Parallel and Flexible Fluorescent Imaging using Two-photon RESOLFT Super-resolution Microscopy ARCHNES with Spatial Light Modulator Control MASACUSETS INSTITUTE by Yi Xue JUL 3 0 2015 B.E., Zhejiang University (2013) LIBRARIES Submitted to the Department of Mechanical Engineering in partial fulfillment of the requirements for the degree of Master of Science in Mechanical Engineering at the MASSACHUSETTS INSTITUTE OF TECHNOLOGY June 2015 @ Massachusetts Institute of Technology 2015. All rights reserved. Signature redacted Author ......................... partment of Mechanica~Engineering May 8, 2015 , ------- -- Certified by..................... Signature redacted Peter T. C. So Professor h)cisS jrvisor Accepted by ..... Signature redacted David E. Hardt Chairman, Department Committee on Graduate Theses MITLibraries 77 Massachusetts Avenue Cambridge, MA 02139 hftp://Iibraries.mit.edu/ask DISCLAIMER NOTICE Due to the condition of the original material, there are unavoidable flaws in this reproduction. We have made every effort possible to provide you with the best copy available. Thank you. The images contained in this document are of the best quality available. Parallel and Flexible Fluorescent Imaging using Two-photon RESOLFT Super-resolution Microscopy with Spatial Light Modulator Control by Yi Xue Submitted to the Department of Mechanical Engineering on May 8, 2015, in partial fulfillment of the requirements for the degree of Master of Science in Mechanical Engineering Abstract High resolution imaging in three dimension is important for biological research. The RESOLFT (Reversible Saturable Optical Fluorescence Transitions) fluorescent microscopy is one technique which can achieve lateral super-resolution imaging. Twophoton microscopy naturally generate high resolution in the longitudinal direction with less background compared to single photon excitation. We combine these two methods to realize three-dimensional high-resolution imaging. This super-resolution method also is limited in imaging speed. We use a spatial light modulator (SLM) as a flexible phase mask of the microscopy. It is used to compensate the system aberration, as well as increasing the imaging speed. The parallel scanning generates multiple super-resolution focuses as an array or in arbitrary positions by phase retrieval calculation. This microscopy combined with SLM control could applied to high throughput 3D imaging or multiple spots tracking in high-resolution. Thesis Supervisor: Peter T. C. So Title: Professor 3 4 Acknowledgments I appreciate the guide and help from my adviser, Prof. Peter So. He gave me advice in every stage of the research and guide me to the correct direction,, based on his profound professional experience. Also, I appreciate the help from Dr. Christopher J. Rowlands, who is a talented researcher in So-Lab. We are in the same sub-group called 'fluorescence'. He especially taught me many technique details during the experiment. I also want to thank all the other members of So-Lab and LBRC (Laser Biomedical Research Center, MIT), and I am glad to be a member of this fantastic group. Also, the cooperation lab, Prof. Ed Boyden in Media Lab, helped me about the biological knowledge. Dr. Kiryl Piatkevich and Mr. Fei Chen provided the early passage cells and plasmid DNA of GFP. I learn a lot from cooperation with them. Finally, I appreciate my parents for their support. After these two years study in MIT, I learned not only the knowledge but also academic attitude. I will continue my research in So-Lab as PhD caididate. I am looking forward to the new start. 5 6 Contents 1 1.1 1.2 2 RESOLFT fluorescent microscopy . . . . . . . . . . . . . . . . . . . . 17 1.1.1 The mechanism to realize super-resolution . . . . . . . . . . . 17 1.1.2 Reversibly photoswitchable fluorophore . . . . . . . . . . . . . 19 Multiphoton fluorescent microscopy . . . . . . . . . . . . . . . . . . . 22 25 Optical System Design 2.1 Start from the Laser . . . . . . . . . . . . . . . . . . . . . . . . . . . 25 2.2 Shutter in the system: Acoustic Optical Modulator (AOM) . . . . . . 29 2.3 2.4 3 17 Introduction 2.2.1 Bragg diffraction and AOM . . . . . . . . . . . . . . . . . . . 29 2.2.2 Measurement of the switching sequence . . . . . . . . . . . . . 31 Spatial Light Modulator (SLM) . . . . . . . . . . . . . . . . . . . . . 33 2.3.1 Liquid crystal retards phase in pixel resolution . . . . . . . . . 33 2.3.2 Correction of gamma curve by Young's double slits experiment 34 Polarization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35 2.4.1 Vectorial Debye theory in high numerical aperture (NA) system 37 2.4.2 Experiments of donut shape focus in different polarization 39 . . 43 SLM Phase Control 3.1 Adapted Optics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43 3.1.1 Optical aberration and Zernike polynomials . . . . . . . . . . 44 3.1.2 Critical metric in adapted optics for focus quality . . . . . . . 45 3.1.3 LabVIEW control and experiments for the aberration correction 46 7 3.2 4 5 Phase retrieval algorithm . . . . . .... .. .. . . .. . .. .. .. . 51 3.2.1 Gerchberg-Saxton iterative algorithm . . . . . . . . . . . . . . 51 3.2.2 LabVIEW control and experiment for arbitrary position spots 52 Two-photon RESOLFT microscopy imaging 57 4.1 PSF measurement: silver nanoparticle scanning 4.2 Parallel scanning: Gaussian spots array and donut shape spots array 59 4.3 Live cell two-photon imaging 60 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conclusion 57 63 8 List of Figures 1-1 Wavelengths, Timing, and Focusing Scheme in a Point-Scanning RESOLFT Setup Employing Dronpa-M159T. [1] 1-2 . . . . . . . . . . . . . . . . . . 18 Parallelized scanning (RESOLFT) nanoscopy using orthogonally and incoherently crossed standing waves [2]. The dark space between the two standing wave overlapping works as dark center of donut shape focus. 1-3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (a) The absorption spectrum of Dronpa (solid line), Dronpa-V157G (rsFastLime, dotted line) and Dronpa-M159T (dash line). (b) The emission spectrum of Dronpa, Dronpa-V157G and Dronpa-M159T [3]. 1-4 19 21 (a) The fatigue curve of Dronpa-M159T, under the condition of 1409=2.1 kW/cm 2 , 1491 =1.5kW/cm 2. It can undergo more than 6000 switching cycles before bleaching down to <50% of the initial signal. (b) switching-off time comparison among Dronpa, Dronpa-V157G and DronpaM159T. a. b. enlarged figure during one unit cycle of Dronpa-V157G and Dronpa-M159T respectively. [1,31 . . . . . . . . . . . . . . . . . . 1-5 21 Schematic diagram of the photoswitching of Dronpa. B, deprotonated form (ON-state); D, an unknown dark state; A 1 , the protonated form (OFF-state) that originally exists in the sample; A 2 , the protonated form that is formed by the photoswitching; I, a nonfluorescent intermediate. The photoswitching pathway from B to A 2 is shown by blue arrows, and that from A 2 to B is shown by red arrows [4J. 9 . . . . . . 22 *2 <gh-photon and two-photon resolution comparison in wide-field mi,rcrscopy (conventional) and confocal microscopy (a) Intensity of the nage of a thin fluorescent edge (lateral resolution); (b) Intensity of the &xial response to a thin fluorescent sheet; (c) Intensity of the image of a thick fluorescent edge; (d) Intensity of the axial response to a thick ; uorescent 2-i layer. [5] . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24 Jhe experiment setup for parallel two-photon RESOLFT microscopy. Mode-locked Ti:Sapphire laser (Spectra-physics, Tsunami) provides the source for 'turn-off' light and excitation light. AOM (acoustic optical modulator, ISOMET) synchronized with LED (Thorlabs, M405L2, center wavelength 405rnm) to control 'turn-on' and 'turn-off' time se- quence. SLM (Spatial light modulator, HOLOEYE PLUTO) provides the phase control of the system. Other main components: CMOS (Point Grey, Flea3 USB), OL (objective lens, Zeiss A-Plan, x100, NA 1.25), DM: dichroic mirror, piezo stage (Queensgate, NPS-XY-100A). 2-2 26 The mode-locked laser pulse reflects back and forth between the mirrors of the resonator. Each time it reaches the output mirror it transmits a short optical pulse. The transmitted pulses are separated by the distance 2d and travel with velocity c. The switch opens only when the pulse reaches it and only for the duration of the pulse. The periodic pulse train is therefore unaffected by the presence of the switch. Other wave patterns, however, suffer losses and are not permitted to oscillate. [61 28 2-3 The common optical path of Ti:Sapphire laser. (Spectra-Physics manual) 29 2-4 Bragg diffraction: an acoustic plane wave acts as a partial reflector of light (a beamsplitter) when the angle of incidence 0 satisfies the Bragg condition. [6] 2-5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . The LabVIEW control of three channel digital output: infrared channel, UV channel and camera. . . . . . . . . . . . . . . . . . . .. .. . 31 10 30 ......36 2-6 Synchronized time sequence of infrared beam and UV light. Channel 1 is the infrared beam while channel 2 is the UV light. Figure (b) is the enlarged period of figure (a). The period time is 12ms. Figure (b) shows more obvious switch-off delay (<500us). . . . . . . . . . . . . . 2-7 32 Molecular orientation of a liquid-crystal cell (a) in the absence of a steady electric field and (b) when a steady electric field is applied. The optic axis lies along the direction of the molecules. [6] . . . . . . . . . 2-8 33 The Gamma curve of the SLM under the experiment condition. LUT: look-up table. This curve represents the relation between gray level and the voltage on the liquid crystal. 2-9 . . . . . . . . . . . . . . . . . . 34 Experiment measured interference fringes shift in the Young's double slits experiment, after the gamma-curve correction. The fringes shift because of the 27r change of phase in one slit. The figure shows each one fringe linearly shifts one period distance during the 2z- phase change. 35 2-10 The key polarization control optical component in the system. HWP: half wave plate (thorlab). GT: Glan-Thompson prism. QWP: quarter wave plate. . . . . . . . . . . . . . . . . . . . . . . . . . . 2-11 The phase change of s-polarized beam and p-polarized beam in reflection.The reflective media reflective index n>1. angle. 0 B is the Brewster's . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36 2-12 Diffraction of a converging spherical wave at a circular aperture: No- tation. [7] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37 2-13 Normalized intensity of polarized light with a vortex 0-2wr phase plate at the focal spot in the horizontal (XY) and longitudinal (XZ) planes. Various polarizations are considered: linear, circular (right-handed and left-handed), right-handed ellipse (with X and Y as the long axis). Intensities of 0 and 1 correspond to black and white, respectively. The axis units are in wavelengths. [8] . . . . . . . . . . . . . . . . . . . . 39 2-14 The vortex 0 - 27r phase and corresponding donut shape focus. The center grayscale for the vortex is 58 and it is a left-hand vortex. 11 . . . 40 2-15 (a) The focus of right-hand polarized beam modified by left-hand vortex 0 - 27r phase plate. (b) The focus of left-hand polarized beam modified by left-hand vortex 0 - 27r phase plate. under the same color bar scale. Two pictures are . . . . . . . . . . . . . . . . . . . . . 41 2-16 Donut shape focus (a) before polarization modify and (b) after polarization modify. The two images are of the same colorbar scale. The left figure is the same as figure 2-14 (b). . . . . . . . . . . . . . . . . 3-1 LabVIEW control panel for system aberration compensation. ...... 3-2 The focus spots under different degree of astigmatism (Zernike mode 47 (-2, 2), Table 3.1) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-3 The sharpness, brightness, and the combined metric v.s. 41 48 bias (b) curves. The combined metric curve will change if change the 'weight' between sharpness and brightness, depending on which one is major. 3-4 The focus spots under fixed astigmatism (b = -0.9) compensation while testing different degree of coma (Zernike mode (3, 1), Table 3.1). 3-5 48 The sharpness, brightness, and the combined metric v.s. 49 bias (b) curves. In the second round testing, the sharpness and brightness have similar trend, which means the aberration of the system converge to a 'local minimize' point under these two phases compensating. 3-6 . . . . . 49 (a) The original Gaussian beam focus without aberration compensation to the system. (b) The focus after aberration compensation. (c) The compensated phase adding to the SLM, which is sum of negative astigmatism and positive vertical coma. . . . . . . . . . . . . . . . . . 3-7 The donut shape focus under Zernike mode (-1, 1) and bias b = -1 50 : 1 modify, selectively shown every other figures. When b = 0.5, the donut shape is the most uniform one . . . . . . . . . . . . . . . . . . . . . . 3-8 The donut shape focus under Zernike mode (3, 3) and bias b = -1 : 1 modify, selectively shown every other figures. . . . . . . . . . . . . . . 12 50 51 3-9 (a) The original donut shape focus without aberration compensation to the system. (b) The focus after aberration compensation. (c) The compensated phase adding to the SLM, which is sum of positive shift and a negative trefoil. (a) and (b) are under the same color scale. 3-10 The principle of the iterative Fourier transform algorithm. 51 . . 191 ..... 52 3-11 The control panel of flexible position focuses. The aimed focuses are selected from the 'new picture' Window . . . . . . . . . . . . . . . . . 53 3-12 . (a) The focuses in random positions, selected from the control panel. (To easy observation, enlarged the focuses area 10 times, each spot here should be one pixel size in the calculation.) (b) The phase mask calculated by G-S algorithm. (c) The camera image of the phase (Rhodamine 123 solution as sample), which is the same as the aimed focuses. 54 3-13 The flexible position of donut spots, for multiple super-resolution particle tracking. The donut shape in the 'input image' is enlarged. Like the Gaussian focuses, one pixel spot to indicate the coordinate value of the position is sufficient. 4-1 . . . . . . . . . . . . . . . . . . . . . . . 55 nanoparticle-scanned PSF of the setup. (a) Gaussian focus. (b) Donut shape focus, with the same intensity scale. (c) Overlay of (a) and (c) to show the colocalization (pseudo color). 4-2 . . . . . . . . . . . . . . . 58 The PSF comparison between single photon excitation (blue dash line) and two-photon excitation (red solid line). (a) The Gaussian focus. (b) The donut shape focus. 4-3 . . . . . . . . . . . . . . . . . . . . . . . . . 59 5x5 spots array, two-photon excitation using Rhodamine 123 sample, detected by the camera. The period of grating on SLM gradually zoom in, and the corresponding image 'zoom out'. 4-4 . . . . . . . . . . . . . . The 5x5 donut spots matrix for parallel super-resolution imaging. 13 . . 60 60 4-5 HEK 293 cell labelled by Dronpa-M159T on tubulin. (a) The field of view (FOV) of the image is 60um x 60um. The scanning step is 500nm. (b) The finc scanning image of the top right region of (a), the FOV is 30um x 30um. The scauning step is 200nm. . . . . . . . . . . 4-6 61 The position change of live cell. (a) 1st time scanning, FOV is 80um x 80um, and step size is lum. (b) 2nd time scanning for the right bottom region of (a), noticing the bottom cell orientation changed during this process. The FOV is 30um x 30um with 200nm step size. ...... 4-7 62 HeLa cell labeled by Dronpa-Tubulin. The enlarged region is increased contrast by ImageJ. The cross profile of the red line region is shown below. The finest structure we can distinguish is about 400nm. The image is 50um x 50um, with 200nm scanning step. 14 . . . . . . . . . . 62 List of Tables 1.1 Comparison of common used photoswitchable fluorophore . . . . . . . 19 3.1 Low-order Zernike mode for optical aberration . . . . . . . . . . . . . 44 15 16 Chapter 1 Introduction High resolution imaging in three-dimension of large area for live specimen is important. For example, we are not only interested in imaging the whole tumor but also want to observe the metastasis of each tumor cell; we are not only want to take image of single synapse event but also want to observe the synapse circuit in a large scale [10]. Sometimes the specimen is sparse labeled, high-resolution image only in the labeled area will be more efficient [11]. Also the requirement for multiple particle tracking simultaneously of high-resolution is important to know the relationship between different particles [121. Thus 3D high-resolution high-speed fluorescent microscopy with flexibly controlled imaging area is a good choice to facilitate the above biological research. 1.1 1.1.1 RESOLFT fluorescent microscopy The mechanism to realize super-resolution There are many methods to achieve super-resolution imaging. One type of methods is based on single molecule imaging, such as STORM (STochastic Optical Reconstruction Microscopy) 113] and PALM (PhotoActivated Localization Microscopy) [141; the other type of methods uses spatial patterned imaging, such as STED (Stimulated emission depletion microscopy) [15], RESOLFT (Reversible Saturable Opti- 17 cal Fluorescence Transitions) [l and SSIM (Saturation Structure Illumination Microscopy) [161. RESOLFT microscopy is one super-resolution technique invented by Hell's group [1]. The technique use reversibly photoswitchable fluorophores to exploit long-live dark-state and fluorescent state. Because the comparably long 'on states' and 'off status' than STED microscopy, the light intensity to achieve sub-diffraction focal spot is several orders lower. In the imaging process, each scanning point is 'turned on' first, and the periphery of the spot is 'turned off' by a donut shape spot, then the same position is illuminated by the excitation b:eam and only the residue part in the center is excited and emits fluorescence (see figure 1-1). This 'on-off-excitation' cycle is played in each scanning point until the whole field of view is imaged and each cycle takes about millisecond [1]. 1. kWlcm 2 1 O FF fluorescence 405nm 3.8 kWIcm i kwM Figure 1-1: 2 250nm Wavelengths, Timing, and Focusing Scheme in a Point-Scanning RESOLFT Setup Employing Dronpa-M159T. [1] To make the RESOLFT imaging more efficient, parallel RESOLFT imaging method is published in 2013 using two incoherently superimposed orthogonal standing light waves [2] (see figure 1-2). This parallel scanning method could take a 120um by 100um super-resolution image within one second. However, this method did not consider about depth resolution, while it used sCOMS camera. So the depth resolution is the same as single photon wide-field imaging. 18 Only First Order Selected Back *7)Focal Plane Polarization direction7 GRID PBS LENS LENS Focal Plane Figure 1-2: Parallelized scanning (RESOLFT) nanoscopy using orthogonally and incoherently crossed standing waves [2]. The dark space between the two standing wave overlapping works as dark center of donut shape focus. 1.1.2 Reversibly photoswitchable fluorophore For RESOLFT fluorescent microscopy, reversibly photoswitchable fluorophore is one of the key part of the technique. Single point scanning RESOLFT much slower than STED is because the photoswitching time is much longer than fluorescent lifetime. Reversible photoswitching of individual molecules has been demonstrated for a number of mutants of the green fluorescent protein. The limited number of switching and switching speed are always concerned. At the same time, the quantum yield and absorption coefficient are important optical factors of the fluorophore. Recently, red color reversible photoswitching fluorophore is also designed [17,18], which extended the available spectrum range and made multi-color RESOLFT possible. However, the red reversible photoswitching fluorophore is not as mature as GFP-like switchable fluorophore. The switching speed, absorption coefficient and quantum yield need further research. For our experiment, because of the laser wavelength we choose (in Chap. 2), we focus on green color reversible fluorophores. Table 1.1 shows the compari- son of characteristics of several commonly used reversible fluorophores in RESOLFT experiments. Table 1.1: Comparison of common used photoswitchable fluorophore 19 Dronpa Dronpa-M159T rsEGFP rsEGFP2 Absorption maximum (nm) 503 489 493 478 Emission maximum (nm) 522 515 510 503 95000 61732 47000 61300 0.85 0.23 0.36 0.3 Switch-off half-time t,/2off (s) $ 263 0.23 ~20 5 Off-state relaxation half-time t1/2rela (min) 840 0.5 180 20 7 6000 1100 2100 Fluorophore Molar absorption coefficient (At - cm-') Quantum yield Fatigue times (<bFL) t$ f Fatigue time means the number of switching cycles to bleach to 50% of the initial fluorescence in the on-state. The switch-off half-time and fatigue switching times depend on experiment conditions, such as the intensity of illumination light. The number is scale to similar intensity scale according to the references [1-3, 19,20]. According to Table 1.1, Dronpa-M159T and rsEGFP2 have similar performance, but Dronpa-M159T can be switched three times more than rsEGFP2. Also, DronpaM159T has higher contrast (intensity of dark state/ intensity of bright state) [21, we choose it as fluorophore in our experiment. Dronpa-M159T is one mutant of Dronpa, in which Met 5" side chain is replaced by a smaller threonine residue. Because Met" interferes with the movement of a six-ring structure upon a light-induced cis-trans isomerization 13], changing it with smaller structure reduces the switch-off half-time. Figure 1-3 and figure 1-4 show more details of Dronpa M159T. The mechanism of photoswitching for Dronpa (mutant Dronpa is similar) is shown in figure 1-5. Photoswitching means Dronpa has two states: 'ON-state' and 'OFFstate'. When the Dronpa is in 'ON-state', it can be excited and emit fluorescence. Otherwise, it cannot be excited by the same light. From the chemical view, these two states actually corresponding to two structure. In 'ON-state', Dronpa is deprotonated; in 'OFF-state', it is protonated. Naturally, pH change also can lead to protonation (acid) or deprotonation, so the photoswitching of Dronpa can be controlled by pH also. The 'turn-on' process is as following: the Dronpa is originally protonated in A 2 , illuminated by UV light at 390nm. Electron will be excited, and it has two 20 31,00.8r 0,6- 0* 0.4. I. 0.2. 0.0- r 250 wavelength [nm] 540 5;0 5;0 600 6 wavelength [n] (a) (b) 350 300 450 400 460 600 50 S0 480 50 520 0640 Figure 1-3: (a) The absorption spectrum of Dronpa (solid line), Dronpa-V157G (rsFastLime, dotted line) and Dronpa-M159T (dash line). (b) The emission spectrum of Dronpa, Dronpa-V157G and Dronpa-M159T [3]. a IF U V c 0.75. W, U ** b U mu U. IF mu 60.50- c4M0 'a 0 0 2000 0.25. [Dints 0 2000 4000 C s600 woUU Wo Cydles t [s] 15W (b) (a) Figure 1-4: (a) The fatigue curve of Dronpa-M159T, under the condition of I405=2.1 kW/cm 2 , 14 91=1.5kW/cm 2 . It can undergo more than 6000 switching cycles before bleaching down to <50% of the initial signal. (b) switching-off time comparison among Dronpa, Dronpa-V157G and Dronpa-M159T. a. b. enlarged figure during one unit cycle of Dronpa-V157G and Dronpa-M159T respectively. [1,3] choices: one is directly from excited state of protonated form to the excited state of deprotonated form, called excited state proton transfer (ESPT); the other is falling to the ground state of deprotonated form (red line in figure 1-5). Because the emission of fluorescence as excited by 390nm is extremely low (<0.01), it means most electrons go through the second way. The quantum yield of switch-on is the 'turn-off' and excitation process are competitive. 21 - 0.37. In 'ON-state', After Dronpa is excited by 488nm light, 65% molecules perform fast on-off blinking (~ show medium-long off-time (order of tens millisecond). OqA ims), and 35% molecules The fast on-off blinking is .ttributcd to transitions into triplet states with a triplet lifetime of 1.2ms. The medium-long transition may be attributed to an unknown dark state D. The switch-off time is not related to excitation intensity [4]. In RESOLFT experiment, normally use longer time illumination to realize switching-off, but using shorter illumination time to realize excitation [2]. But it is not necessary to be higher intensity to switching-off then fluorescent excitation [1,21. ESPT 0.371 65% '*OSW35 3.2 x10-4\ W 3.6 ns Aslow (photoswitched protonated form) % ^kB 0.85 (non-fluorescent intermediate) B Reaction Coordinate Figure 1-5: Schematic diagram of the photoswitching of Dronpa. B, deprotonated form (ON-state); D, an unknown dark state; A 1 , the protonated form (OFF-state) that originally exists in the sample; A 2 , the protonated form that is formed by the photoswitching; I, a nonfluorescent intermediate. The photoswitching pathway from B to A 2 is shown by blue arrows, and that from A 2 to B is shown by red arrows [4]. 1.2 Multiphoton fluorescent microscopy To improve longitudinal resolution, of course 3D super-resolution is one choice. However, the setup for 3D super-resolution imaging, especially RESOLFT method, is very 22 complicate. On the other hand, multiphoton imaging can provide high-resolution in Z-axis. Multiphoton microscopy is widely used in biological tissue imaging, including neuroscience, embryology and oncology [21]. It enables noninvasive study of bio- logical specimens in three dimensions with diffraction limited resolution, especially much higher longitudinal resolution compared to single photon wide-field microscopy. Multiphoton microscopy also reduces specimen photodamage as well as enhances penetration depth for thick tissue. Recently, multiphoton imaging is used to do in vivo imaging of subcortical structures within an intact mouse brain [221, as well as blood flow and neurovascular coupling in the brain [23]. Also, parallel multiphoton microscopy is realized to increase the imaging speed [24], which used grating to generate a 8 x 8 spots array and collected data by a 8 x 8 PMT array. Recently, spots array containing larger number of spots is used in multiphoton microscopy, which can achieve 1kHz throughout rate [25]. Multiphoton microscopy could also combine with other techniques, which melt the advantages of both methods. For example, two-photon excitation STED microscopy improved the spatial resolution of standard two-photon microscopy by a factor of four to six times, which is used to image morphology of dendritic spines and microglial cells of brain slice [26]. One advantage of two-photon microscopy is that it has naturally 3-D high resolution, especially in the Z-axis direction, compared to single photon microscopy. Figure 1-6 shows the comparison of point spread function (PSF) of single photon excitation and PSF of two-photon excitation in wide-field microscopy and confocal microscopy, respectively. The other advantage of two-photon microscopy is that the penetration depth is larger than that of single photon excitation. Because two-photon microscopy uses longer wavelength, according to the Beer-Lambert law, the light is less attenuated by the specimen. Thus, two-photon microscopy has advantage of imaging thick specimen or biopsy. Two-photon iPSF (intensity Point Spread Function) is to the square of singlephoton iPSF. For a 4-f system, the intensity distribution of object can be expressed as 23 2 ( Io(ri) = Ihi(MAYir1) *"oitm- otm cordca oto - ca 0.80- oovn4onlWQphOtS \. 0.2 -1 0.6 012 0 4 0 1 5 I0 to1 15 V U (a) (b) 20 2 0 ~hc OA 0.2 0 .1 0 -6 0 10 S (d) (C) Figure 1-6: single-photon and two-photon resolution comparison in wide-field microscopy (conventional) and confocal microscopy (a) Intensity of the image of a thin fluorescent edge (lateral resolution); (b) Intensity of the axial response to a thin fluorescent sheet; (c) Intensity of the image of a thick fluorescent edge; (d) Intensity of the axial response to a thick fluorescent layer. [51 where M1 is a matrix including demagnification factors for the objective, and hi is the 3-D amplitude PSF for the objective lens. Then after the collector lens with transfer function h 2 and M2 , the image intensity within the detector area D is hi(r) = 1hi(Mr) 2 " [h2 (r)|2 D ( ")] (1.2) n=1, 2, and 3 correspond to single-photon, two-photon, or three photon excitation, respectively. From the iPSF equation 1.2, we know that the two-photon iPSF is square of the single-photon iPSF [27]. 24 Chapter 2 Optical System Design The microscopy is a combination of two-photon microscopy and RESOLFT superresolution microscopy. The experiment setup is shown in figure 2-1. The specimen is illuminated by the UV 'turn-on' light first. Then, the periphery area of the spot is 'turned-off' by the donut shape two-photon light. Finally, the residue center area which is still in 'on-state' is excited by the two-photon Gaussian spot and emit fluorescence. To realize this process, we need to design the optical section of the system. In this chapter, we will talk about the function of each key optical component, including the Ti:Sapphire laser and LED, the acoustic optical modulator (AOM), the SLM, and polarization control components. We will show how these components help us realize the aim of the setup. 2.1 Start from the Laser The theory of two-photon absorption and emission process is developed by Maria Goppert-Mayer around 1930 [27]. In her work, the transition probability of a twophoton electronic process is quadratically related to the excitation light intensity. The process involves the interaction of two photons and an atom via an intermediate 'virtual' state. The first photon induces the transition from the ground state to the virtual state, and the second photon induces the transition from the virtual state to the final excited state. Because the possibility of two-photon absorption process is 25 AOM Filter & collimate DMV Tub lens Filter & collimate OL Piezo stage Figure 2-1: The experiment setup for parallel two-photon RESOLFT microscopy. Mode-locked Ti:Sapphire laser (Spectra-physics, Tsunami) provides the source for 'turn-off' light and excitation light. AOM (acoustic optical modulator, ISOMET) synchronized with LED (Thorlabs, M405L2, center wavelength 405nm) to control 'turn-on' and 'turn-off' time sequence. SLM (Spatial light modulator, HOLOEYE PLUTO) provides the phase control of the system. Other main components: CMOS (Point Grey, Flea3 USB), OL (objective lens, Zeiss A-Plan, x100, NA 1.25), DM: dichroic mirror, piezo stage (Queensgate, NPS-XY-100A). 26 very low, it requires high intensity laser to achieve a measurable effect. In two-photon absorption process, the number na of photons absorbed per fluorophore per pulse can be expressed as n = po2a"r p 2 A2 2 (2.1) . 2hcA where po is the laser power, 6 is the two-photon absorption cross-section (~1058m4 . s), Tr is the pulse duration (~100fs), fp is the repetition rate (~80MHz), A is the numerical aperture (~1.4) [281. Thus, the photon absorbed in two-photon process requires high peak power excitation. For optical domain, the lack of high peak power laser blocks the way. After the femtosecond laser is developed in 1980s, Denk, Strickler and Webb demonstrated two-photon laser scanning fluorescence microscopy in 1990 [28]. They used a collidingpulse, mode-locked dye laser which produced a stream of pulses with a pulse duration of about 100fs at a repetition rate of about 80MHz. In 1990s, Ti-sapphire laser became an optimal for two-photon excitation microscopy until now. The Ti-sapphire lasers are based on Kerr lens mode locking technique to generate femtosecond pulse. Kerr lens effect is caused by inhomogeneous refractive index change of Ti-Sapphire crystal under strongly focused beam. Combined with an aperture on the focus of Kerr lens, only high intensity pulses experience low loss [27]. This is one method to achieve mode-lock. A laser can oscillate on many longitudinal modes, with frequencies that are equally separated by the intermodal spacing VF = c/2d where d is the length of resonator. The modes can be coupled and their phase can be locked together. If M modes in the locked phase, the pulse length dpise = 2d/M, and the peak intensity is I = M2| A1 2 = M 2 1, I is the mean intensity [6]. Thus, the larger M, the higher peak intensity, and the shorter pulse duration in time. On the other hand, the broader envelop in spectrum domain. In a typical Ti-Sapphire laser resonator, when the modes are not locked, the laser is equal to a continuous wave (CW) laser with low intensity (because no Kerr lens effect, the loss is high). Normally the optical system has positive 'chirp' (normal dispersion), which means the speed of longer wavelength wave is slower. The relation 27 Optical switch 2d Reoator Figure 2-2: The mode-locked laser pulse reflects back and forth between the mirrors of the resonator. Each time it reaches the output mirror it transmits a short optical pulse. The transmitted pulses are separated by the distance 2d and travel with velocity c. The switch opens only when the pulse reaches it and only for the duration of the pulse. The periodic pulse train is therefore unaffected by the presence of the switch. Other wave patterns, however, suffer losses and are not permitted to oscillate. [6] between light speed (group velocity) and wavelength is -nvP= + (-) dAI . (2.2) In normal dispersion region, dn/dA < 0 129]. However, prisms are used to achieve negative chirp inside the Ti-Sapphire laser resonator, in order to broaden the spectrum of pulse. For prism, shorter wavelength light has faster speed but also travels longer distance inside prism, so the chirp can be compensated by a set of prisms. According to the above mode-lock analysis, when the spectrum of pulse is extended, the pulse length is shorter in time domain, meaning more mode is phase locked. Figure 2-3 shows the typical optical path inside the resonator of Ti:Sapphire laser. The laser in our lab is manually mode-locking. That is, we adjust the horizontal mirror and vertical mirror to achieve the maximum intensity, and the slit is to select wavelength, as well as a set of prisms to compensate 'chirp'. Once we lose the mode-lock (which happens often), we need to adjust the laser again. During adjusting, the mirror will be rotated, so the output beam will have a slightly shift, causing the realignment of all the following optical path. The laser also needs to be purged by filtered and dried nitrogen gas, which can eliminate the typical problems associated. with dust and 28 Pmdodl e HRPr M Tug Beam 'Splitter otput Brewster P2 Owl(n M3 1er Blewster M4 .PZT (Opfional) Pumpv-1 2 Tirwptie MCde9 305 Rod Duver Eloctronics -AOM ResiduvJ HR searoumnP Opfional Model 3930 Lok-to-Clock Elecitica Figure 2-3: The common optical path of Ti:Sapphire laser. (Spectra-Physics manual) contamination, and tuning discontinuities caused by oxygen and water vapor. The absorption of oxygen and water vapor is right between 900nm to 1000nm most serious, which is right in our working wavelength range (see fluorophore selection table 1.1). 2.2 Shutter in the system: Acoustic Optical Modulator (AOM) This section of chapter talks about the time-sequence control of different light path and camera. We use an AOM as shutter of the infrared beam, and the LED is triggered by a TTL signal. The camera is triggered by a falling edge. The change between Gaussian focus and donut shape focus is controlled by SLM. The specimen scanning is operated by the piezo stage. These signals are synchronized by NI 6009 card and LabVIEW control. The SLM and piezo stage have commercial VI, so in this section we focus on the home-made control of the AOM, the LED and the camera. 2.2.1 Bragg diffraction and AOM For a RESOLFT microscopy, we need to turn-on and turn-off fluorescence in a certain time sequence. Thus, it is necessary to synchronize different beams. The system is controlled by LabVIEW program. In hardware, Acoustic Optical Modulator (AOM) 29 is one the synchronize part, which is the shutter for turni-off and excitation light from the Ti-Sapphire laser. The 1.OM is based on the theory of acousto-optic effect. An acoustic wave creates a perturbation of the refractive index in the form of a wave, thus the medium becomes a dynamic graded-index medium [6]. The key component of our AOM is a Bragg cell, in which the medium reflective index is compressed periodically with a harmonic sound wave. This structure is equal to a grating with period of sound wavelength A. If the incident angle satisfies the Bragg condition, the +1 or -1 order of diffraction light intensity will be maximized. Without the acoustic wave, the light will propagate along the zero order direction. Diffracted light Incident fight A ~~ Sound Transmitted light Figure 2-4: Bragg diffraction: an acoustic plane wave acts as a partial reflector of light (a beamsplitter) when the angle of incidence 0 satisfies the Bragg condition. [6] In figure 2-7, the Bragg angle is sin 6 = 2A The AOM used in our experiment (ISOMET, VA) has deflection efficiency about 60% in the infrared wavelength. Under designed wavelength, the deflection efficiency can reach 90%. AOM efficiency is related to the wavelength of incident light and RF power. The modulation frequency of AOM can reach MHz. Another choice for fast shutter is electro-optics modulator (EOM), such as Pockel cells. For some material, the reflective index changes under electric field, so the phase of incident light will change correspondingly. Thus, the polarization of incident light change. The operation frequency of Pockel cell is about kHz, and the transmission 30 for the designed wavelength can over 99%. If the system concern more intensity than operation frequency, mechanical shutters are more suitable choice. The fast mechanical shutters from Uniblitz or Newport can reach 100Hz frequency, the Uniblitz shutter can reach 400Hz burst frequency of operation. 2.2.2 Measurement of the switching sequence In the experiment, we use NI USB6009 card to synchronize switching signals. Both digital and analog output work, but analog output always has about ims delay between different channel (unknown reason). Digital output doesn't have the artificial delay and also has more channels, so we choose to use digital output. Figure 2-5 shows the LabVIEW control of the synchronization. The first channel is for the infrared beam, and the second channel is for the UV light. The third channel is for the camera trigger. The illumination length can change individually, while the period time of all the three channels also can adjust together. Period [RriAsinle connection wire to the poft~ arthe kotae oUtpuforTL WfV0*60n WV. &----A Ingo- ntrror ou Iftm, Figure 2-5: The LabVIEW control of three channel digital output: infrared channel, UV channel and camera. The maximum trigger frequency of LED is 1kHz, while AOM switching frequency is much higher. The limit of system frequency is from the camera. 31 The camera maximum can achieve 120Hz, and it is controlled by rolling shutter. The minimize line exposure time is 8us. So for a 1280 x 960 image, the minimum exposure time is 8ms (corresponding to 120Hz). When we decrease the period of three channel to about 10ms, we can observe the blinking of focus because the rolling shutter effect. For the outside triggered mode of camera, the maximum frequency is 60Hz. Also, the image acquisition time relates to the fluorescent signal intensity, which decide the minimum exposure time of the camera. Figure 2-6 shows the switching sequence of infrared beam and UV LED, which is collected by photodioded and oscilloscope. This is for the basic two-photon microscopy imaging for photoswitchable fluorescent labelled specimen. Channel 1 is the AOM controlled infrared beam from Ti:Sapphire laser, and channel 2 is the LED light. From the figure 2-6, both channels have switch-off delay, and the UV channel has longer delay time. However, both of them are within 500 us, so it won't have too serious influence. Possibly caused by the frequent turn-on and turn-off, the UV LED light intensity is not stable within the 'on' period, but the slightly intensity drop won't influence switching on the fluorophore. (a) (b) Figure 2-6: Synchronized time sequence of infrared beam and UV light. Channel 1 is the infrared beam while channel 2 is the UV light. Figure (b) is the enlarged period of figure (a). The period time is 12ms. Figure (b) shows more obvious switch-off delay (<500us). 32 2.3 2.3.1 Spatial Light Modulator (SLM) Liquid crystal retards phase in pixel resolution We use a reflection style phase only spatial light modulator (HOLOEYE Photonics AG, Germany) in the setup. The key part of the SLM is a liquid crystal, which molecule orientation changes under a steady electric field, resulting in the modulation of light phase. x 2 E (a) (b) Figure 2-7: Molecular orientation of a liquid-crystal cell (a) in the absence of a steady electric field and (b) when a steady electric field is applied. The optic axis lies along the direction of the molecules. [6] Without external electric field, for a linear polarized light in x or y directions (parallel and perpendicular to the molecular direction), the wave retardation is P = 27r (ne - n0 ) d/A. When apply the electric field along z direction, the resultant electric force tends to tilt the molecule along z direction, but the elastic forces at the surfaces of the glass plates resist this motion. But when the voltage is larger than a certain threshold, most molecules are tilted except those nearing the glass surface. relation between tilt angle and voltage is expressed as aVep - -V 0 2ta 1 V2> V x The 161: (2.3) _V) 0 V < V where V is the critical voltage, and V is a constant. The reflective index is a function of the tilt angle 0 1 n2(0) _ cos 2 0 n. 33 sin 2 (2.4) Thus, the retardation becomes I = 27r (n(O) - n,) d/A. It shows the phase retardation and the electric field voltage is not linear relation. SLM uses gamma-curve to relate grayscale with the actual voltage on each pixel. 2.3.2 Correction of gamma curve by Young's double slits experiment The electric voltage added on the SLM is controlled by gray-scale of image showed on SLM screen. The relation curve between gray-scale and phase retardation called 'gamma-curve' of the SLM. Gamma-curve relates to wavelength, which can be measured and calculated by Young's double slits experiment. In the experiment, SLM screen is divided into two regions: the gray-scale of one region is zero, as reference beam; the gray-scale of the other region is changing from zero to 255, as adding different phase shift to one beam branch of Young's double slits experiment. As the phase difference between the two branches changing, the interference fringes also shift. After measure the fringes shift, the new gamma-curve corresponding to 27r phase change for the certain wavelength can be calculated by the software PhaseCam. Reload the new gamma-curve to the SLM and do the Young's double slits experiment again, adjusting the gamma curve until achieving linear 27r phase. change. 16--- - -- '------------ 12D -<0G A00 A O .. 40& 5M'0'0-A 8096 12 14 Figure 2-8: The Gamma curve of the SLM under the experiment condition. LUT: look-up table. This curve represents the relation between gray level and the voltage on the liquid crystal. At the same time, SLM is also polarization sensitive, which can only modify light 34 polarized in the same direction of liquid crystal molecule direction. In our setup, the SLM axis is in the horizontal direction. So we keep the incident light is horizontally linear polarized. This is the reason we discard the design of perpendicular incident with polarized beam splitter (PBS) and quarter wavelength plate, because circular polarized beam cannot be sufficiently modulated, but introduce distortion. If use non-polarized beam splitter, 75% of energy will lose during this process, so we choose incident with an angle without beam splitter. Figure 2-9: Experiment measured interference fringes shift in the Young's double slits experiment, after the gamma-curve correction. The fringes shift because of the 27r change of phase in one slit. The figure shows each one fringe linearly shifts one period distance during the 27r phase change. Because the SLM has a layer of glass on the top surface of liquid crystal, this layer of glass may introduce interference pattern if the incident angle is too large. That is why we discard the design of 45 degree incident (incident light and reflective light are perpendicular to each other). After the trade of optical path length and incident angle, we choose about 10 degree illumination. At the same time, we add two sets of relay lenses to minimize the phase curvature because of Fresnel diffraction. 2.4 Polarization The polarization control components are shown in figure ??. The part (a) is to make sure the incident light to the SLM is the same direction as the axis of liquid crystal 35 molecule in the SLM, which is along the horizontal direction. So, after the part (a) modulate, the light is purely p-polarized (GT extinction ratio> 100,000:1). According to figure ??), we know that the phase of p-polarized light after reflection may not change or change ir degree. Neither case lead to polarization type or direction change. So the mirrors in the system won't influence the polarization. HWP SLM HWP GT QWP (b) (a) Figure 2-10: The key polarization control optical component in the system. HWP: half wave plate (thorlab). GT: Glan-Thompson prism. QWP: quarter wave plate. 9, 9-O, 0 0 9- Oi Figure 2-11: The phase change of s-polarized beam and p-polarized beam in reflec0 tion.The reflective media reflective index n>1. B is the Brewster's angle. However, the SLM provides different phase retardant in different pixel. Thus, the polarization changed after the SLM. For example, if the SLM works as a vortex 0 -27r phase plate, the polarization of the reflective light becomes similar to 'azimuthally polarized beam' 1301 (not exactly the same). The combination of HWP and GT in section (b) is to get a pure linear polarized beam with maximum intensity. Adding GT in the system is a trade-off: GT will provide a linear polarized beam rather than an 'azimuthally polarized beam', which makes a high-purity circular polarized beam possible after QWP; but part of energy lost. In the former work, some groups use GT to purify the polarization [311 [321, but some are not [331. Because the circular 36 polarization is crucial for a high-NA system to get uniform donut shape spot (will discuss more in Chap2-4-1), we choose to use GT to ensure the purity of polarization. By changing the angle difference between HWP-GT and QWP, we can achieve a circular polarized beam. However, because we did not put the QWP and HWP right before the objective lens, there may be some aberration causing polarization distortion by dichroic mirror. This will be compensated by the spatial light modulator (will discuss more in Chap. 3). 2.4.1 Vectorial Debye theory in high numerical aperture (NA) system Super-resolution fluorescent microscopy usually is a high NA (Numerical Aperture) system, in which the polarization of light influences the electric field of the focus. The electric field vector near the focal spot can be calculated from the vectorial Debye theory [7,8,341. apartum~ Figure 2-12: Diffraction of a converging spherical wave at a circular aperture: Nota- tion. [71 Based on the figure 2-12, the diffracted field near the focus is normally evaluated by Debye integral: U(P) = Af exp(-ikq - R)dQ (2.5) where P is the point near focus 0, Q is the solid angle which the aperture subtends at the focus. q is the unit vector in the direction of OQ, and R is the position vector 37 -f OP. To the paraxial approximation, s - f = -q -,R [71, The vectorial Debye theory considers the electric field at point P as integral of electric field in the surface Lf a reference sphere EO(P), for linear polarized beam, the revised Debye integral is (parameter changed to consist two references): U(P) U(P) exp(-ikq - R)dQ = (2.6) Write in spherical coordinate, U(r, 4, z) = Ui(0, p) exp[-ikr sin 0 cos(p - 4)- ikz cos 0] sin 0d~d o (2.7) Expressing the electric field in spherical coordinate, the incident electric field at pointP is Ui(r) = P(r) cos Vap - P(r) sin Va. (2.8) a. is the tangent direction unit vector, and ap is the radius direction unit vector. After the refraction by the high-NA objective, a. does not change its direction, while ap changes its direction into ao. Thus, P(r) in the above equation becomes P(0), as equ. 2.9 shows below. Ja 0 Ui(r) = P(0) cos Va, - P(0) sin a = cos 0 cos pi + cos 0 sin Cpj + sin 0k (2.9) -sinVi +cosWk a, Thus, substituting Ui(r) to equ. 2.7 to get the vectorial Debye integral [341. According to this theory, only circular polarized beam modified by vortex 0-27r phase plate can focus as a round donut shape. For ellipse polarization or linear polarization, the intensity around the 'dark center' is not uniform or the center has residue light intensity. Also, the circular polarized beam should be the same spiral direction with the vortex phase plate. To achieve the most uniform donut spot as depletion beam, we used a half wave plate (IIWP) and a quarter wave plate (QWP), as well as a pair of Glan-Thompson prism to purify the polarization. The QWP is designed to put right in front of 38 UnwrY 1 .1 ENpse X 1 0 -0 5 .06 , 000 0 0 a0- 0 0 Left44aaded 2-3 ztinsar 00 (D0O' 1 rized g sd .a rriaW an 00 00 ed 00 050.6 00.0 ,O 0 11. Figure 2-13: Normalized intensity of polarized light with a vortex 0-27r phase plate at the focal spot in the horizontal (XY) and longitudinal (XZ) planes. Various polarizations are considered: linear, circular (right-handed and left-handed), right-handed ellipse (with X and Y as the long axis). Intensities of 0 and 1 correspond to black and white, respectively. The axis units are in wavelengths. [81 objective lens, in order to minimize the polarization and phase distortion caused by dichroic mirror surface. However, the illumination beam is about 940nm and UV light, and the fluorescence is about about 515nm. If we also put the QWP right before the objective lens, it should be a very broad band QWP which is expensive. So, we still choose to put is only in the infrared path but combined with adaptive optics to compensate the influence of dichroic mirror. The adaptive optics will be discussed further in Chap. 3. 2.4.2 Experiments of donut shape focus in different polarization In the experiment, SLM, a HWP combined with a Glan-Thompson prism and a QWP mainly control the polarization of the light entering objective lens pupil (see figure ??). Especially the SLM plays an important role. For a fixed SLM phase and HWPGT fast axis direction, as we turning QWP, the polarization after modification could be left-hand polarized or right-hand polarized, or ellipse polarized or linear polarized between these two circular polarized statuses. 39 The focus image is detected by a CMOS camera (figure 2-1). Because of the size of pixels of the camera, the center of the donut is shown as not zero intensity. To get more accurate PSF of donut shape focus, we will use sub-diffraction silver bead to scan the PSF and reconstruct the image, which will be discussed in Chap. 4. The SLM vortex 0 - 2-r phase and donut focus before modify is shown in figure 214. The gray scale of the phase is modified to get the most uniform circular symmetric donut spot. (b) (a) Figure 2-14: The vortex 0 - 27r phase and corresponding donut shape focus. The center grayscale for the vortex is 58 and it is a left-hand vortex. As we turn the HWP-GT fast axis in different direction, the intensity of the focus and quality of donut shape are different. As an example, we randomly choose one direction and show the quality of focuses for left-hand circular polarized beam and right-hand circular polarized beam, respectively (figure 2-15). By turning HWP-GT and QWP, we observe. the donut focus quality to select the best combination of the directions of these two. To measure the exact polarization, we could also use Polarization Analyzing System (PAX5710, Thorlabs), which shows the polarization on a Poincare sphere (not available in our lab). Figure 2-16 compared the donut shape focuses before polarization modify and after polarization modify. After the polarization modify, the focus is more circular symmetric and intensity is more uniform distribution. Coincidently, the start polarization direction in this time test is close to the best combination, so the contrast is not very remarkable. Further 40 (a) (b) Figure 2-15: (a) The focus of right-hand polarized beam modified by left-hand vortex 0 - 27r phase plate. (b) The focus of left-hand polarized beam modified by left-hand vortex 0 - 27r phase plate. Two pictures are under the same color bar scale. (a) (b) Figure 2-16: Donut shape focus (a) before polarization modify and (b) after polarization modify. The two images are of the same colorbar scale. The left figure is the same as figure 2-14 (b). phase modification of donut shape focus will be discussed in Chap. 3. 41 42 Chapter 3 SLM Phase Control The SLM is one of the key components in this system. It plays three roles: first, it generates the vortex 0 - 27r phase plate in order to create the donut shape focus in the RESOLFT super-resolution microscopy; second, it generates the compensate phase aberration mask to optimum the system performance; third, it generates the parallel spots array or arbitrary spots for parallel and flexible imaging. The first function is already discussed in the Chap. 2. In this chapter, we will talk about the algorithm and LabVIEW control for the other two functions of SLM. This chapter relates more to electric and computer sections of the microscopy. 3.1 Adapted Optics As we discussed in the Chap. 2, a uniform circular donut shape spot requires high accuracy in phase, polarization, alignment, and so on. For our setup, one advantage is that we can use adapted optics to correct the system aberration to achieve better Gaussian excitation spot and donut shape 'turn-off' spot. Another advantage is that we can use spherical aberration to slightly reduce the intensity of excitation spot, in order to avoid 'turn-off' happened before enough excitation (the competition between excitation and 'turn-off' is discussed in chap. 1). 43 3.1.1 Optical aberration and Zernike polynomials The aberration of an optical system can be described by Zernike function [291. Because the donut spot and excitation spot are both at about 940nm wavelength, we don't consider about chromatic aberration correction in this setup. Also, we only consider about paraxial system, so the distortion related to high order of departure of axial image is not considered. The Zernike function describes aberrations from low order to high order. However, most of the 'obvious' aberration is caused by low order aberration, while higher order aberration occupies less percentage. Thus we mainly correct the low order aberrations, as listed in Table 3.1. Table 3.1: Low-order Zernike mode for optical aberration Aberration Mode Zi Definition Z2 (-1, 1) 2r cos 0 Figure i Tilt Defocus Z3 (1, 1) 2r sin 0 Z5 (0, 2) V/(2r 2 _ 1) Z 4(-2, 2) v'-/r2 cos(26) Z 5 (2, 2) vf6r 2 sin(20) Zs(-1, 3) 3/2(3r3 - 2r) cos 0 Z(1, 3) 2/2(3r 3 - 2r) dos 0 I: Astigmatism -WI Coma iI 44 3.1.2 trefoil Z7 (-3, 3) 32T/3r' cos 30 trefoil Zio (3, 3) 2/-3 r3 sin 30 Primany Spherical Z1(0,4) v 5 (6r. r- 6r2 + 1) 0 Critical metric in adapted optics for focus quality The system aberration is corrected by adaptive optics method [311. To correct the aberration, M. Booth and his colleagues defined 'intensity' and 'sharpness' as the critical metrics to judge the quality of focus. The intensity of the focal spot is measured, and then the sharpness and brightness of the spot are calculated. The image brightness B is the sum of the pixel values of the image, and the sharpness S is defined as the second moment of the image Fourier transform as S 2 ijL.,n' n,m + m'2 )/Z1:in,m n,m (3.1) where in,m is the Fourier transform of the image, n' and m' are the coordinate relative to the center of the image. pn,m is a low-pass filter with the radius w. To simplify, the combined metric of brightness and sharpness is defined as M = S + OcB + k(S-S) Sharpness is the main concern unless the sharpness is above the threshold (3.2) ST, which is normally chosen as the 90% of the maximum sharpness. a- is equal to +1 or -1 decided by theM is the maximum or minimum value. 3 and k are the experiment value to balance the contribution of sharpness and brightness. 45 - ml -. -=-- Ouins the aberration measurement, add different degree of aberration to the criginal phase input. For example, when the input phase is vortex 0 - 27r#,,. the t i.arg phase is 4 = # + bZi, b E {- 1, 1], Z,. is ith Zernike mode representing aberration. Adding different # to the SLM and calculate the metric M of each image, aod parameter b for the highest M is the compensate factor. I. 3 LabVIEW control and experiments for the aberration correction (Tenerally, different aberration modes, that is, the Zernike modes, are orthogonal to each other in the approximate paraxial condition. That is why it is convenient to mneasure them separately but can add together to modify the phase. However, under high-NA objective lens, the aberration correction causes focus displacement [311. In the other word, different modes are not orthogonal. So, we cannot measure them independently and add together. The former work measured the displacement of each modify mode and calculated the correlation. In contrast, we choose an iterative Atrategy. We check the influence of each aberration first, and then choose the most -erious aberration as 'base mode', then add every other mode to it to check the aew most serious aberration. Iteratively doing this process, we can achieve the least aberration system. The first example is the aberration correction for Gaussian beam focus. After aligning the system, we use, Rhodamine 123 solution as the fluorescent sample, berause of the brightness and uniformity. The illumination light is about 940nm. We ise the camera (see figure 2-1) as detector. We can observe the change of focus while changing the aberration degree. We show the influence of astigmatism to the focus quality (see figure 3-2), which is one of the serious aberration in that test. From the figure 3-2, we can tell that the intensity of focus is higher when compensating the system by large negative astigmatism. Because we use a 100x objective lens in the setup, the entrance pupil is smaller than the beam diameter. So only the top of Gaussian beam can pass through. This way limited the light intensity to the 46 Figure 3-1: LabVIEW control panel for system aberration compensation. focus, but make it more robust to astigmatism. Even though, according to the phase compensation experiments, astigmatism is still one of the most serious aberration in the setup. To quantitative analyze the result, we draw the three curves: brightness, sharpness, and the combination of these two factors, respectively. (Figure 3-3). Figure 3-3 shows the sharpness and brightness have nearly opposite trend. The sharpness curve shows the focus is sharper in positive compensation, but the brightness curve shows the intensity is higher in the negative compensation. If we compare the figures in figure 3-2, we can tell that the low sharpness of negative phase compensation part may be from the saturation intensity in the center. So we choose the brightness as major factor and choose b = -0.9 as the best compensation for the astigmatism. The second step is examine the most serious aberration after the phase correction of the first step. We put Zernike mode (-2, 2) with b = -0.9 as 'base mode', then adding every other aberrations in different degree. After our experiment, the coma 47 Figure 3-2: The focus spots under different degree of astigmatism (Zernike mode (-2, 2), Table 3.1). Adptiv Adaptive Curve: Combination M Adaptive Curve: Brightness Curve: Sharpness 1 -- .i- - - -1 - - Di080.8 . .... .... 0.6..0.6-- . . ... 0.46 .. A.. 0.4 0.2 - 001 -05 0.2 - - .. ...... .......... - 0. -... . .. .. - - 0. 0 Was 0.5 1 -1 -0.5 0 Bias 0.5 1 -1 -0.5 0 Bias 0.5 1 Figure 3-3: The sharpness, brightness, and the combined metric v.s. bias (b) curves. The combined metric curve will change if change the 'weight' between sharpness and brightness, depending on which one is major. (Zernike mode (3, 1), see Table 3.1) is the most serious aberration in this round. The focus quality is shown in figure 3-4 as the system PSF under the synthesis compensation of astigmatism and coma. Because we test the vertical direction coma, it is obvious that the focus is moving along vertical direction. We can tell the quality of focus is better than the first round testing generally (figure 3-2). Then we calculate the sharpness and brightness, as well as combined metric as before. The result is shown in the figure 3-5. According to the 2nd round quantitative comparison, the aberration is least when adding b = -0.9 astigmatism and b = 0.4 coma. The combined phase is shown in figure 3-6 (c). The compare of the focus before and after aberration compensation is in figure 3-6 (a) and figure 3-6(b). We can tell that the focus is brighter and more circular symmetric after aberration compensation. 48 This method cannot guarantee Figure 3-4: The focus spots under fixed astigmatism (b = -0.9) compensation while testing different degree of coma (Zernike mode (3, 1), Table 3.1). Adapve Curve: Sharpness Adapitve Adaptive Curve: CornbinatanM Curve: Brightness 1 1 ...... .....- -0.8 0.8 0.o... 8-.... 00 " .......... 02 0.2 -1 0 0 Bias 0"1 0.6 ........ 0.2 -0.5 0 0.5 1 -1 -0.5 0 05 1 Bias. Bias Figure 3-5: The sharpness, brightness, and the combined metric v.s. bias (b) curves. In the second round testing, the sharpness and brightness have similar trend, which means the aberration of the system converge to a 'local minimize' point under these two phases compensating. the aberration compensation is the least globally, but can make sure this is the local least and already corrected most serious aberrations in the system. Under sufficient time, this phase compensation iterative process could be more accurate. The next example is the aberration correction for donut shape focus. The method of trials to determine the 'local best' donut shape focus is the same as Gaussian focus. However, the quantitative metrics to judge the quality of focus is different. The brightness and sharpness defined as above cannot accurately describe the characteristics of donut shape focus. The intensity of a uniform donut shape spot is less than that of a Gaussian focus because of the diffraction loss from SLM. The donut shape frequency domain is more complicated than the Gaussian focus, so the sharpness calculation is without close relation to uniform donut shape. 49 As we focus on (b) (a) (c) Figure 3-6: (a) The original Gaussian beam focus without aberration compensation to the system. (b) The focus after aberration compensation. (c) The compensated phase adding to the SLM, which is sum of negative astigmatism and positive vertical coma. the uniformity of donut shape, the standard deviation around the dark center is an important metric under consideration. The result is shown below. The experiment figures as different aberration compensations added are shown in figure 3-7. The process is the same as Gaussian beam experiment above. To simplify the result, we choose to show every other result as bias scanning from -1 to 1. For the 1 " round test, Zernike mode (1, 1) has obvious distortion, means the displacement matters to the donut focus. Figure 3-7: The donut shape focus under Zernike mode (-1, 1) and bias b = -1 : 1 0.5, the donut shape is the modify, selectively shown every other figures. When b most uniform one. Based on the 1" round result, Zj-1 at b = 0.5, the 2 " round tests all the other aberration added on this base mode. One representative result is Zernike mode (3, 3), which is the trefoil aberration in Table 3.1. The result is shown in figure 3-8. When b = -0.4, the donut shape is the most uniform one (skipped in figure 3-8 but shown in figure 3-9). After these two round modification, the donut shape is already much more uniform and symmetric than before modification (see figure 3-9). 50 Figure 3-8: The donut shape focus under Zernike mode (3, 3) and bias b = -1 : 1 modify, selectively shown every other figures. As the same as Gaussian beam, this method will achieve better result if more iterative testing processed. (a) (b) (c) Figure 3-9: (a) The original donut shape focus without aberration compensation to the system. (b) The focus after aberration compensation. (c) The compensated phase adding to the SLM, which is sum of positive shift and a negative trefoil. (a) and (b) are under the same color scale. 3.2 3.2.1 Phase retrieval algorithm Gerchberg-Saxton iterative algorithm The phase retrieval of arbitrary position multi-focus uses the classic Gerchberg-Saxton iterative algorithm modified by Fienup [9]. The principle of the iterative Fourier transform (FT) algorithm is figure 3-10. For the signal changing from kth loop to k + 1, the iterative equation is where / is a free parameter, usually chosen close to one. This method has high efficiency to converge. When we already know the aimed position of multiple focuses AT, the phase on the back focal plane is calculated iteratively until converge. Add this retrieved phase to the vortex 0 - 27r phase can generate donut spots in the 51 Ak Back-propagation IFT k Phase space apply ( constraints (SLM phase input) Signal space apply constraints (k) (Target intensity image) YES Ak function satisfied? ak Propagation FT NO Recovered phase Figure 3-10: The principle of the iterative Fourier transform algorithm. [9] corresponding position. Thus, it is possible to locally scanning and do super-resolution. imaging in several sub-regions simultaneous. This flexible imaging method has many 3otential application in biological research. For example the method can be used to image sparse aims in the specimen, which is more efficient than scanning the whole field of view. For small area scanning, it is possible to do live imaging for 3everal sub-regions, such as do multiple super-resolution tracking for several protein .ransportation or interaction. 3.2.2 LabVIEW control and experiment for arbitraryposition spots According to the G-S algorithm, we programmed the control panel by LabVIEW (figure 3-11). The aim position is randomly choose from the 'new picture', and the coordinate value will be collected and process to the phase retrieval calculation. After -he calculation, the recovered phase is sent to SLM screen, so we can get the aimed focuses on the camera. Repeating this process can control the focuses position in real time. There are several potential applications for this method. For example, the 52 '.Gaussian focuses spot can be used as optical tweezers, which can control the motion of molecules. Because we can do several focuses at the same time, it is possible to do multiple particle tracking, in order to track their transportation as well as the interaction between each other: Combined with donut shape spots at the same positions, we can achieve higher resolution. Another application is super-resolution image for several specific areas. The 'new picture' could be a low-resolutiion image, and we can select several regions requiring higher resolution, while other sparse area without scanning again. This can help us get super-resolution images more efficiently. Figure 3-11: The control panel of flexible position focuses. The aimed focuses are selected from the 'new picture' window. In the experiment, we use one pixel point (pulse function) as the aim spot. If we use several pixels as one spot, we should add the extra limit to make sure the intensity of all the pixels inside the spot is identical. But actually, the intensity of different spots are different in the experiment (see figure 3-12 (c)), even the intensity 53 of aimed spots are the same. Adding linit to the difference between focus intensities should help reduce the effect. (a) (b) (c) Figure 3-12: . (a) The focuses in random positions, selected from the control panel. (To easy observation, enlarged the focuses area, 10 times, each spot here should be one pixel size in the calculation.) (b) The phase mask calculated by G-S algorithm. (c) The camera image of the phase (Rhodamine 123 solution as sample), which is the same as the aimed focuses. If combine the retrieval phase with vortex 0- 27r phase mask, we can achieve donut shape focuses on the same aimed position. Figure 3-13 shows the process. Compared to the Gaussian focus, donut shape focus is more difficult because we should consider about the uniformity of the donut. Noticed that the intensity of each spot is not perfectly uniform, this could be improved by more iterative loops, or set the target rule as the standard deviation of intensity is smaller than a certain threshold. Because of the aberration compensation in Chap. 3-1 is for the whole field of view, it is hard to compensate the aberrations for each donut shape focus. 54 Phase recover Vortex 0-2pi phase plate Experiment image (output) * * & Aim position (Input) LabVIEW SLM Camera Figure 3-13: The flexible position of donut spots, for multiple super-resolution particle tracking. The donut shape in the 'input image' is enlarged. Like the Gaussian focuses, one pixel spot to indicate the coordinate value of the position is sufficient. 55 56 Chapter 4 Two-photon RESOLFT microscopy imaging 4.1 PSF measurement: silver nanoparticle scanning The figures taken by camera is actually the real image sampling by camera pixels. Because the 'dark center' of donut spot is very small (<50nm), even enlarged by the objective lens (100x), but still smaller than the pixel size of the camera (about 7um). To get more accurate PSF of the system, we use nanoparticles to scan the PSF. Such as silver nanoparticle, the scatter light from the silver nanoparticle is strong and sensitive to the intensity change. So the measured PSF becomes the convolution of silver particle with the real PSF. The smaller particle we use, the more accurate the PSF it. But scatter light intensity from small particle is weaker as well. At the same time, the scatter light intensity also relates to wavelength. We use 40nm silver nanoparticle (Sigma-Aldrich) in the experiment. One advantage of the setup is the Gaussian beam and donut shape beam share the same beam, but with modulation temporally. So the colocalization of these two beam is accurate. Figure 4-1 shows the bead-scanned PSF of the Gaussian focus, donut focus and the overlay of them. The wavelength is about 940nm. The intensity is the square of the original measured intensity, to show the two-photon PSF distribution (equation 1.2). 57 (a) (b) (c) .gure 4-1: nanoparticle-scanned PSF of the setup. (a) Gaussian focus. (b) Donut lape focus, with the same intensity scale. (c) Overlay of (a) and (c) to show the >c4ocalization (pseudo color). We also measured the PSF of single photon excitation in the wavelength of 532nm. Figure 4-2 shows the comparison of the PSF between single photon excitation and two-photon excitation. For the donut beam, two-photon excitation gave darker center point. But at the same time, the peak-to-peak distance is larger. Considering the experiment condition, it is very hard to achieve 'zero intensity' in the center of donut spot, because of the background, or limited size of NA. However, if the residue intensity over 5% of the peak intensity, the super-resolution image signal to noise ratio will be attenuated [351. From this view, two-photon process can give better quality donut in the experiment. Another concern we have is whether the donut shape will distort as imaging depth increases. If the peak-to-peak distance is larger, maybe the donut spot could tolerate more distortion caused by specimen scatter. This is pending to prove by experiment. 58 .. - - 0.8 0.6 0.8 - 1 06 / 0.4- 0.4 0.2 0.2 0 9500 -50 50 500 0 (b) (a) Figure 4-2: The PSF comparison between single photon excitation (blue dash line) and two-photon excitation (red solid line). (a) The Gaussian focus. (b) The donut shape focus. 4.2 Parallel scanning: Gaussian spots array and donut shape spots array One drawback of RESOLFT microscopy is that it is a point scan method, that is, the imaging speed is proportional to the size of field of view (FOV). However, it is difficult to achieve large FOV as well as high-resolution simultaneously. Parallel scanning is one common method to increase the imaging speed for point scan microscopy [?,2,24]. In our experiment, we generated 5x5 spots matrix, also the corresponding donut spots matrix. By zoom-in and zoom-out the grating on SLM, we can control the separation and position of the spots arrays flexibly (see figure 4-3). While the parallel scanning increases scanning speed, it requires more power to ensure every section of image has good signal to noise ratio. Limited by the maximum energy, we did not get good signal to noise ratio image of donut spots array in twophoton excitation setup. The flowing donut array images are single photon excited using 532nm wavelength laser. We also tried even larger spots matrix (like 11x11), but it is hard to guarantee the quality of every donut shape focus, and the intensity 59 (a) (bi) (c) Figure 4-3: 5x5 spots array, two-photon cxcitation using Rhodamine 123 sample, detected by the camera. The period of grating on SLM gradually zoom in, and the corresponding image 'zoom out'. of each spot is even lower. Figure 4-4: The 5x5 donut spots matrix for parallel super-resolution imaging. 4.3 Live cell two-photon imaging Two-photon RESOLFT microscopy is based on two-photpn microscopy. Without change the phase mask on SLM, this system can be used as a two-photon microscopy (single point scan or parallel scan). First we use HEK 293 ce.l as specimen, labelled 60 by Dronpa-M159T (Table 1.1) on tubulin ". The early passage frozen cell and DNA plasmid are gift from Prof. Edward Boyden's lab. The cell culture follows the typical mammal cell culture protocol, and the transfection uses TransIT 293 reagent (Mirus) and follow the protocol of it. The fluorescent images are taken after 18h-24h of transfection. (a) (b) Figure 4-5: HEK 293 cell labelled by Dronpa-M159T on tubulin. (a) The field of view (FOV) of the image is 60um x 60um. The scanning step is 500nm. (b) The fine scanning image of the top right region of (a), the FOV is 30um x 30um. The scanning step is 200nm. Because we use live cell and the imaging process takes minutes, we can observe the cell position change during continuous two times imaging (figure 4-6). We imaged HeLa cell line also, which microtubule extends to larger area than HEK cells. We can distinguish single microtubule labeled by Dronpa under the two-photon RESOLFT microscopy. The resolution reaches about 400nm (200nm per pixel). 'The color of the image is pseudo color, based on the emission peak wavelength of Dronpa-M159T is 515nm. The contrast is adjusted. 61 (a) (b) Figure 4-6: The position change of live cecll. (a) 1st time scanning, FOV is 80um x 'Oum, and step size is lum. (b) 2nd time scanning for the right bottom region of (a), oticing the bottom cell orientation changed during this process. The FOV is 30uw 3 9 um with 200nm step size. 150 04 ./.. .... Am 0 1000 200 3000 4000 Distante (iXeHlS) .. ...... 4/W 5000 0 00 Figure 4-7: HeLa cecll labeled by Dronpa-Tubulin. The enlarged region is increased contrast by ImageJ. The cross profile of the red line region is shown below. The finest structure we can distinguish is about 400nm. The image is 50um x 50um, with 200nm scanning step. 62 Chapter 5 Conclusion We designed a two-photon RESOLFT microscopy with SLM control, which can apply to high speed 3D high-resolution imaging by parallel scanning, as well as highresolution multiple particle simultaneously tracking by flexible position scanning. The system can achieve less aberration imaging via phase compensation. We showed the algorithm and experiment results of these unique characteristics in the paper. Iterative phase retrieval algorithm is used to calculate the phase mask on SLM for flexible position imaging. aberration. Phase compensation using adaptive optics is applied to reduce As for the experiment result, we showed the PSF and time sequence of the three beams in the RESOLFT microscopy section. For parallel imaging or flexible imaging, both Gaussian spots (excitation beam) and donut spots ('turn-off' beam) experiment images are shown. At the last, the two-photon images of the cell specimen labeled by reversibly photoswitchable fluorophore are demonstrated. At the same time, the design could be further promoted by higher speed scanning, such as using resonant mirror instead of piezo stage. This will improve the image system efficiency, making the live imaging possible. The other drawback is the absolute intensity of super-resolution image is low, because the fluorescence was turned off in the surrounding area and only limited area emits fluorescence. noise ratio reduced compared to the two-photon images. Thus, the signal to sCMOS camera or other more sensitive detector with less background noise could improve the image quality further. After these modification, the large field of view, high-speed, high-resolution 63 microscopy could be realized. This microscopy is potentially used for imaging large area of specimen with 3D high-resolution in detail, such as single synapse event as well as the synapse circuit in the brain s'icc +n vivo. 64 Bibliography . [11 Ilaria Testa, Nicolai T. Urban, Stefan Jakobs, Christian Eggeling, Katrin Willig, and Stefan W. Hell. Nanoscopy of living brain slices with low light levels. Neuron, 75(6):992-1000, 2012. [21 Andriy Chmyrov, Jan Keller, Tim Grotjohann, Michael Ratz, Elisa d'Este, Stefan Jakobs, Christian Eggeling, and Stefan W. Hell. Nanoscopy with more than 100,000 'doughnuts'. Nature methods, 10(8):737-740, 2013. [31 A. Stiel, Simon Trowitzsch, Gert Weber, Martin Andresen, Christian Eggeling, S. Hell, Stefan Jakobs, and M. Wahl. 1.8 a bright-state structure of the reversibly switchable fluorescent protein dronpa guides the generation of fast switching var- ants. Biochem. J, 402:35-42, 2007. [41 Satoshi Habuchi, Ryoko Ando, Peter Dedecker, Wendy Verheijen, Hideaki Mizuno, Atsushi Miyawaki, and Johan Hofkens. Reversible single-molecule photoswitching in the gfp-like fluorescent protein dronpa. Proceedings of the National Academy of Sciences of the United States of America, 102(27), 2005. [51 Min Gu. Principles of three dimensional imaging in confocal microscopes. Singapore: World Scientific, 1996. [61 Bahaa E. A. Saleh and Malvin Carl Teich. Fundamentals of Photonics. Wiley Interscience, 1991. [71 Max Born and Emil Wolf. Principles of optics: electromagnetic theory of propagation, interference and diffraction of light. Cambridge university press, 1999. [81 Hao Xiang, Cuifang Kuang, Tingting Wang, and Xu Liu. Effects of polarization on the de-excitation dark focal spot in sted microscopy. Journal of Optics, 12(11):115707, 2010. [9] J. R. Fienup. Phase retrieval algorithms: a comparison. Applied optics, 21(15), 1982. [101 S. M. Guo, R. Veneziano, R. E. McConnell, S. Gordonov S. Agasti, T. Kulesa, and M. Bathe. Quantitative multiplexed super-resolutionneuronal synapse imaging using dna-paint. Biophysical Journal, 108(2):477a, 2015. 65 [l] Hy'uk-.anq Kwon, Yoon Sung ANam, Dominika M. Wiktor-Brown, Bevin P. En.geiward, and Peter T. C. So. Quantitative morphometric measurements using site selective image cytometry of intact tissue. Journal of The Royal Society Interface, 6, 2009. [12j Emmanuel Adu-Gyamfi, Michelle A. Digman, Enrico Gratton, and Robert V. Stahelin. Single-particle tracking demonstrates that actin coordinates the movement of the ebola virus matrix protein. Biophysical journal, 103(9):L41-L43, 2012. 113] Michael J. Rust, Mark Bates, and Xiaowei Zhuang. Sub-diffraction-limit imaging by stochastic optical reconstruction microscopy (storm). Nature methods, 3(10):793-796, 2006. [14] Eric Betzig, George H. Patterson, Rachid Sougrat, 0. Wolf Lindwasser, Scott Olenych, Juan S. Bonifacino, Michael W. Davidson, Jennifer LippincottSchwartz, and Harald F. Hess. Imaging intracellular fluorescent proteins at nanometer resolution. Science, 313(5793):1642-1645, 2006. /151 Stefan W. Hell and Jan Wichmann. Breaking the diffraction resolution limit by stimulated emission: stimulated-emission-depletionfluorescence microscopy. Optics letters, 19(11):780--782, 1994. [16] Mats GL Gustafsson. Nonlinear structured-illuminationmicroscopy: wide-field fluorescence imaging with theoretically unlimited resolution. Proceedings of the National Academy of Sciences of the United States of America, 102(37):1308113086, 2005. [17/ Flavie Lavoie-Cardinal, Nickels A. Jensen, Volker Westphal, Andre C. Stiel, Andriy Chmyrov, Jakob Bierwagen, Ilaria Testa, Stefan Jakobs, and Stefan W. Hell. Twod iAcolor resolft nanoscopy with green and red fluorescent photochromic proteins. ChemPhysChem, 15(4):655-663, 2014. & [181 Fedor V. Subach, Lijuan Zhang, Theodorus W.J. Gadella, Nadya G. Gurskaya, Konstantin A. Lukyanov, and Vladislav V. Verkhusha. Red fluorescent protein with reversibly photoswitchable absorbancefor photochromic fret. Chemistry biology, 17(7):745-755, 2010. [19] Martin Andresen, Andre C. Stiel, Simon Trowitzsch, Gert Weber, Christian Eggeling, Markus C. Wahl, Stefan W. Hell, and Stefan Jakobs. Structural basis for reversible photoswitching in dronpa. Proceedings of the National Academy of Sciences, 104(32), 2007. /20! Tim Grotjohann., Ilaria Testa, Marcel Leutenegger, Hannes Bock, Nicolai T. Urban, Flavie Lavoie-Cardinal, Katrin L Willig, ChristianEggeling, Stefan Jakobs, and Stefan W. Hell. Diffraction-unlimitedall-6ptical imaging and writing with a photochromic gfp. Nature, 478(7368), 2011. 66 [211 Peter TC So; Chen Y. Dong, 'Barry R. Masters, and Keith M. Berland. Twophoton excitation fluorescence microscopy. Annual review of biomedical engi- neering, 2(1):399-429, 2000. [22] Nicholas G. Horton, Ke Wang, Demirhan Kobat, Catharine G. Clark, Frank W. Wise, Chris B. Schaffer, and Chris Xu. In vivo three-photon microscopy of subcortical structures within an intact mouse brain. Nature photonics, 7(3):205- 209, 2013. [231 Andy Y. Shih, Jonathan D. Driscoll, Patrick J. Drew, Nozomi Nishimura, Chris B. Schaffer, and David Kleinfeld. Two-photon microscopy as a tool to study blood flow and neurovascular coupling in the rodent brain. Journal of Cerebral Blood Flow & Metabolism, 32(7):1277-1309, 2012. [24] Karsten Bahlmann, Peter T. So, Michael Kirber, Robert Reich, Bernard Kosicki, Multifocal multiphoton microscopy William McGonagle, and Karl Bellve. (mmm) at a frame rate beyond 600 hz. Optics express, 15(17):10991-10998, 2007. [25] Paolo Pozzi, Daniela Gandolfi, Marialuisa Tognolina, Giuseppe Chirico, Jonathan Mapelli, and Egidio D 'Angelo. High-throughput spatial light modulation two-photon microscopy for fast functional imaging. Neurophotonics, 2(1):015005-015005, 2015. [26] Philipp Bethge, Ronan Chereau, Elena Avignone, Giovanni Marsicano, and U. Valentin Nagerl. Two-photon excitation sted microscopy in two colors in acute brain slices. Biophysical journal, 104(4):778-785, 2013. [271 Barry R. Masters and Peter T. C. So. Handbook of Biomedical Nonlinear Optical Microscopy. Oxford University Press, 2008. [28] Winfried Denk, James H. Strickler, and Watt W. Webb. Two-photon laser scanning fluorescence microscopy. Science, 248(4951):73-76, 1990. [29] Frank L. Pedrotti and Leno S. Pedrotti. Introduction to Optics, chapter 5. Prentice Hall, 1993. [30] Yi Xue, Cuifang Kuang, Shuai Li, Zhaotai Gu, and Xu Liu. Sharperfluorescent super-resolutionspot generated by azimuthally polarized beam in sted microscopy. Optics express, 20(16):17653-17666, 2012. [31] Travis J. Gould, Daniel Burke, Joerg Bewersdorf, and Martin J. Booth. Adaptive optics enables 3d sted microscopy in aberrating specimens. Optics express, 20(19):20998-21009, 2012. /32] Pedro Felipe GardeazAqbal RodrAfyguez, Yong Wu, Harpreet Singh, Hui Zhao, Ligia Toro, and Enrico Stefani. Building a fast scanning stimulated emission depletion microscope: a step by step guide. Current Microscopy Contributions to Advances in Science and Technology, pages 791-800, 2012. 67 /03] Bevjamin Rarke, Jan Keller, Chaitanya K. Ullal, VkCr 1Westphl Andreas SchA'iinle, and Stefan W. Hell. Resolution scaling in sted micrscopy. Optics express, 16(6):4154-4162; 2008. /34] Mint Gu. Advanced optical imaging theory. Media, 1999. Springer Science and Business |35] Giuseppe Vicidomini Gabriele Lignani Fabio Benfenati Alberto Diaspro Silvia Galiani, Benjamin Harke and Paoio Bianchirti. Strategies to maximie the erformance of a sted microscope. Optics express, 20(?):7362-7374 2008 68