|Home | About | Journals | Submit | Contact Us | Français|
Label-free microscopy based on Raman scattering has been increasingly used in biomedical research to image samples that cannot be labeled or stained. Stimulated Raman scattering (SRS) microscopy, allows signal amplification of the weak Raman signal for fast imaging speeds without introducing the non-resonant background and coherent image artifacts that are present in coherent anti-Stokes Raman scattering (CARS) microscopy. Here we present the Raman-induced Kerr effect (RIKE) as a contrast for label-free microscopy. RIKE allows us to measure different elements of the non-linear susceptibility tensor, both the real and imaginary parts by optical heterodyne detection (OHD-RIKE). OHD-RIKE microscopy provides information similar to polarization CARS (P-CARS) and interferometric CARS (I-CARS) microscopy, with a simple modification of the two-beam SRS microscopy setup. We show that while OHD-RIKE micro-spectroscopy can be in principle more sensitive than SRS, it does not supersede SRS microscopy of heterogeneous biological samples, such as mouse skin tissue, because it is complicated by variations of linear birefringence across the sample.
Recently, a variety of label-free microscopy techniques have been developed for imaging of samples which cannot be labeled or stained with fluorophores or dyes. Vibrational spectroscopy is especially useful for chemical imaging as every molecule has a specific vibrational fingerprint1, 2. However, infrared absorption (IR) is hindered by low spatial resolution and spontaneous Raman by weak signal levels, limiting imaging speed and sensitivity in microscopy.
In coherent anti-Stokes Raman scattering (CARS) microscopy3, 4 the sample is excited with two laser beams at the pump frequency, ωp, and the Stokes frequency, ωS. If the difference frequency Δω = ωp – ωS is tuned into a vibrational frequency Ω of the sample, strong signal is generated at the at the new anti-Stokes frequency, ωaS = 2ωp – ωS, which is enhanced by orders of magnitude compared to spontaneous Raman scattering5, 6 and allows for fast imaging at speeds up to video-rate7.
CARS microscopy has however been limited by the presence of a non-resonant background, which is generated even without resonant molecules in the focus. It causes spectral distortion of the Raman spectrum8, image artifacts, and limited sensitivity9. Quantification of CARS images further suffers from a non-linear dependence on the concentration of the target molecule10 and coherent image artifacts11.
Stimulated Raman scattering (SRS)6, 12, 13 overcomes all these problems14-17. It is excited simultaneously to CARS and manifests as intensity gain (stimulated Raman gain, SRG) and loss (stimulated Raman loss, SRL) of the transmitted excitation beams. To extract the signal with high sensitivity, we have implemented a high-frequency phase-sensitive detection scheme by modulating the Stokes beam intensity at 10-20MHz and detecting the modulation transfer to the pump beam with a lock-in amplifier. Because laser noise and fluctuations due to varying sample transmission during beam-scanning primarily occur at low frequencies, close to shot-noise limited sensitivity can be readily achieved15. Recently, video-rate in vivo SRS imaging in the epi-direction18 and the combination of the high-frequency detection scheme with multiplex excitation19 has been demonstrated, making SRS microscopy a more valuable contrast for microscopy.
The original implementations of SRS microscopy15, 18 has been focused on mapping the distribution of the spontaneous Raman cross-section σ and polarization sensitive measurements, such as the Raman depolarization ratio ρ, are not made. This has previously been possible with polarization-sensitive CARS (P-CARS) microscopy 20, 21 and can provide additional information about the local molecular orientation and symmetries in the sample, e.g. in myelinated axons22 and cellulose fibers23.
SRS also intrinsically probes the imaginary part of the nonlinear susceptibility tensor χ(3) and does not allow accessing its real part as interferometic CARS (I-CARS) techniques24-27. While Raman-based chemical imaging techniques generally aim to extract the imaginary part24, 25, which carries the specific vibrational information only, imaging the real part of the nonlinear response of the sample might have important applications, such as the label-free detection of neuronal activity28.
Here we present the Raman-induced Kerr effect (RIKE)5, 6, 29, which has been widely used in spectroscopy30-33, as a contrast mechanism for microscopy34. Similar to P-CARS, RIKE allows mapping of different elements of χ(3), however without the disadvantages of CARS. Similar to I-CARS, both real and imaginary part of χ(3) can be probed by optical heterodyne detection (OHD-RIKE), however in a more straight-forward two-beam geometry. OHD-RIKE has also been found to be more sensitive than SRS in spectroscopy 6, because the strength of the local oscillator can be adjusted to reduce laser noise.
All CRS techniques, including CARS, SRS and RIKE, can be understood as four-wave mixing processes, in which emission is generated from a third order polarization due to the non-linear interaction of the excitation fields with the sample (Fig. 1) 6, 35. CARS generates an electric field at the anti-Stokes frequency, ωaS, SRS induces an intensity gain (stimulated Raman gain, SRG) and loss (stimulated Raman gain, SRL) of the excitation beams, and the RIKE changes the polarization of one of the excitation beams due to the Raman-induced birefringence in the sample. In both SRS and RIKE the generated emission is thus at the excitation frequencies. In SRS this radiation is emitted parallel to the excitation field and in RIKE the emission is polarized perpendicular to the excitation fields. In this aspect, RIKE is similar to CARS as the emitted radiation does not interfere with the excitation light, while in SRS interference of the new emission results in intensity gain or loss of the excitation fields.
Generally the polarization of the sample can be written as a power series of the excitation fields. The general expression for the third order contribution at the frequency ω4 is
where χ(3) (ω4; ω1, ω2, ω3) is the non-linear optical susceptibility and E(ω1), E(ω2) and E(ω3) are the electric fields of the excitation beams at ω1, ω2 and ω36. In this notation, CARS excitation is at ω1 = ω3 = ωp and ω2 = −ωS and emission is at ω4 = −ωaS. In SRL and its analog in RIKE, excitation is at ω1 = ωp and ω2 = −ω3 = −ωS and emission is ω4 = −ωp. In SRG and its analog in RIKE, excitation is at ω1 = −ωS and ω2 = −ω3 = ωp and emission is at ω4 = ωS. We note that in the notation ωS, ωp and ωaS are positive numbers and the negative signs account for opposite phase as indicated by the direction of the arrows in the energy diagrams Fig. 1A-C.
All CRS techniques share the common feature, that Δω = ω1 + ω2 = ωp −ωS can be tuned to match a vibrational resonance of the sample with center frequency Ω. All spectroscopic information is contained in χ(3), which, far from electronic resonance, has a nuclear contribution and electronic contribution :
where A is proportional to the particular spontaneous Raman cross-section and Γ is the half-width at half maximum35, 36. Because the electronic response does not probe the vibrational resonance and is instantaneous, is independent of Δω and purely real. The resonant contribution has real and imaginary components, which results in a phase shift of the non-linear polarization with respect to the excitation fields. In particular, as indicated by the positive and negative signs in equation (2), on resonance the non-linear polarization has a relative phase of 270° with respect to the excitation fields for CARS and SRL, and 90° for SRG. The imaginary part has a typical Lorentzian line shape of spontaneous Raman spectra and the real part has a dispersive character35, 36.
This non-linear polarization further has to be treated as a tensor , in which α, β, γ and δ are the indices of the polarization components of the electrical field Eα(ω1), Eβ(ω2), Eγ(ω3) and Eδ(ω4) in the x-, y- and z-direction. Depending on the symmetry of the system, only certain combinations are allowed. Specifically, in isotropic samples , , and and their permutations are the only non-vanishing elements and their resonant contributions can be related to σ and ρ6.
In the original implantation of SRS microscopy, parallel and linearly polarized excitation fields were used, i.e. features of were measured (See Fig. 2). It is straightforward to also measure features of by using linearly-polarized pump and Stokes beams with perpendicular polarization. In the SRL analog of RIKE, the emission is perpendicular to the original polarization of the pump beam, which we detect by blocking the transmitted pump beam with a cross-polarizer and measuring the polarization rotation due to the nonlinear interaction with the sample (see experimental section). Traditionally, two different beam geometries have been routinely used: one employs a linearly-polarized Stokes beam at a 45° angle with respect to pump beam (known as linear RIKE), the other uses a circularly-polarized Stokes beam (known as circular RIKE) (see Fig. 2). The RIKE non-linear polarization at the pump frequency is proportional to for the linear RIKE and in case of the circularly RIKE, where i indicates that the polarization is phase-shifted by 90° with respect to the phase of the pump beam6. Thus, the combination of SRS and RIKE microscopy can probe the distribution of all important tensor elements of the non-linear susceptibility in the sample6.
Theoretically RIKE, in contrast to SRS, is a background-free technique as the emission is polarized perpendicular to the excitation light, which is blocked with a cross-polarizer. In practice, however, and especially in a microscopy configuration with a high numerical aperture (NA) objectives20, it is hard to suppress the excitation light sufficiently. It is thus useful to introduce a carefully controlled portion of the excitation light, e.g. by detuning either the cross-polarizer with respect to the excitation beams or the excitation beams with respect to the cross-polarizer. Due to the coherent nature of RIKE, this portion of the transmitted excitation light interferes with the RIKE emission. Such approach is known as optical heterodyne detected RIKE (OHD-RIKE)6, 29.
Optical heterodyne detection is a standard technique in spectroscopy6, 37. The general idea is that the small homodyne signal E(3) is amplified with a second, coherent electrical field at the same frequency ELO, known as the local oscillator. Because of the coherent addition of the two fields, the overall detected intensity is proportional to
If ELO»E(3), the contribution due to the homodyne signal E(3)2 is negligible compared to the mixing term Re[ELO]·Re[E(3)]+ Im[ELO]· Im[E(3)]. If heterodyne detection is further combined with a modulation transfer scheme as used in SRS microscopy (see experimental section), the contribution of the local oscillator ELO.2 is suppressed and only the mixing terms are detected. Depending on the phase of ELO, the real and imaginary part of E(3) can be probed specifically by seting either Re[ELO] or Im[ELO] to zero.
In CRS, the small homodyne signal E(3) is the radiation generated from the non-linear polarization P(3)(ω4). It turns out that E(3) is 90° phase-shifted with respect to P(3)(ω4) (see equation 4.75 in reference 37) and thus
with E(ω1), E(ω2) and E(ω3) being real by definition. As such the signal in optically heterodyne detected CRS is proportional to
In CARS microscopy, optical heterodyne detection has been performed previously by providing the local oscillator in the form of a third beam at ωaS in the excitation light to measure real and imaginary part of the non-linear susceptibility and suppress the non-resonant background21, 24-26. This is challenging experimentally because it requires generation of a third wavelength and precise phase control of three laser beams in a beam-scanning microscope. In SRS and OHD-RIKE, optical heterodyne detection can be performed without the need for a third beam, as the coherent emission from the sample occurs at the excitation frequencies.
SRS is actually a very special case of intrinsic optical heterodyne detection as both frequency and polarization of the new emission generated from the non-linear polarization are identical to that of the transmitted excitation beam. For this reason, the local oscillator is strong and Im[ELO] = 0 by definition. Thus SRS intrinsically probes Im[χ(3)]. This explains why SRS microscopy does not suffer from a non-resonant background as Im due to the instantaneous electronic response. Thus even though , it is not detected in SRS microscopy, since the non-resonant background does not interfere with the excitation fields and thus is not amplified by optical heterodyne detection. It further explains why SRS spectra are identical to those of spontaneous Raman scattering, which also probes Im.6, 14, 15.
In RIKE the emitted homodyne field from the sample has the same frequency as the excitation fields but occurs at the perpendicular polarization and could thus be detected in a background-free manner. In OHD-RIKE some of the excitation beam is rotated to match the emission polarization, allowing us to provide a local oscillator with adjustable strength and phase. In contrast to SRS, in OHD-RIKE the local oscillator strength can thus be optimized and both the real and imaginary parts of the non-linear susceptibility tensor can be probed selectively
The instrumentation required for OHD-RIKE microscopy is based on a typical SRS microscope with high-frequency phase-sensitive detection15. Here we describe how such SRS microscope can be modified to perform OHD-RIKE.
The experimental apparatus is shown in Fig. 3. Synchronized, mode-locked laser pulse-trains are provided by an optical parametric oscillator (OPO) (Levante Emerald, APE-Berlin) synchronously pumped by a frequency doubled Nd:YVO4 laser (picoTRAIN, High-Q, 532nm, 7 ps, 76 MHz repetition rate). Additionally, this laser provides a separate output of the fundamental beam at 1064nm. The 1064nm beam is used as the Stokes beam in combination with the signal beam from the OPO as the pump beam for CRS. The beams are spatially overlapped with a dichroic beam-combiner (Chroma Technology, 1064 DCRB). The Stokes beam is modulated with a home-built Pockel Cell (crystal from Raicol Ltd.) and resonantly driven at 10 MHz by the internal clock of a lock-in amplifier (Stanford Research Systems, SR844RF). This home-built Pockel cell has the advantage that the RF drive signal is amplified within the same electrically-shielded box that contains the modulator crystal to minimize radio-frequency emission that can be detected by the lock-in amplifier. A polarization-analyzer is used to transform the polarization modulation of the Pockel cell into amplitude-modulation.
Pump and Stokes beams are coupled into a modified upright laser scanning microscope (BX61WI/FV300, Olympus) optimized for IR transmission. A 60× 1.2 NA water objective (UPlanApo / IR, Olympus) is used as excitation objective and light is collected in transmission with a 1.4 NA oil condenser (Nikon). After blocking the Stokes beam with a bandpass filter (Chroma, CARS 890/220) the pump beam is detected with a large-area photodiode (FDS1010, Thorlabs) which is reverse-biased to 64V. The photodiode output is bandpass-filtered (Mini-Circuits, BBP-10.7) to suppress the strong high-frequency signal due to the laser pulsing (76MHz) and the low-frequency signal from scanning the beams through the sample with varying transmission. The same high-frequency lock-in amplifier that drives the Pockel cell is used to demodulate the pump intensity.
For imaging, the output of the lock-in amplifier is fed into the A/D-converter of the microscope to provide the pixel intensity. For the micro-spectroscopy, an RS232 computer-controlled interface of the OPO was developed in collaboration with APE GmbH. In brief, the OPO wavelength is tuned with the in cavity Lyot-filter within the phase-matching bandwidth of LBO crystal at a given temperature, allowing up to 200cm−1 tuning range. The microscope is set to illuminate a fixed position in the sample, the OPO is tuned, and the intensity from the lock-in amplifier for each Δω is recorded.
We implement the SRL analog of OHD-RIKE by detecting the polarization rotation of the pump beam due to the non-linear interaction with the Stokes beam in the focus. First, two polarizers (Thorlabs, GTH5M) are used to linearly polarize the pump and Stokes beam. A crossed polarizer (Thorlabs, LPVIS100) is placed in front of the photodiode and positioned to minimize transmission. Due to the inherent loss of depolarization of tightly focused beams, the best extinction ratio that could be achieved is 1:300.
A half-waveplate (Thorlabs, WPH05M-1064) or a quarter-waveplate (Thorlabs, WPQ05M-1064) is placed in the Stokes beam to produce linearly polarized Stokes light at a polarization 45° with respect to the pump beam or circularly polarized Stokes light for linear or circular OHD-RIKE, respectively. Special care was taken to ensure the exact positions of the waveplates. Often the position of the waveplate was optimized using the spectral property of the OHD-RIKE signal, e.g. in circular OHD-RIKE, the non-resonant signal due to cross-phase modulation was zeroed.
To control the amplitude of the local oscillator for heterodyne detection, a half-waveplate (Thorlabs, AHWP05M-980) slightly rotated the polarization of the linear pump beam such that a portion of the light passes through the polarizer in front of the detector. The strength of the local oscillator is measured by recording the DC current from the photodiode.
The phase of the local oscillator can be continuously controlled using a Babinet-Soleil compensator (OFR, SB-10) aligned along the polarization of the local oscillator. Thus by tuning the phase delay it is possible to specifically access the real and imaginary components of the emitted radiation. In linear OHD-RIKE, the non-linear polarization is in phase with the excitation pump field (see Fig. 2) and the imaginary part of can be probed with no phase delay in the local oscillator. In circular OHD-RIKE, the non-linear polarization is 90° phase-shifted with respect to the excitation pump field (see Fig. 2). Thus to probe the imaginary part of the local oscillator has to also be phase-shifted by 90°, i.e. a quarter wave. To probe the real part of and the local oscillator has to be shifted by 90° or 0° in linear and circular RIKE, respectively. Thus in these special cases, the Babinet-Soleil compensator is not necessary, as either no phase-delay is needed at all (0° phase-shift) or a simple achromatic quarter wave plate (Thorlabs, AQWP05M-980) is sufficient (90° phase-shift).
One of the differences of OHD-RIKE compared to SRS is that the strength of the local oscillator can be controlled. Figure 4A shows the circular OHD-RIKE signal of methanol at 2830cm−1 as a function of the intensity of the local oscillator strength. The microscope was configured to probe Im i.e. the phase of the local oscillator was set such that Im[ELO]= 0.
With α = ILO/Ipump, where ILO is the measured intensity of the local oscillator transmitted through the cross-polarizer and Ipump is the total intensity of the pump beam, and , because with increasing intensity of the local oscillator the intensity of the other polarization component of the pump filed is depleted by the same amount. According to (5) the intensity of OHD-RIKE is thus proportional to Im.
With increasing local oscillator strength the measured signal is however not only originating from OHD-RIKE process, but also from the SRS signal of the local oscillator itself, which is proportional to the intensity of the local oscillator α. Since the Stokes beam is circularly polarized, the SRS signal of the local oscillator is further proportional to Im. Thus theoretically the intensity dependence of the overall measured signal is
with proportionality constants SSRS and SRIKE, which can be related to and Im, respectively.
The theoretical prediction describes the experimental results well and fitting the data for the 2830cm−1 vibration of methanol shows that SSRS / SRIKE = 1.9. If the local oscillator is 1% of the pump beam intensity, the signal contribution from RIKE is thus about 5× stronger than the signal from SRS.
By controlling the phase of the local oscillator with the Babinet-Soleil compensator it is possible to access both real and imaginary parts of the non-linear susceptibility. Figure 4B shows both Re and Im of the CH-stretching vibrational spectrum for dodecane measured with circular RIKE with the phase set to 0° and 90°, respectively, and 1% of the pump intensity in the local oscillator. The lock-in amplifier was set such that a loss of the pump intensity appears as positive signal (i.e. SRL causes positive signal). As such the positive and negative signals in Re corresponds to loss or gain of the local oscillator intensity.
In general, chemical imaging aims to detect the resonant response of the sample, i.e. the imaginary part of the non-linear response, which carries the chemically-specific information. Whereas in CARS microscopy, this weak resonant response can be overwhelmed by the non-resonant background, SRS has improved the detection limitation of label-free microscopy by overcoming this notorious background signal. However, another, much weaker, spurious background signal from cross-phase modulation (XPM) can be the limiting factor when pushing the sensitivity limit to low chemical concentrations15, 38.
XPM is detected in SRS microscopy because the modulated Stokes beam intensity causes a modulation of the refractive index in focus due to the Kerr effect. This results in a modulation of the divergence of the transmitted pump beam. If the pump beam, after passing through the focus, strikes any type of aperture (e.g. the edges of the collection optics), this divergence modulation is transformed into an amplitude modulation which is detected by the lock-in amplifier. As such, the modulated change of focusing properties can cause a spurious background signal in SRS microscopy. Using a condenser with numerical aperture (NA=1.4) higher than that of the excitation objective (N.A.=1.2) and a large area photodiode in order to collect all the light from the sample minimize this effect15. However, in strongly scattering samples, this is not always possible and XPM cannot always be fully suppressed.
In OHD-RIKE microscopy set to probe imaginary part of χ(3), the same process can also cause spurious background signal proportional to Re. However, off vibrational resonance, according to Kleinmann symmetry 6. Thus circular OHD-RIKE should be free of spurious background signal from XPM as Re.
In order to check this hypothesis we measure the XPM signal from water (off-resonance of the OH-stretching vibration at 2830cm−1) as function of the local oscillator strength (Fig. 4C). We measure XPM by placing a pinhole in front of the detector to effectively reduce the collection NA. Fitting equation (6) to this data confirms that and all the detected signal originates from the XPM of the local oscillator which probes Re. Thus it is shown here that the off vibrational resonance Kleinmann symmetry obeys for the used mid-IR lasers and that circular OHD-RIKE can indeed overcome XPM.
Fig. 4D shows the ratio of the detected OHD-RIKE signal of methanol over the XPM background signal of water with a closed pinhole in front of the detector as a function of local field strength. At small local fields the majority of the detected signal originates from RIKE and hence the signal-to-background is much better than for SRS.
The inset shows the resulting signal-to-noise ratio of the resonant methanol signal over the noise of the XPM background signal from water with closed pinhole as a function of local field strength. The signal- to-noise at maximal local field corresponds to the one of SRS with circular Stokes beam, which is less than for SRS with minimal XPM with a high-NA objective15. In the regime of large local fields, the noise is mainly determined by the intensity fluctuations of the non-resonant XPM signal due to laser intensity fluctuations. Therefore it follows the trend of the signal-to-background ratio. In the regime of small local fields, it is determined by electrical noise which is independent of signal and the signal-to-noise ratio decreases with decreasing signal. At the maximum the signal-to-noise is improved by about 3× compared to circular SRS. The sensitivity at low local fields could be further improved by using more advanced detection electronics (e.g., resonant tank circuit) or by using an avalanche photo-diode, however, as discussed in the following section, this regime is typically not reached in microscopy.
Fig. 5A shows a circular OHD-RIKE image of a sebaceous gland surrounding a hair follicle in mouse skin at the CH2-stretching vibration of the lipids at 2845cm−1. By tuning off resonance (Fig. 5B) the image contrast vanishes, however a few features do not disappear. They can be explained because of the linear birefringence of the sample which creates a local oscillator of uncontrolled phase. The images of the intensity of the local oscillator (Fig. 5E and F) show that even though local oscillator was set to 1% of the pump intensity in the local oscillator prior to the imaging, locally up to 5% of the pump beam can be transmitted through the cross-polarizer due to spatially varying birefringence of the sample.
Figs. 5C and D show the comparison with SRS imaging of the same region. While XPM introduces a spurious background signal, these effects are weaker than the background effects due to linear birefringence in RIKE and much weaker than the non-resonant background in CARS15.
In this work we have demonstrated the Raman-induced Kerr effect (RIKE) as a contrast mechanism for label-free microscopy and shown how an existing SRS microscope can be readily modified to perform RIKE using only a few polarization optics. In this implementation, we measure the polarization rotation of the pump beam as a result of the non-linear interaction with the Stokes beam, which can be either circularly or linearly at 45° polarized. Optical heterodyne detection is required in the microscopy implementation of RIKE to compensate for the inherent polarization loss in a tightly focused beam. In optimizing the strength of the local oscillator we find that it has to be <1% of the intensity of the pump beam in order to detect an OHD-RIKE signal that is not overwhelmed by the SRS of the local oscillator.
Under these conditions we find that OHD-RIKE can successfully measure different tensor elements of the non-linear susceptibility than SRS and can determine resonant (imaginary part) and dispersive (real part) components. As such, OHD-RIKE combines the advantages of P-CARS and I-CARS microscopy, however it utilizes a more straightforward geometry that does not require a third color beam as local oscillator.
The motivation for us to explore OHD-RIKE was the goal to improve sensitivity compared to SRS, which was suggested by previous spectroscopy studies6. We show that circular RIKE is indeed free from spurious background signal due to XPM and, in principle, offers higher sensitivity than SRS, as the measurement noise scales with the spurious background signal introduced by XPM. However OHD-RIKE microscopy suffers from similar limitations as P-CARS and I-CARS microscopy for the imaging of complex biological samples, i.e. the inherent depolarization and phase-error associated in heterogeneous and birefringent samples. As such, the potential sensitivity advantages of RIKE compared to SRS are difficult to realize in practice in microscopy.
These findings further highlight why SRS15 is a unique contrast for microscopy of biological samples. This is because SRS is a heterodyne detection scheme with a local oscillator that has same optical frequency and polarization of the induced polarization, which avoids the complications.
C.W.F acknowledges a PhD fellowship from Boehringer Ingelheim Funds. M.B.J.R. thanks the FWO (Fondsvoor Wetenschappelijk Onderzoek) for a postdoctoral fellowship and the support from the Belgian American Educational Foundation and the Fulbright Commission - Belgium. This work was supported by the Gates Foundation and the NIH T-R01 award to X.S.X. We thank Shaul Mukamel for his many important contributions to the understanding of nonlinear optical spectroscopy.
†Current address: Department of Chemistry, Katholieke Universiteit Leuven, Belgium