Introduction:
Attosecond pulses capture ultrafast electron and light–matter dynamics on the shortest timescales accessible to controlled experiments, making them indispensable tools for ultrafast spectroscopy, strong-field physics, and high-resolution imaging. Since their first measurement in 2001, attosecond pulse durations have been reduced from 650 as to 43 as, while photon flux, photon energy, and repetition rate have steadily improved. Yet the ultimate performance of attosecond sources remains largely bottlenecked by one crucial element: the driving ultrafast laser.
Recently, the group of Prof. Jinwei Zhang and Prof. Ka Fai Mak at Huazhong University of Science and Technology, together with Prof. Ferenc Krausz's team at the Max Planck Institute of Quantum Optics, published a Review article entitled "Ultrafast lasers for attosecond science" in Light: Science & Applications. Motivated by attosecond generation mechanisms and application needs, the Review systematically surveys the technological landscape and development trends of driving lasers along four main axes—pulse energy, pulse duration, wavelength, and repetition rate—and summarizes the key bottlenecks that must be addressed in the next stage. Xijie Hu and Ka Fai Mak are co–first authors, and Jinwei Zhang is the corresponding author.
Review content:
1. Why do driving lasers set the upper limits of attosecond sources?
Attosecond pulses are currently obtained primarily via strong-field-driven high-harmonic generation (HHG). In essence, an electron is ionized by the intense laser field, accelerated, and then driven back to recollide and recombine with the parent ion, releasing energy as coherent extreme-ultraviolet (XUV) radiation that can support attosecond pulse formation. Consequently, attosecond output is highly sensitive to the driving laser's pulse duration, pulse energy, wavelength, and repetition rate: shorter pulses facilitate the generation of isolated attosecond pulses (IAPs); higher pulse energy increases ionization and can enhance yield/efficiency; longer wavelengths raise the HHG cutoff energy but typically reduce conversion efficiency; and higher repetition rates improve signal-to-noise ratio and acquisition speed, yet are often constrained by limited pulse energy.
Different applications emphasize different key attosecond-source metrics, leading to distinct design trade-offs for the driving laser. For example, in studies of ultrafast dynamics and electron microscopy, robust IAP generation typically requires few-cycle driving pulses with well-controlled carrier–envelope phase (CEP) to enable effective temporal gating and waveform control. In pump–probe spectroscopy and multiphoton ionization experiments, high-energy or high-flux attosecond radiation improves excitation and detection efficiency; achieving this generally relies on HHG driven by higher pulse energy and higher average power, while maintaining acceptable phase matching and beam quality under substantial ionization. To generate attosecond radiation in the X-ray water window—valuable for coherent imaging and time-resolved X-ray absorption spectroscopy—mid-infrared, long-wavelength drivers are often employed to extend the harmonic cutoff and access higher photon energies. In coincidence measurements and photoelectron spectroscopy, where statistical precision is critical, high repetition rates can markedly improve signal-to-noise ratio and data throughput, while lower charge/energy per pulse helps mitigate space-charge effects that degrade energy resolution. The close relationship between driving-laser parameters, attosecond-pulse characteristics, and application requirements is summarized in Fig. 1. Overall, increasingly stringent application demands continue to push attosecond-source performance, in turn driving the development of ultrafast-laser architectures and enabling technologies.
2. Four parameters, four technological routes
1) Higher pulse energy
To overcome the photon-flux bottleneck imposed by the intrinsically low overall conversion efficiency of HHG and to enable high-energy attosecond pulse generation, this Review systematically retraces the evolution of ultrafast driving-laser technologies for attosecond science. The central objective is to simultaneously scale pulse energy and peak power through advanced amplification architectures and coherent/divided-beam combining strategies. Starting from conventional chirped-pulse amplification (CPA), the Review further surveys optical parametric amplification (OPA) and its derivatives, including optical parametric chirped-pulse amplification (OPCPA), dual-chirped OPA (DC-OPA) tailored for broadband mid-infrared amplification, frequency-domain OPA (FOPA), and quasi-phase-matched parametric CPA (QPCPA). It highlights how these approaches mitigate gain narrowing, reduce thermal load, increase damage-threshold margin, and maintain broadband phase controllability. In addition, spatial coherent beam combining (CBC) and temporal divided-pulse amplification (DPA) are discussed as energy-scaling strategies to surpass the physical and engineering limits of single-channel amplifiers—such as thermal effects, nonlinear phase accumulation, and optical damage—thereby providing the driving-laser basis for high-flux attosecond sources in spectroscopy and imaging.
The Review notes that although the CPA architecture has laid the technological foundation for high-energy ultrafast lasers, directly obtaining high-energy, few-cycle mid-infrared pulses remains challenging due to limitations in gain bandwidth, thermal management, and accumulated dispersion/nonlinear phase. As a result, the field has increasingly shifted toward parametric-amplification routes represented by OPA/OPCPA, which combine energy scalability with broadband gain; representative systems have achieved joule-level output energies. To meet the demand for few-cycle mid-infrared drivers, DC-OPA employs dual-chirp management and optimized phase matching to enable broadband, high-energy amplification, with representative demonstrations reporting 53 mJ pulses approaching a single optical cycle. FOPA separates the spectrum into frequency slices and amplifies them in parallel, distributing peak intensity and thermal load across multiple channels to mitigate single-crystal damage and nonlinear limitations. QPCPA, in turn, leverages quasi-phase matching and related schemes to suppress energy back-conversion and improve energy utilization in parametric amplification. Beyond single-channel scaling, the Review also discusses combining strategies such as coherent beam combining (CBC) and divided-pulse amplification (DPA), which enable energy scaling from the millijoule to joule level and peak-power scaling from the terawatt to petawatt regime via coherent addition across multiple channels and temporal energy partitioning, providing viable system pathways toward next-generation attosecond sources for strong-field driving and high-flux HHG.
2) Shorter pulse duration
To generate isolated attosecond pulses (IAPs) capable of resolving ultrafast electron dynamics (e.g., time-resolved X-ray absorption spectroscopy, TR-XAS), the driving laser must overcome the gain-bandwidth limitations of the amplification medium while delivering few-cycle (or even sub-cycle) pulse durations with robust carrier–envelope phase (CEP) stability.
The Review points out that the mainstream approach relies on nonlinear post-compression, in which self-phase modulation broadens the spectrum of high-energy pulses and subsequent compression yields few-femtosecond or even sub-cycle durations. For example, hollow-core fiber (HCF) systems, leveraging the high damage threshold of noble gases and soliton self-compression, have produced the shortest reported pulses to date (~0.9 fs, sub-cycle), but are constrained by the achievable pulse energy (~40 mJ) and practical fiber length. Multi-plate spectral broadening/compression (MPSC) employs segmented focusing to balance self-focusing and diffraction, thereby mitigating solid-state damage; it offers a compact solution but typically at low pulse energies (<1 mJ). Multi-pass cells (MPCs), based on Herriott-type cavity designs, can achieve very high efficiency (>96%) and have demonstrated scaling to kilowatt-class average power and ~150 mJ pulse energy; although dispersion management is more complex, MPCs provide particularly strong overall performance. In addition, coherent pulse synthesis and DC-OPA are also effective routes toward few-cycle pulse generation.
For CEP control, f–2f interferometers (or 0–f interferometers) are required to down-convert optical-frequency information to the radio-frequency domain for measuring CEP drift, combined with active feedback and/or feed-forward stabilization. Implementations include adjusting the oscillator pump or wedge pairs, using an acousto-optic frequency shifter (AOFS) for feed-forward stabilization without intracavity perturbation, and applying an acousto-optic programmable dispersive filter (AOPDF) or fine grating tuning in the amplifier chain. Alternatively, passive all-optical self-stabilization based on difference-frequency generation (DFG) can provide intrinsic CEP stability. Together, these techniques enable high-energy attosecond drivers with precisely controllable waveforms.
3) Longer wavelength
To generate attosecond pulses covering the "water window" (282–533 eV)—enabling chemical-sensitive analysis of biological molecules (e.g., carbon and nitrogen) in aqueous environments—driving-laser technology is shifting from conventional Ti:sapphire systems toward longer-wavelength mid-infrared sources. The Review details three core technological routes toward this goal. First, the most mature approach is optical parametric amplification (OPA) and its cascaded variants: in the 1–5 μm range, broadband amplification is typically realized using oxide crystals such as BiBO and MgO: LiNbO3; beyond 5 μm, non-oxide crystals become essential, including ZGP (requiring ~2 μm pumping) and LiGaS2 (pumpable at ~1 μm). Using these platforms, supercontinua extending into the water-window photon-energy range have been demonstrated. Second, difference-frequency generation (DFG) and intrapulse DFG (IPDFG) provide an alternative route; although the pulse energy is often limited to the microjoule level, the passive carrier–envelope phase (CEP) stability inherent to nonlinear mixing enables highly waveform-controllable seed or driver sources without complex electronic feedback. Finally, considerable attention has focused on direct mid-IR laser technologies based on transition-metal-doped chalcogenides (e.g., Cr:ZnS/Se and Fe:ZnS/Se). Owing to their extremely broad gain bandwidth and favorable thermal properties, these materials are sometimes referred to as "Ti:sapphire in the mid-IR." To date, Cr:ZnS oscillators have directly delivered nearly 2 W average power with pulse durations shorter than three optical cycles and very low phase noise, and are emerging as compact, efficient next-generation attosecond drivers that do not require complex post-amplification stages.
4) Higher repetition rate
To address bottlenecks caused by the low conversion efficiency of HHG (particularly under long-wavelength driving) and by conventional kHz-class drivers—namely insufficient photon flux, long acquisition times, and limited signal-to-noise ratio—the Review notes that although machine learning can assist data reconstruction, developing high-repetition-rate driving lasers is the central route to overcoming space-charge constraints and fundamentally improving the precision of attosecond experiments. At present, two main technological pathways are used to obtain high-repetition-rate attosecond pulses.
The first pathway is resonant enhancement cavities (>10 MHz). In this scheme, phase-stabilized, MHz-rate, low-energy pulses are coupled into a high-finesse passive ring cavity, where constructive interference greatly increases the intracavity peak power to drive HHG. Evolving from early Ti:sapphire implementations to today's Yb-fiber-CPA-driven platforms, this approach has enabled the generation of XUV and EUV frequency combs. However, due to the complexity of output coupling and cumulative plasma effects induced by gas ionization, isolated attosecond pulse (IAP) generation has not yet been experimentally demonstrated in enhancement-cavity HHG, although theoretical work has predicted feasibility using transverse-mode gating (TMG).
The second pathway is direct driving with high-repetition-rate, high-power lasers (~100 kHz to MHz). Compared with conventional Ti:sapphire amplifiers, which are often limited by thermal effects to repetition rates below ~10 kHz, OPCPA systems have shown strong performance and have already demonstrated IAP generation at 100 kHz with microjoule-level pulse energies and pulse durations below 140 as. In parallel, fiber CPA combined with nonlinear post-compression has pushed repetition rates into the MHz regime. In addition, thin-disk oscillators—owing to their compact footprint, low noise, and power-scalability—are promising candidates for next-generation attosecond drivers. Finally, solid-state HHG, which requires substantially lower focusing intensities (down to nanojoule-level pulse energies), offers new possibilities for high-repetition-rate attosecond pulse generation.
Summary and Outlook
As attosecond science expands toward frontier areas such as strong-field physics and bioimaging, driving lasers are rapidly evolving along four key dimensions: pulse energy, pulse duration, wavelength, and repetition rate. This Review summarizes progress in scaling pulse energy via amplification and combining techniques, generating few-cycle waveforms through nonlinear post-compression and CEP locking, extending driver wavelengths into the mid- to far-infrared to access higher photon energies, and developing high-repetition-rate architectures. Looking ahead, it argues that three major "breakthrough directions" and three "hard bottlenecks" will jointly define the roadmap for next-generation attosecond drivers. First, tabletop systems are pushing single-pulse energies toward multi-mJ and beyond, enabling substantially higher peak intensities and photon flux and accelerating the exploration of extreme-field nonlinear interactions in emerging media (including solids and plasmas). Second, longer-wavelength—especially mid-infrared—drivers can, following HHG scaling laws, extend the cutoff into the soft X-ray/water-window range and broaden the "energy and contrast window" for strong-field applications such as molecular orbital imaging. Third, high-average-power ultrafast systems operating at high repetition rates (~100 kHz–MHz) are expected to become essential workhorses for pump–probe experiments as well as coincidence-counting and correlation measurements.
At the same time, three long-standing challenges must be addressed: thermal management and heat load at high average power remain central to stable operation and beam-quality preservation; achieving robust, long-term CEP stability at >10 mJ—particularly in long-wavelength and multi-stage amplification architectures—remains difficult; and the availability and performance of broadband, high-damage-threshold mid-infrared optics (nonlinear crystals, coatings, and compression/dispersion-management components) pose significant constraints, directly limiting energy scaling and system reliability. Overcoming these challenges will not only broaden the impact of attosecond light sources, but also lay the groundwork for exploring the ultimate limits of spatiotemporal resolution, potentially toward zeptosecond lasers.
