Last updated: October 2026 | By Soar Laser
Laser annealing uses a laser beam as the heat source to perform controlled heating of a material surface or thin film on the millisecond down to nanosecond scale — the goal is not to “harden the surface,” but to activate dopants, repair lattice damage, and crystallize amorphous films. From source/drain activation in advanced logic devices and polysilicon channels in 3D NAND, to LTPS low-temperature polysilicon in phone displays, selective emitters in solar cells, and ohmic contacts for SiC / GaN, laser annealing has become an indispensable step in advanced processing. It is most often confused with laser hardening: hardening pursues “rapid cooling,” annealing pursues “controlled temperature,” and their thermal cycles run in almost opposite directions. This article explains its principles, the two dimensions of process classification, key parameters, typical effects, 6 application areas, comparisons with furnace annealing / RTA / flash annealing, and the 7 most common pitfalls.
Key takeaways
- One-sentence definition: Laser annealing is a heat-treatment process that uses a laser as the heat source for controlled heating of a material surface or thin film; the typical goal is not “hardening” but dopant activation, lattice-damage repair, and crystallization of amorphous films.
- Most easily confused with laser hardening: hardening needs rapid quenching (self-quenching to form martensite), annealing needs controlled heating and cooling — the two thermal cycles run in almost opposite directions.
- Understand laser annealing along two dimensions: by mechanism, sub-melt (solid phase, solid-phase epitaxy SPER) versus melt (liquid phase, liquid-phase epitaxy LPE / recrystallization); by time scale, nanosecond / microsecond / millisecond. What engineering calls “millisecond annealing (LSA)” is a mainly solid-phase millisecond scanning process, not a third mechanism alongside solid and liquid phase. Mechanism and time scale together determine junction depth, activation level, and residual defects.
- The key advantage is an extremely low thermal budget: only tens of nanometres to a few micrometres of the surface are heated, while the substrate stays nearly cold — this is the fundamental reason advanced logic devices, 3D memory, and power devices cannot do without it.
- Wavelength determines heating depth: UV (308/355nm) targets the surface, green (532nm) balances surface and shallow layers, near-IR (808–1064nm) heats deeper where absorption permits (note that silicon absorbs 1064nm very weakly at room temperature); choosing the wrong wavelength means choosing the wrong process.
What is laser annealing? A one-sentence explanation
Laser annealing is a heat-treatment process that uses a laser beam as the heat source to heat a material surface or thin film in a controlled way on the millisecond down to nanosecond scale, bringing it to a target temperature and then cooling along a set path to achieve dopant activation, lattice-damage repair, amorphous-to-polycrystalline transformation, or interfacial alloying.
It belongs to the “heat treatment” branch of laser material processing, alongside hardening, cladding, and cutting, but with a completely different purpose: hardening seeks higher surface hardness, annealing seeks repair and optimisation of electrical properties and crystal structure. One sentence to distinguish the two thermal cycles — hardening needs “fast cooling,” annealing needs “controlled temperature”.
Laser annealing’s irreplaceable value lies in its control of the “thermal budget.” Conventional furnace annealing puts the whole wafer in a furnace at over a thousand degrees for tens of minutes; rapid thermal annealing (RTA) compresses this to seconds; laser annealing compresses heating to milliseconds or even nanoseconds, heating only the outermost layer while the substrate barely warms. For a wafer that has already completed front-end steps, “only the surface is hot, the inside is not” means structures already built will not be damaged — the root of its place in advanced processing.
How does laser annealing differ from hardening? Look at the “thermal cycle direction”
The two are often lumped together because both are “laser heat treatments” and both use a laser to heat material. But their thermal-cycle directions are almost opposite, and so are their uses.
1. Hardening: it’s about cooling fast
Laser hardening aims to form martensite in the surface layer of steel by cooling austenite rapidly before it can transform into pearlite. The keyword is therefore “self-quenching” — heat is quickly extracted by the cold bulk, and the cooling rate must exceed the critical cooling rate. See Laser Hardening Encyclopedia.
2. Annealing: it’s about controlling precisely
Laser annealing has the opposite goal — it wants to heat the material to the target temperature and let it cool along the intended path, giving atoms time to return to correct lattice sites (damage repair), letting dopant atoms occupy substitutional sites (activation), and letting amorphous films grow into polycrystalline (crystallization). Note here that mechanism and time scale are two dimensions: the solid-phase route (including sub-melt and millisecond annealing) needs a certain dwell time at high temperature for atoms to reset, and cooling should be smooth and controlled — too fast is bad for defect removal and stress relief; nanosecond melt annealing (ELA / NLA) has no dwell stage — after the surface melts, the unmelted cold substrate acts as a heat sink, solidifying within tens of nanoseconds at cooling rates up to 10⁹–10¹⁰ K/s. So “controlled temperature” in laser annealing means controlled arrival at temperature, cooling dominated by heat diffusion into the substrate, not relying on any critical cooling rate — not “slow cooling.”
3. One table to see the differences
| Dimension | Laser annealing | Laser hardening |
|---|---|---|
| Goal | Dopant activation, damage repair, crystallization, alloying | Surface phase-transformation hardening, higher hardness and wear resistance |
| Thermal cycle | Controlled heating + diffusion cooling (solid-phase processes including millisecond annealing may have a brief constant-temperature stage; nanosecond melt processes have no dwell stage and solidification cooling can still reach ~10⁹ K/s; does not depend on a critical cooling rate) | Rapid quenching (self-quenching, cooling rate must exceed the critical rate) |
| Typical temperature | 600–1400 °C (process-dependent, may exceed the melting point) | 850–1000 °C (above the transformation point, below the melting point) |
| Does it melt? | Depends on the mode: sub-melt does not melt, melt mode deliberately melts | Conventional process does not melt (melting means parameters are over the limit) |
| Depth of action | Tens of nanometres to a few micrometres (thin film / surface layer) | 0.1–2.5 mm (bulk material surface layer) |
| Main targets | Silicon, SiC, GaN, oxide semiconductors, metal films | Medium-carbon steel and above, tool steel, cast iron |
| Typical industries | Semiconductors, displays, photovoltaics, power devices | Molds, guide rails, gears, shafts |
Process classification: how to distinguish sub-melt, melt, and millisecond annealing
Laser annealing can be implemented in many ways; classification should use two orthogonal dimensions and not mix them: by mechanism (whether it melts) into solid phase (sub-melt, solid-phase epitaxy SPER) and liquid phase (melt, liquid-phase epitaxy LPE / recrystallization); by heating time scale into nanosecond, microsecond, and millisecond. “Sub-melt / melt / millisecond annealing (LSA)” below are three common engineering names — the first two describe mechanisms, while LSA describes a millisecond-dwell, scanning time and equipment form; it is itself mainly solid-phase (SPER), not a third mechanism parallel to solid and liquid phase.
| Mode | Heating time | Mechanism | Junction / action depth | Features and suitability |
|---|---|---|---|---|
| Sub-melt (solid phase) | Nanosecond–millisecond | Temperature below the melting point; atoms rearrange in the solid phase; solid-phase epitaxial regrowth | Tens to hundreds of nanometres | Balances damage repair and activation, good surface roughness, no melt-induced morphology change; suitable for ultra-shallow junctions |
| Melt (liquid-phase epitaxy) | Nanosecond–microsecond | Surface melts instantly and recrystallizes by liquid-phase epitaxy using the unmelted substrate as seed | Within tens of nanometres (controllable) | Activation can exceed the equilibrium solid solubility (supersaturation); extremely shallow junction, suitable for source/drain and contact regions |
| Millisecond annealing (LSA) | Millisecond scale | Solid-phase (SPER): scanning implementation with millisecond dwell, does not cross the melting point, may be locally sub-melt; not an independent third mechanism | Tens to hundreds of nanometres | Well matched to production-line takt, scanning process can be scaled, mainstream choice for advanced logic devices |
| Flash annealing (FLA) | Millisecond scale (non-laser) | Broad-spectrum xenon-lamp flash heating | Hundreds of micrometres to whole wafer | Often compared with laser annealing: whole-wafer heating, high throughput, but selectivity and depth control inferior to laser |
How to choose? Three decision steps:
- First, what problem are you solving: dopant activation only → millisecond sub-melt is often enough; ultra-shallow junction + very high activation → nanosecond melt mode;
- Then, how much thermal budget the device can tolerate: the finer the device structure and the closer the metal layers to the surface, the more you need a shorter time scale;
- Finally, production-line takt: scanning millisecond annealing (LSA) matches mass-production rhythm more easily than whole-wafer nanosecond processing.
Key parameters: wavelength, power density, pulse width, and spot
Laser-annealing quality is jointly determined by these parameter groups, which also constrain one another.
| Parameter | What it controls | Key points and typical ranges |
|---|---|---|
| Wavelength | Determines how deep the laser penetrates | UV 308/355nm: absorption depth in silicon only a few nanometres, energy concentrated at the very surface; green 532nm: on the order of hundreds of nanometres, balancing surface and shallow layers; near-IR 808–1064nm: significantly greater absorption depth, but be careful in silicon — silicon’s bandgap is about 1.12 eV and a 1064nm photon carries only about 1.17 eV, so room-temperature absorption is very weak (about 10 cm⁻¹, absorption depth on the millimetre scale), energy is hard to couple; typically one raises energy density, exploits the sharp increase in absorption at high temperature, or uses it for metal / amorphous-silicon films; wide-bandgap materials such as SiC / GaN cannot absorb 1064nm in a single photon either. Note: absorption depth varies with material, temperature, and doping; the above are order-of-magnitude guides — confirm by measurement. |
| Power density | Determines how hot the surface gets | Nanosecond mode is typically 10⁶–10⁸ W/cm² (energy density about 0.1–1 J/cm², pulse width 10–100 ns, melting threshold about 10⁶–10⁷ W/cm²); millisecond mode is on the order of 10⁴–10⁶ W/cm² — the goal is “reach temperature but stay in control.” |
| Pulse width / dwell time | Determines how far heat diffuses | The shorter the time, the more heat is confined near the surface. Heat-diffusion length L≈√(D·t): in silicon (thermal diffusivity D about 0.12 solid / 0.23 molten cm²/s at high temperature), nanosecond scale is about sub-micron to 1 µm, millisecond scale reaches tens to hundreds of micrometres; only low-diffusivity materials such as glass are as short as hundreds of nanometres / tens of micrometres. Note the distinction between diffusion length and actual melt depth: UV ELA melt depth can be limited to tens of nanometres, governed mainly by optical absorption depth (only a few nanometres in UV) and the amorphous-silicon film thickness, not by diffusion length. |
| Spot shape | Determines processing uniformity | Line spots (e.g., long strips) suit scanning crystallization and whole-wafer annealing; square spots facilitate fast production-line scanning; combined with galvo scanners or translation stages for area scanning. |
| Overlap ratio | Determines in-plane consistency | Scanning processes must control overlap, otherwise periodic non-uniformity (“stripes”) appears; typically determined through process trials. |
| Atmosphere / preheating | Determines surface condition | Inert atmosphere prevents oxidation; substrate preheating reduces required laser energy density and thermal shock. |
These parameters relate as follows: wavelength determines “which layer the heat lands in,” power density and pulse width determine “how hot that layer gets and for how long,” spot and overlap determine “whether the whole area is uniform.” Tuning is essentially finding a stable window between “underheating (insufficient activation, residual damage)” and “overheating (runaway melting, interface damage, worsened roughness).”
What can laser annealing achieve? Typical effect ranges
The ranges below are for setting expectations; exact values vary with material, device structure, and equipment conditions, with test wafers plus electrical / structural characterisation as the final basis:
| Goal | Typical effect | Common mode |
|---|---|---|
| Dopant activation | Activation can exceed the equilibrium solid solubility (especially in melt mode), significantly lowering sheet resistance | Melt / sub-melt |
| Ultra-shallow junction formation | Junction depth can be controlled within tens of nanometres, suppressing short-channel effects | Nanosecond melt |
| Implant-damage repair | Implant-induced amorphous layers regrow by solid- or liquid-phase epitaxy, drastically reducing defect density | Sub-melt / melt |
| Amorphous-silicon crystallization | Amorphous-silicon film turns into polysilicon; grain size can be tuned by energy density (including the super-lateral-growth regime) | Excimer (ELA) |
| Interfacial alloying | Forms low-resistance ohmic contacts at lower substrate temperatures, avoiding overall thermal damage from high-temperature furnace steps | Millisecond / nanosecond |
| Grain improvement | Grain size and activation of the polysilicon channel in memory devices improve together | Millisecond (LTA) |
A few empirical rules:
- Shorter time means closer to the surface: the action depth of nanosecond processes is naturally limited by both optical absorption depth and heat diffusion — an advantage (no damage to underlying layers) and a limitation (difficult to activate deep implanted layers);
- Melt mode gives the strongest activation but the highest morphological cost: melt-regrowth enables supersaturated activation, but surface roughness and interface abruptness need extra control;
- Sub-melt is gentler: no melting and good morphology preservation, but a lower activation ceiling than melt mode;
- “Enough is enough”: activation targets are set by device electrical requirements; blindly chasing extreme activation often brings leakage and reliability risks.
Which industries use laser annealing? 6 typical applications
| Application area | Typical targets | Why laser annealing |
|---|---|---|
| Advanced logic devices | Source/drain dopant activation, ultra-shallow junctions, contact regions | Extremely low thermal budget; almost no heating outside the implanted layer; activation can be done at a later process stage |
| Memory devices | Activation and grain improvement of 3D NAND polysilicon channels, peripheral circuits | Three-dimensional structures stacked layer by layer, only laser can selectively heat a specific layer; furnace annealing would destroy completed layers |
| Display panels | Excimer laser annealing (ELA) of LTPS low-temperature polysilicon | Obtains high-mobility polysilicon on a low-temperature substrate — a key step in mass-producing high-end phone displays and AMOLED |
| Solar cells | Selective-emitter (SE) laser doping, local contact optimisation | Heavy doping in the metal-grid contact regions lowers contact resistance and improves conversion efficiency |
| Power and compound semiconductors | Ohmic-contact annealing and ion-implant activation of SiC / GaN | Conventional approaches need thousands of degrees for long anneals; laser accomplishes contact alloying with a low thermal budget |
| Thin films and micro/nano fabrication | Metal-film modification, oxide-semiconductor annealing, sensor films | Local selective heating, compatible with flexible substrates and low-temperature processes |
The common thread is four keywords: low thermal budget, selectivity, locality, and production-line compatibility. Especially since three-dimensional stacked structures appeared (3D NAND, 3D-stacked CMOS), “heating only the layer I want heated” has shifted from a bonus to a requirement.
How to choose among furnace annealing, RTA, and flash annealing?
| Process | Heating method | Time scale | Thermal budget | Advantages and limits |
|---|---|---|---|---|
| Furnace annealing | Whole wafer in furnace, resistive heating | Tens of minutes to hours | Very high | Good uniformity, large batches, low cost; limit: the entire piece is heated, later-stage devices cannot tolerate it |
| Rapid thermal annealing (RTA / spike) | Lamp irradiation of whole wafer | Seconds | High | Mature and stable, highly compatible with production lines; limit: whole-wafer heating, hard to achieve ultra-shallow junctions |
| Laser annealing (LSA / ELA etc.) | Local / scanning laser heating | Millisecond–nanosecond | Very low | Good selectivity, ultra-shallow junctions, usable at later process stages; limit: high equipment investment, in-plane uniformity and takt require dedicated design |
| Flash annealing (FLA) | Broad-spectrum xenon-lamp flash over whole wafer | Millisecond scale | Low (but whole-wafer) | High throughput, whole-wafer processing; limit: broad spectrum, depth selectivity inferior to laser, sensitive to patterned structures |
One-line selection: need throughput, uniformity, and a device that tolerates high temperature → furnace / RTA; need ultra-shallow junctions, low thermal budget, and activation at later stages → laser annealing. In real production lines, laser annealing is often combined with RTA — RTA performs most of the activation first, then laser does surface supplemental activation and junction-depth trimming.
How to configure the equipment? Light source, optical path, and line integration
- Light-source selection (derive wavelength from the application): display-panel ELA almost exclusively uses excimer lasers (XeCl 308nm, line beam); semiconductor activation commonly uses UV solid-state lasers (355nm) and green lasers (532nm); when deeper heating is needed, choose near-IR fibre / solid-state lasers (808–1064nm); see UV Laser Encyclopedia and Solid-State Laser Encyclopedia for principles.
- Spot shaping: line spots use cylindrical lens groups or beam-shaping elements; area scanning uses galvo scanners or high-precision stages, with F-theta lenses ensuring in-plane spot consistency.
- Energy and uniformity monitoring: energy density is the most critical process variable in laser annealing and needs online monitoring with closed-loop control; in-plane uniformity relies on spot shaping and scan-path design (overlap ratio, overlap pattern).
- Optical path and windows: UV and deep-UV place high demands on optics, requiring high-transmission materials such as fused silica; protective windows and optical windows have coatings customized to the wavelength and cannot be interchanged.
- Atmosphere and thermal management: most processes run in inert atmosphere or vacuum to prevent oxidation; equipment thermal management (chillers, cabinet air conditioners) follows the same thinking as laser hardening systems.
7 common myths about laser annealing
- Myth 1: Laser annealing and laser hardening are the same thing. Not at all. Hardening pursues rapid quenching to form martensite; annealing pursues controlled heating and cooling to repair the lattice and activate dopants — the thermal cycles run in almost opposite directions.
- Myth 2: Annealing means “burning the material soft.” “Anneal” here is a heat-treatment term: it uses a controlled thermal history to change a material’s structure and electrical properties, not to soften it.
- Myth 3: The higher the energy density, the better. Too high causes runaway melting, worsened surface roughness, interface abruptness, and higher leakage. Energy density must sit within the window and be paired with online monitoring.
- Myth 4: Laser annealing is only for semiconductors. Displays (LTPS / ELA), photovoltaics (SE laser doping), power devices (SiC / GaN ohmic contacts), and flexible electronics all use it, and many are already large-scale production processes.
- Myth 5: Any wavelength works, it’s all just heating. Wavelength directly determines absorption depth — a few nanometres in UV, hundreds of nanometres to about a micron in green, deeper in near-IR (but silicon absorbs 1064nm very weakly at room temperature, so energy is hard to couple, and deeper is not automatically better). Wrong wavelength means either heat cannot get in, or the wrong place gets heated.
- Myth 6: Laser annealing can fully replace furnace annealing / RTA. It is not a replacement relationship. Furnace and RTA still win on throughput, uniformity, and cost; laser annealing solves low thermal budget and selectivity. Real production lines mostly combine them.
- Myth 7: No need to worry after annealing. Quite the opposite — activation level, junction depth, surface roughness, and stress state all need characterisation; some dopant systems also suffer deactivation in later process steps, which must be accounted for in process design.
Laser annealing FAQ
What is the difference between laser annealing and laser hardening?
Their goals and thermal cycles differ: hardening forms martensite in the steel surface by “rapid quenching” to raise hardness; annealing achieves dopant activation, damage repair, and crystallization through “controlled heating and cooling.” Hardening is judged by cooling rate, annealing by temperature-control precision.
What can laser annealing do, and where is it mainly used?
Four main things: dopant activation, lattice-damage repair, amorphous-film crystallization, and interfacial alloying. Typical scenarios include source/drain activation and ultra-shallow junctions in advanced logic devices, channel annealing in 3D memory, LTPS excimer laser annealing for display panels, selective-emitter laser doping in photovoltaics, and ohmic-contact annealing for SiC / GaN.
Why must laser be used — isn’t furnace annealing enough?
The key difference is “thermal budget.” Furnace annealing and RTA heat the whole wafer (or whole piece of material), which already-built fine structures cannot tolerate; laser heats only tens of nanometres to a few micrometres of the surface, with the substrate barely warming, so it can be used at later process stages.
What wavelength should laser annealing use?
Work backward from the depth you need to heat: UV (308 / 355nm) concentrates heating at the very surface, green (532nm) covers surface and shallow layers, near-IR (808–1064nm) has greater absorption depth and can heat deeper — but note for silicon — silicon absorbs 1064nm very weakly at room temperature and energy is hard to couple (see the wavelength parameters above). Also check the transmission of optical components and coating compatibility at the chosen wavelength.
Does laser annealing melt the surface?
It depends on the mode. Sub-melt mode does not melt, relying on solid-phase rearrangement; melt mode deliberately melts the surface and then recrystallizes by liquid-phase epitaxy using the substrate as seed, for higher activation. Each mode suits different cases — “melting” is not automatically bad.
Are other process steps needed after laser annealing?
Usually yes. Activation level, junction depth, and morphology need characterisation to confirm; many production lines combine laser annealing with RTA and watch for dopant deactivation and stress issues in later steps.
Related products and further reading
Soar Laser supplies optical and thermal-management components for laser-annealing processes: UV solid-state lasers (355nm series), 532nm green solid-state lasers, high-speed galvo scanners and F-Theta lenses, plus chillers for UV laser equipment; optics (fused-silica windows, protective windows, coatings) can be customized by wavelength and energy density, with selection support for your conditions. For more system knowledge, see:
- What is Laser Material Processing? The Optical Processing Revolution in Manufacturing
- Laser Hardening Encyclopedia: Principles, Parameter Window, and Full Comparison
- Laser Principles Encyclopedia: Stimulated Emission and Laser Structure Fundamentals
- UV Laser Encyclopedia / Solid-State Laser Encyclopedia
- Fibre Laser Encyclopedia / Semiconductor Laser Encyclopedia
- Beam Shaping Encyclopedia / Galvo Scanner Encyclopedia / F-theta Lens Encyclopedia
- Laser Protective Window Encyclopedia / Optical Window Encyclopedia
- Optical Coatings & Thin-Film Technology / Fused Silica Encyclopedia
- The Hidden Adversary of Laser Processing: A Complete Guide to Thermal-Effect Control
- Laser Encyclopedia Index
References: F. Cristiano et al., Defect evolution and dopant activation in laser annealed Si (Materials Science in Semiconductor Processing); Mechanistic benefits of millisecond annealing for diffusion and activation of boron in silicon (Journal of Applied Physics, 105, 063514); Dopant Activation of In Situ Phosphorus-Doped Silicon Using Multi-Pulse Nanosecond Laser Annealing (Physica Status Solidi A); Activation Enhancement and Grain Size Improvement for Poly-Si Channel Vertical Transistor by Laser Thermal Annealing in 3D NAND Flash (Micromachines, 14, 230); Silicon and Silicon Carbide Recrystallization by Laser Annealing: A Review (Materials, 16, 7674); Coherent: Lasers in Display Fabrication: Excimer Laser Annealing; Laser Annealing of Ion Implanted Silicon (AIP Conference Proceedings, 50, 275); Thermal Conductivity of Silicon from 300 to 1400 K (Physical Review, 130, 1743); Thermal Diffusivity of Crystalline and Liquid Silicon and an Anomaly at Melting (Japanese Journal of Applied Physics, 30, 2423).



Leave a Reply