Seeing a single molecule vibrate through time-resolved coherent anti-Stokes Raman scattering

Seeing a single molecule vibrate through time-resolved coherent anti-Stokes Raman scatteringSteven Yampolsky1, Dmitry A. Fishman1, Shirshendu Dey1, Eero Hulkko1,2, Mayukh Banik1,Eric O. Potma1* and Vartkess A. Apkarian1

The motion of chemical bonds within molecules can be observed in real time in the form of vibrational wave packetsprepared and interrogated through ultrafast nonlinear spectroscopy. Such nonlinear optical measurements are commonlyperformed on large ensembles of molecules and, as such, are limited to the extent that ensemble coherence can bemaintained. Here, we describe vibrational wave packet motion on single molecules, recorded through time-resolved, surface-enhanced, coherent anti-Stokes Raman scattering. The sensitivity required to detect the motion of a single molecule underambient conditions is achieved by equipping the molecule with a dipolar nano-antenna (a gold dumbbell). In contrast withmeasurements in ensembles, the vibrational coherence on a single molecule does not undergo pure dephasing. It developsphase fluctuations with characteristic statistics. We present the time evolution of discretely sampled statistical states, andhighlight the unique information content in the characteristic, early-time probability distribution function of the signal.

The demonstration of the optical detection of single moleculesunder cryogenic conditions—by means of absorption1 andsoon afterwards via fluorescence2—can be recognized as the

nascence of single-molecule spectroscopy (SMS). The field wasfurther advanced by demonstrations of SMS under ambient con-ditions using near-field3 and far-field optical microscopy4,5. Themost common approach to SMS is the use of chromophores thatstrongly couple to light to detect light–matter interactions througha variety of methods6, including absorption1,7, resonant Raman8,fluorescence induced by linear2 or nonlinear absorption9, andphotothermal effects10. The principle aim of SMS is to interrogatethe properties of individual molecules rather than averages of thevast ensembles typically probed in molecular spectroscopy.Ensemble averaging eliminates outliers, such as the tail of theBoltzmann distribution where most of chemistry hides.

Incoherent spectroscopy of chromophores mainly addressesenvironmental fluctuations that occur on timescales longer thanthe measurement time11. On timescales of a millisecond andlonger over which signals are accumulated, SMS is well suited totrack biomolecular transformations such as enzymatic activity andprotein folding12,13. However, the femtosecond to picosecond scaleover which bonds move, break and form, poses a greater challenge.An important advance in this regard is the recent demonstration ofa series of ultrafast pump–probe measurements using phase-lockedpulse pairs14–17, in which vibrational wave packets and the coherentmanipulation of quantum bits have been demonstrated on singlemolecules. However, the information content of these measure-ments is limited by electronic dephasing, which occurs on time-scales shorter than vibrational periods of motion. Thischaracteristic of chromophores is implicit in their structurelessand broad absorption spectra.

Raman spectroscopy avoids electronic dephasing consider-ations because it does not require evolution in real electronicstates. Accordingly, time-resolved coherent Raman spectroscopyis ideally suited to capturing the motion of molecules in theirground electronic states. Although Raman is a feeble effect, itcan be greatly amplified using plasmonic antennas18 (via the

surface-enhanced Raman scattering (SERS) effect19,20).Following the initial demonstrations of single-molecule sensi-tivity21,22, single-molecule SERS (single molecule (SM)-SERS)has developed rapidly23, with submolecular spatial resolutionrecently achieved at the plasmonic junction of a scanning tunnel-ling microscope24.

The sensitivity of SM-SERS derives from the enhanced localfields concentrated at the junctions of metallic nanostructures.It is not clear whether the principles of SERS can be directly trans-lated to the ultrafast time domain. Molecules subject to fieldgradients may migrate due to optical forces and, at the incidentintensities of 1 × 1010 W cm−2 reached with ultrashort laserpulses, nanoparticles melt, change shape and even fragment25.Surface-enhanced coherent anti-Stokes Raman scattering26–29

and stimulated Raman scattering30 have been demonstratedusing femtosecond lasers in frequency-domain measurementson SERS-active ensembles. Although a time-resolved, surface-enhanced Raman scattering measurement has been reported ina colloidal ensemble, the unique capabilities of coherent Ramantechniques to capture the evolution of vibrational wavepackets31,32 have yet to be demonstrated in the single-moleculelimit. Here, we report this realization.

Through time-resolved coherent anti-Stokes Raman scattering(tr-CARS) studies on individual nano-dumbbells, we capture thevibrational motion of single molecules in real time and highlightthe unique information content of time-resolved measurements inthe single-molecule limit. Rather than pure dephasing, coherencesdevelop phase noise due to discrete sampling statistics, as is familiarin quantum optics33. Most notably, we show that the probability dis-tribution function (PDF) of the noise accumulated during early timeevolution can uniquely distinguish between one, two and many-molecule responses.

Our experiments were carried out on trans-1,2-bis-(4-pyridyl)ethylene (BPE) molecules attached to gold nano-dumbbellsencapsulated in porous silica shells (Fig. 1a). Single-molecule sensi-tivity with SERS at the junction between two gold nanosphereshas been documented in the literature for this and related

1Department of Chemistry, University of California, Irvine, California 92697, USA, 2Department of Chemistry, PO Box 35, FI-40014, University of Jyväskylä,Finland. *e-mail: [email protected]

ARTICLESPUBLISHED ONLINE: 6 JULY 2014 | DOI: 10.1038/NPHOTON.2014.143

NATURE PHOTONICS | VOL 8 | AUGUST 2014 | www.nature.com/naturephotonics650

© 2014 Macmillan Publishers Limited. All rights reserved.

mailto:[email protected]

http://www.nature.com/doifinder/10.1038/nphoton.2014.143

http://www.nature.com/naturephotonics

systems34–36, including an isotopologue test of single-moleculeresponse37,38. In the current experiment, we do not have directknowledge of the location or coverage of the molecules, whichwere adsorbed on the gold spheres before encapsulation. However,signatures of single-molecule behaviour are evident in the spectralfluctuations observed in sequentially recorded SERS spectra on indi-vidual nanostructures. We see meandering Raman trajectories, dra-matic intensity fluctuations and the appearance of lines that cannotbe explained by means of the dipolar Raman response alone (seeSupplementary Section III). These observations parallel an earlierstudy on silver nano-dumbbells, where it was shown that thetensor nature of Raman scattering is in full force in SM-SERS andthat spectral variations accompany changes in the relative orien-tation between the molecule and the local field, and the latterevolve due to structural changes of the irradiated junction39.

The single-particle spectrum shown in Fig. 1c and magnified inFig. 1d, is from a relatively stationary SERS trajectory, obtained atan incident intensity of 30 µW µm−2. Remarkably, the SERS line-widths are significantly broader than those in the ensemble spec-trum of bulk BPE. We will focus on the strong C=C stretchingmodes near 1,600 cm−1, which arise from mixing between the ethy-linic stretch and the pyridine ring breathing modes of the BPE40.These lines have Lorentzian profiles with a full-width at half-maximum (FWHM) of 5 cm−1 in bulk BPE, but appear asGaussians with an FWHM of ∼15 cm−1 for the dumbbell(Fig. 1d). SERS trajectories recorded at a rate of one spectrumevery 10 s show resolution-limited lines of FWHM= 4 cm−1

inside the envelope (see Supplementary Section II). Consistentwith recent tip-enhanced Raman measurements at cryogenic temp-eratures, it is clear that intrinsic SERS linewidths (limited byvibrational dissipation) are much sharper41. Evidently, the broad

Gaussian envelopes represent inhomogeneous distributionssampled by spectral diffusion, and both optical forcesand local heating can be expected to contribute to the observedspectral distribution. The latter effect appears to be greatlyreduced under the pulsed irradiation conditions of ultrafasttr-CARS measurements.

The tr-CARS measurements were carried out on dropcast par-ticles on a 15-nm-thick silicon nitride substrate. After scanning elec-tron microscopy (SEM) mapping and SERS characterization, theparticles were transferred to a femtosecond laser scanning CARSmicroscope. The imaging system was interfaced with a tunable fem-tosecond laser (80 MHz repetition rate), which provided the pump(λpu = 714 nm), Stokes (λSt = 809 nm) and probe (λpr = 714 nm)pulses for the electronically non-resonant CARS process illustratedin Fig. 2a. The difference frequency between the coincident pumpand Stokes pulses, ωpu − ωSt = 1,640 cm−1, was selected to excitethe superposition of bright modes that fall under the spectral band-width (Fig. 2a), to prepare the second-order wave packetf(2) =

∑n an n| ⟩, where av is the coefficient of eigenstate n| ⟩. The

evolving vibrational coherence f(2)(t)∣∣ ⟩⟨f(0)(t)

∣∣ was then interro-gated with the time-delayed probe pulse, and detected through theanti-Stokes shifted photons. Care was taken to keep the average illu-mination intensity below 20 µW µm−2 to avoid photodamage.

A CARS image of a distribution of particles is presented inFig. 2b. The nanoparticles are readily imaged through the electronicCARS response of the metal surface plasmon42–44. The variation intheir intensities is readily associated with the distribution of orien-tations relative to the linear polarization of the excitation. The mol-ecular CARS response is distinguished by its dependence on thedelay of the probe pulse and vibrational resonance (Fig. 2c,d).Beyond 0.2 ps, the instantaneous electronic response of the

1,000 1,250 1,500 1,750 1,600 1,620 1,640 1,660Raman shift (cm−1) Raman shift (cm−1)

Inte

nsity

(a.u

.)

Inte

nsity

(a.u

.)

SERS

Bulk

DFT

SERS

Bulk

N

N

Gold

Silica100 nm

a b

c d

Figure 1 | SERS of BPE molecules on dumbbell antennas. a, Sketch of the dumbbell-shaped gold nanostructures with adhered BPE molecules, encapsulatedin a silica shell. A strong SERS response is expected from BPE molecules in the hotspot between the gold nanospheres. b, TEM image of a single dumbbellstructure. c, Comparison of a SERS spectrum from a single dumbbell, a Raman spectrum from bulk and DFT calculation of the Raman spectrum of BPE.d, Comparison of SERS spectrum and bulk Raman spectrum in the vicinity of 1,600 cm−1.

NATURE PHOTONICS DOI: 10.1038/NPHOTON.2014.143 ARTICLES

NATURE PHOTONICS | VOL 8 | AUGUST 2014 | www.nature.com/naturephotonics 651




plasmon reduces to a constant background over which themodulated molecular contribution appears. The spectrallyresolved tr-CARS signals illustrate that when the differencefrequency, ωpu − ωSt = 1,800 cm−1, is detuned from the molecularresonance, only the constant background of two-beam electronicCARS remains. On resonance, a periodically modulated signal isseen. The period, τ≈ 1 ps, corresponds to the beat between thepair of bright modes separated by ∼35 cm−1 (Fig. 1d). In the casepresented in Fig. 2c, the signal suddenly drops. The particlebecomes SERS-inactive, while its transmission electron microscopy(TEM) image shows an intact dumbbell (Fig. 2b, inset). Themeasurement sequence suggests that the loss of signal is due tothe separation between the molecule and the hot spot. Thiscommon occurrence limits the tolerable exposure time—and there-fore the sampling statistics—of the measurements.

Significant variation is seen in the tr-CARS signals recorded onindividual dumbbells (see Fig. 3a for examples). When all threepulses coincide, at τ = 0, dramatic interferometric fluctuations areobserved, which are blanked out in the figure. At negative delay,when the probe precedes the preparation pulses, the time-indepen-dent two-beam electronic CARS signal of the metal antenna is seen.The fluctuations at negative times establish the noise floor overwhich the molecular response appears upon passing τ = 0 (Fig. 3a,orange and grey curves). No identifiable molecular response isseen from the SERS-inactive structure (Fig. 3a, green curve).Modulation with identifiable periodicity is only seen at positivetimes and only on SERS-active particles. The CARS signal showsidentical features when simultaneously detected in the forward(grey curve) and backward (orange curve) scattering directions,indicating that the oscillation of the signal is an intrinsic responseof the dumbbell–molecule system, independent of detector noise.These signals were acquired with incident peak intensities that

reach ∼1 × 1010 W cm−2. Under these irradiation conditions, wefind surprisingly little change to the dumbbell. Detailed TEMimaging reveals that pre-irradiated dumbbells show a clear junctiongap of <2 nm between nanospheres (Fig. 3b), while post-irradiationimages show the formation of a neck between the spheres (Fig. 3c,d).Although it is unclear at what stage of the illumination this morpho-logical change occurs, the dumbbells remain largely intact and theCARS signal remains remarkably stable during the 1 h duration ofthe measurement.

Examples of time traces that are characteristic of tr-CARS in thesingle BPE molecule limit are shown in Fig. 4, together with thesignal of the ensemble of solid BPE recorded under identical con-ditions. The contrast is striking. The ensemble signal decays in 1 ps(Fig. 4a), while the depth of modulation of the single dumbbellsignals persists for the 10 ps duration of the measurement (Fig. 4b,c).The ensemble signal S, given by the third-order polarization P (ref. 31),

S(τ) = ∫dt ⟨P(3)(τ)⟩∣∣∣

∣∣∣2

(1)

reduces to the damped quantum beats of the prepared vibrationalcoherence:

S(τ) ∝∑

n,n′anan′cos(ωn − ωn′ )e

−(γn+γn′ )τ (2)

with phenomenological dephasing rates γv,v' that arise from the ensem-ble averaging indicated by the angle brackets in equation (1). The sameinformation is contained in the Raman spectrum of the bulk (Fig. 4a,inset), which in the absence of electronic resonances is given by thesame third-order response31. Consequently, the tr-CARS signal canbe retrieved by the windowed Fourier transform of the Raman

3 µm

b

(c)

1 µm 100 nm

CARS TEMtpu = tSt τ

Time

tpu tSt tpr

CA

RS s

igna

l

a

Time (fs)

1,000

6,000 615

650

Time (fs)

1,000

6,000 610

645

On resonance Off resonance

Off

On

Off

0

Wavelength (nm)

Wavelength (nm)

Freq

uenc

y

c

On

Figure 2 | tr-CARS microscopy of individual molecule-dumbbells structures. a, Jablonski diagram of the tr-CARS process, with the time delay τ between thepump/Stokes pair and the probe pulse indicated. b, CARS image of isolated dumbbell structures. Inset: CARS image of a single dumbbell and thecorresponding TEM image. c, Spectrally resolved tr-CARS of a single dumbbell acquired on resonance at ωpu −ωSt = 1,640 cm−1 (left) and acquired off-resonance at ωpu −ωSt = 1,800 cm−1 (right). Note that the clear quantum beats disappear when tuned off-resonance.

ARTICLES NATURE PHOTONICS DOI: 10.1038/NPHOTON.2014.143





spectrum, as shown in Fig. 4a. The broad Lorentzian profile of theRaman lines leads to rapid exponential dephasing of the signal intime. This is clearly not the case for signals recorded on the dumbbell.Given the much broader SERS spectrum (Fig. 4a, inset), the CARSsignal would have been expected to decay even faster than the ensem-ble, on a 0.3 ps timescale. Instead, the depth of modulation of the tr-CARS signal from the dumbbell (Fig. 4b) persists for the duration ofthe measured time delay τ = 10 ps. A vibrational superposition offour states, given by the stick spectrum in Fig. 4a, emulates the signalwith good fidelity. The main beat is that of the two breathingmodes separated by 35 cm−1 (ω = 1,647 cm−1, 1,612 cm−1).The higher frequency beat is reproduced by the inclusion of lines atω = 1,580 cm−1, which are seen in the transient SERS spectrum fromthe same dumbbell (Fig. 4a, inset), and the splitting between thetwo modes of 5 cm−1 reproduces the slow modulation of the signalenvelope. The main features of the signal are satisfactorily repro-duced. Although other interpretations may be given, what is clearis that there is no evidence of dephasing. If more than one moleculecontributes to the signal, then their collective spectral distributionmust remain within ∼0.3 cm−1 for the 1 h duration of the measure-ment. This seems unlikely. The signal is best described as tr-CARS inthe single-molecule limit.

To be more definitive about the single-molecule origin of a signalit is necessary to track the wave packet motion until the onset ofquantifiable noise. An example is provided by the tr-CARS signalof Fig. 4c. Note that each experimental point of the time trace isthe average of 50 measurements, taken with ∼2 µs dwell time per

pixel at 1 s intervals, and we estimate a total of 5–10 signalphotons detected per point. The total accumulation time is ∼1 hfor the full trace. Spectral fluctuations during the measurementintervals generate phase noise. Thus, the discretely sampled normal-ized signal at a given delay time τ can be described with completegenerality as

S(τ)n,m =1M

∑M

m=1

1N

∑N

n=1

1V

∑V

n=1

ane−i(⟨ωn⟩+δn,m,n)τ

∣∣∣∣∣

∣∣∣∣∣

2

(3)

whereM is the number of realized measurements (photons),N is thenumber of molecules, V is the number of vibrational modes, ⟨ωn⟩ isthe mean wavenumber of vibrational mode v, and δv is its stochasticfluctuation around the mean. As suggested by the Gaussian profilesof SERS lines (Fig. 1), δv derives from a normal distributionN (0, σ)with zero mean and variance σ given by the spectral width of thevibrational lines. The phase fluctuation grows with delay time τ,such that at στ ≫ 2π the random phase δτ is uniformly distributedover the full [−π, π] interval. In this limit, the probability distri-bution of observable signal amplitudes reduces to a normaldistribution, with a characteristic mean of

⟨S(τ > 2π/σ)⟩ =1NV

(4)

This is the summation over the coherence terms inside the square in

100 nm 100 nm

100 nm

ab

dc

4,000−4,000 0

Time delay (fs)

Inte

nsity

(a.u

.)

Figure 3 | Stable CARS under minimum photodamaging conditions. a, tr-CARS of individual dumbbell structures, including a SERS-active structure (orangeand grey curves) and a SERS-inactive structure (green curve). The orange curve is the back-scattered CARS and the grey curve is the simultaneouslymeasured forward-scattered CARS. b, TEM of a dumbbell before optical illumination, showing a clear gap between the two spheres. c,d, Examples of TEMimages of dumbbells after SERS optical experiments, showing the formation of a neck between the spheres. Neck formation is in part induced by the electronbeam during the recording of high-resolution TEM images.






equation (3), overNmolecules and Vmodes, which leads to dephas-ing ⟨S(τ)⟩ �0 for large N and V. For a single molecule and asuperposition of two states, the expected value is given by

⟨S(τ > π/σ)⟩ =12

for N = 1,V = 2 (5)

Note that, in a given measurement, the stochastic phase leads to arandom value of the normalized sinusoidal signal S(τ)∈ [1, 0],with mean ⟨S(τ)⟩ = 1/2 and variance 1/

��M

√given by averaging

over theM realizations. This assumes no correlation in spectral fluc-tuations δv between different modes. In the other extreme, assumingfully correlated fluctuations, the signal retains full coherence. Thiscan be clearly seen by considering the superposition of two states

on a single molecule:

S(τ) =12+

12M

∑M

m=1

cos(⟨ωn − ωn′⟩τ + (δn − δn′)mτ) (6)

If in each realizationm, the fluctuation is correlated, δv− δv' = 0, andthe signal does not fluctuate.

More generally, in an open system, spectral fluctuations are sto-chastic. The expected signal for a single quantum beat (V = 2), on asingle molecule (N = 1) and M = 5 realizations sampled from anormal distribution (δn − δn′ ) ∈ N (0, σ = 1.5 cm−1) is illustratedin Fig. 4c (green curve). The assumption of two molecules producesa different trace at long delays, because in this case⟨S(τ > π/σ)⟩ � 1/4 (Fig. 4d). Despite the stochastic nature of the

Inte

nsity

(a.u

.)In

tens

ity (a

.u.)

Inte

nsity

(a.u

.)In

tens

ity (a

.u.)

Time delay (ps)

2 4 6 8 100

Experiment Simulation

One molecule Two molecules

a

b

c

d

Occurrence

Norm

alized intensity (a.u.)N

ormalized intensity (a.u.)

Norm

alized intensity (a.u.)N

ormalized intensity (a.u.)

0.0

0.5

1.0

0.0

0.5

1.0

0.0

0.5

1.0

0.0

0.5

1.0

0.43

0.2

0.46

1: 0.47

2: 0.24

Time delay (ps)

2 4 6 8 100

Time delay (ps)

2 4 6 8 100

Time delay (ps)

2 4 6 8 100

Experiment Simulation

1,550 1,600 1,650

Wavenumber (cm−1)

Figure 4 | tr-CARS traces in the single-molecule limit. a, tr-CARS of bulk BPE (brown curve) together with the windowed Fourier transform of the bulkRaman spectrum (grey curve). Inset: bulk Raman spectrum (pink) and a transient SERS spectrum recorded on the same dumbbell on which the tr-CARStrace (b) was obtained. b, tr-CARS trace of a single structure (brown curve) shows distinct quantum beats. The grey curve is a simulation of the tr-CARSresponse based on the stick spectrum in the inset of a. c, tr-CARS of another isolated nanostructure (brown curve) showing periodic oscillations at an earlytime followed by phase noise at longer time. The grey curve is a simulation based on equation (6) for a single molecule with M = 5 and σ = 1.5 cm−1.d, Comparison between one-molecule (green) and two-molecule (blue) simulated signals. The right panels show the PDFs and their first moments derivedfrom the corresponding temporal traces in the panels to the left.






trajectories, it is possible to assign the prepared superposition to thevibrational coherence of the two states that dominate the ensembleresponse, albeit with a spectral distribution that is significantly nar-rower than the observed SERS linewidths. We learn that there is sig-nificant correlation between the two C=C modes. This highlights animportant difference between CARS and Raman that becomes pro-nounced in the single-molecule limit. Through the squared third-order polarization in equation (1), CARS measures the cross-corre-lation of fluctuations between modes (equation (6)), while linewidthsin Raman spectra measure the dephasing of individual modes.

Stochastic trajectories can be rigorously compared through theirstatistics. Unique assignments are possible through the PDFs,namely the histograms of the time traces shown in the rightpanels of Fig. 4. In the evolution interval 0 < τ < 2π/σ, the observabledistributions and the expected values in signal amplitude are oscil-latory functions that obey phasor statistics45. The characteristicPDFs can be generated numerically, as shown in Fig. 5. By inspec-tion, the histogram of the signal (of the time trace in Fig. 4c) can beuniquely assigned to a single vibrational beat of a single molecule.Moreover, based on the indicated first two moments of the PDFs,the spectral covariance, σ = 1.5 cm−1, can be directly bracketed.The comparison between the experimental PDF and the numericalcharacteristic distributions can be quantified through the non-para-metric Kolmogorov–Smirnov (KS) test46. Based on the KS distancebetween cumulative distributions, it is possible to establish thatthe PDF (N = 1, V = 2, σ = 1.5 cm−1) is the best match to theexperiment, and that the experiment obeys the single-molecule,single-beat statistics with 99% likelihood. More generally, thefigure illustrates that the information content in time-dependentmeasurements of coherences in the single-molecule limit is con-tained in the correlated probability densities of the phase and ampli-tude variations. Although equation (3) yields to rigorous statisticalanalysis, the principle of key relevance to the present is that asingle molecule cannot decohere but develops characteristic phasenoise. In contrast, multiple molecules in a multimode coherencerapidly dephase (Fig. 5). The early-time PDF of the signal yields astatistically meaningful distinction between one, two and many

molecules when a small number of states are prepared and interro-gated. Experimentally, we observe all possible variations: sometraces reflect single-molecule behaviour, while others indicate thepresence of a few to many molecules. Examples are shown in theSupplementary Section IV, together with their KS analysis.

In summary, through the persistent vibrational coherence on asingle nano-antenna (Fig. 4b,c) we have presented the first real-time record of vibrational wave packet motion at the single-moleculelimit, under ambient conditions. To reach single-molecule sensi-tivity, we employed the surface-enhanced Raman effect in thetime domain, at the plasmonic junction of gold nanospheres. Wehave demonstrated the viability of antennaed molecules under ultra-fast irradiation, and have introduced time-resolved, surface-enhanced CARS as a tool to interrogate molecular motions in thesingle-molecule limit. These findings are most directly illustratedin the Supplementary Movie of the breathing of the antennaed mol-ecule. We should note that, being a four-wave mixing (FWM)process, CARS belongs to one of the more flexible multidimensionalnonlinear optical schemes31. It has the attributes of Raman scatter-ing in its generality, and under SERS conditions, dramatic fieldenhancements (E8) are to be expected. As a coherent scatteringprocess, tr-CARS allows the selective preparation and interrogationof superposition states, with time resolution limited by the laserpulses. Complete quantum state reconstruction is possible throughPDFs in electronically resonant FWM47 and it has previously beenrecognized that quantum logic can be mapped on the preparation,evolution and measurement steps of FWM48–50. The manipulationof truly quantum information becomes accessible in the single-mol-ecule limit. The present experimental demonstration of coherentwave packets sustained on single molecules paves the way towardssuch applications.

Received 2 November 2013; accepted 2 June 2014;published online 6 July 2014

References1. Moerner, W. E. & Kador, L. Optical detection and spectroscopy of single

molecules in a solid. Phys. Rev. Lett. 62, 2535–2538 (1989).

V = 2

V = 3

1

2

3

1

1.00.5

0.75 cm−1 1.50 cm−1 3.00 cm−1σ

0.500.067

0.500.046

0.500.035

0.400.080

0.330.056

0.290.042

0.370.083

0.270.059

0.220.042

0.340.025

0.340.032

0.330.041

d = 0.157

N

0.0

0.460.050

1.01.0 0.50.5 0.00.01.00.50.01.00.50.0

1.00.50.01.00.50.01.00.50.0

Figure 5 | PDFs calculated for different realizations of the number of molecules N, number of modes V and variance of the spectral fluctuations σ. Thefirst two moments of the distributions are indicated in each panel. The panel to the right gives the PDF for the tr-CARS trace shown in Fig. 4c. Theexperimental PDF (brown bars) and its two moments uniquely match the simulated PDF for N = 1, V = 2 and σ= 1.50 cm−1 (blue bars). The experimentalcumulative distribution function (CDF; brown dots, top right) and the theoretical CDF for the simulated PDF (blue line) are also indicated. The KS test revealsa maximum distance of d =0.157 between the experimental and theoretical CDFs, translating into a 99% likelihood that the data represent the evolution of astatistical two-state superposition on a single molecule.






2. Orrit, M. & Bernard, J. Single pentacene molecules detected by fluorescenceexcitation in a p-terphenyl crystal. Phys. Rev. Lett. 65, 2716–2719 (1990).

3. Betzig, E. & Chichester, R. J. Single molecules observed by near-fieldscanning optical microscopy. Science 262, 1422–1425 (1993).

4. Nie, S., Chiu, D. T. & Zare, R. N. Probing individual molecules with confocalfluorescence microscopy. Science 266, 1018–1021 (1994).

5. Eigen, M. & Rigler, R. Sorting single molecules: application to diagnostics andevolutionary biotechnology. Proc. Natl Acad. Sci. USA 91, 5740–5747 (1994).

6. Kulzer, F. & Orrit, M. Single molecule optics. Ann. Rev. Phys. Chem. 55,585–611 (2004).

7. Kukura, P., Celebrano, M., Renn, A. & Sandoghdar, V. Single-moleculesensitivity in optical absorption at room temperature. J. Phys. Chem. Lett. 1,3323–3327 (2010).

8. Moerner, W. E. A dozen years of single-molecule spectroscopy in physics,chemistry, and biophysics. J. Phys. Chem. B 106, 910–927 (2002).

9. Plakhotnik, T., Walser, D., Pirotta, M., Renn, A. & Wild, U. P. Nonlinearspectroscopy on a single quantum system: two-photon absorption of a singlemolecule. Science 271, 1703–1705 (1996).

10. Gaiduk, A., Yorulmaz, M., Ruijgrok, P. V. & Orrit, M. Room-temperaturedetection of a single molecule's absorption by photothermal contrast.Science 330, 353–356 (2010).

11. Barkai, E., Jung, Y. & Silbey, R. Theory of single-molecule spectroscopy:beyond the ensemble average. Ann. Rev. Phys. Chem. 55, 457–507 (2004).

12. Weiss, S. Fluorescence spectroscopy of single biomolecules. Science 283,1676–1683 (1999).

13. Xie, S. Single-molecule approach to enzymology. Single Molecules 2,229–236 (2001).

14. Hildner, R., Brinks, D. & van Hulst, N. F. Femtosecond coherence andquantum control of single molecules at room temperature. Nature Phys. 7,172–177 (2011).

15. Brinks, D., Hildner, R., Stefani, F. D. & van Hulst, N. F. Coherent control ofsingle molecules at room temperature. Faraday Disc. 153, 51–60 (2011).

16. Brinks, D. et al. Visualizing and controlling vibrational wave packets ofsingle molecules. Nature 465, 905–908 (2010).

17. Hildner, R., Brinks, D., Stefani, F. D. & van Hulst, N. F. Electronic coherencesand vibrational wave-packets in single molecules studied with femtosecondphase-controlled spectroscopy. Phys. Chem. Chem. Phys. 13, 1888–1894 (2011).

18. Novotny, L. & van Hulst, N. Antennas for light. Nature Photon. 5, 83–90 (2011).19. Jeanmaire, D. L. & Van Duyne, R. P. Surface Raman spectroelectrochemistry:

part I. Heterocyclic, aromatic, and aliphatic amines adsorbed on theanodized silver electrode. J. Electroanal. Chem. 84, 1–20 (1977).

20. Schatz, G. C. Theoretical-studies of surface enhanced Raman scattering. Acc.Chem. Res. 17, 370–376 (1984).

21. Nie, S. & Emory, S. R. Probing single molecules and single nanoparticles bysurface-enhanced Raman scattering. Science 275, 1102–1106 (1997).

22. Kneipp, K. et al. Single molecule detection using surface-enhanced Ramanscattering (SERS). Phys. Rev. Lett. 78, 1667–1670 (1997).

23. Etchegoin, P. G. & Le Ru, E. C. A perspective on single molecule SERS: currentstatus and future challenges. Phys. Chem. Chem. Phys. 10, 6079–6089 (2008).

24. Zhang, R. et al. Chemical mapping of a single molecule by plasmon-enhancedRaman scattering. Nature 498, 82–86 (2013).

25. Link, S., Burda, C., Nikoobakht, B. & El-Sayed, M. A. Laser-induced shapechanges of colloidal gold nanorods using femtosecond and nanosecond laserpulses. J. Phys. Chem. B 104, 6152–6163 (2000).

26. Voronine, D. V. et al. Time-resolved surface-enhanced coherent sensing ofnanoscale molecular complexes. Sci. Rep. 2, 891 (2012).

27. Steuwe, C., Kaminski, C. F., Baumberg, J. J. & Mahajan, S. Surface enhancedcoherent anti-Stokes Raman scattering on nanostructured gold surfaces.Nano Lett. 11, 5339–5343 (2011).

28. Namboodiri, V. et al. Surface-enhanced femtosecond CARS spectroscopy(SE-CARS) on pyridine. Vib. Spectrosc. 56, 9–12 (2011).

29. Ichimura, T., Hayazawa, N., Hashimoto, M., Inouye, Y. & Kawata, S. Tip-enhanced coherent anti-Stokes Raman scattering for vibrational nanoimaging.Phys. Rev. Lett. 92, 220801 (2004).

30. Frontiera, R. R., Henry, A.-I., Gruenke, N. L. & Van Duyne, R. P. Surface-enhanced femtosecond stimulated Raman spectroscopy. J. Phys. Chem. Lett.2, 1199–1203 (2011).

31. Mukamel, S. Principles of Nonlinear Optical Spectroscopy (Oxford Univ.Press, 1995).

32. Karavitis, M., Zadoyan, R. & Apkarian, V. A. Time resolved coherent anti-StokesRaman scattering of I2 isolated in matrix argon: vibrational dynamics on theground electronic state. J. Chem. Phys. 114, 4131–4140 (2001).

33. Loudon, R. The Quantum Theory of Light (Oxford Univ. Press, 2000).

34. Kleinman, S. L. et al. Structure enhancement factor relationships in single goldnanoantennas by surface-enhanced Raman excitation spectroscopy. J. Am.Chem. Soc. 135, 301–308 (2013).

35. Wustholz, K. L. et al. Structure–activity relationships in gold nanoparticledimers and trimers for surface-enhanced Raman spectroscopy. J. Am. Chem.Soc. 132, 10903–10910 (2010).

36. Whitmore, D. D. et al. High sensitivity surface-enhanced Raman scattering insolution using engineered silver nanosphere dimers. J. Phys. Chem. C 115,15900–15907 (2011).

37. Dieringer, J. A., Lettan, R. B., Scheidt, K. A. & Van Duyne, R. P. A frequencydomain existence proof of single-molecule surface-enhanced Ramanspectroscopy. J. Am. Chem. Soc. 129, 16249–16256 (2007).

38. Wang, D., Zhu, W., Best, M. D., Camden, J. P. & Crozier, K. B. DirectionalRaman scattering from single molecules in the feed gaps of optical antennas.Nano Lett. 13, 2194–2198 (2013).

39. Banik, M. et al. Surface-enhanced Raman trajectories on a nano-dumbbell:transition from field to charge transfer plasmons as the spheres fuse. ACS Nano6, 10343–10354 (2012).

40. Yang, W.-H., Hulteen, J., Schatz, G. C. & Van Duyne, R. P. A surface-enhancedhyper-Raman and surface-enhanced Raman scattering study of trans-1,2-bis(4-pyridyl)ethylene adsorbed onto silver film over nanosphere electrodes.Vibrational assignments: experiment and theory. J. Chem. Phys. 104,4313–4323 (1996).

41. Klingsporn, J. M., Sonntag, M. D., Seideman, T. & Van Duyne, R. P.Intramolecular insight into adsorbate-substrate interactions via low-temperature, ultrahigh-vacuum tip-enhanced Raman spectroscopy. J. Am.Chem. Soc. 136, 3881–3887 (2014).

42. Danckwerts, M. & Novotny, L. Optical frequency mixing at coupled goldnanoparticles. Phys. Rev. Lett. 98, 026104 (2007).

43. Wang, Y., Lin, C.-Y., Nikolaenko, A., Raghunathan, V. & Potma, E. O. Four-wavemixing microscopy of nanostructures. Adv. Opt. Photon. 3, 1–52 (2011).

44. Zhang, Y., Wen, F., Zhen, Y.-R., Nordlander, P. & Halas, N. J. Coherent Fanoresonances in a plasmonic nanocluster enhance optical four-wave mixing.Proc. Natl Acad. Sci. USA 110, 9215–9219 (2013).

45. Goodman, J. W. Speckle Phenomena in Optics: Theory and Applications.(Roberts & Co., 2007).

46. Young, I. T. Proof without prejudice: use of the Kolmogorov–Smirnov test forthe analysis of histograms from flow systems and other sources. J. Histochem.Cytochem. 25, 935–941 (1977).

47. Golschleger, I. U., van Staveren, M. N. & Apkarian, V. A. Quantum tomographyof a molecular bond in ice. J. Chem. Phys. 139, 034201 (2013).

48. Zadoyan, R., Kohen, D., Lidar, D. A. & Apkarian, V. A. The manipulation ofmassive ro-vibronic superpositions using time-frequency-resolved coherentanti-Stokes Raman scattering (TFRCARS): from quantum control toquantum computing. Chem. Phys. 266, 323–351 (2001).

49. Bihary, Z., Glenn, D. R., Lidar, D. A. & Apkarian, V. A. An implementation ofthe Deutsch–Jozsa algorithm on molecular vibronic coherences through four-wave mixing: a theoretical study. Chem. Phys. Lett. 360, 459–465 (2002).

50. Glenn, D. R., Lidar, D. A. & Apkarian, V. A. Quantum logic gates in iodine vaporusing time–frequency resolved coherent anti-Stokes Raman scattering: atheoretical study. Mol. Phys. 104, 1249–1266 (2006).

AcknowledgementsThe authors thank R. P. Van Duyne for providing the samples, P. Z. El-Khoury for DFTcalculations for BPE andN. Apkarian for pointing out the KS analysis. SEM and TEMworkwas performed at the Laboratory for Electron and X-ray Instrumentation (LEXI) at UCIrvine. This work was made possible by the National Science Foundation Center forChemical Innovation on Chemistry at the Space–Time Limit (grant CHE-0802913). E.H. issupported by the Academy of Finland Decision no. 265502.

Author contributionsS.Y. and D.A.F. conducted the tr-CARS measurements. M.B., E.H. and S.D. carried out theRaman and SEM/TEM measurements. S.Y. and V.A.A. performed the analysis. E.O.P. andV.A.A. conceived the idea for the work and wrote the manuscript.

Additional informationSupplementary information is available in the online version of the paper. Reprints andpermissions information is available online at www.nature.com/reprints. Correspondence andrequests for materials should be addressed to E.O.P.

Competing financial interestsThe authors declare no competing financial interests.





http://www.nature.com/reprints

mailto:[email protected]



Seeing a single molecule vibrate through time-resolved coherent anti-Stokes Raman scattering

Documents

Transcript of Seeing a single molecule vibrate through time-resolved coherent anti-Stokes Raman scattering