Cavity-enhanced Raman microscopy of individual carbon nanotubes

Abstract

Raman spectroscopy reveals chemically specific information and provides label-free insight into the molecular world. However, the signals are intrinsically weak and call for enhancement techniques. Here, we demonstrate Purcell enhancement of Raman scattering in a tunable high-finesse microcavity, and utilize it for molecular diagnostics by combined Raman and absorption imaging. Studying individual single-wall carbon nanotubes, we identify crucial structural parameters such as nanotube radius, electronic structure and extinction cross-section. We observe a 320-times enhanced Raman scattering spectral density and an effective Purcell factor of 6.2, together with a collection efficiency of 60%. Potential for significantly higher enhancement, quantitative signals, inherent spectral filtering and absence of intrinsic background in cavity-vacuum stimulated Raman scattering render the technique a promising tool for molecular imaging. Furthermore, cavity-enhanced Raman transitions involving localized excitons could potentially be used for gaining quantum control over nanomechanical motion and open a route for molecular cavity optomechanics.

ARTICLE

Received 11 Dec 2015 | Accepted 7 Jun 2016 | Published 12 Jul 2016

Raman spectroscopy reveals chemically specic information and provides label-free insight into the molecular world. However, the signals are intrinsically weak and call for enhancement techniques. Here, we demonstrate Purcell enhancement of Raman scattering in a tunable high-nesse microcavity, and utilize it for molecular diagnostics by combined Raman and absorption imaging. Studying individual single-wall carbon nanotubes, we identify crucial structural parameters such as nanotube radius, electronic structure and extinction cross-section. We observe a 320-times enhanced Raman scattering spectral density and an effective Purcell factor of 6.2, together with a collection efciency of 60%. Potential for signicantly higher enhancement, quantitative signals, inherent spectral ltering and absence of intrinsic background in cavity-vacuum stimulated Raman scattering render the technique a promising tool for molecular imaging. Furthermore, cavity-enhanced Raman transitions involving localized excitons could potentially be used for gaining quantum control over nanomechanical motion and open a route for molecular cavity optomechanics.

DOI: 10.1038/ncomms12155 OPEN

Cavity-enhanced Raman microscopy of individual carbon nanotubes

Thomas Hmmer1,2, Jonathan Noe1,3, Matthias S. Hofmann1,3, Theodor W. Hansch1,2, Alexander Hgele1,3

& David Hunger1,2

1 Fakultat fr Physik, Ludwig-Maximilians-Universitat, Schellingstrabe 4, Mnchen 80799, Germany. 2 Max-Planck-Institut fr Quantenoptik, Hans-Kopfermann-Stra 1, Garching 85748, Germany. 3 Center for NanoScience (CeNS), Ludwig-Maximilians-Universitat, Schellingstrabe 4, Mnchen 80799, Germany.

Correspondence and requests for materials should be addressed to A.H. (email: mailto:[email protected]

Web End [email protected] ) or to D.H. (email: mailto:[email protected]

Web End [email protected] ).

NATURE COMMUNICATIONS | 7:12155 | DOI: 10.1038/ncomms12155 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 1

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms12155

Raman spectroscopy is a powerful technique to identify the molecular structure of a sample. Combined with microscopy, it can provide high-resolution images with chemical

contrast. This offers vast potential to investigate a broad variety of micro and nanosystems of physical, chemical and biological relevance. However, the inherently weak scattering cross-section of Raman transitions poses a challenge, and is particularly severe for spatially resolved measurements of single- or few-particle levels. Therefore, various enhancement techniques have been devised, which either rely on stimulating radiation1 or near-eld enhancement2,3.

Purcell enhancement in optical microcavities4 was suggested early5,6 for its use in Raman spectroscopy. It relies on the stimulating effect of increased vacuum uctuations within a microcavity to enhance the scattering process. It thus does not require stimulating radiation, and thereby avoids high-intensity levels at the sample, and does not introduce any intrinsic background. Detailled knowledge of the cavity mode structure allows to quantitatively evaluate the observed signal strength, which is difcult, for example, for near-eld enhancement techniques. However, due to the need for microscopic mode volumes, Purcell enhancement of Raman signals has only been observed in monolithic systems with limited spatial control or spectral tunability like microdroplets5,7, photonic crystal cavities8 and FabryProt microcavities6,913. This has restricted the spectral coverage and impeded spatially resolved measurements so far, and only large ensembles of molecules have been studied.

Here, we present Purcell-enhanced Raman imaging using an open-access bre-based microcavity14. This cavity platform provides small mode volume and large quality factor leading to signicant Purcell enhancement, and enables spectrally and spatially resolved measurements. We demonstrate cavity-enhanced Raman imaging to study individual single-walled carbon nanotubes (CNTs), a material with widespread applications in elds as diverse as electronics15, photonics16, nanomechanics17 and quantum optics18,19. To obtain information on a CNT samples heterogeneity with regards to their chirality, crystalline quality and local environment one has to characterize multiple isolated CNTs that are spatially well-separated. Raman imaging is particularly sensitive to these differences20, and combined with complementary measurements probing, for example, absorption2123 can enable a detailed characterization. Notably, cavity-enhancement is at the same time present for Raman scattering, absorption and uorescence2426. Our approach thus allows to simultaneously probe the vibrational spectrum and the optical properties of a wide range of samples with enhanced sensitivity.

ResultsScanning-cavity setup. We use a high-nesse bre-based Fabry Prot microcavity14, where the laser-machined concave end-facet of an optical bre is used as a micro-mirror, see Fig. 1a. Altogether with a planar macroscopic mirror that carries the sample and that can be moved laterally, we realize a scanning-cavity microscope2628. To record spatially resolved measurements, we raster-scan the large mirror and detect cavity signals such as the resonant cavity transmission or Purcell-enhanced Raman scattering pixel-by-pixel. The bre mirror is mounted on a piezo actuator that allows to continuously set the cavity length to achieve suitable resonance conditions (see the Methods section). The dielectric mirror coatings lead to a nesse of up to F 56; 000 and a quality factor Qc 2Fd=l depending

on the cavity length d. A small radius of curvature of the concave micro-mirror on the bre and a few-micron cavity length enables small mode volumes down to V 13l3 (see the Methods

section).

Purcell enhancement of Raman scattering. Cavity-enhanced Raman scattering can be understood analogously to uorescence emission in a resonant cavity4,29. The spectral density of the Raman scattered light is enhanced by the (ideal) Purcell factor C0 due to increased vacuum uctuations (proportional to V 1) and an increased density of states (proportional to Qc), combining to C0 (3)/(4p2)l3(Qc)/(V). For our cavities with QcE105 and

VE10l3, the enhancement of the spectral density of the emission reaches values C0E103. In the regime where the cavity resonance is more narrow than a particular Raman feature, the effective Purcell factor C (3)/(4p2)l3(Qeff)/(V) is the gure of merit30,

which accounts for the fact that the cavity enhances only part of the transition. This is included by replacing Qc with the effective quality factor Qeff (Qc 1 Qr 1) 1 (ref. 31), where Qr is the

quality factor of the Raman feature when assuming a Lorentzian spectrum. Figure 1b shows a schematic level scheme of cavity-enhanced Raman scattering and Figure 1d shows a calculation of the ideal and effective Purcell enhancement as a function of the cavity mode volume for different Qr. Typical

Raman features of CNTs have QrE1,0002,000 such that for a cavity with VEl3 one can expect CE100. With narrower Raman lines one can approach C C0. The cavity-enhanced

Raman eld is scattered entirely into the easily collected cavity mode, which, for an asymmetric choice of mirror transmissions (T2cT2) and low loss (LooT2; see the Methods section), can yield a near-unity outcoupling efciency from the cavity

Zc

T2/(T1 T2 L).

Cavity-enhanced extinction microscopy. For our experiments we transfer single-wall CNTs with an average diameter of 1.4 nm grown by chemical vapour deposition onto the planar mirror (see the Methods section). The mirror is designed to have a eld maximum at the surface to ensure optimal coupling of nanotubes to the cavity mode. We study two different cavities (A and B, see the Methods section) with coatings for different wavelength ranges. In a rst step, we perform scanning-cavity-extinction microscopy26 on a large area of the sample. The cavity enhances the extinction signal by a factor 4F=p 7 104

(cavity A), where F is the cavity nesse, and thereby enables

superior sensitivity for absorption microscopy. To acquire an image, we raster-scan the large mirror and tune the mirror separation over a few hundred nanometres at each pixel to sample the cavity resonance at the wavelength of maximum nesse with a narrow-band probe laser (780 nm for cavity A and 890 nm for cavity B, respectively, see the Methods section). We record the resonant cavity transmission Tc on a photodiode to infer the sample extinction B from Tc E4T1T2=T1 T2 L1 L2 2B2. Here, E is the mode

matching between the bre mode and the cavity mode and T1,2,L1,2 are the respective mirror transmission and loss, which are all determined independently (see the Methods section). From the sample-induced loss we deduce the extinction cross-section sext Bpw02/4, where w0 is the e 1 cavity eld

mode waist radius. To account for the anisotropy of the CNT extinction, we take four measurements with different orientations of linear polarization and evaluate the maximum extinction cross-section observed for each pixel. Alternatively, we use circularly polarized light (see the Methods section). Figure 2d shows a measurement with cavity A where we quantitatively image the extinction cross-section of individual CNTs with a background noise level of 5 nm2. We observe continuous elongated structures where a dominant fraction shows an extinction cross-section per length of sl

2=p

p sext=w0 45 nm2 mm 1 or a corresponding

cross-section per carbon atom of 3 10 4 nm 2

when assuming a 1.4-nm tube diameter. This is consistent with

2 NATURE COMMUNICATIONS | 7:12155 | DOI: 10.1038/ncomms12155 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms12155 ARTICLE

a b

v

Excitation

Raman

f

g

c

Wavenumber (cm1)

d

0 500 1,000 1,500

103

Qr>>Qc

104

103

102

Finesse

(eff.) Purcell factor

Qr

2,000

1,000

102

4,000

101

101

760 780 800 820 840 860 880 900 (nm)

0 5 10 15 20 Mode volume (3)

Figure 1 | Principle of cavity-enhanced Raman spectroscopy. (a) Schematic of the cavity formed by the end-facet of an optical bre (top) and a macroscopic mirror (bottom). CNTs (black lines) are deposited on the macroscopic mirror and interact strongly with the light eld when inside the cavity. The large mirror can be positioned laterally by a nanopositioner and the cavity length can be tuned by a piezo actuator (see black arrows). (b) Schematic level diagram for cavity-enhanced Raman scattering showing a ground state (g), a virtual state (v) and a vibrationally excited state (f). The Raman scattering is stimulated by the resonance with the cavity. (c) Calculated cavity nesse from a measurement of the mirror transmission (line) and a direct measurement of the Finesse (blue datapoint) for cavity B (see the Methods section). The upward and downward arrows symbolize excitationand stimulated Raman scattering, respectively. Selective enhancement occurs at a cavity resonance, whose wavelength is set by the cavity length.(d) Calculation of the effective Purcell factor as a function of mode volume for different quality factors Qr of Raman features (solid lines) for QrooQc. The dashed line is the ideal Purcell factor, equivalent to the effective Purcell factor for Qr44Qc. It solely depends on the Q-factor of the cavity and is evaluated for cavity B used in the experiments.

individual CNTs probed away from an absorption resonance21,22 and demonstrates the potential for quantitative analysis of our technique.

Cavity-enhanced Raman spectroscopy. Signicant additional information can be obtained by recording the Raman spectrum of the sample. To do so, we choose a location where absorption consistent with single CNTs is observed. We excite the cavity with a laser at a wavelength where the mirror coating is transparent (690 nm for cavity A, 786 nm for cavity B) such that light can enter the cavity independently of the mirror separation, see Fig. 1c. The Raman scattered light is red-shifted from the excitation laser to wavelengths where the cavity nesse and thus the Purcell factor is high, such that cavity-enhancement occurs (see Fig. 1b,c). This approach allows to access the range of B3003,000 cm 1, covering the important molecular ngerprint region. To record a broadband Raman spectrum, we stepwise tune the mirror separation to sweep a set of cavity resonances across the desired spectral range where Raman features are expected. The number k of resonances contributing simultaneously to probe the spectrum depends on the effective cavity length d and the considered spectral bandwidth Dn (in units of Hz) according to k ceil[2dDn/c]. At each spectral

position of the cavity resonances we record the Purcell-

enhanced scattering rate with a spectrometer and construct a full Raman spectrum by combining the maxima of all spectra. The Raman features of CNTs studied here are the G-band related to tangential axial ( ) and circumferential ( ) vibra

tions of neighbouring carbon atoms, the defect-related D-band, and its overtone the G0-band20. Figure 2a shows a high-resolution measurement of the G-band at small mirror separation using a single longitudinal cavity mode. The measurement is combined from 370 individual spectra which are shown in Fig. 2b in a false-colour map as a function of the cavity length. With such a high sampling it is possible to achieve a spectral resolution beyond that of the spectrometer and to resolve also narrow lines, whereby the cavity-enhancement maintains the Raman lineshape. Figure 2c shows a large-bandwidth spectrum where the G and G0-bands are visible.

Here, resonances with longitudinal mode numbers q 3438

contribute due to the larger average mirror separation. For the measurement we combine 30 spectra where we change the cavity length in 10 nm steps and additionally modulate the length with a comparable amplitude at a high frequency. This provides a spectrometer-limited resolution with a reduced number of exposures. The absence of the D-band expected at 1,350 cm 1 indicates high quality of the material. At the same time, the signal is signicantly enhanced by the Purcell effect (see below).

NATURE COMMUNICATIONS | 7:12155 | DOI: 10.1038/ncomms12155 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 3

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms12155

a

Raman shift (cm1)

c

Raman shift (cm1)

1,500 1,550 1,600 1,650 1,700

Int (a.u.)

q = 38 q = 36

q = 37 q = 35

b

1

Int (a.u.)

Intensity (a.u.)

1,200 1,400 1,600 1,800 2,000 2,200 2,400 2,600

Length (m)

2.80

q = 6

2.75

0

0 0.1 0.2 0.3

Cavity length 14.3 m (m)

0 0.1 0.2 0.3

0 0.1 0.2 0.3

0 0.1 0.2 0.3

0

1,500 1,550 1,600 1,650 1,700Raman shift (cm1)

d e f

400

1

1

300

200

ext (nm

2)

G intensity (a.u.)

0.25

G intensity (a.u.)

100

0

0.15

Figure 2 | Cavity-enhanced absorption and Raman imaging. (a,b) Acquisition of a Raman spectrum by tuning a cavity resonance across a spectral range around the G band and combining individual exposures. (b) Each horizontal line corresponds to an individual spectrum acquired for a particular cavity length. The strong line shows the tuning of the fundamental mode, the second weaker line corresponds to the rst higher transverse mode of the cavity. (a) Spectrum obtained by taking the maximum of all exposures for each spectral position. (c) Broadband spectrum covering the G and G0 band at 1,590 cm 1 and 2,650 cm 1 respectively. The differently shaded areas depict the range which individual longitudinal cavity modes cover at a mirror separation of 14.3 mm. (d) Scanning-cavity microscopy of the extinction cross-section sext of isolated CNTs. (e,f) Raman hyperspectral imaging of the same area showing selectively the strength of the G0 and the G band, respectively. Scale bars, 20 mm.

Cavity-enhanced Raman imaging. We expand this capability to perform hyperspectral imaging. Therefore, we record cavity-enhanced Raman spectra such as the one shown in Fig. 2c at many locations on the mirror. Figure 2e,f show a measurement where the strength of the G and G0-band at 1,590 cm 1 and 2,650 cm 1 is visualized by selecting spectral bands of about 20 cm 1 width, respectively. We observe excellent correspondence of the Raman map and the absorption image. The differences between the G and G0 band partially originate from the spatial variation of sample extinction, which, for this measurement, affects the collected G-band more than the G0-band due to the higher nesse at the spectral position of the G resonance. The spatial resolution of the Raman maps is found to be 2.1 mm, which is higher than for the extinction measurement where it is consistent with the cavity mode waist w0 3.2 mm. This derives

from the fact that the point spread function for Raman imaging is a product of the excitation and collection point spread functions. Due to a residual low-nesse cavity at the excitation wavelength, both are given by the cavity mode, such that the resolution improves by a factor of B 2

p .

Hyperspectral imaging provides a maximum of information about the sample, and can permit one to chart, for example, the diameter distribution of nanotubes (see the subsection Optimized hyperspectal imaging below). The improved signal from the cavity can allow to detect otherwise hidden spatial or spectral structures. The acquisition of a broadband spectrum with the cavity involves several exposures with 100500 ms integration time at each pixel. Much faster Raman imaging can be realized by scanning-cavity microscopy with a cavity stabilized on resonance

with a particular Raman band. The Purcell enhancement of the signal can then provide a net advantage in acquisition time compared with conventional confocal Raman microscopy. We demonstrate this approach in the following using cavity B and a different CNT sample produced in the same way but yielding shorter tubes. We implement a digital optimization algorithm for stabilizing the cavity on resonance with the G -band of CNTs (see the Methods section). Figure 3a shows an example where we raster-scan the mirror with a cavity length of 3.2 mm and use a dwell time of 400 ms per pixel. Here, the smaller effective radius of curvature of cavity B (see the Methods section) leads to a spatial resolution of 1.5 mm. Individual nanotubes yield a signal of typically 2 104 counts s 1 for an excitation power of 14 mW,

signicantly larger than for a comparable confocal measurement.

Quantication of Purcell enhancement. To quantify the cavity-enhancement, we compare the cavity-based measurement of the G band for the smallest achievable cavity length d 2.75 mm

with a corresponding confocal Raman measurement taken with a microscope objective (numerical aperture 0.7) for the same position on the same isolated CNT indicated by a circle in Fig. 3a. We have chosen a location where the extension and orientation of the CNT can be resolved (see the Methods section). We set the polarization of the excitation light to be circular at the sample for both measurements to ensure comparable excitation conditions. Figure 3b shows the signal on the spectrometer for the stabilized cavity and the confocal measurement, normalized to equal excitation intensity at the location of the nanotube. We observe a

4 NATURE COMMUNICATIONS | 7:12155 | DOI: 10.1038/ncomms12155 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms12155 ARTICLE

a b c

Wavenumbers (cm1)

Mode volume/3

10 25 50 100 200

10

104

103

102

101

1,580 1,590 1,600

8

350

Ideal Purcell factor

Eff. Purcell factor

6

300

4

250

2

200

150

0 30,000

Counts per second

Counts per second

897.5 898.0 898.5 899.0 (nm)

0 0 2 4 6 8 10 12 14Cavity length (m)

Figure 3 | Cavity-enhanced Raman imaging and quantication of the Purcell enhancement. (a) Raman map showing the strength of the Gband with the cavity actively stabilized on resonance with the Raman transition. Scale bar, 10 mm. (b) Spectra taken at the location of the circle in a with a confocal microscope (blue data points with Lorentzian t) and the stabilized cavity at a cavity length d 2.75 mm (red data points with Voigt t) normalized to the

same intensity at the nanotube. The grey line shows the intrinsic cavity resonance. (c) Scaling of the integrated cavity-enhanced Raman intensity (red data points) as a function of the effective length of the cavity. The red square is evaluated from the measurement shown in b, and the solid line is the theoretically predicted ideal and effective Purcell factor. The error bar is the estimated experimental uncertainty (95% condence interval).

strongly enhanced Raman signal at the cavity resonance, where the limited spectrometer resolution broadens the cavity line to a non-Lorentzian shape.

From several such measurements on different CNTs, and taking into account the respective collection efciencies (see the Methods section), we extract the effective and ideal Purcell enhancement. For the effective Purcell factor we compare the confocal and cavity spectra integrated over the full G band, yielding C 6.22.5. The error denotes the uncertainty in the

determination of the respective excitation intensities and collection efciencies. A similar statistical uncertainty is obtained when evaluating different nanotubes on the mirror and selecting for high cavity-outcoupling efciency. The ideal Purcell factor can be obtained from the ratio of the peak spectral densities of the measurements. We extract the peak spectral density of the intrinsic cavity line, which has an independently measured linewidth of 0.15 cm 1 at the mirror separation used (see the Methods section), by matching its integral with the integrated count rate measured with the spectrometer. This yields an ideal Purcell Factor of C0 320120. These values are in good

agreement to the calculated values C 8.5 and C0 354 which

were deduced from the cavity mode volume V 13l3, the cavity

quality factor Qc 6.1 104 and the quality factor of the Raman

G band Qr 1.54 103. The Raman Q-factor was obtained

from a t to the confocally collected Raman line shown in Fig. 3b. The enhancement also follows the expected scaling for Purcell enhancement when varying the cavity length and evaluating the respective integrated count rate, see Fig. 3c.

The enhanced signal is furthermore collected very efciently. Evaluated at several different locations, we obtain collection efciencies up to Zc 60%, comparable to the best values

achievable with oil-immersion objectives. Overall, the cavity-enhancement and efcient collection provide a net signal gain, which for example can allow a speedup in Raman imaging and enable the detection of weak Raman features, which otherwise might be hidden in the background.

Optimized hyperspectral imaging. As a nal step, we demonstrate the potential to study the details of spectral features while still beneting from a net enhancement due to the cavity. We use this for efcient CNT characterization by resolving properties of

the G band, where the line splitting is related to the tube diameter, and the linewidth of the G transition allows to separate metallic and semiconducting CNTs20. Figure 4a,b show examples for a semiconducting and a metallic nanotube, respectively. For the measurement we weakly stabilize the cavity to an average length where it is resonant with the G band and modulate the mirror separation with a high-frequency triangular waveform to sweep the cavity resonance across the Raman line. This enables the acquisition of a hyperspectral image covering the G band. Since we restrict the cavity resonance to a spectral region where a Raman signal is observed, we benet from the average enhancement, which for the measurement shown still provides more than a factor of three average Purcell enhancement of these lines.

We t each spectrum with a sum of two Lorentzians and evaluate their separation D and the linewidth of the low-frequency G resonance. Figure 4c shows the tube diameter inferred from20 dt 6:8=

p , and Fig. 4d shows the G linewidth for the same area as shown in Fig. 3a. For most of the nanotubes we obtain a clear indication of their diameter and electronic structure. This technique is particularly robust, since the cavity needs to be stabilized to a mirror separation within a few nanometres only.

DiscussionWe have demonstrated cavity-enhanced Raman spectro-imaging of individual nanoobjects. The accessible spectral range covers the molecular ngerprint region, underlining the general applicability of the technique. Purcell enhancement and the cavity-intrinsic highly efcient collection of signals provide a net gain compared with confocal Raman imaging, and improve the analysis of heterogeneous samples on large areas.

Reducing mode volumes further can improve the Purcell enhancement by about one order of magnitude32, additional enhancement can be achieved by cavity-enhancement of the excitation laser by up to a factor 104 as shown earlier9, and collection efciencies can exceed 90% by slightly optimizing the cavity mirror reectivities. Overall, enhancement factors beyond 106 can be realized in this way. Excitation of Raman transitions via electronic resonances20 can additionally increase the signal. Furthermore, the cavity serves as a spectrally selective element

D

NATURE COMMUNICATIONS | 7:12155 | DOI: 10.1038/ncomms12155 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 5

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms12155

a

c d

Intensity (a.u.)

G+

G

a

a

1,560 1,570 1,580 1,590 1,600 1,610 Raman shift (cm1)

1,560 1,570 1,580 1,590 1,600 1,610 Raman shift (cm1)

b

b

b

Intensity (a.u.)

1.2 2.0

Diameter (nm)

1 5

G FWHM (cm1)

Figure 4 | Mapping of CNT diameter and electronic structure using cavity-enhanced hyperspectral imaging. Example of cavity-enhanced G-band spectra for a semiconducting (a) and a metallic (b) nanotube, taken at the location of the circles shown in c and d. The solid line is a sum of two Lorentzians tted to the data and the dashed lines are the underlying individual Lorentzians of the G and G lines, respectively. (c) Spatial map of the evaluated tube diameter from the G/G line splitting measured on the same area as shown in Fig. 3a. An average diameter of 1.5 nm is observed, in good agreement with atomic force microscopy. (d) Spatial map of the evaluated G linewidth (full width at half maximum, FWHM). Broad lines are expected for metallic nanotubes, while semiconducting nanotubes typically show narrow lines. Scale bars, 10 mm.

without sacricing signal, such that with suitable calibration, no spectrometer is required. This could enable, cost-effectively, a spectral resolution given by the cavity linewidth, as small as0.01 cm 1. The strong cavity ltering could also allow the acquisition of spectra at very low wave-numbers, for example, by exciting at a shallow angle through the mirror where the dielectric coating becomes transparent. The demonstrated combination of enhancement of Raman signals and ultra-sensitive absorption spectroscopy extends the accessible information compared with other techniques. Finally, high-nesse microcavities should enable low-threshold Raman lasers with a variety of novel materials.

Beyond Raman imaging, we believe that our approach could be used to gain optical quantum control over phonons. By harnessing strong coupling between cavity photons and localized excitons in CNTs29,33,34, Raman transitions could be used to create and detect single phonons35,36. This could lead to a hybrid cavity-optomechanical platform37, where quantum control of mechanical motion becomes accessible on a molecular scale38.

Methods

Optical setup. The setup combines a scanning-cavity and a confocal microscope, which share a closed-loop nanopositioning stage (attocube ECS 3030) that can translate the sample-carrying mirror between the cavity and the focus of an objective. This enables us to study the same location on the sample with both techniques. For cavity measurements, we couple grating-stabilized diode lasers (690 nm for Raman in cavity A and stabilization in cavity B, 786 nm for Raman in cavity B and cavity characterization in cavity A, 890 nm for cavity characterization in cavity B) into the cavity bre and collect the cavity transmission with an aspheric lens. The cavity length can be ne tuned across B2 mm with a shear piezo onto which the bre is mounted. We detect the transmitted laser light on a sensitive photodiode (Thorlabs APD120) and Raman signals on a spectrometer (Princeton Instruments Acton SP2500 with Andor ikon-M). The confocal setup contains a long working-distance objective with numerical aperture 0.7 (Mitutoyo Plan Apo NIR HR 100x) oriented with an angle of 38 to the surface normal of the mirror, which we account for in the evaluated excitation intensity and collection efciency.

Optical microcavity. We use two different mirror coatings and bres for experiments covering different wavelength ranges. Cavity A has a coating with a stopband centre at 780 nm with T1 12 p.p.m., T2 60 p.p.m., L1 12, L2 27

p.p.m. and a nesse F 2p=T1 T2 L1 L2 56; 000. The bre of cavity A

has an effective radius of curvature rc 190 mm, leading to w0 2.7 mm for a

mirror separation of d 4.5 mm as used in the experiment shown in Fig. 2d,e.

There, the quality factor amounts to Qc 2dF=l 6:5 105. Cavity B has a

coating which is optimized for low background uorescence and large outcoupling efciency in the presence of CNT absorption. It is centred at 890 nm withT1 77 p.p.m., T2 335 p.p.m., L1 12 p.p.m., L2 20 p.p.m. and F 14; 000. A

smaller rc 90 mm leads to w0 2.1 mm for d 3.2 mm. The edges of the tip of bre

B are mechanically polished off to allow for cavity lengths as small as d 2.75 mm,

where we obtain a mode volume V pw02d/4 13l3. The nesse of both cavities is

measured by probing the cavity transmission with a laser while tuning the mirror separation across more than a free spectral range, and evaluating the resonance linewidth and separation. The effective cavity length d, which includes eld penetration into the coating, is deduced from the free spectral range observed in broadband cavity-transmission spectra. The mode waist w0 is measured by absorption imaging of point-like objects with the cavity.

Cavity stabilization. We use a laser at 690 nm where the mirror has high transmission to provide an error signal for stabilization of cavity B. With the additional transmission signal of the Raman excitation laser at 780 nm, we implement a digital optimization algorithm for locking the cavity on resonance with any Raman transition within the accessible spectral range. Our scheme enables locking also when slip-stick motion of the nanopositioner with signicant crosstalk on the cavity length is involved. The mirror separation is stabilized to below 200 pm and therefore maintains the spectral position of the resonance within 0.31 cm 1.

Count rate comparison. When comparing confocal and cavity-enhanced count rates, we correct for the differences in the respective excitation intensities and collection efciencies. For the cavity, the outcoupling efciency through the planar mirror is given by Zc T2/(T1 T2 L1 L2 2B) and depends on the local

extinction of the sample. We use a scanning-cavity extinction measurement covering the area shown in Fig. 3a to calculate Zc at the locations used for enhancement determination. The collection efciency of the objective is calculated by integrating the interfering direct and reected dipole radiation pattern over the opening angle of the inclined objective, and accounting for the polarization- and angle-dependent complex reectivity of the Bragg mirror. The obtained value depends on the dipole orientation and amounts to 43% (15%) including objective transmission for orthogonal (parallel) orientation with respect to the projection of the inclined optical axis in the mirror plane. Transmission through further optics is directly measured; confocal and cavity measurements share most of the optical path including lters and coupling to the spectrometer. The excitation intensity at the nanotube in the cavity is calculated from the measured mode waist, the transmitted laser power and the measured and calculated outcoupling mirror transmission. For the objective, we measure the excitation point spread function and account for the standing wave due to the mirror. Furthermore, when evaluating signals of extended CNTs we account for a spatial overlap factor x with the cavity mode, entering the evaluation of C0, C and B, which are dened as peak values. The observed average values are, for example, hCi xC, where for a

homogeneous linear CNT larger than the mode diameter x

p=8

p , whereas for a

point-like sample x 1.

6 NATURE COMMUNICATIONS | 7:12155 | DOI: 10.1038/ncomms12155 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms12155 ARTICLE

Sample preparation. Single-wall CNTs were synthesized on SiO2 by chemical vapour deposition following the procedure described in ref. 33. In brief, iron-ruthenium particles spin-coated on SiO2 were used to catalyse the growth of CNTs with methane as carbon feedstock. Before the CVD growth the substrate with catalyst particles was treated by oxygen plasma to yield longer CNTs. Subsequently the substrate was transferred into a standard CVD furnace, heated in an argon hydrogen (95%/5%) atmosphere to 850 C and kept in a ow of 1 slm methane and0.75 slm hydrogen for 10 min before the cool-down in an argonhydrogen atmosphere. The average diameter of nanotubes obtained with this method was 1.4 nm. The nanotubes were transferred by contact-stamping the SiO2 chip with CVD-grown CNTs onto the mirror substrate. Scanning electron microscopy showed isolated CNTs on the cavity mirror.

Data availability. The data that support the ndings of this study are available from the corresponding author on request.

References

1. Camp, Jr C. H. & Cicerone, M. T. Chemically sensitive bioimaging with coherent Raman scattering. Nat. Photon. 9, 295305 (2015).

2. Le Ru, E. C., Blackie, E., Meyer, M. & Etchegoin, P. G. Surface enhanced Raman scattering enhancement factors: a comprehensive study. J. Phys. Chem. C 111, 1379413803 (2007).

3. Hartschuh, A., Snchez, E. J., Xie, X. S. & Novotny, L. High-resolution near-eld Raman microscopy of single-walled carbon nanotubes. Phys. Rev. Lett. 90, 095503 (2003).

4. Vahala, K. J. Optical microcavities. Nature 424, 839846 (2003).5. Snow, J. B., Qian, S.-X. & Chang, R. K. Stimulated Raman scattering from individual water and ethanol droplets at morphology-dependent resonances. Opt. Lett. 10, 3739 (1985).

6. Cairo, F., De Martini, F. & Murra, D. QED-vacuum connement of inelastic quantum scattering at optical frequencies: A new perspective in Raman spectroscopy. Phys. Rev. Lett. 70, 14131416 (1993).

7. Symes, R., Sayer, R. M. & Reid, J. P. Cavity enhanced droplet spectroscopy: Principles, perspectives and prospects. Phys. Chem. Chem. Phys. 6, 474487 (2004).

8. Sumikura, H., Kuramochi, E., Taniyama, H. & Notomi, M. Cavity-enhanced Raman scattering of single-walled carbon nanotubes. Appl. Phys. Lett. 102, 231110 (2013).

9. Petrak, B., Djeu, N. & Muller, A. Purcell-enhanced Raman scattering from atmospheric gases in a high-nesse microcavity. Phys. Rev. A 89, 023811 (2014).

10. Tartakovskii, A. I. et al. Raman scattering in strongly coupled organic semiconductor microcavities. Phys. Rev. B 63, 121302 (2001).

11. Fainstein, A., Jusserand, B. & Thierry-Mieg, V. Raman efciency in a planar microcavity. Phys. Rev. B 53, R13287 (1996).

12. Checoury, X., Han, Z., El Kurdi, M. & Boucaud, P. Deterministic measurement of the Purcell factor in microcavities through Raman emission. Phys. Rev. A 81, 033832 (2010).

13. Fainstein, A. & Jusserand, B. Phonons in semiconductor planar microcavities: A Raman scattering study. Phys. Rev. B 54, 11505 (1996).

14. Hunger, D. et al. A ber FabryPerot cavity with high nesse. New J. Phys. 12, 065038 (2010).

15. Avouris, P., Chen, Z. & Perebeinos, V. Carbon-based electronics. Nat. Nanotech. 2, 605615 (2007).

16. Avouris, P., Freitag, M. & Perebeinos, V. Carbon-nanotube photonics and optoelectronics. Nat. Photon. 2, 341350 (2008).

17. Moser, J., Eichler, A., Gttinger, J., Dykman, M. I. & Bachtold, A. Nanotube mechanical resonators with quality factors of up to 5 million. Nat. Nanotech. 9, 10071011 (2014).

18. Hgele, A., Galland, C., Winger, M. & Imamolu, A. Photon antibunching in the photoluminescence spectra of a single carbon nanotube. Phys. Rev. Lett. 100, 217401 (2008).

19. Jeantet, A. et al. Widely tunable single-photon source from a carbon nanotube in the Purcell regime. Available at http://arxiv.org/abs/1508.06297

Web End =http://arxiv.org/abs/1508.06297 (2015).

20. Dresselhaus, M. S., Dresselhaus, G., Saito, R. & Jorio, A. Raman spectroscopy of carbon nanotubes. Phys. Rep. 409, 4799 (2005).

21. Blancon, J.-C. et al. Direct measurement of the absolute absorption spectrum of individual semiconducting single-wall carbon nanotubes. Nat. Commun. 4, 2542 (2013).

22. Oudjedi, L., Parra-Vasquez, A. N. G., Godin, A. G., Cognet, L. & Lounis, B. Metrological investigation of the (6,5) carbon nanotube absorption cross section. J. Phys. Chem. Lett. 4, 14601464 (2013).

23. Heylman, K. D., Knapper, K. A. & Goldsmith, R. H. Photothermal microscopy of nonluminescent single particles enabled by optical microresonators. Phys. Chem. Lett. 5, 19171923 (2014).

24. Motsch, M., Zeppenfeld, M., Pinkse, P. W. H. & Rempe, G. Cavity-enhanced Rayleigh scattering. New J. Phys. 12, 063022 (2010).

25. Tanji-Suzuki, H. et al. Interaction between atomic ensembles and optical resonators: classical description. Vol. 60 (Elsevier Inc., 2011). Available at http://arxiv.org/abs/1104.3594

Web End =http://arxiv.org/abs/1104.3594 .

26. Mader, M., Reichel, J., Hansch, T. W. & Hunger, D. A scanning cavity microscope. Nat. Commun. 6, 7249 (2015).

27. Toninelli, C. et al. A scanning microcavity for in situ control of single-molecule emission. Appl. Phys. Lett. 97, 021107 (2010).

28. Greuter, L. et al. A small volume tunable microcavity: development and characterization. Appl. Phys. Lett. 105, 121105 (2014).

29. Miura, R. et al. Ultralow mode-volume photonic crystal nanobeam cavities for high-efciency coupling to individual carbon nanotube emitters. Nat. Commun. 5, 5580 (2014).

30. Kaupp, H. et al. Scaling laws of the cavity enhancement for nitrogen-vacancy centers in diamond. Phys. Rev. A 88, 053812 (2013).

31. Auffves, A. et al. Controlling the dynamics of a coupled atom-cavity system by pure dephasing. Phys. Rev. B 81, 245419 (2010).

32. Kelkar, H. et al. Sensing nanoparticles with a cantilever-based scannable optical cavity of low nesse and sub-l3 volume. Phys. Rev. Appl. 4, 054010 (2015).

33. Hofmann, M. S. et al. Bright, long-lived and coherent excitons in carbon nanotube quantum dots. Nat. Nanotech. 8, 502505 (2013).

34. Ma, X., Hartmann, N. F., Baldwin, J. K. S., Doorn, S. K. & Htoon, H. Room-temperature single-photon generation from solitary dopants of carbon nanotubes. Nat. Nanotech. 10, 671675 (2015).

35. Wilson-Rae, I., Galland, C., Zwerger, W. & Imamolu, A. Exciton-assisted optomechanics with suspended carbon nanotubes. New J. Phys. 14, 115003 (2012).

36. Restrepo, J., Ciuti, C. & Favero, I. Single-polariton optomechanics. Phys. Rev. Lett. 112, 013601 (2014).

37. Stapfner, S. et al. Cavity-enhanced optical detection of carbon nanotube brownian motion. Appl. Phys. Lett. 102, 151910 (2013).

38. Roelli, P., Galland, C., Piro, N. & Kippenberg, T. J. Molecular cavity optomechanics as a theory of plasmon-enhanced Raman scattering. Nat. Nanotech. 11, 164169 (2016).

Acknowledgements

We thank Jakob Reichel and Matthias Mader for helpful discussions, Hanno Kaupp for assistance with fabricating the bres, and Philipp Altpeter and Reinhold Rath for assistance in the clean-room. We gratefully acknowledge nancial support by the European Research Council under ERC Grant Agreement no. 336749, the DFG Cluster of Excellence NIM (Nanosystems Initiative Munich), the Center for NanoScience (CeNS) and LMUInnovativ. T.W.H acknowledges funding from the Max-Planck Foundation.

Author contributions

D.H., A.H. and T.W.H. conceived the experiment. T.H. and D.H. designed and performed the experiments. T.H., D.H and A.H. analysed the data. J.N. and M.S.H. prepared the sample. D.H., T.H. and A.H. wrote the article. All authors discussed the manuscript.

Additional information

Competing nancial interests: The authors declare no competing nancial interests.

Reprints and permission information is available online at http://npg.nature.com/reprintsandpermissions/

Web End =http://npg.nature.com/ http://npg.nature.com/reprintsandpermissions/

Web End =reprintsandpermissions/

How to cite this article: Hmmer, T. et al. Cavity-enhanced Raman microscopy of individual carbon nanotubes. Nat. Commun. 7:12155 doi: 10.1038/ncomms12155 (2016).

This work is licensed under a Creative Commons Attribution 4.0 International License. The images or other third party material in this article are included in the articles Creative Commons license, unless indicated otherwise in the credit line; if the material is not included under the Creative Commons license, users will need to obtain permission from the license holder to reproduce the material. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/

Web End =http://creativecommons.org/licenses/by/4.0/

r The Author(s) 2016

NATURE COMMUNICATIONS | 7:12155 | DOI: 10.1038/ncomms12155 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 7