1. Introduction

Over the past two decades, few-cycle pulse (FCP) lasers have revolutionized the field of strong-field physics with the advent of single attosecond pulses [1] probing electron dynamics in atoms and molecules, inducing sub-cycle changes in solids [2], relativistic optical mirror harmonic generation [3], etc. Due to its immense promise of opening doors to new physics in relativistic regimes, there has been a concerted effort [4] to construct next-generation ultra-intense FCP sources, such as the Petawatt Field Synthesizer (PFS) [5], Appolon [6], and ELI [7]. In the PFS design, at one stage, the compressor consists of 26 chirped mirrors with 52 bounces, where a damage in an earlier optics in the chain could cause all of the mirrors down the chain to be damaged.

While chirped mirror performance is critical, perhaps just as important is the performance of low-dispersion high reflectors which can support the bandwidth of these ultra-broadband systems. So far, beam routing of amplified femtosecond pulses with energies in the mJ range has relied on dielectric quarter-wave multilayers. By maximizing the index ratio between the coating materials, bandwidths as large as 200 nm centered at 800 nm could be achieved at normal incidence, and >250 nm at 800 nm achieved at 45° with s-polarized light. With the advent of sources capable to deliver many-mJ pulses with durations approaching the few-cycle regime, beam routing optics will have to fulfill new demands. This increased fluence calls for employing coating materials with larger energy bandgap in order to increase the laser-induced damage threshold (LIDT). However, these materials have lower refractive indices, which in turn results in limited bandwidths, insufficient to support the ever-increasing spectral witdth of the pulses.

Silver layers are known to have laser induced damage threshold values comparable to dielectric high-index materials [8], in the many-cycle pulse regime. Equipped with properly designed dielectric over-coatings, Ag-based mirrors can provide high reflectance (99% and above depending on angle of incidence and polarization) and extremely low dispersion (of the order of a few fs²) over bandwidths capable to support less than two-cycle pulses centered at 800nm. Owing to these properties, dielectric-enhanced silver (DE-Ag) mirrors are ideal candidates for the beam routing of the rapidly-emerging next-generation few-cycle amplifiers and knowledge of their LIDT in the 5-fs regime is of paramount importance.

However, to date, there has been no systematic study of single- or multiple-pulse FCP LIDT of ultra-broadband optics, lack of which significantly hampers optimal design, construction and operation of these new, intense FCP systems. This is partly because FCPs require highly accurate dispersion management when propagating through anything other than vacuum. Previous FCP LIDT works were mainly on bulk materials (UVFS, copper, etc.) [9–11]. Additionally, there has been controversy about whether LIDT is different in vacuum vs. air environment, due to the onset of non-linear phase accumulation in air. In this work, we present a systematic study of FCP LIDT of commercially-available ultra-broadband optics, in which we compare and contrast the effects of air vs. vacuum environment in both single- and multi-pulse regimes.

2. Experimental setup

A 3 mJ, 500 Hz, 35 fs homemade liquid-nitrogen-cooled Ti:Sapphire laser system was used to generate FCPs by way of an argon-filled hollow-core fiber and chirped-mirror compressor system (Kaleidoscope, Spectra-Physics), and a schematic of the experiment is shown in Fig. 1. Ultra-broadband pulses with energy 0.35 mJ and nominal central wavelength 770 nm were obtained from a 0.5 mJ input into the fiber, and the dispersion of the experiment was carefully managed to achieve and maintain FCP compression on the target inside the experimental chamber. Initial pulsewidth measurement was performed in situ by a Femtometer interferometric autocorrelator (Spectra-Physics), and was used to set the optimal number of chirped mirror bounces. Subsequently, a fused silica wedge pair was inserted to tune dispersion more finely than a single chirped mirror bounce (as characterized in situ by a dispersionless scanning autocorrelator), and was adjusted when switching between air and vacuum tests to account for the difference of dispersive pathlength through air. Dispersion-managed energy control was achieved by a combination of dispersive mirrors, an achromatic waveplate (Spectra-Physics FemtoOptics, OA228) and a Brewster reflection from an uncoated fused silica prism, and a calibrated photodiode which measured the energy of each pulse using a pick-off from a one-side AR-coated, 1 mm-thick, fused silica window. The beam was coupled into the experimental chamber with a 2 mm fused silica Brewster window. All reflectors used for beam transport were DE-Ag mirrors with an optimized set of dielectric layers to enhance reflectivity from 600–950 nm and maintain ultra-low GVD in that bandwidth (FemtoOptics). In this way, a pulsewidth of 5–6 fs was maintained at the interaction region. For comparison, some tests were performed at 35 fs by evacuating the hollow-core fiber chamber.

 

Fig. 1 Schematic of the FCP LIDT setup. Input pulses are coupled into an argon filled hollow-core fiber to generate hundreds of nm of bandwidth. The output is collimated, and then is compressed by a set of chirped mirrors. Pulse energy is varied with a waveplate and Brewster-reflection from a prism, and measured with a calibrated photodiode. Pulse duration is optimized with a wedge pair, the beam enters the vacuum chamber via a Brewster window, and is focused onto a target. The 5-axis target stage can be rotated for testing at both 0° and 45° incidence. A HeNe laser illuminates the interaction region, which is monitored in situ via a 10× microscope objective. The inset shows the configuration for focal spot analysis, along with a typical beam profile.

Download Full Size | PDF

The experimental setup within the vacuum chamber allows LIDT measurement for near-0° and 45° angle of incidence (AOI), where a f = 150 mm spherical DE-Ag mirror at small incidence angle was used to focus the p-polarized laser pulses onto the target. The target optics are mounted in a 1″ kinematic mirror mount, and the target position can be manipulated by 3-axis motorized control. The sample may be translated completely out of the beam, allowing the focus to be characterized in situ in both air and vacuum, and the typical focus has a Gaussian e⁻² waist 2w₀ = 30 μm. However, the reported peak fluences were calculated by directly integrating the local fluence as function of the measured beam profile, and comparing with the peak pixel value, in order to avoid a Gaussian fit to a slightly irregular spatial mode. The in situ damage area monitoring system with < 1 micron resolution is illuminated by a CW HeNe laser, for quick detection of damage. A dry scroll pump maintains the vacuum pressure of < 0.5 torr in the chamber, where high vacuum cleanliness protocol is maintained.

The samples tested in this experiment include various commercially-available optics which have thin-film coatings designed to be used with ultra-broadband FCP systems such as the above. Specifications of the various samples can be found in Table 1. In summary, there are two different DE-Ag mirrors, one chirped mirror, two beamsplitters with different splitting ratios (all from the Spectra-Physics FemtoOptics product line), and a fused silica substrate as a control sample (MTI Corp.).

Table 1. Samples Tested

View Table

Damage threshold was determined in a similar manner to the standard probabilistic method [12,13], where 10–20 fresh sites at single fluence were exposed to the laser pulses, and damage probability vs peak beam-normal fluence was plotted for each test (see Fig. 2 for examples). While some test conditions resulted in smooth transitions between damaged and undamaged (containing several test fluences of partial probability 0 < P < 1), many tests yielded more sharply-defined LIDTs where only one or zero tested fluences lie in this partial-probability regime. In order to achieve a consistent method for determining LIDT among these varying probability distributions, the LIDT fluences were reported as the average between the highest undamaged (P = 0) fluence and the lowest damaged (P > 0) fluence. In the case of fused silica, the LIDT was determined by fitting crater depth vs. natural logarithm of fluence. Surface damage was determined by optical microscope, Wyko NT9100 interferometric surface profiler, atomic force microscope (AFM) and scanning electron microscope (SEM) for silver mirrors.

 

Fig. 2 Example plots of damage probability. Some samples or test conditions had more smoothly-defined LIDT (left), while others had more sharply-defined LIDT (right). The dashed line indicates the reported LIDT, while the shading indicates the size of the error bar.

Download Full Size | PDF

3. Damage threshold results

Results of the LIDT testing are shown in Fig. 3, and several relationships emerge, despite the fact that not all tests were performed on all samples. First, it is apparent that under otherwise identical conditions, multiple pulse LIDT fluence F₁₀₀₀ is significantly lower than that of single pulse F₁ (by 20 − 50%, depending on the sample). In the case of the bare UVFS sample, this ratio F₁₀₀₀/F₁ = 0.57 for both air and vacuum. This value is higher than previous measurements with non-FCPs [14–16], but is perhaps pulsewidth-dependent and would be consistent if F₁ is more intensity-dependent than F₁₀₀₀. This is plausible because ionization is in the nonlinear regime (photon order 5–8, which corresponds to 600–900 nm). LIDT is found to be pressure-independent for single- and multi-pulse damage of fused silica, and for single-pulse damage of single-layer oxide films, consistent with previous work [14]. However, for multi-pulse damage of single-layer oxide films, Nguyen et al reported decreasing LIDT in vacuum for > 300 pulses (influenced by the lack of ambient O₂ or H₂O). We have observed this trend for the 50–50 beamsplitter and possibly an DE-Ag mirror, but not across all samples – possibly due to the complexities of the multi-layer designs. Additionally, vacuum-multi-pulse LIDT lowering may not be as pronounced in this work due to using only rough vacuum (as opposed to high), as well as only testing with 10³ pulses, since Ref. [14] found this lowering to saturate near ∼ 10⁵ pulses.

Within a particular test type, the primary trend is that LIDT tends to increase as the overall reflectivity decreases. This leaves the chirped mirror and DE-Ag mirrors with the lowest LIDTs due to their > 98% reflectivity, where the remaining LIDT differences are presumably determined by other factors of the testing or coating design. For example, the top layer of the DE-Ag type I mirror has a higher refractive index compared to the top layer of the DE-Ag II mirror. High-index materials typically have lower LIDT than low-index materials when the primary damage mechanism is related to nonlinear ionization [15], so perhaps having a low-index material on the surface slightly increases single-pulse LIDT, consistent with the results of the DE-Ag mirrors. However, our multi-pulse damage results seem to show the reverse trend, where thin-film optics with low-index outer layer appear to have a reduction of LIDT in vacuum environment. See Section 4.2 for morphological evidence which may explain this trend.

 

Fig. 3 LIDT fluences vs. test type. Fluences are reported as peak, beam-normal. Refer to Table 1 for details of each sample. As expected, multi-pulse LIDT (left) is always significantly lower than the single-pulse LIDT (right) under otherwise identical conditions. Within a given test type, the LIDT increases approximately with decreasing sample reflectivity. While the air vs. vacuum environment does not appear to have an effect on LIDT for the bare UVFS sample, some coated optics are influenced by the environment. Furthermore, the environmental effect on multi-pulse LIDT trends in the opposite direction of the effect on single-pulse LIDT. UVFS results have been scaled down by a factor of 10 for plot readability.

Download Full Size | PDF

The DE-Ag II mirror was also tested at 35 fs in vacuum for the sake of pulsewidth comparison, since a FCP here generates ∼ 6× more intensity for a given fluence, and the decreased intensity led to an increased LIDT. For the non-FCP pulse, the single- and multi-pulse LIDT is increased by ∼ 1/3 and ∼ 1/2 respectively.

4. Damage morphology results

Beyond simply determining LIDT, damage sites were examined with AFM, SEM, and optical profiler in order to study damage morphology, and stark differences were observed between air and vacuum for multi-pulse damage. In this study, we focus on fluences below the single-pulse threshold (F₁ > F > F₁₀₀₀), in an effort to study the mechanisms involved in the onset of FCP multi-pulse damage. For the optics tested at 45°, laser-induced periodic surface structures (LIPSS) oriented ‖ to laser polarization are observed to form in air when overall reflectivity is high, and formation is suppressed as reflectivity decreases. In this way, the air-generated LIPSS are most prevalent in the DE-Ag mirror, still clearly present in the 50–50 beamsplitter, less clear in the 37–63 beamsplitter, and absent from the UVFS at these fluences (though an alternate structuring is observed). In vacuum test environment, if LIPSS form at all they were observed to form ⊥ to polarization. No LIPSS were observed in the chirped mirror for damage sites formed either in air or in vacuum, which may be related to the fact that no material was ablated even at 2F₁₀₀₀. However, the damage sites show complex morphological changes pertaining to interaction of FCP with specific layers which are beyond the scope of this paper and will be presented in a future work.

4.1. Dielectric-enhanced silver mirror type II

The air/vacuum morphological difference is most clearly observed in the DE-Ag mirrors, where resulting LIPSS are rotated by 90° depending on test environment (see Fig. 4). In the case of air (Fig. 4a), the LIPSS are primarily oriented parallel to the laser polarization and are reasonably well-explained by Sipe theory [17], a topic which has been widely studied [18]. The LIPSS period in this case is Λ = 1.11 ± 0.06 μm (measured with AFM), which is consistent with the Sipe model for AOI θ = 45°P and air refractive index n = 1, following ${\mathrm{\Lambda }}_p=\lambda /\sqrt{n^2-{\text{sin}}^2\theta }$ [19]. This implies that the surface-scattered light propagates primarily in the air above the surface before interfering on surface. In the case of vacuum, the periodic structures are perpendicular to the polarization, as in Fig. 4b, and the LIPSS period is different on the right side compared to the left. This is consistent with an interference pattern generated at nonzero AOI where there is forward- and backward-propagating surface wave solutions (surface plasmon polaritons (SPPs)), yielding two possible solutions Λ_⊥ = λ/(λ/λ_s ±sin θ) for SPP wavelength λ_s [20]. The forward and backward LIPSS periods are Λ_f = 1.25 ± 0.08 μm and Λ_b = 0.417 ± 0.008 μm, but these periods cannot be generated by the same λ_s. If λ_s is allowed to vary freely, the corresponding λ_s values for the measured Λ give (λ/λ_s)_f = 1.32 and (λ/λ_s)_b = 1.14.

 

Fig. 4 SEM images of DE-Ag II mirror, showing the contrast between air and vacuum for 1000 pulse damage at 0.20 J/cm². LIPSS generated in air (a) are primarily ‖ to laser polarization, whereas in vacuum (b) they are ⊥ to polarization. Laser is incident from the left at 45°P.

Download Full Size | PDF

The simplest expression for SPP wavelength is known to be obtained by considering the interface between two semi-infinite media ${\lambda }_s=\lambda /\sqrt{(∊+{∊}_d)/∊{∊}_d}$, with dielectric functions ∊ of the metal and ∊_d of the dielectric medium. Here the dielectric side is complicated due to the multi-layer dielectric stack, but can be instructive if we consider the case where this complex design is represented by a single effective ∊_d, which would be influenced by the layer materials and thicknesses. The ∊_d values which correspond to the measured Λ are ∊_d ≈ 1.6 forward and ∊_d ≈ 1.2 backward, for which $n=\sqrt{{∊}_d}=1.26$ and 1.1 are both lower than that of any typical coating material. This suggests that the dielectric function is reduced through ionization within the Drude model ∊ = c − n_ee²/m_eff∊₀ω², where ∊_c is the unexcited dielectric function, n_e is the electron plasma density, e is the electron charge, m_eff is the effective electron mass, ∊₀ is the vacuum permittivity, and ω is the laser frequency [21]. Then, the forward/backward spatial non-uniformity of ∊_d could result from the local laser fluence as determined by the spatial mode of the focal spot. Further evidence of this spatial non-uniformity is visible in Fig. 4, where the curvature of the LIPSS correlates with bending of SPPs outwards towards the refractive index gradient (since waves bend toward the direction of slower propagation). Considering that the LIPSS extend further to the right than the left, the intensity peak of the laser is closer to the backward-propagating LIPSS, which could correspond to this increased ionization and therefore decreased ∊_d.

It is worth noting that structures similar to the forward-propagating solution are also visible in the case of air, though they are not the dominant feature, but suggesting that ⊥-LIPSS mechanism is more intrinsic and independent of ambience. In this way, it is likely that SPPs are generated at the silver layer (underneath the outer dielectric layers), since a metal dielectric interface with roughness naturally can support plasmons [22].

The single-pulse damage morphology is very similar between air and vacuum, showing ablation in the second layer (high-index) at higher fluences and showing blistering (swelling) or delamination of the outer layer nearer to F₁.

4.2. 50–50 beamsplitter

Fig. 5 shows representative profiles of the 50–50 beamsplitter for both vacuum and air. In Fig. 5a, the optical profile clearly captures the lateral scale of LIPSS, though the depth information is only qualitative since the sample is a stack of thin films which render the device’s depth information inaccurate (hence, depth scale is omitted). Here the LIPSS are observed to be splayed off to the right side of the laser peak intensity (in the direction of propagation), consistent both with the DE-Ag II mirror results as well as the Sipe model for nonzero AOI. Fig. 5b shows an AFM profile of the same site, where the damage debris is so tall (300–400 nm) that it somewhat obscures the underlying LIPSS trenches. The LIPSS period is Λ = 0.96 ± 0.04 μm, which is slightly less than the DE-Ag II mirror and for the beamsplitter implies that the scattered wave travels through an effective average index of n = 1.07 instead of n ≈ 1. It is common for the selvedge region to have 1 < n < n_sample due to the roughened surface being some combination of the two materials at the interface.

 

Fig. 5 Depth profiles of the 50–50 beamsplitter, showing the contrast between air (a,b) and vacuum (c,d) for 1000 pulse damage. Optical profile in (a) clearly shows LIPSS (parallel to polarization), though depth scaling is inaccurate due to interference effects from the multi-layer thin-film design of the optic. AFM in (b) shows the same site as (a) (white arrow guides eye to an example LIPSS groove) with quantitative depth information, and shows very tall debris. AFM in (c) shows vacuum blister morphology, with onset of more dramatic damage visible at the central peak fluence location, which is magnified in (d). Depth contrast in (c) is enhanced to increase visibility of blister contour, while true depth scaling is maintained in (d). Laser is incident from the left at 45°P.

Download Full Size | PDF

Comparison with the vacuum damage morphology (Fig. 5c) of this sample gives some insight into why the LIDT is different between vacuum and air testing ambience. In vacuum, no LIPSS are present, but instead only a crater or a pre-crater swelling or “blister” which is not observed in the air tests. Blister formation in oxide films with femtosecond pulses has been previously studied in of case of SiO₂ on Si [23]. In that work, the authors speculated/proposed that energy absorbed underneath the surface film caused a softening of the film which allowed relief of residual compressive stress of the film, leading to surface expansion into a blister. For this site, the central region is swollen by ∼ 10 nm, has a few isolated bumps with a height of 10’s of nm, and a small trench located at the peak fluence. This trench and surrounding debris bumps seem to indicate the onset of more dramatic damage where the outer (low-index) layer is ruptured and ejects molten material from underneath. Comparing the fluences in Figs. 5b (0.25 J/cm²) and 5c (0.23 J/cm²) with the vacuum and air LIDTs (0.20 and 0.24 J/cm² respectively) suggests that F₁₀₀₀ in air is defined by this more dramatic (ablative) mechanism, where the material beneath the surface layer has sufficient pressure to locally rupture the low-index surface, ejecting material outward. Whereas in vacuum, F₁₀₀₀ is lower and is defined by the blistering mechanism (non-ablative damage in a subsurface layer). The vacuum ambience could perhaps induce the blistering mechanism simply due to the external stress of the 1 atm pressure differential or possibly be aided by aforementioned mechanisms such as oxygen deficiency, but further study would be necessary to determine the exact cause.

4.3. Discussion of air vs. vacuum

For the beamsplitters and the DE-Ag mirrors, multi-pulse damage in air forms LIPSS parallel to laser polarization (suggesting the surface-scattered-wave or Sipe model), whereas in vacuum LIPSS are perpendicular to polarization (DE-Ag, suggesting the SPP model) or LIPSS are absent (beamsplitters). In vacuum, the presence or absence of LIPSS correlates with the presence or absence of a metal layer in the coating design, and the absence of LIPSS in purely dielectric coatings (beamsplitters) points to the fact that multi-pulse damage (in both vacuum and air) is happening below the laser-metallization threshold for SPPs (∊ < −1) of the dielectric layers. Then for the surface-scattered-wave model to hold, air needs to somehow influence the interaction to generate localized scattering sites. Following the Sipe model, the roughened “selvedge” region generated by previous pulses is generally assumed to be the source of this scattering, however in the vacuum test case (which also has roughened surface), LIPSS parallel to polarization are not observed. Instead of the roughened surface, perhaps nanoparticulate debris is the dominant source of scattering which dominates LIPSS formation in air ambience. In this case, air would tend to hinder ablated particles from escaping the surface, and between pulses perhaps some could re-settle upon the damage site (see Fig. 5b). For the vacuum case, the reduced pressure would favor free expansion, so that debris particles could ballistically escape, leaving the target region perhaps more clean than in the air case [24]. If so, then the air case would have many more scatterers to form the surface-scattered wave for Sipe-LIPSS generation (‖ to polarization). We would not expect this mechanism to have much influence on the LIDT (consistent with our DE-Ag mirror results), since it first requires damage to generate nanoparticles. It is worth mentioning that the generated nanoparticles in all cases here would be composed of dielectric materials, and thus would natively have only limited scattering (as compared with metallic nanoparticles). However, since they are sitting in an intense FCP, local field enhancement due to the nanostructures could increase local ionization and transiently metallize the particles to further enhance the scattering for LIPSS formation.

4.4. Fused silica

LIPSS are known to form on UVFS for fluences F > F₁ where there is sufficent ionization to enter the SPP LIPSS formation regime [25], but in this work we have focused on F < F₁ and did not observe LIPSS. However, there is a notable morphology difference between air and vacuum, where in air there is a collection of aligned nano-ridges oriented ⊥ to polarization (Fig. 6a) which are suppressed in vacuum (Fig. 6b). At this fluence of 1.44 J/cm², the ridges are typically ∼ 200 nm wide and vary from 1 − 15 nm in height, and at higher fluences (still < F₁) these features grow taller. While these ridges seem to be aligned and have characteristic lateral size, they are not arranged periodically as LIPSS.

 

Fig. 6 AFM profiles of the UVFS, 1000 pulses. Test conditions are (a) FCP, air; (b) FCP, vacuum; (c) 35 fs, air. No LIPSS are observed for neither air nor vacuum, though the air case has nano-scale ridges aligned ⊥ to laser polarization. In (a), the air-formed ridges are ∼ 200 nm wide and 1 − 15 nm tall at this fluence.

Download Full Size | PDF

For comparison, damage tests were also performed with the 35 fs pulses (Fig. 6c), and the ridges were found to form with much greater prevalence in both air and vacuum ambience, though again less prominently in vacuum. Though these features merit further investigation, we speculate that the formation of these surface structures could be related to so-called “nanogratings” which form in the interior of bulk fused silica [26, 27]. The alignment ⊥ to polarization also suggests that this is a different mechanism than the Sipe theory, but matches the orientation of nanogratings. Additionally, nanograting formation efficacy has been found to decrease with decreasing pulsewidth [26], which is consistent with what we observe. Nonetheless, this effect is neither present on thin films, nor is it enhanced by FCPs.

5. Conclusion

In this work, we have performed a damage study of commercially-available ultra-broadband optics using few-cycle laser pulses, where single- and multi-pulse LIDTs were measured for both air and vacuum test environment (ranging from 0.1–0.4 J/cm²). As sample reflectivity increases, LIDT tends to decrease. For the bulk fused silica and many of the tested optics, LIDT was independent of air/vacuum ambient condition. However for some thin-film coatings such as the 50–50 beamsplitter, the multi-pulse LIDT was decreased in vacuum, and morphological evidence supported this result. Multi-pulse damage morphology was studied for fluences below the single-pulse LIDT, and a stark morphological contrast was observed between air and vacuum damage sites. The air tests resulted in LIPSS patterns consistent with a surface-scattered-wave model, but they were found to be absent from vacuum tests. We attributed this effect to the observed increase of laser-generated damage debris in air ambience, where this high debris content on the surface can act as scattering sites for subsequent laser pulses. In this way, the multi-pulse air tests are consistent with the standard Sipe model of LIPSS formation, but the vacuum case is not. We also found that for the specific instance of dielectric-enhanced Ag mirror testing in vacuum, multiple pulse interaction couples its energy into the surface via excitation of surface plasmon polaritons. To gain further insight into the fundamental mechanisms of these processes, future work underway includes studying FCP damage morphology in air vs. vacuum for the case of single-layer films.