Research progress and prospects of laser coating technology

The rapid advancement of high-power laser systems has enabled their extensive application in numerous scientific and industrial domains, such as the detection and elimination of space debris^1,2, laser shot peening³, laser materials processing⁴, and laser nuclear fusion⁵. Optical coatings that have the versatility to modulate the intensity, spectrum, polarisation, dispersion, and propagation direction of a laser serve as key components in high-power laser systems. The output laser quality and energy depends on their performance, which limits the development of high-power laser technology^6,7.

The theory of optical coatings can be traced back to the 17^th century, when Newton’s ring phenomenon and the interference of light were discovered. The amplitude, energy, and phase relationships at the interface were deduced based on a combination of electromagnetic and wave theories. Consequently, a direct theoretical basis for optical coatings was established. In the early 20^th century, in a study by Alexander Smakula⁸, multilayers were fabricated to satisfy the requirements for antireflective (AR) coatings, which marks a significant advancement in optics. Modern optical coatings have expanded from one-dimensional (1D) layers to multidimensional multilayers, gratings, metasurfaces, metamaterials, and their combinations. Their capabilities to adjust the beam shape, power, polarisation, and bandwidth guarantee the output laser performance and promote the development of modern technologies and disciplines.

Although the basis for optical coatings can still accurately describe the propagation behaviour of the light field, designing and fabricating the required optical coatings remains challenging because of insufficient control capability of the light-field and laser-induced damage. In the past decade, analytical and computational strategies such as finite element analysis, machine learning, and artificial intelligence have been introduced to support structural design and multiphysical field simulation. The emergence of rigorous coupled wave analysis (RCWA) and finite-difference time-domain (FDTD) method, combined with near- and far-field characterisations, has helped to demonstrate the mechanism of laser-induced damage. The development of fabrication techniques that can adjust the material bandgap and control defects can further improve the performance of high-power laser coatings.

This paper provides the basic concepts of light-field modulation, laser-induced damage mechanism of high-power laser coatings, and research progress towards the performance requirements. Finally, the future challenges and opportunities for high-power laser coatings are discussed.

In optics, a multilayer is a structure composed of alternating thin films of materials with different refractive indices, typically arranged in a periodic or quasiperiodic manner. The basic intention behind multilayer coatings is to exploit the interference effects. Constructive or destructive interference occurs when light interacts with the boundaries between layers of different refractive indices, which depends on the optical thickness of the layers and the wavelength of the incident light. This enables the multilayers to be tailored for specific optical functionalities, such as enhancing the reflection at certain wavelengths or reducing it at other wavelengths.

The overall reflectance and transmittance in multilayers are calculated by considering the cumulative effect of each interface and the phase shifts owing to the layer thicknesses. Historically, multilayer calculations have employed the recursion method based on the Fresnel coefficient formula to determine the reflectance and transmittance across surfaces, progressing iteratively to an equivalent for all the layers. The advent of the admittance matrix method has simplified this calculation process and has been predominantly used⁹. The characteristic matrix $ {M}_{i} $ for a layer with thickness $ {d}_{i} $, refractive index $ {n}_{i} $, incident angle $ {\theta }_{i} $, and wavelength $ \lambda $ is,

where, $ {\mathrm{\delta }}_{i}={(2{\text π}{n}_{i}{d}_{i}cos{\theta }_{i})}/{{\lambda }} $ is the phase thickness of the layer and $ {\eta }_{i}=\left\{\begin{array}{l} {n}_{i}/cos{\theta }_{i},\quad p-polari{\textit z}ed\\ {n}_{i}\mathrm{cos}{\mathrm{\delta }}_{i},\quad s-polari{\textit z}ed\end{array}\right. $ is the admittance. The total system matrix for $ N $ layers is the product of the individual layer matrices.

The reflectance $ R $ and transmittance $ T $ of the multilayer system can be extracted from the total matrix by solving for the electric and magnetic fields at the interfaces as follows:

The key applications of multilayers in high-power laser systems include AR and highly reflective (HR) coatings, edge/bandpass filters, beam splitters, polarisers, and dispersive components, as shown in Fig. 1a. These requirements emphasise precision, control, and stability in terms of refractive-index contrast, spectral or angular response, and dispersion management. In high-power laser systems, these characteristics are essential for optimising the performance and ensuring the durability of the multilayer components.

Fig. 1  Optical manipulation of optical coatings. a Spectral performance of typical multilayer mirrors. b Porous¹⁰ and textured nanostructures^11,12. c Nanostructures evolving from pillars to circular cones for various applications^13–15. d Multilayer gratings for broadband efficiency¹⁶. e Quasi-3D microstructures by integrating metasurfaces with multilayers¹⁷.

Multilayer coatings are characterised by stratified perpendicular structures that are uniform across the surface. An alternative approach involves lateral modifications known as microstructures. These structures enable precise control of light-matter interactions at the subwavelength scale, which facilitates arbitrary manipulation of the wavefront and polarisation distribution across a broad electromagnetic spectrum and offers opportunities to enhance the intrinsic properties of the coating materials. Based on the design principles, microstructures can be categorised into four types: effective medium nanostructures, gratings, metamaterials, and metasurfaces; each has distinct optical capabilities.

Porous or textured effective medium nanostructures are designed by combining substrate materials with air at the subwavelength scale using modulated fill factors to control the effective refractive index, as shown in Fig. 1b^10–12. Integrating these nanostructures with multilayers, either as alternatives or in combination, has introduced new dimensions to the development of optical coatings, particularly in the visible and near-infrared (IR) regions. Tolenis et al.¹⁰ demonstrated the use of glancing-angle deposition (GAD) to fabricate all-glass porous-nonporous high-low (HL) refractive index stacks. Varying the angle of incidence of the incoming vapour particles modulated the porosity within the layers, thereby controlling the effective refractive index with the comparative optical performance of traditional HL stacks for HR coatings. Building on the effective medium theory, periodic or random textured surfaces were introduced as alternatives to conventional multilayers. Researchers have discovered a strong correlation between the bandwidth and the unit shapes of textured surfaces, which indicates new optimisation directions. For example, Yu et al.¹² fabricated cone-shaped periodic silicon structures using trilayer resist nanoimprint lithography and achieved reflectivity below 3% in the visible spectrum. Nishijima et al.¹⁸ reported cascading composite nanostructures prepared on Si using a combination of wet and dry etching processes and realized a refractivity of less than 1% for visible wavelength light at normal incidence. Yoo et al.¹³ used solid-state diffusional dewetting for the self-assembly of gold nanoparticle (NP) ensembles, followed by reactive ion beam etching (RIBE), which resulted in random nanostructures that evolved from pillars during shorter etching times to circular cones with longer etching times (Fig. 1c). As the sidewall slopes reached an aspect ratio of approximately 12, the reflectance dropped below 0.5% across the 350–2000 nm range¹⁹. These effective medium nanostructures can be easily fabricated, and their refractive indices can be tuned by adjusting the etching time, which demonstrates the potential for large-aperture optics and structural freedom beyond periodic designs. Further research combining NP masking with angled etching has yielded nanostructures with angled features that induce optical anisotropy and birefringence, which makes them suitable for applications such as waveplates¹⁵.

Gratings are microstructures with one lateral freedom that are widely used in chirped-pulse amplification (CPA) systems²⁰. They stretch the input pulse before amplification to reduce the peak intensity and prevent damage to the gain medium. Pulse compression gratings (PCGs) facilitate high-power, high-brightness spectral beam combining while maintaining a diffraction-limited performance²¹. Combining multilayers with gratings, as shown in Fig. 1d, can offer phase modulation and improve the efficiency. However, achieving near-100% diffraction efficiency (DE) across a wide spectral bandwidth remains a challenge, particularly for short pulses exceeding 100 nm. Furthermore, the polarisation dependence of the gratings limits their performance in high-power laser beam combining. Single-layer grating structures can only modulate a limited set of parameters, such as the refractive index, duty cycle, and grating height, which makes it difficult to optimise the DE for both transverse electric (TE) and transverse magnetic (TM) polarized light. Although double-layer gratings provide increased design flexibility, they still struggle to meet the broadband high-efficiency requirements. As high-power laser systems continue to evolve, there is a growing demand for the independent modulation of broadband efficiency and polarisation to meet performance expectations.

The next stage of microstructures deals with metamaterials with adjustable three-dimensional (3D) structural parameters. These materials can achieve arbitrary permittivities and permeabilities, thereby prompting novel optical phenomena such as negative refraction and electromagnetic stealth²². Although 3D metamaterials have been applied in the microwave domain, their development in the optical frequency range is limited by challenges in fabrication at subwavelength scales.

Therefore, two-dimensional (2D) metasurfaces have been proposed to address this issue. Using subwavelength antennas to control light, these metasurfaces integrate both strong wavefront modulation capability at the subwavelength scale and feasibility of fabrication. Owing to the advancements in micro-nano manufacturing technology in the recent years, numerous metasurface devices made from various materials have been reported, including meta-lenses²³, invisible cloaks²⁴, absorbers²⁵, vortex beam generators²⁶, and holography systems²⁷. A metasurface can achieve complete control over scalar and vectorial optical fields^28,29, which indicates that a single metasurface platform can replace multiple conventional optical components. Owing to their powerful capabilities and multiparameter optimisation freedom, metasurfaces have gained increasing attention because of their potential applications in high-power laser systems.

Metasurfaces offer compactness that is far superior to traditional optical components. However, achieving a high efficiency remains a significant challenge. Researchers are investigating the physical mechanisms that limit efficiency, focusing on impedance distribution and surface wave interactions. Although metasurfaces have demonstrated high reflection, the efficiency demands of high-power laser systems is yet to be satisfied. A straightforward approach involves integrating metasurfaces with more mature technologies and functional devices. For instance, integrating metasurfaces with multiple layers can potentially balance phase control and reflectivity (Fig. 1e)¹⁷. Continued research on the interaction between lasers and quasi-3D microstructures is crucial for unlocking their full potential.

The complex interactions between the laser and the material in optical coatings result in laser-induced damage. When optical coatings are exposed to high-power laser radiation, the material changes manifest as recoverable modifications, which affects the optical performance before visible damage occurs. Once permanent damage is initiated, the damage site expands with repeated laser exposure, degrades the optical performance, and eventually causes device failure. In high-power laser systems, laser-induced damage to optical coatings is considered the primary limitation for increasing the output power³⁰ and has been a major focus of research for several decades³¹.

The damage mechanism in optical coatings is a complex process influenced by various factors, such as laser parameters (wavelength, pulse width, frequency, and energy) and film properties (material type, defects, etc.)^32–34. Particularly, the laser pulse width plays a crucial role in determining the interaction between the laser and the material, as shown in Fig. 2a^35–37.

Fig. 2  Laser-induced damage mechanisms and morphological behavior. a The interaction between electrons and photons across fs to ns timescales³⁷. b Typical damage morphologies on multilayers induced by fs, ps, and ns laser pulses³⁸. c Typical damage morphologies on multilayers dominated by thermal melting³⁹, mechanical damage⁴⁰, and plasma scalding⁴¹ under nanosecond laser pulses. d Schematic of the three principal damage precursors: electric field enhancement and visible and non-visible defects⁴². e Typical damage morphologies on multilayers originated from nodule defects⁴³ and nano-absorption centers⁴⁴. f Typical damage morphologies on gratings originated from electric field enhancement, nodule defects, and nano-absorption centers⁴⁵. g Growth sequences on MDGs owing to picosecond pulses at 1053 nm⁴⁶.

Energy is deposited rapidly in short-pulse lasers (sub-picoseconds to femtoseconds), which heats the material instantaneously. Because the damage morphology (Fig. 2b)³⁸ is clean with minimal residue from melting, mechanical damage, or material splashing, field-induced damage is termed “cold” damage⁴⁷. This process is driven by nonlinear absorption mechanisms, including multiphoton absorption and avalanche and tunnel ionisations, which result in field-induced damage because the strength and efficiency of nonlinear processes are directly related to the electric field intensity (EFI) within the material. As the pulse separation time is comparable to the lattice heat relaxation time of the coating materials, the incubation effect is strong for short-pulse lasers, which eventually results in the accumulation of defect sites^48,49. Multiphoton ionisation exhibits a power-law dependence on the laser intensity⁵⁰ and the number of laser shots⁴⁸, and is influenced by the pulse separation time, which, in effect, makes the absorption process deterministic. Nonlinear refractive index and absorption changes induced by femtosecond lasers have been identified using the two-photon absorption process^51,52 and have emerged as limiting factors, even at light intensities below the laser-induced damage threshold (LIDT)⁵³.

Conversely, the damage induced by long-pulse lasers (picoseconds to nanoseconds) is complex. The damage process begins primarily with defect-induced linear heat absorption, causing the material to melt³⁹, vaporize, and ionise, which eventually forms a plasma. This plasma also absorbs laser energy, which results in a sharp increase in temperature and rapid expansion, thereby producing shock waves that cause impacts and mechanical damage⁴⁰. This could result in material cracking, grain expansion, stripping, and ejection, thereby inducing complex damage morphologies (Fig. 2c)³⁸. Absorbing defects can generate dense plasma, thereby inducing thermal-plasma coupling damage⁴¹. Even when the energy is below the vaporisation threshold, lasers can induce structural or chemical alterations, causing the accumulation of defect sites over multiple pulses, eventually causing damage⁵⁴. Consequently, the LIDT for long-pulse lasers is inversely related to the pulse duration. When the electron and lattice temperatures reach equilibrium and can be described by the same temperature, the relationship between the ablation threshold $ {F}_{th} $ and pulse duration $ {\tau }_{p} $ can be expressed as $ {F}_{th}\sim \sqrt{{\tau }_{p}} $⁵⁵. However, predicting the LIDT for long-pulse lasers is more challenging than that for short-pulse lasers because of the complex interplay between thermal and mechanical damage processes.

For continuous laser, thermal accumulation prevails^56,57, and the incubation effect is not considered: Ablation and mechanical damage are the typical morphological characteristics of continuous laser damage, because absorption by the layer material and localized defects alter the temperature field of the film, which induces thermal stress damage; when the temperature generated by the thermal effect reaches the melting point of the film, the film exhibits thermal melting damage. The LIDT of a continuous laser can be predicted or calibrated by applying a thermal diffusion equation.

The damage precursors for high-power lasers can be broadly categorised into three principal types: intrinsic electric field enhancement⁴⁵, visible defects, and non-visible defects⁵⁸, as shown in Fig. 2d. Studies on multilayers reveal that the primary contributors to interface damage are the standing wave electric fields that are formed within the coatings with maximum values typically occurring at interfaces close to air. Li et al.⁵⁹ compared the cross-sections of the damage sites and LIDTs with the corresponding EFI distributions for HfO₂/SiO₂ HR coatings and determined that the damage initiated near the peak of the EFI, with a negative linear relationship between the LIDT and maximum EFI. Research on gratings has highlighted the significance of electric fields. Néauport et al.⁶⁰ fabricated multiple samples with different electric-field distributions inside multilayer dielectric gratings (MDG). Detailed optical and atomic force microscopy (AFM) inspections of the damage sites revealed that the damage occurred when the electric field was the highest. Further research⁶¹ confirmed that the LIDT of the MDGs was determined by the value of $ {E}^{2} $. Although experimental LIDT studies are limited for metasurfaces, it has been established that the electric field enhancement in these structures is theoretically larger. To extract the influence of EFI from the overall LIDT, Sozet et al.⁶² proposed using the intrinsic LIDT, $ {LIDT}_{\rm{int}}={EFI}_{\max}\times {LIDT}_{\rm{eff}} $, to compare different samples.

In addition to the intrinsic electric field enhancement, optical coatings contain defects ranging from submicron to submillimetre scales, including substrate imperfections, rogue particles within coatings, nanoscale defects, such as metal and dielectric nanoclusters that form and aggregate during deposition or result from the destruction-redeposition of chemical bonds during dry-etching processes, and post-fabrication surface contaminants. These defects are often categorised as visible or non-visible, depending on the difficulty of detection. With the continuous development of defect-free polishing techniques, visible substrate scratches, pits, and digs are no longer a problem, and are only considered for large-aperture optics. However, nodular defects originating from particulates or seeds remain critical. Nodules developed into an inverted conical shape with a domed top protruding above the surface of the coating (Fig. 2d-f)^54,56. Both theoretical and experimental studies have confirmed the significant role of nodule defects in enhancing the electric field inside the nodules, amplifying local energy absorption, and increasing the likelihood of damage. Dijon et al.⁶³ demonstrated that shallow nodules result in flat pits, whereas deeper nodules produce conical pits owing to splintering. This behaviour is explained by the thermal damage model, in which energy absorption at different points on the curved nodule surface causes localised overheating and eventual damage. Chris et al.^64,65 employed 3D FDTD simulations to illustrate the link between nodular damage and electric field enhancement inside a nodule, which varied according to the reflection band and light polarisation. Zhang et al.⁶⁶ developed a dynamic model coupling photoionization, impact ionisation, electron collisions, heating, and refractive-index modifications to analyse the effects of deep and shallow nodules on MDGs. They found that deep nodules were more likely to cause internal coating damage, and overlapping nodule-induced scattering and a defect-free MDG electric field were the primary causes of damage distribution. Interestingly, while the electric field model predicts damage initiation at the grating pillars, the experimental results suggest that it begins with nodule ejection, thereby highlighting the importance of considering the mechanical weaknesses in nodules. Poulingue et al.⁶⁷ employed purely mechanical methods and demonstrated that nodules of 4 μm or larger could easily cause mechanical damage to the film. Ma et al.⁶⁸ performed studies using electron-beam evaporation (EBE) and ion-assisted deposition (IAD) combined with annealing to prepare Ta₂O₅/SiO₂ film artificial nodules with different boundary continuities. For nodules of the same size, the damage threshold for the IAD-processed films was four times higher than that of the EBE-processed films, which indicates better durability with strong boundary continuity. However, damage growth threshold exhibited the opposite behaviour: for nodules in EBE reflector films, the average damage growth threshold was measured at 11.6 J/cm², which is approximately 2.9 times higher than the initial damage threshold. Conversely, for the IAD reflector films, the average damage growth threshold was 7 J/cm², which is 2.6 times lower than the initial damage threshold. This indicates that, although the nodule boundary continuity improves the initial damage resistance, the films are more susceptible to damage growth over time.

Although Non-visible defects are smaller and have minimal impact on EFI and mechanical strength compared to nodules, they introduce intra-bandgap states that form nano-absorption centres. These nano-absorption centres can trigger plasma formation, resulting in thermal explosions and ultimately causing damage (Fig. 2d-f)^{53,55,56,69,70}. However, detecting nonvisible defects is challenging because of their low density and sub-micrometre size, which makes meaningful comparisons between experimental results and theoretical models difficult. To address this issue, artificial gold NPs are often used to simulate nano-absorption centres because they have well-documented optical and thermal characteristics, narrow size distributions, and chemical stability, and are available as colloids or powders. Papernov et al.⁷¹ reported an excellent correspondence between nano-absorbed gold particles and crater location and indicated that absorption is solely responsible for damage initiation. Even gold particles of 1.9 nm in diameter embedded at a depth of 60 nm significantly reduced the threshold by one-third. Comparisons of energy absorbed by gold particles, $ {E}_{\text{abs}} $ (based on Mie theory), and the energy required for crater formation, $ {E}_{\text{cr}} $ (measured through AFM), confirm that $ {E}_{\text{abs}}\le {E}_{\text{cr}} $, implying absorption delocalisation occurs beyond the particle volume, which provides direct evidence for plasma-ball formation, validating the thermal explosion mechanism.

Optical coatings with defects significantly affect the laser resistance. Nodules influence the electric fields, mechanical strength, and LIDT, whereas nano-absorption centres primarily affect the LIDT without altering other coating properties. Owing to their focusing effect, nodules are highly sensitive to electric fields and polarisation in both the femtosecond⁶³ and nanosecond⁵⁴ regimes, thereby rendering HR coatings and polarisers particularly critical. Conversely, nano-absorption centres play a more significant role in damage initiation in AR coatings and coatings designed for short wavelengths, such as 532 nm, 355 nm, and the near-ultraviolet region^72,73. Pre-treating gold-particle-embedded samples with subthreshold fluences can disperse gold into the surrounding matrix, providing insights into the possible mechanism of UV laser conditioning⁷⁴. Predictably, laser coatings are particularly vulnerable to damage when EFI is enhanced near defects such as multilayer interfaces or microstructure ridges, emphasising the importance of localised EFI control.

The lifetime of optical coatings applied in high-power laser systems is limited by two different phenomena: damage initiation for a given fluence and the damage growth tendency under iterative shots. Different types of initial damage in the multilayers and microstructures result in varying damage-growth characteristics under subsequent laser irradiation. Damage growth in multilayer films is influenced by the film stack design, coating process, and irradiation conditions. In the early stages, Genin et al.⁷⁵ systematically studied the growth of typical damage morphologies and highlighted the impact of nodule defects. Gallais et al.^76,77 photographed the damage growth process of nodule defects under femtosecond and sub-picosecond pulse-laser irradiation and suggested that during the linear phase, the damage area evolved as $ {S}_{n}=\alpha n+{S}_{0} $, where, $ {S}_{0} $ and $ {S}_{n} $ were the initial damage area and the damage area after $ n $ laser pulses, respectively, $ \alpha $ is the linear growth coefficient, which increased with the fluence set during the growth sequence.

For multilayer gratings, the damage often involves the removal of several adjacent grating pillar sections without affecting the underlying multilayer structure. This type of damage was more consistent with the effects of stress waves⁷⁸ and eruptive pressure¹⁶. Hao et al.⁵⁷ investigated the damage growth in MDGs under picosecond pulses at 1053 nm. Once initiated, the damaged area grows linearly with the number of laser shots and saturates after a sufficient number of shots owing to the Gaussian spot distribution, which is similar to the damage growth characteristics observed in multilayers (Fig. 2g). Furthermore, the damage growth rate along the laser propagation direction was greater than that in the reverse direction, indicating that this asymmetry in the growth behaviour was caused by the asymmetrical intensity modulation caused by the damage sites.

Laser-induced damage in optical films is a multifaceted process that is affected by material properties, laser parameters, defects, and impurities. Understanding these interactions is the key to improving the laser resistance of optical coatings and ensuring the long-term durability of optical components, particularly in high-power laser systems. Although the electric field distribution primarily governs the initial damage and damage growth, a full explanation should involve electrical, thermal, and stress fields and the evolution of the morphology and phase under non-equilibrium conditions.

The widespread use of high-power lasers has increased the demand for optical coatings with diverse spectral performances, which ranges from standard HL stacks to more complex, nonperiodic multilayers. Recent advancements in effective medium nanostructures and micro-nano fabrication techniques have broadened the light-modulation capabilities of optical coatings, offering opportunities to enhance system compactness, beam quality, and laser output power. This section explores the latest progress in improving the optical performance of high-power laser multilayers and multilayer gratings, and how metasurfaces unlock new functionalities beyond traditional laser coating applications.

In high-power laser systems, coatings play a crucial role in the spectra, polarisation, and phase control achieved through components such as multilayer filters, polarisers, polarising beam splitters, dispersive mirrors, and PCGs. These functionalities often intersect rather than being independently achieved or required. Thin films and gratings are specifically tailored to match the pulse duration and wavelength of the laser, which influence the crucial material properties, such as absorption and LIDT. Designing multilayers involves optimising the materials, thicknesses, and stacking sequences from standard HL stacks to complex nonperiodic structures to satisfy the demands of high-power applications.

Spectral control is the most widely applied high-power laser coating technique and is mostly controlled by filters. A rapid transition between high transmission and high reflection over a narrow spectral range is often required in edge or bandpass filters, where a sharp cut off between the passed and blocked wavelengths is required (Fig. 3a)⁷⁹. Pervak et al.⁸¹ theoretically and experimentally demonstrated a band filter with two reflection and broadband transmission ranges obtained using standard two-material technology. The fabricated filter had better transmission and reflectivity characteristics than those achieved using rugate technology. Using two-material multilayers is the simplest method for designing notch filters. However, two-material designs often have very thin layers that are difficult to control precisely and in turn to control the notch filters. Tikhonravov et al.⁸² utilized constrained optimisation techniques to improve the design of quasi-rugate coatings that feature smooth and continuously varying refractive index profiles. The constrained optimisation process ensures that extended high-transmission zones with suppressed sidelobe reflectance ripples are achieved, while preserving the continuous gradient of the refractive index. Zhang et al.⁷⁹ proposed a design for notch filters using multilayer structures with high transmittance (no back-side reflections included) in the wavelength ranges of 400-500 nm and 550-700 nm and high reflectivity in the spectral region from 500-550 nm, optimised through constrained algorithms. Some layers near the substrate were further optimised to suppress the transmittance ripples in the passband. Moreover, all the layers of the obtained design had optical thicknesses close to the half-wave of the central rejection wavelength, and therefore, were suitable for accurate monitoring during deposition.

Fig. 3  Typical applications of optical coatings in high-power laser systems. a A notch filter using multilayer structures with steep spectral cut off at 100 and 350 nm⁷⁹. Reprinted with permission from⁷⁹ © Optical Society of America. b A long wavelength infrared narrow-band reflection filter⁸⁰. c A highly-dispersive mirror in the wavelength range of 775–825 nm. d A MDG with broadband diffraction.

Achieving nearly complete transmission or reflection at specific wavelengths with minimal leakage is essential to ensure high monochromaticity (Fig. 3b)⁸⁰. Although this is relatively straightforward for transmission filters that use strategies such as Fabry-Pérot (F-P) cavities, it remains a challenge for reflective designs. Recently, narrowband filters operating in the long-wavelength infrared region have leveraged subwavelength structures based on guided mode resonance (GMR) and surface plasmon resonance (SPR)⁸², thereby reducing the thickness and material cost. A key approach is to use metallic film materials combined with F-P filter designs, which provide precise control over the reflectivity in the visible spectrum⁸³. However, when using GMRs and SPRs with metals, typically, multiple transmission or reflection peaks exist over a wide spectral range. Furthermore, because of the strong absorption of light by metals, filters utilising SPR are usually accompanied by a wide spectral linewidth, which is detrimental to narrowband filtering. Recently, SPR, such as the Fano resonance⁸⁴ using dielectric materials, have been used to realise narrowband reflection or transmission. In the dielectric mirrors used in laser resonators, high reflectivity over a narrow band ensures that the laser operates efficiently at a single wavelength without losses to neighbouring wavelengths. Recent work on multilayer dielectric films (MDFs) has focused on optimising the refractive index contrasts to maximise the reflectivity over a narrow wavelength range. These approaches use advanced modelling techniques to fine-tune the optical properties for applications such as lasers, spectroscopy, and other high-precision optical systems.

Polarisers are optical components designed to selectively transmit light of a specific polarisation, while reflecting light with orthogonal polarisation. The extinction ratio and bandwidth are critical metrics for evaluating polarisation selectivity.

Birefringent crystals such as calcite, quartz, and lithium niobate can generate polarised light with high extinction ratios and operate across a broad wavelength range, from UV to IR, depending on the material used. However, their performance is limited by angle sensitivity, intrinsic material properties affecting the LIDT, and challenges associated with large-aperture fabrication owing to their bulky size.

Thin-film polarisers provide excellent LIDT and scalability and are particularly suitable for high-power laser systems. These polarisers function through interference and reflection principles, which facilitate applications such as beam splitting, pulse picking, frequency doubling, and optical isolation. Stolz⁸⁵ explored various initial designs for Brewster's angle polarisers, focusing on the bandwidth, layer thickness error sensitivity, and extinction ratio. One design approach typically optimises long-wave-pass edge filters by leveraging the polarisation-splitting properties at non-normal incidence angles, while maintaining mirror symmetry relative to the central layer. Alternately, the F-P bandpass filter design achieves higher extinction ratios with comparable design thicknesses. Although the F-P design has a narrower angular bandwidth than typical long-wave pass designs, it is less sensitive to random thickness errors, which makes it a more robust alternative. Zhang et al.⁵⁰ proposed a combination of long-wave pass and short-wave pass stacks to create broadband thin-film polarisers and achieved an s-polarised transmittance exceeding 98% and a p-polarised transmittance below 1% at 1064 nm, with a bandwidth of 50 nm. This performance was twice that of the traditional thin-film polarisers. Zhu et al.⁸⁶ demonstrated HfO₂/SiO₂ Brewster’s polarisers at 1064 nm using the LWP and F-P bandpass approaches and obtained a theoretical extinction ratio of over 30 dB. Enhancing laser-induced damage resistance is crucial for polarisers in high-power laser systems, as detailed in the following section.

Gratings offer another avenue for polarisation selection. Au-on-silica grating structures can span the visible and near-infrared spectra and achieve extinction ratios exceeding 100 dB when the grating pillar width matches the period⁸⁷. A four-layer Au subwavelength metallic grating separated by silica demonstrated both high transmission (40% antireflection effect) and high extinction ratios (~27 dB) in the wavelength range 1100-1600 nm, reaching an impressive 70 dB at wavelengths beyond 1450 nm⁸⁸. Yuan et al. ⁸⁸ proposed that increasing the number of grating layers could simultaneously enhance the transmittance and extinction ratio. Jin et al.⁸⁹ designed reflecting polarising beam splitters using a multilayer metal-dielectric grating structure. Their simplified model achieved extinction ratios of 62.2 dB and 66.4 dB for the −1st and 0th diffraction orders, respectively. Beyond polarisation selection, vectorial optical fields with spatially inhomogeneous polarisation states have garnered significant attention in areas such as nanofabrication, the investigation of novel light-matter interactions, and quantum optics. Polarisation gratings with inhomogeneous polarisation distributions and periodic vector-beam structures have opened new avenues for innovative applications in these fields⁹⁰.

CPA delivering femtosecond laser pulses with millijoule energy at kilohertz repetition rates constitute a major workhorse in nonlinear optics and ultrafast science. A large amount of dispersion comparable to that applied in CPA systems is also required in high-energy femtosecond laser oscillators operating at the microjoule level. These systems rely on complex, lossy, and alignment-sensitive stretchers and compressors. Dispersive mirrors with HL dielectric layers provide a compact and efficient solution for broadband dispersion control. These mirrors chirped the thickness of the coating layers to change the Bragg wavelengths within the structure, thereby affecting the penetration depth and group delays (GD) for different light wavelengths (Fig. 3c). Pervak et al.⁹¹ demonstrated that the required dispersion of up to the order of 10⁵ fs² could be introduced by a set of highly dispersive chirped MDFs, which provides several advantages, including simplicity, alignment insensitivity, and the potential for increased efficiency. Alqattan et al.⁹² reported a high-power four-channel attosecond light-field synthesiser with a broadband spectrum spanning 200-1000 nm, sub-femtosecond temporal duration, and attosecond resolution.

Nevertheless, given the constraints imposed by the fabrication tolerances, both the total number of coating layers and their cumulative physical thicknesses were limited, thereby constraining the achievable GD range. This results in a compromise between the extent of GD dispersion compensation and the usable operational bandwidth. Exploring the possibility of an attainable GD range effectively distributed across multiple, distinct frequency bands to increase the amount of achievable GD dispersion, rather than being uniformly applied across a single broadband, holds the promise of advancing ultrafast applications that target specific spectral ranges or multicolour excitations. Chia et al.⁹³ presented an innovative multiband design approach to optimise the dispersion control across preselected targeted frequency bands. This enabled more efficient pulse compression in highly dispersive optical systems by enhancing the compensation capacity per bounce and reducing the number of mirror bounces required. Advancing HDMs to even higher negative dispersion benefits the development of compact, user-friendly, high-power femtosecond oscillators, and may even open the prospect of simplifying pulse stretching-compression schemes in chirped-pulse amplifier systems.

PCG is also a widely applied dispersion element in the CPA systems. The key challenge for high-power applications of PCGs is to achieve diffraction gratings with near 100% DE for both TE and TM polarized light across a wide spectral range (Fig. 3d). So far, three main types of PCGs have been developed: metal gratings, MDGs, and metal MDGs. Metal gratings demonstrate good DE over a broad bandwidth of up to approximately 300 nm⁹⁴ and are still commonly used in femtosecond compressors for high-power lasers. However, their low LIDT owing to the intrinsic absorption properties of the metals limit their performance in high-power environments. Dielectric materials with negligible energy absorptions inherently exhibit high LIDT. In 1995, Perry et al.⁹⁵ introduced the concept of all-dielectric diffraction gratings for CPA systems. These MDGs consist of an HR coating and a grating structure, which balances the high passband efficiency and minimal depolarisation. Compared with metal gratings, dielectric gratings provide higher thermal capacity, improved efficiency, and greater resistance to laser-induced damage, making them ideal for high-power laser systems. Subsequent research, such as the study by Wang et al.⁹⁶ explored the design of MDGs using materials like HfO₂ and SiO₂ and achieved diffraction efficiencies above 97.5% for broadband applications. The primary focus was on improving the bandwidth and DE by optimising the structural parameters. Inserting a metal layer into MDGs is a promising solution that combines a broad bandwidth with a high LIDT. Metal MDGs have pulse durations compressed to the femtosecond range, which enables pulse bandwidths of 100 or even 200 nm^96–98. However, owing to mechanical stress, further investigations into manufacturing technology are still underway.

Metasurfaces, with their multidimensional design freedom, enable precise control over the amplitude, wavefronts, and polarisation distributions, which facilitates the planarization, miniaturisation, and integration of optical devices and systems. This capability has garnered significant global attention. However, metasurfaces typically achieve high efficiency at a single wavelength or within a narrow bandwidth, and their efficiency rapidly declines when they deviate from the designed wavelength. However, multilayers possess strong amplitude-control capabilities, which makes them ideal for applications that require a spectral cut off. By coupling multilayers with metasurfaces to form a quasi-3D microstructure, a high efficiency across a broader range can be achieved. Over the past few decades, the progressive research on metasurfaces and multilayer metasurfaces has significantly expanded the application domain of laser coatings, making them promising tools for advanced optical technologies.

The compression of light-element pellets using shock waves has broad significance in material science. By shaping laser pulses into concentric rings and focusing them convergently, the excitation efficiency can reach 10⁴ GPa/J, which enables tabletop laser amplifier systems to replace expensive large-scale facilities. Studies have demonstrated that synchronising and superimposing multiple converging shock waves can generate pressures that are greater than 20 times higher than those of a single source. To this end, the beams should be shaped and focused into microscale concentric rings, which conventional optics struggle to accomplish owing to their low LIDT and millimetre-scale resolution.

To address this limitation, Kai et al.⁹⁹ utilised a “Death Star” cavity featuring a four-mirror cyclic design where the last mirror is a partial reflector (PR). The PR allows part of the laser input to exit and offset the reflected beam horizontally for another round trip, thereby creating an output of $ N $ horizontally spaced pulses. This beam array was transformed into multiple rings on the sample surface using a single optical phase object (Fig. 4a). The metasurface included the optical phase of a refractive lens with a focal length of 20 mm to focus the beams on a common optical centre at the sample surface. The desired fluence distribution was achieved in the rings by adjusting the transmission coefficient. A dielectric metasurface with an LIDT of 1.1 J/cm² was applied, which replaced conventional optics and generated pristine high-fluence laser rings with excellent repeatability and efficiency, thereby advancing the ability to generate high shock pressures with precision and control. This innovative approach of using dielectric metasurfaces and the “Death Star” cavity demonstrates a significant advancement in achieving high-pressure shock waves with precise control, thereby marking a promising direction for future developments in metasurface for high-power laser applications.

Fig. 4  Metasurfaces towards potential applications in high-power laser systems. a High-power laser beam shaping by TiO₂ nanopillars to generate laser rings without any visible spurious hot spots in the center⁹⁹. b Broadband depolarized quasi-3D microstructure by HfO₂ and SiO₂ with perfect Littrow diffraction for beam combining¹⁷. c Perfect anomalous reflector by all-dielectric quasi-3D microstructure for beam scanning¹⁰⁰.

Littrow diffraction is the fundamental mechanism for controlling light waves in diffractive optics. When light is incident on a grating at a Littrow angle, the negative first-order diffracted light aligns with the incident direction, which makes it an effective retroreflector. This technique has been extensively studied and applied to various optical systems. Moreover, broadband Littrow diffraction is crucial for applications in micro-spectrometers, optical communication, pulse shaping, and laser spectral synthesis. However, the performance of Littrow gratings such as metal-blazed gratings and MDGs is limited by their low broadband efficiency and significant polarisation dependence.

Dong et al.¹⁷ addressed these challenges by defining the physical conditions necessary to achieve high efficiency for single-wavelength depolarisation. They demonstrated that when the phase difference between two Bragg modes in the grating layer is π under Littrow incidence, both polarisation states can be satisfied simultaneously (Fig. 4b). Additionally, by adjusting the broadband dispersion of the Bragg mode phase difference to 0, they ensured that the phase difference remains constant at π across a wide bandwidth, which results in high-efficiency broadband depolarisation. Based on this principle, they proposed a quasi-3D microstructure that combined a 1D all-dielectric multilayer with a 2D free-form metasurface. This microstructure offers the freedom to control the phase difference and phase dispersion of the Bragg modes within the structure. Through topological optimisation of the microstructural shape, perfect Littrow diffraction with broadband depolarisation that reached an efficiency greater than 99% in the wavelength range of 1030–1090 nm was achieved. Multilayer films were fabricated using EBE and IAD techniques, and a metasurface was created through electron beam direct writing combined with atomic layer deposition (ALD). The resulting quasi-3D multilayer metasurface demonstrated an unpolarized broadband efficiency as high as 98%, thereby effectively overcoming the “broadband efficiency” bottleneck. This advancement is expected to significantly enhance the development of optical metasurface-based laser spectral synthesis.

Anomalous reflection is a fundamental technique for controlling light waves through metasurfaces and serves as the basis for various advanced applications, such as beam scanning in spectrometers and lidar systems. Currently, the anomalous reflection efficiency of optical metasurface devices is typically less than 90%, which falls short of the efficiency required (> 99%) for use in high-performance laser systems. This creates an urgent need to advance the scientific understanding of the anomalous deflection efficiency and develop innovative methods to achieve perfect anomalous deflection at optical frequencies.

Historically, the scientific understanding of anomalous reflection efficiency has progressed in two main stages. Initially, phase gradients were used to control the direction of the reflected beam. However, controlling only the phase gradient is insufficient to achieve the desired efficiency of approximately 100%. Recent studies have demonstrated that both the phase and amplitude must be co-regulated to achieve perfect anomalous reflection. In particular, regulating the amplitude requires the introduction of gains and losses in different regions of the metasurface, which poses a new challenge in the design of optical metasurfaces.

He et al.¹⁰⁰ addressed this challenge by defining the energy flow distribution necessary for perfect anomalous reflection. They proposed a new quasi-3D subwavelength structure that suppressed the second-order energy flow term while matching the first-order term. This structure efficiently coupled incident light to Bloch waves within the metasurface and transmitted waves through the spacer layer. By adjusting the reflection amplitude and phase within the multilayer film, the energy flow distribution generated by the Bloch and transmitted waves was controlled to achieve the energy flow required for a perfect anomalous reflection.

For example, they designed a structure to achieve a perfect anomalous reflection at 40° under normal incidence. This design uses a gradient metasurface and a highly reflective film as the starting point. By fine-tuning the phase responses $ {{\phi }}_{0} $ and $ {{\phi }}_{1} $ of the multilayer film, the energy flow distribution necessary for perfect anomalous reflection was realised. The final quasi-3D subwavelength structure, comprising a multilayer film and a metasurface, achieved an anomalous reflection efficiency greater than 99% at a wavelength of 1550 nm. The metasurface was fabricated using magnetron sputtering and electron-beam direct-writing technologies, and the resulting quasi-3D wavelength-scale device (Fig. 4c) was tested, which demonstrated an anomalous reflection efficiency of 98%. The strong agreement between the experimental and theoretical design validates the accuracy and reliability of the fabrication and testing processes. Furthermore, they introduced designs for polarisation-independent and polarisation-selective anomalous reflection devices, highlighting the immense potential and efficient control capabilities of quasi-3D wavelength structures in advanced optical systems.

Metasurfaces, as a cutting-edge generation of optical devices, enable near- and far-field wavefront regulation at the subwavelength scale and have attracted significant attention since their introduction less than a decade ago. We highlight their potential applications in high-efficiency laser-beam shaping, combining/splitting, and steering. Specific design strategies such as silicon cross-resonators with linear phase gradients¹⁰¹ facilitate polarisation-independent beam steering, whereas meta-atoms with birefringence and rotation angles provide complete and independent control of the optical amplitude and phase¹⁰². Despite their potential, the integration of metasurfaces into high-power laser systems remains challenging. The key requirements include achieving high efficiency, high LIDT, and large apertures to meet the demands of practical applications.

Laser damage in optical coatings has long been a primary limitation of high-power laser systems. Decades of research on damage mechanisms have realized that partially recoverable pre-damage degradation, initial damage, and damage growth are strongly linked to three factors: the EFI, imperfections within the optical coatings, and the bandgap of the coating material. Numerous studies have explored strategies to improve the laser resistance by addressing these factors.

It has been established that the most fragile areas in these coatings are the surface, interface, and microstructure ridges. Reducing the EFI at these critical locations can substantially improve the LIDTs.

The highest EFI in multilayers typically occur at the topmost interfaces and within the nodules. Zhu et al.⁸⁶ designed four HfO₂/SiO₂ Brewster’s polarisers for 1064 nm and compared their EFI, absorption, and LIDT. For both P- and S-polarised light, the lower the peak EFI and the further the maximum EFI layer was from the air, the higher was the LIDT. The electric field enhancement became more significant in defects and microstructures when their geometric scale was comparable to the wavelength of the incident light. EFI simulations and damage morphology analyses of two types of broadband low-dispersion (BBLD) mirrors revealed that the nodules intensified the light in the middle of the boundary between the nodule and surrounding coating, which enabled the nodules to bypass the protective outer layers and reach the vulnerable layers, resulting in a much lower LIDT¹⁰³. The experimental results for the MDGs consistently demonstrated a strong correlation between the damage sites and EFI peaks. Kong et al.¹⁰⁴ proposed the use of an electron production model involving photoionization and avalanche ionisation to calculate the theoretical LIDT based on the average EFI of the grating ridges. Liu et al.¹⁰⁵ combined this with a nonlinear ionisation model involving the transient distribution of the lattice temperature field to calculate the LIDT of MDGs. The model predictions of the LIDT were in reasonable agreement with the reported experimental data. Moreover, the results of the electron density simulation suggested that, under laser irradiation, the imaginary parts of the refractive index within the grating pillars were significantly enhanced. This resulted in the strong absorption of laser energy, which ultimately caused permanent damage to the structure¹⁰⁶.

Because low-refractive-index materials generally have a larger bandgap than high-refractive-index materials, positioning the EFI peak within the low-refractive-index layer can enhance the overall laser damage resistance (Fig. 5a)¹⁰⁷. Bellum et al.¹⁰⁹ reported on a TiO₂/SiO₂ BBLD mirror designed to reduce EFI in high-index materials. This mirror performed well for both S- and P-polarisation scenarios, and laser damage tests under sub-nanosecond and femtosecond pulses demonstrated improved LIDTs. Willemsen et al.¹⁰⁷ also demonstrated that replacing the binary Ta₂O₅ layer, where the EFI was highest, with a nanolaminate sequence or a ternary composite layer with a larger bandgap reduced the maximum field intensity by 190% and 200%, respectively, thereby further enhancing the LIDT. For dispersive MDFs experiencing unexpectedly intense nonlinear responses, Razskazovskaya et al.¹¹⁰ and Gui et al.¹¹¹ demonstrated that positioning the H layers in a weak electric field and the L layers in a strong electric field can impede the nonlinear absorption and temperature rise. Adjusting the EFI distribution within nodule defects can be realized by modifying the focusing and penetration behaviour of light, and has been applied to multilayers. Cheng et al.^112,113 designed a wide-angle reflective HfO₂/SiO₂ film with a reflection angle bandwidth for P- and S-polarised lights wider than the typical incidence angle for nodal defects. This broadband structure prevented light from penetrating the reflective layer where the nodule defect was located, thereby significantly reducing the electric-field enhancement effect (Fig. 5b). This design caused a 75% reduction in EFI compared to conventional quarter-wavelength HfO₂/SiO₂ films and nearly doubled the LIDT. Additionally, Ristau et al.⁵¹ and Khabbazi Oskouei et al.¹¹⁴ demonstrated that reducing EFI can limit THG.

Fig. 5  Structure optimization to minimize EFI and improve LIDT. a Replacing small bandgap layers with large bandgap materials¹⁰⁷. b Changing the light penetration behaviour by broadband structure. c The damage morphology of NS¹⁰⁸. Reprinted with permission from ¹⁰⁸ © Optical Society of America.

Adjusting the EFI within the microstructures is significantly more complex than adjusting it within multilayers. Optimisation to a low EFI can only be accomplished using a merit function during numerical DE calculations. Consequently, the choice of the merit function has a direct effect on the outcome. One preferred method for defining the merit function is to use EFI to normalise the DE, thereby balancing performance and damage resistance, such as $ MF={{\eta }_{1}}/{{E}_{1}^{2}} $^115,116 for single wavelength, $ MF={1}/{N}{\sum }_{{\lambda }_{1}}^{{\lambda }_{N}}({DE\left({\lambda }_{i}\right)})/({{Emax}_{{\lambda }_{i}}^{2}}) $¹¹⁷ for broadband, $ MF= {1}/{N}{\sum }_{{\lambda }_{1}}^{{\lambda }_{N}}[1-{DE}_{-1R}({\lambda }_{i})+Emax({\lambda }_{i}\left)\right] $¹¹⁸ for DE and EFI considered equally in broadband optimisation and $ MF\{f,d,{h}_{1},{h}_{2}\}= \{{1}/{N}{\sum }_{{{\lambda }}_{1}}^{{{\lambda }}_{N}}[{\left({{{\lambda }}_{i}}/{4}-\left({\Delta }{n}_{1}^{TM}{h}_{1}\,+\,{\Delta }{n}_{2}^{TM}{h}_{2}\right)\right)}^{2}\, + {\left({{{\lambda }}_{i}}/{2}-\left({\Delta }{n}_{1}^{TE}{h}_{1}+{\Delta }{n}_{2}^{TE}{h}_{2}\right)\right)}^{2}]\}^{1/2} $⁸⁹ for broadband depolarisation. Although they are all reasonable, efficient, and comprehensive, a compromise between EFI and DE is intrinsically indicated in the formula, and they rely heavily on full-electric-field calculations through FDTD or RCWA. However, these methods are computationally demanding, particularly for multi-dimensional microstructures. Reduced-dimension and reduced-order algorithms are required to optimise the electric fields more efficiently, representing a key area for further exploration.

The concept of microstructures introduces a new dimension for adjusting the LIDT by altering its geometric characteristics. Porous/nonporous all-glass multilayers by GAD and all-glass NS by NP combined etching demonstrated the capability to create scalable, designable coatings that were monolithic to the substrate; therefore, their LIDT was close to that of bulk glass. The porous/nonporous all-glass HR coating achieved a damage probability that did not exceed 20%, even for fluences as high as 65 J/cm² (the highest available fluence of the measurement system)¹⁰, and the all-glass NS AR coating achieved an LIDT of 39 J/cm² with sample cleaning and 74 J/cm² after laser conditioning¹⁰⁸, which indicates an outstanding potential intrinsic LIDT value. These techniques enable large-scale applications, rendering them feasible for metre-scale optics. Moreover, laser damage to NS surfaces exhibits a non-growing failure mode, where the initial damage site does not propagate further, as shown in Fig. 5c¹⁰⁸, thereby reducing the demand for managing optics in high-power laser systems. Furthermore, NS enables the control of additional surface properties beyond optical performance, such as reducing microparticulate adhesion, improving water resistance, being mechanically durable, being capable of withstanding the cleaning processes such as sonication, and offering practical advantages for lasers operating in harsh environments.

With advances in layer deposition and postprocessing technologies, along with a deeper understanding of defect-induced damage mechanisms, the concept of material engineering has gained prominence in improving the optical coatings for high-power laser applications. By tailoring the material properties through precise engineering, numerous strategies have emerged to improve the bandgap, mitigate the effects of defects, and strengthen the mechanical stability of optical coatings in demanding laser environments.

Off-stoichiometric and structural imperfections owing to rapid film growth often results in an intra-bandgap state, which increases the absorption and reduces the LIDT^67,119. High-temperature annealing has proven to be effective in improving stoichiometry and structural uniformity. Pu et al.¹²⁰ studied the annealing of Ta₂O₅/SiO₂ filters and revealed the number of absorption points, owing to sub-stoichiometry, was reduced in annealed films compared to as-deposited films, and the LIDT could be improved nearly threefold, from 59.32 to 158.87 J/cm², when absorption was reduced by a factor of five, from 39.99 to 7.2 ppm at 600 °C. Lin et al.¹²¹ annealed plasma-enhanced ALD (PEALD) HfO₂ coatings in nitrogen and oxygen atmospheres at different temperatures and investigated the effects on the properties of the coatings. When the HfO₂ coating was annealed in an oxygen atmosphere, the carbonates and amines decomposed. The C and N contents in the coatings decreased or even disappeared. The oxygen-enriched environment enabled the elimination of oxygen deficiencies in the coatings. The LIDT of the PEALD HfO₂ coating annealed at 700 ℃ in an oxygen atmosphere was nearly three times that of the as-deposited coating. However, this process can also cause crystallisation or recrystallisation (Fig. 6a)¹²², altering the volume of the film and introducing internal stress, which can cause cracks and compromise the performance.

Fig. 6  Material engineering strategies. a Annealing towards reduced absorption and increased stress owing to crystallization¹²². b Laminating to avoid crystallization at high annealing temperature¹²³. c Nanolaminates and their theoretical optical characteristics towards high LIDT¹²⁴. d Rugate filter with improved EFI and LIDT¹²⁵. e Sandwich-like structure to improve the mechanical and thermal duability¹²⁶.

Artificial nanocomposite materials have been proposed as a solution to address the conflicting requirements of annealing and crystallisation of natural materials. These engineered materials can be designed to maintain stability during high-temperature processes, such as annealing, without crystallisation, which is essential for enhancing the durability and performance of optical coatings in high-power laser systems. Niu et al.¹²³ proposed the use of co-evaporation technology to dope SiO₂ into HfO₂ thin films, which not only effectively inhibited crystallisation and reduced absorption but also broadened the bandgap and increased the intrinsic damage threshold of the coatings (Fig. 6b). The transmittance spectra of these films revealed that the short-wave cut off edge of the Hf_0.7Si_0.3O₂ hybrid film significantly surpassed that of pure HfO₂, which indicates a wider bandgap. Moreover, the relationship between the SiO₂ content and bandgap demonstrated that as the SiO₂ content increased, the bandgap also increased, and Hf_0.60Si_0.40O₂ achieved a maximum bandgap of 6.01 eV. These findings highlight the potential of Hf_1-xSi_xO₂ films to improve the bandgap and LIDT. Willemsen et al.¹⁰⁷ employed ion beam sputtering (IBS) to prepare HfAlO_x and TaSiO_x mixed materials, resulting in 188% and 20% increases in the damage thresholds under nanosecond and picosecond lasers, respectively. Shi et al.¹²⁷ used high-temperature annealing to reduce the oxygen vacancy density in Hf_0.5Al_0.5O_x, thereby further increasing the LIDT by 20% without crystallisation.

Lamination is another effective strategy to maintain the amorphous structure after annealing. Thin layers of amorphous materials with low absorption and refractive indices were periodically inserted into high-refractive-index materials to create laminated films. This structure inhibited crystallisation, modified the microstructure, reduced absorption, and broadened the bandgap. The alternating HL layers in laminated films are not only easy to model and prepare but also cost-effective, generating significant interest. Studies show that thinner sub-layers of materials like TiO₂ raise the crystallisation temperature, which makes the film less prone to crystallisation, thereby maintaining the amorphous state even after annealing¹²⁸. However, in nanocomposites and laminates these modifications have inherent limitations and cannot surpass the fundamental characteristics of bulk materials; for instance, the refractive index and energy of the absorption edge are intrinsically linked. Quantised nanolaminates (QNL), which were first proposed in 2017¹²⁹, represent a new class of engineered materials. In QNLs, thin sub-nanometre layers of high- and low-refractive-index materials are stacked, resulting in quantisation effects, as shown in Fig. 6c. These effects enable independent shifts in both the refractive index and bandgap, unlike classical dielectric materials. Successful fabrication of QNLs has been reported using ALD¹²⁴, IBS¹²⁴, and magnetron sputtering¹³⁰.

In conjunction with these developments, a significant focus has been placed on gradual interfaces, which are being explored as alternatives to traditional stacks of discrete material layers that create binary step-like variations in dielectric constants (Fig. 6d). Laser damage tests by Jupé et al.¹²⁵ showed that the LIDT of HR Rugate filters using TiO₂ and SiO₂ was improved by more than a factor of ten compared with the standard components in nanosecond lasers at 1064 nm, despite the fact that rugate filters exhibit higher EFI and absorption values. In addition, a comparison between rugate filters and HL quarter-wave stack HR coatings at 1064 nm using Ta₂O₅ and SiO₂ deposited by ion-beam sputtering¹³¹ showed that although there was no significant difference in the initial LIDT or damage morphology, the rugate filters demonstrated greater resistance to delamination at higher laser powers. Zeng et al.¹²⁶ noted that the mechanical stress within the gradient structure and the simulated temperature rise likely contributed to the improved laser resistance, as illustrated in Fig. 6e. They also developed a dichroic mirror using Al₂O₃-HfO₂ as the H material in a superimposed sandwich structure. This design effectively doubled the LIDT by leveraging both nano-compositing/nano-laminating techniques and gradient structures. Their subsequent work¹³² on a mixture-based design strategy demonstrated a 2–2.6 times improvement in LIDT. Future trends suggest combining gradient layers with multi-objective optimisation strategies, incorporating interface EFI, or utilising nanolaminates to push the theoretical optical gap limit, which could further enhance the LIDT of coatings.

Although the improvement in the LIDT for gradual interfaces over traditional layer stacks is evident, the exact magnitude of this effect remains unclear because of the inability to fully separate the material properties from design effects. Further research is required to address the absorption and defect issues in gradient interfaces for an optimised performance.

The preparation of optical coatings involves multiple processes such as substrate polishing, multilayer deposition, photoresist mask preparation, and RIBE. These steps can introduce defects such as imperfections within the surface, subsurface, and multilayer films, photoresist residues, and etching-induced damage, including chemical bond destruction and redeposition. While polishing scratches and residual surface/subsurface damage can be minimised through optimised cleaning and polishing, particularly for large-aperture applications, the most commonly observed problems include nodules and nano-absorption centres.

Although nodules are ejected at a low fluence, the ejected pits are very stable up to a high fluence⁵⁸. Therefore, conditioning optical coatings with low-fluence lasers is an efficient method for eliminating nodules. It has been successfully applied in HR coatings and polarisers. Recently, Shuai et al.¹³³ integrated nanosecond laser conditioning into various manufacturing stages of gratings on MDFs and final grating products, increasing the picosecond LIDT of nodular ejection pits by 40% compared to untreated nodular defects, and the surrounding grating structure was preserved, thereby avoiding additional losses owing to nodular defects. However, laser conditioning causes plasma burns or nodule-ejected pits, necessitating a trade-off between the optical performance and LIDT. McCauley et al.¹³⁴ investigated the suitability of ion etching of substrates to mitigate the impact of nodules. While the process effectively removed nodules, it also introduced small defects on the substrate surface, along with a noticeable “shadow” where they were removed, and the multilayer around nodules was often peeled. Despite the LIDT measurements showing no significant reduction in substrate durability, the pits caused an increase in optical loss.

Nodule planarization has been proposed to eliminate nodules while maintaining surface flatness. This approach smooths nodules rather than removing them, prevents surface roughness, and minimises scattering losses, thereby balancing nodule mitigation while preserving the optical quality. Stolz et al.¹³⁵ applied ion-beam etching to reduce the size of the nodules during deposition, as shown schematically in Fig. 7a. In the dual IBS system, the main ion source deposits the film, while an auxiliary ion source perpendicular to the substrate performs the etching. By adjusting the beam pressure and current, the etching rate could be increased at oblique incidence angles (45-50°), which was twice as fast as that at the vertical incidence. This method reduced the size of the nodules, resulting in a smoother surface. After laser testing, an LIDT exceeding 100 J/cm² under 10 ns pulse laser irradiation was achieved, which was a nearly a 20-fold increase. Cheng et al.¹³⁸ performed systematic studies on nodule flattening using spherical seed sources under real-world conditions. Tests with seed diameters of 0.5, 1, 1.5, 2, and 3 μm and SiO₂ flattening layers of 1.25 and 2.5 μm showed that seed sources of 1 μm were fully flattened, resulting in a smooth surface. Larger seeds were only partially flattened, however, the nodule size was significantly reduced. This technique effectively minimises the size of the nodule and decreases the interfacial correlation, enabling the deposition of high-quality mirror coatings, although it is mainly applicable to IBS.

Fig. 7  Strategies to eliminate nodules and nano-absorption centers. a IBS-based nodule planarization and NDR process¹³⁶. b Ultrasonic cleaning and oxygen plasma cleaning for metasurfaces¹³⁷. c Additive manufacturing strategy for MDGs¹¹⁷.

IBS-based planarization flattens the nodules during layer deposition, whereas another planarization method focuses on post-polishing. Liu et al.¹³⁶ proposed a nodule dome removal (NDR) strategy, as shown in Fig. 7a, in which anhydrous ethanol was applied to a polishing pad, which created a uniform ethanol film between the pad and the coating. As the coating rotated on the polishing pad, the increasing capillary force between the surfaces acted as a normal load, gradually polishing the nodule dome via friction. This method increased the LIDT of UV mirror coatings by approximately 1.9 to 2.2 times for coatings without artificial nodules and with artificial nodule seeds of 1 μm diameter, respectively. Compared to nodule planarization, the NDR method offers greater flexibility across different coating techniques. However, the use of polishing solutions introduces the risk of penetration into porous structures, potentially leading to corrosion. This means that the NDR method may not be suitable for all types of optical coatings.

The size of the nano-absorption centres is beyond the detection limit of most conventional detection methods (AFM, profilometry, etc.), and their impact is severe, especially for short wavelengths, as discussed in the previous section. Chemical cleaning, including dry and wet processes, such as plasma cleaning and RCA cleaning, is most widely used to remove nanoscale absorption defects on optical coatings. Wet chemical cleaning uses specific cleaning solutions such as the piranha solution¹³⁹ and electrostatic repulsion to remove contaminants, which is cost-effective and widely applied in semiconductor industries. Liu et al.¹³⁷ studied ultrasonic and oxygen plasma cleaning methods for metasurfaces (Fig. 7b). The optimal cleaning method significantly increased the LIDT to 56 J/cm², which is comparable to that of bare-fused silica. Other cleaning techniques such as plasma cleaning¹⁴⁰ and gas-phase chemical cleaning have also been applied. However, wet cleaning methods introduce new penetrating impurities, whereas dry cleaning may increase the surface roughness. Furthermore, dry etching frequently introduces nano-absorption centres at the microstructure ridges, which are also the positions that experience the highest EFI. Therefore, their impact on the LIDT was strengthened. To address these issues, Xie et al.¹¹⁷ proposed an additive manufacturing strategy for MDGs that integrated laser interference lithography, nanoimprinting, ALD, and RIBE (Fig. 7c). The MDG achieved almost 90° sidewall angles, reduced the EFI enhancement, and prevented nano-absorbing defects at the ridges. In addition, the self-limiting surface reaction during ALD ensured precise control of the stoichiometric ratios and atomic-level thickness, minimising the structural degradation of high depth-to-width ratio structures and absorptive defects, and further increasing the efficiency.

In conclusion, the control and mitigation of defects in optical coatings are critical for achieving a high LIDT and maintaining optical performance. Optimised cleaning and polishing techniques that can eliminate scratches, pits, nodules, and nano-absorption centres remain challenging. Advances in nodule planarization, NDR, and additive manufacturing have shown great promise in minimising these defects and significantly improving the surface quality and LIDT of optical coatings. Continued refinement of these techniques and their combinations is essential for further enhancing the durability and efficiency of coatings in high-power laser applications.

Aperiodic multilayers and metasurfaces exhibit exceptional optical performance, however, they significantly increase the design complexity owing to the larger number of parameters that require optimisation. For instance, Tikhonravov et al.^141,142 developed an optical coating design software based on a needle optimisation technique (Fig. 8a) with a merit function containing complex spectral quantities. More recently, optimisation methods such as genetic algorithms¹⁴⁵, particle swarm optimization¹⁴⁶ and ant colony algorithms¹⁴⁷ have been applied in the field, as shown in Fig. 8b. Similarly, Lininger et al.¹⁴⁸ used convolutional neural networks (a deep learning subset) to solve the inverse design problem for metamaterials comprising thin-film stacks, demonstrating their ability to explore vast design spaces and map relationships between the metamaterial structure and spectral properties. Han¹⁴³ introduced an innovative inverse design approach that applies backpropagation to a networked transfer matrix (Fig. 8c) using gradients to optimise layer thicknesses without additional learning processes, thereby significantly reducing the calculation time and enhancing the efficiency in thin-film design, which emphasises the need for more advanced optimisation frameworks with enhanced search capabilities. Analytical methods that enable dimensionality reduction, such as the method proposed by He et al.¹⁰⁰ for abnormal deflection devices (Fig. 8d) using eigenmode expansion and scattering matrix mode transfer, have also been widely studied. However, many of these algorithms assume that the film materials are not optically dispersive, resulting in a decline in the practical broadband efficiency. For certain applications, electric fields, material gaps, and extinction coefficients must also be considered.

Fig. 8  Inverse designing algorithms for optical coatings. a The needle method. b The particle swarm optimization and colony algorithms. c The neural network algorithm with backpropagation and gradient descent¹⁴³. d The eigenmode expansion and scattering matrix mode transfer method¹⁰⁰. e The deep Q-learning with Markov decision process¹⁴⁴.

Addressing numerous parameters transforms the optimisation problem from a multiparameter-single output challenge to a multiparameter-multi output scenario, where traditional algorithms fall short. In such cases, weighting each output parameter in the merit function is crucial. This complexity has been increasingly addressed using artificial intelligence (AI). Reinforcement learning, which has recently been applied to several physical problems, is promising for the multi-objective optimisation of optical coatings. By treating the total optical film optimisation process as a Markov decision process, RL can solve multi-objective challenges and extend beyond multilayer films to complex microstructures (Fig. 8e)¹⁴⁴. It has already been successfully applied to optimise selective solar absorber thicknesses and holds the potential for broader use in multi-objective optimisation within the field of high-power laser coatings.

The fabrication techniques for optical multilayers and microstructures face challenges in balancing precision, scalability, and application-specific requirements. In high-power laser systems, achieving low optical loss and high LIDT is critical for maintaining the efficiency, imaging quality, and system stability. The choice of deposition technique depends on the application demands and often requires a combination of methods.

EBE can control the layer thickness to several nanometres and is widely used owing to its high-throughput and large-aperture fabrication capabilities. However, EBE produces films with columnar microstructures and low densities, which adversely affect their mechanical properties. Ultra-low optical loss has become a critical requirement for applications such as laser gyroscope^149,150, gravitational-wave observatory¹⁵¹, and optical clocks¹⁵². IBS enables modern coatings to achieve absorption levels in the ppm range, with thickness control at the nanometre scale, surface roughness below 0.05 nm, and high conformality, making it suitable for high-quality coatings like ultra-low-loss multilayers, filters, and dispersive mirrors. While their total integral scattering results still range from 2 × 10⁻⁶ to 5 × 10⁻⁶ at a wavelength of 633 nm, which suggests that further reductions in surface roughness are reaching their limits. A focus on reducing the interface scattering loss is required^153–155. Techniques have been developed based on the principles and models^156–158. Trost et al.¹⁵⁹ and Zhang et al.¹⁶⁰ demonstrated that, in oblique-angle deposition, the cross-correlation of interfaces decreases, thus reducing scattering. Additionally, Peverini et al.¹⁶¹ reported that etching each newly formed layer after IBS can alter the surface morphology and PSD trends, suggesting that combining IBS with etching can effectively eliminate interface correlations, thereby presenting new possibilities for the further reduction of interfacial scattering. For high LIDT demands, continuous advancements in defect control during IBS, along with a deeper understanding of film growth and scattering in defects and microstructures, are driving improvements in IBS.

Chemical deposition methods excel at handling complex geometries and large-aperture production. Films produced by chemical vapour deposition (CVD) feature three-dimensional network structures with fewer off-stoichiometric defects, resulting in higher LIDT¹⁶². However, issues with uniformity and precise thickness limit their use. ALD addresses these challenges with atomic-level precision and conformality, which are ideal for protective coatings and additive manufacturing of microstructures¹⁶³; however, it faces difficulties in controlling roughness owing to nanocrystal formation in thick layers¹⁶⁴ and slow growth rates. Additionally, the growth rate, refractive index, density, crystallisation ratio, and LIDT of ALD layers are highly dependent on the precursor type and reaction temperature and vary with the layer thickness^165–167. The sol-gel method, which has been used for over three decades¹⁶⁸, produces large-aperture high-LIDT AR coatings. Zhang et al.¹⁶² reported HfO₂ films via sol-gel with an LIDT of 31.6 J/cm² at 1064 nm, 12 ns, surpassing those from EB¹⁶⁹ and IBS¹⁷⁰. Recently, UV-controlled sol-gel methods have shown promise for fabricating microscale microstructures without a photoresist, offering a low-cost and scalable approach¹⁷¹, showing promise for large-aperture applications. However, poor thickness-control precision, weak mechanical resistance, and pollutant adsorption are challenges that must be addressed for long-term applications. Advanced deposition techniques promise to improve material performance by capitalising on their unique advantages. Precise control over the stoichiometric ratio of mixed materials is a common problem that requires a comprehensive understanding of the growth characteristics of each layer and careful in situ monitoring and control of the layer thickness.

Microstructural fabrication can be classified into subtractive and additive manufacturing methods. Nanostructures for AR coatings are typically produced using subtractive methods, such as self-assembly and etching. These processes do not require complex lithography and are low-cost, scalable, and high-throughput processes, making them suitable for large apertures, although they offer limited design flexibility. The subtractive manufacturing of metasurfaces has also been widely reported and often involves lithography, layer deposition, and reactive-ion etching. This approach allows precise patterning with resolutions below 10 nm for both isolated features and periodic structures^172,173, making it applicable to advanced imaging and trace detection systems.

By contrast, additive manufacturing uses layer deposition instead of etching to transfer patterns. This technique offers better aspect ratios and fewer etching defects, particularly at the ridges, making it ideal for high LIDT applications¹¹⁷. The additive process for metasurfaces typically involves lithography, CVD, and etching. Although widely used in nanoelectronics, storage, and imaging, commercial challenges remain, such as improving throughput, reducing defects, and ensuring precise pattern alignment. In addition, the application of additive manufacturing in high-power laser systems is limited because of the stringent requirements for defect-free subwavelength structures over areas greater than 50 mm², which are necessary to withstand high energy densities and enhanced EFI.

Defects, particularly those located at the boundaries and near the top surface, such as internal nodules within the multilayer, photoresist residues on the sidewalls and tops of the structures, and chemical bond disruptions caused by etching, are linked to laser-induced damage. Achieving full process control depends on accurate detection, which remains a challenge when conventional methods are used.

Infrared spectroscopy (IR) and Raman spectroscopy (RS) are powerful techniques that provide molecular information; however, they have limitations owing to their low absorption contrast and spatial resolution. Recent advancements in AFM-based imaging techniques, such as AFM-IR (Fig. 9a)¹⁷⁴, scanning near-field infrared microscopy (Fig. 9b)^175,176, and tip-enhanced RS¹⁷⁸, show promise for addressing these detection challenges in a nondestructive mode with a high signal-to-noise ratio. The first technique detects the light absorbed by a sample and offers sensitivity to IR radiation at specific depths within the material. Its absorption spectrum closely aligns with the FTIR spectrum and is free from thickness-dependent peak shifts, making chemical analysis and compositional mapping using AFM-IR data straightforward. By contrast, the other two techniques focus on scattered light, which primarily reflects surface-level interactions. In these methods, the signal is influenced by the complex optical properties of the tip, sample, and underlying substrate.

Fig. 9  Advanced detection techniques with defect sensitivity. a Resonance-enhanced AFM-infrared spectroscopy¹⁷⁴. b Scanning near-field infrared microscopy^175,176. c Angle-resolved scattering (ARS) 3D mapping¹⁷⁷.

Although nodules have a detectable geometric scale, their morphologies blend into the background, and surface morphology detection is limited in coverage, making isolated defect identification challenging¹⁷⁹. Studies have shown that even one nodule per square millimetre in HR coatings can cause double or triple scattering losses, indicating that scattering-based techniques are sensitive to nodules. In addition, Munser et al.¹⁸⁰ proposed the sensitivity of the ARS to subsurface-related differences. Therefore, ARS mapping (Fig. 9c)^{177,179–182} offers the potential for the detection of nodule defects and subsurface damage in a fast, nondestructive mode. In addition, the original texture of the substrates from processes such as grinding and polishing, combined with coating growth patterns, often leads to anisotropic surface and interface morphologies featuring defects. These features can also be measured sensitively using the ARS technique.

Optical coatings play a crucial role in high-power laser systems by facilitating the various functions of the main laser, alignment laser, and stimulated Brillouin scattering. Microstructures offer alternatives to conventional AR, HR, and passband filters, adding new capabilities such as replacing PCGs and DOEs, while increasing system compactness, particularly in spaceborne and airborne applications.

Advances in light modulation and computational science enable flexible design processes, merging traditional optical merits such as spectrum performance, polarisation, and beam shaping, with considerations for error tolerance, nonlinear effects, EFI distribution, and fabrication feasibility. These enhancements have spurred the need for diversified merit functions, multi-objective optimisation strategies, and design models that combine physical, analytical, and experimental results.

Laser resistance remains a key limitation in the output power, prompting research into damage mechanisms and methods for improving coatings. Material advancements such as post-annealing, mixing, and laminating aim to modulate the bandgap and enhance LIDTs. Further studies have focused on reducing the damage in critical areas such as multilayer interfaces, microstructure ridges, and defect-prone zones. Simulation tools, such as FDTD and RCWA, have aided in calculating the electric field distribution around defects, revealing how electric field enhancement contributes to damage at interfaces and nodules. Techniques such as wide-angle HR coatings, laser conditioning, planarization, and additive manufacturing have emerged to mitigate these challenges. Novel approaches, including effective medium nanostructures, aim to make the LIDT of coatings comparable to those of bulk materials.

Despite progress, the practical performance often falls short of theoretical designs owing to fabrication errors. Non-quarter-wave multilayers and subwavelength microstructures face challenges in terms of thickness monitoring and structural integrity during fabrication. These issues highlight the need for designs that consider fabrication feasibility, optical performance, and laser resistance. With the rapid advancements in fabrication technologies, multilayers, nanostructures, and gratings are now accessible for industrial and commercial applications. Despite their commercial applications in nanoelectronics, the large-scale fabrication of metasurfaces for high-power laser applications remains a significant challenge because of the limitations in defect mitigation, high-resolution template fabrication, and precise pattern alignment. In addition, detection techniques for nanoscale defects must be improved, as conventional methods struggle to identify nano-absorption centres or distinguish subsurface defects from surface morphology. Emerging detection techniques such as near-field nanoimaging, AFM-IR, and scanning near-field infrared microscopy promise resolutions below 10 nm by combining AFM with physicochemical-sensitive methods. To meet the demands of larger apertures, lower optical loss, higher LIDT, advancements in multi-objective optimisation, continuous development of layer deposition, microstructure manufacturing techniques with a deeper understanding of the layer growth mechanism, and defect identification advancements integrating near- and far-field detection require further exploration.

In summary, addressing the challenges of low optical efficiency and limited laser load capacity requires continued research on the design, optimisation, material engineering, and defect elimination. Quasi-3D microstructures show promise for replacing traditional optical components and enhancing laser system performance. Future research should focus on refining the multi-objective optimisation and developing advanced fabrication and detection techniques.

This research was supported by the National Natural Science Foundation of China (Nos. 62192772, 62475193, 62275196, 62305251, 6201101335, 62020106009, and 62192770), Fundamental Research Funds for the Central Universities, and National Key R&D Program of China (2022YFA1603403 and 2023YFF0715602). We gratefully acknowledge the collaborators who provided samples and measurements described in this review.