Raman spectroscopy

Raman spectroscopy identifies and characterizes materials by illuminating them with a narrow-linewidth laser and measuring the small fraction of scattered light whose frequency has changed. The change equals the frequency of a molecular vibration or crystal phonon, so the spectrum is a fingerprint of chemical bonds and structure. Shifts are reported in wavenumbers relative to the laser, typically from about 100 to 3500 cm⁻¹: about 1000 cm⁻¹ for the ring-breathing mode of benzene derivatives, 520.7 cm⁻¹ for crystalline silicon and near 2900 cm⁻¹ for C–H stretches. The underlying process is described under Raman scattering; this entry covers the measurement.

Raman shift and wavelength

The Raman shift is the difference between the excitation and scattered wavenumbers (wavenumber $\tilde\nu = 1/\lambda$):

$$\Delta\tilde\nu = \frac{1}{\lambda_0} - \frac{1}{\lambda_s}.$$

With $\lambda_0$ in nm, the Stokes line lies at

$$\lambda_s = \frac{10^7}{10^7/\lambda_0 - \Delta\tilde\nu}\ \text{nm}.$$

For 785 nm excitation, $10^7/785 = 12\,739$ cm⁻¹; a 1000 cm⁻¹ shift puts the Stokes line at $10^7/11\,739 \approx 851.9$ nm, and the anti-Stokes line at about 727.9 nm. The same 1000 cm⁻¹ band lies at 561.9 nm with 532 nm excitation and at 1190.7 nm with 1064 nm excitation. Because the shift is fixed in wavenumber, a given band moves to a different wavelength whenever the laser changes, which is why spectra are plotted against Raman shift.

The anti-Stokes side is much weaker at room temperature because few molecules start in an excited vibrational state. For a 1000 cm⁻¹ mode at 295 K the Boltzmann factor is about 0.008, so most instruments record only the Stokes side.

Choosing the excitation wavelength

Scattered power scales roughly as the fourth power of frequency (closer to the third power when counting photons), so moving from 785 nm to 532 nm increases the signal by about $(785/532)^4 \approx 4.7$, and going from 1064 nm to 532 nm by a factor of 16. The competing effect is fluorescence: many organic and biological samples, and impurities in them, absorb in the visible and emit a broad background that can be orders of magnitude stronger than the Raman lines and swamps them. Common choices are:

• 532 nm: strong signal, good for inorganic materials, carbon and semiconductors; prone to fluorescence from organics.
• 785 nm: the usual compromise for pharmaceuticals, polymers and biological samples, with much less fluorescence and silicon detectors still usable.
• 1064 nm: lowest fluorescence, but weak signal and Stokes lines beyond the silicon cutoff, so InGaAs arrays or Fourier-transform Raman instruments are used.

A related pitfall: at 785 nm, a 3000 cm⁻¹ C–H band falls at about 1027 nm, where the response of silicon CCDs drops steeply, so the high-shift region is often noisy.

Instrument

A Raman spectrometer has three essential parts. The laser must be single-frequency or narrow, since its linewidth adds directly to the width of every Raman band, and is passed through a clean-up filter to remove spontaneous emission from the laser diode. The elastically scattered (Rayleigh) light, roughly a million times stronger than the Raman signal, is removed with a long-pass edge filter or a notch filter of optical density 6 or more; good filters allow measurement to within about 50–100 cm⁻¹ of the laser line. The remaining light is dispersed by a grating spectrometer onto a cooled CCD (see CCD vs CMOS). Wavenumber resolution follows from the wavelength resolution: 1 nm at 852 nm corresponds to about 14 cm⁻¹, and 0.2 nm to about 2.8 cm⁻¹. Coupling the system to a microscope gives Raman microscopy with micrometer spatial resolution and, with a pinhole, depth sectioning as in confocal microscopy. Choosing the laser, filters and spectrometer for a working instrument, with a worked resolution example and calibration steps, is covered in Raman spectroscopy setup.

Calibration and pitfalls

The wavenumber axis is calibrated with neon or argon lamp lines and checked against a standard: the 520.7 cm⁻¹ phonon line of a silicon wafer is the routine daily check, and polystyrene or cyclohexane give multiple reference bands. The excitation wavelength must be known accurately, because an error of 0.1 nm at 785 nm shifts every band by about 1.6 cm⁻¹. High laser irradiance can heat or burn dark samples and shift bands; cosmic-ray spikes on the CCD must be removed; and fluorescence backgrounds are usually subtracted with a polynomial baseline, which can distort broad bands.

Common questions

Why is the Raman signal so weak?

Raman scattering is a second-order process with a small cross-section; typically only around one in a million or fewer of the scattered photons is Raman shifted. Long integration times, efficient collection optics and cooled detectors compensate.

What is SERS?

Surface-enhanced Raman spectroscopy places molecules on roughened or nanostructured gold or silver, where plasmonic field enhancement raises the signal. Average enhancement factors of about 10⁴–10⁸ are commonly reported, with much higher values claimed at individual hot spots; reproducibility between substrates is the main practical difficulty.

How does Raman compare with infrared spectroscopy?

Both probe vibrations, but with different selection rules: Raman responds to changes in polarizability and infrared absorption to changes in dipole moment, so the two give complementary spectra. Raman handles water-containing samples and glass containers easily; FTIR spectroscopy is usually more sensitive for polar groups.

References: E. Smith, G. Dent, Modern Raman Spectroscopy: A Practical Approach, 2nd ed. (Wiley, 2019); D. A. Long, The Raman Effect (Wiley, 2002); E. C. Le Ru, P. G. Etchegoin, Principles of Surface-Enhanced Raman Spectroscopy (Elsevier, 2009).