Structuring of thin films by ultrashort laser pulses

Modern life and global communication would not be possible without technologically tailored thin films; they are omnipresent in daily life applications. In most cases, the films are deposited entirely at the carrying substrates in a specific processing step of the device or sample. In some cases, however, removal or modification must be performed locally, i.e., site-controlled and material selective through an additional laser processing step. For that ultrashort laser pulses with durations in the femtosecond and picosecond range can provide unique advantages and capabilities in industrially scalable schemes. This article reviews the current state of the research and corresponding industrial transfer related to the structuring of thin films by ultrashort pulsed lasers. It focuses on the pertinent historic developments, reveals the relevant physical and chemical effects, explores the ultimate limits, and discusses selected industrial and scientific applications.


Introduction
The structuring of thin films is of paramount importance for many daily life applications, where a local post-processing of functional coatings is required to adapt, adjust or improve their contour to the demands of the devices 2D-or 3D-geometry. Such coatings can enable or control optical, electrical, or mechanical surface properties for applications in semiconductor electronics, the field of renewable energy, friction and wear management, or for surface decoration or protection. Moreover, the laser-induced "structuring" can involve either complete film removal or the modification of its intrinsic structural or electronic state [1]. For all this, laser-processing offers a simple, fast, and reliable technology that allows to re-shape, modify, and tailor even very thin, extremely soft, hard, or brittle film materials in a clean and contactless manner, while featuring lateral and vertical machining precision in the sub-micrometer scale [2,3].
On the other hands, thin films are the backbone for many optical applications in fields of optical devices, photovoltaics, and modern laser technology. Here, particularly the development of damage resistant optical coatings with antior high-reflective performance is desired [4]. For optimizing these optical coatings, a deep understanding of the manifold linear and nonlinear interaction mechanisms of optical radiation with matter as well as a profound knowledge of the materials solid state physics and chemistry is desired.
In this review article, we present an up-to-date literature overview on the laser processing of thin films. In most cases, the focus is laid on single film layers and the treatment by single or multiple ultrashort laser pulses with durations in the picosecond (ps) and femtosecond (fs) range. For that, the particularities of laser-matter interactions in the presence of additional film-or substrate interfaces are pointed out, the impact of intrinsic optical or adhesional mechanical film properties are revealed, and relevant applications are discussed. Since the literature published in this field over decades is already very exhaustive, our article aims to focus on the groundbreaking early publications, while simultaneously shedding some more light on specific, less known aspects in ultrashort laser processing and their historic development. To aid the reader selecting suitable laser processing conditions, special emphasis will be given to the discussion and consideration of the intrinsic influence of the film material (metallic, semiconducting, dielectric), the film thickness (in relation to the wavelength and focusing depth), and the physical properties of the substrate material (optical reflectivity, heat conductivity, thermal expansion, etc.).
The article is organized as follows: Sect. 2 briefly recalls the benefits of ultrashort laser pulses (ULP) in material processing. In Sect. 3, the ULP laser structuring is discussed, including possible processing geometries (3.1), interaction mechanisms with films made of metals, semiconductors, or dielectrics (3.2), and relevant transient effects along with their time scales and with a focus on important differences between thin and thick films (3.3). Following the historic developments in the field, Sect. 4 presents selected industrial and scientific applications of ULP processing of single-layers and multi-layer film systems, including photolithographic mask repair (4.1), laser cleaning aspects (4.2), avoiding laser-damage of optical coatings (4.3), laser-printing through laser-induced forward transfer (4.4), and thinfilm depth-profiling analyzes for material sciences (4.5). Section 5 provides a brief summary along with an outlook to future developments.

Implications of ultrashort laser pulses for materials processing
Before discussing the specifics of thin film processing with ultrashort laser pulses in the following sections, the advantages and implications of using ultrashort laser pulses for materials processing should be briefly reviewed. While the literature on that topic is already mature and very exhaustive, we try to break this down to a few key aspects only. For additional details, the reader is referred to some books and review articles here [1,[6][7][8][9][10][11][12][13][14][15][16].
• Since the pulse duration is small compared to the typical electron-phonon interaction times (τ e-ph ), heat diffusion into the material is insignificant during the time scale of the laser pulse [1]. • The extremely rapid energy deposition leads to a high local confinement of the laser pulse energy resulting in reduced threshold fluences of melting and ablation when compared to longer pulse durations. • The strong energy confinement results in a small heat affected zone (HAZ, extent typically ≤ 100 nm [17]) and consequently in a high lateral and vertical machining precision. Compared to the ns-pulse irradiation, the minimal achievable ablation depth per pulse is reduced. Significant smaller features with sizes even below the beam diameter can be generated. • Nearly every material (metals, semiconductors, ceramics, polymers, dielectrics, and composite materials) can be ablated successfully by employing intensities between 10 12 and 10 14 W/cm 2 which are easily obtained for fslaser pulse durations. • For all materials, a distinct damage threshold, i.e., the minimum fluence F th (areal energy density in J/cm 2 ) can be observed above which a permanent material modification (structural, chemical, or ablative) can be induced upon laser irradiation. For bulk materials, such thresholds are often ruled by specific thermodynamic mechanisms, such as melting, ablation, or oxidation [18][19][20]. For dielectrics irradiated by femtosecond laser pulses, the damage threshold is more deterministic compared to longer laser pulses since the electrons seeding the avalanche dielectric breakdown are generated by the pulse itself [5,6]. For dielectrics and semiconductors damage is then initiated when a critical carrier density of the order of 10 21 cm −3 in the conduction band is exceeded. • For a single-pulse irradiation event, the formation of the ablation plasma occurs after the ultrashort laser pulse, i.e., temporally separated from the intra-pulse absorption of the energy in the solid [8,21,22]. Due to the lack of plasma-shielding (for low repetition rate laser pulse trains) the ablation rates are less dependent on the focusing conditions (spot size) and the laser pulse energy can be used more efficiently. • Femtosecond laser pulse irradiation in the ablative regime can lead to a very rapid (< 50 ps) disintegration of the excited ablating material into nanometer-sized fragments [23,24]. • In most materials, two different regimes of "gentle ablation" (for fluences below a few J/cm 2 ) and "strong ablation" (for fluences above several J/cm 2 ) have been observed [25,26]. For dielectrics and semiconductors Coulomb-explosion [27,28] accounts for the gentle ablation regime, while for all materials the thermodynamic phase-explosion [29][30][31][32] has been discussed. • In comparison with longer pulse durations (τ ≥ ns), the expulsion of material and the amount of redeposited material in the surrounding of the irradiated zone is generally reduced.
Particularly the small HAZ along with the reduced threshold fluence and increased ablation precision make ultrashort laser pulses potentially interesting for the processing of thin films. However, care must be taken to avoid or reduce thermal heat-accumulation effects that may destroy the benefits of ultrashort pulse durations at (too) high laser pulse repetition rates exceeding several tens to hundreds of kilohertz. 3 Single-layer structuring

Processing geometries
The laser-processing of thin films is performed either in the substrate geometry, where the laser beam is focused by a suitable optical element (lens, mirror, axicon, diffractive optical element) directly onto the surface of the film that was previously deposited on a carrier substrate or in the superstrate geometry where the laser beam is focused through the transparent substrate into the film/carrier interface (Fig. 1).
In the substrate geometry (Fig. 1a), the ablation products (atoms, ions, molecules, clusters, nano-and microparticles, fragments) and the incident laser beam are counter-propagating at the same half-space, that my cause detrimental effects such as plasma-shielding for long pulse durations [21,33] or at high laser pulse repetition rates [34]. Moreover, the ablated material can be partly redeposited as debris at the film surface in regions later accessed by the laser-processing.
In the superstrate geometry (Fig. 1b), such interactions with the debris can be avoided since the film/carrier-interface is neatly protected from debris through the film itself. Also, plasma-shielding does not occur since the incident laser beam and the ablated material are spatially separated completely through the film. On the other hand, the laser processing in this geometry is only possible for carrier materials being transparent for the laser wavelength. Additional difficulties may arise for ultrashort high-intensity laser pulses, where transient defocusing, self-focusing, or beamfilamentation through nonlinear beam propagation effects (such as the Kerr-effect) can antagonize against a defined localized confinement of the optical energy in the film/carrier-interface region-finally preventing a controlled laser processing then. Thus, the superstrate geometry may not be accessible in all laser processing cases.

Interaction mechanisms
Laser-induced damage is referred to as an irreversible and permanent modification of the film material or surface morphology. Typically, it manifests at laser fluences exceeding a sharply defined threshold value, referred to as damage threshold fluence. The laser-induced damage of thin films can be caused by different physico-chemical intra-or interpulse mechanisms. While the first ones occur already upon irradiation by single laser pulses during the pulse duration, the latter ones manifest on longer time scales and result in changes of the film surface topography, its chemistry, or the intrinsic material structure through multiple laser pulses. Through this, multi-pulse irradiation often involves damage accumulation phenomena (often referred to as incubation) that can rely on the laser-induced formation of electronic defects in the film material or on heat accumulation when the laser-induced surface temperature does not cool down to the environmental temperature between successive laser pulses [1].
The laser-induced damage of single layered films typically can be associated with one of four different scenarios, as sketched in Fig. 2. In the first scenario (Fig. 2a), the film damage occurs at the air/film interface. This usually manifests for materials absorbing strongly the laser radiation, such as for metals. For these materials via linear absorption the absorbed laser intensity decays then exponentially within the film materials according to the Lambert-Beer law, typically at depths of a few tens of nanometers only. Thus, in most cases, the laser radiation does not reach the film/carrier interface.
In the second scenario (Fig. 2b), the laser-induced damage manifests at the film/carrier interface. This often occurs for thin transparent films on strongly absorbing carrier materials or for material combinations, where film and carrier exhibit very different thermal expansion coefficients.
In the third scenario ( Fig. 2c), both film interfaces (air/ film and film/carrier) become damaged. This may occur for transiently changing optical absorption mechanisms triggered upon irradiation with high-intensity ultrashort laser pulses and for weakly absorbing film materials with moderate and small band gap energies. In such a scenario, the low-intensity rising pulse slope can reach the film/carrier interface, while at the high-intensity pulse maximum, nonlinear absorption reduces the energy deposition depths and transiently confines most of the absorbed optical pulse energy to the air/film interface.
Finally, in the fourth scenario (Fig. 2d), the laser-induced damage occurs within the volume of the film material. Since typically for a given film material the interface damage threshold lies distinctively below the volume damage threshold, this scenario seldomly manifests for homogeneous thin film materials along with low or moderate laser fluences. Fig. 1 Approaches for laser processing of thin films on a carrier through a focusing element (e.g., lens, concave mirror, diffractive optical element, etc.). a Substrate geometry: laser beam is incident to the air/film interface. b Superstrate geometry: laser beam is incident through the transparent carrier to the carrier/film interface At extremely high irradiation intensities, however, the film material may be driven to highly excited thermodynamic states and may undergo homogeneous melting or phaseexplosion [30,32]. For (semi-)transparent films of medium or large thickness on high-reflecting carrier materials, a less known and often overlooked interference-based mechanism can rule the film damage threshold (see Sect. 3.3.2).

Interaction with metals
We follow the scenario according to Fig. 2a for high absorbing metals. In this chapter, results for gold and nickel films are presented because gold and nickel are excellent examples for materials with weak (gold) and strong (nickel) electron-phonon coupling [1].
In metals, energy absorption is governed by free electron transitions within the conduction band. Within the electron subsystem, the energy is spread on a time scale between tens of femtoseconds up to a picosecond. The thermalization between the electron subsystem and the lattice takes place typically on time scales between one and some tens of picoseconds.
In this paragraph, important findings for nanosecond pulse laser interaction with metal films are briefly described. For gold and nickel films on poorly heat-conductive fused silica substrates, Matthias et al. determined single-pulse ablation thresholds in the thickness range (h) from 50 nm to 7 μm for pulses with a pulse duration of τ = 14 ns at a wavelength of λ = 248 nm [35]. For small film thicknesses h < L th , they found a linear increase of the damage threshold fluence up to h = L th : = (2Dτ) 0.5 , with D being the thermal diffusivity of the irradiated film material. For thicknesses h > L th , the ablation threshold fluence is constant. The thermal diffusion length L th was determined to be 1.05 μm for gold and 0.73 µm for nickel, respectively. All thicknesses h investigated were greater than the small-signal optical penetration depth α −1 . Here, α denotes the linear absorption coefficient. The investigations demonstrated that for material removal a certain minimal energy per unit volume must be deposited in the material. For a thermally well conductive material, F abl /L th is decisive and not the quotient F abl /α −1 , because energy dissipates out of the excited volume element already during the nanosecond laser pulse. The characteristic scaling of the ablation threshold with τ 0.5 for pulses with durations exceeding the electron-phonon relaxation time τ e-ph is just a consequence of intra-pulse 1D heat transport into the bulk of the material [1]. For layer thicknesses h < L th the ablation threshold fluence F abl is, thus, reduced by the factor h/L th since the heat reaches the film/substrate interface already during the laser pulse and stays widely confined within the thin film due to the inefficient cooling by the underlying, poorly heat conductive dielectric substrate. For such thin films the deposited optical energy is restricted to the film and gives rise to a more efficient heating of the film material compared to films of thicknesses h exceeding L th , thus reducing the ablation threshold fluence. For both metals (Au, Ni), it was found that the single-pulse ablation threshold fluence increases linearly with film thickness up to the thermal diffusion length of the film. Beyond this point, it remains independent of film thickness [35] since the additional film thickness can then support the normal "bulk cooling mechanism" via 1D heat (phonon) transport into depth.
In the femtosecond time domain, light absorption through valence electrons and succeeding hot electron diffusion govern the energy deposition. The electron gas is practically thermally insulated from the lattice of the solid. That means that the lattice initially stays "cold", while the electrons may be heated by ultrashort pulsed laser radiation [36]. Depending on the strength of the electron-phonon coupling, hot electrons can penetrate deeply (compared to the optical penetration depth) into the material before any interaction with the lattice takes place. When an electron temperature (T e ) is established, the heat transport inside the metal can be described by the two temperature model (TTM) [36][37][38]. It consists of a set of two partial differential equations for the electron and the lattice temperature (T l ), coupled by a Fig. 2 Four different scenarios of laser-induced film damage. a Surface damage: the damage is induced at the air/film interface. b Interface damage: the damage is induced at the film/carrier interface. c Surface + interface damage: the damage is induced simultaneously at both the air/film and the film/carrier interfaces. d Bulk film damage: the damage is induced in the bulk of the film material term describing the strength of the electron-phonon interaction (electron-phonon coupling constant g) multiplied by the temperature difference (T e -T i ). Hence, any temperature difference between the electrons and the lattice drives an energy exchange among both sub-systems until a thermal equilibrium is reached. Along with suitable boundary conditions (laser, sample geometry, material properties, etc.) the TTM equations can be solved numerically to gain temporal solutions of the temperature fields of the electrons and the lattice, respectively.
The TTM is the backbone of most computational studies of the interaction of ultrashort laser pulses with solids and has been combined already with other advanced numerical simulations (in 2D and 3D), such as molecular dynamics (MD) or finite-difference time-domain (FDTD) calculations. Explicit representations of the TTM along with numerical solutions can be found in the persistent literature [1,25,[36][37][38][39][40][41][42][43][44][45][46]. Here, selected results applying the TTM are shown in the following. The time dependence of the electron and the lattice temperatures predicted by the TTM for gold and nickel films are depicted in Fig. 3 [38]. Absorption of a single laser pulse with a duration of 200 fs at 400 nm wavelength and a laser fluence of 0.023 J/cm 2 is sufficient to reach the melting temperature (T m ) of the lattices in both cases. The electron and lattice temperatures at the front and rear surfaces of the 100 nm thick films on fused silica substrates are plotted for both metals through solid or dashed lines, respectively. The transfer of energy to the lattice proceeds about ten times faster for nickel due to a much stronger electron-phonon coupling in nickel compared to gold. It is obvious that no significant heat transport through the nickel film to the rear surface takes place, where nearly no temperature difference to the initial sample temperature is calculated even when the front surface reaches the melting point (see the bottom part of Fig. 3). In contrast, for gold (top part of Fig. 3), front and rear surface temperatures are indistinguishable. In this case, the absorbed optical energy can quickly spread through the ballistic transport and diffusion of electrons through the entire 100 nm thick gold film, being confined to it through the poorly electrically conductive fused silica substrate, before the excited electron's energy can be converted to heat via the release of phonons through electron-phonon scattering. Thus, both the front and the rear surface exhibit the same temperature. Figure 3 shows very clearly that the equilibration time for electrons and lattice depends on the electron-phonon coupling constant g [38] explaining the large differences between Ni (g(Ni) = 3.6 × 10 17 W/(m 3 K) [1]) and Au (g(Au) = 2.1 × 10 16 W/(m 3 K) [1]).
Corkum et al. [37] demonstrated in a fundamental work that for bulk metals a critical laser pulse duration τ c , above which the damage threshold fluence shows the usual τ 0.5 dependence for longer (picosecond and nanosecond) pulses, is inverse proportional to the electron-phonon coupling constant g. For laser pulse durations below τ c , the damage threshold remains constant due to linear absorption and energy confinement during the pulses.
Multi-pulse damage threshold fluences of gold films were investigated by Stuart et al. for a 200 nm thick gold grating with 600 pulses per spot at 1053 nm wavelength and pulse durations in a range from 140 fs to 1 ns [6]. The damage threshold was detected by scanning electron microscopy. Later, Wellershoff et al. presented a TTM fit to the experimental data of Stuart et al. [38] (see Fig. 4). In addition to the "standard" TTM, they considered transient absorption due to a change of reflectivity with electron temperature. A good agreement between experimental data and the TTM prediction is found in the entire range of pulse durations if transient optical absorption is included. An arrow indicates Single-and multi-pulse ablation threshold values were investigated for 28-fs laser treatment at 793 nm wavelength [47]. Gold films with different thicknesses between 31 and 1400 nm were deposited on BK7 glass substrates. Damage (ablation) was detected by means of an optical microscope. Figure 5 gives ablation threshold values in dependence on the gold film thickness for multi-pulse laser treatment (N = 100, 1,000, and 10,000 pulses per spot) [47]. For films with thicknesses smaller than 180 nm, increasing threshold values with rising thickness are observed. For thicker films (> 180 nm), the ablation threshold value is constant and remains at its bulk value. A characteristic penetration depth of the pulse energy into the material of L c ≈ 180 nm is observed. Here, L c exceeds the optical penetration depth ∝ −1 of about 12 nm by more than one order of magnitude. This is a direct consequence of the weak electron-phonon coupling in the case of gold discussed above. The resulting electron thermal diffusion length (before interaction with the lattice) is the crucial quantity for fs-laser material processing, especially for the damage threshold.
The specification of multi-pulse damage threshold fluences is suitable particularly for industrial users, since this definition includes the amplification and visualization of small (single-pulse-induced) defects and also includes possible incubation phenomena. Incubation means that for a fixed laser fluence one or more pulses can be applied to the material without macroscopic material removal. After these (few) incubation pulses, which irreversibly modify the material (chemically, surface properties, color center generation, etc.), the material removal starts with the following laser pulses. Incubation effects are well known for many materials and a large variety of laser parameters. It was also observed in the context of the above-mentioned ablation of gold films. The ablation threshold fluence changed between 0.7 J/cm 2 (single pulse) and 0.2 J/cm 2 (10,000 pulses per spot) [47].
The pulse number (N) dependence of the laser-induced damage threshold fluence of (bulk) metals can be described by a power law [48] with F th (1) and F th (N) are the single-and multi-pulse damage threshold fluences, respectively, and ξ is a materialdependent incubation parameter. The model considers mechanical fatigue damage caused by repetitive stressinduced strain and has been found to successfully describe the laser treatment of gold films in a thickness range from 90 nm to 1.5 µm (on fused silica substrates) with femtosecond laser pulses (τ = 200 fs, λ = 400 nm). ξ = 0.92 was calculated [49]. In another femtosecond laser investigation (τ = 28 fs, λ = 793 nm) of gold films, ξ = 0.87 was found [47].
In addition to the investigations on metal films described above, an important experimental finding, which was determined on bulk material, completes this section. Nolte et al. described multi-pulse experiments on the ablation of copper targets with laser pulse durations between 150 fs and 30 ps [25]. For laser pulses shorter than 1 ps and at low laser fluences (close to the ablation threshold), a "gentle ablation regime" was found. Here, the average ablation rate is determined by the optical penetration depth α −1 . For higher fluences, "strong ablation" is observed, where the effective heat penetration depth determines the average ablation rate. For laser pulses with durations longer than 1 ps, the "gentle regime" is not present [25]. It should be noted that the electron-phonon coupling constant of copper amounts to g = (0.9 ± 0.1) × 10 17 W(m 3 K) [50] which indicates a strong coupling of the electrons to the lattice.
Apart from the already discussed optical constraints and influence of the electronic properties of the irradiated solids, also thermophysical effects that determine the relaxation dynamics of fs-laser excited metal films, e.g., via interfacial mechanical stress, hydrodynamic melt flows of the film material, or even substrate de-wetting effects, may have to be considered [51]. In this context, the irradiation of several tens of nanometer thin gold films upon single, tightly focused fs-laser pulses should be illustrated. That topic gained attention around 2004, when a Japanese and a German group independently reported the formation of submicrometric nanojet-like morphologies frozen at the surface of thin gold films [52,53]. Figure 6 assembles a series of side-view scanning electron micrographs of a 60 nm thick magnetron sputtered gold film on a quartz glass substrate after exposure to single Ti:sapphire laser pulses (τ = 30 fs, λ = 800 nm) of varying energies (fluences) that were focused by a 36 × Schwarzschild microscope objective (NA = 0.5) in substrate geometry [53]. For all laser pulse energies between 6 and 13 nJ a pedestal bump with heights up to ~ 800 nm is seen at the film surface. At pulse energies exceeding 7-8 nJ, an additional central protruding nanojet-feature manifests at the surface, exhibiting a pulse energy dependent aspect ratio along with heights up to 1-1.5 µm.
These publications have triggered further systematic theoretical and experimental research, including thermoelastic continuum modeling [54] and atomistic molecular dynamics simulations [55,56], as well as time-resolved measurements [57,58]. The numerical modeling revealed that the bump is caused by thermoelastic stress causing a delamination of the metal film from the dielectric substrate, while the nanojet itself forms from molten film material through intertia accompanying the hydrodynamic melt dynamics and is limited in its height by surface tension effects through the Rayleigh-Plateau instability of the melt [54,56,58]. Side-view shadowgraphic measurements indicated that the complete melting and re-solidification dynamics is terminated withing ~ 250 ns [58]. From their experiments, the authors have drawn the following physical scenario: approximately 5 ns after the impact of the fs-laser pulse, the thin gold film bulges and exhibits a bump of ~ 2.4 μm in height and of ~ 4.5 μm in diameter. At a delay time of ~ 35 ns, the bulge structure collapses and a jet of molten material starts to emerge at its center. While the bump height further decreases, the central jet grows to a length of ~ 4 μm and is continuously expanding until a delay time of ~ 125 ns. After ~ 150 ns, the jet detaches from the bump. Caused by the Rayleigh-Plateau instability the jet disintegrates into two components, i.e., into a small spherical droplet at its tip followed by a larger elongated droplet. Upon detachment, a very thin sub-micrometric spike remains at the base of the structure. At even larger delay times, the shape of the droplets turn spherical through the surface tension of the molten gold, while they are moving away from the substrate at velocities between 20 and 30 m/s [58]. This dynamic evolution explains the strong fluence dependence of the remaining nanojet-like surface spikes observed in Fig. 6.
In view of these time-resolved shadowgraphy experiments and the molecular dynamics simulations, the formation of the permanent nanojet surface features shown in Fig. 6 can be seen as the residuals of a laser-induced film material transfer that is similar to the laser-induced forward transfer (LIFT, see Sect. 4.4 below), with the only difference that here the irradiation is performed in substrate geometry and not in superstrate geometry (as for the LIFT).

Interaction with semiconductors and dielectrics
For "weakly" absorbing materials such as dielectrics or even semiconductors, whose band gap E g (minimum energy difference between valence band (VB) and conduction band (CB)) is larger than the energy hν of the irradiating photons, a different concept for describing material destruction has proven successful: in many of such materials, free charge carriers in the conduction band are present only in small numbers under normal conditions (typically 10 8 -10 10 cm −3 ), but can be generated by photo-excitation. This is done for high intensities in the range between 10 11 and 10 14 W/cm 2 on the one hand by (1) nonlinear multi-photon absorption (MPA) or on the other hand by (2) absorption of the radiation by electrons already being present in the CB (free-carrier absorption), followed by impact ionization. In this context, "ionization" means the generation of an electron-hole pair. The mechanism (2) can lead to an avalanche-like increase of the number of electrons in the CB still during the laser pulse excitation and is, therefore, also called avalanche ionization (AI). An optical breakdown ("laser-induced breakdown'') manifests when a critical electron density N cr e is exceeded in the conduction band of the solid as a result of the irradiation which then ultimately leads to material destruction [5]. This characteristic charge carrier density is material dependent and typically lies between 10 19 and 10 21 cm −3 [1].
For a quantitative understanding of these effects, the time evolution of the charge carrier density N e of the electrons in the conduction band as consequence of the exposure to a laser pulse with an intensity profile I(t) can be described by a single differential rate-equation which takes into account the charge carrier generation in the conduction band by AI (first term) as well as by MPA (second term) [6] (2) Ξ av denotes the avalanche coefficient and γ m the coefficient for the absorption of m photons. Here, m is the minimum number of photons required for a transition between valence and conduction bands (m ⋅ hν ≥ E g ). In the rate equation, loss mechanisms (e.g., carrier diffusion or recombination) have been neglected, which is specifically justified for ultrashort laser pulse durations. It is very instructive to analyze separately the two limiting cases of pure AI or pure MPA. Equation (2) can then be solved analytically by separating the variables. This procedure allows in both cases a statement about the scaling behavior of the materials damage threshold fluence F th with the laser pulse duration τ. However, for high laser intensities, the two effects (AI and MPA) often occur simultaneously, so that the rate equation must then be solved numerically. In particular, for large bandgap materials such as fused silica (SiO 2 : E g ~ 8.3 eV), MPA at very small pulse durations (τ ~ 10 fs) provides an efficient mechanism to generate seed electrons for the subsequent avalanche process. Note that for such short pulse durations and at high intensities also tunnel ionization effects may have to be considered [59].
In the first limiting case of pure avalanche ionization (γ m = 0), the expression N e (τ) = N 0 ⋅ exp Ξ av ⋅ ∫ 0 I(t)dt is obtained immediately after laser pulse excitation, i.e., an avalanche-like (exponential) increase of the electron concentration occurs, explaining the name of this intra-pulse ionization process. The quantity N 0 is the number of electrons already being present in the CB at the beginning of the laser pulse. The expression ∫ 0 I(t)dt represents, by its definition, the fluence F of the laser pulse. Thus, the critical electron density N cr e is just reached at the fluence threshold In the case of pure AI, the damage threshold fluence is, therefore, independent of the laser pulse duration.
In the second limiting case of pure m-photon absorption (Ξ av = 0), rate Eq. (2) yields the solution N e (τ) = N 0 + γ m ⋅ ∫ 0 I m (t)dt . If one assumes a temporal rectangular pulse of intensity I 0 to simplify the mathematical analysis, there is a linear relationship between fluence F and intensity I 0 of the pulse (F = F 0 = I 0 τ). Thus, the pulse duration dependence of the damage threshold can be directly derived: Thus, the damage threshold fluence F th scales with a power law τ (m−1)/m . For m = 1 (linear absorption) the threshold is independent of the pulse duration (consistent with the results gained for metals in the frame of the TTM [37], see Sect. 3.2.1). This result is also in line with the above result for pure AI, which is also based on a cascade of linear absorption processes. It is interesting to note at this point that pure two-photon absorption (m = 2) also leads to the characteristic √ scaling law as it is also obtained by the distinct physical process of 1D heat transport for longer psto ns-pulse durations. In case that several different orders m of nonlinear absorption simultaneously contribute to the carrier generation in the CB, Eq. (2) may be solved numerically but some general trends may be summarized as follows: linear absorption mechanisms produce a constant damage threshold, while multi-photon absorption leads to decreasing thresholds for shorter laser pulses with durations below the electron-phonon relaxation time. Larger values of m manifest in a stronger reduction of the damage threshold fluence with decreasing τ. Therefore, an experimental measurement of F th (τ) with ultrashort laser pulses allows conclusions to be drawn about the physical damage mechanisms being involved [6,19,60,61].
Mero et al. studied the single-pulse damage threshold of five different ion beam sputter-coated oxidic thin dielectric films with band gap energies E g between 3.3 and 8.3 eV (TiO 2 , Ta 2 O 5 , HfO 2 , Al 2 O 3 , SiO 2 ; film thickness h ranging between 500 and 800 nm) on fused silica substrates employing Ti:sapphire ultrashort laser pulses with durations between 25 fs and 1.3 ps [62]. Figure 7a plots the corresponding data of the damage threshold fluence F th as function of the pulse duration τ, while in Fig. 7b the damage threshold fluences of τ = 30 fs (full circles) and 1.2 ps (open circles) laser pulses are plotted vs the band gap energy E g of the film materials. The pulse duration scaling in Fig. 7a suggest the significant involvement of multi-photon absorption processes, whereas a strictly linear dependence on the band gap energy is evident from Fig. 7b. Thus, the damage threshold is determined by intrinsic material properties rather through defects and impurities caused by imperfections in the film deposition process.
Combining their measurements of the pulse duration and the band gap scaling with a rate-equation model extended from the damage models of Stuart et al. [6] and Keldysh [59] by adding a carrier density decay term considering relaxation and re-absorption effects through electronic defects, such as self-trapped excitons (STEs), the authors proposed a phenomenological damage scaling law F th (τ, E g ) according to where c 1 , c 2 and κ are fitting parameters [62]. Values of the exponent κ ranging between 0.27 and 0.33 were found for all five oxidic film materials-rather independent of the actual band gap energy. Based on their retrieved values of the fit parameters to this model, the authors attributed this observation to a significant contribution of AI to the carrier excitation-even at pulse durations as short as a few 10 fs. As consequence, the power law (Eq. (5)) and the exponent κ are rather insensitive to the specific values of the multiphoton and avalanche coefficients. At constant pulse duration the damage threshold fluence exhibits an approximately linear scaling with the band gap energy, pointing toward photoionization as process controlling this behavior [62].
An extended consortium around these researchers studied also the pulse-number (N) dependence of the same set of single-layer film materials, having a special focus on the reduction of the damage threshold for multiple fslaser pulses (780-800 nm, 13-150 fs, 1 Hz-1 kHz), i.e., incubation effects [63]. The decrease of the damage threshold with N was explained with the accumulation of STEs having a lifetime exceeding tens of minutes. The formation of STEs occurs on a time scale much faster than 1 ms and explains the observation of the independence of the damage threshold on the pulse repetition frequency in the range from 1 Hz to 1 kHz. The authors developed a detailed rate equation model for the CB electron density that includes STEs and shallow traps as electronic inter-band defect states. The corresponding set of two coupled differential rate equations was solved numerically to explain the multiple-pulse damage behavior. According to this incubation model, the STEs control the dependence of the damage threshold fluence on N, while the shallow traps are involved in thermal heating effects at low laser fluences. Figure 8 exemplifies the capabilities of this incubation model for a ~ 550 nm thick Ta 2 O 5 film on fused silica upon irradiation at 1 kHz pulse repetition rate (800 nm, 30 fs) [63]. The data points (full squares) were obtained in experimental N-on-1 measurements of the damage threshold fluence, normalized to the single-pulse threshold value that accounts to F th (1) = 0.6 J/cm 2 here. The solid line represents a numerical fit to the complete set of the two rate equations (for details see [63]), while the dashed line is based on a least-squared-fit to the approximating analytical model presented in Eq. (6): In this power law, m is again the order of the MPA process and accounts to m = 3 for the Ta 2 O 5 material. The last term in round brackets on the right-hand side controls the asymptotic approach to the value F m th (∞) at a very large number of laser pulses. It is determined by the ratio of the band-to-band relaxation time τ 1 and the creation time τ 2 of the absorbing inter-band electronic states, as well as by the ratio of the electron density n 1 generated by a single laser pulse and the maximum density of trap states n 2 .
In 2015, Sun et al. introduced a generalized incubation model for the irradiation of film and bulk materials that is based on two physical mechanisms, i.e., (i) laser pulse induced change of the optical absorption and (ii) the laserinduced decrease of the specific energy that is required to damage the sample [64]. Given the very general nature of these mechanisms, the model is applicable to a broad class of materials (metals, semiconductors, dielectrics) and can explain changes of the damage threshold fluence as a function of the laser repetition rate. Details about the mathematical implementation of the model and a software to fit experimental data can be found online [65].
Driven by the potential of industrial applications, largearea surface structuring by lasers is continuously moving into the focus of scientific and technical investigations. For dielectric materials, this is complicated due to the large band gap energies, typically requiring nonlinear laser-matter interactions for material excitation when the single photon energy is significantly smaller than the band gap energy. Along with the material-specific incubation effects discussed above, this makes a homogeneous large-area surface structuring very challenging and highly dependent on the laser processing parameters, such as the spatial beam shape, peak fluence, pulse repetition frequency, laser scanning velocity (both define the spot-overlap within a scan line), and the interline separation. One way to overcome this difficulty can lie in adding a very thin but strongly absorbing cover-layer on the surface of the transparent dielectric to facilitate resonant coupling effects of the laser radiation into the material underneath the cover-layer.
As an example, Kunz et al. [66] demonstrated that through this approach employing a 20 nm thick gold coverlayer large surface areas can be laser-processed by nearinfrared fs-lasers (1025 nm, 300 fs, 100 kHz) on fused silica glass samples to be homogeneously covered by subwavelength surface nanostructures, so-called laser-induced periodic surface structures (LIPSS, for more details on their formation and applications the reader is referred to [67][68][69][70]). A scheme of their approach is shown in Fig. 9a along with the generated HSFL in Fig. 9b.

Transient effects
In this section, complex transient phenomena are pointed out that may affect the film processing with ultrashort laser pulses. This includes extremely fast (non-thermal) intrapulse effects that can occur already during the laser pulse, but also slower inter-pulse processes that are just initiated by the laser pulses but may occur on much longer time scales up to the microsecond range. This manuscript part is divided in two sub-sections, where specific effects of "thin film" laser-matter interaction are discussed (Sect. 3.3.1) and complemented to effects being relevant for "thick films" (Sect. 3.3.2).

Thin films: surface ablation and interfacial spallation
For thicknesses h < λ < Rayleigh length (DOF) films will be considered as "thin" here. As already discussed in Sect. 3.2.1, for strong absorbing materials such as metals there may be a characteristic linear dependence of the ablation threshold fluence on the film thickness if h is smaller than the thermal diffusion lengths of the electrons. However, for transparent films that are deposited on strongly absorbing carriers, the sample damage typically follows other pathways. This originates from the fact that the laser radiation is not significantly absorbed or scattered in the transparent film before being deposited in the carrier material close to the film/substrate interface. Additionally, the transparent film may act as classical Fabry-Perot resonator that is affecting the amount of back-reflected radiation via the coherent superposition and interference of the Fresnel reflections occurring at the air/film and film/substrate interfaces. Moreover, the presence of the transparent thin films may hinder the excited substrate material from being ablated by covering the laser-excited material, leading to a situation of "frustrated ablation". As a consequence, for a sufficiently thick transparent film, the ablation threshold fluence of such a film-substrate combination may be even higher than the threshold value of the strong absorbing substrate material itself.
The dynamics of laser ablation by single fs-laser pulses has been studied by McDonald et al. through complementary fs-time-resolved pump-probe side-view shadowgraphy and top-view microscopy on the ps-ns time scale for a set of silicon wafer samples covered by different thermally grown oxide thicknesses varying between a few tens of nanometers up to h = 1.2 µm [71]. Figure 10a compiles the corresponding side-view shadowgraphy images for five selected pump-probe delays Δt (columns) of 0.21 ns, 2.35 ns, 5.01 ns, 7.68 ns, and 10.35 ns as well as for five different oxide layer thicknesses (rows) of 20 nm, 54 nm, 147 nm, 300 nm, and 1,200 nm, respectively, upon irradiation with The dark/blue regions represent ablating material as well as the sample surface. The vertical fringes around the surface are caused by edge diffraction of the probe beam radiation. For the 20 nm thin oxide film sample, at large delays (10 ns) the front of a shock wave formed in air is visible as a bright semi-circular boundary. It can be distinguished from the dark appearing material front that lags behind. For medium oxide film thicknesses of ~ 50 nm to ~ 150 nm, the ablated material front appears fractured at that time delay. For even larger film thicknesses (300 nm, 1,200 nm), the spatial expansion of the material is strongly reduced. From these side-view shadowgraphy measurements, for each oxide layer thickness, the position of the material front was extracted and plotted as a function of the time delay Δt (Fig. 10b). For all data sets, a linear scaling with Δt can be seen that allows to quantify the velocity of the ablating material as the slope of the least-squares-fits. Values ranging between v film = 3010 ± 360 m/s (h = 20 nm), exceeding the speed of sound in air by a factor of ~ 10, down to 200 ± 20 m/s (h = 1200 nm) were obtained [71].
In 2013, Rapp et al. extended the studies of the same material system (SiO 2 films of different thickness on Si wafers) to the earlier (ps-ns) and later (10-100 ns) ablation stages through fs-time-resolved pump-probe top-view microscopy of the surface reflectivity upon single fs-laser pulse irradiation [72]. This technique was initially developed by Downer et al. in 1985 [73] and further improved by Sokolowski-Tinten and von der Linde [74]. This led to a significantly improved understanding of the physical processes involved in ultrashort pulsed laser ablation, including the transient appearance of Newton-fringes caused by the formation of a stress-related rarefaction wave [8]. Meanwhile, fs-time resolved microscopy (fs-TRM) has been extended into many different powerful imaging modalities (see Refs. [75][76][77] and references therein). Figure 11a presents a collage of brightfield fs-TRM images of spots at SiO 2 -film covered (100)-silicon wafer surfaces upon irradiation by single fs-laser pulses (1053 nm, 660 fs, 1.0 J/cm 2 ) [72]. For two samples of different oxide film thicknesses (upper row h(SiO 2 ) = 100 nm; lower row: h(SiO 2 ) = 500 nm), the images were acquired after ten specific delay times Δt of the illuminating probe pulses ranging between 0 ps and 100 ns. The right image in each row labeled with "∞" displays the permanent laser-induced surface modification for comparison.
At a delay time of ~ 1 ps the reflectivity of the laser-irradiated spot appears bright due to the melting of the silicon (Fig. 11a). After ~ 10 ps, the reflectivity in the center of the spots drops again and develops into a transiently changing ring pattern that is persistent between several hundreds of picoseconds and a few nanoseconds. It is caused by interference effects through Newton fringes. The general principle of their formation through the interference of a reflection at the ablation front and a reflection by the sample surface after passing twice through the semi-transparent ablating material is sketched in Fig. 11b (taken from [8]). Note that the characteristic Newton ring pattern develops later for the thicker oxide layer, which indicates that a thickness (and fluence) dependent bulging of the oxide layer occurs here. After several nanoseconds, the Newton rings disappear due to the fragmentation of the ablated SiO 2 layer into flying particles and fragments.
Based on these results, Fig. 12 sketches the temporal evolution of the single fs-laser pulse induced ablation of thermally grown transparent SiO 2 layer from the underlying silicon carrier [72].
An important remark must be made here for the case of fs-laser ablation of extremely thin (native) oxide layers  [8]. Reprinted from [8], Appl. Surf. Sci., Vol. 154-155, von der Linde, D. and Sokolowski-Tinten, K., The physical mechanisms of short-pulse laser ablation, 1-10, Copyright (2000), with permission from Elsevier usually being present on semiconductor wafers, where a distinct fs-laser-induced ablation behavior is observed. Such native oxide layers typically exhibit a few nanometer thickness only. For silicon, they grow once the material is exposed to the ambient air environment and are self-limiting in thickness (h ~ 3 nm). Detailed post-irradiation analyses by atomic force microscopy (AFM) and spectroscopic imaging ellipsometry (SIE) proved that the single fs-laser pulse destruction threshold of the native oxide layers on silicon lies above the melting threshold of the semiconductor but below its ablation threshold [20,78]. The effect is fully in line with thermodynamic considerations, where the enthalpy of decomposition of the native oxide layer lies above the enthalpy of melting of the semiconductor but below its enthalpy of evaporation [78]. In other words, the native oxide layer is thermally removed via indirect heating of the semiconductor underneath. This scenario was also directly confirmed by fs-TRM for germanium wafers through the observation of low-contrast transient Newton fringes manifesting at peak fluences F m ≤ F 0 ≤ F abl below the ablation threshold of the semiconductor [79] and will be illustrated in the following.
Th effect is demonstrated in Fig. 13, which reveals the removal of the native oxide layer from a germanium wafer surface in the fluence regime of thermally melting the germanium through fs-TRM performed at delay times Δt = 1 ns and Δt = 10 ns (Fig. 13a) [79]. In this sub-ablative fluence range no crater is formed permanently at the germanium surface, in contrast to peak fluences F 0 exceeding the ablation threshold (Fig. 13b, compare the images labeled by "∞"). For both cases (fluences), Newton fringes can be seen at Δt = 1 ns. At the lower fluence (F 0 = 0.63 F abl , Fig. 13a) in the melting regime, re-solidification occurred almost completely after 10 ns, while at the higher fluences (F 0 = 2.35 F abl , Fig. 13b) the ablated material still reduces the reflectivity by screening the probe radiation from the surface. However, interference in the form of Newton fringes does not occur anymore since the layer initially ablated with sharp interfaces (Fig. 13c, 1 ns) started to disintegrate at the longer delay times (Fig. 13c, 10 ns). The disappearance of the spatial fringes in fs-TRM for high fluences and long delays can be attributed to the modification of the mass density profile of the ablating layer that is then featuring a decreased sharpness of the interfaces [79,80], preventing the interference effect to occur.
The intrinsic geometrical asymmetry of a thin film-coated transparent carrier allows to distinguish between the two scenarios of fs-TRM in which the fs-laser excitation of the film by the pump beam is carried out through the air environment or through the transparent substrate, while the probe beam path is kept identical. Such a complementary approach, where the film is excited either through the air/ film or the substrate/film interface, can reveal the dynamics of differences in the reflectivity contrast and through that allows to distinguish ultrafast (non-thermal) electronic material changes from significantly slower thermal modification effects or even from mechanical deformation, e.g., through thermal melting, shock wave propagation, blister formation, etc. Moreover, the substrate-side irradiation geometry is relevant for applications such as laser-induce forward transfer, LIFT (see Sect. 4.4).
For this reason, Domke et al. studied in detail the ablation of 470 nm thick, highly absorbing metal (Mo) films on transparent glass substrates through fs-TRM upon both, film-side and substrate-side fs-laser pulse excitation (1053 nm, 660 fs) employing peak fluences between 0.5 and 1.0 J/cm 2 [81]. Their time-and space-resolved observations in the glass-side irradiation scenario revealed phase-explosion, generating a liquid-gas mixture in the Mo/glass interface approximately ten picoseconds after the impact of the pump laser pulse, as the driving mechanism behind the lift-off of a coin-shaped Mo-film disk. Subsequently, a shock wave and gas expansion cause the Mo-film to bulge and form a blister at the Mo/glass interface, while the enclosed liquid-gas mixture cools and may condense at delay times in the hundreds of picosecond range. The bulging of the Mo-film to heights up to ~ 1.4 µm emerges for approximately twenty nanoseconds, until an intact Mo disk of a few tens of micrometers in diameter shears and lifts off at a velocity of ~ 70 m/s, leaving behind a pronounced hole in the film, without significant debris or any substrate material damage [81].
Except transient non-thermal, subsequent thermodynamic, or mechanical effects, also active optical emission may transiently occur in thin films during its , who studied optical damage and fracto-emission effects upon irradiation of 60 nm thick amorphous diamond-like carbon films on glass or silicon wafers upon irradiation with single fs-pulses of an amplified colliding-pulse mode-locked rhodamine 6G dye laser (620 nm, 120 fs) [82]. The effect is exemplified in Fig. 14 via a collage of top-view brightfield optical micrographs (left column) and corresponding wavelength integrated self-emission images (right column) assembled for irradiation at four different peak fluences ranging between 0.27 J/cm 2 (top) and 0.35 J/cm 2 (bottom). At higher peak fluences exceeding 0.4 J/cm 2 , no fragmentation or fracto-emission was observed in the center of the laser-excited region [82], but instead the material-independent universal ablation behavior previously described in Fig. 13b.
It is evident that the laser-irradiated but non-removed film fragments still attached in border of the ablated hole area exhibit a strong optical emission. Optical filters indicated broadband emission from the blue into the near infrared spectral range. The temporal evolution of the optical emission was analyzed by replacing the imaging CCDcamera with a spatially integrating photodiode-oscilloscope detection system. The measurements (data not shown here for brevity) identified a weak but "fast" emission on the 100-200 ns time scale followed by a much stronger but "slower" main emission peaked around 0.5-1 µs that can be associated with ~ 90% of the totally emitted light [82]. The results revealed that the optical emission occurs mainly after fracture of the thin film and the re-arrangement of its fragments. Emission from a laser-generated free-electron plasma or ablation plasma could be safely excluded. Instead, the authors proposed that the optical emission can be explained by the high internal compressive stress of the order of a few GPa in the films caused by the employed pulsed laser deposition process. The fs-laser-induced impulsive and isochoric heating sufficiently reduces the adhesion forces to the substrate and, thus, the film is removed from the substrate without transition to the gas phase. The internal mechanical stress causes also the destruction of the film and the ejection of its fragments. The authors suggested that the early emission peak can be associated with the initial detachment and breakup of the film, while the slower emission component may be related to the final relaxation of the internal stress after the formation of the individual film fragments [82].
Except transient non-thermal, subsequent thermodynamic, or mechanical effects, also active optical emission may transiently occur in thin films during its ultrashortpulsed removal. This was demonstrated by Sokolowski-Tinten et al., who studied optical damage and fracto-emission effects upon irradiation of 60 nm thick amorphous diamond-like carbon films on glass or silicon wafers upon irradiation with single fs-pulses of an amplified collidingpulse mode-locked rhodamine 6G dye laser (620 nm, 120 fs) [82]. The effect is exemplified in Fig. 14

via a collage of top-view brightfield optical micrographs (left column) and
corresponding wavelength integrated self-emission images (right column) assembled for irradiation at four different peak fluences ranging between 0.27 J/cm 2 (top) and 0.35 J/ cm 2 (bottom). At higher peak fluences exceeding 0.4 J/ cm 2 , no fragmentation or fracto-emission was observed in the center of the laser-excited region [82], but instead the material-independent universal ablation behavior previously described in Fig. 13b.
These examples show that a plethora of complex physical processes can transiently manifest during the structuring of thin films by ultrashort laser pulses. However, additional chemical or structural changes of the films may take place between successive laser pulses hitting the irradiated material. An obvious inter-pulse chemical effect is the laserinduced oxidation that may occur if the fs-laser irradiation is conducted in the ambient air since the laser-heated surface meets a reactive chemical environment. Such laser-induced effects were explored in detail around the turn of the millenium for the fs-laser irradiation of TiN hardcoatings as used for improving the lifetime of mechanical tools [3,17,18]. It was found that a specific sub-ablative laser fluence range exists in which the TiN is transferred into sub-stoichiometric TiO 2-x . Ion-sputter depth-profiling revealed in combination with X-ray photo-electron spectroscopy (XPS) [17] and spectroscopic Auger electron microscopy (AEM) [17,18] that the TiN film material may be chemically altered up to a depth of a few hundreds of nanometers below the residual film surface upon multi-pulse fs-laser processing.
In the following, we want to exemplify another inter-pulse structural effect that was observed upon single fs-laser pulse irradiation (790 nm, 35 fs) of ~ 0.9 µm thick hydrogenated amorphous carbon films (a-C:H) that were deposited by chemical vapor deposition on 300 µm thick (111)-silicon wafer substrates [83]. Such a-C:H films typically contain 15-20% (at.) of sp 3 hybridized carbon. The optical penetration depth at the laser wavelength was evaluated through ellipsometric measurements and accounts to 1/α ~ 380 nm for the pristine a-C:H films. A series of irradiation spots was generated at the film using different peak laser fluences between ~ 0.2 and ~ 0.4 J/cm 2 and subsequently characterized by optical microscopy (OM), white light interference microscopy (WLIM), micro Raman spectroscopy (µ-RS), and microscale mechanical indentation tests for evaluating the microscopic, topographical, structural, and mechanical properties after the fs-laser treatment. Figure 15 demonstrates by combining results of OM, WLIM, and µ-RS that structural changes can be generated through laser-induced graphitization in the film material at laser fluences below the ablation threshold. In Fig. 15a, a top-view differential interference contrast optical micrograph is provided along with a superimposed cross-sectional WLIM topography profile at the bottom. An ablation depth of ~ 25 nm of the film material is evident in the center of the irradiated spot, surrounded by a region that is protruding up to ~ 90 nm above the original film surface plane. Figure 15c, d provides µ-RS spectra taken at different locations in the fs-laser-irradiated spot (labeled "B" and "C" in Fig. 15a) along with a reference spectrum of the non-irradiated surface (Fig. 15b, recorded at spot "A").
The characteristic signature of the so-called D-and G-bands of the nanocrystalline carbon system is visible in all Raman spectra, with broad peaks located around ~ 1350 cm −1 (D) and ~ 1600 cm −1 (G) here. Due to the fs-laser irradiation, the Raman signal intensity ratio I(D)/I(G) of the two bands increases toward the center of the fs-laser-irradiated spot (see Fig. 15c, d). At the same time, the spectral width of both bands decreases, indicating a higher degree of structural order in the laser-modified areas. Both is indicative of a fs-laser-induced graphitization of the a-C:H material, i.e., a decrease of the sp 3 and an increase of the sp 2 fraction in the films. This explains also the swelling of the film material observed in the WLIM topography through a reduced specific mass density. A breakage of the sample through a suitable spot and subsequent cross-sectional inspection in SEM confirmed that no delamination occurred at the film/ substrate interface in the laser-irradiated regions here. Thus, a blister formation could be ruled out and it was concluded that the elevated film surface topographies were solely generated by structural material changes in the film material [83]. The fs-laser-induced graphitization was further confirmed by complementary mechanical indentation measurements that indicated a significantly reduced material hardness (∼50%) in the irradiated regions, as expected by a reduced content of sp 3 hybridized carbon [83].

Thick films: bulk damage through interference effects
The ablation of thick films is usually dominated by other effects if h > Rayleigh length (DOF) > λ is fulfilled. Particularly for strong absorbing materials such as metals and semiconductors, it is clear that in the limit of very thick films (h ≫ λ, DOF) the processing/ablation must be identical to that of the corresponding bulk material (as already revealed in Sect. 3.2). However, for (semi-)transparent materials, if the film thickness is of the order of the wavelength and smaller than the Rayleigh length (DOF > h ≥ λ), another often neglected intra-film interference effect may affect and limit the film ablation for specific film/substrate combinations, as it will be detailed in the following.
To the best of our knowledge, the intra-film interference effect was reported first in the context of laser-material processing in 2001 in post-irradiation experiments of a Japanese group studying the single-pulse ablation of ~ 15 µm thick organic m-MTDATA films (amorphous substituted triphenylamine: 4,4′,4ʺ-tris(3-methylphenyl phenylamino) triphenylamine) being transparent for the laser irradiation wavelength (780 nm, 170 fs) from the surface of quartz substrates [84]. In stylus-profilometric measurements, the authors observed fluence dependent step-like crater profiles as reprinted in Fig. 16a. The authors proposed an interference effect, where the front part of the fs-laser pulse is partially reflected at the interface between the organic film and the quartz substrate and can interfere with its following tail. Hence, a standing light wave is formed transiently within the film along the light propagation direction, exhibiting maxima at a spatial separation of d = λ/(2n). For the refractive index n ~ 2 of the m-MTDATA polymer film material, d is expected to be ~ 195 nm here, reasonably supported by the step-height data derived from the crater profiles presented in Fig. 16b.
About four years later, the same standing light-wave based mechanisms was identified to manifest in in-situ b Depth of ablation steps as a function of the peak fluence. Reproduced from Hosokawa, Y. et al. [84], with permission from Japan Laser Processing Society time-resolved reflectivity measurements during the fs-laser irradiation of ~ 1 µm thick triazenepolymer (TP) films on glass substrates [61,85]. The effect is responsible for the intricate reflectivity behavior observed only in a very narrow fluence range below the permanent film damage threshold, as detailed in the following. Figure 17a sketches the experimental setup of the real-time resolved reflectivity measurements. Single fs-laser pulses (800 nm, 130 fs) are focused under normal incidence to the TP-films in substrate geometry to a beam spot diameter (1/e 2 ) of ~ 100 µm [86]. The probing cw-radiation of a single-mode 514.5 nm Arion laser is focused to a ~ 30 µm spot in the center of the fs-laser-irradiated spot at an angle of incidence of θ = 18°. The reflected and re-collimated and spectrally cleaned probe beam is simultaneously monitored on short and long time scales via a combination of a streak-camera (time window 50 ns, sub-ns resolution) and a photodiode (time window 2 µs, few ns resolution). Figure 17b, c exemplify three typical transients of the normalized reflectivity change ΔR/R 0 (normalized with the reflectivity value R 0 of the non-excited film) for three different peak fluences below the value of the single-pulse damage threshold of the film (F th = 0.5 J/cm 2 ). Above that threshold fluence complex and rather chaotic transient reflectivity changes are emerging (data not shown here; for details see [61,85]). Below the damage threshold, generally the reflectivity rapidly drops to a minimum (R 1st peak , see curves I to III in Fig. 17b) before recovering to the initial surface reflectivity (ΔR/R 0 = 0) on the microsecond and millisecond time scale. For peak fluences close to the damage threshold, a second reflectivity minimum (R min , see curve III in Fig. 17c) may be observed on the sub-microsecond time scale. Interestingly, in a very narrow fluence range, the reflectivity is transiently increased (see curves II). This becomes visible when plotting the three characteristic normalized reflectivity values R 1st peak , R min , and R 1.6 µs as a function of the peak laser fluence in Fig. 17d, see regime II. The reflectivity increase can be explained by a transient intra-film Bragggrating that is formed at a suitable fluence when the standing light wave modifies the refractive index of the polymer just at its intensity maxima being spaced at d = λ/(2n), see the explanatory scheme in Fig. 17d [86]. The coherently superimposed Fresnel reflections from the equidistant spatial refractive index modulations can then constructively interfere in the direction of the probe beam featuring a transient Fabry-Perot mirror effect here.
In 2014, a Canadian group transferred this concept of "light sheet ablation" (see Fig. 18) to inorganic materials and systematically explored the effect for SiN x films with thicknesses varying between 0.02 and 1.5 µm deposited on silicon wafers and for various ultrashort-laser wavelengths [87,88]. Remarkable narrow intra-film damage regions of a few tens of nanometers depth only can be reached through the involvement of nonlinear absorption effects triggered by the ultrashort laser pulses (522 nm, 200 fs) along with a few micrometer confined circular top-hat profile (Fig. 18).
Finally, some general aspects of the interference-based intra-film structuring mechanism should be summarized: (i) Only films with thicknesses h > λ/(2n) are showing this effect. (ii) Spatial confinement of the standing light wave pattern must be supported by a suitable ratio between the Rayleigh length and the film thickness (DOF > h). (iii) Temporal confinement must be enabled, i.e., the coherence of the laser beam should be sufficiently large to allow forming the intra-film standing light wave. (iv) The effect manifests only in specific fluence windows, where material modifications form selectively at the maxima of the standing light wave. In case the ablation threshold is locally exceeded, the film can disintegrate through the removal of some of the topmost intra-film segments, potentially leaving behind some blistered sub-surface regions. (v) A highly reflective substrate material and a spatial top-hat laser beam profile may help Fig. 18 a Scheme of light sheet ablation of a 500 nm thick transparent silicon nitride film on silicon, sketching the segmentation of the focal interaction volume of the incident 522 nm laser radiation through several thin interaction zones (orange discs) as a result of the interference of Fresnel reflection of the incident light from the film/ substrate interface. b The modulated intensity profile (solid line) calculated for a Gaussian-shaped beam of ~ 1 µm (1/e 2 ) diameter along with the electron density profile (orange shaded) expected inside the film at the exposure of 9 × 10 12 W/cm 2 average incident intensity. The electron density exceeds the critical plasma density (n cr ) at the interference maxima positions, manifesting in the intra-film structuring simultaneously observed in scanning electron micrographs in (c). The irradiation results in the ejection of film segments S1 and S2, and the blistering of the segments S3 and S4 to form a nanovoid at the fourth fringe maximum, and blistering of segment S5 overlying a second nanovoid. As indicated by red lines, the positions of the cleavage planes align with the fringe maxima positions [87]. Reprinted from [87] by permission from Springer Nature: Light: Sci. Appl. 3:e157 (Quantized structuring of transparent films with femtosecond laser interference, Kumar, K. et al.), Copyright (2014) to optimize the interference contrasts and, thus widen the observed fluence window. (vi) The effect occurs only if a sufficient part of the laser pulse energy is being reflected at the film/substrate interface. Thus, the linear transparency regime of the film material is favored but an interplay with nonlinear absorption or incubation effects at the intensity maxima may help to confine the damage regions to very thin "light sheets". (vii) The intra-film interference effect may prevent that the films can be completely ablated from the substrate material [85].

Repair of photolithography masks
Due to an increasing demand of highly integrated semiconductor circuits and the simultaneous decrease of their structural sizes, suitable tools for photolithographic mask repair are needed by the semiconductor industry. Binary photolithographic masks are typically consisting of metallic chromium (Cr) films of ~ 100 nm thickness on quartz substrates. Defects on the photolithographic masks often manifest in the form of excess absorber (Cr). A transfer of the pattern to the wafer would then result in wafer-level wiring errors which render the integrated circuit inoperative. Since the manufacturing process of a photolithographic mask is very expensive, a defect removal is highly desirable [89] and the relatively high cost for a fs-laser system can be afforded.
Usually, the repair of these binary photomasks is done by the focused ion beam (FIB) technique which has a high lateral resolution but is limited by the ion-induced damage of the quartz substrate. To overcome this drawback, two concepts of fs-laser-based optical mask-repair systems have been realized first by two companies in 1998/1999, employing a far-field [90] and a near-field [91] based structuring technique. Both methods take benefit from the substantially different ablation threshold fluences of chromium films and the underlying dielectric material, i.e., F abl (Cr) < F abl (quartz).
The near-field structuring of chrome films was realized by Liebermann et al. (Nanonics Lithography Ltd.) by coupling the third harmonic of Ti:sapphire femtosecond laser pulses (260 nm) into a light guiding hollow, micropipette (chromium coated fused silica, aperture diameter ~ 400 nm) which was used then in the scanning near-field mode [91]. Laterally protruding excess chrome defects with sizes of 3 µm were carved-off from a 100 nm thick Cr-layer with an accuracy of approximately 50 nm.
The far-field structuring technique was realized by Haight et al. (IBM) imaging a movable rectangular aperture (projection mask) placed in the beam of a femtosecond oscillator Fig. 19 Top-view SEM micrographs of IBM photomasks repaired with ns-laser pulses (top) and fs-laser pulses (bottom) using the MARS I system [89], High resolution material ablation and deposition with femtosecond lasers and applications to photomask repair, Haight (800 nm, 100 fs) directly to the sample surface (photomask: Cr on glass) [90,92]. For that purpose, a high numerical aperture objective was used (projection mask demagnification factor 100). Sub-diffraction limited ablation spot sizes were realized capable to remove 600 nm wide mask defects. The first generation of this device (Mask Repair System, MARS I) was installed in April 1998 in IBM's Burlington Mask House for routine manufacturing operation [92]. Figure 19 compares the repair of a 100 nm thick Cr-film on glass based binary mask with nanosecond laser pulses (top micrograph) and with fs-laser pulses (bottom micrograph) employing the MARS I system [89]. In the top SEM micrograph (ns-repair), a bridge of chromium connecting two horizontal chromium lines was removed, resulting in some redeposited chromium particles evident throughout the entire picture. For direct comparison, the bottom SEM micrograph (fs-repair) displays a photomask region that was ablated with a 100-fs laser in a square-shaped and a rectangular area. Very sharp chromium edges and no debris or metal splatter are evident, demonstrating the superior quality of the mask repair through ultrashort fs-laser pulses compared to longer ns-laser pulses.
The second generation of the mask repair system (MARS II) has been installed in late 2001 as IBM's primary repair tool for 248 nm and 193 nm wavelength chrome on glass phase shift photolithographic masks [93]. In this system, a frequency-tripled Ti:sapphire femtosecond laser beam (260 nm, 100 fs, 1 kHz) is focused to a diffraction limited 150 nm Gaussian spot and scanned over the defect to remove it. An ablation resolution of 80 nm together with an increased flexibility in ablating complex structures were demonstrated. Mask features were trimmed to a root-meansquare precision of ~ 5 nm [93].
Later, the experiments at IBM were extended to generate sub-200-nm fs-laser radiation, using a regeneratively amplified Ti:sapphire system (800 nm, 30 fs pulses 1 mJ, 1 kHz) [94]. After frequency-doubling the fundamental Ti:sapphire laser radiation to 400 nm, the laser pulses passed through an appropriate delay line and were mixed with remaining 800 nm light in a hollow-core glass fiber resident inside an argon-gas-filled cell, finally converting the ultrashort pulsed input radiation via sum-and difference-frequency mixing into wavelengths around 160 nm, 193 nm, and 266 nm. After spectral separation, pulse energies up to 0.5 µJ were realized at 193 nm wavelength, confined in ~ 30 fs pulse duration. As for the ablation using 266 nm wavelength (100 fs) pulses, the chromium layer was completely removed without any observable damage to the underlying fused silica substrate, while the shorter 193 nm wavelength bear the potential of further improved laser processing precision [94].

Laser cleaning
Laser cleaning with the aim of removing an (unwanted) thin (or even thick) film or contaminants from a surface without damaging the latter was already reported about 50 years ago. In this context, questions of the restoration of cultural assets were addressed [95]. This seems surprising only at first glance. Due to possible aggressive environmental conditions, the preservation of cultural assets is often a race against time. Hence, a reliable non-contact method like laser processing is highly welcome [96]. Thin films in that respect have thicknesses up to the order of 100 µm as in the case of total paint film thicknesses in the aircraft and automotive industry. In restoration, e.g., aged varnishes, overpaintings, or encrustations are typically involved. Here, the thickness range covers both thin (< 100 µm) and thick films (> 100 µm) up to the millimeter domain.
The general principle of laser cleaning to uncover buried surfaces by means of laser radiation is based on the selective absorption of laser radiation in the surface-covering layer to be removed (compare substrate geometry in Fig. 1a). Cleaning is possible if the applied laser fluence can cause an ablation process of the layer. This happens when the energy absorbed per unit volume minus the losses (e.g., due to heat conduction for longer laser pulses) is above a material-dependent fluence threshold value. For thin films, the ablation threshold might depend on the film thickness (see Sect. 3.2.1).
A promising situation for laser cleaning is when the optical and thermal properties of the contamination layer and the object to be preserved are very different. This is the case, for example, when an absorbent (black) film should be removed from a highly reflective (white) substrate, e.g., for contaminated paper or parchment. Ideally, the object to be preserved should have significantly greater thermal resistance to the laser radiation than the film. The benefit of using lasers results from the facts that processing is done in a contact-less and site-controlled manner. The laser parameters (wavelength, pulse duration, repetition rate, focusing conditions, fluence) can be adapted to the specific laser cleaning application. Moreover, often there is the possibility of self-limiting of the process, i.e., one can reach an "etch stop" at the interface to the object. When applying a scanprocessing approach, where a (focused) laser beam is moved across the sample surface, large surface areas can be cleaned from contaminants. Up-to-date laser and scanner technology enables efficient laser cleaning rates at the m 2 /s level in an industrial environment [97].
Laser cleaning is not limited to cultural assets. It plays an important role for many technical applications when contaminations are to be removed or when the surface of workpieces must be prepared for further processing. Up to now, a large variety of objects and technical surfaces consisting of diverse basic materials were treated with lasers emitting at various wavelengths in different pulse duration regimes. Here, only an excerpt from the extensive literature that is considered as representative can be briefly discussed.
Metal surfaces were cleaned from organic contaminants [98][99][100][101] and from paint coatings [102][103][104]. The latter has great application potential in the aircraft, automotive, and shipbuilding industry. The removal of a 20 µm thick zinc coating from steel was also reported [105]. And vice-versa, the cleaning of surfaces to allow for subsequent painting is discussed too, e.g., to prepare carbon fiber reinforced polymer (CFRP) surfaces [106].
Nowadays, laser cleaning can be considered as a wellestablished technique in the field of stone conservation. Cooper et al. demonstrated the capabilities of Nd:YAG nanosecond laser radiation to remove the black crust from a historic limestone sculpture. They also showed that the addition of a water layer prior to laser irradiation might be helpful for the removal of material [107]. The potential for practical laser graffiti-removal systems on the basis of Nd:YAG lasers was shown [108]. Undesired graffiti-as a cultural side effect-is a growing problem worldwide, affecting a large number of buildings and monuments worthy of conservation. Recently, femtosecond pulse laser cleaning was discussed in this context [109].
During the "early phase" of the application of laser cleaning, excimer lasers played an important role, e.g., for the restoration of paintings [110] and the processing of organic materials such as parchment [111] and paper [112]. However, the high photon energies of UV emitting lasers can lead to photochemical degradation of the cleaned objects [113] that often must be prevented. Hence, lasers emitting in the optical or infrared spectral range moved into the focus, particularly when the low-repetition rate gas-laser technology was replaced by high-repetition rate solid-state laser systems.
Since the turn of the millennium, also femtosecond lasers were explored for cleaning purposes. Initially, a specific problem within painting conservation, namely the removal of glue-paste adhesive from the reverse side of a painting on canvas, was addressed [114]. Ultrafast laser pulses in the femtosecond range, in some cases also with ultraviolet wavelength, were used for cleaning on a variety of objects from very different material classes, e.g., metals [115,116], painted artifacts [117][118][119], granite [120], parchment [121], paper [122,123], and wooden artworks [124].
Despite the multitude of options in the choice of laser irradiation parameters, including the availability of ultrashort laser pulse durations that enable a highly localized and precise material treatment, the surface cleaning of cultural heritage objects is still a delicate and often irreversible process. The optimization of the laser cleaning parameters for the object to be treated plays a major role to avoid any undesired side effects like, e.g., yellowing, discoloration or darkening [113,[125][126][127].
Generally, laser damage threshold fluences are higher for longer laser pulses, regardless of the irradiated type of object (compare Sects. 3.2.1 and 3.2.2). For a fixed laser wavelength, then the laser cleaning thresholds for the removal of the soiling or contamination should be lower than the damage thresholds for the objects independently on pulse duration. The "cleaning window", i.e., the working range with respect to the laser fluence, for a successful laser cleaning procedure might be larger for longer (nanosecond) pulses than for ultrashort (femtosecond) laser pulses [122]. When the cleaning efficiency and quality is comparable in both cases (or even better for higher usable laser fluences in the nanosecond case and below the corresponding damage threshold of the object), then also the application of the longer laser pulses can be more reasonable. Figure 20 demonstrates the capabilities of laser cleaning through an optical micrograph that allows directly comparing an untreated soiled silk fabric region (right) with a Nd:YAG ns-laser cleaned silk area (left) [128].
A complementary "cleaning application" is enabled by the precision and minimal thermal damage (heat-affected zone) to the material in ultrashort pulse laser processing (see Sect. 2): often the precise cutting of cross-sections through delicate paint layers of artworks poses a major problem when using mechanical scalpels that are resulting in crumbling, delamination and compression of the paint layer. A Japanese group has demonstrated that this obstacle can be overcome when using a femtosecond laser to create controlled cross-sections of paint layers with minimal damage to the surrounding artwork [129]. The selection of the laser source for restorative applications should be based on the principle that the cleaning success is also accompanied by a minimization of undesirable side effects. In individual cases, this may also mean that longer laser pulses are the better choice.
Finally, the question should be answered, where the ultimate limits are with respect to the laser-induced removal of extremely thin films and how small a laser-induced film perforation can be made? Here, we consider the thin films as "closed", i.e., a continuous layer of a certain material that is covering another material. Note that this situation is distinctively different from the laser-induced desorption of single atoms, molecules, or clusters from a bulk material surface through electronic excitation and sputtering effects [1,130].
The above-mentioned question was addressed by Hartmann and co-workers within the frame of the DFG Priority Programme SPP 1327 "Optically generated sub-100 nm structures for biomedical and technical applications" of the German Research Foundation (DFG) [131]). The authors explored the nonlinear processing and multiphoton ablation of self-assembled monolayers (SAMs) for application as ultrathin resists and in biochemical sensors [132]. Figure 21 (top panel) summarizes their results obtained for the fluence dependence of the fs-laser-induced diameters d Si , d rim , d ripples , and d SAM of different surface topographic features observed at the sample surface for an octadecylsiloxane (ODS) self-assembled monolayer deposited on surface-oxidized silicon wafers after irradiation with single focused fs-laser pulses (800 nm, < 30 fs) [133]. The laser pulses were focused tightly to the sample in substrate geometry according to Fig. 1a using a Schwarzschild microscope objective with a numerical aperture NA = 0.5. Out of the different listed characteristic surface topography diameters, Reprinted from [133] with permission d SAM denotes the ablation threshold of the ODS-SAM film, here. Minimum diameters for the selective laser processing of the ODS layer are ~ 300 nm, i.e., a direct sub-wavelength patterning close to λ/3 is feasible. Corresponding surface height profiles taken by atomic force microscopy (AFM) reveal crater depth of only 0.3 nm below the thickness of the ODS-SAM of typically h SAM ~ 2 nm.
It can be concluded that under optimum conditions, a remarkable precision with sub-wavelength and sub-monolayer film alteration is feasible for thin film processing by ultrashort laser pulses.

Avoiding laser-damage of optical coatings
Another obvious application of thin films in interaction with laser pulses is the development of damage resistant optical coatings. The topic is as old as the experimental realization of the laser itself, since in 1960 Maiman used silver-coated ruby rods as laser resonator (see the introducing historical note of Soileau in Chapter 1 of Ref. [4]). Nowadays, in most cases the coatings consist of multi-layer stacks made of dielectric materials that are tailored for the targeted application, e.g., acting as high-reflective mirrors, for compensation of spectral phase shifts (chirp) for ultrashort pulses, etc. Such optimized coatings are key for the development of modern high-power laser systems on scientific or industrial levels. With meter-scale dimensions [135] they may be of relevance for peta-and exawatt large-scale laser facilities [136] and in future laser-fusion experiments.
Given this high technological relevance of the subject, a whole scientific community is studying the laser-induced damage of optical coatings for more than five decades already, as reflected through the traditional annual Boulder Laser Damage Symposium, first time organized in 1969 in the USA. Since 1984 Round-Robin laser damage test sequences are being organized and measurement methods for quantifying laser damage and its standardization are actively improved, as reflected through the international standards ISO 11254 developed in the 1990s [137,138] and the more recent ISO 22154 dated from 2011 [139,140].
The state-of-the-art knowledge about laser-induced damage in optical materials until the year 2015 was summarized in a comprehensive reference book edited by Ristau [4]. In the following section, we just highlight specific aspects being relevant for the ultrashort pulsed laser damage of thin film based optical coatings. For more details, the reader is referred to that book reference.
In various aspects, the damage of optical coatings upon ultrashort pulsed laser radiation (fs-ps duration) differs from that of short pulsed laser radiation with nanosecond durations or even longer. One reason lies in the high peak intensities that can be easily reached for ultrashort laser pulses at quite moderate pulse energies. Secondly, optical absorption, interference, and scattering effects take place on time scales before significant thermal processes and heat flows can manifest (see Sect. 3). Thus, two demands can be formulated as requirements for ultrashort pulsed laser radiation [4]: i. The optical component has to be capable to handle peak powers up to the petawatt range. This refers to its damage threshold fluence but also to potential nonlinear effects triggered by the large peak powers/intensities. ii. The optical components often require special optical properties, which are usually not necessary for longer pulse durations or continuous wave radiation.
One important requirement is related to the ultrashort pulse durations that, according to the Fourier-transform limit, come with a certain minimum spectral bandwidth. Thus, the optical element must be able to properly manage all spectral components in a controlled manner. Important examples are pulse stretching or compressing elements, such as chirp-mirrors, or coated prisms, or gratings. Moreover, the large spectral bandwidth of fs-laser pulses generates additional difficulties for the design of non-phase-sensitive optical components like mirrors or windows.
According to Jupé et al. [141], the calculation of the damage threshold fluence of a complex multi-layer stack is based on three different and inter-dependent physical entities: (1) The intrinsic damage threshold of the layer material.
This material-specific threshold can be determined from the threshold F th,layer of a single layer and then be normalized by the electric field strength |E| inside the single layer, according to: F th,layer ×|E|= const. (2) The sequence of layers and the electric field strength have to be considered. Note that both entities are influencing each other. (3) The energy flux has to be considered for the calculation of the threshold characteristics in the layered stack and is affected by its refractive index profile. Figure 22 exemplifies the energy flux (solid line, left abscissa) vs the thickness of the multi-layer system along with the refractive index profile of a standard high-reflective (HR) mirror made of alternating quarter-wave (λ/4) layers of SiO 2 (refractive index n ~ 1.5) and TiO 2 (n ~ 2.35) [141]. The laser-induced damage threshold (LIDT) of the entire optical coating is determined by the lowest internal LIDT. Damage is then observed at the interface between the first and second layers on the TiO 2 side, where the energy flux is maximum.
For the improvement of the damage threshold fluence of optical coatings for ultrashort laser pulses, different approaches have been developed. One relies on the use of ternary oxides instead of binary oxides, allowing for a band gap engineering resulting in more flexibility in the coating designs. Another approach is referred to as the refractive index steps down (RISED) concept [142]. Here, ternary oxides (e.g., TiO 1-x Si x O 2 ) and SiO 2 may be used for deposition of stacks of alternating layers, where the refractive index of the TiO 1-x Si x O 2 layers are continuously reduced with depth below the coating surface via the titanium oxide content ranging from 85% at the surface to 5%. This RISED approach allowed to increase the LIDT by more than a factor of two [142]. For more details on the engineering of damage resistant coatings for ultrashort laser pulses, the reader is referred to that reference.

Laser-induced forward transfer (LIFT)
LIFT is a method using the superstrate geometry according to (Fig. 1b). It was first demonstrated by Bohandy et al. in 1986 [143] and represents an additive direct-write technique: single excimer laser pulses at 193 nm wavelength were used to irradiate copper layers on fused silica substrates through the transparent dielectric. The distance between the copper donor layer and the silicon acceptor was below 10 µm. For suitable laser pulse energies, small copper deposits were transferred onto the silicon in vacuum. Two years later, Bohandy et al. demonstrated the process also in air environment for copper and silver layers transferred by 532-nmpulses of a Nd:YAG laser and fused silica substrates as acceptor [144]. Figure 23 shows a scheme of the basic LIFT-setup. The incident laser pulse propagates through the transparent substrate (ideally without losses) and is absorbed at the backside of the donor layer that was previously deposited on the substrate material. When the laser fluence exceeds a material-specific threshold fluence, material is ejected from the donor film and moves to the acceptor substrate, resulting in a "laser-printed" deposit. The aim of the process is a localized deposition of the donor material on an acceptor substrate with spatial "resolution" in the micrometer range. For avoiding disintegration of the transferred deposits, the use of a top-hat spatial beam profile may be beneficial compared to a Gaussian beam, keeping the initial velocity of the ablating material widely constant.
The capabilities of LIFT will be demonstrated in the following by an example. In thin-film photovoltaics, microconcentrator solar cells [176] are a promising approach to improve the efficiency of solar energy conversion. The photovoltaic active area is realized as an array of miniaturized, sub-millimeter-sized solar cells. The incident sunlight is focused on these cells via micro-lenses. The principle of this concept is shown in Fig. 24. The cells consist of solar absorber islands on a metallic back contact (here a thin Mo film that is mechanically stabilized by a glass substrate), embedded in an electrical insulator (green). The solar cells electric front contact layer (gray) covers both the insulator and the absorber islands.
Here, the case is to be considered, that the absorber islands are made of copper-indium-gallium-diselenide (CIGSe) on a molybdenum back contact on a glass substrate. Two different femtosecond laser-based methods to produce such micro-sized precursors of CIGSe solar absorber islands on molybdenum back contacts were investigated.
Method 1 (Laser-induced nucleation of indium): In a "classical" direct-focusing procedure, multiple pulses of a 30-fs laser at 790 nm center wavelength were used for subtle structuring (roughening) of the molybdenum film surface (320-420 nm thickness). A subsequent physical vapor deposition (PVD) of indium was utilized for the siteselective nucleation of indium droplets at the laser-ablated surface spots, serving as precursors for the production of CIGSe absorbers [178,179]. Figure 25 visualizes a regular array of such indium droplets on a molybdenum back contact exclusively grown by the PVD at the laser-induced surface modifications. The inset of the figure shows a scanning electron micrograph (SEM) of a typical indium island along with a stylus profilometric scan across the center of the indium droplet.
Method 2 (LIFT of metallic precursors): Using the same fs-laser system, LIFT was performed with single laser pulses. Donor films consisting of both, indium and copper layers were utilized. Arrays of combined Cu-In-precursors of micro-absorbers were site-controlled deposited on the molybdenum back contact by moving the sample relative to the laser beam [180]. In a later work, even complete Cu-In-Ga layered stacks were successfully LIFTed as donor material [181].
The application potential of the LIFT deposition technique can be seen in Fig. 26a, where a 5 × 5 array of solar micro absorber islands is depicted. The micro absorbers were produced by subjecting the LIFTed Cu-In-Ga-precursor deposits to selenization through rapid thermal processing [182]. Figure 26b shows a top-view energy dispersive X-ray (EDX) analysis of one of these micro absorbers, visualizing the spatial distribution of the involved elements. It is obvious that all absorber components like indium, gallium, copper, and selenium are present in the micro absorber, while the molybdenum signal is reduced through the covering absorber island. Additionally, Fig. 26c presents a crosssectional SEM and an EDX analysis of a micro absorber after deposition of the solar cell ZnO front contact layer. As Fig. 24 Scheme of micro-concentrator solar cell concept [177] Fig. 25 Top-view optical micrograph of an array of indium islands on molybdenum-coated glass fabricated with site-selective shallow fslaser ablation of the molybdenum layer surface (on glass substrate) and subsequent PVD of indium nucleating at the laser processed spots [177]. Insets: Tilted-view SEM of a typical indium droplet (top) and corresponding stylus profilometric scan through the center of the droplet (bottom) [179]. Insets reprinted from [179], Appl. Surf. Sci., Vol. 418, Ringleb, F. et al., Growth and shape of indium islands on molybdenum at micro-roughened spots created by femtosecond laser pulses, 548-553, Copyright (2016), with permission from Elsevier 14 Page 28 of 38 expected, zinc and oxygen are found at the top and molybdenum in the back contact, respectively. The copper signal is homogeneously distributed across the micro absorber, while indium is depleted in proximity to the back contact and gallium toward the front contact. The control of stoichiometry of the chalcopyrite materials is challenging. Here, the advantage of the LIFT technique may lie in the flexibility to tailor the precursor composition by adequate choice of the donor layer(s) and the sequence of transfer. Additionally, the number of processing steps for manufacturing micro solar cells can be reduced. Moreover, LIFT is a single-pulse process while the alternative strategy of In-island growth bases on a multi-pulse laser treatment of the molybdenum back  contact layer. Thus, the laser processing time is in favor of the LIFT-process here.
For both fs-laser processes discussed above (Methods 1 & 2), an efficient and material saving bottom-up fabrication step of precursors for fully functional and working micro-concentrator solar cells was successfully demonstrated [183]. Figure 27 depicts a cross-section of a breakage through a single CuInSe 2 (CISe) micro solar cell produced by the In-nucleation approach of Method 1. The main components are labeled and marked by arrows. The polycrystalline grain structure of the CISe Island sandwiched between the ZnO front and Mo back contact is evident.
The CISe micro absorber cells (Fig. 27) were electrically connected in a parallel manner and conversion efficiencies between 0.15% and 2.9% under 1 sun illumination strength were demonstrated. Under concentrated illumination conditions exceeding this value, significant efficiency enhancements could be achieved [183].

Thin-film depth-profiling for material analyses
The idea to characterize materials by removing and analyzing a microscopic part of a sample through focused laser radiation is almost as old as the laser itself [184]. Since that time several concepts and variants of laser ablation based chemical analysis methods, such as laser-induced breakdown spectroscopy (LIBS) or laser ablation mass spectrometry (LA-MS), have been developed for solids and liquids were steadily improved during the last decades [185][186][187][188][189]. These approaches are based on the examination of material removed by the laser beam (usually the ablation plasma containing electrons, atoms, ions, clusters, or nano-and microparticles), which is subsequently analyzed in detail either by optical or by mass spectrometric techniques. All of them take benefit from advantages provided by laser ablation process, i.e., allowing to probe microscopically small sample volumes in a contactless and rapid manner. Around the turn of the millennium, these techniques have been successfully combined with femtosecond laser sources since their application seemed to be promising and beneficial with respect to the possible sensitivity, background signal levels, and depthresolution (for the reasons briefly discussed in Sect. 2).
The specific advantages of fs-laser ablation for application in chemical analysis of solids can be summarized as follows [186,190,191]: (1) No sample preparation is needed.
(2) Conductive and non-conductive arbitrary shapes can be analyzed directly. (3) Spatial (lateral) resolutions down to a few micrometers or even below can be obtained, controlled via the diameter of the focused laser beam. (4) A rapid analysis of multiple elements is possible. (5) Stoichiometric and very precise lateral and vertical material sampling can be achieved by employing ultrashort laser pulse durations.
However, if the fs-laser ablation is performed in reactive environments, such as air, care must be taken since interpulse chemical reactions (e.g., oxidation) may alter the surface chemistry [17,18]. Top-hat spatial beam profiles are superior to Gaussian shaped ones for avoiding spectroscopic signal intermixing from different sample ablation depths. Moreover, the laser irradiation conditions must be carefully adjusted to realize small ablation rates of the order of 10 nm/ pulse while simultaneously reducing "topographic crosstalk", e.g., potentially arising from self-ordered micro-and nanostructures, such as laser-induced periodic surface structures, micro-grooves or -spikes [19,67,70,192] or surface roughness.

fs-LIBS
In fs-LIBS [193,194], an ablation plasma is induced by fslaser irradiation of a solid sample material either in air or an inert buffer-gas environment at atmospheric or reduced pressures. The time-gated and in most cases spatially integrated optical emission of this ablation plasma is used for the analytical proof of elements in the plasma which is related to the local sample composition.
First studies on fs-LIBS were published 2000 and 2001 by different groups in Germany [195][196][197], the United States of America [198], and Canada [199]. Margetic et al. [195] performed a comparative study of fs-and ns-LIBS of brass samples in Argon inert gas (775 nm, 170 fs/6 ns, max. 10 Hz, 1-1000 mbar). They found significant influences of the buffer-gas pressure and differences in the temporal evolution of the atomic and ionic line intensities for fs-and ns-pulses. The more reproducible fs-laser ablation leads to a reduced ablation threshold fluence, an improved ablation precision and higher sensitivity. In a later study on the same material (brass, 775 nm, 170 fs, max. 10 Hz, 40 mbar Ar) [196], more attention was paid to the nonlinearity of calibration curves as they can be found in ablation measurements in alloys.
Additional work has been performed in this group with respect to the depth-resolved analysis of multi-layer samples. In a comparative study between fs-LIBS (Ar buffer-gas, 1-1000 mbar) and fs-LA-ToF-MS (vacuum < 10 -6 mbar) samples consisting of 600 nm thick Cu-Ag sandwich layers on silicon could be resolved by the fs-LIBS technique (775 nm, 170 fs, max. 10 Hz) [197]. The spatially Gaussian shaped beam profile limited the contrast of the detected depth profiles. Angel et al. [198] reported about fs-, ps-and ns-LIBS measurements on copper in air at atmospheric pressure (810 nm, 140 fs, 1-1000 Hz). Compared to nspulse excitation (1064 nm, 7 ns, 5 Hz), the ultrashort laser pulse generated ablation plasmas showed a much lower background signal and a more rapid decay of atomic line emissions. The authors state that due to the relatively low background, a temporally non-gated optical signal detection is very effective. High femtosecond laser pulse repetition rates (1000 Hz) were shown to increase the LIBS signal. Le Drogoff et al. [199] performed comparative studies on time-resolved space-integrated fs-(800 nm, 100 fs, 10 Hz) and ns-LIBS (1064 nm, 8 ns) of aluminum alloys in air at atmospheric pressure as well. Besides the observation of the faster decay of continuum and spectral lines and a shorter plasma lifetime in the fs-case, the analytical performance was evaluated. The fs-LIBS detection limits were found to be element dependent. They vary between 2 and 80 parts per million (ppm).
New dual-pulse LIBS geometries combining the material excitation with combinations of fs-and ns-laser pulses (for pre-ablation spark generation or plasma re-heating) are reported by Scaffidi et al. [200]. Large signal enhancements up to 80 were observed for copper and aluminum bulk samples for ablation with 100-fs (800 nm) laser pulses and plasma re-heating with subsequent ns-laser pulses (1064 nm, 7 ns).
Further aspects of fs-LIBS were reviewed in detail by Labutin et al. [194], including reports on depth-resolutions of the order of a few tens of nanometers when employing top-hat spatial beam profiles.

fs-LA-MS
Wechsung et al. [201] presented in 1978 the method of laser microprobe mass analysis (LAMMA) that is combining the laser ablation process with mass spectroscopy (LA-MS). Nowadays, femtosecond laser alation mass spectrometry (fs-LA-MS) has been developed into three main variants, i.e., the femtosecond laser ablation inductively coupled plasma mass spectrometry (fs-LA-ICP-MS), the femtosecond laser ablation time-of-flight mass spectrometry (fs-LA-ToF-MS), and as femtosecond matrix-assisted laser desorption/ ionization time-of-flight mass spectrometry (fs-MALDI-ToF-MS). Since the latter is-by its principle-not made for analyzing thin films, the focus in this review article will lie solely on the first two methods, as briefly discussed in the following.
In fs-LA-ToF-MS [186,188] the ionic species of the laser-induced plasma are analyzed with respect to their arrival time to a mass sensitive detector. Figure 28 provides a scheme of a fs-LA-ToF-MS presented in Cui et al. [202]. This technique is advantageous compared to the LIBS method with respect to the sensitivity and the simultaneous detection capability of ions of all elements removed from the surface by a single laser shot [188]. In a first study of Margetic et al. [197], a depth-resolved fs-LA-ToF-MS analysis (pressure < 10 -6 mbar, laser excitation: 775 nm, 170-200 fs, max. 10 Hz) was performed on a TiN-TiAlN layered system deposited on an iron substrate. The layered structure with a thickness of 280 nm of each layer could be clearly resolved.
In a later study of this group, the fs-LA-ToF-MS was successfully applied to the depth-resolved analysis of multi-layer samples (pressure < 10 -6 mbar, laser excitation: 775 nm, 170-200 fs, max. 10 Hz) [190]. An ion-implanted silicon wafer (150 nm shallow Co-implanted layer, 10 17 ions/ cm 3 ) and a thin film metal standard (NIST 2135c consisting of alternating Cr and Ni layers, having a thickness of 56 and 57 nm, respectively) have been analyzed. Lateral resolutions of 30 µm -40 µm and a depth resolution of ~ 10 nm (limited by the pulse-to-pulse energy stability of ±5%) combined with a mass-resolution of m/Δm ~ 300 were demonstrated in that work. Meanwhile, better lateral resolution, significantly increased mass resolutions exceeding 10,000, and detection limits of a few tens of parts per billion (ppb) or single-pulse detection limits of 10 -15 g (fg) were demonstrated in selected applications (see Tulej et al. [188] and references therein).
Later, systematic depth-profiling analyses were performed on semiconductor wafers and metal foils [202,Fig. 28 Illustration of a fs-LA-ToF-MS system [202] consisting of the ion optics, sample stage, laser focusing system, and a vacuum chamber including the differential vacuum wall. . Although no attempt was made to calibrate the results against the number of applied laser pulses per spot, an estimate of the depth-resolution ranges from a few nanometers to a few tens of nanometers in the laser fluence window between 0.6 and 1 J/cm 2 [188].
In fs-LA-ICP-MS [204,205] the femtosecond laser generated plasma is injected into a stream of inert gas (e.g., Ar) which transports the ablated species to an inductively coupled mass spectrometer (ICP-MS). When using a fslaser source, nanoparticles with a size distribution centered around 100 to 200 nm are generated that can be easily decomposed in the ICP-torch. However, since the technique involves this crucial aerosol transport, it requires a very careful adjustment of operation parameters and data evaluation. Some limitations during the aerosol transport, and the partial digestion in the plasma of the Ar torch [206], make this technique at first view more promising for the investigation of bulk samples than for thin film analysis. Nevertheless, high depth-resolutions of several tens of nanometers were demonstrated for metal films studied by fs-LA-ICP-MS with optimized ablation conditions (laser excitation: 400 nm, 150 fs, 1 Hz, homogenized beam) [207], making this method also suitable for depth-profiling applications of thin films [188,191]. Moreover, a remarkable sensitivity is reached that even allows to identify isotopes and monitor their ratios and, thus, can be used in archaeological and geological investigations [205,208].
In conclusion, both technologies, fs-LIBS and fs-LA-MS, have reached a mature state and can be routinely used for thin-film depth-profiling for material analyses in specialized laboratories.

Summary and outlook
An approach to classify film damage mechanisms by ultrashort laser pulses at the lowest fluence leading to a permanent material alteration regarding possible damage mechanisms is presented in Table 1, ordering for the different film material classes (metals, semiconductors, dielectrics) the possible damage mechanisms regarding to the film thickness (small, medium, large), the carrier substrate materials (low or high reflective at the laser wavelength), and the laser processing geometry (substrate vs. superstrate). The table may aid and guide the reader to identify potentially relevant laser damage scenarios (SD: surface damage, ID: interface damage, BD: bulk damage) for their specific experimental conditions.
In conclusion, ultrashort laser pulses have proven to be a valuable tool for structuring thin and thick films when contactless and precise material removal or surface modification is desired. With very few exceptions, their use is superior to that of laser pulses with longer pulse duration (of the same wavelength and spatial beam profile) by "transcribing" the reduced heat-affected zone or nonlinear absorption mechanisms into an improved sub-micrometer machining precision. Just in very few cases, where detrimental highintensity-related limiting effects manifest, e.g., through  localized electromagnetic field-enhancement in the bulk of the films or at its surface topography, potential care must be taken. Given these unique advantages and the already broad availability of reliable high-repetition rate sources, ultrashort laser pulses will further conquer industrial thin film processing for tailored scientific and real-life applications.
Funding Open Access funding enabled and organized by Projekt DEAL.

Conflict of interest
The authors declare that there are no conflicts of interest related to this article.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.