Far-Field Super-Resolution Vibrational Spectroscopy | Analytical

Jun 28, 2019 - PDF (4 MB) ..... The top row represents a cross section of the simulated images shown in .... suppression offer great theoretical poten...
1 downloads 0 Views 4MB Size
Feature pubs.acs.org/ac

Cite This: Anal. Chem. 2019, 91, 8723−8731

Far-Field Super-Resolution Vibrational Spectroscopy Potential label-free alternatives to super-resolution fluorescence techniques have been the focus of considerable research due to the challenges intrinsic in the reliance on fluorescent tags. In this Feature, we discuss efforts to develop super-resolution techniques based on vibrational spectroscopies and address possible sample applications as well as future potential resolution enhancements.

Downloaded via 5.188.217.216 on July 19, 2019 at 15:18:51 (UTC). See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.

Christian T. Graefe, David Punihaole, Celina M. Harris, Michael J. Lynch, Ryan Leighton, and Renee R. Frontiera* Department of Chemistry, University of Minnesota, Minneapolis, Minnesota 55455, United States However, Abbe’s diffraction limit restricts spatial resolution to hundreds of nanometers when visible light is used, as it is in most optical microscopies.8 This is described by the equation d=

(1)

where d is the smallest distance apart that two different point sources of light can be while still being resolved from one another, λ is the wavelength of light, n is the refractive index of the imaging medium, and θ is the maximum half angle of the cone of light that can enter or exit the optical system.9 The quantity n sin θ defines the numerical aperture (NA) of the system. Using this definition, resolution is limited to 200 nm with 400 nm light and an NA of 1. Equation 1 was amended slightly by Lord Rayleigh such that

N

anoscale heterogeneities often contain important structural and dynamic information that impacts macroscopic processes in many material, biological, and chemical systems. For example, the cell membrane contains highly localized environments composed of different lipids, proteins, and other biomolecules where crucial cellular processes such as signal transduction, ion and molecular transport, and small molecule binding and activation occur.1 These processes are vitally important in areas such as drug development and delivery as membrane proteins were the targets of over 60% of marketed pharmaceuticals as of 2006.2 However, it is not well understood how the surrounding nanoscale membrane environment can impact function. In material samples, nanoscale defects and boundaries are frequent, impacting properties such as exciton transport, electrical conductivity, and magnetic conductivity.3,4 In photovoltaics, efficient and consistent charge transfer is necessary for a number of material classes to perform optimally, and grain boundaries can significantly hinder the efficiency of charge transport, reducing photovoltaic efficiency. Despite their impact, little is known about what occurs at these boundary interfaces.5 It is challenging to isolate interface properties from bulk properties because of the size of these interfacial domains. A major challenge in broadly determining nanoscale structure−function relationships arises from the limited number of techniques available to probe dynamic and living systems on the 10−100 nm length scales. Cryogenic electron microscopy (cryo-EM) offers unprecedented spatial resolution down to angstrom scales; however, samples must be flash frozen, limiting cryo-EM’s ability to provide dynamic information. Additionally, samples sustain damage from the electron beam due to the high energy and ionizing effects of the electrons.6,7 Optical microscopy techniques are useful tools for determining the composition and structure of a sample. © 2019 American Chemical Society

λ 2n sin θ

d=

0.61λ n sin θ

(2)

Lord Rayleigh’s definition differs from Abbe’s in that it defines the resolution limit as the distance from the center of an Airy disk diffraction pattern to its first minimum. The invention of super-resolution techniques that break the optical diffraction limit has opened the door for research on the crucially important nanometer length scale. Pioneering fluorescence-based techniques, such as photoactivated localization microscopy (PALM),10 stochastic optical reconstruction microscopy (STORM),11 and stimulated emission depletion (STED) microscopy,12 have become particularly widespread and the techniques were recognized with the 2014 Nobel Prize in Chemistry.13 These techniques have benefited the fields of neuroscience and structural biology, among many others.14−18 While these super-resolution techniques have been revolutionary, fluorescence-based techniques have some inherent disadvantages that make them poorly suited for certain applications. Most problems stem from the requirement for samples to be tagged with fluorescent probes. According to PALM developer Eric Betzig, “For fluorescence microscopy it always comes back to labeling, this hampers everything.”19 Fluorophores frequently photobleach, particularly in oxygenrich environments.20,21 The resulting loss of observed signal leads to difficulties in tracking dynamics and working with live Published: June 28, 2019 8723

DOI: 10.1021/acs.analchem.9b01731 Anal. Chem. 2019, 91, 8723−8731

Feature

Analytical Chemistry

Figure 1. Figures depicting the concepts described in this Feature to achieve subdiffraction-limit resolution. (a) In infrared photothermal heterodyne imaging (IR-PHI), a visible beam is used to measure the reflectivity of the sample after excitation with an IR beam, resulting in resolution at the visible diffraction limit. Adapted from ref 35. Copyright 2017 American Chemical Society. (b) Standing wave attenuated total reflection uses two evanescent pump waves to create an interference pattern, saturating local vibrations. A probe beam is focused between the two most intense fringes, resulting in subdiffraction limit resolution. Adapted with permission from ref 36. Copyright 2017 Elsevier B.V. (c) In structured illumination microscopy (SIM), the sinusoidally patterned excitation light allows higher frequency spatial modes to be detected. Adapted with permission from ref 37. Copyright 2005 National Academy of Sciences, U.S.A. (d) Spatial light modulators (SLMs) alter the shape or phase profile of beams in order to shrink the central lobe of the point spread function (PSF).38 (e) Raman suppression methods use a doughnut-shaped pulse to deplete Raman signal in a spatially defined area, leaving a smaller area in the middle of the doughnut where signal can be generated. Adapted from ref 39. Copyright 2016 American Chemical Society.

the sample, and tips can be perturbative to the point of driving chemical reactions.34 We focus on far-field vibrational superresolution microscopy methods due to their greater potential for reproducibility and ease of use. Raman spectroscopy is a vibrational technique that has great potential to access the sub-100 nm length scale as it can be performed using much shorter wavelengths than infrared (IR) absorption spectroscopy. However, the significantly lower Raman scattering cross sections compared to IR absorption or fluorescence cross sections present a notable challenge. The use of coherent Raman-based techniques, such as stimulated Raman scattering (SRS) and coherent anti-Stokes Raman scattering (CARS), that boost the signal intensity are promising for super-resolution Raman imaging. These fourwave mixing (4WM) techniques employ multiple beams with specific frequency differences to enhance the likelihood of Raman transitions.40,41 Efforts to develop and optimize various super-resolution techniques based on far-field vibrational spectroscopies are ongoing. In this Feature, we discuss several developing IR spectroscopy techniques and focus on three approaches that can achieve subdiffraction-limit resolution with Raman methods (Figure 1). We discuss potential applicability of the techniques and challenges that they face. As this field is relatively young, the techniques we discuss are in the early stages of development or theoretical evaluation as compared to the field of super-resolution fluorescence, which has been under development for decades and has become widespread. As such, the resolution of these techniques does not yet compare to that which is achievable by fluorescence. In some cases, proof-of-concept results indicate resolution comparable to that of fluorescence-based methods is achievable. These results represent encouraging initial steps toward label-free super-resolution techniques, which would open the door to new potential avenues of research due to their chemically

samples. Super-resolution images can be relatively slow to acquire,22 and long acquisition times make photobleaching increasingly likely. In addition to photobleaching, fluorescent labels can disrupt the structure or dynamics of the system of interest.23,24 Finally, multiplexing is challenging as each analyte of interest must be bioorthogonally labeled with a unique fluorophore, which requires probing with multiple excitation lines, filter sets, and detectors. Techniques that are sensitive to an analyte’s inherent properties and composition would allow researchers to observe multiple analytes without the need for extra sample preparation. Label-free methods that attain comparable levels of spatial resolution would avoid the specific difficulties that come from working with fluorophores, and new systems would be opened to research on the nanoscale. Vibrational spectroscopies are a class of techniques that offer promising alternatives to fluorescence-based techniques. When combined with microscopy, these techniques give spatially resolved, chemically specific information about the molecular vibrations inherent in a sample. These vibrational signals can also be sensitive to the local chemical and physical environment, giving information about how the analyte’s structure changes in response to a perturbation or how it is influenced by its surroundings.25−30 This information can be used to investigate the impact of structural defects and heterogeneities on a material’s properties, for example. Vibrational superresolution microscopy techniques would be particularly useful with material samples for which labeling would hinder or disrupt packing, such as polycrystalline films. Near-field processes, such as tip-enhanced vibrational spectroscopies, can be used to enhance spatial resolution well beyond the farfield diffraction limit. Near-field scanning optical microscopy (NSOM) techniques were among the first optical techniques to break the diffraction limit and use either a tip or an aperture to confine light and improve spatial resolution.31,32 However, atomic-scale changes in tip fabrication make quantitative measurements challenging,33 tips must be extremely close to 8724

DOI: 10.1021/acs.analchem.9b01731 Anal. Chem. 2019, 91, 8723−8731

Feature

Analytical Chemistry specific signal, less perturbative nature, and greater potential for multiplexing.

sample, limiting its use to thin and thermally conductive samples. An IR-related technique that also brings spatial resolution into the visible diffraction-limited range was demonstrated by Hanninen et al.44 They developed a third-order sum frequency generation (TSFG) scheme that involves an IR transition followed by a two-photon hyper Raman transition. Unlike typical sum frequency generation (SFG), which is a surfacespecific technique,45 TSFG signals are generated in bulk samples. As the hyper Raman transition can be driven by visible or near-IR light, the spatial resolution of TSFG is better than in standard IR microscopy, while still providing information about IR-active modes. Another super-resolution IR method in development aims to surpass the diffraction limit by using an attenuated total reflection (ATR) configuration.36 This technique uses two identical pump beams directed at a prism above the sample from opposite directions, with incident angles greater than the critical angle (Figure 1b). The two resulting evanescent waves create an interference pattern that saturates local vibrational transitions, allowing for a probe beam aimed between the two most intense fringes to obtain subdiffraction-limited resolution. Though this method has yet to be experimentally demonstrated, resolution enhancements predicted by simulations show great promise for subdiffraction-limit IR imaging. Simulations have shown that this method could potentially achieve a spatial resolution of 100 nm using a laser with a wavelength of 3.5 μm. This technique would be well suited for looking at thin samples or surface features; however, the penetration depth is limited by the use of evanescent waves, making it unsuitable for imaging deeper into thick samples. These approaches offer great promise for surpassing the spatial resolution limits of traditional IR absorption techniques in a far-field manner and reaching the resolution of techniques in the visible region. However, they are currently far from achieving the resolution available with super-resolution methods utilizing fluorescence, such as STORM, PALM, and STED. In addition, IR-PHI and the ATR method are restricted to relatively thin samples. These drawbacks prohibit their usefulness for research involving nanostructured heterogeneity. Finally, water has a large IR absorption signal, making IR poorly suited for biological applications. Since Raman does not experience this interference, it is often a better choice when water cannot be avoided.



INFRARED-BASED SUPER-RESOLUTION TECHNIQUES Due to the use of significantly longer wavelengths, high spatial resolution in far-field IR absorption is fundamentally more challenging compared to microscopies performed in the visible range.42 In addition, the Cassegrainian objectives typically used for IR microscopy have a lower NA than the objectives commonly used with lasers in the visible spectrum.43 However, IR absorption processes can have reasonably large optical cross sections, allowing for a high signal-to-noise ratio. Here, we focus on several exciting new approaches that circumvent the diffraction limit for IR-based methods and allow for spatial resolution comparable to that achieved with visible light methods. One method reduces the spatial resolution of IR imaging down to resolution achievable with visible light through the use of infrared photothermal heterodyne imaging (IR-PHI).35 An IR pump beam is focused onto one side of the sample, while a visible probe beam is focused onto the other side (Figure 1a). When the sample absorbs the IR pump light, it heats up. This changes the reflectivity of the sample, which is monitored by measuring the intensity of the probe beam reflected off the back of the sample. The probe beam utilizes visible laser light with a high NA objective, allowing for the effective replacement of the diffraction limit of the IR beam and its Cassegrainian objective with that of the visible beam and its objective. Using this method, researchers imaged a 100 nm polystyrene bead with a spatial resolution of 300 nm (Figure 2).35 This is almost 7 times better resolution than the diffraction limit of the pump beam alone, which was 2060 nm. However, although this setup removes the inherent wavelength and NA limitations from IR spectroscopy, resolution is still limited by the diffraction limit of the probe beam. Additionally, this method requires heat to be transmitted through the



STRUCTURED ILLUMINATION MICROSCOPY When discussing structured illumination microscopy (SIM), it is helpful to consider a sample’s structure as a superposition of spatial Fourier modes, or k-modes, that are diffracted. According to Abbe’s diffraction limit, an image is formed only by the k-modes that are accepted by the collection optics in the microscope.37,46 Higher frequency k-modes are missed, and thus, the resolution limit of a diffraction-limited microscope is determined by the spatial frequency bandwidth of the microscope. SIM improves resolution by scattering the sample’s normally undetectable high spatial frequency kmodes back into the observable field of view. This is typically done by illuminating a sample with a periodic sinusoidal intensity pattern containing a spatial frequency k′ that can mix with k-modes of the sample. The resulting Moiré interference pattern contains low-frequency fringes at each k − k′ difference frequency, allowing high frequency k-modes to be indirectly observable (Figure 3a).37 Images are compiled as the

Figure 2. (a) IR-PHI image of a 100 nm polystyrene bead that was recorded with a step size of 50 nm. (b) The line profile extracted from panel (a) showing a full width at half-maximum (fwhm) of 300 nm. Adapted with permission from ref 35. Copyright 2017 American Chemical Society. 8725

DOI: 10.1021/acs.analchem.9b01731 Anal. Chem. 2019, 91, 8723−8731

Feature

Analytical Chemistry

Figure 3. Resolution enhancement through fluorescence SIM. (a) Spatial frequencies in the sample mix with patterned illuminating light, resulting in a Moiré interference. (b) In nonlinear SIM techniques, higher harmonic fringes appear in the illumination pattern, increasing the effective NA of the microscope compared to conventional, linear SIM techniques. Adapted with permission from ref 37. Copyright 2005 National Academy of Sciences, U.S.A.

illuminating pattern is rotated or shifted (Figure 1c), and numerical deconvolution methods can be used to extract the higher frequency k-modes and reconstruct the image with enhanced resolution. While SIM increases the detectable spatial frequency by k′, the resolution cannot be infinitely enhanced by illuminating the sample with a spatial pattern that contains as high of a k′ as possible. This is because the k′ frequencies contained in the illuminating light field are still limited by diffraction, like the observed k frequencies diffracting from the sample. Therefore, k′ can at most be equal to the maximum observable frequency using standard microscopy methods. This generally limits the resolution enhancement to a factor of 2 in linear SIM techniques.37 Figure 1c depicts the observable frequency space. Modes that are observable with linear SIM techniques are represented by the medium blue circles around the darker central circle, which represents the k-modes that are observable without SIM. Higher spatial frequencies (represented by the light blue circles) are accessible with nonlinear microscopy techniques.37 While SIM techniques were originally developed with fluorescence microscopy,47−52 their success has inspired its application with other types of microscopy, including Raman. In 2015, Watanabe et al. pioneered label-free SIM using spontaneous Raman spectroscopy. To do this, they used structured line illumination in a slit-scanning Raman microscope.53 Watanabe et al. achieved a 1.4-fold improvement in resolution along the slit illumination direction compared to the theoretical limit of a widefield Raman microscope.53 They demonstrated this improved resolution on fixed mouse brain slices (Figure 4). The resolution in Raman SIM can theoretically be improved further by using higher frequency illumination fringes. However, this is difficult in practice because spontaneous Raman signals are typically very weak, making it difficult to separate high frequency spatial components in Fourier space. SIM techniques that are linear with power, such as those based on fluorescence and spontaneous Raman scattering, are generally limited in achievable spatial resolution improvement.

Figure 4. Raman microscopy images of a mouse brain tissue slice comparing (a) a diffraction-limited image taken with a conventional line illumination (LI) Raman microscope and (b) an image showing improved resolution taken with a structure line illumination (SLI) Raman microscope. Red represents the 1682 cm−1 amide-I protein mode and green represents the 2848 cm−1 lipid CH2 stretching mode. Adapted with permission from ref 53. Copyright 2015 Springer Nature Publishing AG.

However, this can be overcome by exploiting nonlinear or saturated optical responses from the imaging sample.54 Compared to linear SIM, higher frequency k-modes are scattered back into the detection plane through the creation of wavelets containing higher harmonic waves with spatial frequencies that are integer multiples of k′ (Figure 3b). Nonlinear SIM concepts have been applied to fluorescence microscopy,37 and Huttunen et al. have shown these principles could be applied to other nonlinear spectroscopies, such as second (SHG) and third harmonic generation (THG).55 Similarly, utilizing SIM in combination with nonlinear Raman techniques such as SRS or CARS could lead to subdiffractionlimit resolution.



SPATIAL LIGHT MODULATION AND WAVE MIXING Nonlinear optical processes can also be used to enhance resolution by generating a smaller point spread function (PSF) central lobe due to the interaction of multiple light fields. Modifying the intensity or phase profile of a beam can further shrink the PSF of the signal. Illuminating patterns can be formed using a variety of methods. Toraldo di Francia described an early method of shaping the illuminating light using a series of ring-shaped apertures to shrink the central lobe of the PSF.38,56 Because this process also creates intense side lobes, a balance must be found to both minimize the ratio 8726

DOI: 10.1021/acs.analchem.9b01731 Anal. Chem. 2019, 91, 8723−8731

Analytical Chemistry



of the side lobe intensity to the central lobe intensity and minimize the central lobe width (Figure 1d). Raghunathan and Potma theoretically evaluated the use of Toraldo filters and the multiplicative nature of 4WM to achieve super-resolution, specifically for use in CARS.57 They describe two approaches: multiplicative focal volume shaping and subtractive focal volume shaping.57 In multiplicative focal volume shaping, the pump and Stokes beams, with different phase and amplitude profiles, are multiplied together at their coincident foci, producing a narrow centroid with suppressed side lobes. The CARS signal central lobe width can be reduced by 50% while maintaining reasonable side lobe intensities of 5% of the central lobe intensity. While there is no theoretical limitation to making the centroid as narrow as desired, complications arise from the side lobes becoming brighter and larger while the centroid becomes dimmer. In subtractive focal volume shaping, the pump beam is split in two, referred to as p1 and p2. p2 is given a transverse phase profile by a spatial light modulator (SLM), while p1 is unchanged. Before being recombined, p1 and p2 are separated by a phase delay of π/2. Under these circumstances, interferometric methods can be used to extract out the CARS signal from a subdiffraction spot. The width of the central lobe in the subtractive excitation profile is determined by the relative intensity of the two split pump beams and by the phase pattern given to p2. Kim et al. utilized these methods of pulse shaping in nonlinear spectroscopy to attain super-resolution CARS images of 100 nm diameter Si nanowires and 300 nm polystyrene beads (Figure 5).58 The super-resolution image shows that individual beads can be distinguished from the rest of the grouping, unlike in the diffraction-limited image. These results experimentally demonstrate the success of combining pulse shaping with nonlinear wave mixing techniques.

Feature

SIGNAL SUPPRESSION

Another potential approach to resolution improvement is using a spatially shaped beam to selectively suppress signal generation in a defined area. Recent advances have included both theoretical descriptions59−68 and experimental implementations39,69 of approaches to obtain subdiffraction-limit resolution with Raman microscopy. They all propose the use of a doughnut-shaped beam to suppress Raman signal (Figure 1e), which is inspired by the fluorescence-based STED method.12 Suppression via Signal Saturation. Rieger et al. proposed using ground state depletion with a doughnutshaped beam in order to achieve suppression of spontaneous Raman scattering in a spatially defined area.59 In their proof of concept work, they utilized tris(bipyridine)ruthenium(II) (Ru(bpy)32+), which readily transitions from ground to longlived excited electronic states upon illumination, enabling effective ground state population depletion. Using this method, the authors were able to achieve scattering suppression of nearly 50%. Two scans of the region of interest are taken using different pulse combinations, and their difference results in a subdiffraction image (Figure 6). Although imaging results have yet to be published with this technique, simulations illustrating imaging applications are promising.59,60 However, the ground state depletion process depends linearly on laser power and it is challenging to achieve excited state populations above 50%, limiting resolution improvements to at most a factor of 2. It is also worth noting that this technique is not suitable for all systems as the suppression efficiency and thus possible resolution enhancement depends strongly on molecular properties such as excited state lifetime and absorption cross section. Similar techniques were proposed by Gasecka et al.69 and Gong and Wang61 using CARS and SRS, respectively. These techniques avoid the sample constraints that face ground state bleaching methods as they do not rely on depleting the ground state population into a long-lived excited state but rather on the change in the Raman pump intensity. Competition-Induced Suppression. The Cho group accomplished suppression of SRS and CARS signal by promoting an alternative SRS pathway that competes for pump photons.62 They have successfully demonstrated suppression of Raman signal using several related approaches. In general, the goal is to make two Raman modes compete for shared photons that drive the Raman scattering processes, resulting in suppressed signal in the target mode. The scheme that resulted in the highest signal suppression efficiency (∼97%) involved introducing an SRS process to compete with a CARS process that uses the same pump beam (Figure 7).63 They suggest using a doughnut-shaped beam to drive SRS processes that deplete the CARS signal on the edges of the focal spot, resulting in a smaller PSF central lobe. Numerical simulations from the group found that the PSF of this subdiffraction-limit point will be capable of being narrowed by increasing either the intensity ratio between the Stokes and depletion beams or the Raman gain coefficients, although no imaging experiments have been performed to date.64 This version of the technique from the Cho group is highly successful at depleting the observed signal because the CARS signal is proportional to the square of the pump intensity, making it more sensitive to decreased power than if the relationship was linear. This is significant because a nonlinear dependence on depletion power allows for better resolution

Figure 5. CARS images of 300 nm polystyrene beads. (a) Performed with standard PSFs and therefore no resolution enhancement. (b) Performed over the same area using a Stokes beam shaped by an SLM resulting in resolution enhancement. Adapted with permission from ref 58. Copyright 2012 Optical Society of America. 8727

DOI: 10.1021/acs.analchem.9b01731 Anal. Chem. 2019, 91, 8723−8731

Feature

Analytical Chemistry

Figure 6. Simulation from Rieger et al. demonstrating enhanced resolution using ground state bleaching. The top row represents a cross section of the simulated images shown in the bottom row. (a) Scattering sample with a diameter of 50 nm. PSF after scanning the sample in part a with (b) a diffraction-limited Gaussian beam, (c) a doughnut-shaped beam, and (d) both beams. (e) Reconstructed image showing enhanced resolution composed by subtracting part d from part c. Adapted with permission from ref 59. Copyright 2016 Optical Society of America.

copy (FSRS).39 In our experimental setup, we utilize three beams: a broadband probe beam, a pump beam, and a depletion beam. The depletion beam is used to disrupt the vibrational coherence generated by the sample’s interaction with the pump and probe beams. We demonstrated Raman signal suppression of 97% for the cyclohexane ring-breathing mode (Figure 8a). As depletion beam powers were increased,

Figure 7. Figures from Choi et al. demonstrating competing CARS and SRS processes for super-resolution Raman imaging. Jablonski diagrams indicating (a) the conventional CARS process and (b) the competing SRS process. (c) Schematic that shows the CARS signal being depleted due to fewer available pump photons when the SRS process is induced simultaneously. (d) CARS signal of the ring breathing mode of benzene depleted by a competing SRS process. When the depletion beam energy is 250 nJ, the CARS signal is depleted by 97%. Adapted with permission from ref 63. Copyright 2018 Royal Society of Chemistry.

Figure 8. (a) Cyclohexane FSRS signal is depleted with an efficiency of 97% with a Gaussian-shaped depletion beam. The inset shows that depletion of cyclohexane FSRS signal depends nonlinearly on the power of the Gaussian-shaped depletion beam. (b) Resolution improvement demonstrated using a 40 μm thick piece of CVDdiamond. Adapted with permission from ref 39. Copyright 2016 American Chemical Society.

Raman signal decreased nonlinearly (Figure 8a, inset). The nonlinear dependence on depletion power enhances the superresolution potential as it allows for signal suppression even with relatively low depletion beam powers.70 We demonstrated improved resolution with this method by scanning the beams across a 40 μm thick diamond plate on a glass slide using a doughnut-shaped depletion beam (Figure 8b). We also performed two-dimensional scans of a 10 μm thick CVD diamond plate, showing resolution improvement of 40%, well below the diffraction limit of this microscope.39 While this technique is somewhat challenging to implement due to the requirement of overlapping three ultrafast pulses spatially and temporally, it has the potential to achieve

improvements at lower laser powers as compared to linear suppression schemes. Additionally, the requirement of having two highly Raman-active modes in a sample that can compete for pump photons is a common characteristic of many biologically relevant molecules, making this technique very promising for biological applications if the high pulse energies that are required can be mitigated. Suppression of Vibrational Coherence. In 2016, our group published results demonstrating subdiffraction-limit resolution based on femtosecond stimulated Raman spectros8728

DOI: 10.1021/acs.analchem.9b01731 Anal. Chem. 2019, 91, 8723−8731

Feature

Analytical Chemistry Table 1. Summary of Techniques Mentioned in This Feature Developers

Method to Achieve Resolution Enhancement

DiffractionLimited Resolution

Achieved Experimental Resolution

Li et al.35

Measuring change in reflectivity with a visible probe

2060 nm

Hendaoui et al.36 Watanabe et al.53 Raghunathan and Potma57 Kim et al.58 Rieger et al.59

Bleaching IR transitions with evanescent waves

2500 nm

Structured line illumination Raman

Unspecified

SLM and wave mixing

340 nm

SLM and wave mixing Ground state depletion Raman

300 nm 536 nm

Gong and Wang61 Gasecka et al.69

SRS signal saturation with a doughnut-shaped beam

226 nm

Subtracting image taken with doughnut-shaped pump beam from image taken with Gaussian pump beam Competition SRS and CARS processes induced by doughnut-shaped beam

330 nm

180 nm

Vibrational coherence depletion induced by doughnutshaped beam

1370 nm

Not yet implemented in microscope 780 nm

Choi et al.63 Silva et al.39

Diffraction limit of the visible probe beam 100 nm

Unspecified

Factor of 2 improvement Factor of 1.5 improvement (higher if no potential phase interference)

130 nm 175 nm (or better with significantly higher power) 64 nm (or better with significantly higher power)

Limited only by sample photodegradation Limited only by sample photodegradation

frequency region. In the case of biological samples, using small Raman tags that vibrate in the “cell silent region” (∼1800− 2800 cm−1) can allow otherwise undetectable analytes to be tracked. Examples of these include alkynes, nitriles, and B−H and C−D stretches.72,73 While in these cases, the technique is not truly label-free, these vibrational tags cannot bleach and are smaller than fluorescent dye molecules and much smaller than fluorescent proteins, making them less likely to cause unwanted perturbations. Multiplexing is also easier with vibrational tags than with fluorescent tags due to their narrower peak bandwidths. Wei Min’s group has developed sets of numerous Raman tags that can be differentiated from one another despite vibrational signatures in similar spectral regions.74,75 While we have discussed many approaches for improving the lateral spatial resolution in vibrational microscopy, demonstrated improvements in the axial direction are still lacking. To address this, inspiration could again be taken from the fluorescence community. For example, applying the principles of 4Pi microscopy to Raman has been shown to improve axial resolution.76 The use of a double-helix PSF, as well as SIM methods, has been shown to improve fluorescence axial resolution and could potentially be applied to Raman as well.77 In this Feature, we discussed various approaches to improving spatial resolution in far-field IR and Raman microscopies beyond the diffraction limit. We evaluated their potential for further improvements in resolution as well as breadth of applicability and impact. Table 1 compares key characteristics of the techniques discussed here. As many of these methods are in the developmental stage, no far-field vibrational super-resolution technique has experimentally reached the levels of resolution achievable with fluorescencebased super-resolution. However, vibrational techniques that rely on nonlinear signal suppression offer great theoretical potential for resolution on the sub-100 nm length scale. Also, while SIM-based methods are limited in their potential resolution enhancement, they offer a more modest improvement that can be implemented relatively easily. Researchers can maximize the resolution improvements of these techniques by using high NA optics and shorter wavelengths. Finally, it is

resolution on the sub-100 nm length scale due to the nonlinear relationship between depletion power and FSRS signal depletion. Additionally, due to the use of a broadband probe pulse, multiplex detection can be performed without the need for additional optics or pulse adjustments. This technique could be applicable to a wide variety of systems due to the lack of significant sample requirements beyond having an identifiable Raman signature. In the case of STED, spatial resolution has improved to the level of a single macromolecule, making the size of the labels themselves the limiting factor.71 Similarly, highly nonlinear super-resolution Raman techniques, such those mentioned above from the Cho group and our group, can theoretically reach spatial resolution limits down to a single Raman active bond. Of course, this is only true as long as the depletion power can continue to be increased and practical limitations prevent resolution on this scale. Beam powers can only be increased to a point before sample damage occurs. Additionally, the signal-to-noise ratio will decrease as the PSF of the signal shrinks. Therefore, longer acquisition times may be required to achieve very high resolution.



Theoretical Resolution Limit

300 nm

OUTLOOK

The techniques described in this Feature represent a range of approaches for improving spatial resolution in far-field vibrational super-resolution microscopy, with diverse potential sample applications ranging from thin films to biomaterials. However, despite the listed advantages, barriers to commercialization remain for techniques that have been experimentally tested. Many of these techniques involve overlapping multiple laser pulses in space and time, which requires significant expertise. In the case of SRS and CARS based techniques, the lasers capable of producing the short duration but high-power pulses required to generate the signal are also relatively expensive. As discussed above, vibrational spectroscopies have an advantage over fluorescence in that they provide chemically specific information about the analyte without the need for tags. However, the analyte of interest must have a vibrational signature that can be separated from other peaks in the same 8729

DOI: 10.1021/acs.analchem.9b01731 Anal. Chem. 2019, 91, 8723−8731

Feature

Analytical Chemistry

(4) van der Zande, A. M.; Huang, P. Y.; Chenet, D. A.; Berkelbach, T. C.; You, Y.; Lee, G.-H.; Heinz, T. F.; Reichman, D. R.; Muller, D. A.; Hone, J. C. Nat. Mater. 2013, 12, 554−561. (5) Wong, C. Y.; Cotts, B. L.; Wu, H.; Ginsberg, N. S. Nat. Commun. 2015, 6, 5946. (6) Bai, X.; McMullan, G.; Scheres, S. H. W. Trends Biochem. Sci. 2015, 40, 49−57. (7) Cheng, Y.; Glaeser, R. M.; Nogales, E. Cell 2017, 171, 1229− 1231. (8) Abbe, E. K. Arch. für mikroskopische Anat. 1873, 9, 413−418. (9) Schermelleh, L.; Heintzmann, R.; Leonhardt, H. J. Cell Biol. 2010, 190, 165−175. (10) Betzig, E.; Patterson, G. H.; Sougrat, R.; Lindwasser, O. W.; Olenych, S.; Bonifacino, J. S.; Davidson, M. W.; Lippincott-Schwartz, J.; Hess, H. F. Science 2006, 313, 1642−1645. (11) Rust, M. J.; Bates, M.; Zhuang, X. Nat. Methods 2006, 3, 793− 796. (12) Hell, S. W.; Wichmann, J. Opt. Lett. 1994, 19, 780−782. (13) Gevaux, D. Nat. Nanotechnol. 2014, 9, 878. (14) Sahl, S. J.; Hell, S. W.; Jakobs, S. Nat. Rev. Mol. Cell Biol. 2017, 18, 685−701. (15) Liu, Z.; Lavis, L. D.; Betzig, E. Mol. Cell 2015, 58, 644−659. (16) Tønnesen, J.; Katona, G.; Rózsa, B.; Nägerl, U. V. Nat. Neurosci. 2014, 17, 678−685. (17) Xu, K.; Zhong, G.; Zhuang, X. Science 2013, 339, 452−456. (18) Szymborska, A.; de Marco, A.; Daigle, N.; Cordes, V. C.; Briggs, J. A. G.; Ellenberg, J. Science 2013, 341, 655−658. (19) Graydon, O. Nat. Photonics 2016, 10, 504−505. (20) Greenbaum, L.; Rothmann, C.; Lavie, R.; Malik, Z. Biol. Chem. 2000, 381, 1251−1258. (21) Bernas, T.; Zarębski, M.; Cook, R. R.; Dobrucki, J. W. J. Microsc. 2004, 215, 281−296. (22) Hein, B.; Willig, K. I.; Hell, S. W. Proc. Natl. Acad. Sci. U. S. A. 2008, 105, 14271−14276. (23) Veatch, S. L.; Leung, S. S. W.; Hancock, R. E. W.; Thewalt, J. L. J. Phys. Chem. B 2007, 111, 502−504. (24) Skaug, M. J.; Longo, M. L.; Faller, R. J. Phys. Chem. B 2011, 115, 8500−8505. (25) Punihaole, D.; Workman, R. J.; Upadhyay, S.; Van Bruggen, C.; Schmitz, A. J.; Reineke, T. M.; Frontiera, R. R. J. Phys. Chem. B 2018, 122, 9840−9851. (26) Wang, Y.; Purrello, R.; Georgiou, S.; Spiro, T. G. J. Am. Chem. Soc. 1991, 113, 6368−6377. (27) Myshakina, N. S.; Ahmed, Z.; Asher, S. A. J. Phys. Chem. B 2008, 112, 11873−11877. (28) Punihaole, D.; Jakubek, R. S.; Dahlburg, E. M.; Hong, Z.; Myshakina, N. S.; Geib, S.; Asher, S. A. J. Phys. Chem. B 2015, 119, 3931−3939. (29) Miura, T.; Takeuchi, H.; Harada, I. J. Raman Spectrosc. 1989, 20, 667−671. (30) Punihaole, D.; Jakubek, R. S.; Workman, R. J.; Asher, S. A. J. Phys. Chem. Lett. 2018, 9, 1944−1950. (31) Zeisel, D.; Nettesheim, S.; Dutoit, B.; Zenobi, R. Appl. Phys. Lett. 1996, 68, 2491−2492. (32) Keilmann, F.; Hillenbrand, R. Philos. Trans. R. Soc. London. Ser. A Math. Phys. Eng. Sci. 2004, 362, 787−805. (33) Huang, T.-X.; Huang, S.-C.; Li, M.-H.; Zeng, Z.-C.; Wang, X.; Ren, B. Anal. Bioanal. Chem. 2015, 407, 8177−8195. (34) Christopher, P.; Xin, H.; Linic, S. Nat. Chem. 2011, 3, 467− 472. (35) Li, Z.; Aleshire, K.; Kuno, M.; Hartland, G. V. J. Phys. Chem. B 2017, 121, 8838−8846. (36) Hendaoui, N.; Mani, A.; Liu, N.; Tofail, S. M.; Silien, C.; Peremans, A. Opt. Commun. 2017, 382, 574−579. (37) Gustafsson, M. G. L. Proc. Natl. Acad. Sci. U. S. A. 2005, 102, 13081−13086. (38) Toraldo di Francia, G. Atti Fond. Giorgio Ronchi 1952, 7, 366− 372.

important to note that since most of the work represented in Table 1 is proof-of-concept, acquisition times and power values may differ significantly when these techniques are applied to actual systems of interest. For example, Choi et al. point out that a depletion power of 2 TW cm−2 is incompatible with biological material. However, less efficient depletion will still allow them to see improvements in resolution.63 The results and simulations discussed here represent promising steps toward this goal and toward offering alternative microscopy options to fluorescence-based methods.



AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. ORCID

Renee R. Frontiera: 0000-0001-8218-7574 Notes

The authors declare no competing financial interest. Biographies Christian T. Graefe received his B.A. in Chemistry from St. Olaf College in 2013 and is now a chemistry Ph.D. student at the University of Minnesota. His research focuses on the development and optimization of a new super-resolution imaging technique based on stimulated Raman spectroscopy. David Punihaole received his B.S. in Molecular Biology and Ph.D. in Molecular Biophysics and Structural Biology at the University of Pittsburgh. He is currently a postdoc at the University of Minnesota, where he is developing super-resolution stimulated Raman microscopy techniques for biological imaging. Celina M. Harris is a Chemistry Ph.D. candidate at the University of Minnesota. She obtained her B.S. in Chemistry from Gettysburg College in 2017. Her research interests are focused on studying protein−lipid interactions using Raman spectroscopy. Michael J. Lynch graduated summa cum laude with a B.A. in Chemistry from the University of Colorado Boulder in 2017. He is an NSF GRFP fellow working towards a Ph.D. in Chemistry from the University of Minnesota where his work focuses on imaging neurons with high-resolution Raman spectroscopy. Ryan Leighton has a B.S. in Chemistry and a B.A. in Physics from the University of Minnesota Duluth. He is currently a graduate student at the University of Minnesota Twin Cities. Renee R. Frontiera is a McKnight Land-Grant Assistant Professor in the Department of Chemistry at the University of Minnesota. Her research group focuses on advanced applications of Raman microscopy and spectroscopy.



ACKNOWLEDGMENTS Funding for this work was provided by the National Institutes of Health, Grant 5R35-GM119441 (D.P., M.J.L., R.L., R.R.F.) and the National Science Foundation, Grant CHE-1552849 (C.T.G. and C.M.H.). C.T.G. acknowledges the Torske Klubben Fellowship for support, D.P. gratefully acknowledges the Ford Foundation for postdoctoral fellowship funding, and M.J.L. acknowledges the NSF GRFP for support.



REFERENCES

(1) Engelman, D. M. Nature 2005, 438, 578−580. (2) Yin, H.; Flynn, A. D. Annu. Rev. Biomed. Eng. 2016, 18, 51−76. (3) Clark, K. W.; Zhang, X.-G.; Vlassiouk, I. V.; He, G.; Feenstra, R. M.; Li, A.-P. ACS Nano 2013, 7, 7956−7966. 8730

DOI: 10.1021/acs.analchem.9b01731 Anal. Chem. 2019, 91, 8723−8731

Feature

Analytical Chemistry (39) Silva, W. R.; Graefe, C. T.; Frontiera, R. R. ACS Photonics 2016, 3, 79−86. (40) McHale, J. Molecular Spectroscopy, 2nd ed.; CRC Press/Taylor & Francis: Boca Raton, FL, 2017. (41) Prince, R. C.; Frontiera, R. R.; Potma, E. O. Chem. Rev. 2017, 117, 5070−5094. (42) Freudiger, C. W.; Min, W.; Saar, B. G.; Lu, S.; Holtom, G. R.; He, C.; Tsai, J. C.; Kang, J. X.; Xie, X. S. Science 2008, 322, 1857− 1861. (43) Nasse, M. J.; Walsh, M. J.; Mattson, E. C.; Reininger, R.; Kajdacsy-Balla, A.; Macias, V.; Bhargava, R.; Hirschmugl, C. J. Nat. Methods 2011, 8, 413−416. (44) Hanninen, A. M.; Prince, R. C.; Potma, E. O. IEEE J. Sel. Top. Quantum Electron. 2019, 25, 1. (45) Hunt, J. H.; Guyot-Sionnest, P.; Shen, Y. R. Chem. Phys. Lett. 1987, 133, 189−192. (46) Barsi, C.; Fleischer, J. W. Nat. Photonics 2013, 7, 639−643. (47) Bailey, B.; Farkas, D. L.; Taylor, D. L.; Lanni, F. Nature 1993, 366, 44−48. (48) Hell, S.; Stelzer, E. H. K. J. Opt. Soc. Am. A 1992, 9, 2159− 2166. (49) Gustafsson, M. G. L.; Agard, D. A.; Sedat, J. W. J. Microsc. 1999, 195, 10−16. (50) Heintzmann, R.; Cremer, C. G. Proc. SPIE 1998, 3568, 185− 196. (51) Gustafsson, M. G. L. J. Microsc. 2000, 198, 82−87. (52) Gustafsson, M. G. L.; Shao, L.; Carlton, P. M.; Wang, C. J. R.; Golubovskaya, I. N.; Cande, W. Z.; Agard, D. A.; Sedat, J. W. Biophys. J. 2008, 94, 4957−4970. (53) Watanabe, K.; Palonpon, A. F.; Smith, N. I.; Chiu, L.; Kasai, A.; Hashimoto, H.; Kawata, S.; Fujita, K. Nat. Commun. 2015, 6, 10095. (54) Heintzmann, R.; Jovin, T. M.; Cremer, C. J. Opt. Soc. Am. A 2002, 19, 1599−1609. (55) Huttunen, M. J.; Abbas, A.; Upham, J.; Boyd, R. W. J. Opt. 2017, 19, 085504. (56) Ranfagni, A.; Mugnai, D.; Ruggeri, R. J. Appl. Phys. 2004, 95, 2217−2222. (57) Raghunathan, V.; Potma, E. O. J. Opt. Soc. Am. A 2010, 27, 2365−2374. (58) Kim, H.; Bryant, G. W.; Stranick, S. J. Opt. Express 2012, 20, 6042−6051. (59) Rieger, S.; Fischedick, M.; Boller, K.-J.; Fallnich, C. Opt. Express 2016, 24, 20745−20754. (60) Rieger, S.; Würthwein, T.; Sparenberg, K.; Boller, K.-J.; Fallnich, C. J. Chem. Phys. 2018, 148, 204110. (61) Gong, L.; Wang, H. Phys. Rev. A: At., Mol., Opt. Phys. 2014, 90, 013818. (62) Kim, D.; Choi, D. S.; Kwon, J.; Shim, S.-H.; Rhee, H.; Cho, M. J. Phys. Chem. Lett. 2017, 8, 6118−6123. (63) Choi, D. S.; Rao, B. J.; Kim, D.; Shim, S.-H.; Rhee, H.; Cho, M. Phys. Chem. Chem. Phys. 2018, 20, 17156−17170. (64) Cho, M. J. Chem. Phys. 2018, 148, 014201. (65) Beeker, W. P.; Groß, P.; Lee, C. J.; Cleff, C.; Offerhaus, H. L.; Fallnich, C.; Herek, J. L.; Boller, K.-J. Opt. Express 2009, 17, 22632− 22638. (66) Liu, W.; Niu, H. Phys. Rev. A: At., Mol., Opt. Phys. 2011, 83, 023830. (67) Cleff, C.; Groß, P.; Fallnich, C.; Offerhaus, H. L.; Herek, J. L.; Kruse, K.; Beeker, W. P.; Lee, C. J.; Boller, K.-J. Phys. Rev. A: At., Mol., Opt. Phys. 2013, 87, 033830. (68) Wang, D.; Liu, S.; Chen, Y.; Song, J.; Liu, W.; Xiong, M.; Wang, G.; Peng, X.; Qu, J. Opt. Express 2017, 25, 10276−10286. (69) Gasecka, A.; Daradich, A.; Dehez, H.; Piché, M.; Côté, D. Opt. Lett. 2013, 38, 4510−4513. (70) Harke, B.; Keller, J.; Ullal, C. K.; Westphal, V.; Schönle, A.; Hell, S. W. Opt. Express 2008, 16, 4154−4162. (71) Donnert, G.; Keller, J.; Medda, R.; Andrei, M. A.; Rizzoli, S. O.; Lührmann, R.; Jahn, R.; Eggeling, C.; Hell, S. W. Proc. Natl. Acad. Sci. U. S. A. 2006, 103, 11440−11445.

(72) Wei, L.; Hu, F.; Chen, Z.; Shen, Y.; Zhang, L.; Min, W. Acc. Chem. Res. 2016, 49, 1494−1502. (73) Messina, M. S.; Graefe, C. T.; Chong, P.; Ebrahim, O. M.; Pathuri, R. S.; Bernier, N. A.; Mills, H. A.; Rheingold, A. L.; Frontiera, R. R.; Maynard, H. D.; Spokoyny, A. M. Polym. Chem. 2019, 10, 1660. (74) Chen, Z.; Paley, D. W.; Wei, L.; Weisman, A. L.; Friesner, R. A.; Nuckolls, C.; Min, W. J. Am. Chem. Soc. 2014, 136, 8027−8033. (75) Wei, L.; Chen, Z.; Shi, L.; Long, R.; Anzalone, A. V.; Zhang, L.; Hu, F.; Yuste, R.; Cornish, V. W.; Min, W. Nature 2017, 544, 465− 470. (76) Diaz Tormo, A.; Khalenkow, D.; Saurav, K.; Skirtach, A. G.; Thomas, N. Le. Opt. Lett. 2017, 42, 4410−4413. (77) Pavani, S. R. P.; Thompson, M. A.; Biteen, J. S.; Lord, S. J.; Liu, N.; Twieg, R. J.; Piestun, R.; Moerner, W. E. Proc. Natl. Acad. Sci. U. S. A. 2009, 106, 2995−2999.

8731

DOI: 10.1021/acs.analchem.9b01731 Anal. Chem. 2019, 91, 8723−8731