Coherent Raman and Nonlinear Molecular Spectroscopies

Spontaneous Raman scattering hands us one of the sharpest fingerprints in spectroscopy — a vibrational spectrum governed by the same polarizability derivative $\partial\alpha/\partial Q$ , the same vibrational and rotational selection rules, and the same Born–Oppenheimer separation of electronic and nuclear motion that the molecular-structure chapter builds up. It has one fatal flaw: it is feeble. Only about one incident photon in $10^{6}$ to $10^{8}$ is scattered inelastically, radiated incoherently into the full $4\pi$ of solid angle. Coherent Raman is the engineering answer, and it is pure wave physics. Take two synchronized laser fields — a pump at $\omega_p$ and a Stokes at $\omega_s$ — and tune their difference frequency onto a molecular vibration, $\omega_p-\omega_s=\Omega_{\mathrm{vib}}$ . The optical beat note then acts like a single macroscopic force that pushes every illuminated molecule at its vibrational frequency, and pushes them all in step. Instead of a crowd of independently ringing oscillators with random phase, we build one coherent, macroscopic material vibration whose dipoles radiate in phase along a phase-matched direction. Their amplitudes add rather than their intensities, so the signal leaps by orders of magnitude and emerges as a laser-like beam. Everything downstream — the sensitivity gain, the directional signal, label-free images of living tissue, single-shot flame thermometry — flows from that one act of coherence.

Key takeaways

Coherent Raman uses two laser fields whose difference frequency $\omega_p-\omega_s$ is tuned onto a vibration $\Omega_{\mathrm{vib}}$ ; the optical beat drives all molecules in phase, turning weak incoherent scattering into a strong, directional signal.
Every modality reads out one quantity — the third-order susceptibility $\chi^{(3)}$ — whose resonant part is a complex Lorentzian: real part dispersive, imaginary part a symmetric peak.
CARS detects a new blue-shifted colour at $\omega_{as}=2\omega_p-\omega_s$ ; its intensity $\propto|\chi^{(3)}|^2$ mixes the resonant term with a flat non-resonant background, giving an asymmetric, peak-shifted line.
SRS measures energy transfer between the beams (Stokes gain or pump loss); it isolates $\mathrm{Im}\,\chi^{(3)}$ , so its line is undistorted, linear in concentration, and background-free.
Coherence turns the incoherent $I\propto N$ into $I\propto N^2$ along the phase-matched direction $\Delta\mathbf k=2\mathbf k_p-\mathbf k_s-\mathbf k_{as}=0$ — the origin of the huge enhancement.
Key idea: two laser fields whose difference frequency matches a vibration coherently phase-lock every molecule; the resulting macroscopic $\chi^{(3)}$ polarization radiates a directional, intensity-boosted signal scaling as $N^2$ rather than the incoherent $N$ of spontaneous Raman.

The core insight

Two laser fields whose difference frequency matches a vibration coherently phase-lock all molecules; the resulting macroscopic $\chi^{(3)}$ polarization radiates a directional, intensity-boosted signal that scales as $N^2$ instead of the incoherent $N$ of spontaneous Raman.

Where this lives today

The most visible modern home of coherent Raman is the operating room. Stimulated Raman histology, built on SRS microscopy from the Xie group and now a deployed clinical instrument, images fresh unstained surgical tissue in minutes by tuning $\omega_p-\omega_s$ across the protein and lipid CH bands near 2850–2930 cm⁻¹. A picosecond pump and Stokes pair — typically $\sim800$ nm and a synchronized $\sim1040$ nm beam, a beat of $\sim2850\ \mathrm{cm^{-1}}$ for lipids — are raster-scanned through a biopsy, and lock-in detection of the tiny stimulated-Raman-loss builds a virtual haematoxylin-and-eosin image at the bench, letting a surgeon read tumour margins with no slide, stain, or wait. Because SRS reports $\mathrm{Im}\,\chi^{(3)}$ , the contrast is quantitative and free of the CARS background, so lipid-rich white matter and protein-rich tumour separate cleanly. At the other extreme, CARS remains the reference non-intrusive thermometer for combustion. A single nanosecond shot records the whole nitrogen $Q$ -branch envelope near 2330 cm⁻¹; because its shape is set by the Boltzmann populations of rotational and hot-band levels, fitting one spectrum returns gas temperature to a few percent and species mole fractions in the heart of a turbulent flame where no thermocouple survives — the same directional signal that lets the blue anti-Stokes beam be collected against the flame's glare.

The physical picture

Picture every molecule as a tiny oscillator that can be pushed at its vibrational frequency $\Omega_{\mathrm{vib}}$ . In spontaneous Raman each molecule is kicked independently and rings with random phase, so intensities add and the signal is proportional to the molecule number $N$ . Coherent Raman changes the drive. When two fields overlap, their product $E_p E_s^{*}$ contains a term oscillating at the difference frequency $\omega_p-\omega_s$ ; tune that beat onto $\Omega_{\mathrm{vib}}$ and it becomes a metronome that pushes every molecule in the focus with the same phase. The emitters are now phase-locked, their fields add in amplitude, and along the direction where the wavelets march in step the intensity climbs as $N^2$ . That single fact — amplitudes add, not intensities — is the whole engine. The family splits by how you read out the vibration you have driven. CARS makes a new colour: two pump photons plus one Stokes photon emit a photon at $2\omega_p-\omega_s$ , blue-shifted above the pump so it sits clear of the red-shifted fluorescence that swamps ordinary Raman. SRS makes no new colour — it transfers energy between the beams you already have, amplifying the Stokes (stimulated Raman gain) and attenuating the pump (stimulated Raman loss). Because SRS is a small difference measured on strong beams, its signal is directly proportional to concentration and free of the interference distortion built into CARS.

Figure

CARS four-wave mixing: two pump photons at 532 nm and one Stokes photon at 607 nm drive the N₂ vibration at Ω = 2330 cm⁻¹, then the coherence emits a blue-shifted anti-Stokes photon at 473 nm, cleanly separated from red-shifted fluorescence

The whole family lives inside one complex number, the third-order susceptibility $\chi^{(3)}(\omega_{as};\omega_p,\omega_p,-\omega_s)$ . Its imaginary part is a clean symmetric Lorentzian peaked on resonance — what SRS measures. Its modulus squared, contaminated by a flat, real non-resonant background from the near-instantaneous distortion of the electron cloud, is what CARS measures — and that contamination is why CARS lines are asymmetric and their peaks shifted, as the second figure shows.

Figure

What CARS and SRS actually measure near the CH₂ stretch at 2850 cm⁻¹: SRS reads Im χ³, a symmetric Lorentzian centred on resonance, while CARS reads the modulus of χ³ squared, whose interference with the flat non-resonant background pulls the peak down to 2845 cm⁻¹ and digs a dip below background near 2862 cm⁻¹

The core equations

The starting point is the nonlinear polarization. Where an ordinary dielectric responds linearly, an intense field drives higher powers of $E$ ; the third-order term is the lowest that can mix two pump fields and a Stokes field to radiate at the anti-Stokes frequency — four-wave mixing, three photons in and one out, with the tensor $\chi^{(3)}$ carrying all the molecular physics.

P^{(3)}_i(\omega_{as}) = \varepsilon_0\,\chi^{(3)}_{ijkl}(\omega_{as};\omega_p,\omega_p,-\omega_s)\,E_j(\omega_p)E_k(\omega_p)E_l^{*}(\omega_s) \qquad (1)

where: $P^{(3)}$ — third-order polarization density that radiates the anti-Stokes field; $\chi^{(3)}$ — third-order susceptibility tensor; $\omega_p,\omega_s$ — pump and Stokes angular frequencies; $\omega_{as}$ — anti-Stokes frequency; $\varepsilon_0$ — vacuum permittivity. The conjugate on the Stokes field subtracts $\omega_s$ rather than adding it. Photon energy must balance, which fixes the anti-Stokes colour, and the process only becomes resonant when the difference frequency lands on a real vibration. These are the tuning knobs of every CARS experiment.

\omega_{as} = 2\omega_p - \omega_s,\qquad \omega_p-\omega_s=\Omega_{\mathrm{vib}} \qquad (2)

where: $\Omega_{\mathrm{vib}}$ — the vibrational (or rotational) transition being probed. The first relation is energy conservation for the four photons; the second is the resonance you reach by tuning $\omega_s$ . The susceptibility has two pieces. A resonant part, sharply peaked when the beat hits $\Omega_R$ , carries the molecular fingerprint; a non-resonant part, flat in frequency, comes from the electronic response and is always present. The first derivation below shows the resonant term is a complex Lorentzian.

\chi^{(3)}(\Omega) = \chi^{(3)}_{\mathrm{NR}} + \frac{A}{\Omega_R - (\omega_p-\omega_s) - i\Gamma} \qquad (3)

where: $\chi^{(3)}_{\mathrm{NR}}$ — real, frequency-flat non-resonant electronic background; $A$ — Raman coupling amplitude, $A\propto N(\partial\alpha/\partial Q)^2$ ; $\Omega_R$ — vibrational resonance frequency; $\Gamma$ — Raman linewidth (HWHM, set by vibrational dephasing). Note that $A$ inherits the very $\partial\alpha/\partial Q$ that governs spontaneous Raman activity. CARS detects the anti-Stokes intensity, so the signal scales as the modulus squared of $\chi^{(3)}$ , grows quadratically with interaction length when the four waves stay in step, and is penalized by a sinc-squared factor for any wavevector mismatch. Perfect phase matching needs $\Delta\mathbf k=2\mathbf k_p-\mathbf k_s-\mathbf k_{as}=0$ , which dispersion forbids for exactly collinear beams and the BOXCARS crossing geometry restores vectorially.

I_{\mathrm{CARS}} \propto |\chi^{(3)}|^2\, I_p^2\, I_s\, L^2\, \mathrm{sinc}^2\!\left(\tfrac{\Delta k L}{2}\right) \qquad (4)

where: $I_p,I_s$ — pump and Stokes intensities; $L$ — interaction length; $\Delta k=|2\mathbf k_p-\mathbf k_s-\mathbf k_{as}|$ — phase mismatch, each $\mathbf k_j$ of magnitude $n_j\omega_j/c$ . The $I_p^2$ is the two pump photons and $I_s$ the one Stokes photon; the $L^2$ at $\Delta k=0$ is the fingerprint of coherent, in-phase addition. Stimulated Raman instead reads the energy flowing from pump to Stokes as the coherence builds, governed by the imaginary part of $\chi^{(3)}$ alone — why it returns an undistorted Lorentzian, strictly linear in concentration.

\frac{\Delta I_s}{I_s} = \frac{8\pi^2\,\omega_s}{n_p n_s c^2}\,\mathrm{Im}\,\chi^{(3)}(\Omega)\, I_p L \;\equiv\; g_R\, I_p L \qquad (5)

where: $\Delta I_s/I_s$ — fractional Stokes gain; $g_R$ — Raman gain coefficient, $g_R\propto\mathrm{Im}\,\chi^{(3)}$ ; $n_p,n_s$ — indices at pump and Stokes. The gain has no dependence on $\chi_{\mathrm{NR}}$ — the background-free advantage of SRS. Finally the headline scaling, from direct superposition. Summing $N$ phase-locked dipoles along the phase-matched direction gives $E_{\mathrm{tot}}=Na$ and intensity $N^2|a|^2$ , whereas random phases make the cross terms average away, leaving $\langle I\rangle=N|a|^2$ — the correlation the beat field imposes is exactly what spontaneous Raman lacks.

I_{\mathrm{coh}} \propto N^2 \quad\text{versus}\quad I_{\mathrm{spont}} \propto N \qquad (6)

where: $N$ — number of coherently driven molecules in the focal volume. The factor $N$ between the two is the raw coherent-enhancement ceiling, before phase matching and finite coherence trim it.

Derivation AWhere the resonant Lorentzian comes from — the driven molecular vibration

The goal is to earn equation (3) from mechanics rather than assert it: to show that a single Raman-active coordinate, treated as a damped classical oscillator driven by the optical beat force, responds with a complex-Lorentzian amplitude, and that this amplitude — fed back through the polarizability — is what makes $\chi^{(3)}$ resonate. The reduced mass, the derivative $\partial\alpha/\partial Q$ , and the dephasing rate $\Gamma$ each take up their role along the way.

Step 1. Model the normal coordinate $Q$ as a damped oscillator of frequency $\Omega_R$ and reduced mass $m$ , driven by the difference-frequency force: $\ddot Q + 2\Gamma\dot Q + \Omega_R^2 Q = \tfrac{1}{2m}(\partial\alpha/\partial Q)\,E_pE_s^{*}\,e^{-i(\omega_p-\omega_s)t}$ . Why: the force on a Raman mode is the gradient of $-\tfrac12\alpha E^2$ in $Q$ , and only the cross term $E_pE_s^{*}$ oscillates near $\Omega_R$ and drives it resonantly.

Step 2. Seek a steady response at the drive frequency, $Q\propto e^{-i(\omega_p-\omega_s)t}$ , giving $Q(\Omega)\propto E_pE_s^{*}\big/\big[\Omega_R^2-(\omega_p-\omega_s)^2-2i\Gamma(\omega_p-\omega_s)\big]$ . Why: a linear oscillator answers at its drive frequency, so the differential equation becomes algebraic and the denominator carries the resonance.

Step 3. Near resonance factor $\Omega_R^2-\Omega^2\approx 2\Omega_R(\Omega_R-\Omega)$ and keep the resonant pole only. Why: the counter-rotating term is off by $\sim2\Omega_R$ and negligible — the rotating-wave approximation turns the quadratic denominator into the single pole $\Omega_R-(\omega_p-\omega_s)-i\Gamma$ .

Step 4. The polarization at $\omega_{as}$ acts with $\partial\alpha/\partial Q$ once more on this $Q$ , so $\chi^{(3)}_{\mathrm R}\propto N(\partial\alpha/\partial Q)^2\big/\big[\Omega_R-(\omega_p-\omega_s)-i\Gamma\big]$ — a complex Lorentzian atop the flat $\chi_{\mathrm{NR}}$ . Why: the same derivative couples the drive in and the radiation out, so it enters squared, tying coherent-Raman strength to ordinary Raman activity and to number density through $A\propto N(\partial\alpha/\partial Q)^2$ . *The result is the driven-oscillator Lorentzian: HWHM $=\Gamma$ and $\mathrm{Im}\,\chi^{(3)}_{\mathrm R}$ peaks at $\Omega_R$ .*

Go deeper · Derivation B · 1 minThe asymmetric CARS lineshape from non-resonant interference

Derivation BThe asymmetric CARS lineshape from non-resonant interference

The goal is to derive, term by term, why a CARS line is not the symmetric peak you would draw for a resonance but a shifted, dispersive shape riding a background — and why that makes CARS awkward for quantitation while SRS stays clean. We expand $|\chi^{(3)}|^2$ with the resonant Lorentzian on the flat non-resonant background and read off three distinct contributions.

Step 1. Write $\chi^{(3)}=\chi_{\mathrm{NR}}+A/(\delta-i\Gamma)$ with detuning $\delta\equiv\Omega_R-(\omega_p-\omega_s)$ , and split $A/(\delta-i\Gamma)=A\delta/(\delta^2+\Gamma^2)+iA\Gamma/(\delta^2+\Gamma^2)$ . Why: the real part is odd in $\delta$ (dispersive, crossing zero on resonance), the imaginary part even (a symmetric Lorentzian) — and only the imaginary one is what SRS reports.

Step 2. Form the modulus squared: $|\chi^{(3)}|^2=\chi_{\mathrm{NR}}^2+\dfrac{2A\chi_{\mathrm{NR}}\,\delta}{\delta^2+\Gamma^2}+\dfrac{A^2}{\delta^2+\Gamma^2}$ . Why: squaring a sum makes a cross term $2\chi_{\mathrm{NR}}\,\mathrm{Re}\,\chi^{(3)}_{\mathrm R}$ that multiplies the flat background against the odd dispersive part — the origin of all CARS asymmetry.

Step 3. The three pieces are a constant background $\chi_{\mathrm{NR}}^2$ ; an odd interference term that adds on one side and digs a dip below background on the other, pulling the peak off resonance; and a pure Lorentzian $A^2/(\delta^2+\Gamma^2)$ . Why: for the CH₂ example they place the peak near 2845 cm⁻¹ and a dip near 2862 cm⁻¹ though the true resonance is 2850 cm⁻¹ — exactly the second figure — and since the interference term is linear in $A\propto N$ while the pure term is quadratic, a CARS peak height is neither simply $N$ nor $N^2$ , the ambiguity that SRS sidesteps. *As expected, the lineshape collapses to a pure symmetric Lorentzian as $\chi_{\mathrm{NR}}\to0$ .*

Limiting cases

Case	Result	Meaning
Weak-field / spontaneous limit	$I_{\mathrm{sig}}\propto N$ (incoherent)	One field only, or weak unlocked fields: coherence is lost and one recovers spontaneous Raman, linear in molecule number.
Perfect phase matching	$\Delta k L\to0\ \Rightarrow\ I_{\mathrm{CARS}}\propto L^2$	Along $\Delta\mathbf k=0$ the signal builds quadratically with length; dispersion forbids exact collinear matching in bulk, so BOXCARS crossing angles are used.
Resonant background-free limit (SRS)	$\Delta I\propto\mathrm{Im}\,\chi^{(3)}_{\mathrm R}\propto N$	SRS isolates the imaginary susceptibility: a true Lorentzian, linear in concentration, no interference.
Non-resonant limit (CARS off resonance)	$I_{\mathrm{CARS}}\to\|\chi^{(3)}_{\mathrm{NR}}\|^2$	Far from any vibration the signal is the flat electronic background — good for normalization, bad for lineshapes near resonance.

Common trapA CARS peak is just a spontaneous-Raman line made brighter — read its position and height the same way.

It is not, and this is the trap that ruins careless CARS quantitation. Because CARS measures $|\chi^{(3)}|^2=|\chi^{(3)}_{\mathrm R}+\chi^{(3)}_{\mathrm{NR}}|^2$ , the resonant molecular term interferes with the ever-present, real non-resonant background. That does three things a spontaneous-Raman line never does: it makes the line asymmetric, it shifts the apparent peak off the true frequency (to 2845 cm⁻¹ for a resonance actually at 2850 cm⁻¹ in the worked figure), and it can push the signal below background on one wing, a dip that mimics a second feature. Worse, the peak height is not proportional to concentration: it interpolates between linear ( $\propto N$ , the cross term) and quadratic ( $\propto N^2$ ) as concentration rises. Never read a CARS peak position as a vibrational frequency, nor its amplitude as a concentration, without accounting for $\chi_{\mathrm{NR}}$ — or measure SRS, which returns $\mathrm{Im}\,\chi^{(3)}$ and is free of all three.

Worked examples

1 · Colours of a CARS thermometry shot (confidence builder)

A CARS combustion probe uses a doubled Nd:YAG pump at $\lambda_p=532.0$ nm and tunes the Stokes to the nitrogen $Q$ -branch, $\Omega_{\mathrm{vib}}=2330\ \mathrm{cm^{-1}}$ . Find the Stokes and anti-Stokes wavelengths, and confirm the signal is blue-shifted clear of fluorescence.

Step 1. Convert the pump to wavenumbers, where Raman shifts add: $\tilde\nu_p=10^7/\lambda_p[\mathrm{nm}]=10^7/532.0=\mathbf{18\,797\ cm^{-1}}$ .

Step 2. The Stokes beat equals the vibration, so $\tilde\nu_s=18\,797-2330=16\,467\ \mathrm{cm^{-1}}$ , i.e. $\lambda_s=10^7/16\,467=\mathbf{607.3\ nm}$ (deep orange-red) — the beam you sweep to reach resonance.

Step 3. The anti-Stokes carries $\tilde\nu_{as}=2\tilde\nu_p-\tilde\nu_s=\tilde\nu_p+\Omega=21\,127\ \mathrm{cm^{-1}}$ , so $\lambda_{as}=10^7/21\,127=\mathbf{473.3\ nm}$ (blue). Since $473.3<532<607.3$ nm the signal lies to the blue of both drive beams and of red-shifted flame luminosity and fluorescence — why CARS survives in a glowing flame where spontaneous Raman drowns. λ_s = 607.3 nm, λ_as = 473.3 nm (blue-shifted by 2330 cm⁻¹).

2 · How long the coherence lives — dephasing from linewidth

A liquid Raman line has full width at half maximum $\Delta\tilde\nu=5.0\ \mathrm{cm^{-1}}$ . Estimate the vibrational coherence (dephasing) time $T_2$ an impulsive-Raman experiment would see.

Step 1. The driven coordinate rings as $Q(t)\propto e^{-\Gamma t}\cos\Omega_R t$ , whose transform is a Lorentzian of half-width $\Gamma$ ; the $1/e$ amplitude time is $T_2=1/\Gamma$ .

Step 2. Convert the width to a rate: the angular FWHM is $2\Gamma$ and $\omega=2\pi c\tilde\nu$ , so $\Gamma=\pi c\,\Delta\tilde\nu$ . With $c=3.0\times10^{10}\ \mathrm{cm/s}$ , $\Gamma=\pi(3.0\times10^{10})(5.0)=4.7\times10^{11}\ \mathrm{s^{-1}}$ .

Step 3. Invert: $T_2=1/\Gamma=1/(4.7\times10^{11})=\mathbf{2.1\ ps}$ . The phase-locked vibration stays coherent only a couple of picoseconds — why coherent-Raman drive pulses and probe delays live on the sub-picosecond-to-picosecond scale, and why narrower gas lines ( $\sim1\ \mathrm{cm^{-1}}$ ) give longer, $\sim10$ ps coherences. A 0.1 fL focal volume holds $N\approx3\times10^{9}$ molecules, so the ideal $N^2/N=N\approx3\times10^{9}$ enhancement ceiling is enormous, though phase matching and finite coherence hold the delivered gain to $10^4$ – $10^5$ . T_2 ≈ 2.1 ps.

Physics of the portrait

Coherent Raman is a child of the laser: the moment intense coherent light arrived, the third-order response of matter became a whole toolbox, and one physicist did more than anyone to make the field rigorous.

Origins · 1 minNicolaas Bloembergen and how this was discovered

Nicolaas Bloembergen

1920–2017 · nonlinear optics and coherent Raman spectroscopy

Nicolaas Bloembergen was the architect of nonlinear optics as a coherent discipline, and coherent Raman lives entirely inside the framework he built. A Dutch-American physicist at Harvard, he had already reshaped magnetic resonance and invented the three-level maser scheme (1956) before turning, once ruby lasers arrived, to the nonlinear response of matter to intense light. In the early 1960s he and coworkers wrote the systematic theory of the nonlinear susceptibilities $\chi^{(2)}$ , $\chi^{(3)}$ and beyond, and derived the coupled-wave equations and phase-matching conditions that govern how a nonlinear polarization radiates a new field. Every equation here — the four-wave-mixing $P^{(3)}$ , the $\mathrm{sinc}^2(\Delta k L/2)$ phase-matching factor, the resonant $\chi^{(3)}$ that CARS and SRS read out — is a direct application of that machinery. His 1965 monograph Nonlinear Optics became the field's foundation text, and his insistence that the susceptibility is a causal, complex, dispersive object obeying Kramers–Kronig relations is why we can speak of $\mathrm{Re}\,\chi^{(3)}$ and $\mathrm{Im}\,\chi^{(3)}$ as the two things CARS and SRS respectively measure. He shared the 1981 Nobel Prize in Physics with Arthur Schawlow for laser spectroscopy; the coherent-Raman diagnostics now imaging living cells and probing flames are the applied edge of the theory he laid down.

History

1928

— Raman and Krishnan report inelastic light scattering, the effect that names the field

1962

— Woodbury and Ng discover stimulated Raman scattering in a Q-switched ruby laser

1965

— Maker and Terhune at Ford demonstrate the third-order four-wave-mixing process behind CARS

1974

— Begley, Harvey and Byer coin CARS and develop it as a practical spectroscopy

1999

— Zumbusch, Holtom and Xie demonstrate laser-scanning CARS microscopy for chemical imaging

Going further

Texts. Boyd, Nonlinear Optics, 3rd ed., Ch. 10 (stimulated Raman and CARS) · Demtröder, Laser Spectroscopy, 4th ed., Ch. 3 (nonlinear spectroscopy) and Ch. 8 (CARS) · Loudon, The Quantum Theory of Light, 3rd ed., Ch. 8 (nonlinear optics and Raman scattering) · Saleh and Teich, Fundamentals of Photonics, 2nd ed., Ch. 21 · Hecht, Optics, 5th ed., §13.4. The key numerical checks are: $\lambda_p=532.0$ nm with $\Omega=2330\ \mathrm{cm^{-1}}$ gives $\lambda_s=607.3$ nm and $\lambda_{as}=473.3$ nm; $T_2=1/(\pi c\,\Delta\tilde\nu)=2.1$ ps for $\Delta\tilde\nu=5\ \mathrm{cm^{-1}}$ . The CH₂ (2850 cm⁻¹) and N₂ (2330 cm⁻¹) frequencies are standard tabulated values.

Builds on. The Harmonic Oscillator Model of Vibration · Selection Rules for Molecular Transitions · Vibration-Rotation Spectra and P, Q, R Branches · The Born-Oppenheimer Approximation. Leads to. Rotational Spectra and the Rotational Constant · Anharmonicity and the Morse Potential.

Ask the tutor: Ask me to walk through why an SRS line stays a clean Lorentzian while the CARS line beside it bends into an asymmetric, peak-shifted shape.

Why is spontaneous Raman so weak in the first place?

Raman is a two-photon, off-resonant process: the incident photon is not absorbed by a real state but only briefly polarizes the molecule through a virtual level, and the inelastic piece — modulated by the vibration through $\partial\alpha/\partial Q$ — is a tiny fraction of the elastic Rayleigh scattering. Only about one photon in $10^6$ to $10^8$ is scattered with a vibrational shift, incoherently into all directions. Coherent Raman does not change any molecule's cross section; it changes the phase relationship between molecules, so their weak wavelets add in amplitude along one direction instead of averaging away.

What is the difference between CARS and SRS in one sentence?

CARS generates a new beam at $2\omega_p-\omega_s$ and detects its intensity $\propto|\chi^{(3)}|^2$ , so it mixes the molecular resonance with a non-resonant background; SRS keeps the two original beams and detects the tiny energy transfer between them $\propto\mathrm{Im}\,\chi^{(3)}$ alone. So CARS gives a bright, colour-shifted, easily separated signal but a distorted, background-riding lineshape, while SRS gives a background-free, undistorted, concentration-linear signal at the cost of measuring a small change on a large beam — which demands high-frequency modulation and lock-in detection.

Where does the non-resonant background actually come from?

It is the near-instantaneous distortion of the whole electron cloud by the intense fields — a four-wave-mixing response with nothing to do with any vibration and essentially flat across the spectrum. Being electronic rather than nuclear, it responds without the delay and dephasing that give the resonant term its Lorentzian shape, so $\chi_{\mathrm{NR}}$ is real and frequency-independent to good approximation. In CARS it interferes with the resonant term and distorts the line; in time-resolved CARS you suppress it by probing a few hundred femtoseconds after the pump, once the electronic response has died but the vibrational coherence still rings. SRS avoids it structurally, since a real flat $\chi_{\mathrm{NR}}$ contributes nothing to $\mathrm{Im}\,\chi^{(3)}$ .

Why can't the beams simply be collinear for CARS?

Phase matching requires the vector condition $\Delta\mathbf k=2\mathbf k_p-\mathbf k_s-\mathbf k_{as}=0$ , and the magnitudes $|\mathbf k_j|=n_j\omega_j/c$ carry the medium's dispersion. In a dispersive bulk the index differs across the three wavelengths, so collinear beams give a nonzero mismatch and the $\mathrm{sinc}^2(\Delta k L/2)$ factor throttles the signal past a short coherence length. Crossing the beams at a small chosen angle closes the vector triangle exactly (BOXCARS), restoring phase matching and sending the anti-Stokes off to be spatially filtered from the drive beams. In tightly focused microscopy the broad cone of wavevectors relaxes the requirement, which is why collinear CARS microscopy works.

How does a single CARS spectrum measure a flame's temperature?

Temperature sets the Boltzmann populations of the rotational and vibrational levels, and each populated level adds its own resonant term to $\chi^{(3)}$ . As the flame heats, population spreads into higher rotational states and vibrational hot bands, so the shape of the $Q$ -branch envelope changes calculably. Fitting a measured single-shot spectrum against a library of computed spectra returns the temperature, usually to within a few percent, plus species mole fractions. It works because the lineshape, not just the overall intensity, carries the thermodynamics, and because the coherent signal is bright enough to record in one nanosecond shot.

Is the N² scaling something you actually get in the lab?

It is the ceiling, not the delivered number. The $N^2$ law assumes every molecule is driven to the same coherence and radiates perfectly in phase along the matched direction, which is never fully true: only molecules within a coherence length add in phase, the driven amplitude is minute, and shot noise limits detection. So while the mechanism is genuinely quadratic — why a coherent signal can be $10^4$ to $10^5$ times a spontaneous one — the achieved enhancement is far below the naive $N\sim10^9$ ceiling of a focal volume. Read $N^2$ as the reason coherent Raman is strong and directional at all, with phase matching and coherence setting how much you collect.