Significance Comment

This is an article that has undoubtedly had a very large impact, with (at time of writing, according to Google Scholar) over 3400 citations. It (along with Probing single molecules and single nanoparticles by surface-enhanced Raman scattering, Nie, S., & Emory, S. R., Science, 275, 5303,(1997)) is responsible for demonstrating the possibility of Single-Molecule SERS (SM-SERS), and for kick-starting that field.

Thus, this paper is important, even if it does contain errors which detract from it.

Quality Comment

This article contains some serious errors, and while many in the field acknowledge them as errors, some of the erroneous claims are still repeated by some. This is despite many more recent papers which have used more reliable methods, and is a result of this paper's high Impact/Quality ratio.

Starting from the abstract, errors appear. "Spectra observed in a time sequence for an average of 0.6 dye molecule in the probed volume exhibited the expected Poisson distribution for actually measuring 0, 1, 2 or 3 molecules." This claim regarding Poisson distributions has now been thoroughly rebutted, as a consequence of insufficient samples. In fact, the claim of a Poisson distribution is equivalent to claiming quantised enhancement factors, which as anyone familiar with how electric fields are distributed near a metallic sphere will know is an unsupportable claim. (Electromagnetic enhancement is the primary mechanism of SERS enhancement.)

The other major error in this paper is the claim of enhancement factors of "about $10^{14}$", a figure based on comparing the signal from crystal violet to that of methanol, and assuming that the non-SERS Raman cross-sections of those two molecules are the same, and comparing ratios of numbers of molecules. Methanol is a small molecule (CH$_3$OH) compared to Crystal Violet (C$_{25}$N$_3$H$_{30}$Cl), and as such has a smaller Raman cross section. Assuming that the two have equal cross sections leads to an overestimation of the SERS enhancement factor, by several orders of magnitude (it is hard to say what the EF that they observed was, but SM-SERS could be observed with an EF of $10^8$, and claims above $10^{11}$ are a stretch).

This is not to question the claim of single-molecule detection; I believe that single molecules were detected, and the problem rather is in the analysis performed to support that claim.

Summary

This is a seminal paper in the history of SM-SERS, and reports seeing SM-SERS from Crystal Violet adsorbed on silver colloids, using a NIR (830nm) laser. The concentrations used were such that the expected number of CV molecules in the scattering volume is low (0.6), which is responsible for the Poisson distribution claim, as this leads to scarce statistics. More recent techniques (BiASERS, eg Proof of Single-Molecule Sensitivity in Surface Enhanced Raman Scattering (SERS) by Means of a Two-Analyte Technique, E. C. Le Ru, M. Meyer, and P. G. Etchegoin, J. Phys. Chem. B 110, 1944 (2006)) use higher concentrations, with a mechanism for determining when it is likely that more than one molecule is being probed, which allows for more events to be observed. As well, diluting to the $10^{-14}$M concentration used has many potential problems, such that the final concentration might not be entirely reliable.

It should be highlighted that this review was written 17 years after the paper, and as one would expect the field has advanced in this time, so some of the now-obvious shortcomings may not have been so apparent at the time.