows.

To analyze each device, the measurement system is configured to provide a linearly polarized input beam inducing a pass (transmissive) state (Figs. 5(a) and 5(c)) or block (reflective) state (Figs. 5(b) and 5(d)) in the WGP under test. For both pass and block states the pass polarization angle, $\varphi$, was oriented vertically, horizontally, and diagonally as described previously. At each of these orientations, the incidence angle is varied from $-50^\circ$ to $50^\circ$ in $5^\circ$ increments. At each incidence angle, reflectance and transmittance data are separately averaged to suppress etalon effects. Wavelength averaging over an 8 nm wavelength range was used for polarizers 1, 4 and 5, while angle averaging was used for polarizers 2 and 3. The results for each of these polarizers is plotted (light gray) in addition to the average over the five samples for each pass polarization angle (marked/colored) in Fig. 5. Reflectance measurements were not recorded below $\pm 10^\circ$ incidence, to ensure the detectors did not obstruct the incident beam path. Consequently, a gap has been left in the averaged reflectance data at angles smaller than this value. The T+R sum is also plotted for both the pass state and block state (dotted magenta lines) for angles where both a T and R measurement were recorded.

 

Fig. 5. Measured (a) pass state transmittance (b) block state reflectance (c) pass state reflectance and (d) block state transmittance as a function of incidence angle for five different polarizers. The average reflectance and transmittance are indicated by lines in color, with the markers referring to pass polarization angle, $\varphi$. The data for each individual device is displayed in the gray curves. Magenta curves in (a) and (b) indicate the sum of the average transmittance and reflectance for each polarizer orientation, indicated by the markers. Note that not every polarizer was tested at every incidence angle for every system state, so the average transmittance and reflectance shown were calculated from the available data.

Download Full Size | PDF

The polarizers demonstrated a pass state transmittance between 92% and 95% at normal incidence, with an average just below 94%. This is slightly lower than the typical transmittance of 96% at 1550 nm wavelength published by the manufacturer [10]. The pass state near normal incidence reflectance was found to vary $\pm 0.5\%$ around the center value of 3.5%. In the block state, the reflectance near normal incidence varied from 91% - 96% between devices, with an average near 93%. The measured block state transmittance at normal incidence was below 0.15% for all devices, with the average below 0.1%. As stated above, use of the half-wave plate to rotate the state of polarization of the input light and use of the pick-off plate to monitor the input beam intensity slightly degrades the DoLP of the incident laser light. For the horizontal and diagonal input polarization orientations, we find the contrast ratio of the transmitted light slightly below the manufacturer specifications, and attribute this result to the limitations of the characterization system. Thus, the manufacturer stated contrast ratio could only be verified for horizontal polarizer orientation $\varphi = 90^\circ$, in which vertical input polarization $\delta = 0^\circ$ is blocked. The data collected in this orientation produced a transmission contrast ratio near 6000 near normal incidence, which decreased to just below 5000 at incidence angle of $\theta = \pm 50^\circ$. This data agrees with the nominal contrast ratio value of 5000 provided from the manufacturer [10]. From the data provided in Fig. 5, it is clear that the transmission contrast ratio is much higher than the reflection contrast ratio. At near normal incidence ($\theta = \pm 10^\circ$) we calculate a reflective contrast ratio of approximately 27. This result is consistent with physical characterization and simulations performed by other investigators addressing the PBS capabilities of these grating structures [4].

In the pass state at larger incidence angles, differences on the scale of 5% to 10% were seen in both the transmittance and reflectance, depending on the azimuthal orientation of the polarizer. This variation is attributed in part to the Fresnel interaction at the substrate-air interface on the back surface of the WGP, an effect that has been noted previously [14]. In addition, the surface supporting the wire grid also exhibits a Brewster effect with the angle for minimum TM reflectance dependent not only on the substrate material but also on the dimensions of the metallic grating [4], and may therefore be different from Brewster’s angle for the glass-air interface on the back side. For both surfaces, however, the transmittance for horizontal incident polarization (TM) is expected to increase and the reflectance will decrease as the angle of incidence grows toward Brewster’s angle for these surfaces. This $\varphi$ orientation dependency of transmission and reflection for the pass state at oblique incidence is not observed in the block state. In the case of the pass state, each pair of curves from the transmission and reflection complemented one another and summed to a total conserved beam power of about 97% at near normal incidence, decreasing to 96% at $\pm 50^\circ$ (Fig. 5(a)). In comparison, the total sum of beam powers from the block state only reached values close to 92% (Fig. 5(b)), as might be expected from a metallic mirror [11].

In many applications, light will be incident on the PBS from either side. The plate wire grid polarizing beam splitter is not symmetrical, with the wire grid lying on one side of the substrate. It is therefore important to characterize the PBS performance with light incident from either direction. We chose to test this on a single polarizer (polarizer 3), by performing incidence angle measurements for the same three polarizer orientations used in Fig. 5. The comparison of the two incidence directions is shown in Fig. 6 for both forward incidence (grating first, solid lines) and backward incidence (substrate first, dashed lines).

 

Fig. 6. Comparison of pass-state transmission propagating forward (grating first) and backward (substrate first) through polarizer 3 as a function of incidence angle.

Download Full Size | PDF

The direction of propagation through the device provided insignificant (<1%) deviations on vertical and horizontal pass state transmission. However, for the diagonal orientation, the backward propagating beam (WGP on second face) showed a sharper decrease at higher incidence angles than its forward counterpart (WGP on first interface). This resulted in a maximum deviation due to the direction of propagation of approximately 3% at a $\pm 50^\circ$ angle of incidence. This is potentially explained by the different propagation media before the wire grid (free-space vs Eagle XG Glass), which should result in both a different local incidence angle on the polarizer structure, as well as different projections of the incident electric field into components parallel and perpendicular to the wire grids. In all, it was determined that these devices work at a similar level of performance with respect to their pass-state transmission in the forward or backward orientation.

4. Conclusion

We have described a method of measuring the performance of a WGP functioning as a PBS over a range of incidence angles ($-50^{\circ } \le \theta \le 50^{\circ }$). We collected transmittance and reflectance data from 5 samples of the Meadowlark Versalight IR WGP at varying incidence angles and polarizer orientations. For the pass state performance, a transmittance of approximately 94% $\pm$2% over the five tested devices was observed for incidence angles up to $\pm 30^\circ$. At larger incidence angles, the transmittance depended on the orientation of the wire grating with respect to the incidence plane. Vertical orientation (TE pass-state) showed the largest decrease in transmittance at higher incidence angles. The orientation dependency is explained by Brewster’s effect, and suggests performance will be enhanced at higher angles for TM pass state, which was observed. In the block state, the reflectance was observed to peak at 92% $\pm$2% around $\pm 10^\circ$ incidence (the minimum angle measured) with a slow decrease down to 90% at incidence angles of $\pm 50^\circ$. Unlike the pass state, Brewster’s effects were not discernible in the measured block state data. The high transmission contrast ratio of these WGPs was confirmed by block state transmission measurements for vertical (TE) input polarization. The input polarization purity limits our measurement of contrast ratio for horizontal (TM) and diagonal (equal TE and TM projection) incident polarization orientations. The measured pass state reflectance did not drop below 3.5% at normal incidence and was similarly dependent on the polarizer azimuth orientation at oblique incidence like its coupled pass-state transmittance. This resulted in a reflective contrast ratio much lower than its transmissive counterpart, never reaching above 27:1 near normal incidence and significantly varied at incidence angles above $\pm 30^\circ$.

Due to limited sample size, we cannot accurately summarize the statistics of the performance spread of these devices. Our data found a 3% spread between the five tested devices with respect to their normal incidence pass state transmittance. Additionally, the block state near-normal incidence reflectance of these devices varied across a range of 5% within our sample group.

Overall, our characterization of the commercial WGPs found good consistency in the performance of the five individual WGPs tested, as well as good agreement with the nominal performance specifications provided by the vendor. In addition, these devices were found to demonstrate high efficiency in both pass state transmittance and block state reflectance over a range of incidence angles and with differing input polarization configurations, confirming their usability as PBS’s with variable incidence angle and polarizer orientation.