Dual Laser Beam Processing of Semiconducting Thin Films by Excited State Absorption

We present a unique dual laser beam processing approach based on excited state absorption by structuring 200 nm thin zinc oxide films sputtered on fused silica substrates. The combination of two pulsed nanosecond-laser beams with different photon energies—one below and one above the zinc oxide band gap energy—allows for a precise, efficient, and homogeneous ablation of the films without substrate damage. Based on structuring experiments in dependence on laser wavelength, pulse fluence, and pulse delay of both laser beams, a detailed concept of energy transfer and excitation processes during irradiation was developed. It provides a comprehensive understanding of the thermal and electronic processes during ablation. To quantify the efficiency improvements of the dual-beam process compared to single-beam ablation, a simple efficiency model was developed.


Introduction
In 1994, Hell et al. theoretically described the stimulated emission depletion (STED) microscopy, which was experimentally demonstrated a few years later and today enables to resolve structures smaller than 10 nm [1][2][3]. Here, a Gaussian laser beam excites carriers, which are subsequently depleted back to their ground state by a second donut-shaped STED beam with a specific material-dependent wavelength. Although the focusability of both laser beams is still limited by diffraction, the resulting fluorescent volume is significantly decreased to the zero-intensity center of the STED beam [4].
This approach is now also used for structuring different photopolymers with feature sizes in the low double-digit nm range [5]. In this case, polymerization in the outer, high-intensive region of the donut-shaped beam is suppressed by stimulated depletion of the electrons required for polymerization or by activation of inhibitors. Unfortunately, the technique is limited to certain photopolymers and specific light sources. As an alternative, the structures can be transferred to other classes of materials by replica casting. However, direct sub-diffraction structuring of semiconductors and dielectrics using the STED-technique would improve flexibility, reduce process complexity, and lower overall manufacturing costs [6,7].
In this context, an approach proposed by Engel et al. [8] is very promising for optically active materials like direct semiconductors. Briefly, comparable to STED microscopy, a diffraction limited Gaussian laser beam with sufficient photon energy E P is used to excite electrons into the conduction band (CB). A second temporally and spatially aligned donut-shaped beam with suitable E P in the range of the optical band gap energy E BG depletes these excited states and transfers the electrons back to the valence band (VB) by stimulated emission. The depletion only occurs in the high-intensive ring-shaped region of the beam, whereas the excited electrons in the zero-intensity center keep unaffected. This excited central region can subsequently absorb a third Gaussian laser beam by excited state absorption (ESA) at a wavelength otherwise transmitted in non-excited regions. Based on several processes such as intra-band absorption, avalanche effects, or coulomb explosion, the excess energy provided by this ESA-based process step is potentially suitable for subdiffraction structuring, although the single beams are subject to diffraction limited focusing.
Based on previous studies [9], the focus of the present work is on the ESA-based ablation as an important element of the proposed sub diffraction direct laser writing technique ( Figure 1a) [8]. Several studies already implemented an ESA-based microfabrication by a hybrid laser process of wide bandgap semiconductors such as fused silica, sapphire, or quartz [7,[10][11][12][13]. The non-thermal ablation by the excitation of electrons above the vacuum energy improves ablation quality and decreases debris formation on the surface. For these dual-beam processes, Excimer and Raman lasers with wavelengths in the UV-range and (ultra-)short pulse lasers are most commonly used to increase absorption through photodissociation and changes in transient absorption [7]. This process is also applicable to lower bandgap semiconductors like gallium nitride, lithium niobate, and silicon carbide [13][14][15][16][17]. In this case, the single laser beams already exceed E BG , but the excitation beyond vacuum energy for improved non-thermal ablation is only achieved by the dual-beam setup. Similarly, an increased absorption of a second, otherwise transmitted laser beam with E P < E BG , without exceeding the vacuum energy enables energy and cost efficient processing as already demonstrated for silicon and silicon carbide [18,19]. The key factor here is that commercially available, compact, and less expansive lasers in the visible and infrared spectrum can be used [18]. als 2021, 14, x FOR PEER REVIEW 2 of 14 by stimulated emission. The depletion only occurs in the high-intensive ring-shaped region of the beam, whereas the excited electrons in the zero-intensity center keep unaffected. This excited central region can subsequently absorb a third Gaussian laser beam by excited state absorption (ESA) at a wavelength otherwise transmitted in non-excited regions. Based on several processes such as intra-band absorption, avalanche effects, or coulomb explosion, the excess energy provided by this ESA-based process step is potentially suitable for sub-diffraction structuring, although the single beams are subject to diffraction limited focusing. Based on previous studies [9], the focus of the present work is on the ESA-based ablation as an important element of the proposed sub diffraction direct laser writing technique ( Figure 1a) [8]. Several studies already implemented an ESA-based microfabrication by a hybrid laser process of wide bandgap semiconductors such as fused silica, sapphire, or quartz [7,[10][11][12][13]. The non-thermal ablation by the excitation of electrons above the vacuum energy improves ablation quality and decreases debris formation on the surface. For these dual-beam processes, Excimer and Raman lasers with wavelengths in the UV-range and (ultra-)short pulse lasers are most commonly used to increase absorption through photodissociation and changes in transient absorption [7]. This process is also applicable to lower bandgap semiconductors like gallium nitride, lithium niobate, and silicon carbide [13][14][15][16][17]. In this case, the single laser beams already exceed EBG, but the excitation beyond vacuum energy for improved non-thermal ablation is only achieved by the dual-beam setup. Similarly, an increased absorption of a second, otherwise transmitted laser beam with EP < EBG, without exceeding the vacuum energy enables energy and cost efficient processing as already demonstrated for silicon and silicon carbide [18,19]. The key factor here is that commercially available, compact, and less expansive lasers in the visible and infrared spectrum can be used [18]. Based on this dual-beam ablation technique, Zinc(II) oxide (ZnO) as a direct bandgap semiconductor was used. Especially as thin films on an low absorbing substrate like fused silica, ZnO is commonly used in state-of-the-art applications like optoelectronics, metamaterials, or photonics, and is therefore a promising candidate for the ESA-and potential Based on this dual-beam ablation technique, Zinc(II) oxide (ZnO) as a direct bandgap semiconductor was used. Especially as thin films on an low absorbing substrate like fused silica, ZnO is commonly used in state-of-the-art applications like optoelectronics, metamaterials, or photonics, and is therefore a promising candidate for the ESA-and potential STED (in combination with ESA)-based ablation [20,21]. Thus, we focused on the selective, high quality thin film ablation at optimized processing parameters and cost efficiency. To analyze the suitability of the dual-beam setup for thin film structuring, the ablation process is studied with respect to the laser peak fluences, temporal delay, and wavelengths of the respective beams.

Sample Preparation and Characterization
Fused silica substrates (Infrasil 301, Heraeus, Hanau, Germany) with a thickness of 1 mm were coated by RF-magnetron sputtering of 200 nm thin ZnO layers using a sputtering time of 37 min at a pressure of 0.295 Pa, a gas flow (argon with 2% oxygen) of 6 cm 3 /min, and a power of 150 W. The coated and uncoated samples were optically characterized by UV/VIS transmittance spectroscopy (MCS/100-3, J&M Analytik, Essingen, Germany).
The photoluminescence emission spectra of the coated samples were measured at room temperature with a HeCd laser (IK3202R-D, Kimmon Koha, Tokyo, Japan) at a laser wavelength λ = 325 nm and a continuous wave power of 17 mW. The fluorescence was detected at an angle of 45 • by an optical spectrometer (Maya2000 Pro, Ocean Insight, Rostock, Germany). A 350 nm longpass filter (XUL0350, Asahi Spectra Co., Tokyo, Japan) was utilized in front of the detector to restrict the detection range above E BG , where only weak emission is expected [22].

Laser Processing Setup
The dual-beam setup consists of two Nd:YAG lasers (SLI-10, Amplitude Systems) with a fundamental wavelength λ = 1064 nm. The third harmonic of one of them was used as pump beam with λ pump = 355 nm and a pulse duration of τ pump = 4.7 ns. The second Nd:YAG laser equipped with an additional optical parametric oscillator provides selected ESA beam wavelengths λ ESA = 450, 500, 550, and 600 nm with τ ESA = 4.8 ns. The beams were collinearly focused on the sample surface from opposite sides. Consequently, the ESA beams must pass the fused silica substrate (Figure 1b). For synchronization and temporal peak-to-peak delay, adjustments of both beams at the sample surface, a digital delay generator (DG535, Stanford Research Systems) and a silicon photodetector (ET-2000, EOT) were used. A digital oscilloscope (XDS3302 Plus, OWON Technology Inc., Zhangzhou, China) with a rise time < 1.2 ns monitored the temporal pulse shape (Figure 1c).
The ablation process was investigated using single laser pulses of the respective pump and ESA beams in order to exclude accumulation and incubation effects that would lead to an altered absorption. These include steady sample heating, contamination of the surface by resolidified ZnO, or the formation of deep level defects and color centers [11,23,24]. To ensure optimum laser operation with highest output stability, the laser was operated at its standard repetition rate of 10 Hz during the measurements. The separate pulses were then selected by mechanical shutters, synchronized with the delay generator already used for laser operations. Furthermore, the lasers were operated at their optimum output energy E pump = 160 mJ and E ESA = 27-29 mJ and attenuated by neutral density filters (FW2AND and NDC-100C-4M, Thorlabs Inc., Newton, MA, USA). To further reduce the impact of pulse energy fluctuations and sample inhomogeneities, the results of 5 independent ablation spots fabricated with identical processing parameters were averaged for all performed single-and dual-beam ablation measurements.

Generation and Analysis of Ablation Spots
The ablation behavior of the ZnO layer for the single laser beams was characterized by optical microscopy (VH-Z100, Keyence, Osaka, Japan) and white light interferometry (CCI HD, Ametek Inc., Berwyn, IL, USA) as a function of their respective pump-pulse energy E pump ≤ 13 µJ and ESA-pulse energy E ESA . The latter was labeled according to the utilized ESA-wavelengths λ ESA as E 450 ≤ 19 µJ, E 500 ≤ 12 µJ, E 550 ≤ 25 µJ, E 600 ≤ 31 µJ. The single-beam ablation threshold energy E th and the beam diameter 2ω f (1/e 2 -diameter) were determined using the method proposed by Liu [25]. For this purpose, the squared ablation diameters D 2 were plotted in semilog plot versus the pulse energy E. According to the intersection of the linear fit at D 2 = 0 and its slope determines E th and 2ω f , respectively. Using these parameters, the laser peak fluence at the ablation threshold was calculated by F th = 2E th /πω f 2 . Hereinafter, all fluence values refer to laser peak fluences. For fluences below F th in order to prevent single-beam ablation, ESA-based laser processing with a combination of pump and ESA beam was analyzed. The impact of the different parameters for optimal ablation quality and efficiency was characterized by varying the ESA beam fluence F ESA (F 450 ≤ 1.6 J/cm 2 , F 500 ≤ 4.5 J/cm 2 , F 550 ≤ 5.7 J/cm 2 , F 600 ≤ 5.3 J/cm 2 ) below their individual F th at a constant pulse delay of ∆t = 5 ns and several pump beam fluences at F pump = 0.2 J/cm 2 , 0.1 J/cm 2 , and 0.02 J/cm 2 . Using the highest pump beam fluence F pump = 0.2 J/cm 2 and the same values of F ESA , ∆t was varied from -10 to 500 ns to investigate the influence of the pulse delay. Here, a negative temporal delay (∆t < 0) implies that the ESA beam precedes the pump pulse ( Figure 1c).

Optical Characterization of the ZnO Thin Film
The UV/VIS transmittance spectrum of the uncoated fused silica substrate T S (Figure 2a) shows no absorption for the applied λ and thus no interaction with the radiation. The deviation to T = 1 is caused by reflectance losses at the interfaces [26]. The transmittance of the 200 nm thin ZnO layer on the substrate T ZnO shows a single steep absorption edge at λ ≈ 380 nm with a direct optical band gap at E BG = 3.26 eV [20,27], determined by a Tauc-plot [28]: where G is a constant that depends on the transition probability. The parameter R is determined by the optical transmittance type and is R = 0.5 for directly allowed optical transmissions [29,30]. Plotting αE P over E P and using a linear fit of the straight portion of the absorption edge, the intersection at αE P = 0 determines E BG (Figure 2b). The results demonstrate optimal conditions for a strong fundamental single photon absorption of the pump beam (λ pump = 355 nm). The optical penetration depth is determined by d p = 1/α with the absorption coefficient α = ln(T −1 )/d L . d p = 62 nm and corresponds by definition to the depth at which the beam intensity has dropped to 1/e. This value is smaller than the layer thickness d L and thus almost all of the pulse energy is used to promote electrons to the CB, which are available for the subsequent ESA-based ablation process. By analyzing the sinusoidal interference fringes in the weak and medium absorbing region of the T ZnO -spectrum by the envelope method of Swanepoel [26], the interference free transmittance T E was derived. Here, the T ZnO -maxima and minima were used to fit an upper T M and lower T m envelope function, respectively. Both envelopes were computer-generated using the program OriginPro 2020b (Origin-Lab Corp.) [30]. At shorter wavelengths in the strong absorbing region above E BG , the envelope functions converge to a single curve, merging with the measured values of T ZnO . The interference free absorption coefficient was calculated from the determined T E values (Table 1) [31]. Table 1. Optical parameters (T E , n, α, d p ) at the utilized λ calculated from the transmittance data by the method of Swanepoel (Figure 2a) [26], as well as ablation parameters for single-beam ablation (E th , 2ω f , F th ) determined by the method of Liu (Figure 3a) [25].   Below the band gap energy EBG, n and α steadily increase and the small variation of λ below EBG indicates the comparably weak absorption of the ESA wavelengths without additional pump excitation. The difference between TE and TS is caused primarily by the deviating n, as the upper envelope function that intersects the fringe maxima almost coincides with TS [26]. With increasing absorption at EBG, α increases rapidly.
The sinusoidal interference fringes in the spectrum are caused by reflection losses due to interference effects of the radiation between the air-film and film-substrate interfaces [33]. This indicates homogeneous, smooth, and low scattering surfaces [34]. Thus, only at sufficiently high beam fluences, an independent ESA beam ablation is expected by either a multi-photon absorption or processes including various intrinsic point defects. The latter include deep level defects inside the ZnO layer or shallow donors that are located typically at the surface [20,22,35,36]. As illustrated in Figure 2c, the location of the Furthermore, the amplitude and oscillation of the interference fringes contain infor-184 mation about the ZnO-layer thickness dL and the refraction index n, respectively. The re-185 fraction index of the substrate was derived from the transmittance of the uncoated sub-186 strate. The interference free refraction index in the weak absorbing region was calculated 187 using the envelope functions as demonstrated by Swanepoel [26]. By using λ and n of 188 adjacent fringe maxima or minima and the order number, dL was calculated to be 203 nm. 189 The measured and calculated parameters for the utilized λ are listed in Table 1 [26,31]. ). The dotted lines labeled T M , T m , and T E are the fits of the upper and lower envelope function as well as their geometric mean, respectively [26]. In the emission spectrum of the ZnO film ( Furthermore, the amplitude and oscillation of the interference fringes contain infor-184 mation about the ZnO-layer thickness dL and the refraction index n, respectively. The re-185 fraction index of the substrate was derived from the transmittance of the uncoated sub-186 strate. The interference free refraction index in the weak absorbing region was calculated 187 using the envelope functions as demonstrated by Swanepoel [26]. By using λ and n of 188 adjacent fringe maxima or minima and the order number, dL was calculated to be 203 nm. 189 The measured and calculated parameters for the utilized λ are listed in Table 1 [26,31].  Below the band gap energ λ below EBG indicates the comp additional pump excitation. Th ). The linear fit ( Furthermore, the amplitude and oscillation of the interference fringes contain infor-184 mation about the ZnO-layer thickness dL and the refraction index n, respectively. The re-185 fraction index of the substrate was derived from the transmittance of the uncoated sub-186 strate. The interference free refraction index in the weak absorbing region was calculated 187 using the envelope functions as demonstrated by Swanepoel [26]. By using λ and n of 188 adjacent fringe maxima or minima and the order number, dL was calculated to be 203 nm. 189 The measured and calculated parameters for the utilized λ are listed in Table 1 [26,31]. Below the band gap energy EBG, n and α steadily increase and the small variation of 202 λ below EBG indicates the comparably weak absorption of the ESA wavelengths without 203 additional pump excitation. The difference between TE and TS is caused primarily by the 204 ) at the straight portion of the band edge allows to determine E BG by the intersection at αE P = 0. (c) Schematic of the location of intrinsic point defect levels within the band gap by full-potential linear Muffin-tin orbital method calculations according to Xu et al. [32]. The numbers on the lines indicate ∆E P of the defect states compared to the energy of the valence band (VB)-maximum with filled ( coefficient was calculated from the determined TE values (Table 1) [31]. 183 Furthermore, the amplitude and oscillation of the interference fringes contain infor-184 mation about the ZnO-layer thickness dL and the refraction index n, respectively. The re-185 fraction index of the substrate was derived from the transmittance of the uncoated sub-186 strate. The interference free refraction index in the weak absorbing region was calculated 187 using the envelope functions as demonstrated by Swanepoel [26]. By using λ and n of 188 adjacent fringe maxima or minima and the order number, dL was calculated to be 203 nm. 189 The measured and calculated parameters for the utilized λ are listed in Table 1 [26,31]. Furthermore, the amplitude and oscillation of the interference fringes contain information about the ZnO-layer thickness d L and the refraction index n, respectively. The refraction index of the substrate was derived from the transmittance of the uncoated substrate. The interference free refraction index in the weak absorbing region was calculated using the envelope functions as demonstrated by Swanepoel [26]. By using λ and n of adjacent fringe maxima or minima and the order number, d L was calculated to be 203 nm. The measured and calculated parameters for the utilized λ are listed in Table 1 [26,31].
Below the band gap energy E BG , n and α steadily increase and the small variation of λ below E BG indicates the comparably weak absorption of the ESA wavelengths without additional pump excitation. The difference between T E and T S is caused primarily by the deviating n, as the upper envelope function that intersects the fringe maxima almost coincides with T S [26]. With increasing absorption at E BG , α increases rapidly.
The sinusoidal interference fringes in the spectrum are caused by reflection losses due to interference effects of the radiation between the air-film and film-substrate interfaces [33]. This indicates homogeneous, smooth, and low scattering surfaces [34]. Thus, only at sufficiently high beam fluences, an independent ESA beam ablation is expected by either a multi-photon absorption or processes including various intrinsic point defects. The latter include deep level defects inside the ZnO layer or shallow donors that are located typically at the surface [20,22,35,36]. As illustrated in Figure 2c The emission spectrum in Figure 2a shows a weaker emission in the UV-range from 360-500 nm (close to E BG ), compared to λ = 550-800 nm. The sharp band emission at λ 1 = 380-390 nm overlaps with the band edge and is caused by recombination of photoinduced charge carriers through exciton-exciton collision processes (Figure 2c) [21,22,37]. The specific emissions at λ 2 = 430 nm and λ 3 = 480 nm are attributed to Zn i → VB and Zn i → V Zn transitions, respectively [22]. Their weak emission indicates low roughness and homogeneous surface quality, in line with the findings in the T ZnO spectrum as these transitions occur primarily at the surface [22]. The strongest emission in the visible region corresponds to deep level defects in the bulk of the layer [22,32,36]. Here, the broad emission is comprised of several transitions at λ 4 = 625 nm, λ 5 = 650 nm, λ 6 = 700 nm, and λ 7 = 750 nm, induced by Zn i → O i , Zn i → O Zn , CB → O Zn , and Zn i → V 0 Zn i [22].

Single-Beam Ablation
It becomes evident from Table 1 that F th of the ESA beams is about one order of magnitude larger than the pump beam. The pump beam with λ pump = 355 nm is absorbed fundamentally by single photon absorption, indicated by the linear increase in D 2 (at the semilog scale in Figure 3a) and the transmittance in Figure 2a (E pump > E BG ) [38,39]. In contrast, the ESA beams require either populated defect levels that absorb the laser wavelengths and promote electrons to CB [35], or a multi photon process (E ESA < E BG ) [7]. However, such nonlinear processes would cause an exponential deviation from the linear trend [38] and is therefore unlikely. Thus, the ESA at defect levels is expected as the primary absorption mechanism. The steeper slope angle of the pump beam indicates a larger beam diameter 2ω f = 28 µm [40]. It is comparable or larger than the ESA beams (Table 1) and therefore it provides ideal conditions for optimal utilization of the pulse energy for the ESA process.
at Fth of the respective ESA-beams. Additionally, to compare the dual-beam process with single-beam ablation, the diagonal dashed lines show the linear fits of the Fth calculations, determined by the fits of Eth in Figure 3a.
Decreasing FESA results in a decreased D 2 as less additional energy is available to further excite the already excited electrons. This limits the ablation to an area closer to the beam center, where the intensity of the Gaussian beam profile is still sufficient for ablation. Here, the utilizable fluence range for ESA-based ablation extends about one order of magnitude down to FESA ≈ 0.3 J/cm² for λ = 450 nm. Slightly below these values, the ablation spots are characterized by their non-circular shape. By further decreasing FESA, inhomogeneously distributed substructures were observed within the irradiated area. These surface modifications at low laser fluences are caused by locally enhanced carrier densities that are generated by either an inhomogeneous laser beam profile or locally enhanced ablation at dust, scratches, or crystal defects [41]. It has to be noted that the analysis only includes ablated structures that resulted in distinct and well pronounced ablation spots, excluding the aforementioned value-ranges.
The comparison of the different ESA wavelengths reveals a similar behavior of the beams with the highest photon energy at λ = 450 and 500 nm showing highly increased D 2 when compared to λ = 550 and 600 nm at similar fluences (Figure 3b-e). Higher EP allow an increased number of populated defect levels to absorb these ESA beams and to populate the CB by an ESA (Figure 2c). It has to be noted that a single photon absorption of VB-electrons directly to the CB is not possible at λESA with EP < EBG. In addition, a multiphoton absorption is negligible at the low applied fluences. Furthermore, a transition to another defect level is unlikely due to the specific required wavelength. Thus, besides the intra-band transition of CB-electrons, a transition from defect levels close to the VB into the CB is the most likely absorption mechanism. The existence of a considerable amount of these defect levels is confirmed by the characteristic fluorescence (Figure 2a), where the highest number of emitted photons is attributed to these defect level transitions [42][43][44].  is the most likely absorption mechanism. The existence of a considerable amount of these defect levels is confirmed by the characteristic fluorescence (Figure 2a), where the highest number of emitted photons is attributed to these defect level transitions [42][43][44]. Thus, with the highest EP, using λESA = 450 nm is optimal to generate the biggest ablation spots at the lowest FESA (Figure 3b). Here, at Fpump = 0.2 J/cm² (red markers), even less additional FESA was necessary than the single pump beam requires for ablation. This is indicated by values above the purple line (linear Fth-fit of Figure 3a).
To analyze the effect of Fpump on the ESA-based ablation, Δt = 5 ns at the identical values of FESA for each wavelength were used. As seen in Figure 3d,e, below Fpump = 0.2 J/cm² no distinct ablation spots can be observed for λ = 550 and 600 nm. The utilizable FESA range reduces drastically with decreasing Fpump. Without a sufficient excited electron concentration, the ESA-based ablation process stops abruptly. Only for λ = 450 nm at Fpump = 0.02 J/cm² distinct ablation spots can be observed (Figure 3b,c).

Ablation Efficiency
When compared to single-beam ablation, the ESA-based process allows to decrease the amount of laser energy required for ablation (Table 1). This improvement was evaluated based on the method of Liu [25]. For this purpose, D 2 of the ESA-based ablation spots in is the most likely absorption mechanism. The existence of a considerable amount of these defect levels is confirmed by the characteristic fluorescence (Figure 2a), where the highest number of emitted photons is attributed to these defect level transitions [42][43][44]. Thus, with the highest EP, using λESA = 450 nm is optimal to generate the biggest ablation spots at the lowest FESA (Figure 3b). Here, at Fpump = 0.2 J/cm² (red markers), even less additional FESA was necessary than the single pump beam requires for ablation. This is indicated by values above the purple line (linear Fth-fit of Figure 3a).
To analyze the effect of Fpump on the ESA-based ablation, Δt = 5 ns at the identical values of FESA for each wavelength were used. As seen in Figure 3d,e, below Fpump = 0.2 J/cm² no distinct ablation spots can be observed for λ = 550 and 600 nm. The utilizable FESA range reduces drastically with decreasing Fpump. Without a sufficient excited electron concentration, the ESA-based ablation process stops abruptly. Only for λ = 450 nm at Fpump = 0.02 J/cm² distinct ablation spots can be observed (Figure 3b,c).

Ablation Efficiency
When compared to single-beam ablation, the ESA-based process allows to decrease the amount of laser energy required for ablation (Table 1). This improvement was evaluated based on the method of Liu [25]. For this purpose, D 2 of the ESA-based ablation spots in is the most likely absorption mechanism. The existence of a considerable amount of these defect levels is confirmed by the characteristic fluorescence (Figure 2a), where the highest number of emitted photons is attributed to these defect level transitions [42][43][44]. Thus, with the highest EP, using λESA = 450 nm is optimal to generate the biggest ablation spots at the lowest FESA (Figure 3b). Here, at Fpump = 0.2 J/cm² (red markers), even less additional FESA was necessary than the single pump beam requires for ablation. This is indicated by values above the purple line (linear Fth-fit of Figure 3a).
To analyze the effect of Fpump on the ESA-based ablation, Δt = 5 ns at the identical values of FESA for each wavelength were used. As seen in Figure 3d,e, below Fpump = 0.2 J/cm² no distinct ablation spots can be observed for λ = 550 and 600 nm. The utilizable FESA range reduces drastically with decreasing Fpump. Without a sufficient excited electron concentration, the ESA-based ablation process stops abruptly. Only for λ = 450 nm at Fpump = 0.02 J/cm² distinct ablation spots can be observed (Figure 3b,c).

Ablation Efficiency
When compared to single-beam ablation, the ESA-based process allows to decrease the amount of laser energy required for ablation (Table 1). This improvement was evaluated based on the method of Liu [25]. For this purpose, D 2 of the ESA-based ablation spots in Figure 3b-e (for Fpump = 0.2 J/cm² at Δt = 5 ns) was plotted in a semilog plot as a function of FESA. The linear extrapolation allows to calculate the effective ESA-beam diameter 2ωf eff ) were used. F ESA was varied below their respective F th (Table 1) The behavior at λ ESA = 450 nm deviates slightly from the other wavelengths, where F th is smaller and 2ω f is larger. This suggests a partially different excitation process due to the high E P closest to E BG . This might be explained by the excitation of VB-electrons directly to defect levels close to CB. Moreover, defect levels could be able to absorb the beam and excite electrons directly to CB (Figure 2c), which would lower F th [32,35]. As the defect concentration is expected to be highest at the ZnO-fused silica interface, most of the energy would be absorbed here. At sufficient F ESA , the abrupt absorption at this interface can cause an explosion-like ablation of the whole layer without the necessity to heat the entire ZnO inside the laser spot. For single-beam ablation of the ESA-beams, the resulting spots are mostly inhomogeneous at low F ESA (Figure 4b,g) and the further increase of F ESA leads to substrate damage (Figure 4a,f). Here, an inhomogeneous ablation with ablation depths d a of several microns were observed in the dark areas of the ablation spot. The intense absorption at the fused silica interface exceeds the substrate ablation threshold and limits the F ESA -range significantly. Thus, for the F th calculations, F ESA was decreased and the spots at higher F ESA exceeding the damage threshold of the fused silica substrate were excluded from analysis, as the measured D 2 are highly inconsistent and do not represent the ablation characteristic of the ZnO-layer. Figure 4a,b show selected single-beam ablation spots for the utilized ESA wavelengths. At high FESA, the absorption of defect levels at the fused silica-ZnO interface leads to substrate damage (Figure 4a,f). At FESA close to Fth, the ablation spots become increasingly inhomogeneous (Figure 4b). As described earlier, the further decrease of FESA results in only partially interconnected spots inside the beam. In comparison, the pump beam hits the ZnO-air interface first (Figure 1b) and the fundamental absorption leads to an almost circular and homogeneous single-beam ablation of the entire ZnO layer ( Figure  4c). When reaching the ZnO-substrate interface, the pump beam is already highly attenuated preventing abrupt heating and damaging of the substrate (Figure 4h). The maximum da observed in the depth profiles (Figure 4g-j) are in good agreement with the calculated layer thickness dL = 203 nm. Compared to single-beam ablation, the spot diameters are decreased by the ESAbased ablation process, indicated by decreased values of 2ωf eff when compared to 2ωf (Tables 1 and 2). The ablation area is restricted by the superposition of both pulses. This limits the ablation to the mostly uniform high-intensive center of the pulses and results in an enhanced ablation quality with circular ablation spots. In comparison to single pump beam ablation, the absorbed energy and induced heat at the ZnO-air interface is reduced and homogeneously distributed throughout the entire ZnO-layer. This results in a reduced melt formation at the edges protruding above the layer surface (Figure 4g-j), as well as decreasing the surface contamination by resolidified ZnO-droplets around the ablation spots (Figure 4d,e). By increasing Δt, these effects are reduced even further, as the heat dissipates and an ESA at defect levels dominates the ablation process (Figure 4i,j).

Ablation Quality
The optimization of the parameters allows the complete removal of the ZnO layer without damaging the fused silica substrate (Figure 4d,i) and a precise control of da (Figure 4e,j). The reduced FESA prevents the fused silica-ZnO interface from abruptly heating to the point of substrate damage (Figure 4a,f). As the fundamentally absorbed pump beam

Influence of the ESA Beam Fluence
At first, F ESA was varied (Figure 3b-e) and the values for optimum ablation conditions were set to a small positive temporal delay of ∆t = 5 ns and F pump = 0.2 J/cm 2 . In the case of an ablation process solely based on ESA, the pump beam induces the transient absorption of the spatially and temporally superimposed ESA beam (Figure 1). This requires to prevent surface modification of the individual beams by ablation [8]. Therefore, all measurements were performed using laser peak fluences below their respective F th (Table 1). This is indicated in Figure 3b-e where all values are below the vertical dashed lines, originating at F th of the respective ESA-beams. Additionally, to compare the dual-beam process with single-beam ablation, the diagonal dashed lines show the linear fits of the F th calculations, determined by the fits of E th in Figure 3a.
Decreasing F ESA results in a decreased D 2 as less additional energy is available to further excite the already excited electrons. This limits the ablation to an area closer to the beam center, where the intensity of the Gaussian beam profile is still sufficient for ablation. Here, the utilizable fluence range for ESA-based ablation extends about one order of magnitude down to F ESA ≈ 0.3 J/cm 2 for λ = 450 nm. Slightly below these values, the ablation spots are characterized by their non-circular shape. By further decreasing F ESA , inhomogeneously distributed substructures were observed within the irradiated area. These surface modifications at low laser fluences are caused by locally enhanced carrier densities that are generated by either an inhomogeneous laser beam profile or locally enhanced ablation at dust, scratches, or crystal defects [41]. It has to be noted that the analysis only includes ablated structures that resulted in distinct and well pronounced ablation spots, excluding the aforementioned value-ranges.
The comparison of the different ESA wavelengths reveals a similar behavior of the beams with the highest photon energy at λ = 450 and 500 nm showing highly increased D 2 when compared to λ = 550 and 600 nm at similar fluences (Figure 3b-e). Higher E P allow an increased number of populated defect levels to absorb these ESA beams and to populate the CB by an ESA (Figure 2c). It has to be noted that a single photon absorption of VB-electrons directly to the CB is not possible at λ ESA with E P < E BG . In addition, a multiphoton absorption is negligible at the low applied fluences. Furthermore, a transition to another defect level is unlikely due to the specific required wavelength. Thus, besides the intra-band transition of CB-electrons, a transition from defect levels close to the VB into the CB is the most likely absorption mechanism. The existence of a considerable amount of these defect levels is confirmed by the characteristic fluorescence (Figure 2a), where the highest number of emitted photons is attributed to these defect level transitions [42][43][44].
Thus, with the highest E P, using λ ESA = 450 nm is optimal to generate the biggest ablation spots at the lowest F ESA (Figure 3b). Here, at F pump = 0.2 J/cm 2 (red markers), even less additional F ESA was necessary than the single pump beam requires for ablation. This is indicated by values above the purple line (linear F th -fit of Figure 3a).
To analyze the effect of F pump on the ESA-based ablation, ∆t = 5 ns at the identical values of F ESA for each wavelength were used. As seen in Figure 3d,e, below F pump = 0.2 J/cm 2 no distinct ablation spots can be observed for λ = 550 and 600 nm. The utilizable F ESA range reduces drastically with decreasing F pump . Without a sufficient excited electron concentration, the ESA-based ablation process stops abruptly. Only for λ = 450 nm at F pump = 0.02 J/cm 2 distinct ablation spots can be observed (Figure 3b,c).

Ablation Efficiency
When compared to single-beam ablation, the ESA-based process allows to decrease the amount of laser energy required for ablation (Table 1). This improvement was evaluated based on the method of Liu [25]. For this purpose, D 2 of the ESA-based ablation spots in Figure 3b-e (for F pump = 0.2 J/cm 2 at ∆t = 5 ns) was plotted in a semilog plot as a function of F ESA . The linear extrapolation allows to calculate the effective ESA-beam diameter 2ω f eff (slope of the fit) and the effective ESA-beam threshold peak fluence F th eff (fit intersect at D 2 = 0) ( Table 2). Table 2. Calculation of dual-beam ablation parameters (2ω f eff, F th eff ) based on the method of Liu [25] and dual-beam efficiency parameters (F E , F ET ) determined from Equations (3) and (4). Here, the values of 2ω f eff for λ = 450 and 500 nm are close to their respective 2ω f of single-beam ablation. Consequently, almost the entire area of the fluence profile exceeds the ablation threshold and contributes to the ESA-based ablation. This suggests an effective utilization of the pulse energy. For λ = 550 and 600 nm, the values deviate significantly by a factor of two, which becomes evident by the decreased slope of the fits in Figure 3d,e.
Similarly to F th that defines the minimum required fluence for single-beam ablation, F th eff defines the minimum F ESA at given F pump that is required for ESA-based dual-beam ablation. These values decreased by a factor of 5-10, but are still proportional to F th . With the highest E P , λ ESA = 450 nm can utilize the energy, induced by the pump, most optimally. Here, F th eff = 0.26 J/cm is even below F th = 0.49 J/cm 2 of the single pump beam, indicated by values above the purple line, as described earlier (Figure 3b).
To evaluate the improvements of the ESA-based process, compared to single-beam ablation, fluence efficiency and total fluence efficiency as the sum of the normalized single-beam fluences  (3) to F th of the ESA-beams, allows the comparison of the efficiency improvement with the ESA-beams. In order to exclude an ordinary accumulation effect of the pump and ESA beam, F ET was defined. For F ET < 1, an additional induced transient absorption must be considered as part of the ablation process (Table 2). Figure 4a,b show selected single-beam ablation spots for the utilized ESA wavelengths. At high F ESA , the absorption of defect levels at the fused silica-ZnO interface leads to substrate damage (Figure 4a,f). At F ESA close to F th , the ablation spots become increasingly inhomogeneous ( Figure 4b). As described earlier, the further decrease of F ESA results in only partially interconnected spots inside the beam. In comparison, the pump beam hits the ZnO-air interface first (Figure 1b) and the fundamental absorption leads to an almost circular and homogeneous single-beam ablation of the entire ZnO layer (Figure 4c). When reaching the ZnO-substrate interface, the pump beam is already highly attenuated preventing abrupt heating and damaging of the substrate (Figure 4h). The maximum d a observed in the depth profiles (Figure 4g-j) are in good agreement with the calculated layer thickness d L = 203 nm.

Ablation Quality
Compared to single-beam ablation, the spot diameters are decreased by the ESA-based ablation process, indicated by decreased values of 2ω f eff when compared to 2ω f (Tables  1 and 2). The ablation area is restricted by the superposition of both pulses. This limits the ablation to the mostly uniform high-intensive center of the pulses and results in an enhanced ablation quality with circular ablation spots. In comparison to single pump beam ablation, the absorbed energy and induced heat at the ZnO-air interface is reduced and homogeneously distributed throughout the entire ZnO-layer. This results in a reduced melt formation at the edges protruding above the layer surface (Figure 4g-j), as well as decreasing the surface contamination by resolidified ZnO-droplets around the ablation spots (Figure 4d,e). By increasing ∆t, these effects are reduced even further, as the heat dissipates and an ESA at defect levels dominates the ablation process (Figure 4i,j).
The optimization of the parameters allows the complete removal of the ZnO layer without damaging the fused silica substrate (Figure 4d,i) and a precise control of d a (Figure 4e,j). The reduced F ESA prevents the fused silica-ZnO interface from abruptly heating to the point of substrate damage (Figure 4a,f). As the fundamentally absorbed pump beam is focused on the ZnO-air surface, the initial abrupt heating together with a high concentration of excited carriers diffuse gradually and homogeneously through the ZnO-layer (d p = 62 nm). Optimal ablation conditions were found at ∆t ≈ 5 ns. The ZnO layer is already excited by the pump beam, in particular at the ZnO-substrate interface, where the defects are located primarily. It thus requires only a low excess of additionally absorbed ESA photons and energy to ablate the ZnO layer from the fused silica substrate.

Influence of Temporal Pulse Delay
To determine the importance of defect levels compared to absorption solely based on intra-band transition, the temporal delay between the pulses was varied for different F ESA at the fixed optimum of F pump = 0.2 J/cm ( Figure 5). At the highest F ESA , for λ = 450 and 500 nm (Figure 5a,b), delays above 300 ns still resulted in a measurable ablation. Compared to λ = 550 and 600 nm (Figure 5c,d), this delay is much larger with ∆t mostly below 50 ns. As the fluorescent lifetime of ZnO band gap electrons is well below 10 ns [42,44], an ablation at these high ∆t cannot be explained solely by intra-band transition-based absorption. Thus, the defect levels with their increased fluorescent lifetime serve as the main driving force [35]. First, the pump beam promotes electrons into CB (Figure 1a). From there, besides relaxation back to the VB, electrons transfer rapidly to deep defect levels, which are suitable as absorption centers for the subsequent ESA beam (Figure 2c). The decrease of D 2 and the ∆t-range is almost proportional to F ESA . Only F ESA -values close to their respective F th deviate from this trend. Especially for λ = 550 and 600 nm, these values increase overproportional at the highest F ESA , where the ESA-beams can almost ablate the ZnO solely by single-beam ablation.
strongest ESA is expected. This is in good agreement with the ablation maximum measured at a delay of about 5 ns.
Furthermore, the rapid D 2 -decrease in the delay range 20-50 ns ( Figure 5) can be attributed to heat dissipation and correlates with literature values [45,50]. As 2ωf is about 100 times larger than dL and µ for the time scale in consideration, one-dimensional heat conduction can be assumed [47,51,52]. Here, the heat should be mostly dissipated. Consequently, the defect level absorption is now the dominant factor for ESA, and it is no longer superimposed by pump beam heating. At higher Δt, the ablation slowly decreases due to the finite fluorescent lifetime of the defect level electrons that slowly lowers the available excited electron density for the ESA [43,53].

Conclusions
In the present study, we investigated a dual-beam laser process based on ESA by applying several wavelengths (λpump = 355 nm, λESA = 450, 500, 550, 600 nm) at various Fpump and FESA. The ESA-beams with the highest photon energies produced the largest ablation spots at the lowest beam fluences. A small temporal delay of +5 ns is optimal for the dualbeam ablation process, as it utilizes a combination of heat and excited carriers for increased absorption. Longer temporal delays result in heat dissipation and quickly lower the ablation diameters. Now, the ESA beam absorption is determined solely by the excited carrier concentration. The quality of the ablation spots was improved by reducing the debris and melt formation and the required laser energy was decreased. Furthermore, a  At low temporal delays (∆t < 10 ns), the impact of λ ESA is less pronounced ( Figure 5). Here, the effect of an intra-band transition and sample heating dominates as the beams still overlap. For λ = 550 and 600 nm, an increasing ∆t leads to a rapidly decreasing D 2 because the defect level absorption emerging as the dominant process is only utilized effectively at λ = 450 and 500 nm. As shown in Figure 1c, at ∆t = +5 and −5 ns, the temporal overlap is similar, but the resulting D 2 differ significantly, as prior excitation of electrons by the pump beam is required for ESA.
Exemplary, for λ = 500 nm at F ESA = 4.51 J/cm 2 , ∆t was varied in one ns-steps between ∆t = −5 to +10 ns to analyze the area of temporal beam overlap in more detail (Figure 5b). At the optimum conditions around ∆t ≈ 5 ns, the highest number of prior excited electrons is available for an ESA (Figure 1c). At ∆t = 0 ns, where both beams overlap optimally, the excited electrons are only partially used for the ESA, as parts of the radiation excite carriers after the highest intensity of the ESA beam reaches the sample. At negative temporal delays and further decreasing beam overlap, only minor ablation is observed for F ESA close to F th . Here, the number of excited electrons is insufficient to reach the required ablation threshold. Increasing ∆t above the optimum also leads to a decreasing D 2 -values. However, the rate is much lower than for negative delays.
At ∆t < 50 ns, the heat generated by the pump excitation must be considered, as it effects the ablation process considerably, particularly close to the temporal beam overlap (Figure 1c). Here, increased sample temperatures act as an offset for the ESA, as less additionally absorbed energy is required for ablation. Thus, a wider area of the ESA beam surpasses F th , resulting in increased D 2 . F pump = 0.2 J/cm 2 , well below F th = 0.49 J/cm 2 was used to reduce sample heating.
For polycrystalline ZnO-layers, thermal conductivities k th < 10 W/m·K are found in literature, which is well below k th~3 0-100 W/m·K for epitaxially grown or bulk ZnO [45][46][47][48][49]. The thermal diffusion length is determined by µ = 2(α th ·∆t) 0.5 with α th = k th /c V where c V is the volumetric heat capacity with about 2.4-2.8·10 6 J/m 3 ·K [47,49]. With d p = 62 nm, most of the heat is created by the pump beam at the ZnO-air interface. It takes about 5 ns to reach the ZnO-fused silica interface, where the highest defect concentration and thus, the strongest ESA is expected. This is in good agreement with the ablation maximum measured at a delay of about 5 ns.
Furthermore, the rapid D 2 -decrease in the delay range 20-50 ns ( Figure 5) can be attributed to heat dissipation and correlates with literature values [45,50]. As 2ω f is about 100 times larger than d L and µ for the time scale in consideration, one-dimensional heat conduction can be assumed [47,51,52]. Here, the heat should be mostly dissipated. Consequently, the defect level absorption is now the dominant factor for ESA, and it is no longer superimposed by pump beam heating. At higher ∆t, the ablation slowly decreases due to the finite fluorescent lifetime of the defect level electrons that slowly lowers the available excited electron density for the ESA [43,53].

Conclusions
In the present study, we investigated a dual-beam laser process based on ESA by applying several wavelengths (λ pump = 355 nm, λ ESA = 450, 500, 550, 600 nm) at various F pump and F ESA . The ESA-beams with the highest photon energies produced the largest ablation spots at the lowest beam fluences. A small temporal delay of +5 ns is optimal for the dual-beam ablation process, as it utilizes a combination of heat and excited carriers for increased absorption. Longer temporal delays result in heat dissipation and quickly lower the ablation diameters. Now, the ESA beam absorption is determined solely by the excited carrier concentration. The quality of the ablation spots was improved by reducing the debris and melt formation and the required laser energy was decreased. Furthermore, a model for the evaluation of the ESA-based efficiency improvement was developed. Based on these findings, future investigations on a combination of STED and ESA with regards to STED direct writing processes were advanced.

Conflicts of Interest:
The authors declare no conflict of interest.