A polarisation-analysing CMOS image sensor for sensitive polarisation modulation detection

In this study, a polarisation-analysing CMOS image sensor is fabri- cated for sensitive polarisation modulation detection. Although the image sensor with on-pixel polarisers can image the incident polarisation collectively, its sensitivity to a weak polarisation change is not high. With the proposed method, an external polariser is used to enhance the polarisation modulation sensitivity with a polarisation image sensor. The performance of this highly sensitive polarisation image sensor and imaging experiments are evaluated using a ﬂow channel. was ﬂowing in the channel, and a change in polarisation of approximately 0.06 ◦ was analysed when D-menthol was ﬂowing. A similar change in polarisation can also be seen in the imaging results of

Introduction: Polarisation is a fundamental property of light and is useful for sophisticated imaging applications. Polarisation imaging is expected to be used for calculating normal vectors to the surfaces [1], enhancing the contrast under hazy conditions [2]; non-contact fingerprint pattern detection [3]; material classification [4]; and the discrimination of optical isomers of chemical substances [5,6], among other uses.
An image sensor with an on-pixel polariser structure, in which polarisers based on a wire grid structure formed by wiring layers or phonic crystals are placed on the pixels of a CCD or CMOS image sensor, can image incident polarised light collectively [7][8][9][10], although the sensitivity to weak changes in polarisation is not high. In a previous study, weak polarisation detection was achieved by processing data from a large number of polarisation pixels [5,6]. However, this method is unsuitable for imaging.
To detect weak polarisation, a large amount of incident light is required. However, photodiodes in image sensors are small and have a low capacitance, and thus are prone to pixel saturation. In addition, commonmode noise originating from incident light and a power supply causes a poor signal-to-noise ratio (SNR). In this study, we designed, fabricated, and evaluated a sensor to solve these problems, and demonstrated its sensitive polarisation imaging.
Proposed method used by multilayer sensor with a polariser structure: Figure 1(a) shows the conceptual diagram of a polarisation measurement conducted using the proposed method. The on-pixel polariser is fabricated using a metal wiring layer through a CMOS process, as shown in Figure 1(b). As with our previous studies [5,6] , the use of a metal wiring layer during the CMOS process eliminates the need for a post-process to mount the polariser on the pixel. In addition, a pixel array composed of polarisation pixels for different angles is suitable for signal processing.  Thus, polarisation information can be extracted from the obtained image [6,7,[11][12][13]. An external polariser with a high extinction ratio is placed directly above the image sensor at 45 • to the on-pixel polarisers, which are orthogonal to each other on the image sensor, as shown in Figure 1(a). Of these on-pixel polarisers, those in the horizontal direction are called 0 • pixels, and those in the vertical direction are called 90 • pixels.
During the measurement, the observation target is irradiated with linearly polarised light at a polarisation angle orthogonal to the external polariser. This linearly polarised light is rotated by the observation target, and the light enters the sensor. First, the external polariser reduces the linearly polarised light component irradiated to the observation target. Next, the orthogonal polarisation component generated through the polarisation rotation is passed through the external polariser. Thus, the total light intensity is reduced and the pixel saturation can be avoided. Common mode noise is removed by acquiring the difference between orthogonal pairs of polarisers on the chip. These structures make it possible to measure polarisation changes with high sensitivity. The details of this method will be reported elsewhere. Figure 2 show schematic diagrams of the proposed method. In the figure, the boundary of pixel saturation is indicated by a circle. The incident light is shown as a vector along the y-axis. In this case, the polarisation to be measured is the red vector. This vector is larger than the pixel saturation, and thus it cannot be measured. Therefore, the incident linear polarisation component is reduced by an external polariser. As a result, the x-axis component of the incident light is reduced, and pixel saturation can be avoided. Furthermore, because the polarisation component generated by the measurement object passes through, only the polarisation component parallel to the incident light is reduced. The reduced power is compensated by increasing the incident light intensity. Thus, the ratio of the rotated component is increased and the SNR is improved.
After passing through the external polariser, the polarisation is analysed by the wire grid polariser on the pixel. Figure 3(a) shows the simulation results of the 0 • and 90 • pixel outputs. The polariser on the pixel shifts the minimum point, resulting in a difference in output. Figure 3(b) shows a plot of the difference between the two outputs. This difference in value corresponds to the change of the polarisation angle from the reference. The amount of change is almost linear within the range where the sine function is linearly approximated. By normalising this value, the change in polarisation can be estimated from the sensor output signal. In addition, because the difference in value is obtained, the common mode noise component can be reduced.  Polarisation imaging CMOS sensor overview: In this study, we designed and fabricated the sensor using a 0.35-μm 2-poly 4-metal standard CMOS process. Table 1 shows the specifications of the sensor. Figure 4(a) shows a block diagram of the prototype sensor, and Figure 4(b) shows a micro-graph of the sensor and its pixel layout. Figure 4(a) shows a block diagram of the prototype sensor, and Figure 4(b) shows a microphotograph of the sensor and the layout of the pixels. The size of the pixels is 15 × 15 μm and the sensor has a pixel resolution of 160 × 120. Because the orthogonal pixels are arranged next to each other, a polarisation image with a pixel resolution of 80 × 120 pixels can be acquired. In this sensor, an operation for taking the difference between orthogonal pairs of polarisers was created using a CMOS circuit. Figure 4(c) shows the circuit. By installing a differential amplifier in each of two columns, the difference can be taken in parallel, and the process of image acquisition can be omitted, enabling real-time measurements. An external ADC with a resolution of 14-bit was used to digitise the output signal. The difference circuit is based on a switched capacitor circuit. First, Set-SW is turned on and the output of Pixel-OUT-A is sampled. Then, Set-SW is turned off and Sample-SW is turned on. By doing so, the difference in potential between V Pixel-OUT-A and V Pixel-OUT-B can be obtained. Furthermore, by applying a V Bias reference voltage of 1.5 V to the + terminal of the op-amp, it is possible to output the difference between the two-pixel outputs with V Bias as the centre voltage. Therefore, the output from the difference amplifier is as shown in Equation (1).
In the designed switched-capacitor amplifier, capacitors are provided to double the amplification factor. In addition, 80 differential amplifiers are installed as column amplifiers, and the output column of the differential amplifier can be selected by select-SW.

Measurement results and imaging experiments:
To evaluate the function of the sensor, measurements were made by changing the linear A red high-intensity LED with a peak wavelength of 633 nm was used as the incident light source. The polariser used to generate the linear polarisation was a high extinction ratio polariser with an extinction ratio of 10 5 :1. For the external polariser on the extinction ratio sensor, a wire grid polariser film with an extinction ratio of approximately 500:1 was used. The linear polarisation incident on the sensor was varied within the range of ±0.5 • in 0.004 • steps, and 96 frames were acquired at each angle. In this measumrement, we averaged 60 × 60 pixels and 96 frames. The output from the difference amplifier changes linearly with the change in polarisation. In this case, the error was calculated as 2.5 × 10 −3• within the range of ±0.5 • , as shown in Figure 5(c). This value includes the mechanical error of the polariser holder that changes the linear polarisation, and the measurement accuracy of the sensor itself is estimated to be higher than this.
For the polarisation imaging experiments, we set up an experimental system, as shown in Figure 6 The analysed polarisation is opposite for the L-menthol and Dmenthol solutions. In fact, a change in polarisation opposite to the experimental image was detected. Furthermore, the amount of change in polarisation analysed differs depending on the concentration of the solution. Figure 7 shows the results of the temporal change in polarisation when L-menthol and D-menthol were flowed into the channel. After ethanol flowed in the channel, L-menthol flowed, ethanol flowed again to clean the channel, D-menthol flowed, and ethanol flowed again to clean the channel. In this measurement, we averaged a pixel area of 40 × 40 the consecutive frames acquired from among the pixels in the channel, and plotted the moving average among 32 frames on a graph. A change in polarisation of approximately 0.09 • was detected when L-menthol was flowing in the channel, and a change in polarisation of approximately 0.06 • was analysed when D-menthol was flowing. A similar change in polarisation can also be seen in the imaging results of  Figure 6(b,c). The spikes in the plot are caused by air bubbles that are introduced when the liquid flowing in the channel is changed and they pass through the channel.

Conclusion:
We fabricated a polarisation-analysing CMOS image sensor for sensitive polarisation modulation detection. The proposed method avoids the limitations of polarised image sensors such as a pixel saturation and low extinction ratio of on-chip polarisers, and realises a highly sensitive imaging of changes in polarisation. In addition to the flow path imaging demonstrated in this paper, the proposed approach is expected to be applied to real-time high-frequency electric field imaging based on the electrooptic effect [14].