Same Raman D/G ratio, different defect

If you have read a materials paper about carbon, you have seen the move. The Raman spectrum shows a G peak and a D peak, the authors divide one height by the other, and the ratio I_D/I_G becomes "the defect density." Higher ratio, more defects. It is quick, it needs no sample prep, and it turns a spectrum into a single number you can put in a table.

That number is not useless. It is just smaller than it looks. Report 013 covered the case where it breaks down completely, in heavily oxidized graphene. This report is about the cases where it works, and what it still leaves out.

Where the ratio comes from

The G band is allowed in perfect graphite. The D band and a smaller one called D' are not: in a flawless lattice the scattering cannot conserve momentum, so they stay dark. A defect supplies the missing momentum and switches them on. That is why the D peak is a defect signal at all, and why the ratio of D to G has been used since the 1970s, when Tuinstra and Koenig tied it to crystallite size (the idea behind the X-ray sizing in Report 192, by a different route).

Problem one: the same ratio, two densities

The ratio does not climb forever. Eckmann and colleagues (University of Manchester, with Freie Universität Berlin and the Karlsruhe Institute of Technology) summarized the earlier ion-bombardment work this way: at low defect density I_D/I_G rises as defects get closer together, and past a turning point it falls again, because the defects are now so dense that they cut the electron lifetime the D peak depends on. They call these Stage 1 and Stage 2. In the work they cite, the turn came at a ratio of about 3, with a 2.41 eV (514.5 nm) laser, when defects were about 2 to 5 nm apart. In their own oxidized graphene it came at about 4.

So a ratio below the peak value matches two samples: one lightly damaged, one heavily damaged. The ratio alone cannot tell you which side you are on. You need another clue, such as the width of the G peak or how the sample was made.

Problem two: how much, not what kind

This is the paper's central result. The team made graphene with different kinds of defects: sp³ sites, where an atom such as fluorine or oxygen bonds onto the sheet and pulls a carbon out of its flat bonding; vacancy-like damage from argon-ion bombardment, taken from earlier studies; and grain boundaries in polycrystalline graphite, also from the literature. They added their own flakes made by a method called anodic bonding, whose defects nobody had characterized.

Their Figure 2 is the one to remember. A fluorinated flake and an anodic-bonded flake gave I_D/I_G of about 2.3 each, with the same G peak width (about 24 cm^-1), and their D, G and 2D peaks matched almost perfectly. By the usual reading, the same material. The only clear difference was the D' peak, the shoulder on the high side of G.

That led to the number the paper is cited for. In Stage 1, theory says both D and D' grow in proportion to the number of defects, so their ratio I_D/I_D' should depend on the kind of defect and not on how many there are. Fitting all the samples, the authors found I_D/I_D' was largest, about 13, for sp³ defects, about 7 for vacancy-like defects, and smallest, about 3.5, for grain boundaries in graphite.

They then checked it with a microscope that did not depend on Raman at all. The anodic-bonded flakes fell on the vacancy line, and atomic force microscopy showed them riddled with holes, typically below 100 nm across. The fluorinated flakes showed patches of low conductivity about 20 to 30 nm wide, which the authors attribute to fluorinated clusters. Two methods agreed on what the single D/G ratio had hidden.

Where the D' trick stops

The paper is careful about its limits, and they are worth carrying around:

Stage 1 only. In Stage 2 the authors found no clear dependence on defect type, probably because the defects are so dense that the information about any single one is lost. At high disorder the D' peak also merges into G and cannot be separated.

D' is small. At moderate disorder it can reach about a third of the G peak's height, but it is often a faint shoulder. Its value depends entirely on how you fit it, and the team used a different line shape for D' (a Fano profile) than for the other peaks.

Theory disagreed. The ab initio calculations available then predicted about 10.5 for vacancy-type defects and 1.3 for sp³, which is not only different numbers but the opposite order. The authors attribute this to idealized defects in the calculations. The 13, 7 and 3.5 are empirical slopes from specific samples, not constants of nature.

One laser, one material. These are values measured at 2.41 eV on graphene and graphite. The D band depends on the excitation laser wavelength, so carrying ratios between instruments with different lasers is an assumption.

In bulk carbons it gets harder

Much of the carbon in the literature is not a clean graphene flake. It is fiber, char, hard carbon for batteries, or soot, and the spectrum is a broad hump that has to be split into peaks before any ratio exists. A 2026 paper by Yamada and colleagues in the Journal of Materials Science, the same paper behind Report 116, lists the reasons this is hard. Among them: the broad D region in defective carbons contains vibrations besides the classic breathing mode, and the number of peaks people fit varies widely. Their own survey: Li and colleagues used nine peaks between 1000 and 1800 cm^-1 in 2006, Smith and colleagues nine in 2016, Brubaker and colleagues argued in 2021 for at least five for carbon fiber, and Yamada's group has used eleven and twelve.

Their own calculations add one more wrinkle: the features at about 1500 to 1550 cm^-1, between D and G, come from carbon double bonds in hexagonal rings affected by nearby non-hexagonal rings and by oxygen groups such as cyclic ethers. In a spectrum like that, D' is one peak among many fitted ones, and the value you report for it depends on how many neighbors you allowed it.

How to read a D/G ratio

Which stage? If the paper does not say whether the sample is lightly or heavily disordered, the ratio is ambiguous.

Heights or areas? Eckmann's team found the two equivalent in Stage 1 but not in Stage 2, where peaks broaden. Papers use both.

Which laser? A ratio without the excitation wavelength cannot be compared with anyone else's.

How many peaks were fitted? In a broad spectrum the D, G and D' heights are outputs of a fit, and a different peak count gives different numbers.

Does the paper claim a defect type? D/G cannot support that alone. D/D' can, in Stage 1, on graphene-like material, with an independent check. Without one of those, "defect-rich" is the most the ratio supports.

The honest limits on this

I read the 2012 paper in its arXiv version (posted July 2012, same title and authors as the Nano Letters article); I did not compare it line by line against the published journal text or read the supplement. The literature points it makes about earlier work, and the peak-count survey from Yamada, are those papers' summaries of other studies, not studies I opened. The 2012 result has been cited widely since, and later work may refine the 13, 7 and 3.5 values; I have not reviewed that literature here. My own research is in microwave spectroscopy, not Raman, and none of my work is involved in either paper.