Multiphoton Spectroscopy

January 1, 2011


Volume 26, Issue 1

This month's column discusses the various multiphoton spectroscopy techniques and the lasers required for each approach.

There are many types of nonlinear optical techniques, and several depend on the spatial and temporal overlap of multiple optical pulses at different wavelengths on the target. For example, with techniques such as coherent anti-Stokes Raman scattering (CARS) spectroscopy and stimulated Raman scattering (SRS) spectroscopy, which are gaining in popularity, the ability to alter at least one of the wavelengths is essential. Many different systems exist today addressing the many microscopic and spectroscopic techniques being practiced; however, most are dedicated to one single application. Fiber lasers are finding their way into applications that in the past were only addressable by the larger classic laser systems. Multiphoton spectroscopy is an area that only recently began benefiting from the advances made in fiber laser development and the potential value they offer. Development activity in this area is growing at a rapid rate.

Raman Spectroscopy

Scattered photons from molecules are grouped in two forms depending on the type of scattering: elastic scattering and inelastic scattering (1). In elastic scattering (also known as Rayleigh scattering), the scattered photons all have the same energy and frequency. In the latter case, however, we can observe small amounts of scattered photons at a lower or higher frequency than the injected photons after interaction with the samples (molecular vibrations). Most of the incident photons on samples experience Rayleigh scattering in the spontaneous Raman process and only a small number of photons contribute Raman signals (2). Each molecule exhibits a unique spectral response, making Raman scattering a very useful spectroscopic tool (2–4).

Because Raman scattering is a phenomenon occurring between photons and the chemical bonds between atoms or molecules, it does not require various fluorophores for enhancing the imaging contrast, which are usually used to label the molecules for fluorescent imaging. In fluorescent imaging, measurement sensitivity is usually limited by photobleaching and phototoxicity (5,6).

Raman techniques (including CARS and SRS) are being developed for optical biopsy applications. It is indeed possible to distinguish cancer cells from normal cells using these techniques. Incoherent Raman is slow and lacks sensitivity, however, and the other techniques mentioned below are at least 10,000-fold more sensitive, enabling in-vivo high-speed imaging for diagnostics.

Two-Photon Excitation and Second Harmonic Generation Microscopy

In the two-photon excitation (TPE) and second harmonic generation (SHG) processes, two photons are fused or absorbed together to generate blue-shifted signals (6–8). SHG is a coherent process in which energy exits the sample in the form of a single photon. TPE loses energy through absorption into the media even if no linear absorption exists (transparent materials). An energy level diagram of TPE and SHG is shown in Figure 1.

Figure 1: Energy level diagram showing transition states for two-photon excitation (TPE), second harmonic generation (SHG), coherent anti-Stokes Raman scattering (CARS) and stimulated Raman scattering (SRS).

There are several advantages of TPE and SHG microscopy when compared to conventional techniques such as fluorescent confocal microscopy, which is based upon linear and one-photon absorption. In confocal microscopy, fluorophores above and below the focal plane are also excited; photobleaching takes place in whole planes and, therefore, is not detected (8,9). Signals from TPE are collected only at the vicinity of the focal points, which reduces the phototoxic effects and photobleaching. In the case of SHG, excitation of fluorophores is not necessary, which makes it free from any phototoxic effects and photobleaching (10,11). Another advantage is related to increased depth of images. Near-IR lasers deployed in TPE and SHG microscopy applications can propagate photons much deeper inside biomedical tissues, which enables high resolution images in thicker samples (7,8,12,13).

The principles of TPE and SHG microscopy extend to cover multiphoton microscopy and higher harmonic generation microscopy (7,8).

CARS Spectroscopy

CARS is based on four-wave mixing and is a third-order nonlinear phenomenon where typically two lasers are deployed (14–16). If the pump and Stokes fields (denoted as ωp and ωs, respectively) are spatially and temporally overlapped in molecules with vibrational frequency Ω matching ωp – ωs, vibration of the molecules becomes enhanced at ωp – ωs and the pump field becomes scattered to produce the anti-Stokes signal ωas at 2ωp – ωs. CARS signals are quadratically proportional to the pump fields and are linearly dependent on the Stokes field. By virtue of the coherent properties of CARS, the resulting signal is orders of magnitude higher than the signal obtained through the spontaneous Raman process, which makes molecular real time imaging possible. The advantages of CARS are that the signal is at a higher energy than that of the generating laser and, most importantly, the photons are in the visible range and thus can be detected with conventional highly sensitive photodetectors such as photomultiplier tubes. Hyperspectral information can also be obtained by deploying two lasers; one should have a wide wavelength tuning capability while remaining synchronized with the other.

The intrinsic properties of CARS differ from Raman spectroscopy because of the existence of a nonresonant background that is unrelated to any vibrational resonances. Because the nonresonant background provides no useful information from the samples, it must be suppressed as much as possible (17–19). Typically picosecond pulses with narrow spectral width of a few wavenumbers (denoted in cm-1 ) are preferred for CARS spectroscopy because they can simultaneously offer a higher spectral resolution as well as higher contrast between the CARS signals and the nonresonant background (20–22).

SRS Spectroscopy

SRS spectroscopy was realized by adopting another interesting four-wave mixing process, stimulated Raman scattering (23,24). To realize SRS in vibrational molecules, two fields are again used, the pump and the Stokes, which are synchronized and focused on the samples with vibrational frequency Ω matching ωp – ωs. Instead of measuring anti-Stokes signals as in CARS, stimulated Raman loss (SRL) of pump field or stimulated Raman gain (SRG) of Stokes field is used.

Unlike CARS spectroscopy, no background noise is present in SRS because the SRG and SRL happen only when the frequency difference between pump and Stokes is equal to the vibrational frequency. The SRG and SRL are proportional to the product of the pump and Stokes intensities. Measurement of the SRL is achieved by modulating the Stokes field. Likewise, measurement of the SRG is achieved by modulating the pump field. SRS measurements require the use of lock-in amplifiers for extracting the signal.

Pump-Probe Spectroscopy

Reflection, transmission, absorption, and other characteristics of a sample can be examined by use of pump-probe spectroscopy. Pump-probe spectroscopy is a useful tool in research areas such as biology, chemistry, and physics (25). The temporal response of the sample is probed after the pump beam reaches the sample while the delay time of the probe beam is adjusted. The time resolution in pump-probe spectroscopy is dependent upon the pulse width. It is also possible to implement two-color pump-probe spectroscopy by deploying two lasers.

Fluorescence lifetime imaging microscopy (FLIM) is one of the established technologies based on the two-color pump-probe scheme (26,27). FLIM was developed as a technique that provides an enhanced image contrast over fluorescence microscopy by adding the fluorescence lifetime component to the measurement. In conventional fluorescence microscopy, multiple fluorophores tagging different molecules are excited by photons within their absorption bands, which causes them to emit fluorescence in the emission bands. However, with fluorescence microscopy, when the emitted fluorescent spectra are placed close each other, each molecule can no longer be visualized properly. FLIM, on the other hand, can identify fluorophores through fluorophore-dependent lifetime measurement. With FLIM based on pump-probe techniques, the pump beam excites fluorophores and the delayed probes experience stimulated emissions. Instead of measuring the probe beam, the intensity variation of the fluorescence is measured. This technique does not require a high intensity pump to saturate fluorophores and consequently reduces the risk of photobleaching, which can destroy cells or render them inviable (incapable of cellular mitosis).

Terahertz Imaging and Sensing

Radiation in the terahertz range (wavelengths from 30 µm to 3000 µm and frequencies from 0.1 THz to 10 THz) penetrate clothing and many other organic materials and offers spectroscopic information, especially for materials that impact safety such as explosives and pharmacological substances (28). It is also notable that terahertz waves have the characteristics of millimeter waves and IR waves. Various compounds respond differently when exposed to radiation in the 0.1–3 THz range and each thus can be uniquely identified. Terahertz spectroscopy has been actively under research for applications in chemical sensing, noninvasive molecular imaging, and carrier dynamics measurement of semiconductors, dielectrics, and nanomaterials. Terahertz signals also can be used for communication, astronomy, and security.

Regarding terahertz sources, several different techniques have been considered promising in the realization of terahertz waves. When considering direct methods, quantum cascade lasers, gas lasers, Schottky diodes, and Gunn diodes fall into this range. Lasers that emit terahertz waves directly typically require cryogenic cooling (28,29).

Terahertz signals can also be generated indirectly by difference frequency mixing of two lasers at wavelengths that are very close together or by the use of femtosecond lasers. In both schemes, photoconductive antennas consisting of either a semiconductor substrate or a nonlinear optical crystal and onto which a metallic antenna structure is deposited are gated by the applied optical beating signals or the femtosecond laser pulses. Terahertz pulses usually are generated by the use of Ti:sapphire lasers or femtosecond fiber lasers whereas continuous wave terahertz sources are generated by the use of two continuous wave semiconductor lasers or two continuous wave fiber lasers (29).

Optical Coherence Tomography

Optical coherence tomography (OCT) is a technique used to acquire the cross-sectional and noninvasive tomographic image of biological samples with high axial resolution at high speeds (30). The axial resolution is determined by the bandwidth of the optical source. A number of different types of OCT techniques have been developed to acquire different pieces of information from the biological samples, including spectroscopic OCT, phase-sensitive OCT, optical Doppler tomography (ODT), polarization-sensitive OCT, optical coherence microscopy (OCM), optical coherence phase microscopy, and many more (31). Two implementation methods are available for OCT systems: time domain and Fourier domain. Fourier-domain OCT is further categorized into two methods: spectral domain and swept source OCT (SSOCT).

Applications depending upon the acquisition of video-rate OCT images require both a fast A-scan rate and high sensitivity. Fourier-domain OCT, which exhibits 20–30 dB better sensitivity than time-domain OCT, is commonly used for video-rate imaging.

Laser System Requirements: What Exists Today

Each aforementioned application technique requires a specific laser type and configuration. For example, SSOCT requires lasers with wavelength tuning speeds at rates of tens of kilohertz for real-time imaging applications. A longer coherence length (narrow instantaneous linewidth) is also required to achieve deeper imaging for depths of several millimeters or sometimes up to tens of millimeters depending on the sample. Wavelengths also need to be chosen carefully depending on the properties of the samples to be analyzed. The optical bandwidth of the laser is important because the axial resolution is linearly improved by optical bandwidth; ideally, a bandwidth greater than 100 nm is required. There are a few companies today producing lasers that meet the specifications for SSOCT. These lasers usually are composed of a semiconductor gain medium and fast wavelength filters.

However, lasers adapted for SSOCT applications cannot be deployed in applications requiring nonlinear responses from the samples because of the quasi-continuous wave signals that are generated from SSOCT lasers. Conventional lasers in use today for nonlinear spectroscopy applications consist of Ti:sapphire lasers because of the wide optical bandwidth they offer (799–1000 nm), narrow pulse width (down to a few femtoseconds), and high peak power (higher than a few kilowatts). However, Ti:sapphire lasers are notorious for being difficult to operate and require a controlled environment including an optical table. For nonlinear spectroscopy applications requiring two lasers such as in CARS, SRS, and two-color pump probe applications, optical parametric oscillators (OPO) pumped by a Ti:sapphire laser are deployed as the sources.

Synchronization between two lasers is achieved by placing free-space, mechanically activated optical delay lines in either of the two laser output paths. The lasers and setups in use today associated with this approach are large and complex and must be installed in a controlled and stable environment. Also, wavelength tuning of such systems requires either a temperature adjustment or crystal alignment that requires longer than 1 s. Spectroscopy and hyperspectral imaging experiments remain fundamentally limited by the relatively slow wavelength tuning speeds of lasers. Schemes based on supercontinuum generation provide a broadband spectrum that is then mixed with a narrowband pump, and a spectrometer is used to resolve the vibrational spectra (32). Spectral images have been achieved with simple systems but they lack the efficiency to image thick tissue. The development of laser sources with characteristics tailored for the needs of nonlinear imaging and spectroscopy systems remains an active area of research.

For applications in life sciences and more specifically in clinical applications, fiber-based lasers are very appealing because they are durable, robust, compact, and simpler to manage and they facilitate light delivery to the sample. However, techniques based upon coherent Raman spectroscopy require a laser source capable of generating picosecond pulses that have a spectral width well matched to the Raman linewidth of the vibrational molecules. So far, only a few technologies based on fiber lasers have been proposed for nonlinear spectroscopy techniques such as CARS and SRS imaging (33–35).

A system consisting of two integrated fiber lasers has been developed for nonlinear spectroscopic applications such as those discussed (Genia Photonics). The synchronized programmable laser is a small-footprint system that comprises a programmable, dispersion-tuned, actively mode-locked fiber laser and a fiber master oscillator power amplifier, with the picosecond pulse train output of each laser synchronized both spatially and temporally through a single output fiber. The capability to sweep wavelengths in arbitrary manner all while maintaining synchronization of the pulses make it useful for multiphoton spectroscopy and hyperspectral imaging.


The field of multiphoton spectroscopy includes many techniques that apply to a broad range of applications. Laser systems or laser setups today typically are capable of addressing a single application and a major effort is required when reconfiguring them for other applications.

Efforts toward development of systems that can be easily adapted for use in multiple applications are highly desired and one such area is in clinical applications of spectroscopic imagery. Fiber lasers are robust and provide a smaller footprint, transportability, and ease of use. We are seeing a growing trend in the use of fiber lasers for applications that formerly were the domain of larger and more-complex systems.


(1) R.W. Boyd, Nonlinear Optics, third edition (Academic Press, New York, 2008).

(2) K. Kneipp et al., Chem. Rev. 99, 2957–2975 (1999).

(3) J.T. Motz et al., J. Biomed. Opt. 10(3), 031113 (2005).

(4) A.S. Haka et al., J. Biomed. Opt. 14(5), 054021 (2009).

(5) J.W. Lichtman et al., Nat. Methods 2, 910–919 (2005).

(6) W. Denk et al., Science 248, 73–76 (1990).

(7) L. Moreaux et al., Biophys. J. 80, 1568–1574 (2001).

(8) W.R. Zipfel et al., Nat. Biotechnology 21, 1369–1377 (2003).

(9) J.-A. Conchelle et al., Nat. Methods 2, 920–931 (2005).

(10) P.J. Campagnola et al., Nat. Biotechnol. 21, 1356–1360 (2003).

(11) D.A. Dombeck et al., Proc. Natl. Acad. Sci. U.S.A. 100, 7081–7086 (2003).

(12) F. Helmchen et al., Nat. Methods 2, 932–940 (2005).

(13) J. Mertz, Curr. Opin. Neurobiol. 14, 610–616 (2004).

(14) J.X. Cheng et al., J. Phys. Chem B 108, 827–840 (2004).

(15) F. Ganikhanov et al., Opt. Lett. 31, 1292–1294 (2006).

(16) A.F. Pegoraro et al., Opt. Express 17, 2984–2996 (2009).

(17) M. Jurna et al., Opt. Express 16, 15863–15869, 2009.

(18) A. Volkmer et al., Phys. Rev. Lett. 87, 023901 (2001).

(19) E.O. Potma et al., Opt. Lett. 31, 241–243 (2006).

(20) K. Kieu et al., Opt. Lett. 34, 2051–2053 (2009).

(21) E.R. Andresen et al., Opt. Express 15, 4848–4856 (2007).

(22) G. Krauss et al., Opt. Lett. 34, 2847 2849 (2009).

(23) C.W. Freudiger et al., Science 322, 1857–1861 (2008).

(24) P. Nandakumar et al., N. J. Phys. 11, 033026 (2009).

(25) K.L. Hall et al., Opt. Lett. 17, 874–877 (1989).

(26) C.Y. Dong et al., J. Biophys. 69, 2234–2242 (1995).

(27) J.R. Lakowicz, Principles of Fluorescence Spectroscopy, Third Edition (Springer, New York, 2006).

(28) P.H. Siegel, IEEE T. Microw. Theory Tech. 50, 910–928 (2002).

(29) M. Tonouchi, Nat. Photonics 1, 97–105 (2007).

(30) I. Hartl et al., Opt. Lett. 26, 608–610 (2001).

(31) A.F. Fercher et al., Rep. Prog. Phys. 66, 239–303 (2003).

(32) T.W. Kee and M.T. Cicerone, Opt Lett. 29(23), 2701–2701 ((2004).

(33) K. Kieu, B.G. Saar, G.R. Holtom, X.S. Xie, and F.W. Wise, Opt. Lett. 34, 2051– 2053 (2009).

(34) M. Marangoni, A. Gambetta, C. Man zoni, V. Kumar, R. Ramponi, and G. Ce-rullo, Opt. Lett. 34, 3262–3264 (2009).

(35) G. Krauss, T. Hanke, A. Sell, D. Trautlein, A. Leitenstorfer, R. Selm, M. Winterhalder, and A. Zumbusch, Opt. Lett. 34, 2847–2849 (2009).

Youngjae Kim and Joseph Salhany are with Genia Photonics Inc., Lasalle, Québec, Canada. The authors can be contacted at the following e-mail address: