The feature article “Surface-Enhanced Raman Scattering” by Katrin Kneipp (Physics Today, Physics Today 0031-9228 60 11 2007 40 https://doi.org/10.1063/1.2812122 November 2007, page 40 ) presents views and results that are largely outdated; some are no longer generally accepted by the SERS research community, and some have been proven wrong. We counter Kneipp’s opinion by focusing on two important aspects of the article: the magnitude of the SERS enhancement factor and the proof of single-molecule sensitivity.
Some of Kneipp’s claims are based on interpretations that are more than 10 years old. For example, the claim of the “remarkable 14-order-of-magnitude signal enhancement” is now known to be an overestimation of the actual enhancement factors achievable in SERS and necessary for single-molecule detection. 1 The origin of Kneipp’s claim can be traced back to the improper normalization of SERS signals: The SERS intensity of a preresonant molecule (crystal violet) was compared with the non-SERS intensity of a nonresonant molecule (methanol), an error that magnifies the enhancement by an artificial factor. 1 Simple characterizations of the bare Raman cross sections of these molecules show that using methanol as a reference induces an error in the enhancement by a factor of about 1200 at 785-nm laser excitation, and it is still a factor of about 350 at 1064 nm, further in the near-IR.
In 1999 Hongxing Xu and coworkers showed that a proper normalization with respect to the correct Raman cross section results in a maximum SERS enhancement factor of approximately 1010, even for single-molecule conditions. 2 There is currently no strong experimental or theoretical evidence for SERS enhancements larger than around 1010–1011. Moreover, SERS has been used more recently, with even smaller enhancements, for single-molecule detection of resonant molecules. Andreas Otto, in addition, argued on purely theoretical grounds that enhancement factors of approximately 107 are already sufficient to see single molecules. 3 Recent experiments have confirmed those estimates. 1
Similar comments can be made regarding the claim that a Poisson distribution exists in the single-molecule SERS regime, as shown in figure 2 of Kneipp’s article. Simple arguments can show that this apparent Poisson distribution is an artifact of the poor sampling of events over a long-tail distribution of SERS enhancement factors. 4 Statistics over a much larger set of data would wash out completely the discrete peaks in the figure’s histogram. Unfortunately, over the years, there has been a remarkable lack of disposition to substantiate those claims with more reliable statistics beyond a few hundred spectra. As a consequence, that approach cannot be used as proof of single-molecule sensitivity in SERS. Bruno Pettinger and coworkers discussed that issue in detail for the related technique of tip-enhanced Raman spectroscopy (TERS). 5
Fortunately, the single-molecule detection capability of SERS has since been proven unambiguously using other more reliable approaches, among them the Langmuir–Blodgett films used by Ricardo Aroca and coworkers, 6 the technique of bi-analyte SERS developed by our group, 1 and TERS.
In short, we believe that Kneipp’s article ignores the dynamic nature of the understanding in the field and the huge amount of work from many contributors over the past 10 years—work that has dramatically improved and in many cases radically corrected the interpretations of earlier pioneering studies.
