Article pubs.acs.org/JPCB
Chemically Selective Imaging of Overlapping C−H Stretching Vibrations with Time-Resolved Coherent Anti-Stokes Raman Scattering (CARS) Microscopy Anne Kotiaho, Pasi Myllyperkiö, and Mika Pettersson* Department of Chemistry, Nanoscience Center, University of Jyväskylä, P.O. Box 35, FI-40014 Jyväskylä, Finland S Supporting Information *
ABSTRACT: Chemically selective imaging of spectrally overlapping compounds is studied with a time-resolved, femtosecond approach on coherent anti-Stokes Raman scattering (CARS) microscopy taking advantage of timedependent oscillating CARS amplitude which is sensitive to different chemical components at different time points. Chemically selective imaging is demonstrated for composite material of polypropylene (PP) matrix and omPOSS (octamethyl polyhedral oligomeric silsesquioxane) microparticles having partly overlapping CH stretching vibrations. Inverse Fourier transformation (IFT) was applied to Raman spectra of PP and om-POSS, indicating that the oscillatory structures of the vibrational decays differ markedly and allow selective imaging, with minimally using one time point per spatial point, which is also confirmed by the CARS measurements. CARS decays measured additionally for lipid films of cholesterol and DOPC (1,2-dioleyl-sn-glycero-3phosphocholine) indicate selective detection of cholesterol at a specific probe delay time. The results of this study show that the tr-CARS technique has potential for chemically selective, nonresonant background free imaging using overlapping vibrations. wavelengths. Spectral focusing9 based on pulse chirping and controllable time delay has been implemented for more rapid scanning over different vibrational resonances. Multiplex CARS10 combines spectrally broad (fs) pulses with spectrally narrow (ps) pulses, which makes it possible to detect a CARS spectrum in each imaged point, since many vibrations are excited simultaneously. CARS imaging is complicated by the nonresonant response, no matter which of the techniques described above is used. The nonresonant contributions are seen in CARS images as a background signal, which is not chemically selective and prevents quantification of a number of resonant molecular oscillators. The shapes of the CARS spectra are distorted compared to Raman spectra, because of the nonresonant response. Several experimental and mathematical approaches have been developed for removal of the nonresonant background (NRB).11 Another coherent Raman technique, stimulated Raman scattering (SRS) imaging, does not suffer from a similar nonresonant background.12 Time-resolved CARS (tr-CARS), also called Raman free induction decay, is one of the experimental methods suitable for removal of NRB in CARS microscopy.13 The relaxation of the coherently oscillating, vibrationally excited state is measured as a function of the probe delay time (Figure 1C). As shown in Figure 1D, the removal of NRB is based on delaying the probe
1. INTRODUCTION Imaging based on vibrations has two advantages: each molecule has a unique vibrational spectrum and vibrations are inherent to the molecule, leading to chemically selective and label-free imaging. An example of vibrational microscopy is coherent antiStokes Raman scattering (CARS) microscopy.1,2 In comparison with more conventional vibrational microscopies, IR and Raman, CARS imaging offers increased imaging speed and improved spatial resolution and sensitivity, and due to the nonlinear nature of the CARS process, three-dimensional sectioning is possible.3 Biological imaging, especially imaging of lipid structures, is an important example of application of CARS microscopy.3−6 The CARS process is presented in Figure 1. The pump and Stokes pulses are overlapped in time and are selected so that their difference matches the frequency of the Raman active molecular vibration(s) to be studied (see Figure 1A). The coherently oscillating, excited vibrational state is then probed by a third pulse, creating the anti-Stokes signal. In addition to this resonant contribution, the CARS signal comes always with a nonresonant contribution (Figure 1B), which is due to the electronic response of the material and independent of the Raman response. CARS microscopy can be implemented using several techniques. Single frequency CARS7 is based on spectrally narrow (ps) pulses, which excite a single vibrational resonance. Imaging is fast,8 but in the case of complex samples, detecting one vibration is not necessarily enough and measurement of the CARS spectrum requires scanning of one of the excitation © 2014 American Chemical Society
Received: February 19, 2014 Revised: April 1, 2014 Published: April 4, 2014 4363
dx.doi.org/10.1021/jp5017642 | J. Phys. Chem. B 2014, 118, 4363−4369
The Journal of Physical Chemistry B
Article
2. EXPERIMENTAL SECTION The Raman spectra were measured with a home-built Raman spectrometer in a backscattering geometry. The excitation wavelength was 532 nm (Alphalas, Monolas-532-100-SM). The excitation beam was focused on the sample with a microscope objective (Nikon LU Plan ELWD 50×/0.55), and the backscattered light was collected with the same objective. The Rayleigh scattering was attenuated with an edge filter (Semrock). The scattered light was then focused on the entrance of a spectrograph (Acton SpectraPro 2500i) and dispersed with a grating of 600 grooves/mm. An entrance slit of 50 or 100 μm was used. An EMCCD camera (Andor Newton) was used for detecting the Raman spectra. The acquisition time was 2−4 min per spectrum. The fs-CARS setup is based on an amplified integrated femtosecond Ti:sapphire laser (Coherent Libra) with an output energy of ∼1 mJ at 800 nm with a 1 kHz repetition rate. The second harmonic radiation at 400 nm is used to pump three home-built noncollinear parametric amplifiers (NOPA), which generate the required pump (525 nm, fwhm 10 nm), Stokes (620 nm, fwhm 20 nm), and probe (585 nm, fwhm 18 nm) pulses. The difference between pump and Stokes pulse frequencies corresponds to 2920 cm−1. The pulses are compressed by a prism compressor to compensate for the temporal broadening of the pulses particularly in the microscope objective. The probe beam is delayed by a computer controlled delay line (Aerotech). The beams are directed collinearly into a microscope objective (Nikon L Plan SLWD 100×/0.70) using beam splitters. The energy per pulse was set to ∼1 nJ for each of the three beams, corresponding to a total energy of 3 nJ, for which no sample degradation was observed. The CARS signal (500 nm, fwhm 10 nm) is collected in the forward direction with a lens (N.A. 0.4) and filtered with a short-pass dichroic beamsplitter, a bandpass filter (Semrock), and a monochromator (Spex Minimate). The CARS signal is detected using a cooled photomultiplier tube and a boxcar integrator (EG & G Princeton Applied Research). The sample is moved during imaging by piezo positioners (AttoCube systems). A ring light source was installed on the microscope objective, and the backscattered light from the sample was directed trough an ocular onto a web camera in order to visually inspect the sample. A step of 0.5 μm was used in CARS imaging, and in the figures presented, the data has been interpolated for easier visualization. The CARS signal collection time was 0.3 s per pixel. Sheets of polypropylene (PP) and PP/om-POSS (octamethyl polyhedral oligomeric silsesquioxane) composite were obtained from a melt state compounding process described earlier.20 The composite studied contains 3 wt % of om-POSS microparticles in PP matrix. Cholesterol (Sigma) and DOPC (1,2-dioleyl-sn-glycero-3-phosphocholine, Avanti) were used for preparation of lipid films. The films were prepared on microscope glass slides by placing a drop of lipid chloroform solution on the surface and then evaporating the solvent. Lipid mixtures with the following cholesterol content in DOPC were prepared: 17, 29, 38, and 44 mol %.
Figure 1. (A, B) Energy level diagrams for resonant and nonresonant CARS signals, respectively. In the case of the resonant signal, the probe pulse probes coherent molecular vibrations with a specific lifetime, whereas the nonresonant signal comes from virtual states which decay instantaneously. When spectrally broad pulses are used, more than one vibrational state is excited and probed. (C, D) Principle of time-resolved CARS measurement. When the probe is overlapped in time with pump and Stokes pulses, the CARS signal includes a contribution of the nonresonant background. By increasing the probe delay, only a vibrationally coherent signal is detected.
pulse so that the instantaneous NRB has decayed and only the vibrationally resonant signal is detected. The suppression of NRB and strength of the resonant signal can be optimized by suitable modification of the probe pulse shape and duration.14,15 In hybrid or broadband CARS, use of probe delay leads to CARS spectra resembling the Raman spectra.14,16 Usefulness of tr-CARS goes beyond just removing the nonresonant background: all the spectroscopic information contained in the Raman spectrum is coded in the time domain vibrational dynamics. The CARS transients can be converted into spectral information by Fourier transform, as has been demonstrated for halomethanes and their mixtures with a single-pulse CARS microscope using a pulse shaper.17 In order to retrieve full spectral information from the time-domain signal, a sufficiently long time scan with proper sampling must be performed. An example of this strategy is interferometric Fourier transform CARS.18,19 Imaging based on vibrational dephasing times has been demonstrated for two chemical species (toluene and polystyrene) with the same Raman transition frequency but different CARS decay profiles.16 Despite the potential of tr-CARS for chemical imaging, it has been much less studied than frequency domain CARS. In the present study, we demonstrate the capability of tr-CARS to selectively image different species with overlapping spectra in a composite sample.
3. RESULTS AND DISCUSSION The three input laser pulses are spectrally sufficiently broad to coherently excite all the CH stretching vibrations in the sample. At first, this seems to indicate that spectral resolution within the used bandwidth is lost. However, as we will show, the selectivity 4364
dx.doi.org/10.1021/jp5017642 | J. Phys. Chem. B 2014, 118, 4363−4369
The Journal of Physical Chemistry B
Article
Figure 2. (A) Raman spectra of PP sheet and om-POSS powder. The inset shows the convolution of the pump and Stokes pulses used in CARS. (B, C) Comparison between IFT of the Raman spectrum and measured CARS decays for PP and om-POSS, respectively, and (D) measured CARS decays for PP and om-POSS. The nonresonant response was measured from a BBO crystal. Vertical lines are shown at probe delays of 0.10 and 0.48 ps.
symmetric stretching (2869 and 2884 cm −1), and CH3 asymmetric stretching (2954 and 2959 cm−1). The Raman spectrum of om-POSS has three distinct peaks, which can be assigned to symmetric (2912 cm−1) and asymmetric (2972 and 2991 cm−1) CH3 stretchings. The Raman spectra of PP and om-POSS contain information about the vibrations in frequency domain, and the time domain information is extracted from them by taking an inverse Fourier transform (IFT). The IFT amplitude squared, or power, of the frequency domain response simulates the tr-CARS response. The comparison between the IFT obtained from the Raman spectrum and the measured CARS decay is shown in Figure 2B and C for PP and om-POSS, respectively. The time resolution of the CARS experiment is determined from a nonresonant response from a BBO crystal, yielding ∼80 fs (fwhm). In the case of a single Raman transition with a Lorentzian line shape, the CARS decay is exponential with the dephasing time dependent on the line width: decay is faster for broader Raman bands. When multiple transitions are excited, the decay becomes multiexponential and is superimposed with quantum beats, the beat frequency corresponding to a difference in frequencies of the excited modes. The CH stretching band is broad because of overlapping bands, and thus, a fast CARS decay is observed. However, with a high enough time resolution (
Chemically Selective Imaging of Overlapping C−H Stretching Vibrations with Time-Resolved Coherent Anti-Stokes Raman Scattering (CARS) Microscopy Anne Kotiaho, Pasi Myllyperkiö, and Mika Pettersson* Department of Chemistry, Nanoscience Center, University of Jyväskylä, P.O. Box 35, FI-40014 Jyväskylä, Finland S Supporting Information *
ABSTRACT: Chemically selective imaging of spectrally overlapping compounds is studied with a time-resolved, femtosecond approach on coherent anti-Stokes Raman scattering (CARS) microscopy taking advantage of timedependent oscillating CARS amplitude which is sensitive to different chemical components at different time points. Chemically selective imaging is demonstrated for composite material of polypropylene (PP) matrix and omPOSS (octamethyl polyhedral oligomeric silsesquioxane) microparticles having partly overlapping CH stretching vibrations. Inverse Fourier transformation (IFT) was applied to Raman spectra of PP and om-POSS, indicating that the oscillatory structures of the vibrational decays differ markedly and allow selective imaging, with minimally using one time point per spatial point, which is also confirmed by the CARS measurements. CARS decays measured additionally for lipid films of cholesterol and DOPC (1,2-dioleyl-sn-glycero-3phosphocholine) indicate selective detection of cholesterol at a specific probe delay time. The results of this study show that the tr-CARS technique has potential for chemically selective, nonresonant background free imaging using overlapping vibrations. wavelengths. Spectral focusing9 based on pulse chirping and controllable time delay has been implemented for more rapid scanning over different vibrational resonances. Multiplex CARS10 combines spectrally broad (fs) pulses with spectrally narrow (ps) pulses, which makes it possible to detect a CARS spectrum in each imaged point, since many vibrations are excited simultaneously. CARS imaging is complicated by the nonresonant response, no matter which of the techniques described above is used. The nonresonant contributions are seen in CARS images as a background signal, which is not chemically selective and prevents quantification of a number of resonant molecular oscillators. The shapes of the CARS spectra are distorted compared to Raman spectra, because of the nonresonant response. Several experimental and mathematical approaches have been developed for removal of the nonresonant background (NRB).11 Another coherent Raman technique, stimulated Raman scattering (SRS) imaging, does not suffer from a similar nonresonant background.12 Time-resolved CARS (tr-CARS), also called Raman free induction decay, is one of the experimental methods suitable for removal of NRB in CARS microscopy.13 The relaxation of the coherently oscillating, vibrationally excited state is measured as a function of the probe delay time (Figure 1C). As shown in Figure 1D, the removal of NRB is based on delaying the probe
1. INTRODUCTION Imaging based on vibrations has two advantages: each molecule has a unique vibrational spectrum and vibrations are inherent to the molecule, leading to chemically selective and label-free imaging. An example of vibrational microscopy is coherent antiStokes Raman scattering (CARS) microscopy.1,2 In comparison with more conventional vibrational microscopies, IR and Raman, CARS imaging offers increased imaging speed and improved spatial resolution and sensitivity, and due to the nonlinear nature of the CARS process, three-dimensional sectioning is possible.3 Biological imaging, especially imaging of lipid structures, is an important example of application of CARS microscopy.3−6 The CARS process is presented in Figure 1. The pump and Stokes pulses are overlapped in time and are selected so that their difference matches the frequency of the Raman active molecular vibration(s) to be studied (see Figure 1A). The coherently oscillating, excited vibrational state is then probed by a third pulse, creating the anti-Stokes signal. In addition to this resonant contribution, the CARS signal comes always with a nonresonant contribution (Figure 1B), which is due to the electronic response of the material and independent of the Raman response. CARS microscopy can be implemented using several techniques. Single frequency CARS7 is based on spectrally narrow (ps) pulses, which excite a single vibrational resonance. Imaging is fast,8 but in the case of complex samples, detecting one vibration is not necessarily enough and measurement of the CARS spectrum requires scanning of one of the excitation © 2014 American Chemical Society
Received: February 19, 2014 Revised: April 1, 2014 Published: April 4, 2014 4363
dx.doi.org/10.1021/jp5017642 | J. Phys. Chem. B 2014, 118, 4363−4369
The Journal of Physical Chemistry B
Article
2. EXPERIMENTAL SECTION The Raman spectra were measured with a home-built Raman spectrometer in a backscattering geometry. The excitation wavelength was 532 nm (Alphalas, Monolas-532-100-SM). The excitation beam was focused on the sample with a microscope objective (Nikon LU Plan ELWD 50×/0.55), and the backscattered light was collected with the same objective. The Rayleigh scattering was attenuated with an edge filter (Semrock). The scattered light was then focused on the entrance of a spectrograph (Acton SpectraPro 2500i) and dispersed with a grating of 600 grooves/mm. An entrance slit of 50 or 100 μm was used. An EMCCD camera (Andor Newton) was used for detecting the Raman spectra. The acquisition time was 2−4 min per spectrum. The fs-CARS setup is based on an amplified integrated femtosecond Ti:sapphire laser (Coherent Libra) with an output energy of ∼1 mJ at 800 nm with a 1 kHz repetition rate. The second harmonic radiation at 400 nm is used to pump three home-built noncollinear parametric amplifiers (NOPA), which generate the required pump (525 nm, fwhm 10 nm), Stokes (620 nm, fwhm 20 nm), and probe (585 nm, fwhm 18 nm) pulses. The difference between pump and Stokes pulse frequencies corresponds to 2920 cm−1. The pulses are compressed by a prism compressor to compensate for the temporal broadening of the pulses particularly in the microscope objective. The probe beam is delayed by a computer controlled delay line (Aerotech). The beams are directed collinearly into a microscope objective (Nikon L Plan SLWD 100×/0.70) using beam splitters. The energy per pulse was set to ∼1 nJ for each of the three beams, corresponding to a total energy of 3 nJ, for which no sample degradation was observed. The CARS signal (500 nm, fwhm 10 nm) is collected in the forward direction with a lens (N.A. 0.4) and filtered with a short-pass dichroic beamsplitter, a bandpass filter (Semrock), and a monochromator (Spex Minimate). The CARS signal is detected using a cooled photomultiplier tube and a boxcar integrator (EG & G Princeton Applied Research). The sample is moved during imaging by piezo positioners (AttoCube systems). A ring light source was installed on the microscope objective, and the backscattered light from the sample was directed trough an ocular onto a web camera in order to visually inspect the sample. A step of 0.5 μm was used in CARS imaging, and in the figures presented, the data has been interpolated for easier visualization. The CARS signal collection time was 0.3 s per pixel. Sheets of polypropylene (PP) and PP/om-POSS (octamethyl polyhedral oligomeric silsesquioxane) composite were obtained from a melt state compounding process described earlier.20 The composite studied contains 3 wt % of om-POSS microparticles in PP matrix. Cholesterol (Sigma) and DOPC (1,2-dioleyl-sn-glycero-3-phosphocholine, Avanti) were used for preparation of lipid films. The films were prepared on microscope glass slides by placing a drop of lipid chloroform solution on the surface and then evaporating the solvent. Lipid mixtures with the following cholesterol content in DOPC were prepared: 17, 29, 38, and 44 mol %.
Figure 1. (A, B) Energy level diagrams for resonant and nonresonant CARS signals, respectively. In the case of the resonant signal, the probe pulse probes coherent molecular vibrations with a specific lifetime, whereas the nonresonant signal comes from virtual states which decay instantaneously. When spectrally broad pulses are used, more than one vibrational state is excited and probed. (C, D) Principle of time-resolved CARS measurement. When the probe is overlapped in time with pump and Stokes pulses, the CARS signal includes a contribution of the nonresonant background. By increasing the probe delay, only a vibrationally coherent signal is detected.
pulse so that the instantaneous NRB has decayed and only the vibrationally resonant signal is detected. The suppression of NRB and strength of the resonant signal can be optimized by suitable modification of the probe pulse shape and duration.14,15 In hybrid or broadband CARS, use of probe delay leads to CARS spectra resembling the Raman spectra.14,16 Usefulness of tr-CARS goes beyond just removing the nonresonant background: all the spectroscopic information contained in the Raman spectrum is coded in the time domain vibrational dynamics. The CARS transients can be converted into spectral information by Fourier transform, as has been demonstrated for halomethanes and their mixtures with a single-pulse CARS microscope using a pulse shaper.17 In order to retrieve full spectral information from the time-domain signal, a sufficiently long time scan with proper sampling must be performed. An example of this strategy is interferometric Fourier transform CARS.18,19 Imaging based on vibrational dephasing times has been demonstrated for two chemical species (toluene and polystyrene) with the same Raman transition frequency but different CARS decay profiles.16 Despite the potential of tr-CARS for chemical imaging, it has been much less studied than frequency domain CARS. In the present study, we demonstrate the capability of tr-CARS to selectively image different species with overlapping spectra in a composite sample.
3. RESULTS AND DISCUSSION The three input laser pulses are spectrally sufficiently broad to coherently excite all the CH stretching vibrations in the sample. At first, this seems to indicate that spectral resolution within the used bandwidth is lost. However, as we will show, the selectivity 4364
dx.doi.org/10.1021/jp5017642 | J. Phys. Chem. B 2014, 118, 4363−4369
The Journal of Physical Chemistry B
Article
Figure 2. (A) Raman spectra of PP sheet and om-POSS powder. The inset shows the convolution of the pump and Stokes pulses used in CARS. (B, C) Comparison between IFT of the Raman spectrum and measured CARS decays for PP and om-POSS, respectively, and (D) measured CARS decays for PP and om-POSS. The nonresonant response was measured from a BBO crystal. Vertical lines are shown at probe delays of 0.10 and 0.48 ps.
symmetric stretching (2869 and 2884 cm −1), and CH3 asymmetric stretching (2954 and 2959 cm−1). The Raman spectrum of om-POSS has three distinct peaks, which can be assigned to symmetric (2912 cm−1) and asymmetric (2972 and 2991 cm−1) CH3 stretchings. The Raman spectra of PP and om-POSS contain information about the vibrations in frequency domain, and the time domain information is extracted from them by taking an inverse Fourier transform (IFT). The IFT amplitude squared, or power, of the frequency domain response simulates the tr-CARS response. The comparison between the IFT obtained from the Raman spectrum and the measured CARS decay is shown in Figure 2B and C for PP and om-POSS, respectively. The time resolution of the CARS experiment is determined from a nonresonant response from a BBO crystal, yielding ∼80 fs (fwhm). In the case of a single Raman transition with a Lorentzian line shape, the CARS decay is exponential with the dephasing time dependent on the line width: decay is faster for broader Raman bands. When multiple transitions are excited, the decay becomes multiexponential and is superimposed with quantum beats, the beat frequency corresponding to a difference in frequencies of the excited modes. The CH stretching band is broad because of overlapping bands, and thus, a fast CARS decay is observed. However, with a high enough time resolution (
