Raman Microscopy and Mapping

Introduction

A bulk Raman spectrum tells you what a sample contains on average. Many real questions, however, are about where things are: is the drug evenly spread through a tablet, which grains in a rock are calcite and which are quartz, is a graphene sheet single-layer everywhere or only in patches? Raman microscopy answers these questions by coupling a Raman spectrometer to an optical microscope, so that each spectrum comes from a spot around a micrometre across. Scanning that spot builds a chemical image in which every pixel carries a full vibrational fingerprint.

Core explanation

The instrument. A laser (commonly 532, 633 or 785 nm) is focused through a microscope objective onto the sample. The same objective collects the back-scattered light. A notch or edge filter removes the intense Rayleigh line, and the remaining Stokes Raman light is dispersed by a grating onto a cooled CCD detector. Because Raman scattering is so weak (roughly one photon in 10⁶–10⁸ is inelastically scattered), a high-NA objective that gathers light from a wide cone is essential.

Lateral resolution. The smallest spot a lens can form is limited by diffraction. A common estimate of lateral resolution is

d ≈ 0.61λ ÷ NA

With λ = 532 nm and NA = 0.9, d ≈ 360 nm; with a 785 nm laser the same objective gives about 530 nm. Shorter wavelengths therefore give sharper images, and also stronger scattering because intensity scales approximately with ν⁴.

Confocality and depth. A pinhole placed in an image plane of the focal spot passes light from the focus but blocks most light from above and below it. This gives optical sectioning: depth resolution is typically 1–2 µm in transparent samples, so layers in a polymer film or inclusions inside a crystal can be probed without cutting the sample. Refraction at the surface stretches the probed depth, so apparent depths in a medium of higher refractive index must be corrected.

Building a map. The stage (or the laser spot) is moved in a raster pattern, and a spectrum is recorded at every point. The result is a hyperspectral data cube: two spatial axes and one Raman-shift axis. A 100 × 100 map therefore contains 10 000 spectra. Fast "line-focus" and electron-multiplying CCD methods reduce acquisition time to milliseconds per pixel for strong scatterers.

Turning spectra into images. The simplest image plots the intensity of one characteristic band, for example the 1086 cm⁻¹ carbonate stretch. Richer images plot band position (sensitive to strain or composition), band width (sensitive to crystallinity) or intensity ratios. When components overlap, multivariate methods such as principal component analysis, cluster analysis or classical least-squares fitting against reference spectra assign each pixel a composition.

Practical limits. Tight focusing concentrates milliwatts into a sub-micrometre spot, giving very high power densities. Dark or absorbing samples can heat, change phase or burn, so laser power must be reduced and checked by repeating a spectrum at the same point.

Formulae

Lateral resolution: d ≈ 0.61λ ÷ NA. Numerical aperture: NA = n sin θ, where n is the refractive index of the medium between lens and sample and θ is the half-angle of the collection cone. Map acquisition time ≈ number of pixels × time per pixel.

Step-by-step reasoning

To plan a Raman map of a heterogeneous sample:

1. Record single spectra of each expected component and pick a band unique to each. 2. Choose the laser wavelength: shorter for resolution and signal, longer to reduce fluorescence. 3. Choose the objective NA and step size; steps finer than about half the resolution add time but no real detail. 4. Test the laser power on one spot for damage, then set the exposure time. 5. Acquire the map, then remove cosmic-ray spikes and subtract the background. 6. Build univariate or multivariate images and check them against the raw spectra of representative pixels.

Visual explanation

Imagine a stack of thin coloured cards, each card an image of the sample at one Raman shift. Pulling out the card at 1086 cm⁻¹ lights up every calcite grain; the card at 465 cm⁻¹ lights up quartz. Overlaying the two in red and blue produces a false-colour chemical map of the rock.

Real-world analogy

A Raman map is like a survey of a city in which an inspector visits every house on a grid and records a full list of its contents. Afterwards you can colour the map by any item on the list — every house with a piano, every house with a bicycle — without going back to collect more data.

Real-world example

Pharmaceutical companies map tablets to confirm that the active ingredient is evenly dispersed and that it is in the intended polymorph. Two polymorphs of the same drug have identical chemical formulae but different lattice vibrations below about 200 cm⁻¹, so a map can reveal small regions where an unwanted crystal form has appeared during processing.

Why?

Why does Raman microscopy give chemical contrast where ordinary light microscopy often cannot? A white powder of calcite and a white powder of a drug may look identical by eye, because both scatter visible light elastically. Their vibrational spectra, however, depend on bond strengths and atomic masses, so the inelastically scattered light is different for each and distinguishes them pixel by pixel.

Common misconception

"A confocal pinhole lets Raman microscopy image features far below the diffraction limit." Confocal detection improves depth sectioning and slightly sharpens lateral resolution, but the spot size is still set by λ and NA. Features much smaller than about 300 nm are averaged unless near-field methods such as tip-enhanced Raman spectroscopy are used.

Worked example

Question: A 50 µm × 50 µm area is mapped with a 532 nm laser and a 0.90 NA objective, using a 0.5 µm step and 0.2 s per spectrum. Estimate the resolution, the number of spectra and the acquisition time.

Reasoning: d ≈ 0.61 × 532 nm ÷ 0.90 ≈ 360 nm, so a 0.5 µm step is reasonable. Pixels along each side = 50 ÷ 0.5 = 100, giving 100 × 100 = 10 000 spectra. Time ≈ 10 000 × 0.2 s = 2000 s, about 33 minutes (plus stage overheads).

Answer: About 0.36 µm lateral resolution; 10 000 spectra; roughly 35 minutes.

Quick check

1. Why does switching from a 532 nm to a 785 nm laser make a Raman map blurrier with the same objective? Answer: Diffraction-limited resolution is proportional to wavelength, so the spot grows by a factor of about 785/532, roughly 1.5.

Exam focus

Be ready to calculate resolution from d ≈ 0.61λ/NA, to explain confocal depth sectioning, and to state the trade-off between shorter wavelengths (better resolution and signal) and longer wavelengths (less fluorescence and often less photodamage). Explain why band position maps report strain or composition.

Advanced insight

In silicon devices, the 520.7 cm⁻¹ phonon band shifts by roughly 1–2 cm⁻¹ per GPa of stress, so position maps reveal stress around transistors and through-silicon vias. In graphene, maps of the 2D-to-G intensity ratio and 2D band shape distinguish single-layer from multilayer regions, while the D band maps defects and edges. Tip-enhanced Raman spectroscopy combines a plasmonic probe tip with scanning-probe microscopy to reach resolutions of tens of nanometres or better.

Summary

Raman microscopy focuses a laser through a high-NA objective and collects scattered light from a sub-micrometre spot, with lateral resolution d ≈ 0.61λ/NA and confocal depth resolution of about 1–2 µm. Raster scanning produces a hyperspectral cube from which band intensity, position, width or multivariate images are built. Resolution, signal, fluorescence and sample heating must be balanced when choosing wavelength, power and exposure.

Practice questions

1. Calculate the lateral resolution of a 633 nm Raman microscope fitted with a 0.75 NA objective. Answer: d ≈ 0.61 × 633 ÷ 0.75 ≈ 515 nm, about 0.5 µm. 2. What extra information does a map of band position give compared with a map of band intensity? Answer: Band position reports local strain, stress, temperature or composition, whereas intensity mainly reports how much of a component is present. 3. Why is laser power usually tested on a single spot before a map of a dark sample is acquired? Answer: Tight focusing gives very high power density, and absorbing samples can heat, transform or burn, which would distort the spectra. 4. Two components have strongly overlapping spectra. Suggest how to produce a reliable chemical map. Answer: Use multivariate analysis, such as least-squares fitting to reference spectra or principal component analysis, that uses the whole spectrum rather than a single band.