1. Introduction

There is a large body of work that depicts the benefits of ultrasonic excitation in material processing. Acoustic energy has been shown to be effective in liquid phase reactions [1–4] and in bulk solid reactions [5–14] where enhanced microstructure within the solid has been attributed to the generally agreed principle that acoustic energy enhances diffusion rates via mechanical forces. Recent experiments with laser generated ultrasound (> 10 MHz) have shown that diffusion rates of molecular adsorbates on crystalline surfaces can also be enhanced and the enhancement is not the result of a mechanical “knock-on” force, but due to the alteration of the potential energy barrier that binds the adsorbate [15]. The physical perspective, derived by modeling, suggests that the transient atomic strain that is apparent in a propagating ultrasonic (i.e., atomic displacement) wave (e.g., surface acoustic or Rayleigh waves) induces the action [16–21].

The goal of this work is to explore the potential for laser-induced ultrasound to enhance a simple order-to-disorder phase transition by non-thermal means. Enhanced crystallization rate of metallic glasses by bulk ultrasonic (US) excitation (~1 MHz) has been demonstrated in the literature by Ichitsubo et al. [11,22]. In that work stochastic resonance is suggested as the dominant mechanism; in that scenario, the relevant resonance frequency is not associated with the time scale of atomic vibrations (typically 10¹²-10¹⁴ Hz in solids); rather, it is believed to be associated with the time scale of atomic jumps. The utility of ultrasonic excitation to influence atomic diffusion is well-established in the context of applications (e.g., welding [12,13] and wire bonding [6]). For example, in the work of Gunduz et al. [5] on ultrasonic-assisted welding of zinc and aluminum, diffusivity enhancement on the order of 10⁵ have been reported in the presence of ultrasound at 20 kHz frequency. The generation of vacancy defects induced by high strain-rate oscillations is frequently cited as a mechanism for enhanced diffusion rates.

In the light of mechanisms most often cited in the literature, there are three reasons to believe that laser-induced ultrasound may offer unique benefits in material processing. Surface acoustic waves (SAWs) generated by pulsed laser excitation are characterized by (1) high bandwidth (approximated by the inverse of the laser pulse width; i.e.., the bandwidth for a 1 ns pulse is 1 GHz), (2) large strain rate, and (3) precise control over the “source” location. The last point is critical because the delivery of short-wavelength strain waves is limited by attenuation as a function of distance. The attenuation increases with frequency as f² or f⁴ in crystalline, poly-crystalline, and glassy systems [23–25]. A laser-based ultrasonic source also permits a degree of spatial separation (50 μm in this experiment) from traditional bulk excitation approaches (e.g., piezoelectric, typical separation of mm or greater) and thereby allow the effect of the US wavelength to be explored.

Amorphous MoS₂ is the candidate material chosen for this investigation because of the variety of industrial applications that rival graphene [26,27]. Atomically thin MoS₂ has been a subject of great interest because of its unique electrical (i.e., 2D material) and optical properties [27,28]. Even though the growth of thin film crystalline MoS₂ over large areas is possible with chemical vapor deposition (CVD) [29], deposition temperatures above 500°C are required for good crystallinity [30]. These temperatures are not workable for some commercially relevant substrates. For example, recent work by McConney et al. demonstrated laser crystallization of sputtered MoS₂ thin films is possible on polymer substrates [31]. Also, laser processing of exfoliated MoS₂ flakes [32,33] have further shown the utility of lasers to modify films that are already in the crystalline state.

Laser direct-write processing enables phase conversion at low bulk temperatures. With the addition of process control, this approach enables control over the local properties of the film and opens the door to a number of interesting device applications that integrate band-gap engineered photonics and 2D electronics, possibly on flexible media [28,34–36]. To enable such technology requires a combination of controlled laser processing technologies [37], direct-write patterning capability, and process development for laser crystallization and material removal at low bulk temperatures.

In this work, we demonstrate that the use of ultrasonic waves can reduce a process temperature. Using an RF sputtered thin film, where uniform coverage and density can be reliably produced, we show the effect by characterization of the threshold laser power for crystallization and the subsequent increase in crystalline size and density with continued heating. High resolution Raman spectroscopy is used to characterize crystal formation and we relate these results with those from optical microscopy and optical profilometry for possible future use as an in situ process control. The crystalline properties are characterized by Raman and TEM, and thermal modeling (COMSOL) is used to ensure that the applied energies (i.e., thermal, US) are separable. Our results show that there is a measurable reduction in the laser power required for crystallization and densification processes. The experiment demonstrates the effect of ultrasonic excitation on material transformation in a definitive manner which should help to elucidate the underlying physics.

2. Methods

Laser processing: The crystallization experiments were performed in a chamber under low vacuum (~10 Torr). A 473-nm heating laser is focused to a spot size less than 3 μm diameter on the sample using a Mitutoyo M Plan APO NUV 20x microscope objective. The spot size measurement is based on the smallest resolvable crystalline line produced at the lowest heating laser power above the crystallization threshold. The laser beam from a pulsed NdYVO₃ (Spectra Physics J40-BL6S-355Q, 355 nm wavelength, 35 kHz, 6 ns pulse width) laser is split into two beams of equal intensity using a + 1/-1 order phase mask (Ibsen FBG 850 nm); both beams are directed to the center of the microscope objective and straddle the heating laser beam with equidistant offset (approximately 50 μm – measured by generating and measuring burn marks on Kapton tape with an optical microscope, just above the damage threshold). Moreover, the spatial intensity distribution of the ultrasonic laser beams is altered by a long focal length cylindrical lens placed prior to the phase mask. At the focus, the beam shape is an elongated ellipse (approximately 20 μm length) that is oriented parallel to the stage-scan direction. Consequently, the intensity distribution at the process region includes heating laser with Gaussian profile that is framed by two elliptically shaped beams (ultrasonic laser sources). The laser power in both the heating and ultrasonic lasers is set and controlled by a fast (~1 MHz) Pockels (CONOPTICS 350-105 with CW laser; CONOPTICS 350-50 with pulsed laser) that acts as an electro-optic shutter. In addition, the power in both laser beams is independently measured and the information stored after a sequence of pattern-processing runs. The experimental processing station which comprises, the lasers, a computer controlled XYZ motion stage (Aerotech AVL 125(Z), ALS 130-150H (X,Y), < 5 μm positioning accuracy), power meters, and control systems is a fully automated “lights-out” system. The process-patterns are designed using a CAD program and converted to CNC machine code, G-code (MasterCam v9.1), and the G-Code is further post-processed by a compiler that inserts control commands within the pattern-command structure for shuttering and measuring the laser power. The automated approach insures processing uniformity and verification for sample areas up to 100 cm², but in these experiments the sample area processed is closer to 0.25 cm². A single exposure pattern consists of a line 200 μm in length and 3-25 μm in width. All exposed areas are patterned at a velocity of 25 μm/s.

The experiment is performed over a range of heating laser powers. Each heating laser power data set consists of at least two processed lines. (1) A processed line where only the heating laser power is present (called the “unassisted” run) and (2) a processed line where both the heating laser and ultrasonic laser are present (with the ultrasonic laser fixed at 30 mW total incident power at 35 kHz; called the “assisted” run). A typical total sample series has 60 processed lines.

This configuration is chosen to ensure that the steady-state thermal contribution from the ultrasonic lasers, at the heat location, is small compared to that of the heat laser. To verify this, a heat transfer model (COMSOL) is used to estimate the temperature profile of the heat laser. The thermal parameters of the film and substrate are calculated using low temperature values from the literature for amorphous MoS₂ [38] and single-crystal YSZ [39], neglecting variation with temperature as a first-order approximation. Moreover, the model was calibrated by a physical experiment where the front surfaces of several optically transparent samples are coated with temperature-sensitive paint (Tempilaq Temperature-Indicating Paint). The paint liquefies at the specified temperature, irreversibly altering its optical properties. The physical calibration experiment involved drawing a line pattern across the sample through the back (uncoated) surface with the heat laser set at power levels commensurate with the experiments (210 mW). For a given temperature sensitive paint, laser power, and pattern velocity, a visible track (a region with a halo due to the liquification/solidification process) is produced for parts of the painted area that reach the defined temperature. The results are shown in Fig. 1(a).

 

Fig. 1 Calibration of thermal profile and crystalline quantity. (a) A temperature profile calculated using COMSOL determined by measuring halo generated on sample coated with Tepilaq Temperature Indicating Liquid (example shown in inset). Model was calibrated using paint with transition temperatures at 200°F (366 K), 300°F (422 K), and 500°F (533 K) paint, with isotherm size denoted as red points. The green point indicates a point at the edge of the crystalline region (12 μm from the center of the laser spot) (b) A similar experiment using the pulsed laser; no measurable halo was observed. (c) A comparison between a white light image of the crystalline line (top) and a map of the spectrally integrated Raman intensity (bottom, with corresponding color-bar in arbitrary units on the right). (d) A plot of the spectrally- and spatially-integrated Raman intensity as a function of crystallized line width for 30 mW of tapping power (blue) and no tapping (red), with corresponding linear fits. The width values are averages of ten measurements across each line at a given heating laser power; the error bars indicate the standard deviation of these values.

Download Full Size | PDF

These calibration experiments were also used to refine the results of the COMSOL model with regards to the local temperature produced by the heating source laser. First, temperature calibration experiments were performed using sensitive paint at 200°F (366 K), 300°F (422 K), and 500°F (533 K) for the same incident laser power of 210 mW. Second, the width of the three calibrated isotherms were measured using an optical microscope. Third, the information from the width data is used to calibrate the COMSOL model by varying the background temperature and absorption coefficient until the measured location of the three isotherms approximately matched the corresponding temperature values in the model. The result of the calibrated COMSOL model is shown in Fig. 1(a). These results show that 50 μm from the heat source (where the ultrasonic laser is positioned), the local temperature rises from room temperature approximately 100 K.

A similar calibration experiment was performed with the pulsed ultrasonic laser, but no halo was observable, presumably because the heat is localized to the laser spot (as expected). To better observe and measure the halo region, the experiment is performed on the front surface of a film on a Foturan glass substrate (whose thermal conductivity is lower than YSZ and would lead to more spread out, more easily resolvable isotherms). Half the sample is coated with the lowest temperature sensitive paint (200°F), then a single ultrasonic laser spot is focused to pattern a line at a glancing angle at the edge of the painted region. Figure 1(b) shows the paint is removed directly beneath the laser spot via ablation, but no halo is observed. The lack of a halo surrounding the patterned line indicates that the ultrasonic laser does not significantly (above 200°F) raise the steady-state temperature of the film away from the laser spot.

The thermal properties of the film change significantly going from the amorphous phase to the crystalline phase [38], making it difficult to model the temperature directly in the vicinity of the laser spot. However, the COMSOL model is likely to be accurate outside the crystalline region, where the material is amorphous and having thermal properties as such. Consequently, it can be used to estimate the temperature reached at the edge of the crystallized line. The temperature extrapolated to a point at the edge of the crystalline line (indicated by a green dot on both the temperature profile and inset image in Fig. 1(a)) is very close to the minimum temperature required to crystallize amorphous MoS₂ film (around 750 K; near the lowest process temperature where crystallinity is observed in CVD grown films by Laskar et al. [29], around 500°C). Note that at the feed rate under consideration, the system is in quasi-steady state conditions.

Laser ultrasonic excitation: The abrupt deformation of a material, in the thermoelastic regime, generates a propagating acoustic wave that permeates the material. Wave modes that travel into the material are called bulk modes (i.e., shear and longitudinal) while modes that travel along an interface are called surface acoustic waves or Rayleigh waves. Both interface and bulk ultrasonic excitations are produced with each having a characteristic velocity. In this experiment, the substrate is crystalline (YSZ, 300 μm thick, (100) orientation) and the incident angle of the ultrasonic laser is placed slightly oblique (0.31° - determined by tracking the in-plane “drift” of the pulsed laser spots ± 200 μm away from the focal plane in 100 μm steps, measured via burn marks on Kapton tape) to insure that the Poynting vector of the bulk reflected wave from the substrate rear surface affects an area outside the phase conversion area [40]. Consequently, we anticipate the primary source of ultrasonic energy at the phase conversion area to be in the form of interface waves (i.e., surface acoustic waves – SAWs). For a thin film on a thick substrate, the majority of the SAW energy will be located within the substrate material; in essence, the propagating excitation mode is likely to be a traveling SAW at the YSZ surface which is perturbed by the thin amorphous MoS₂ over layer [41]. The spatial intensity distribution of the incident US laser is made elliptical (~7 μm x 20 μm) to enhance the propensity for generating US plane waves, however, calculations of the Rayleigh distance show that the surface waves in fact do not propagate as plane waves but have some curvature even at the upper frequency limit of our pulsed laser source (~125 MHz). This fact abates the notion that every laser spot size (~3 μm) is perturbed by US excitation pulses only when the heating laser is present. The curvature of the propagating US wave affects the sample areas before and after the laser heated spot, albeit at a lower US wave amplitude and local temperature.

Sample growth: The thin films are grown using RF sputtering [42]. A typical sample is nominally comprised of 10 nm MoS₂ sputtered on a 300 μm thick coupon of yttria-stabilized zirconia (YSZ). The thickness is based on a calibration curve of deposition time to thickness, derived based on previous measurements of films with a stylus profilometer [43]. Laser crystallization experiments have been performed on a number of different substrates and it was found that the power threshold for crystallization is strongly dependent on the thermal conductivity of the substrate material. YSZ is selected for the following reasons: (1) the laser power necessary to crystallize the film is appropriate based on the maximum power output of our heating laser, (2) it can be commercially procured as a polished/flat single crystal (MTI corp. 1 cm², (100), both sides polished) to insure uniform surface coverage during deposition and to make TEM measurements easier, (3) MoS₂ exhibits significant absorption at 473 nm [27], while the YSZ substrate does not [44]; consequently, most of the absorption and heating is thought to originate in the MoS₂ film. The amorphous nature of the deposited film has been confirmed by the lack of Raman peaks associated with crystalline MoS₂.

Line width, height, and crystalline quantity: Two approaches have been used to measure the spatial properties of the crystallized material and one for characterizing crystalline formation: (1) measurement of the line width is performed using an optical microscope (Olympus BX51) in the bright field mode with an Olympus NEO Splan 100x objective. A U-25LBD filter is used to improved contrast, (2) for height measurements, an optical profilometer is used (ZYGO NewView 6K) which is outfitted with a large array camera (992x992 pixels) and a standard 50x objective with 2x zoom lens. Crystalline quantity and quality is verified by Raman spectroscopy. Raman measurements are time consuming for large area analysis; we have verified that the spatial width of the line as measured under a white light optical microscope can be used as an alternative metric for crystal growth and thereby expedite data analysis. This relationship is demonstrated qualitatively in Fig. 1(c) by comparing the optical microscope image with a spatial map of the integrated Raman intensity; and is verified quanitatively in Fig. 1(d) by plotting the integrated Raman intenisty versus the measured spatial width from the white light image.

Raman spectroscopy: Raman measurements are performed using a Renishaw inVia Raman Microscope system with a 100x objective. The excitation source is a 514 nm laser delivering 2 mW power on target. A 1800 mm⁻¹ grating with a 50 μm slit provides a spectral resolution of ~1 cm⁻¹. For each laser power, the integrated Raman signal was measured across a 5 μm × 30 μm strip through the center of the line. Large area mapping measurements are typically performed with 5 s integration time (e.g., 100 acquisitions at 5 s in Fig. 2(c), 1 acquisition per pixel at 5 s in Figs. 2(d) and 2(e)). However, to get cleaner signals for peak fitting, a higher integration time is used (3 acquisitions at 20 s integration time in Fig. 6). Long integration time can result in partial crystallization of the amorphous film. However, we detected no measurable change in the spectrum as a result of the Raman laser. The Raman spectrum of crystalline MoS₂ observed in our experiments is characterized by two primary Raman peaks and the four disorder-activated peaks [45]. To compare Raman data between processed pattern lines and arrive at a value that is descriptive of the total crystalline quantity (without bias for/against disorder) the spectrum is first normalized and is then integrated from 300 cm⁻¹ to 500 cm⁻¹ after subtracting a polynomial fit baseline between these two points. The integrated sum includes the two primary Raman peaks and the four nearby disorder peaks. A spatial map of this quantity indicates the presence of crystalline matter (as shown in Fig. 2(d)). To obtain a single number indicative of the total amount of crystalline material generated by a given laser treatment, the spectrally integrated Raman signal is then spatially integrated across some representative slice through the crystalline line.

 

Fig. 2 Impact of US on crystalline quantity. (a) A cartoon image of the configuration of laser spots, showing the heating laser (473 nm, CW) flanked by two pulsed tapping laser spots (355 nm, 6 ns pulses at 35 kHz). (b) Shows a typical process region; the three laser spots are moved across the sample parallel to the vertical direction, creating a crystalline line beneath the heating laser spot and a trench beneath each tapping laser spot. (c) A typical Raman spectrum (taken at the center of the line shown in (b)). (d) A spatial map of the integrated Raman spectrum (Across the spectral range shown in (c)) within the dotted line designated in (b). (e) A plot of the integrated Raman signal squared, spatially integrated across 5 μm × 30 μm boxes for crystalline lines generated at a range of heating laser powers; unassisted (heating laser only, blue circles) and assisted (heating laser with tapping laser present – red diamonds) data is shown. The intensity is squared to apply a model fit, as discussed in Section 3.3; this fit is shown for each curve. Based on the intercepts of the fitted curves, crystallization occurs at lower heating laser power in the assisted experiment (155.8 mW ± 5.6 mW, 95% CI) compared to the unassisted experiment (171.1 mW ± 5.3 mW, 95% CI).

Download Full Size | PDF

TEM: TEM specimens were prepared by first using a gallium focused ion beam (FIB) lift-out method on a FEI Strata 400. Final thinning was performed at 2 kV.

3. Results

3.1 Estimated characteristics of induced SAWs

Surface acoustic waves are generated in the film and substrate by rapid heating induced by absorption of the pulsed laser. It can be assumed that absorption occurs entirely within the film (for bulk MoS₂, α = 8.8 × 10⁵ cm⁻¹ [46]) given that YSZ is transparent at 355 nm (below the band gap, 5.8 eV, absorption will occur primarily via disorder and impurities) [47].

The relevant time and length scales associated with the thermal profile can be estimated based on the thermal diffusivity of MoS₂ (3 × 10⁻⁷ m²/s for amorphous MoS₂ [38]) and YSZ (7 × 10⁻⁷ m²/s) [39]. During the length scale of a single laser pulse (6 ns), the thermal diffusion length is 84 nm in MoS₂ and 130 nm in YSZ; during the time between pulses (29 μs), the thermal diffusion lengths are 5.8 μm and 9 μm, respectively.

A single laser pulse is divided between two ellipitical spots, approximately (~7 μm x 20 μm), over a total area of 880 μm². The total volume heated by the laser pulse will be approximately be defined by the thermal diffusion length (84-130 nm; resulting in 80-120 μm³). The effective heat capacity over this area will be a compromise between that of MoS₂ (1.8 J/cm^3⋅K [38]) and YSZ (2.7 J/cm^3⋅K [39]). Using low-temperature values and ignoring the temperature dependence of thermal parameters, a single pulse (0.86 μJ) would raise the local temperature of the material by 850-2200 K, estimating 37% absorption based on the optical parameters of MoS₂. Clearly the film would be damaged in this regime, however YSZ is robust at high temperature and would survive the lower end of this range. Under the assumption that the response of the YSZ surface can be treated in the thermoelastic regime, the strain rate would be estimated at 1.0-2.6 × 10⁶ s⁻¹ (using 7 × 10⁻⁶ Κ⁻¹ for the coefficient of thermal expansion of YSZ [48]). Although this is likely an over-estimate, it is not an unreasonable number for laser-induced stress waves (~10⁷ s⁻¹ was reported for 3.9-ns pulses on metallic films [49], albeit at significantly higher fluence).The depth of an acoustic wave traveling at the surface of the sample would be significantly longer than the film thickness and will be governed primarily by the acoustic properties of the YSZ. Along the [110] direction (parallel to the surface of the substrate), the speed of sound in YSZ is 7 km/s for longitudinal waves and 3 km/s for transverse waves [50], corresponding to wavelength components on the order of 70 μm and 30 μm for SAWs with a bandwidth of 100 MHz.

3.2 Rate and onset of crystallization

Figure 2 presents a composite image that describes key elements of the experiment and results. Figure 2(a) depicts a 3D model of the multi-laser interaction zone with two pulsed US laser spots straddling a CW heat source laser (center). Figure 2(b) shows a high magnification optical microscope image of the crystallized area (the line depicted by the rectangular box). Figure 2(c) shows the Raman spectrum (blue) taken of the line in the box (Fig. 2(b)). Crystalline MoS₂ is often characterized by the presence of two distinct Raman peaks: the E¹_2g peak near 383 cm⁻¹ and the A_1g peak near 408 cm⁻¹ [51]. Figure 2(c) also shows the spectral decomposition into bands which are symmetry forbidden in a single crystal material, but can become activated by the presence of disorder or small crystallite size; as identified by Mignuzzi et al. [45], these peaks correspond to the TO(M) mode near 357.7 cm⁻¹, the LO(M) mode near 377.0 cm⁻¹, the ZO(M) mode near 411.9 cm⁻¹, and the second-order LA(M) mode near 455.2 cm⁻¹.

Figure 2(d) presents a 3D spatial map of the normalized and integrated Raman data within the box (details to be found in the Experimental Section). The results show that the line is near uniformly crystallized. Using the integrated Raman data as a measure of crystalline content, Fig. 2(e) presents data of the crystalline content as a function of CW heat source laser power for the case of US assisted and unassisted crystallization. The lines fitted to the data are based on a model where the boundary between the crystalline and amorphous region is defined by the regions at which a minimum threshold temperature for crystallization has been reached. The width, w of the line can be related to the total Integrated Raman intensity, I. The temperature profile, T, as a function of distance, r, from an incident laser spot is expected to fit a Gaussian profile, as it would be true for a slowly moving laser beam incident on a homogeneous material with absorption at the surface [52]. Therefore, we can write the following relation,

 

Fig. 3 Growth of crystalline width with heating power. (a) A plot of line width as a function of CW heating power assisted by pulsed laser tapping (red curve, 30 mW average power) and unassisted (blue curve). Points A, B, and C are three sets of assisted/unassisted points, each pair at roughly equivalent heating power (185 mW, 192 mW, 216 mW, respectively); properties of these lines will be compared and contrasted in Figs. 4 and 6.

Download Full Size | PDF

3.2 Changes in surface topography and thickness

The results in Fig. 2 show that in the higher laser power range (200-220 mW) both the US assisted and unassisted data merge. There is a plateau which is followed by a merging of the two data sets. The data suggest that the ultrasonic induced effect subsides with continued added CW laser power (i.e., heating). Measurement of the size of the crystallized area does not capture the entire physical picture in this regime. Densification (at low power) and removal of oxide (at higher power) are also expected to occur, changing the thickness of the film within the process region.

To provide a better understanding of the underlying physics, three data points, labeled A, B, and C are identified in Fig. 3 and the associated crystallized lines are further analyzed using optical profilometry (Fig. 4), TEM (Fig. 5), and Raman (Fig. 6).

 

Fig. 4 Surface topography of crystalline lines. (a) Cross-sections of the film thickness as measured with an optical profilometer, shown for a range of heating laser powers in the assisted and unassisted experiments; the baseline for each cross-section is offset by the heating laser power (in mW) to show the variation of topography with power. (b) 3D optical profilometer images of the laser-crystallized lines identified in Fig. 3. Tapping-assisted (bottom) and unassisted (top) at heating power values – 185 mW, 192 mW, and 216 mW, corresponding to points designated A, B, and C in Fig. 3, respectively. A TEM cross-section of the unassisted line B in Fig. 6(b) illustrates that the film maintains contact with the substrate after crystallization, showing this topography is not caused by delamination.

Download Full Size | PDF

 

Fig. 5 TEM images. Image of sample processed with 216 mW laser power; corresponding to point C (heating-laser only) on Fig. 2. Representative pictures, (a) away from the laser processed region (as-sputtered film) and (b) on the laser line. Analysis of the layers in processed region was found to have a periodicity of around 0.64 nm, as would be expected for crystalline MoS₂.

Download Full Size | PDF

 

Fig. 6 Evolution of the Raman spectrum with varying heating laser power. Shown for assisted (red diamonds) and unassisted (blue circles) data sets; each point is based on a fit to a spectrum taken at line center, using a six-peak model that includes the nearby disorder-activated peaks. (a) and (b) show the Raman peak positions of the A_1g and E¹_2g bands, respectively.

Download Full Size | PDF

Figure 4 presents cross-sectional data across several lines as measured by an optical profilometer. Moreover, for the three marked regions A, B, and C we present the full 3D map of the topography for both the unassisted and assisted cases. The cross-section data are plotted as a function of CW heat laser power encompassing the range just above the mark C in the data shown in Fig. 3 (each curve represents film height in nm, but is offset on the y-axis as a function of laser power to show variation with changing power). One obvious surprise is the fact that the crystallized material appears to have swelled, raising concern for delamination. Figure 5(b) presents a cross-sectional TEM image of the film, processed at the laser power associated with C (216 mW), which show the film to be intact on the substrate but comprised of a disordered array of polycrystalline matter. The highly misoriented small crystallites indicated by Fig. 5(b) and swelling of the film suggests an increase in the porosity of the film from laser processing in this power region. Figure 5(b) can be compared to Fig. 5(a), where the unprocessed material is shown to be dense, but lacking crystalline “stacking” features which are indicative of crystallization.

The following trend can be discerned for the cross-section results of Fig. 4: (1) at low laser power, initial crystallite formation coincides with swelling of the film, (2) with further increase in the heating power, the center of the crystallized film thins (i.e., reduction in the swelling) forming a canyon shape structure (starting at marking locations A and B). Figure 4 illustrates that both transformations – the initial crystallite formation and the “canyon” formation – corresponding to densification – appear at lower CW power when US excitation is present.

Figure 5(b) shows a cross-section TEM of the “canyon” region which shows the material undergoing an ordering process. Given that the MoS2 density typically increases from the polycrystalline (3.8 g/cm³ or 3.95 g/cm³) to single crystal state (4.93 g/cm³) [53], this change in film thickness is unsurprising.

3.3 Effect of thermal power on Raman signal

Work in the literature has demonstrated that the positions, linewidths, and splitting of the two prominent Raman bands of MoS₂ provide information about the properties of the crystalline matter within the film [45,51,54,55]. High resolution Raman measurements were produced at line-center as a function of CW laser heat power; Fig. 6 is a composite data of Raman signal features plotted as a function of heating laser power. Plotted for the primary Raman bands E¹_2g and A_1g are the peak positions (Fig. 6(a), 6(b)), splitting (Fig. 6(c)), peak linewidths (Figs. 6d and 6(e)) and the total (all bands) Raman intensity (Fig. 6(f)). Moreover, the data is plotted for both the US assisted case (red diamonds) and unassisted case (blue circles). Simple fits are used to draw attention to trends. With increasing CW laser heat power, there is a blue-shift of the primary Raman active peaks (Figs. 6(a) and 6(b)), which coincides with a reduction in the band peak separation (Fig. 6(c)). The former can be indicative of increasing compressive strain [55], while the latter is consistent with thinning of the film [51]. Increased heating power also corresponds to narrowing of the Raman-active band width (Figs. 6(d) and 6(e)), which is consistent with increased crystallite size [45]. The trend lines do suggest that the US assisted results exhibit a narrower band width than the unassisted data set (Figs. 6(d) and 6(e); data in red are lower than blue), however the difference is clearly not large relative to the resolution of the measurement (quantified by the expected resolution of the system – ~1 cm⁻¹ – and the scatter of the data), so this conclusion – when considered in isolation – must be viewed with skepticism. On the other hand, the key result of this measurement is that increased laser power does coincide with crystallite growth for both the assisted and unassisted data sets. This provides confirmation that the densification process demonstrated in the optical profilometer data does at least partly correspond with crystallite growth. Figure 4 shows enhancement of thedensification process due to ultrasound, providing evidence that the crystallite growth process is also enhanced by the presence of ultrasonic excitation.

4. Discussion

The data indicate that the presence of ultrasonic excitation is beneficial in both (1) reducing temperature / power necessary for initiation of crystallization (indicated by the shift in the intercept evident in Figs. 2 and 3) and (2) increasing the amount of crystalline matter present at fixed temperature / power above the crystallization threshold. Furthermore, the crystallized film appears denser (demonstrated in Fig. 4) and exhibits less disorder and / or larger crystallite size (demonstrated by the lowered Raman peak broadening in Fig. 6).

The crystallization process requires both nucleation of initial crystallites and growth of the ordered phase, so any mechanistic explanation likely involves one or both steps. The lowering of the initial crystallization threshold indicates that US is beneficial to initial nucleation of crystallites, possibly by enhanced diffusion rates or by dynamically lowering the activation energy barrier for seed crystal formation. Crystallite growth is governed by diffusion, so enhancement of the diffusion rate is an important mechanism to consider at that stage of the process as well. Analysis of improved crystallization above threshold is more difficult – upon crystallization of the material, the thermal and optical properties change, which may increase the steady state temperature of the film for a given CW laser power.

The current data are not sufficient to precisely pin down the governing mechanism. However, the results presented here, evaluated in the context of the literature, suggest a few clues for further investigation. First, waves of sufficiently high amplitudes can generate vacancy defects – as discussed in Gunduz et al. [5]. A high concentration of defects would serve as nucleation sites and perhaps allow crystallization to occur more readily. Defects are not well-defined in an amorphous material, but the discussed mechanisms primarily rely on strain-induced distortion of the short-range order, which will be similarly present in the amorphous state. Second, US excitation could exhibit a transient effect on the energy landscape by lowering energetic barriers to enhance atomic motion or temporarily increase the process temperature via transient delamination from the substrate. Third - the threshold behavior observed here matches results seen by Ichitsubo et al. in metallic glasses [22]. Ichitsubo and associates observed crystallization of a metallic glass system below the normal glass transition temperature in the presence of US excitation.

A few key experiments would be beneficial in clarifying the physical mechanisms governing the effect observed in this work. The most important would be a thorough investigation of the crystallization enhancement as a function of US excitation bandwidth, i.e., laser pulse width. This analysis would help determine whether the enhancement exhibits resonant behavior or is characterized by some other clear frequency dependence. Similar experiments can also be performed with a mechanical transducer at fixed frequency for comparison.

5. Conclusion

In summary, this work has shown that laser-generated ultrasound can lower the temperature for crystallization for an amorphous nanoscale film. Moreover, the ultrasound-assisted process initiates densification of the polycrystalline film at lower heating power / process temperature. The unique properties of the laser-generated ultrasonic pulses (very high strain rate, high bandwidth) in conjunction with the precise spatial and timing control that can be provided by pulsed laser processing techniques, opens the door to novel material processing applications, especially with direct-write capability. While we have demonstrated the phenomenon experimentally, the fundamental microscale dynamics of the effect remain to be proven.