Raman Spectroscopy of Carbon Nanomaterials: Reading the D, G and 2D Bands

Why Raman is the routine check for carbon materials

Every other technique on the characterisation menu asks for something. Electron microscopy wants an electron-transparent specimen and a vacuum. X-ray diffraction wants a hundred milligrams of dry powder. Surface-area analysis wants an overnight degas. Raman wants a speck of material on a glass slide and about thirty seconds, and for carbon nanomaterials it answers two questions nothing else answers as cheaply: how defective is this, and how many layers is it.

That accessibility is also why Raman numbers are quoted more loosely than any other characterisation figure in the trade. A supplier datasheet that reports a single intensity ratio, with no laser wavelength and no spectrum attached, has given you a number that cannot be compared to anybody else's. This guide is about reading the spectrum rather than the ratio.

What the instrument measures

Inelastic scattering, in one paragraph

A monochromatic laser illuminates the sample. Almost all the light scatters elastically at the incident wavelength — Rayleigh scattering — and is filtered out. A very small fraction exchanges energy with a vibrational mode of the material and comes back shifted in frequency. That shift, conventionally reported in wavenumbers (cm^-1), is a direct measurement of vibrational energy, and because the vibrational modes depend on bonding and symmetry, the spectrum is a structural fingerprint.

Why carbon is unusually loud

Raman scattering is normally weak. In graphitic carbon it is not, because the laser energy is resonant with real electronic transitions in the material's continuum of pi states. Every visible laser line is resonant with something in graphene or a nanotube. The result is a strong signal from a very small amount of material — and, less conveniently, an intensity that depends on which laser you used.

The four features you need to recognise

The G band, near 1580 cm^-1

The G band is the in-plane stretching of sp²-bonded carbon pairs. It is present in every graphitic material — graphite, graphene, nanotubes, carbon black, graphene oxide — because it does not require any particular long-range order, only sp² bonds. Its position shifts with doping and strain, which makes it a useful reporter for functionalised material, and it is the reference against which the defect band is measured.

The D band, near 1350 cm^-1

The D band is defect-activated. In a perfect infinite graphene sheet it is forbidden; it needs a defect — a vacancy, an sp³ carbon, a functional group, a grain boundary or simply an edge — to conserve momentum. That is the whole basis of using it as a quality metric: no defects, no D band.

Two consequences follow immediately and are routinely missed. First, an edge is a defect as far as this measurement is concerned, so small flakes give a D band even when the lattice inside them is perfect. Second, the D band is dispersive: its position moves with the excitation energy, by something in the region of 50 cm^-1 per electronvolt of laser photon energy. Quoting a D-band position without the laser line is quoting an incomplete number.

The 2D band, near 2700 cm^-1

Also written G' in older literature. It is the second-order overtone of the D band, and — this is the important part — it is not defect-activated. Two phonons can conserve momentum between them, so the 2D band appears in defect-free material. It is the strongest feature in good monolayer graphene, and its shape is the standard layer-counting tool.

The radial breathing mode, 100-350 cm^-1

Unique to single-walled nanotubes: the whole tube expanding and contracting radially. Its frequency scales inversely with tube diameter, and the relation is usually written as a constant divided by the diameter plus an offset. The practical caveat is that the published parameterisations of that relation disagree with one another, because the offset depends on the tube's environment — bundled, suspended, surfactant-wrapped or on a substrate. Use it to see that you have a distribution of diameters and roughly where it sits; do not treat a diameter derived this way as a calibrated measurement unless the parameterisation and environment are stated.

Multi-walled tubes do not show a usable radial breathing mode. The outer walls are too large and too stiff, and the signal from the innermost tube is buried.

I(D)/I(G): the ratio everyone quotes

The intensity ratio of the D and G bands is the standard defect metric. It is genuinely useful and it is also the single most over-interpreted number in the field, for three reasons.

It is not monotonic

In the nanocrystalline regime — a graphitic material with a well-defined in-plane crystallite size — the ratio rises as the crystallite size falls, which is the behaviour described by the Tuinstra-Koenig relation and the behaviour everyone assumes. Push the disorder further, into genuinely amorphous carbon, and the trend reverses: with few enough intact aromatic rings left to scatter, the D band weakens again. A sample with a low ratio is therefore either very good or very bad, and the ratio alone cannot tell you which. The G-band position and width break the tie — amorphous carbon shows a broad, upshifted G band, good graphite a narrow one.

Area or height changes the answer

Some groups report peak-height ratios, some peak-area ratios from a fit. These are not the same number, and they diverge exactly when the bands are broad — that is, in the defective samples you most wanted to compare. A ratio reported without stating which one was used is not comparable to one that states the other.

It depends on the laser

Because the resonance conditions differ, the same sample measured at 488, 532, 633 and 785 nm gives four different ratios. Any comparison between a supplier's figure and your own measurement requires the same excitation wavelength. If the datasheet does not state it, the figure is not usable for comparison and you should ask.

The D' shoulder

A shoulder near 1620 cm^-1, on the high-wavenumber side of G, is another defect-activated band. Its ratio to the D band carries information about what kind of defect dominates — boundary-like, vacancy-like or sp³. It requires clean spectra and careful fitting, but it is the route from "defective" to "defective in this particular way", which is the more useful statement.

Counting graphene layers from the 2D band

For mechanically exfoliated or CVD graphene, the 2D band shape is the most reliable quick layer count available:

• Monolayer — a single sharp, symmetric Lorentzian, typically a few tens of wavenumbers wide, and usually taller than the G band.
• Bilayer — visibly broader and asymmetric; the electronic structure splits and the band is properly fitted with four components rather than one.
• Few-layer — broadens further and shifts upward, and the 2D-to-G intensity ratio falls below one.
• Graphite — a characteristic two-component shape, quite distinct from the monolayer's single peak.

Two limits to keep in mind. This diagnostic assumes Bernal-stacked layers; turbostratic material, where the layers are rotationally misaligned, decouples electronically and can present a monolayer-like 2D band with several layers present. And it assumes a low defect density: a strong D band broadens everything and the layer count degrades into a guess. For liquid-exfoliated graphene powders, which are usually turbostratic and small-flaked, do not use the 2D shape as a layer count at all — that is a job for AFM, TEM or a statistical optical method.

Graphene oxide and reduced graphene oxide

Heavily oxidised material gives broad, strongly overlapping D and G bands, a high intensity ratio, and a weak, poorly defined 2D band — the last of which is the diagnostic that matters. Reduction restores conjugation, and the spectrum responds: the 2D band recovers definition. What does not happen is the intensity ratio falling neatly back towards the graphite value. Reduction creates many small sp² domains rather than restoring large ones, so the ratio often stays high or even rises. Reading that as "reduction failed" is a common and expensive misreading; the 2D band and the electrical conductivity are the honest checks. We take this apart in the guide to choosing between graphene, GO and rGO.

How the measurement goes wrong

Laser-induced damage

The laser is being focused into a diffraction-limited spot on a strongly absorbing black material. It is entirely possible to oxidise, anneal or ablate the sample during the measurement and then report the spectrum of the damaged material. Graphene oxide and small-diameter nanotubes are the most sensitive. The standard defence is to run a power series — measure at successively lower laser powers until the spectrum stops changing — and to work below that threshold. A spectrum acquired at unstated power on a beam-sensitive material carries an unquantified risk of being a spectrum of something you made in the microscope.

Fluorescence background

Organic contamination, some surfactants, and certain substrates fluoresce far more strongly than the sample scatters, burying the spectrum under a sloping background. Longer excitation wavelengths — 633 or 785 nm — usually suppress it, at the cost of signal.

Sampling statistics

The probed volume is roughly a micron across. A powder is not homogeneous at that scale. A single spectrum from a bulk nanomaterial powder is an anecdote; a defensible number comes from a map or from many points across several areas, reported with a spread. Ask suppliers for the number of points behind any quoted ratio.

What to ask a supplier for

• The spectrum itself, not only a ratio. It costs them nothing and it is the only way you can check the assignment.
• Excitation wavelength and laser power at the sample. Without the first, no ratio is comparable; without the second, no spectrum is provably undamaged.
• Whether the ratio is height or area, and how the baseline was subtracted.
• How many points were measured, and the spread across them.
• For graphene products, the 2D band shape and whether the material is claimed to be Bernal-stacked.

Those five lines are what separate a Raman figure that supports a purchase decision from one that decorates a datasheet. The same principle applied across the whole document is how to read a certificate of analysis.

Standards worth knowing

ISO/TS 21356-1 specifies structural characterisation of graphene from powders and dispersions, and it exists precisely because the label on the bottle has proven unreliable. ISO/TS 80004-13 is the vocabulary standard for graphene and related two-dimensional materials — worth reading once, because it defines terms such as "few-layer graphene" and "graphene nanoplatelet" that are used interchangeably in commerce and are not interchangeable.

Related guides

• Deciding which measurements a material needs at all: the full characterisation workflow.
• Raman sees bonding; XRD sees the crystalline domain. For that, see X-ray diffraction for nanomaterials.
• The residual-catalyst question Raman cannot answer: TGA for nanomaterials.