a n a l y t i c a c h i m i c a a c t a 6 0 6 ( 2 0 0 8 ) 119–134
avai lab le at www.sc iencedi rec t .com
journa l homepage: www.e lsev ier .com/ locate /aca
Review
Achievements in resonance Raman spectroscopyReview of a technique with a distinct analyticalchemistry potential
Evtim V. Efremov, Freek Ariese ∗, Cees GooijerDepartment of Analytical Chemistry and Applied Spectroscopy, Laser Centre Vrije Universiteit Amsterdam, The Netherlands
a r t i c l e i n f o
Article history:
Received 19 July 2007
Received in revised form
31 October 2007
Accepted 2 November 2007
Published on line 26 November 2007
Keywords:
Bioanalytical applications
Carotenoids
Metalloproteins
Carbon nanotubes
Excitation profiles
Fiber optics
Surface-enhanced Raman
Tip-enhanced Raman
Fluorescence rejection
Gated detection
a b s t r a c t
In an extended introduction, key aspects of resonance Raman spectroscopy (RRS) such as
enhanced sensitivity and selectivity are briefly discussed in comparison with normal RS.
The analytical potential is outlined. Then achievements in different fields of research are
highlighted in four sections, with emphasis on recent breakthroughs: (1) The use of visible
RRS for analyzing carotenoids in biological matrices, for pigments and dyes as dealt with in
art and forensics, and for characterizing carbon nanotubes. (2) The use of RRS in the deep
UV (excitation below 260 nm) in the bioanalytical and life sciences fields, including nucleic
acids, proteins and protein–drug interactions. Metalloproteins can be studied by visible RRS
in resonance with their chromophoric absorption. (3) Progress in theoretical calculations of
RRS excitation profiles and enhancement factors, which ultimately might facilitate analy-
tical RRS. (4) Instrumental and methodological achievements including fiber-optic UV–RRS,
coupling of RRS to liquid chromatography and capillary electrophoresis. Sensitivities can
approach the single-molecule level with surface-enhanced RRS or tip-enhanced RRS. Last
but not least, promising fluorescence background rejection techniques based on time-gated
detection will be presented. This review ends with a concluding section on future expecta-
tions for RRS, in particular its potential as an analytical technique.
Raman spectroscopy (RS) is based on the inelastic scatteringof radiation (usually in the visible of near-infrared region) by asample, which can be a solid, a liquid or a gas. By far the largestfraction of the scattered light has the same wavelength as thelaser light (Rayleigh scatter). However, in the most commonform of Raman scattering (Stokes Raman) also at longer wave-lengths weak scattered light is observed: its photon energyis reduced by a vibrational energy quantum of the scatteringmolecule. In other words, RS is a vibrational spectroscopictechnique – complementary to IR spectroscopy – but excitationand emission involve higher-energy photons, similar to thoseused in electronic absorption and fluorescence spectroscopy.A highly monochromatic light source must be used. Originallythis was commonly a mercury lamp with a bandpass filter, butnowadays lasers are almost exclusively used for excitation.Since practically all materials have Raman-active vibrations,the solvent or matrix often shows up in the spectra, whichcomplicates the analysis of (dilute) samples with conventio-nal RS. Fortunately, water is a relatively poor Raman scatterer,so that RS is the vibrational technique of choice for bioana-lytical applications. As early as 1928 Sir C.V. Raman reportedthe effect that was later named after him [1], but only in thelast two decades has RS developed into a mature techniquefor chemical analysis in a broad range of application areas.The state-of-the-art in conventional (non-resonance) Ramanis extensively covered in the monographs by McCreery [2] andPelletier [3]. The emphasis of the present review is on a specialmode of RS: resonance Raman spectroscopy (RRS).
It should be stressed that the basic feature of RS is scat-tering, which can occur at any laser wavelength. There is achange in the direction of the light, but no photon annihila-tion takes place as in an electronic absorption or fluorescenceprocess. The selection rules are very different from those inIR: Raman signals are associated with vibrations that causea change in polarizability. Normally Raman emission is ratherweak, but it can be enhanced by several orders of magnitude inthe special situation where the laser wavelength is close to an
electronic absorption band (resonance). The smaller the fre-quency difference between laser and electronic transition thestronger the RRS intensity, but as a result of the finite widthof the absorption band (homogeneous and inhomogeneous
broadening), this difference will never be zero and the RRSintensity will not increase to infinity. In RRS only the vibrationscoupled to the chromophoric group are intensified in the spec-trum. The Jablonski energy diagram in Fig. 1 shows the processof resonance Raman and also illustrates why fluorescence canbe expected to cause interference.
To the best of our knowledge, resonance enhancement wasfirst experimentally observed in 1946 by Harrand and Lennuierat Sorbonne University [4]. They noticed (with experimentsthat required 360-h exposure times!) that in the case of adichloronitrobenzene solution there is strong Raman enhan-cement if the excitation wavelength is chosen fully within theabsorption band, and much less if excitation occurs in theflank. A non-enhanced solvent vibration (CCl4) was used asinternal standard. They also noticed that only a few vibrationsshowed enhancement and postulated that the correspondingfunctional groups were also responsible for the absorptionband. In the following years the phenomenon was extensi-vely studied, especially in the Soviet Union, with emphasis onphysical chemistry and theoretical chemistry aspects [5]. The(bio)analytical potential became clear in the 1970s [6], soonafter the introduction of the laser. Since resonance effectsare only expected if the laser line coincides with an electro-nic absorption, RRS combines sensitivity and selectivity. Inprinciple, it enables one to selectively observe a chromopho-ric solute in a dilute solution or study a particular Ramanscatterer in a complex matrix. For example, in the case ofheme proteins we can choose a laser wavelength close to theabsorption maximum of the porphyrin group (ca. 410 nm) andobserve the latter with minimal interference. The remainderof the protein will not show any enhancement (being off-resonance) and therefore its Raman signals, which in normalRS would dominate the spectrum, will be relatively weak.
There are several reasons why the acceptance of RRS as ananalytical method has been relatively slow. First of all, manylaser systems still suffer from a lack of tunability. The free-dom of excitation wavelength choice is essential for optimumsensitivity and selectivity. Secondly, RRS often suffers fromfluorescence background, which can fully obscure the Raman
signals. As illustrated in Fig. 1, fluorescence is likely to occurif the laser light is absorbed by the analyte or chromophoricgroup, which is of course always the case in RRS. Further-more, in real sample analysis, background fluorescence from
a n a l y t i c a c h i m i c a a c t a 6
Fig. 1 – Jablonski energy diagram illustrating the process of(Stokes) resonance Raman scattering. The difference inphoton energy between the excitation source and theobserved Raman line corresponds to a specific vibrationalmode of the electronic ground state S0 of the probedmolecule. The diagram also illustrates that when theexcitation energy overlaps with the first electronictransition (S –S ), fluorescence from the analyte is likely toi
opfcelsybcsrartattl(ba
cWnf
(
0 1
nterfere with the RRS emission.
ther components or from the matrix can also be a majorroblem. Matrix fluorescence typically decreases upon goingrom the UV via the visible to the near-infrared, so that inonventional RS a trade-off is commonly made: at very longxcitation wavelengths the fluorescence background is veryow, more than compensating the fact that the Raman inten-ity decreases with a factor �−4 and that detector quantumields will be lower. Obviously, in RRS such an approach cannote followed. Rather unexpectedly, fluorescence interferencean also be avoided by using very short wavelengths: Asherhowed that matrix and analyte fluorescence hardly play aole below 260 nm [7]. Recently other options have becomevailable to suppress fluorescence in the near-UV and visibleange, as will be discussed in Section 5.4. A third obstacle tohe acceptance of RRS as an analytical method is the enhancednalyte photodecomposition under resonant excitation condi-ions, especially in the case of stationary samples. The abovehree points are to some extent interrelated. If laser wave-engths are created by frequency doubling or tripling of pulseddye) lasers at a low repetition rate, the peak powers tend toe relatively high so that photodecomposition is difficult tovoid.
It is the purpose of this review to highlight the analyti-al potential of RRS with emphasis on recent developments.e hope to provide a balanced overview, but given the broad-
ess of the topic it will necessarily be far from complete. Theollowing topics will be discussed:
1) Visible RRS. First we will discuss visible RRS as a well-known tool for analyzing strongly absorbing organic
components such as carotenoids, and an appropriatetechnique for pigments and dyes in areas such as art,archeology and forensics. It is also particularly suitablefor the characterization of a new class of materials,
0 6 ( 2 0 0 8 ) 119–134 121
single-walled carbon nanotubes, which show very sharpabsorption features in the visible and NIR.
(2) RRS in life sciences. Secondly, we will focus on recentachievements in the bioanalytical and life sciences field,including nucleic acids, proteins and protein–drug interac-tions. UV–RRS has been utilized (i) as a tool to study nativenucleic acids and complex biological assemblies contai-ning either DNA or RNA; (ii) as a tool to study drug–proteininteractions by monitoring the spectra of tryptophan andtyrosine residues; (iii) at about 200 nm monitoring theamide band of the peptide bond to study protein and pep-tide folding. For some metalloproteins, RRS in the visiblerange is widely used.
(3) Computational aspects. Since in RRS only specific vibra-tions are strongly enhanced while others remain weak,the resulting spectra often look very different fromconventional (off-resonance) Raman spectra. Theoreticalcalculations of excited-state potentials can be an impor-tant help to predict the enhancement factors. Attentionwill also be paid to the observation of intense overtones.
(4) Instrumental breakthroughs. Finally, we will discuss someimportant instrumental and methodological develop-ments achieved in the past decade. Laser systems havebecome more flexible, rugged and user-friendly, allo-wing a wide choice in excitation wavelengths and (peak)powers. Fiber-optic probes have been adapted to RRSand UV–RRS. UV–RRS was developed for detection andidentification in liquid chromatography (LC) and capillaryelectrophoresis (CE). Surface-enhanced resonance Ramanspectroscopy (SERRS) and more recently tip-enhancedresonance Raman spectroscopy (TERRS) at or near thesingle-molecule sensitivity level have been reported, thelatter also in combination with an unprecedented spa-tial resolution, well below the diffraction-limited spot size.Last but not least, new developments will be discussedin the area of fluorescence suppression. By using picose-cond excitation pulses and time-gated detection via Kerrgating or fast intensified CCD detectors, the fluorescenceinterference can be strongly reduced.
In a concluding section, we will present our expectationsregarding the future importance of RRS in chemical analysis.
2. RRS in the visible range
2.1. Carotenoids
As mentioned above, the analytical potential of RRS becameobvious in 1970, when Gill et al. [6] detected lycopene and beta-carotene in intact plant samples (carrot and tomato), as well asin carrot juice and tomato sauce. The resonance enhancementwas sufficiently strong to overcome losses due to absorptionof the exciting and scattered photons, as well as interferencefrom fluorescence. The authors noted that small shifts invibrational frequencies between the plant spectra and those
of reference samples in n-hexane were probably due to solventeffects, and suggested that this phenomenon could probablybe used to probe the local environment. In another study thesame authors used all available emission frequencies of the
a c t a
122 a n a l y t i c a c h i m i c a
argon-ion laser to obtain the excitation profile (EP) of the twopigments [8]. EPs can provide additional information about theelectronic structure of a given molecule. In fact, measuring EPsintroduced a new type of molecular spectroscopic experimentand instigated further development of the theory of RRS effect[9,10], see also Section 4. The development of reliable and fastmethods for determining the concentrations of carotenoidsin various food products with no or minimal sample prepara-tion remains important today. A study on RRS for quantitativeanalysis of intact fruit and vegetable samples, validated by LCanalysis of extracts from the same samples, was presented byBhosale et al. [11].
Carotenoid-type molecules are essential components ofmost light receptors in animals and humans, as well as precur-sors for vitamin A synthesis. Therefore, considerable effort hasbeen devoted to biophysical studies of such systems [12–15].An intriguing contribution is the work of Carreira et al. onthe EP of the coherent anti-Stokes resonance Raman spec-trum of beta-carotene [16]. A time-resolved resonance Raman(TR3) approach was successfully used in a number of workson excited-state characterization of beta-carotene [17,18] andlycopene [19]. An overview on the subject was written by Dal-linger et al. [20]. Such studies continue today, helped by thedevelopment of femtosecond laser systems and fast detectors[21–23]. A recent review on the ultrafast dynamics of carote-noid excited states was presented by Polivka and Sundstrom[24].
An overview on carotenoid photooxidation in photosystemII and the spectral information that could be obtained fromRRS studies was compiled by Tracewell et al. [25]. Interestingly,the absorption spectrum of the radical cationic photoproductCar+ (kept stable in a frozen matrix at 85 K) extends so far intothe NIR region that Fourier-transform RRS could be carried outat 1064 nm.
Carotenoids are thought to play a significant role in theskin’s anti-oxidant defense system, and may help preventmalignancy. RRS was demonstrated to be an accurate, non-invasive technique for in vivo detection in all skin types [26].By means of selective excitation, lycopene and beta-carotenecan be determined separately [27]. The validity of in vivo RRSwas confirmed by LC analysis of skin extracts. The carote-noid content of carcinoma tissue was found to be significantlylower than in healthy tissue [28].
2.2. Minerals and pigments
In the world of archeology, art history and conservation, theubiquity of colored materials means that RRS in the visiblerange is often the preferred method to identify its chemi-cal constituents. The non-invasive character of RRS (providedlow photon fluxes are used) is obviously a major advantage.Sometimes it is also possible to microsample objects (removefragments of the order of tens of micrometers) for analysisin the laboratory under a Raman microscope. In the contextof this review, we highlight only a few recent, interestingexamples.
Coupry et al. [29] described the case of a Roman wall pain-ting found at Tivoli (Italy), depicting a hunter. RRS analysisof the blue pigment showed that it consisted of ultramarine,unusual for the period. The identification of the green pig-
6 0 6 ( 2 0 0 8 ) 119–134
ment as phthalocyanin (patented in 1936) and white pigmentas Ti(IV) oxide definitely proved the work to be a 20th centuryforgery.
Bell [30] demonstrated the importance of selective reso-nance excitation for the identification of the yellow naturaldye that was used to color the paper of the Diamond Sutraand other 9th century Chinese documents, believed to bethe World’s oldest printed works. Under blue excitation theRRS spectra were overwhelmed by fluorescence, whereas withlong-wavelength off-resonance excitation the spectra weredominated by the paper matrix. With �exc = 363.8 nm excita-tion better RRS spectra were obtained. The non-random noiseof the background (mainly pixel-to-pixel variation of the CCDdetector) was further decreased by a chemometric approach(shifted, subtracted RS), in which the grating position of thespectrograph was moved by 20 and 40 cm−1 and the resultingspectra subtracted.
Krampelas et al. used RRS at various visible and NIR wave-lengths to identify the pigments in colored freshwater pearls[31]. The C C and C–C stretch vibrations indicated polyenicstructures (but different from carotenoids), with 6–14 conju-gated double bonds. In most colored pearls mixtures of thesecompounds were observed. The usefulness of RRS in the foren-sic laboratory was tested for ballpoint inks on paper by Seifaret al. [32] and for dyed textile fibers by a consortium of forensicinstitutes [33].
2.3. Carbon nanotubes
Since the early 1990s carbon nanotubes and in particularsingle-walled carbon nanotubes (SWCNTs) are intensively stu-died in various areas of science as a result of their uniquemechanical, electrical, and chemical properties. SWCNTs canbe seen as a rolled-up sheet of a single layer of graphite (gra-phene), seamlessly fused and with a typical diameter of about1 nm. The length is often orders of magnitude greater. Manyproperties of SWNCTs, for instance whether they are semicon-ductive or metallic, are related to the diameter and chiralityangle, usually given by the indices (n, m).
Unfortunately, SWCNTs are produced as mixtures, witha distribution of (n, m) values. This hampers the thoroughcharacterization of their (photophysical) properties and theiroptimal use. Selective methods are required to analyze mix-tures and to assess the result of purification procedures.Suitable methods include absorption, Raman spectroscopy,and photoluminescence (most but not all species show emis-sion). SWCNTs have strong, narrow absorption features in thevisible and NIR range, the maxima depending on the (n, m)indices. Tuning the laser to these narrow absorption bands canprovide enormous selectivity in RRS. Jorio and coworkers pre-sented RRS excitation-emission matrices that showed sharpRaman spectra at different excitation frequencies. They useda tunable Ti–sapphire system and a series of Ar/Kr lines; atriple monochromator provided sufficient stray light rejectionso that low-frequency vibrations could be observed close toeach laser wavelength [34]. The radial breathing modes (RBMs)
in the 100–400 cm−1 range are useful to identify the geome-try of the compound [35]. Based on theoretical considerationsand comparisons with absorption and luminescence data RRScross-sections were derived and used to estimate the relative
t a 6
altcDas(c
3
3
Tn1cwdcRpAibpu
arpccabt1btatiotat
Dgsmd
nbFUl
small, and furthermore the marker bands overlap greatly withthe capsid protein bands. In this case resonance excitation at257 nm is of great help. Since the DNA bands are selectivelyenhanced, these spectra provide much information about the
Fig. 2 – UV–RRS spectra of the filamentous virus fd in water,
a n a l y t i c a c h i m i c a a c
bundance of the different species in the sample. The totalevel of semiconductive SWCNTs was about 11 times higherhan that of the metallic species [34]. In a follow-up publi-ation the same approach was used to test the selectivity ofNA wrapping as a purification method. The method yieldedmixture enriched in semiconductive SWCNTs [36]. Luo et al.
howed that a set of weak vibrations in the 400–600 cm−1 range“intermediate frequency modes”) can provide extra identifi-ation power [37].
. RRS in life sciences
.1. Nucleic acids
he nucleic acids DNA (deoxyribonucleic acid) and RNA (ribo-ucleic acid) have already been studied by RS since the late960s [38]. This interest can be readily understood: the geneticode in DNA contains the instructions for protein synthesis,hich is decoded with the help of different forms of RNA. So,etailed information about their structures under physiologi-al conditions is of main importance for biochemical studies.RS quantitation of nucleic acids can also be used for finger-rinting, for instance to identify bacterial strains (see below).s a result of the absorption properties of nucleic acids, RRS
s only possible in the deep UV, directed at the four DNAases adenine (A), cytosine (C), guanine (G) and thymine (T)resent as base pairs A-T or G-C. In RNA, T is replaced byracil (U).
Several modes of RS have been used to study nucleic acidsnd their complexes, including conventional RS and diffe-ence spectroscopy, polarized RS and UV–RRS. In the presentaper we highlight some recent applications of RRS in thisontext, with emphasis on analytical chemistry features. Aomprehensive review, describing in detail the state-of-the-rt and underlining recent progress, was recently publishedy the group of Thomas [39]. High-quality RRS spectra ofhe individual nucleosides were already published in the late970s, applying 266 nm excitation. The spectra are influencedy incorporating the bases into nucleic acid duplex struc-ures: not only from the electronic effects of base pairingnd stacking, but also from vibrational coupling to modes ofhe ribose-phosphate backbone. These effects were studiedn detail by Fodor and Spiro [40]. They compared RRS spectraf the deoxynucleotides dAMP, dTMP, dGMP and dCMP withhose of their copolymers poly(dA-dT) and poly(dG-dC) as wells calf thymus DNA, using 266-, 240-, 218- and 200-nm excita-ion.
In fact, Raman spectra of the nucleotide constituents ofNA and RNA are composed of many bands and their assi-nment is complex since the vibrational coupling with theugar moiety is not negligible. This may also hold for pro-inent bands previously denoted as marker bands, as clearly
emonstrated in the case of thymine [41,42].As regards analytical applications, the interaction of
ucleotides with various ions, as well as the effect of hydrogen
onding on RRS spectra have been investigated extensively.or this purpose, Takeuchi and coworkers have recordedV–RRS difference spectra of deuterium-labeled and unlabe-
ed purines. The information was successfully used to study
0 6 ( 2 0 0 8 ) 119–134 123
complex formation between the antibiotic actinomycin D anda DNA fragment [43].
RRS can also be used to study the interaction of DNA withsmall molecules, such as carcinogens or dyes. DNA-bindingcompounds have potential value as gene-directed therapeuticagents. In fact, rational drug design requires insight not onlyinto the binding characteristics, but also into the structuralchanges of genomic DNA caused by the interactions. RRS doesnot provide explicit information on the DNA backbone confor-mation, but it can be either directed at the DNA bases or atthe drug itself, provided that it has appropriate chromopho-ric properties. Examples of compounds that were successfullystudied include adriamycin [44], distamycin [45] and hypericin[46].
When applying UV–RRS to study nucleic acid–protein inter-actions it should be realized that wavelength selection isoften crucial. In fact, the frequently used 266 nm line provi-ded by the fourth harmonic of the Nd:YAG laser is very suitableand the same holds for the continuous-wave (CW) output ofthe frequency-doubled argon-ion laser at 257 nm. However atshorter wavelengths – for instance the 229 nm line that canalso be obtained from the same Ar+ ion laser – the aroma-tic amino acid residues of the protein become much morepronounced in the spectra (UV–RRS of proteins will be dis-cussed in the next section). This is illustrated in Fig. 2, wherethe UV–RRS spectra of a virus are shown. Whereas at 257 nmthe spectrum is dominated by viral DNA, excitation at 229 nmprimarily reveals the tryptophan and tyrosine residues [47].
It should be noted that for studying viruses, UV–RRS has aninherent advantage compared to conventional RS. In the lattermode the Raman markers of the single-stranded DNA genomeare very weak, since its contribution to the total virion mass is
excited at 229 nm (top) and 257 nm (bottom). Peak codes A,C, G and T indicate the DNA bases contributing to thespectra; W = tryptophan, Y = tyrosine. Na2SO4 was used asinternal standard (981 cm−1) (adapted from Wen et al. [47]).
124 a n a l y t i c a c h i m i c a a c t a
Table 1 – Experimental and calculated RRS intensities ofprominent bands for Eschericia coli
Raman bands (cm−1)
1240 1334 1485 1575 1618
Peak components T + U A A + G A + G Trp + TyrIntensities measured 3.47 10.3 22.2 9.15 3.78
Intensities calculated 3.96 13.9 27.1 11 1.3
Intensities in arbitrary units (from Wu et al. [50]).
base environments and their interaction with the viral capsid[48].
Whereas the Thomas group extensively developed the useof UV–RRS to study viruses, Nelson and coworkers focusedtheir attention on the study of living cells, especially thoseof bacteria, in particular Eschericia coli [49,50]. Changes in thenucleic acid contribution as a function of culture conditionswere observed. Furthermore, G-C/A-T molar ratios were rela-ted to the genus to which the bacteria belong [50,51]. For thesestudies, reproducible and quantitative UV–RRS informationwas crucial, especially regarding spectral intensities. To thisend a high duty-cycle picosecond tuneable UV laser sourcewas used in combination with a sensitive CCD detector [50].Optimum selectivity was obtained with 251 nm excitation, awavelength carefully selected in between the typical CW Ar+
wavelengths 244 nm (where interference from aromatic aminoacid spectra can be quite strong) and 257 nm (where fluores-cence interference begins to come into play). The challengewas to calculate the Raman spectral intensities for bacte-ria based on the known numbers and kinds of bases withinthe nucleic acids, simply using the free base cross-sections.Of course this approach is based on several assumptions:spectral shifts (hypochromism) and internal absorption canbe ignored and furthermore the spectra of e.g. guanine inDNA exactly matches the guanine spectrum of RNA. None-theless, the experimental and calculated intensities appearedto be quite close. This is obvious from Table 1. Using a simi-lar approach, bacterial growth could be monitored via 244-nmRRS focusing on specific bands reflecting the protein/nucleicacid ratio. The amount of nucleic acids is more or lessconstant in a cell whereas the protein content increases withgrowth [52].
Recent results on the detection and identification of sepa-rate nucleotides by means of UV–RRS coupled to separationmethods will be described in Section 5.2.
3.2. Proteins
Also in case of proteins, deep UV is the wavelength regionof choice for RRS. Metalloproteins, which often show strongabsorption also in the visible range, will be discussed in Sec-tion 3.3. UV–RRS of proteins has been applied extensively inlife sciences studies. Information can be obtained not only onthe aromatic amino acids phenylalanine (Phe), tyrosine (Tyr)and tryptophan (Trp), but also on the peptide bonds via the
amide frequencies. In the early 1990s a number of compre-hensive reviews have appeared, underlining the progress inthis field during the previous decade [53,54]. This occurred inspite of the fact that available laser sources were not easy to
6 0 6 ( 2 0 0 8 ) 119–134
handle and – even more serious – provided only output at avery low repetition rate (the first generation typically 20 Hz).This implied very high pulse energies and the risk of samplephotodamage. The use of CW lasers or high-repetition rate,low peak power Ti:sapphire lasers can significantly reducephotodegradation [55]. Recently a tunable system based onfourth harmonic generation of a Ti:sapphire laser was presen-ted for denaturation studies [56]. It could be tuned over the193–205 nm range.
N-Methylacetamide is an appropriate model for the pep-tide bond. There are two relevant absorptions, a weak oneat ca. 210 nm (n�* transition) and a stronger one at 186 nm(��*). The latter plays the main role in UV–RRS. Laser excita-tion around 200 nm reveals prominent amide modes at around1230–1300 cm−1 (amide III), 1390 cm−1 (amide S), 1550 cm−1
(amide II) and 1640–1655 cm−1 (amide I). The 1390 cm−1 bandis not observed in off-resonance RS; it is assigned to a vibra-tional mixing of a C–H deformation mode and the amide IIImode. As regards structural information: in UV–RRS the atten-tion is usually on the intensities of the amide II and the amideS modes, which correlate with the �-helical content of theproteins.
In the case of proline (Pro), the peptide bond contains noamide hydrogen – so that one is dealing with an imide instead– and as a result the 186 nm absorption band is red-shiftedover some 10 nm. As a consequence, by tuning the laser above210 nm X-Pro is selectively excited and also the vibrationalmodes are different from the normal amide modes. Gene-rally in UV–RRS a single band between 1460 and 1500 cm−1
is observed – denoted as the imide-II band – which is sensitiveto hydrogen bonding to the oxygen atom. As early as 1990 itwas used by Takeuchi and Harada for conformational studiesof short peptides in solution [57].
The aromatic amino acids Phe, Tyr and Trp all absorb stron-gly in the 195–230 nm region. Their vibrational modes are wellunderstood, in large part owing to the normal mode calcula-tions on model compounds. The tyrosine modes are sensitiveto hydrogen bonding of the phenolic OH group. At 218 nmexcitation the enhancement pattern in the tyrosine UV–RRSspectrum is similar to that observed for phenylalanine; howe-ver by exciting at longer wavelengths (230–240 nm), a highselectivity for tyrosine is observed. In view of the structure ofPhe – being unable to form hydrogen bonds – its Raman bandfrequencies are not influenced by the protein environment orsolvent. Also the presence of tyrosinate can be readily establi-shed; excitation in the 245–255 nm range has been shown togive high selectivity.
In tryptophan-containing proteins the Trp vibrationalmodes are prominent in the RRS spectra applying excitationbetween 218 and 230 nm. Trp exhibits a strong absorptionband at 218 nm (the Bb band), and a weaker absorption isfound at around 280 nm (comprising the close lying La andLb bands). Spectra are typically recorded at 220 and 260 nm(longer wavelengths are more difficult to handle because offluorescence background problems). The 880 cm−1 bond (W17mode), exhibiting a downshift upon forming hydrogen bonds,
is a useful marker for hydrogen bonding. The strong band at1555 cm−1 (W2 mode) is sensitive to the torsional angle of thechromophore-backbone chain, shifting to higher wave num-bers if the angle increases. Over the angle range from 60 to 120◦
t a 6
ttTa
amihih
3
Mcacmavppact
etatsε
mftRtm
tTloalsimpbptlfincdSo
a n a l y t i c a c h i m i c a a c
he shift is some 15 cm−1. This explains why in hemoglobinhe W3 band is a doublet, the main band is at 1558 cm−1 (tworp residues with the same angle) and a shoulder at 1550 cm−1
rises from a third Trp-residue.The relative intensities of the 1360 and 1340 cm−1 peaks
re influenced by the polarity around the indole ring, the for-er one being more intense in non-polar environments. Other
ntensity effects to probe changes in the protein environmentave been reported as well; they have been attributed to shifts
n absorption maxima, changes of local refractive index andypochroism.
.3. Metalloproteins
etalloproteins usually contain one or more coordinationomplexes of transition metal ions and protein ligands (suchs porphyrins), forming biologically active sites. These sitesan be involved in electron transfer, binding of exogenousolecules and/or catalysis. Usually in such complexes there
re strong electronic absorption bands in the near-UV orisible range, so that in principle selective excitation can beerformed without interferences from the remainder of therotein (300 nm is the longest absorption wavelength of aminocid residues, i.e. Trp). When there are several different metalenters in the protein, different laser wavelengths can be usedo excite these separately [58].
If the metal ion has a partially filled d-shell, three types oflectronic transitions can be discerned: d–d transitions, chargeransfer (CT) transitions (ligand-to-metal or metal-to-ligand)nd ��* transitions located on the ligand. The d–d transi-ions are too weak to produce useful RR enhancement [58],ince as a rule of thumb the RRS intensities are related to2, where ε is the molar extinction coefficient [59]. The muchore intense CT transitions are more appropriate. Especially
or Fe(III) and Cu(II) the ligand-to-metal transitions (being inhe visible region) received extensive attention in the earlyRS literature. If ligands such as CO, NO and O2 are boundo the metal ion, CT transitions in the opposite direction, i.e.
etal-to-ligand, are also important [58].In the case of metalloporphyrins such as heme proteins,
he ��* transitions of the ligand are of main importance.wo bands in the electronic absorption spectrum are avai-able, i.e. the Q-band located at about 550 nm (ε of therder of 10,000 M−1 cm−1) and the very strong Soret bandt about 400 nm (ε being 100,000 M−1 cm−1 or more). In theatter band the RRS spectrum is dominated by the totallyymmetric modes of the porphyrin ring, whereas excitationn the Q-band merely enhances the non-totally symmetric
odes. As a consequence, by combining these results a com-lete assignment of the porphyrin vibrational modes cane given [58,60]. Even though the highest-frequency por-hyrin ring modes do not directly involve the metal ion,hey nevertheless provide detailed information about theigation chemistry. Various marker bands have been identi-ed that are affected by the spin state or the coordinationumber (both in Fe(II) and Fe(III) complexes). In favourable
ases also vibrational modes of the axial ligands can beetected in the RRS spectra, as in case of CO, NO and O2.mulevich et al. [61] have recently compiled an overviewf heme peroxidase studies and described how RRS could
0 6 ( 2 0 0 8 ) 119–134 125
shed light on the importance of H-bonding on heme liga-tion.
Current research in this field is devoted to the details ofthe interactions of these small gaseous ligands XO (whereX = C, N, O) and heme sensor proteins. How does the hemedomain discriminate between CO, NO and O2 and how doesthe binding event lead to activation? In a recent paper theSpiro group gave a detailed interpretation of RRS results obtai-ned using the 413.1 nm and the 406.7 nm lines of a Kr ion laser[62]. They demonstrated anticorrelation between the stret-ching FeC and CO vibration frequencies in the CO adduct andalso for the FeN and NO vibrations in the NO adduct. In the lat-ter case, large changes were observed in the presence of distalH-bonds or positive charges. These differing RRS vibrationalpatterns for CO and NO adducts may help to elucidate themechanism of ligand discrimination and signaling in hemesensor proteins. Ibrahim et al. [62] supported their experimen-tal results with advanced theoretical calculations (i.e. densityfunctional theory, DFT), which underlines the increased powerand relevance of computational methods in current RRS (seealso Section 4). Furthermore it should be emphasized thatlaser-induced dissociation of the axial base (a rather specificexample of photodamage) was circumvented by using frozensolutions when recording RRS spectra.
Recently Bonifacio et al. [63] developed a small-volumespectro-electrochemical cell that could be used to recordspectra of cytochrome c while varying the potential of the elec-trode. The cell was moved to avoid photodegradation underthe Raman microscope. Changes in the Raman marker bandsindicated the reversible oxidation/reduction of the Fe-hemegroup (see Fig. 3). The RRS spectra were extra strong owingto the surface-enhancement effect (SERRS; also to be dis-cussed in Sections 5.2 and 5.3) from the silver electrode. Aself-assembled monolayer of mercaptopropionic acid on theelectrode prevented direct contact and possible denaturationof the protein on the Ag surface.
3.4. Drug–protein interactions
As outlined above, some vibrational bands of Tyr and Trp aresensitive to the microenvironment, i.e. polarity and hydro-gen bonding. A specific example is the intensity ratio of theso-called Fermi resonance doublet of Tyr, which is sensitiveto the extent of hydrogen bonding of the phenolic hydroxylgroup. Such dependencies can in principle be utilized to studydrug–protein interactions by UV–RRS. At present, there are stillfew applications. Hashimoto et al. [64] studied the interactionsof warfarin, ibuprofen and palmitate with human serum albu-min. They showed that warfarin interacts with Trp at position214, whereas ibuprofen interacts with the phenolic hydroxylof Tyr at position 411.
Couling et al. focused on drug binding in dihydroxyfolatereductase (DHFR), gyrase and catechol O-methyltransferase(COMT). They applied the intracavity frequency-doubledargon-ion laser output at 244 nm to record UV–RRS spectra[65]. An illustration of their results is reproduced in Fig. 4, sho-
wing the 244 nm RRS spectra of the gyrase/novobiocin system.Gyrase is a bacterial enzyme that catalyzes the introductionof negative supercoils into closed circular DNA. DNA gyrase isthe target of two groups of antibacterial agents, the quinolines
126 a n a l y t i c a c h i m i c a a c t a
Fig. 3 – Spectro-electrochemistry of cytochrome c inaqueous buffer (pH 7.4) on a silver electrode, coated with amonolayer of mercaptopropionic acid. At positive potential(a, 0.1 V) the oxidized form (six-coordinated low spin)predominates; lowering the potential results in a gradual(reversible) conversion to the reduced form, as indicated byseveral marker bands (b = 0 V; c = −0.1 V; d = −0.3 V;
colas, who applied it with success to different molecules
e = −0.5 V); excitation, 514 nm (from Bonifacio et al. [63]).
and the coumarins, such as novobiocin, the drug consideredhere. It is expected that novobiocin (which has a highly flexiblemolecular frame) is forced into a conformation when it bindsto the protein, a change that should be reflected in its vibra-tional spectrum. It should be noted that also novobiocin isresonantly enhanced at 244 nm (Fig. 4b). Therefore, to distin-guish between the spectral changes of the enzyme and thoseof the drug, spectra have to be subtracted as shown in Fig. 4d.Upon binding, both in the gyrase and the novobiocin spectrasubstantial changes are observed. A detailed analysis of thespectra can be found in Couling et al. [65].
Also for the other protein–drug systems studied, i.e. DHFR-trimethoprim and COMT-dinitrocatechol, detailed informa-tion could be obtained from 244-nm RRS. However, definitivestructural interpretation of the observed spectral changeswould require calibration experiments, which has not yet
been feasible. Therefore, structural analysis still relies on X-ray diffraction data (which were available for the drug–proteinsystems investigated in this study [65]).
6 0 6 ( 2 0 0 8 ) 119–134
As regards the potential of UV–RRS in protein–drug interac-tions, it should be emphasized that the spectra were recordedat around the 1 mg mL−1 level, low enough to avoid proteinaggregation. The technique should be considered complemen-tary to fluorescence spectroscopy, which is usually dominatedby Trp fluorescence, whereas UV–RRS also provides infor-mation about Tyr residues. Furthermore, selective excitationshould be performed especially in cases where the drugabsorption spectrum extends to longer wavelengths and canbe probed without exciting the protein. Nevertheless, it shouldbe mentioned that such experiments are not straightforward,especially when the compound’s extinction coefficient (ε) isnot very high or in case of a strong fluorescence background.
4. Theoretical aspects of RRS
In RRS different vibrational modes are selectively enhan-ced depending on the excitation wavelength, which meansthat identification is not always straightforward. Vibrationsthat undergo little or no enhancement are likely to becomeinvisible in the spectrum. This effect results in decreasedinformation content of the spectrum compared to normalRaman. In other words, the increase in sensitivity comes atthe price of reduced discrimination power, especially whensimilar molecules are under investigation. Also, conventionalRaman spectra can no longer be used for fingerprint iden-tification, and reference spectra recorded under the sameresonance excitation conditions must be available. On theother hand, the relative simplicity of the RRS spectra mightbe an advantage when the molecular species of interest is in acomplex environment or part of a very large molecule. Other-wise, in conventional RS such situations would give rise to veryovercrowded spectra.
For these reasons, a theoretical model to calculate thepositions and most importantly the relative intensities ofthe resonance-enhanced modes would be extremely valuable.In the case of very similar species, such calculations couldbe used to predict the RRS profiles for different excitationwavelengths and select the excitation conditions that wouldprovide optimal enhancement for those vibrations that showdifferences, thus enabling discrimination.
Contrary to ground state vibrational calculations fornormal Raman or infrared absorption spectroscopy, the cal-culation of RRS profiles is not yet a standard, user-friendlyprocedure. However, a significant development of differenttheoretical approaches has been made over the years.
The importance of a pioneer paper by Hizhnyak and Tehver[66] was not immediately recognized. In 1979, Tonks and Pagefurther developed the Kramer-Kronig relation between thepolarizability tensor and the optical absorption [67]. This ledto calculations of excitation profiles for different vibrationalmodes of molecules like beta-carotene and cyanocobalamin.Blazej and Peticolas carried out similar calculations for pyri-midine nucleotides [68]. Further investigations led to therefinement of the transform theory of RRS by Rush and Peti-
of interest: uracil [69], metalloporphyrins and heme pro-teins [70]. Mroginski et al. [71] used this approach to studylinear tetrapyrroles, constituents of the chromophoric sites
a n a l y t i c a c h i m i c a a c t a 6 0 6 ( 2 0 0 8 ) 119–134 127
Fig. 4 – UV–RRS spectra, illustrating the binding of the drug novobiocin to the bacterial enzyme gyrase. Spectrum (a)corresponds to the free protein and (b) to the free drug. Spectrum (c) is that of the 1:1 complex, and after subtraction of theenzyme spectrum (a) the difference spectrum (d) corresponds to the bound drug. It is not identical to that of the free drug;c 5]).
ottwl
L[fstmg
tg[cR
tctoaR
hanges are indicated with an asterisk (from Couling et al. [6
f various biological photoreceptors. However, the transformheory requires calculation of the exact geometries of not onlyhe ground state but also the electronically excited states,hich is time-consuming and requires extensive expertise in
ight of the many assumptions that need to be made.A time-dependent treatment of the Raman scattering by
ee and Heller [72] and subsequent studies by the same group73] led to the present-day use of time-dependent densityunctional theory (TD-DFT) for optimization of excited-tate geometries. A comparative study of the TD-DFT andhe Hartree–Fock configuration interaction single-type (CIS)
ethods for excited-state optimization was presented by Neu-ebauer and Hess [74].
Moreover, in some cases excited-state geometry optimiza-ion can be simplified by calculating only the excited-stateradients and then approximate the position of the minima72,73,75]. Neugebauer et al. showed how well this approachan predict the observed resonance enhancement factors ofRS spectra of substituted pyrenes excited at 244 nm [76].
Other features of RRS are not yet well predicted by theore-ical approaches: pyrene solutions excited under resonanceonditions show unusually strong overtones and combina-
ion bands, especially when exciting into a vibrational regionf an absorption band, and much less so when exciting inton origin band [77,78]. As an illustration, Fig. 5 shows theRS spectra excited at 229 nm (top; vibronic region of S0–S4
transition), and excited at 257 nm (bottom; vibronic region ofS0–S3 transition). Both spectra show very intense overtonesand combination bands, which are hardly visible when thesame solution is excited at 244 nm (origin of S0–S4 transi-tion; not shown) [78]. The fundamental regions of the twospectra in Fig. 5 are very different, since different vibrationswith other symmetry properties are enhanced at those twowavelengths. The frequencies of the bands extending up to5000 cm−1 agree very well with the possible linear combi-nations of the strongest fundamental peaks. The reason forenhanced overtones upon excitation into a vibronic band iscurrently not yet fully understood, but appears to be related tothe slope of the potential energy surfaces; preresonance withhigher states could also cause a relative suppression of thefundamental bands through interference [79]. So far, overtonestudies were mostly carried out to investigate excited-statedynamics in small molecules [79,80], but we believe that theovertone region may also provide extra analytical selectivityfor the distinction of similar compounds [78].
To summarize, there has been significant progress in thetheory of resonance Raman scattering, which allows us to pre-dict with a reasonable accuracy the RRS profiles of a large
variety of molecular species. However, there are different cal-culation methods available, all of which are based on a numberof assumptions and approximations. Which of those must betaken into account and exactly which theoretical approach
128 a n a l y t i c a c h i m i c a a c t a
Fig. 5 – RRS spectra of pyrene (2.5 × 10−3 M in methanol): (A)excitation at 229 nm (vibronic region of S0–S4 transition)and (B) excitation at 257 nm (vibronic region of S0–S3
transition). Different vibrations are enhanced, dependingon the electronic transition overlapping with the laserwavelength. The spectra show exceptionally strong
overtones and combination bands. *Solvent bands (adaptedfrom Efremov et al. [78]).
should be used in a particular case still requires extensiveexpertise. At the moment this is beyond the capabilities ofmost spectroscopists. Hopefully, in the near future the cal-culation of RRS profiles will become a routine procedure instandard computational software packages.
5. Instrumental and methodologicaldevelopments
5.1. Fiber-optic UV–RRS
The use of fiber optics to design affordable, small-sized andeasy-to-handle instrumentation for conventional RS is wellknown in the analytical chemistry literature [81,82]. Fiberoptics are generally combined with NIR diode laser excitationsources – which are low-cost, while fluorescence background
is minimal – and small, efficient spectrographs in combinationwith CCDs to reduce size and complexity. Such instrumenta-tion has strongly stimulated interest in analytical applicationsof RS: remotely located samples – even in traditionally inac-
6 0 6 ( 2 0 0 8 ) 119–134
cessible environments – can be measured while fiber-opticprobes also minimize sampling and alignment problems. Inthe case of large objects (e.g. a patient, artwork) it is easierto move the probe head across the sample than to move thesample relative to the laser beam. In most fiber-optic probedesigns excitation light is delivered by a central fiber and theRaman signals are collected by one or several separate collec-tion fibers. This way extra filters can be added to remove fiberRaman signals at the exit of the excitation fiber, and reject thelaser line at the entrance of the collection fiber(s).
When designing fiber-optic probes for RRS and/or UV–RRS,several aspects should be carefully considered. First of all,in conventional designs the volume that is most effectivelyprobed is typically several millimeters away from the fibers,depending on the geometry and the fiber acceptance angles.However, in RRS not only the excitation light but also theRaman scattered light is usually within the analyte’s absorp-tion band. In order to avoid extensive transmission losseswith strongly absorbing liquid samples the optical pathlengthshould be very short—less than a few hundred microns. Adesign that offers such short optical pathlengths, with an exci-tation fiber surrounded by six collection fibers, is shown inFig. 6. Secondly, in the case of UV excitation, high-qualityquartz fibers are crucial to assure sufficient throughput ofexcitation light and RRS signal, and to minimize backgroundfrom impurities in the fiber. The fibers should be resistantto solarization: most conventional fiber materials slowly turnopaque under prolonged UV radiation. Thirdly, fiber-opticprobes provide a means to reduce UV-induced photochemis-try in cases where spatial resolution is not critical. Withfiber-optic probes the typical irradiated volumes are muchlarger than under conventional UV–RRS conditions. In the lat-ter case, the spectrometer slit width and thus the samplesize is typically 100–200 �m; in the fiber-optic setup descri-bed by Barbosa et al. [83] an excitation fiber with a 600 �mcore diameter was used, allowing about 16 times higher laserpower without increasing the power density at the sample. Anintracavity frequency-doubled argon-ion laser was used, pro-viding CW output at 248.2 and 257.2 nm. High-quality spectracould be obtained in 10 s. To illustrate the utility of fiber-optic UV–RRS a particularly photosensitive enzyme–substratesystem was successfully studied, i.e. the extradiol dioxyge-nase DHBD (2,3-dihydroxybiphenyl-1,2-dioxygenase) with andwithout the substrate 2,3-dihydroxybiphenyl present.
5.2. RRS for detection in liquid separation techniques
It needs no further clarification that because of its identifica-tion power RS would be an attractive detection method forseparation methods such as liquid chromatography (LC) orcapillary electrophoresis (CE). However, only in recent yearssignificant progress has been made in this field, as reviewed byDijkstra et al. [84]. A main obstacle is the poor analyte detecta-bility of conventional RS, which was recently addressed by thedevelopment of liquid-core waveguide (LCW) detection cells[85,86]. With an LCW the optical pathlength can be increa-
sed up to typically 50 cm, while the internal volume is stillonly 20 �L, compatible with regular-size LC. In reversed-phaseLC–RS, there is an additional problem: most organic modi-fiers of the eluent display very strong Raman bands. That is
a n a l y t i c a c h i m i c a a c t a 6 0 6 ( 2 0 0 8 ) 119–134 129
Fig. 6 – Schematic drawings (side and top view) of right-angle probe geometry for RRS, which minimizes the opticalp nsityB
wtace
lreb(
s(c–ttsstpflesitmatlctbt
au
athlength through an absorbing liquid with high optical dearbosa et al. [83]).
hy Cooper et al. [87] used completely deuterated eluents,hus shifting the solvent Raman bands to lower frequenciesnd reducing interference with the analyte vibrations. Alsohemometrical approaches have been developed to improveluent background subtraction routines [88].
Nonetheless, the most promising development in couplingiquid separations with Raman detection is by making use ofesonance excitation, thus improving the selectivity over theluent background. This is preferably done in the deep UV,ecause with visible excitation the fluorescence background
from a variety of sources) is difficult to overcome [89–91].Our group used LC–UV–RRS at 244 nm (60 mW), with a Z-
haped flow cell in back-scatter mode and acetonitrile/water70/30; v/v) as eluent [92]. For a mixture of polyaromatic hydro-arbons – fluorene, phenanthrene, fluoranthene and pyrenegood RRS spectra were obtained on-the-fly, with very lit-
le fluorescence background. Using routine trace enrichmentechniques the LODs were as low as 50–200 ng mL−1 injected. Ithould be emphasized that for RRS a standard flow cell usuallyuffices; the penetration depth of the excitation light is limi-ed by analyte absorption. Under such conditions the variableenetration depth also means that calibration curves tend toatten at higher levels (as the concentration is increased theffective pathlength becomes shorter and roughly the sameignal intensity is obtained). Therefore, RRS is often used fordentification, and quantitation can be carried out with for ins-ance a UV–vis detector. Furthermore, the Raman emission
ay partially overlap with the absorption spectrum of thenalyte, and corrections have to be made in order to obtainhe proper relative peak intensities [92]. However, for veryow concentrations of absorbing species it can be useful toouple long LCW detector cells to RRS. Tian et al. demons-rated that with such a setup and visible (514 nm) excitationeta-carotene can be detected in aqueous samples down to
he 2.5 × 10−10 M level [93].
In CE, analyte identification with RRS would be even morettractive than in LC, since most often aqueous buffers aresed, without any organic modifiers. In 1988, Chen and Morris
. Inner core diameters are approx. 600 and 400 �m (from
[94] coupled CE and visible RRS (442 nm, 40 mW HeCd laser) todetect the model compounds methyl red and methyl orangeand obtained LODs of 2.5 �M, but little progress has been repor-ted since. Apart from the fluorescence background problem,another disadvantage of visible excited RRS is that only a limi-ted number of analytes absorb in this wavelength range. Chenand Morris [95] overcame this problem in the case of epine-phrine; it was off-line oxidized to adrenochrome, which couldthen be measured by RRS with an LOD of about 10 �M.
To make the technique more widely applicable, we explo-red CE–UV–RRS at 244 and 257 nm, the strongest CW linesof a frequency-doubled argon-ion laser [96]. Fig. 7 showsthe UV-detected electropherogram of a 500 �g mL−1 mix-ture of the four nucleotides TMP, AMP, GMP and UMP. Goodspectra of the first three compounds are depicted; theiridentification limits are in the 1–50 �g mL−1 range. The RRSspectra provided sufficient details to distinguish between thenucleotides.
Extremely high sensitivity can be obtained through thejoint effects of resonance enhancement with surface enhance-ment: SERRS. A comprehensive discussion of SERRS is beyondthe scope of this paper; for recent overviews see [97,98]. SERRSis a heterogeneous technique, requiring the adsorption ofthe analytes to a rough solid metal surface – usually silveror gold – in the form of vacuum deposited metal films oninert nanospheres, etched surfaces or colloidal suspensionsof nanoparticles. In order to achieve surface enhancement,the laser wavelength should be tuned to the surface plas-mon vibrations of the selected metal. For the coupling ofSERRS and separation methods usually aggregated colloidsare applied and the surface plasmon bands are typically inthe visible range for Ag and in the near-infrared for Au. Thislimits the applicability of SERRS to analytes that have chro-mophoric moieties absorbing in those wavelength regions.
However, the group of Smith and Graham demonstrated deri-vatization with specific SERRS labels (that bind strongly to Agand have excellent chromophoric properties) to detect speci-fic DNA sequences with very high sensitivity. Because of the
130 a n a l y t i c a c h i m i c a a c t a 6 0 6 ( 2 0 0 8 ) 119–134
Fig. 7 – Capillary electrophoresis with UV–RRS detection. Top left: electropherogram (conventional UV–vis detection;� = 260 nm) of a mixture of 0.5 mg mL−1 of each TMP, AMP, GMP and UMP, using an injection volume of 23 nL.
sharp lines a number of different labels can be distinguished,more easily than with fluorescent labels [99,100]. Due to lack ofplasmon resonance, UV–SERRS is not possible, certainly not inthe deep UV. As an advantage of SERSS, fluorescence interfe-rence from analyte and/or impurities adsorbed on the metallicsubstrate is efficiently quenched.
Being a heterogeneous technique, SERSS is not readily cou-pled on-line with LC. In particular problems associated withaggregation inside the LC system need to be avoided. Althoughsome encouraging results in this field have been reported[97,101,102], the at-line approach is less cumbersome andallows one to use longer signal accumulation times to improvethe signal-to-noise. In that approach the LC effluent is depo-sited on the surface of a moving TLC plate while the eluent issimultaneously evaporated, so that an immobilized LC chro-matogram is obtained. Next, small amounts of silver colloidare added to the analyte spots and SERRS spectra are recorded[103].
Similarly, also for CE on-line coupling with SERRS has beenrealized [104], but again the at-line approach seems to be morepromising. Since the CE flow rates are very low (nL min−1),it should be possible to deposit the analyte in very smallspots, compatible with Raman microscopy. However, in CE it isessential that during deposition the electrical current be main-tained. Appropriate devices have been developed by Devault
and Sepaniak [105] and He et al [106]. In the latter case thecapillary outlet is metallized to maintain electrical contact andthe effluent is deposited by capillary forces. Recently, a morerugged interface with a stainless steel needle as (grounded)
t), AMP (bottom left) and GMP (bottom right). Excitationfrom Dijkstra et al. [96]).
cathode was developed [107]. The outlet end of the CE capillarywas inserted into this metal needle and CE buffer touching theneedle tip served as electrical contact for the CE separation.Much effort is devoted to finding the most appropriate sub-strate in relation to deposition characteristics, such as etchedsilver foil and vapor-deposited silver film on inert nanospheres[108]. Ruggedness, reproducibility and sensitivity still have tobe further improved.
5.3. Sensitivity at the single-molecule level
Using a highly focused laser beam, efficient collection opticsand appropriate silver or gold particle clusters, SERRS spectracan be recorded at extremely low concentrations in extre-mely low volumes or on extremely small surface areas. Fromthe statistical nature of the signal intensity, single-moleculedetection can be inferred. Enhancement factors of up to 14orders of magnitude have been observed for specific “hots-pots”. Contrary to fluorescence, where the ns lifetime of theexcited state limits the number of excitation events per unittime, higher photon fluxes can be used in a Raman expe-riment without such saturation effects. Under such conditionsstrong anti-Stokes Raman bands can be observed since vibra-tional states become extra populated following a Raman event.Single-molecule detection can be important if one would like
to rule out analyte inhomogeneity. On the other hand, mole-cules outside such a hotspot will remain undetected. Anoverview of single-molecule Raman scattering was recentlypublished [109].
a n a l y t i c a c h i m i c a a c t a 6
Fig. 8 – Tip-enhanced resonance Raman spectra ofdecreasing amounts of MGITC dye on a gold surface,excitation 632.8 nm. The calculated number of probedm 3
(
dseotppama[hi(lc
5
Aici
or the photostability of ink jet prints [119]. The effectivenessof the Kerr gate and reduction in fluorescence (from analyteas well as matrix) is illustrated in Fig. 9. For these studiesthe excitation wavelength could be optimized using a tunable
Fig. 9 – Fluorescence suppression using Kerr gating. Thetop spectrum shows the non-gated spectrum of yellow inkjet dye on paper, dominated by fluorescence. The middlespectrum was obtained with time-gated detection. The
olecules varies from 1.3 × 10 (spectrum A) to 5–6spectrum F) (adapted from Domke et al. [111]).
Lateral resolution as small as 20 nm, well below theiffraction-limited spot size that normally determines thepatial resolution of optical methods, can be achieved by tip-nhanced Raman spectroscopy (TERS, or TERRS if excitationccurs under resonance conditions). This nanoscopic surfaceechnique, in which surface-enhanced Raman and scanningrobe microscopy (SPM) techniques are combined, holds greatromise for the semiconductor and communication industriesnd especially for life sciences. Use is made of a metal (or aetal-coated) tip featuring ‘plasmon resonance’ at the sharp
pex, which results in enhanced and highly localized signals110]. Recently Domke et al. [111] demonstrated that TERRSas almost single-molecule detection capabilities, as shown
n Fig. 8. Decreasing levels of malachite green isothiocyanateMGITC) dye were probed on a smooth gold surface by a HeNeaser (632.8 nm) focused on a gold tip. The lowest spectrum Forresponds to only five to six dye molecules.
.4. Fluorescence rejection by time-gated detection
s mentioned above, in many RRS applications fluorescencenterference is a major obstacle. It may originate from matrixonstituents or solvent impurities, from the compound ofnterest itself, or from the optical components. Avoiding fluo-
0 6 ( 2 0 0 8 ) 119–134 131
rescence by working with very long excitation wavelengthswill usually not be possible, since only very few compoundsabsorb in that region. Since RRS is based on excitation at orclose to the absorption maximum of the molecule of inter-est (see Fig. 1), the probability of fluorescence emission uponrelaxation from the exited state increases substantially. Fur-thermore, when using visible or near-UV excitation sources,fluorescence from matrix components will also increase. Asmentioned above, Asher have shown that in RRS with exci-tation in the deep UV (<250 nm) fluorescence interference isnegligible [7], but at the cost of diminished selectivity.
For these reasons, fluorescence rejection is crucial for thedevelopment of RRS in the visible/near-UV range. Pelletiergave an overview of available options for fluorescence rejec-tion [3]. Chemometric approaches using second derivatives[112] or wavelength shifts [113] have been used in order to dis-tinguish the sharp Raman lines from the broad fluorescencebackground. However, the shot noise associated with thestrong background will not be removed. The most promisingoption is to make use of time discrimination between ins-tantaneous Raman photons and fluorescence that will occurover a few ns time interval. This approach was extensivelytested in the 1980s by Everall and co-workers [114,115], butwith a ns-order time resolution the detection system was notyet fast enough to obtain significant fluorescence rejectionfactors. More recently, researchers at the Rutherford Apple-ton Laboratories (UK) developed a time-gated system basedon picosecond Kerr gating [116]. Several research groups haveused that setup since to carry out RRS studies under stronglyfluorescent conditions, such as the formation of lignin radi-cals in wood cell walls [117], the yellowing of paper pulp [118],
bottom spectrum shows the reference spectrum of theyellow dye in solution, also with gated detection.Excitation = 470 nm; spectra are not on the same scale(adapted from Vikman et al. [119]).
132 a n a l y t i c a c h i m i c a a c t a
Fig. 10 – Fluorescence suppression using an ICCD camera.RRS spectrum of the strongly fluorescent flavoproteinPAMO, without (top) and with (bottom) time-gated detection
r
(adapted from Efremov et al. [122]).
laser system, in order to achieve resonance enhancement ofthe target compounds while minimizing the contribution fromthe complex matrix. In the case of extremely strong fluores-cence (Rhodamine 6G!) one has the option to remove most ofthe remaining fluorescence by means of the shifted excitationapproach [120].
Recently an alternative instrument was developed for time-gated Raman, using an approach that seems to be moreeasily implemented. A commercially available, ultrafast-gatedintensified charge-coupled device (ICCD) camera was applied,with a gate closing speed of about 80 ps [121,122]. Sincethe ICCD can be operated at a repetition rate as high as76 MHz, low pulse energies can be used to conduct the RSexperiments, thus minimizing sample photodegradation. Afrequency-doubled or -tripled Ti:sapphire laser can providefreedom in wavelength selection over most of the UV–visrange.
Efremov et al. used such a laser system with a pulse widthof 3 ps [122]. It should be emphasized that shorter pulsesare not appropriate: the accompanying increase in bandwidthwould strongly affect the spectral resolution. With this setupthe rejection factor is limited by the 80-ps closing time of theICCD gate (considerably slower than the 3-ps Kerr gate), but itcan be operated at much higher repetition rates (with corres-ponding low peak powers). The overall fluorescence rejectionfactor depends of course on the lifetime of the fluorescencebackground; a factor of 50–100 is obtained for compoundswith a lifetime of a few ns. The feasibility of this approachis demonstrated in Fig. 10, showing the effectiveness of ICCDgating on the RRS spectra of phenylacetone monooxygenase, aflavoprotein that gives a strong auto-fluorescence background.In conclusion, we expect that the availability of user-friendly
instrumentation and in particular fluorescence suppressionoptions will stimulate the development of more RRS applica-tions in the 250–600 nm region.
6 0 6 ( 2 0 0 8 ) 119–134
6. Conclusions
It is obvious from this review that significant progress in RRShas been made during the last decade, mainly by dedicatedexperts in Raman spectroscopy, while currently the numberof pure analytical chemistry studies and real-life applicationsis still relatively low. This is in spite of the fact that theanalytical potential of RRS is very high, as should be clearfrom the variety of topics reviewed above, for instance in thebioanalytical field including nucleic acids, proteins and theirinteractions with drugs.
However it can also be concluded that the main hindrancesof transforming RRS into an analytical tool have been over-come during the last decade, paving the way for RRS to enterthe analytical research and application laboratories. First ofall, easy-to-use instrumentation is available to exploit deepUV–RRS, a technique that is very suitable for bioanalyticalchemistry, not hindered by fluorescence background. Its com-bination with fiber optics and also with LC and CE, only quiterecently demonstrated, should be fully exploited. Spectacu-lar sensitivities have been reported, especially under surfaceenhancement conditions. User-friendly models and softwareto predict RRS spectra as a function of excitation wavelengthwill soon become available.
As regards visible RRS, two main obstacles have been domi-nating, i.e. wavelength tunability to fully exploit the inherentselectivity of RRS and fluorescence background, which stron-gly reduces the applicability of RRS. It should be emphasized,however, that in spite of these restrictions there are severalinteresting fields in which visible RRS has been successfullyinvoked using standard CW Kr or Ar ion lasers, such as the ana-lysis of carotenoids in a variety of matrices and the interactionof heme proteins with ligands. In the latter case the 406.7 and413.1 nm lines of the Kr ion laser are in resonance with thestrong Soret absorption band of the porphyrin group, whilefluorescence background of such samples is extraordinarilyweak.
A broad choice of excitation wavelengths must be avai-lable in order to make optimal use of selective excitation.A good example is the selective RRS analysis of mixturesof single-walled carbon nanotubes. Almost complete wave-length tunability is provided by picosecond Ti:sapphire lasers.Their 3 ps pulses are long enough to avoid extra broa-dening of the Raman spectra. A challenge is the furthersuppression of fluorescence background, although time-gated detection with a Kerr gate or a fast-gated intensifiedCCD camera is a major step forward in the developmentof RRS as a technique that is broadly applicable to awide variety of real-life problems. From this review weconclude that the number of analytical chemistry papersdealing with RRS is likely to increase steeply in the decadeahead.
e f e r e n c e s
[1] C.V. Raman, K.S. Krishnan, Nature 121 (1928) 501.[2] R.L. McCreery, Raman Spectroscopy for Chemical Analysis
(Chemical Analysis Volume 157), Wiley, New York, 2000.
[4] M. Harrand, R. Lennuier, Comptes Rendus Academie desSciences 223 (1946) 356.
[5] J. Behringer, J. Brandmuller, Z. Elektrochemie 60 (1956) 643.[6] D. Gill, R.G. Kilponen, L. Rimai, Nature 227 (1970) 743.[7] S.A. Asher, Anal. Chem. 65 (1993), 59A and 201A.[8] L. Rimai, R.G. Kilponen, D. Gill, J. Am. Chem. Soc. 92 (1970)
Khachik, W. Gellermann, L.K. Pershing, J. Investig.Dermatol. 115 (2000) 441.
[29] C. Coupry, A. Lautie, G. Sagon, F. Froment, in: H.G.M.Edwards, J.M. Chalmers (Eds.), Raman Spectroscopy inArcheology and Art History, Royal Society for Chemistry,Cambridge, 2005, p. 207.
[30] S.E.J. Bell, in: H.G.M. Edwards, J.M. Chalmers (Eds.), RamanSpectroscopy in Archeology and Art History, Royal Societyfor Chemistry, Cambridge, 2005, p. 292.
[31] S. Krampelas, E. Fritsch, J.Y. Mevellec, J.P. Gauthier, S.Sklavounos, T. Soldatos, J. Raman Spectrosc. 38 (2007) 217.
[33] G. Massonnet, P. Buzzini, G. Jochem, M. Stauber, T. Coyle, C.Roux, J. Thomas, H. Leijenhorst, Z. van Zanten, K. Wiggins,C. Russell, S. Chabli, A. Rosengarten, J. Forensic Sci. 50(2005) 1028.
[34] A. Jorio, A.P. Santos, H.B. Ribeiro, C. Fantini, M. Souza, J.P.M.Vieira, C.A. Furtado, J. Jiang, R. Saito, L. Balzano, D.E.
Resasco, M.A. Pimenta, Phys. Rev. B 72 (2005), Art. No.075207.
[35] M.S. Dresselhaus, G. Dresselhaus, M.A. Pimenta, P.C.Eklund, in: M.J. Pelletier (Ed.), Analytical Applications of
[36] C. Fantini, A. Jorio, A.P. Santos, V.S.T. Peressinotto, M.A.Pimenta, Chem. Phys. Lett. 439 (2007) 138.
[37] Z.T. Luo, F. Papadimitrakopoulos, S.K. Doorn, Phys. Rev. B 75(2007), Art. No. 205438.
[38] R.C. Lord, G.J. Thomas Jr., Spectrochim. Acta A 23 (1967)2551.
[39] J.M. Benevides, S.A. Overman, G.J. Thomas Jr., J. RamanSpectrosc. 36 (2005) 279.
[40] S.P.A. Fodor, T.G. Spiro, J. Am. Chem. Soc. 108 (1986) 3198.[41] A. Toyama, Y. Takino, H. Takeuchi, I. Harada, J. Am. Chem.
Soc. 115 (1993) 11092.[42] M. Tsuboi, K. Komatsu, J. Hoshi, E. Kawashima, T. Sekine, Y.
Ishido, M.P. Russell, J.M. Benevides, G.J. Thomas Jr., J. Am.Chem. Soc. 119 (1997) 2025.
[43] A. Toyama, Y. Miyagawa, A. Yoshimura, N. Fujimoto, H.Takeuchi, J. Mol. Struct. 598 (2001) 85.
[44] G. Smulevich, A. Feis, J. Phys. Chem. 90 (1986) 6388.[45] C.A. Grygon, T.G. Spiro, Biochemistry 28 (1989) 4397.[46] E. Kocisova, L. Chinsky, P. Miskovsky, J. Biomol. Struct. Dyn.
[54] J.C. Austin, T. Jordan, T.G. Spiro, in: R.J.H. Clark, R.H. Hester(Eds.), Advances in Spectroscopy 20, BiomolecularSpectroscopy, Part A, 1993, pp. 55–127.
[55] R.E. Littleford, D. Cunningham, P. Matousek, M. Torie, A.W.Parker, I. Khan, D. McComb, W.E. Smith, J. RamanSpectrosc. 36 (2005) 600.
[56] I.K. Lednev, V.V. Ermolenkov, W. He, M. Xu, Anal. Bioanal.Chem. 381 (2005) 431.
[57] H. Takeuchi, I. Harada, J. Raman Spectrosc. 21 (1990) 509.[58] T.G. Spiro, R.S. Czernuszewicz, in: K. Sauer (Ed.),
Biochemical Spectroscopy (Methods in Enzymology, vol.246), Academic Press, San Diego, 1995, p. 416.
[96] R.J. Dijkstra, E.V. Efremov, F. Ariese, U.A.Th. Brinkman, C.
Gooijer, Anal. Chem. 75 (2003) 5697.[97] A. Ruperez, J.J. Laserna, in: J.J. Laserna (Ed.), Modern
Techniques in Raman Spectroscopy, Wiley, New York, 1996,p. 227.
6 0 6 ( 2 0 0 8 ) 119–134
[98] K. Kneipp, M. Moskovits, H. Kneipp (Eds.),Surface-enhanced Raman Scattering—Physics andApplications, Springer, Berlin, 2006.
[99] W.E. Smith, K. Faulds, D. Graham, in: K. Kneipp, M.Moskovits, H. Kneipp (Eds.), Surface-enhanced RamanScattering—Physics and Applications, Springer, Berlin,2006, p. 381.
[100] K. Faulds, F. McKenzie, W.E. Smith, D. Graham, Angew.Chem. Int. Ed. 46 (2007) 1829.
