Chinese Physics Letters, 2020, Vol. 37, No. 1, Article code 015202 Partially Overlapped Dual Laser Beams to Reduce Ablation Craters * Meng-Han Wang (王梦涵)1,2, Jun-Le Qu (屈军乐)1**, Ming Zhu (祝铭)1,3 Affiliations 1Key Laboratory of Optoelectronic Devices and Systems of Ministry of Education and Guangdong Province, College of Optoelectronic Engineering, Shenzhen University, Shenzhen 518060 2Shenzhen Yuanqing Environmental Technology Service Co., Ltd, Shenzhen 518071 3Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences, Shenzhen, 518000 Received 16 September 2019, online 23 December 2019 *Supported by the National Natural Science Foundation of China under Grant Nos 61525503 and 41603059, the National Key Research and Development Project of China under Grant No 2016YFC14000701, and the Shenzhen Basic Research Project (JCY20160428092445411).
**Corresponding author. Email: jlqu@szu.edu.cn
Citation Text: Wang M H, Qu J L and Zhu M 2020 Chin. Phys. Lett. 37 015202    Abstract We present our experimental ablation results for partially overlapped dual nanosecond laser beams (PO-DB) on metal and glass surfaces. Numerical simulations are performed to evaluate the crater reduction potential of the PO-DB setup. Damage probability experiments proved the collaboration of two beams within the overlap region. Bright-field and three-dimensional profile measurements verify the reduced ablation area from the proposed PO-DB scheme. Laser-induced plasma is generated when transparent glass is ablated. Atomic emission of Na I ($\sim$589.95 nm) shows comparable signal between the PO-DB set and the traditional single laser beam set. The proposed PO-DB ablation mechanism could also be applied to femtosecond laser systems. DOI:10.1088/0256-307X/37/1/015202 PACS:52.38.Mf, 42.62.Fi, 61.80.-x © 2020 Chinese Physics Society Article Text Laser induced breakdown spectroscopy (LIBS) is a noncontact approach for microanalysis that uses a highly energetic laser pulse as the excitation source. LIBS differs from many other spectroscopy methods so that it can be applied for multiple simultaneous atomic analyses.[1] A recent study has shown that LIBS could detect small amounts of substance, as little as 75 fg, using an optical trap.[2] However, LIBS can cause significant sample damage. Therefore, reducing the ablation size is essential to improve spatial resolution and to minimize sample effects. Reducing ablation area to the scale of single cell ($\sim $15 µm) would provide a foundation for cellular level detection and could realize in vivo LIBS analysis.[3] The minimum damage area is larger than the spot size when laser is tight focused, where the minimum spot diameter ($L$) is dictated by the diffraction limit, $$ L=1.27\times f\times \lambda \times \frac{M^{2}}{\it\Phi},~~ \tag {1} $$ where $f$ is the objective focal length, $\lambda$ is the irradiating laser wavelength, $M$ is the laser mode, and ${\it\Phi}$ is the incident laser beam diameter. A short-wavelength (smaller $\lambda)$ single-mode (smaller $M^{2}$) laser, or large numerical aperture (NA) objective (i.e., better focus) can help to reduce the diffraction limit of the system and consequently reduce the ablation area. This study proposes a partially overlapped dual-beam (PO-DB) optical scheme to reduce the area where the laser energy reaches the material fluence threshold, as shown in Fig. 1. Numerical simulations using PO-DB with multimode laser beams and flat-top laser beams show the potential for reducing the ablation area. Energy superposition of the dual-beam within the overlap area is validated by the ablation experiment with pure aluminum and the breakdown experiment of soda-lime glass. These two experiments demonstrate the reduction of ablation craters. The possibility of extending the PO-DB set to the femtosecond laser is discussed.
cpl-37-1-015202-fig1.png
Fig. 1. Spatial and temporal overlapped dual laser beam ablation. (a) Spatial overlap of the two 532 nm laser beams is shown with yellow and blue circles, the overlapping area is shown in green. (b) Temporal irradiance within and out of the overlap area is shown. The material ablation threshold is shown with the dashed black line. Fluence within the overlapping area exceeds the threshold, while fluence within the independent region is insufficient to ablate the material.
The proposed theory is shown in Fig. 1. When laser energy does not reach the damage threshold of the material, temporal evolution of the residue beam will be in a Gaussian shape. If the laser energy reaches the damage threshold, the residue energy of the laser beam will be cut off by optical breakdown. Two pulse beams arrive successively at the target surface with picosecond intervals of less than 1% of the pulse duration by using a Michelson interferometer system. The focus spots are overlapped partially (Fig. 1(a)), with the energy in the independent region being lower than the material fluence threshold and the energy in the overlap area exceeding the threshold (Fig. 1(b)). Thus, the ablation area is determined by the degree of overlap, and could be reduced significantly for the optimum optical path, compared with the individual focus spot size. Figure 2 shows the experimental setup. LIBS measurements were obtained using a single multimode Q-switched Nd:YAG laser (Nimma-600, Beamtech) operating at 532 nm, with an 8-ns pulse duration, 8-mm beam diameter, 0.5-nm line width, and $ < $1 mrad divergence. The laser beam was expanded 5-fold and limited by a 3-mm variable iris to restrict the beam central region. The restricted beam was redirected by a beam splitter to generate two temporally separated beams. Because the coherence laser length was approximately 0.35 mm, we set the path-length difference between the separated beams as 2 mm, or 6 ps. Thus, the discrepancy period of two laser beams would be sufficient to avoid interference, whereas the energy lost would be neglected.
cpl-37-1-015202-fig2.png
Fig. 2. Experimental setup for partially overlapped dual laser beam ablation. Restricted laser profile before influxing objective is shown at the top right with apparent multimode characteristics.
The separated beams passed through another beam splitter (30% transmission, 70% reflection) and were focused on the sample surface with a 20$\times$ objective. The sample was placed on a 13-mm $X$–$Y$–$Z$ translation stage driven with a 10-µm interval. Bright-field image and plasma-emission signals were collected from the reflection path of the 30/70 beam splitter and switched by a mirror mounted on a flip adaptor. The spectrometer was coupled with a gated ICCD camera (iStarDH320 T, Andor), using a 1200 l/mm grating, which provided a 0.1-nm spectral resolution with a 50 µm entrance slit and a 10 µs integration time. A CMOS camera (DCC1645 C, Thorlabs) was used to collect micrographs. Mirror R1 was used to adjust the overlapping area by changing the incidence angle of the beam. The combination of a half-wave plate and a Glan–Taylor polarizer was used to adjust the laser fluence continuously. It was ideal that the fluence of a single-beam laser was less than the material fluence threshold, whereas the fluence within the overlap exceeded the threshold. Figure 2 also shows the incident laser fluence distribution, measured with a Spiricon laser beam profiler, which is clearly of multimode. The laser fluence profile at the focal plane is similar to the near-field profile.[4] Thus, the irradiance at the focal plane could be viewed to be similar to the theoretical prediction, as shown in Fig. 1. To validate the proposed set, we performed experiments using aluminum (analytical grade: 99.99%) and common soda-lime glass (microscope slide). Because ablation and breakdown mechanisms are different for these two materials, we discuss the ablation processes separately. Single-beam LIBS always ablates a larger area than the theoretical minimum (diffraction limit) for opaque materials, such as metals. For nanosecond pulses, the laser energy is first absorbed by free electrons and heats the target surface for melting and then vaporization.[5] The energy transported to the metal can be expressed as $$ C_{\rm e} \frac{\partial T}{\partial t}=\frac{\partial \left(k_{0} \frac{\partial T}{\partial z}\right)}{\partial z}+I(t)A\alpha e^{-\alpha z},~~ \tag {2} $$ where $C_{\rm e}$ is the lattice heat capacity per unit volume, ${T}$ is the lattice (and electron) temperature, ${k}_{0}$ is the thermal conductivity, ${z}$ is the normal depth to the target surface, and ${I(t)}$ is the laser irradiance of a temporal Gaussian beam, $$ I(t)=I_{0} \exp [-4(\ln 2)t^{2}/\tau^{2}],~~ \tag {3} $$ $I_{0}$ is the irradiance at the peak of pulse, $\tau$ is the pulse duration (full width at high maximum), ${A}$ and $\alpha$ are surface transmissivity and material absorption coefficient, respectively. Ablation with a long laser pulse will provide sufficient time for the thermal wave to propagate into the target when the irradiance fulfills the condition $$ I>I_{\rm damage} \sim \frac{\rho {\it\Omega} D^{1/2}}{\tau^{1/2}},~~ \tag {4} $$ where $\rho$ is the material density of material, ${\it\Omega}$ is the specific heat of melting or evaporation, and ${D}$ is the heat diffusion coefficient, the material within the focal area would be melted or evaporated. Subsequent evaporation or plasma generation creates a relatively large ablated volume. Thus, within the nanosecond regime, the energy transfer from laser to metals could be viewed as the process of continuous heating. If multiple beams reach the same position in the same time period, they would heat the target simultaneously. In the case of PO-DB, ablation would be restricted within the overlap region when single laser-beam irradiance ${I}$ fulfills the condition $$ 2I>I_{\rm damage} >I.~~ \tag {5} $$ The fluence distribution of the restricted laser was measured using a laser-beam profiler (Figs. 2 and 3(a)). The multimode laser exhibited an irregular angular fluence distribution, with several high fluence bands separated by low fluence areas. Effective irradiance $I_{\rm E}$ within the overlap region could be simplified as a numerical summation of two independent laser beams. The PO-DB fluence distribution was simulated numerically. Two beams were arranged diagonally, then the distance between the beam centers was reduced gradually, which increased the overlap area (Fig. 3 C). A rough estimation of ablation area was calculated based on the simulation, where the smallest ellipse that could cover all areas above the assumed threshold $I_{\rm damage}$ was viewed as the representative ablation area (Fig. 3(f)). $I_{\rm damage}$ was set as 55% of the peak irradiance. Similar simulation was also conducted on flat-top laser beams (six order square super-Gaussian) to infer the area reduction potential of PO-DB (Figs. 3(b) and 3(d)). Figure 3(e) compares the laser fluence profiles among different laser beams and setups. The minimum area that exceeds the set threshold value when PO-DB was applied using a flat-top beam could be 12.25% of that with a single multimode beam.
cpl-37-1-015202-fig3.png
Fig. 3. Numerical simulation of laser beam overlap. (a) Fluence profile of laser used in the experiment, (b) flat-top laser fluence profile calculated from 6-order square super-Gaussian function, fluence distribution of overlapped beams from laser in experiment (c) and flat-top laser (d). (e) Fluence along the diagonal of A, B, C, D and fluence profile of the Gaussian laser beam. (f) Areas with fluence exceeding threshold when the distance between two beam centers increased.
Interaction between the laser and material can result in complex effects, including plasma core, vapor, melt expulsion, melt deposits, and heat conduction to surrounding material.[6] These phenomena are common for irradiated metal samples, and the proposed partial PO-DB LIBS system could melt a larger ablation area than single-beam LIBS if both create plasma. We chose laser energy to ablate only the metal and avoid plasma generation. The ablated material (without generating plasma) could be used as the raw material for secondary plasma generation by introducing an orthogonal second pulse. Thus, the proposed PO-DB setup could be more valuable if combined with other system. For example, PO-DB could act as the ablation laser in an orthogonal dual-pulse LIBS setup.[7,8] The R1 beam center was adjusted $\sim $5 µm from the R2 path center. Because the reduction in ablation area is proportional to the original ablation area, and a defocusing laser beam would ablate a larger material area, laser-beam defocusing could present the reduction effect of the PO-DB better. Therefore, we set the target surface slightly out of focus. Figure 4 shows ablation craters under different setups. The energy of a single laser beam (R1 and R2) was 3.6 µJ. When the PO-DB system was applied, the energy of each beam was set to 2 µJ. Thus, when the fluence within the overlapping area exceeded the aluminum ablation threshold, fluence within the independent region was insufficient to ablate the aluminum. Table 1 shows the two-dimensional ablated area parameters. The ablated area was reduced significantly compared with the single-beam system.
cpl-37-1-015202-fig4.png
Fig. 4. Bright-field microscope image of ablation area with 20$\times$ objective. Top to bottom: R1 path, R2 path, and PO-DB. Left to right: single and 10 pulse accumulated ablation. Fluence densities and ablation area dimensions are listed in Table 1. Overlapped paths produced smaller ablation area.
Table 1. Ablated dimensions for single and 10 accumulated pulses for R1, R2, and PO-DB pulses. Overlapped pulses ablate an ellipse area, whereas the single beam, either R1 or R2, ablates a circular area.
Ablated dimension Fluence density
1 pulse 10 pulses
R1 path 8.29 µm 10.02 µm 7.2 J/cm$^{2}$
R2 path 8.05 µm 8.47 µm 9.3 J/cm$^{2}$
PO-DB (Ellipse) 4.94 and 4.93 µm 5.02 and 4.38 µm Overlap Independence
9.2 J/cm$^{2}$ 4 and 5.2 J/cm$^{2}$
The ablation and breakdown processes for optically transparent materials are quite different, with no free electrons available to absorb the photon energy. For pulses longer than a few tens of picoseconds, the generally accepted ablation model for transparent dielectrics involves conduction band electron avalanche instigated by the incident radiation and transfer of this energy to the lattice.[9,10] Because the controlling rate is that of thermal conduction through the lattice, this model predicts a $\tau^{1/2}$ dependence of the threshold damage fluence. Conceptually, conduction band electrons collide with the matrix many times during each optical cycle, driven by the laser light electric field, and gain sufficient energy to free some electrons.[11] The damage fluence for pulse duration ${F}_{\rm damage}(\tau)$ can be approximated as[12] $$ F_{\rm damage} (\tau)=F_{\rm damage,ref} \sqrt {\tau_{\rm ref} \times \tau } +F_{\rm damage,cw},~~ \tag {6} $$ where ${F}_{damage, ref}$ is the reference damage fluence of the pulse duration $\tau_{\rm ref}$ and ${F}_{\rm damage, cw}$ is the damage fluence for continuous radiation. The damage irradiance $I_{\rm damage}(F_{\rm damage}/\tau)$ is inversely proportional to the square root of the pulse duration,[12] and the frequency dependence of the breakdown field can be expressed as $$ E_{\rm B} (\omega)=E_{\rm B}^{\rm dc} (1+\tau_{\rm e}^{2} \omega^{2})^{1/2},~~ \tag {7} $$ where ${E}_{\rm B}^{\rm dc}$ is the dc field breakdown strength, $\tau_{\rm e}$ is the electron phonon collision time, and $\omega$ is the laser angular frequency. The combined beam irradiance is $$ I(P)=I_{1} (P)+I_{2} (P)+2\sqrt {I_{1} (P)I_{2} (P)} \cos \delta (P),~~ \tag {8} $$ where the distance discrepancy of the two beams is larger than the coherent length, which is the case for the current study. The beams can be considered as independent incoherent lasers. Therefore, the cosine term in Eq. (6) would be 0 through the pulse duration, and irradiance within the overlapping area simplifies to $$ I(P)=I_{1} (P)+I_{2} (P),~~ \tag {9} $$ and the electric field to $$ E(P)=\sqrt {I_{1} (P)+I_{2} (P)} =\sqrt {E_{1}^{2} (P)+E_{2}^{2} (P)}.~~ \tag {10} $$ If the electric field within the overlapping area ${E}_{\rm overlap}$ exceeds the threshold damage field strength, and the electric field in the independent region ${E}_{\rm independence}$ is insufficient for breakdown, then the ablation area is determined by the overlapping area. Heat diffusion may be neglected with the laser focal spot size,[11] and the damage area will be restricted within the overlap area. The laser-induced damage probability was measured by adjusting the laser energy of single and PO-DB sets (Fig. 5). The initial laser fluence density was adjusted as 25 J/cm$^{2}$. In such a case, one laser shot of the single beam and PO-DB pulses both could damage soda-lime glasses with 100% probability. When the laser fluence density was reduced to $\sim $20 J/cm$^{2}$, the single beam laser occasionally failed to damage the samples while PO-DB pulses kept 100% damage probability. When the laser fluence density was reduced to $\sim $16 J/cm$^{2}$, the single beam laser can no longer damage the sample, while PO-DB pulses held 93% damage probability. A clear fluence discrepancy exists between the single beam and PO-DB to damage the samples. The results indicate that within the overlap area, two incoherent laser beams worked together and ablated the glass surface.
cpl-37-1-015202-fig5.png
Fig. 5. Damage probability curve for soda-lime glasses by one laser shot of single beam and PO-DB pulses. Normalized laser fluence of 1.0 corresponds to fluence density of 25 J/cm$^{2}$, and then reducing the laser fluence by adjusting polarization through half-wave plate and Glan–Taylor polarizer.
cpl-37-1-015202-fig6.png
Fig. 6. Examples of ablation craters on soda-lime glass, as measured with 3D interferometer profiler for (1) PO-DB, (2) single beam, and (3) ER-DB. Ablated crater 3D shape is shown beneath the bright-field photo. Dimensions and Na I atomic line emission intensities are shown in Table 2.
Figure 6 shows the three-dimensional (3D) profile measurements for craters on the microscope slide glass surface, with spatial parameters and a signal intensity shown in Table 2. We chose a Na atomic emission at 590 nm as the probe for LIBS detection. The proposed PO-DB system ablated a significantly smaller area compared with single-beam ablation. The energy distribution of the multimode laser beam is unstable, and some experimental cases ablated the entire region of the dual-beam (ER-DB). However, PO-DB and ER-DB craters were significantly shallower than the SP ablation. Signal intensity ER-DB $>$ single beam $\approx$ PO-DB LIBS, whereas PO-DB LIBS ablated less material than single-beam LIBS. This effect results from the overlapping area absorbing laser energy more efficiently than the single beam area. When the critical plasma density is achieved, the PO-DB plasma absorbs more photons from two laser beams because the plasma dimension is larger than the ablated material volume.
Table 2. Ablated crater dimensions on soda-lime glass surface for R2, PO-DB and ER-DB cases (average and standard deviation values of three runs). Na I emission intensity at 589.95 nm is also listed.
X-sample plane Standard deviation Y-sample plane Standard deviation Ablated area Depth Na I emission intensity
(µm) (µm) (µm) (µm) (µm$^{2}$) (nm) (counts)
PO-DB 4.13 0.37 3.50 0.42 45.41 98.79 541
R2 path 5.06 0.05 4.66 0.07 72.49 411.56 554
ER-DB 5.50 0.22 5.76 0.31 99.44 165.51 612
Single-mode Gaussian laser beams are unsuitable for PO-DB ablation because the spatial fluence distribution weakens the two beams' spatial addition. Multimode lasers have a more even energy distribution, although craters ablated by single beams are somewhat larger. The energy distribution of the multimode laser was not as stable as the single-mode laser, which resulted in occasional PO-DB and ER-DB craters. The ability of the laser fluence to reach the damage threshold of the material is usually used to evaluate the laser's ability to ablate material. The fluence distribution determines the range of ablation. If the distance between high fluence regions is sufficiently large (e.g., when two beams irradiate different positions), then the area in the high fluence area will remain undamaged. If multiple high-fluence regions are synchronized in time and are sufficiently spatially close, the area in the high-fluence area would be damaged (e.g., when a single multimode beam is applied). When the low-fluence region is established outside the high-fluence region, the ablation area can be limited to the high-fluence region, such as the PO-DB in this study. The multimode laser in this experiment is distributed in an annular mode, but the energy of every specific region is unstable. The regional instability causes an instability of energy addition. Ideally, the fluence of the overlap region is significantly (two times) higher than that of the peripheral independent areas, which results in a reduced ablation area. However, in some cases the energy difference between the overlap region and the peripheral area is not very large, which results in ER-DB. Thus, a flat-top laser beam would be ideal for PO-DB because the energy addition, and hence the ablation area, would be more precisely predictable. Processing materials, including metals and transparent materials, with short duration pulse could result in a smaller ablation area. Because laser energy absorption is increased by multiphoton processes for opaque materials, the energy cannot instantly transfer from the electron gas to the ion grid. Thus, the plasma vapor and melting model become invalid, and the phase explosion creates a high pressure mixture of liquid droplets and vapor nearly simultaneously. In this case, $C_{\rm e} {T}_{\rm e} /t \gg \gamma {T}_{\rm e}$, and Eq. (2) should be expressed as $$ \frac{C_{\rm e} \partial T_{\rm e}^{2} }{\partial t}=2I(t)A\alpha \exp (-\alpha z),~~ \tag {11} $$ where the irradiance threshold for the $I_{\rm damage}$ ablation is $$ I>I_{\rm damage} \sim \frac{\rho {\it\Omega} }{\tau }.~~ \tag {12} $$ When ablating transparent materials, multiphoton ionization becomes more important as the laser duration decreases. The fluence threshold ${F}_{\rm damage}$ corresponding to the plasma critical density ${n}_{\rm cr}$ is $$ F_{\rm cr} =\frac{2}{\alpha }\ln \left(\frac{n_{\rm cr} }{n_{0} }\right),~~ \tag {13} $$ where $n_{0} =\int_{-\infty }^\infty {P(I)dt}$ is the total number density of electrons produced by multiphoton ionization. For a temporal Gaussian laser beam, the avalanche is assumed to start at the peak of the pulse, $$\begin{alignat}{1} n_{0} &=N_{\rm s} \int_{-\infty }^\infty {\sigma_{m} (\frac{I(t)}{\hslash \omega })}^{m}dt\\ &=\sigma_{m} N_{\rm s} \left(\frac{I_{0} }{\hslash \omega }\right)^{m}\left(\frac{\pi }{\ln 2}\right)^{1/2}\frac{\pi }{4}~~ \tag {14} \end{alignat} $$ where $\sigma_{m}$ is the cross section of $m$-photon absorption, ${N}_{\rm s}$ is the solid atom density, and $\omega$ is the laser frequency. When laser energy reaches the threshold, multiphoton ionization creates an electron density, i.e., the critical plasma electron density, and laser light is strongly absorbed. Femtosecond lasers could be used with the proposed PO-DB system, in which coherence should be avoided to obtain a restricted ablation area. Two femtosecond laser beams with different colors[13,14] could achieve the PO-DB concept. For opaque materials, heating of the overlapping area would be the result of both laser beams, whereas for dielectric materials, ablation occurs when the electron cloud produced by multiphoton ionization from both beams reached the critical density. The proposed PO-DB LIBS set was verified to reduce the ablation crater size and to increase the LIBS spatial resolution for micro-elemental analysis. The two laser beams were focused on partially overlapped locations on the material surface, such that irradiance within the overlapping area could reach the material fluence threshold, whereas the fluence of the independent region was lower than the threshold. A numerical simulation of the energy distribution of the PO-DB predicted the constrained overlap ablation region, with a damage threshold expressed as 55% of the highest fluence of the region. The area with a fluence was higher than the threshold changes along with the distance between two beams. We show that the ablation area for aluminum could be reduced to 12.25% of the single-beam ablation area. A glass damage probability experiment proves that within the overlap region, two incoherent beams work together to generate the plasma. The ablation volume of the glass surfaces could be reduced without decreasing the LIBS signal. In principle, superior results could be obtained with incoherent flat-top femtosecond lasers.
References Laser-Induced Breakdown Spectroscopy Microanalysis of Trace Elements in Homo Sapiens TeethAtomization efficiency and photon yield in laser-induced breakdown spectroscopy analysis of single nanoparticles in an optical trapEffect of Sample Preparation on the Discrimination of Bacterial Isolates Cultured in Liquid Nutrient Media Using Laser-Induced Breakdown Spectroscopy (LIBS)Laser beam profile influence on LIBS analytical capabilities: single vs. multimode beamFemtosecond, picosecond and nanosecond laser ablation of solidsMetal Ablation with Short and Ultrashort Laser PulsesSilver jewelry microanalysis with dual-pulse laser-induced breakdown spectroscopy: 266 + 1064 nm wavelength combinationMicroanalysis of silver jewellery by laser-ablation laser-induced breakdown spectroscopy with enhanced sensitivity and minimal sample ablationOptical ablation by high-power short-pulse lasersNanosecond-to-femtosecond laser-induced breakdown in dielectricsBulk and surface laser damage of silica by picosecond and nanosecond pulses at 1064 nmDetermination of Absolute Cross-Sections of Nonresonant EUV-UV Two-Color Two-Photon Ionization of HeSelective bond breaking of CO 2 in phase-locked two-color intense laser fields: laser field intensity dependence
[1] Alvira F C et al 2010 Appl. Spectrosc. 64 313
[2] Purohit P et al 2017 Spectrochim. Acta Part A 130 75
[3] Gamble G R et al 2016 Appl. Spectrosc. 70 494
[4] Lednev V et al 2010 J. Anal. At. Spectrom. 25 1745
[5] Chichkov B N et al 1996 Appl. Phys. A 63 109
[6] Leitz K H et al 2011 Phys. Procedia 12 230
[7] Mo J et al 2014 Appl. Opt. 53 7516
[8] Mo J Y et al 2014 Chin. Opt. Lett. 12 083001
[9] Stuart B C et al 1996 J. Opt. Soc. Am. B 13 459
[10] Stuart B C et al 1996 Phys. Rev. B 53 1749
[11] Smith A V and Do B T 2008 Appl. Opt. 47 4812
[12]Menzel 2007 Photon. Linear Nonlinear Interact. Laser Light Matter 2nd edn (Berlin: Springer)
[13] Fushitani M et al 2014 19th International Conference on Ultrafast Phenomena (Optical Society of America, 2014) paper 08.Tue.P2.5
[14] Endo T et al 2017 Phys. Chem. Chem. Phys. 19 3550