High-resolution imaging of graphene by tip-enhanced coherent anti-Stokes Raman scattering

Coherent anti-Stokes Raman scattering (CARS) is able to enhance molecular signals by vibrational coherence compared to weak Raman signal. The surface or tip enhancement are successful technologies, which make it possible for Raman to detect single molecule with nanometer resolution. However, due to technical di±culties, tip-enhanced CARS (TECARS) is not as successful as expected. For single molecular detection, high sensitivity and resolution are two main challenges. Here, we reported the ̄rst single atom layer TECARS imaging on Graphene with the highest resolution about 20 nm, which has ever been reported. The highest EFTECARS=CARS is about 10, the similar order of magnitude with SECARS (EF of tip is usually smaller than that of substrates). Such resolution and sensitivity is promising for medical, biology and chemical applications in the future.


Introduction
Raman spectroscopy provides molecular vibrational ngerprint information and has been widely used in medicine, biology, materialogy, chemical engineering and bromatology. Compared to weak spontaneous Raman signal, coherent anti-Stokes Raman scattering (CARS) can enhance molecular signals by vibrational coherence. 1,2 CARS is a third-order nonlinear Raman imaging modality, 3 which needs three incident¯elds including a pump°ied (! 1 ), a Stokes°ied (! 2 ; ! 2 < ! 1 ), and a probe°ied (! 1 ). With a high numerical aperture (NA) objective lens, the phase matching condition can be satis¯ed at the focused light spot. 4,5 The anti-Stokes signal (2! 1 À ! 2 ) will be generated if the di®erence frequency of°ied pump and Stokes (! 1 À ! 2 ) coincides with molecular vibrational frequency (Fig. 1). *, † Corresponding authors. This is an Open Access article published by World Scienti¯c Publishing Company. It is distributed under the terms of the Creative Commons Attribution 4.0 (CC-BY) License. Further distribution of this work is permitted, provided the original work is properly cited.
CARS microscopy can achieve submicron scale three-dimensional imaging. 6,7 However, the resolution is still limited by the light di®raction (which is about 300 nm). 6 To achieve both higher resolution and signal sensitivity, several novel methods have been proposed such as structured illumination in wide-¯eld CARS microscopy, 8-10¯b er probe tip near-¯eld scanning, 11 and also electric¯eld enhancement e®ect 12 including surface enhancements [13][14][15][16][17] and tip enhancements. 18 The¯eld enhancement e®ect is attributed to the excitation of the local mode of the surface plasmon polaritons, 12 and was wildly applied to the amplication of light emission such as: two-photon-excited°uorescence, 19 infrared absorption, 20,21°u orescence, 22 and Raman spectroscopy. In Raman, this technique is well known as surface enhanced Raman scattering (SERS) [23][24][25] and tip enhanced Raman scattering (TERS). [26][27][28] In the past decade, TERS becomes a promising technique. With a \hot" tip apex, it is possible to detect the Raman signal from a single molecule, and even from a single chemical bond 29,30 with spatial resolution down to 1 nm. 31,32 However, due to technical challenges, the surface enhanced CARS (SECARS) and tip enhanced CARS (TECARS) didn't show a higher sensitivity and resolution as expected, and the enhancement e®ect is one of the challenges: The SECARS enhancement factor (EF) over CARS (EF SECARS=CARS ) can theoretically reach 10 8 -10 24 and EF SECARS=Raman can even reach 10 14 -10 30 , 33 which has not yet been experimentally demonstrated. In general, the near-¯eld enhancement of tip is smaller than that of substrates, and because of which, the EF TECARS=CARS has not even been measured in the recent reports of TECARS. 17,18,34,35 Due to technical di±culties, the reported spatial resolution in TECARS microscopy is far less than expected (20 nm for DNA cluster 18 ; $80 nm for cell membrane 34 and $60 nm for carbon nanotube clusters 17,35 ). The single molecule detection by TECARS has never been achieved yet.
In this paper, graphene sample was imaged by TECARS, and this is the¯rst TECARS report imaging single atom layer at one of the highest resolutions of 20 nm by Au tip. The highest EF TECARS=CARS is about 10 4 , which is even close to SECARS 36,37 and had never been seen in TECARS experiments before. With technological improvement, it is expected that there is a bright future for biological, chemical and medical application of TECARS.

Methods and Materials 2.1. TECARS system and measurement
The TECARS microscopy system is shown in Fig. 2, which is a combination of a homemade CARS microscope and an atomic-force microscope (AFM).
The homemade CARS microscope is equipped with two mode-locked Ti:sapphire lasers pulse (Coherent Inc., Mira HP-F laser, pulse duration 2 ps, pulse repetition rate 76 MHz), and an inverted optical microscope (Nano¯nder HE, Tokyo Instruments, Inc.) with the built-in 532 nm Raman laser, an EMCCD (Newton DU920P-BEX2-DD), an avalanche photodiode (APD, Martock Design Ltd) and AFM (AIST-NT, Combiscop TM -1000 SPM). ! 1 and ! 2 beams are collinearly overlapped in time and space (! 1 ¼ 861 nm and ! 2 ¼ 756 nm), and introduced into the microscope system with an objective lens (immersion: air, magni¯cation: 60Â, Numerical aperture: 0.9, RMS60X-PFC, Olympus) focused onto the sample surface. The optical pulses synchronization was controlled by synchro-lock advanced performance system (Synchro-Lock AP controller, Coherent Inc.). The AFM controlled probe tip contacts the sample surface with the constant force and is illuminated by the focused spot. The repetition rate of the excitation lasers is controlled by a pulse picker (Conoptics Model 25D). CARS emission is collected with the same objective lens and detected with the APD. During measurement, the repetition rate was reduced to about 25 MHz and the average power is about 1 mW. The exposure time of each pixel is 50 ms.
Raman spectrum was excited by the built-in 532 nm Raman laser at about 1 mW. 60 Â objective lens (immersion: air, Numerical aperture: 0.9, RMS60X-PFC, Olympus) was used for both exciting and detection. CCD detector was calibrated by the built in calibration laser.

Tip and sample preparation
AFM tip was purchased from Nanosensors (ATEC-NC-50), and metal coating of the 20 nm Au was carried out by E-beam evaporation (Edwards, BOC-500).
Graphene was grown through chemical vapor deposition (CVD) method, as reported previously, 38 and in brief: 10% HCl cleaned Cu foil was inserted into the corundum tube inside a horizontal furnace and heated in H 2 /Ar at 1000 C. Graphene nuclei was formed with 0.05% methane in Ar for 10 min. Then the°ow rate of Ar is increased for 15 min in order to enlarge the graphene nuclei size. After growth, the furnace cooled down quickly.
Then, graphene was transferred onto glass substrate by wet-transfer method 39 : In general, one side of the graphene/Cu foil was spin-coated with (Poly (methyl methacrylate) (PMMA), (950 PMMA, MicroChem) and heated at 120 C for 1 min. The other side of the graphene was removed by O-plasma. Cu foil was etched overnight by 0.1 M (NH 4 Þ 2 S 2 O 8 (Sigma Aldrich). After cleaning by deionized water, the graphene/PMMA¯lm was picked up by glass substrate, and dried in vacuum. Finally, hot acetone was sued to dissolve PMMA for 1 h.

Graph imaging
One of the main challenges of TECARS is the single molecular detection, which has never been reported before. Here, graphene is chosen for TECARS measurement. As a two-dimensional building block High-resolution imaging of graphene by tip-enhanced coherent anti-Stokes Raman scattering 1841003-3 of carbon allotropes, graphene plays an important role since its electronic properties. [40][41][42][43][44] The Raman spectra of graphene sheet is shown in Fig. 3(a), the 2D band is characterized by a single peak and the intensity is about 4 times stronger than the G peak, which indicated that the graphene sample is a single layer graphene. 45 The AFM imaging are shown in Fig. 3(c). The height of graphene is about 1 nm (Fig. 3(d)), which again demonstrated the single graphene layer.
TECARS measurements of graphene G band were carried out. The coherent anti-Stokes signal of G band (at 1612 cm À1 ) was excited by 861 nm (Stokes) and 756 (pump and probe), and the CARS signal was detected at 674 nm by APD (Figs. 3(b) and 4). The 20 nm gold covered tip was used for signal enhancement.
CARS process is a third-order nonlinear scattering, which need the spatial overlap and time synchronization of pump, probe and Stokes pulses. Figure 4 shows the CARS signal was generated only when the tow incident laser pulses are synchronized, which suggested the signal was generated by vibrational coherence.

The enhancement factor
Another technical di±culty of enhanced CARS is that the EF is always not as high as expected. Surface EF over CARS are denoted as EF SECARS=CARS / jg p j 4 jg S j 2 jg aS j 2 , where g p , g S and g aS are the near-¯eld EFs of the pump, Stokes and anti-Stokes signal, which can theoretically reach 10 8 -10 24 , and EF SECARS=Raman can reach 10 14 -10 30 in Ref. 33, but never been reached in experiments. The near-¯eld EF of tip is smaller than that of substrates, it is expected that the EF TECARS=CARS should be lower than SECARS and may be hard to measure, and the reported TECARS experiments has never shown the EF. Here, we experimentally found the highest EF TECARS=CARS can reach 10 4 which is even almost near the EF of SECARS. 36,37 Figure 5 shows the TECARS imaging of graphene with tip approaching the sample (Figs. 5(a) and 5(b)). Compared to control experiment (tip only, without graphene. Figures 5(c) and 5(d)), the  TECARS signal intensity of graphene and background signal near the tip apex are about 3:8 Â 10 5 and 9:6 Â 10 4 Counts, respectively (Fig. 5(e)). With the background subtracted, the intensity is about 2:9 Â 10 5 Counts (3:8 Â 10 5 À 9:6 Â 10 4 ). The signal intensity without enhancement e®ect can be obtained in the region distant from tip which are about 18 and 8 Counts corresponding to the Graphene and background, respectively (Fig. 5(f)), and after deducting the background, the CARS signal intensity is 10 Counts. The EF TECARS=CARS is about 3 Â 10 4 , close to SECARS EF, 36,37 calculated by the equation EF ¼ I nf =N nf I ff =N ff , where I nf and I ff are the neareld and far-¯eld intensities and N nf and N ff are the corresponding numbers of molecules contributing to each signal. 46

Spatial resolution
One of the most important technical challenges in TECARS, which attracts the most attention, is spatial resolution. To evaluate the resolution, a smaller scanning step as 10 nm was set. The spatial resolution can be computed as the lateral distance between 20% and 80% of the edge's height. 47 Figure 6 shows the highest resolution is about 20 nm, which is the same as the highest resolution of TECARS imaging which has ever been reported (with DNA sample clusters). 18

Conclusion
In conclusion, here we¯rst reported the single atom layer TECARS imaging, since the single layer graphene can be characterized by both AFM imaging and Raman spectra. Before this, single molecular detection by TECARS has only been theoretically demonstrated. Sensitivity, EF and resolution are the major challenges of TECARS technique and our data show the highest EF TECARS=CARS is about $10 4 , which has never been measured in TECARS reports and have the similar order of magnitude with SECARS 36,37 (EF of tip is usually smaller than that of substrates). The highest resolution is about 20 nm, the same as the highest TECARS resolution which have ever been reported. 18 TECARS imaging can be a promising tool for studying morphological, biochemical of materials and live cells at the molecular and subcellular levels without labeling. It is anticipated that imaging quality can be improved signi¯cantly by optimizing the tip, polarization laser pulse shaping and also optical path optimization to perform more accurate TECARS imaging in the future.

Con°ict of Interest
The authors declare no competing¯nancial interest.