Open Access
Add to BibSonomy
Abstract
We describe a mini-endoscope design that uses a new type of electrically tunable liquid crystal lens array enabling the dynamic increase of spatial resolution by adjusting the working distance in various zones of interest over a relatively large field of view (FoV) without mechanical movement. The characterization of the system is performed by using uniform fluorescent films, fluorescent micro spheres and a tissue sample expressing the fluorescent calcium indicator GCaMP6s. Lateral resolution of up to 2 µm over the FoV between 300 µm - 400 µm is experimentally demonstrated.
© 2024 Optica Publishing Group under the terms of the Optica Open Access Publishing Agreement
1. Introduction
The development of miniature microscopes, also known as mini-endoscopes, has opened new avenues for brain study [1–3]. Their use has become increasingly popular for studying brain activity at cellular and network levels in different brain regions of freely behaving animals. Currently, it is a preferred method for researchers due to its ability to provide a clear and objective look at brain activity. This technique enables longitudinal investigations while animals carry the microscope on their head and perform a wide range of behavioral tasks.
Significant developments have been reported during the last decade to increase the functionality of these mini-endoscopes, including their capability of depth scanning by using approaches such as mechanical shift [4], liquid lenses [5], or electrically tunable liquid crystal (LC) lenses (TLCLs) [6,7] based on nematic LC (NLC) materials [8].
Among these approaches, TLCLs are very attractive thanks to several advantages procured by NLC materials. In fact, NLC based spatial light modulators have already been used to enable many adaptive optical designs [9,10]. However, their large size and high cost prohibit their use in mini-endoscopes. In addition, their control is not simple, and the optical path difference (OPD), they can generate, is rather small. In the case of lenses with one clear aperture (CA), this OPD limitation is partially overcome by using relatively thick LC layers ($d_{LC} = 50\; {\mathrm{\mu}}{\textrm m}$) [6,7]. However, the maximum achievable optical power (OP) variation range of TLCLs with large CA is still limited by the optical birefringence of the NLC, $\Delta n = n{\parallel }- n_{\perp }$, and by the fact that their maximal OP is inversely proportional to the square of the CA [11]:
Thus, the capability of changing the focus (the inverse of OP) is reduced dramatically when we consider TLCLs with larger CAs, which are typically needed in mini-endoscope design with a larger field of view (FoV). For instance, the Finchscope mini-endoscope [12] (which features a larger field of view) utilizes a 1.8 mm GRIN lens as its objective. This implies that a TLCL with a numerical aperture of at least 1.8 mm would be required in front of the objective for focusing adjustments, in contrast to the TLCL with a numerical aperture of 0.5 mm used in our previous model (which features a smaller FOV and uses a 0.5 mm objective) [7]. Using a typical birefringence value ($\Delta n =0.2$) for an NLC, we obtain an OP of 350 diopters (D) for a CA of 0.5 mm, but it decreases to 25 D for a CA of 1.8 mm.
At the same time, many brain investigations require the view of an ensemble of neurons, and it is highly desired to be able to "resolve" fine details of individual neuronal activities in specific zones of interest, (see e.g., [7]). Traditional approaches, based on classical "single-aperture" TLCLs, are not efficient here due to the above-mentioned inversed quadratic dependence.
Recently, a new type of TLCL was introduced [13] for ophthalmic or augmented/virtual reality applications [14], which enables operation with optical systems having very large CA. This solution was inspired by foveal human vision. Namely, the activated area here is local, but can dynamically be moved over the entire CA of the device to image the zone of interest at a given moment.
In the current study, we describe the incorporation of the above-mentioned foveal TLCL in a scheme that should enable the fabrication of a mini-endoscope that can address the above-mentioned problem by enabling relatively large FoV and high spatial resolution in dynamically variable zones of interest. Preliminary performance characteristics of this prototype are studied by using a fluorescent uniform film and fluorescent beads. Prototype characteristics include, among others, a lateral resolution of 2.12 $\pm$ 0.04 ${\mathrm{\mu}}{\textrm m}$ over a FoV of approximately 400 ${\mathrm{\mu}}{\textrm m}$, and a magnification of 7.2. An example of an application of the developed system is also presented with real brain tissue slices.
2. Materials and components
The mini-endoscope’s design that has inspired our work was previously described in [7]. Like most mini-endoscope designs [1,15], it is composed of two main parts: one referred to as the illumination path, which delivers the excitation light to the sample under study, and the other one as the emission path, which collects the sample’s fluorescent emission and directs it onto a camera recording the image of the sample. The illumination path consists of an excitation source, the light of which is delivered to the mini-endoscope through a multimode fiber and an excitation filter. The emission path includes an objective, an imaging lens, an emission filter, and a camera. A unique feature of that design was its subcellular resolution, a capability that was previously exclusive to larger two-photon systems. However, this feature was achieved by using a small diameter gradient index (GRIN) rod-lens objective (with 0.5 mm diameter), resulting in a smaller FoV.
In the design presented here, we combine the foveal TLCL with a larger objective lens to achieve a larger FoV compared to the previous setup. The design of the foveal TLCL as well as its operation principle were recently reported in details [14]. It is made of two indium tin oxide (ITO) coated glass substrates (inner surfaces only), separated by 40 ${\mathrm{\mu}}{\textrm m}$ (using spacers) to form a sandwich, which is filled with an NLC. The ITO electrodes are patterned into a continuous (from one edge to the other of the same substrate) linear serpentine shape. The serpentine ITO lanes have a width of $w = 5\; {\mathrm{\mu}}{\textrm m}$ and are separated with a gap of $g = 5\; {\mathrm{\mu}}{\textrm m}$. A limited number (20 per surface) of individually controllable external electrical contacts allow a) to choose the position of the local lens to be excited, b) to vary its CA (from 100 ${\mathrm{\mu}}{\textrm m}$ up to 2 mm), and c) to vary dynamically its OP. Standard planar alignment of the NLC is obtained by coating (using a spin coater) both substrates with a polyimide layer and by rubbing their dried surfaces in "anti-parallel" directions [16]. Two substrates are then rotated at 90 degrees (one with respect to the other) to obtain main linear lanes of serpentine electrodes (of two substrates) to be aligned orthogonally to each other.
The preliminary optical characterization of our device (including the foveal TLCL) was made by using fluorescent uniform films (doped by Coumarin 7) and 1 ${\mathrm{\mu}}{\textrm m}$ diameter fluorescent beads with a peak emission at 520 nm (FS03F, Bangs Laboratories). Then, we used biological tissue samples. All experiments were approved by the Animal Welfare Committee of CHU de Québec and Université Laval in accordance with the Canadian Council on Animal Care policy. As previously described [7], an AAV2.9-CAG-Flex-GCaMP6s (Molecular Tool Platform of the Université Laval) was injected in the motor cortex of 2-month-old male or female VGluT2-IRES-Cre transgenic mice (The Jackson Laboratory, RRID: IMSR_JAX:016963). 3 weeks after injection, brain tissues were harvested, cut at a cryostat, and mounted on slices for imaging.
3. Mini-endoscope and experimental procedures
The preliminary experimental setup (see Fig. 1), used for the demonstration, is similar to a modified fluorescence microscope. A linearly polarized laser diode (operating at 455 nm) is used as an excitation source. The light source will be conveyed to the mini-endoscope using fibers in the miniaturized set-up. The excitation beam passes through a half-wave plate (HWP) to control the input polarization angle and to align it with the ground state orientation plane of the local optical axis of the NLC, described by the director vector (averaged local orientation of long molecular axis of the NLC [8]). Then a beam expander is used to illuminate the entire aperture of the foveal TLCL (eFoveal) placed on the excitation path. The L1 lens forms a telescope with the objective lens to project the illumination plane (ePlan, located just after the eFoveal) onto the sample after being reflected by the dichroic mirror. The GRIN lens (diameter 0.5 mm, length 6.0 mm) is used as an implant (to be inserted into the animal brain). It is serving as a relay lens prolonging the optical path.
Fig. 1. Schematic set-up of a mini-endoscope design with two foveal TLCLs. In the excitation path, a foveal TLCL (eFoveal) is added to locally focus light on the sample, resulting in a spotlight effect. In the imaging path, a second foveal TLCL (iFoveal) is added allowing the local adjustment of the focal plan on the sample.
For the imaging path, the fluorescent light, emitted by the sample, passes through the same relay GRIN lens and the objective lens, which operates as a telescope with the lens L2. For this particular demonstration, the foveal TLCL had only one NLC layer, and thus, it could affect (focus) only light polarized in one direction (the extra-ordinary polarized light) [17]. This is the reason why a polarizer was used to polarize the fluorescent light [17] and to match its direction with the ground state director of the NLC of the imaging foveal TLCL (iFoveal), which was placed before the image plane (iPlan) of our system. We have employed a 1:1 relay system to project the image plane onto the CMOS camera (DCC1645C, Thorlabs) without additional magnification. This relay system was added to increase the available space between iFoveal and iPlan to further accommodate the emission filter and the camera.
As explained in [14], the foveal lens is driven by two superposed electric signals, one with high frequency (HF) and another one with low frequency (LF). In our case, for both foveal lenses and through the entire experiment, the root mean square (RMS) amplitude of the HF signal was changed (to achieve the desired variations of the OP), but its frequency was kept constant at 1000 Hz. The RMS amplitude of the LF signal (2.0 V, AC sin shaped) as well as its frequency (at 50 Hz) were also kept constant.
4. Experimental results
4.1 Characterization of the excitation path
Using the eFoveal, we were able to dynamically control the light intensity distribution over the sample (a sandwich cell filled by coumarin 7 to obtain a uniformly fluorescent sample). By applying the appropriate excitation signal on chosen electrodes of the eFoveal, it was possible to focus the excitation light on the areas of interest everywhere in the FoV, locally (selectively) increasing the intensity of excitation (Fig. 2 for an example with 3 distinct positions).
Fig. 2. Examples of selective excitation of different areas of interest, demonstrated with a cell filled by coumarin 7. The image a) is obtained when the eFoveal is switched OFF, images b), c) and d) demonstrate the shift of the excitation spot by the eFoveal. All images were recorded with the same exposure parameters. Scale bar represents 20 ${\mathrm{\mu}}{\textrm m}$.
Moreover, as we can see from Fig. 2 and Fig. 3, the spot intensity can be controlled by tuning the focus. Namely, without the activation of the eFoveal, the surface of the sample was more or less uniformly illuminated (see, Fig. 2(a)). However, when the eFoveal was activated, we were able to increase the excitation intensity of light in specific areas (arbitrarily chosen), giving rise to a spotlight effect. In addition to this, there was also an undesired cross-like illumination, which may be eliminated later (also explained in [14]). Indeed, we observed slightly increased "cross-shaped" lines of light intensity along the activated contact electrodes due to the index variations around these electrodes. It was more visible at low or high voltages, when the light was focused in front or behind the sample plane, as the relative intensities between the center and periphery were similar. Adjusting the HF voltage (applied to the TLCL) allowed us to control the intensity of the spotlight.
Fig. 3. Demonstration of selective excitation of the same area of interest with different intensities, controlled by the driving voltage of the eFoveal. Coumarin 7 doped cell is used for this demonstration. All images are recorded with the same exposure. The voltage was increased from 1.0 V to 2.8 V, with the intensity of the spot being measured and plotted (on the right). The scale bar represents 50 ${\mathrm{\mu}}{\textrm m}$.
As the spot became narrower, it also became brighter, as seen in Fig. 3, with a maximum at 2.0 V of the HF driving voltage. For this voltage, most of the rays, passing through the TLCL, cross on the sample plan. For this reason, the main increase was observed in the central section of the area (see the inset on the right, around 150 ${\mathrm{\mu}}{\textrm m}$ in Fig. 3). The obtained dynamic increase of the excitation intensity in the zone of interest allows the reduction of the overall level of excitation and, consequently, a significant reduction of the undesired fluorescence generated by the excitation of the surrounding zones.
As we have already mentioned, the eFoveal not only allowed us to choose the position and the intensity of the spot, but also its size (Fig. 4). To achieve this, we need to increase the local CA of the eFoveal by activating a larger separation of control electrodes [14]. However, increasing the size of the spot inevitably reduces the number of positions available on the FoV and creates overlaps.
Fig. 4. Demonstration of the dynamic change of the spot size, controlled by varying the clear aperture of the foveal lens. The clear apertures are a) 0.3 mm, b) 0.6 mm and c) 1.2 mm. Coumarin 7 doped cell is used. Exposure parameters have been modified for image clarity. The scale bar represents 50 ${\mathrm{\mu}}{\textrm m}$.
4.2 Characterization of the imaging path
To demonstrate the improvement of imaging resolution from the local focal length adjustment (Fig. 5), we used 1 ${\mathrm{\mu}}{\textrm m}$ diameter fluorescent beads dispersed in antifade reagent. Comparing the distances between the beads in Fig. 5, the magnification changed when using the iFoveal with a different driving voltage (for example by a factor of $\approx 0.80$ with a driving voltage of 2.0 V, see also Fig. 8 for other driving voltages). At the bottom of the figure, we quantitatively demonstrate the effect of the iFoveal by using the cross-section histograms of one of beads along the small tilted white line (shown in pictures). Let us emphasize again that the proposed "foveal" approach allows us to overcome the above-mentioned difficulties encountered with TLCLs in this specific context of mini-endoscope application.
Fig. 5. Demonstration of the resolution optimization within the area of interest. On the left, the sample is in focus. In the center, the sample is mechanically shifted approximately by 35 $\mathrm \mu$m to be out of focus. On the right, the iFoveal is locally activated. The scale bar represents 20 ${\mathrm{\mu}}{\textrm m}$. Bottom graphs show the histogram across one bead.
This can be explained by using basic geometrical optics equations. When activating the eFoveal, the effective focal length of the doublet L2 + iFoveal is modified. Using the thin lens equation and the equation of the magnification [18], the magnification of the equivalent lens formed by L2 + iFoveal depends on its focal length $f$ as:
With $p$ being the distance between the equivalent lens (formed by the doublet L2 and the iFoveal). In our experiment, $p$ was fixed, but $f$ changed because of the iFoveal lens. Using the Eq. (2), if the focal length decreases, the magnification also decreases. For a doublet, the effective focal distance is given by:
where $d$ is the distance between the two lenses, $f_{L2}$ the focal length of the lens L2, and $f_{Fov}$ the focal length of iFoveal. $d$ and $f_{L2}$ are fixed, only the iFoveal will have a varying focal length ( $f_{Fov}$). The new magnification will depend upon the focal set to the iFoveal.Theoretically, using the smallest possible lens on the foveal lens ("unit" element on the array) would result in the possibility of $19 \times 19 = 361$ lens positions. However, such small lenses ($CA = {0.1}\; {\textrm {mm}}$) tend to introduce more aberrations and are more difficult to work with because of their very high OP ($OP_{max} \approx {8000}\; {\textrm D}$). Thus, for our experiment, we used a lens that was four times the size of the smallest lens ($CA = {0.4} {\textrm {mm}}$). This reduced the number of possible positions to $15 \times 15 = 225$, with positions that overlapped (the step being 0.1 mm). An example of the iFoveal activated with that configuration at three different positions of the FoV can be seen on Fig. 6.
Fig. 6. Activation of different lens positions with the iFoveal is demonstrated by using fluorescent beads with a diameter of 1 ${\mathrm{\mu}}{\textrm m}$. The scale bar represents 20 ${\mathrm{\mu}}{\textrm m}$.
The ability to retrieve information from small biological objects, which are out of focus, was demonstrated also by using a brain slice with neurons expressing the fluorescent calcium indicator GCaMP6s (Fig. 7). The first column of images depicts neurons mechanically brought in focus, while the second column shows displacement by 35 ${\mathrm{\mu}}{\textrm m}$ from the original point of focus. In the third column, the spatial information was recovered in the activation area (dotted circle) of the iFoveal. This is better illustrated, on Fig. 7, by using the cross section histogram of the intensity of emission by a biological object. Using this local adjustment of the focal length allows us to minimize the undesired fluorescence noise generated by the surrounding area (being now out of focus).
Fig. 7. Example of local focus adjustment allowing to recover a small biological object from a sample moved 35 ${\mathrm{\mu}}{\textrm m}$ away from the focal plan. Brain tissue expressing the fluorescent calcium indicator GCaMP6s is used. The HF voltage was 2.0 V.
Fig. 8. Influence of the optical power of the iFoveal on the image quality. Tissue expressing the fluorescent calcium indicator GCaMP6s is used.
The obtained results are summarized in the Table 1. It is worth mentioning that the FoV with the iFoveal is only limited by the available components at the time as its size in the FoV depends on the lens used to project the iFoveal on the sample.
Table 1. Impact of the iFoveal on the system properties.
We were able to shift the focal plan up to 244 ${\mathrm{\mu}}{\textrm m}$ by increasing the driving voltage of the iFoveal to 5.0 V (Fig. 8). This helps to recover high spatial frequency details and distinguish two different neurons. Nevertheless, the contrast was reduced compared to smaller focal shifts because of higher aberrations. The increase of aberrations with the rise of the driving amplitude is well known for TLCLs [14].
4.3 Simultaneous activation of illumination and imaging foveal lenses
Figure 9 shows the impact of simultaneous use of both eFoveal and iFoveal, aiming two neurons in the FoV (identified by two white arrows on the left). At first (in a)), the sample was placed in focus (by mechanical movement), then it was moved mechanically 35 ${\mathrm{\mu}}{\textrm m}$ out of focus in b). The activation of the iFoveal allows to recover fine details in d). In the images e), the sample was back in the focal plan illustrating the capacity of the eFoveal. In f), the sample was moved again 35 ${\mathrm{\mu}}{\textrm m}$ out of focus and both eFoveal and iFoveal were activated. The obtained dynamic increase of the excitation intensity in the zone of interest allows the reduction of the overall level of excitation and, consequently, a significant reduction of the undesired fluorescence generated by the excitation of the surrounding zones.
Fig. 9. Application of two foveal lenses on tissue expressing the fluorescent calcium indicator GCaMP6s. a) An image of the complete field of view, with the cells in focus. b) Region of interest (ROI), zoom centered on two neurons. c) The sample is moved approximately 35 ${\mathrm{\mu}}{\textrm m}$ out of the focal plan. d) Keeping the previous position, the iFoveal is activated. e) The sample is placed back in focus and isolated by activating the iFoveal. f) The sample is at 35 $\mathrm \mu$m from the focal plan and both eFoveal and iFoveal are activated. The scale bars represent 20 ${\mathrm{\mu}}{\textrm m}$.
5. Discussions and conclusions
Taking advantage of the high OP of small aperture (local) TLCLs, we have demonstrated that foveal TLCLs can be used in mini-endoscopes with large objective apertures to locally adjust the focal length with no moving parts. We also have demonstrated that we can significantly increase the signal/noise ratio by locally exciting the zone of interest and by locally imaging the same zone, which decreases the peripheral background.
It is worth mentioning that the use of a diode laser in our study was necessary for two reasons. First, the light needs to be collimated, but this is not the primary concern, since our mini-endoscope design has the advantage of transmitting the input light through fiber, which provides us with flexibility in choosing the light source. Second, a relatively powerful source of illumination was necessary to maintain the required (for adequate imaging) fluorescence level since we are using a polarizer in the imaging path. This issue can be mitigated by employing polarization-independent foveal lenses, which are fabricated by using two similar lenses with perpendicular rubbing directions.
The combination of foveal lenses in both the excitation and emission paths of the mini-endoscope procures properties that are similar to confocal microscopes, without mechanical movements. From a biological viewpoint, such capabilities in a small lightweight optical system allow ones to deepen and broaden our understanding of neural circuits in small animals during various behavioral tasks. Moreover, using solely the foveal lens in the excitation path of the mini-endoscope also offers the possibility of probing the local network connectivity by delivering a small light spot in specific areas in the FoV. We can imagine alternative designs where the movements of optical elements (lenses, diaphragms, etc.) might be used to locally change the light intensity distribution. However, such systems will be extremely difficult to integrate into head-mounted mini-endoscopes.
These results, obtained using a macroscopic model of a mini-endoscope, proved that foveal lenses could be used to locally adjust the focal length and increase the signal/noise ratio. Encouraged by the obtained results, we have optimized the opto-mechanical design of our new head-mounted mini-endoscope and we are currently working on its fabrication. The projected dimensions (validated by Zemax simulations) are 10 x 12 x 21 mm. With a 1.8 mm diameter GRIN lens (64520, Edmund Optics) as the objective, the maximum imaging depth of the system will be approximately 120 ${\mathrm{\mu}}{\textrm m}$ when using a 0.5 mm GRIN probe (relay lens) for exploring deeper regions of the brain. The axial resolution of the macroscopic system may not directly mirror the axial resolution of the ultimate mini-endoscope setup. Again, according to Zemax estimations, the axial resolution is anticipated to be approximately 25 ${\mathrm{\mu}}{\textrm m}$ with 520 nm light, primarily attributed to spherical aberrations introduced by the GRIN lenses. The corresponding results will be reported soon.
Funding
Natural Sciences and Engineering Research Council of Canada (05888, 2018-06218); Canada Research Chairs (230212); Fonds de Recherche du Québec - Santé (284011).
Acknowledgments
We thank Dr. Josée Seigneur for technical help. We thank LensVector Inc. (San Jose, CA, USA) for its material and financial support of this work. LT thanks the SMAART program. Natural Sciences and Engineering Research Council of Canada (NSERC; 05888) and Canada Research Chair in Liquid Crystals and Behavioral Biophotonics (CRC, 230212) to TG. Natural Sciences and Engineering Research Council of Canada (NSERC; 2018-06218) and Fond de Recherche du Québec en Santé scholarship (FRQS; 284011) to FB.
Disclosures
The authors declare no conflicts of interest.
Data availability
Data underlying the results presented in this paper are not publicly available at this time but may be obtained from the authors upon reasonable request.
References
1. D. Aharoni and T. M. Hoogland, “Circuit investigations with open-source miniaturized microscopes: Past, present and future,” Front. Cell. Neurosci. 13, 1 (2019). [CrossRef]
2. N. J. Beacher, K. A. Washington, C. T. Werner, et al., “Circuit investigation of social interaction and substance use disorder using miniscopes,” Front. Neural Circuits 15, 1 (2021). [CrossRef]
3. C. T. Werner, C. J. Williams, M. R. Fermelia, et al., “Circuit mechanisms of neurodegenerative diseases: A new frontier with miniature fluorescence microscopy,” Front. Neurosci. 13, 1 (2019). [CrossRef]
4. W. Piyawattanametha, E. D. Cocker, L. D. Burns, et al., “In vivo brain imaging using a portable 2.9 g two-photon microscope based on a microelectromechanical systems scanning mirror,” Opt. Lett. 34(15), 2309–2311 (2009). [CrossRef]
5. B. N. Ozbay, G. L. Futia, M. Ma, et al., “Three dimensional two-photon brain imaging in freely moving mice using a miniature fiber coupled microscope with active axial-scanning,” Sci. Rep. 8(1), 8108 (2018). [CrossRef]
6. A. Bagramyan, T. Galstian, and A. Saghatelyan, “Motion–free endoscopic system for brain imaging at variable focal depth using liquid crystal lenses,” J. Biophotonics 10(6-7), 762–774 (2017). [CrossRef]
7. A. Bagramyan, L. Tabourin, A. Rastqar, et al., “Focus-tunable microscope for imaging small neuronal processes in freely moving animals,” Photonics Res. 9(7), 1300–1309 (2021). [CrossRef]
8. P.-G. de Gennes and J. Prost, The Physics of Liquid Crystals, International Series of Monographs on Physics) (Oxford University Press, 1995) Volume 2, p. 4.
9. G. D. Love, “Wave-front correction and production of zernike modes with a liquid-crystal spatial light modulator,” Appl. Opt. 36(7), 1517–1524 (1997). [CrossRef]
10. X. Wang, B. Wang, I. McManamon, et al., “Spatial resolution limitation of liquid crystal spatial light modulator,” in Advanced Wavefront Control: Methods, Devices, and Applications II, vol. 5553J. D. Gonglewski, M. T. Gruneisen, M. K. Giles, et al., eds., pp. 46–57.
11. T. V. Galstian, Smart Mini-cameras (CRC Press, 2013).
12. W. A. Liberti, L. N. Perkins, D. P. Leman, et al., “An open source, wireless capable miniature microscope system,” J. Neural Eng. 14(4), 045001 (2017). [CrossRef]
13. T. Galstian, “Electrode structure for creating electrical potential gradient,” US Patent App. 17/783,539 (2023).
14. L. Begel, B. Khodadad, and T. Galstian, “Adaptive lens for foveal vision, imaging and projection over large clear apertures,” Opt. Express 31(2), 2877 (2023). [CrossRef]
15. K. Chen, Z. Tian, and L. Kong, “Advances of optical miniscopes for in vivo imaging of neural activity in freely moving animals,” Front. Neurosci. 16, 1 (2022). [CrossRef]
16. K. Takatoh, M. Sakamoto, R. Hasegawa, et al., Alignment Technology and Applications of Liquid Crystal Devices (CRC Press, 2005).
17. A. Bagramyan and T. Galstian, “Dynamic control of polarization mismatch and coma aberrations in rod-grin assemblies,” Opt. Express 27(10), 14144–14151 (2019). [CrossRef]
18. W. J. Smith, Modern Optical Engineering : The Design of Optical Systems (McGraw-Hill Professional, 2008), 4th ed.
