Laser-Assisted Etching (LAE) Overview

• Laser-Assisted Etching (LAE) is a family of laser-enabled processes that precisely modify material regions for subsequent selective wet etching.
• LAE techniques vary by activation method—from direct photochemical reactions in silicon to femtosecond and picosecond patterning in glass—each with unique etch selectivity and trade-offs.
• Applications include micro-optics, device fabrication, and semiconductor processing, leveraging controlled optical localization to drive differential etching.

Laser-Assisted Etching (LAE) denotes a family of laser-enabled material-removal processes in which irradiation either directly activates etching chemistry at the illuminated site or creates a latent, selectively etchable volume that is removed in a subsequent wet-chemical step. In the literature represented here, LAE includes electrodeless photochemical dissolution of silicon in HF under laser illumination, femtosecond- and picosecond-laser-written selective wet etching of fused silica and related glasses, laser-induced deep etching of internally modified fused silica, and laser-enabled selective wet etching of previously modified crystalline silicon. The same corpus also includes broader boundary cases—direct laser micro-etching, laser-ablation-based thinning, and interlayer-mediated lift-off—that are often discussed alongside LAE because they solve closely related selectivity and damage-control problems (Saxena et al., 2014, Li et al., 2018, Casamenti et al., 2021, Wimmer et al., 2024, Borra et al., 2023, Dey et al., 22 Apr 2025).

1. Terminology and process scope

The term LAE is not used uniformly. In silicon, "laser-induced etching" (LIE) is presented as an electrodeless photochemical process in which laser illumination generates the carriers required for localized dissolution in HF; in the terminology of the present topic, that LIE is explicitly a laser-assisted etching variant (Saxena et al., 2014). In fused silica and related glasses, the dominant form is a two-step process: ultrafast-laser exposure first writes a modified zone inside the transparent substrate, and a later wet etch preferentially removes that modified region. Papers in this class use labels such as femtosecond laser-assisted chemical etching, selective laser etching (SLE), femtosecond selective laser etching, and Laser-Induced Deep Etching (LIDE) (Li et al., 2018, Casamenti et al., 2021, Casamenti et al., 2024, Wimmer et al., 2024, Bang et al., 21 Oct 2025).

This literature therefore separates naturally into three process classes. The first is simultaneous photochemical LAE, in which the laser is present during etching and locally supplies carriers or otherwise activates dissolution; the n-Si/HF LIE process is the clearest example (Saxena et al., 2014). The second is laser modification followed by selective wet etching, which dominates bulk-glass and laser-modified-silicon work (Li et al., 2018, Borra et al., 2023). The third is a broader usage of “etching” for highly localized laser-driven removal, including photonic-jet machining of metals and semiconductors, direct near-IR etching of $\beta$-Ga$_2$O$_3$, Raman-laser-induced thinning of graphene, and ultrafast thin-film lift-off; these studies are closely related in application space but are explicitly described as ablation, multi-photon-absorption-driven removal, local burning, or lift-off rather than conventional wet-chemical LAE (Pierron et al., 2018, Shuvro et al., 2023, Piazzi et al., 2012, Kim et al., 2020).

A plausible implication is that LAE is best understood not as a single mechanism but as a process family organized by where the laser enters the removal chain: at the moment of dissolution, at the stage of writing etch selectivity, or at the boundary between etching and ablation.

2. Electrodeless photochemical LAE in silicon

A direct and comparatively simple LAE implementation is the electrodeless laser-induced etching of porous silicon from commercially available n-type Si(100) wafers with resistivity $3$–$5\ \Omega\cdot\text{cm}$ in $40\%$ HF, with no electrical bias and with etching activated by irradiation from an argon-ion laser of photon energy $2.41\ \text{eV}$ at a reported power density of $1.76\ \text{kW/cm}^2$ (Saxena et al., 2014). The wafers were cleaned in acetone and ethanol, mounted on two Teflon plates, immersed in HF in a plastic vessel, and irradiated while immersed. The etched region was controlled by the laser spot, giving an LIE sample area around $\sim 100\ \mu\text{m}$, and two etch durations were used: $45\ \text{min}$ and $_2$0. A defining experimental control was explicit: without laser illumination, no etching occurs and porosification is not possible (Saxena et al., 2014).

The mechanism is photochemical carrier-assisted dissolution. Irradiation of the n-type crystalline Si substrate in HF generates photoexcited holes at the illuminated surface, and those holes initiate local silicon dissolution. In compact conceptual form, the sequence may be written as

$_2$1

followed by fluoride-assisted dissolution often summarized as

$_2$2

The paper itself emphasizes not reaction balancing but the role of photogenerated holes as the indispensable trigger (Saxena et al., 2014).

The morphology is governed by the optical intensity profile. With a Gaussian beam,

$_2$3

the local carrier-generation rate and hence the local etch rate vary with $_2$4. The maximum intensity at the beam center gives the highest etching rate, while the beam periphery etches more slowly. The reported consequences are central to silicon LAE: pore-size variation across the irradiated spot, broader pores than in metal-induced etching (MIE), and substantial lateral etching in addition to vertical advance (Saxena et al., 2014). Top-view SEM showed porous Si with characteristic pore size around $_2$5 for both $_2$6 and $_2$7 min etches, while the longer-etched sample showed wider and more connected pores. Because the irradiated area was small, cross-sectional SEM of the LIE sample was not obtained, which itself illustrates a recurrent LAE trade-off between spatial selectivity and ease of structural metrology (Saxena et al., 2014).

The same comparison establishes the strengths and weaknesses of this LAE mode. Relative to MIE, LIE is electrodeless, confined to the illuminated area, and suitable for localized porous-Si patterning. Relative to MIE, it also yields broader, less uniform, more laterally expanded pores because the optical beam profile directly imprints the etch morphology. The paper further notes that earlier LIE work showed a double photoluminescence peak, whereas the MIE sample measured here showed a single broad visible peak centered at $_2$8 with width about $_2$9; the authors interpret that contrast as evidence that LIE tends to produce multiple dominant nanostructure sizes, again consistent with Gaussian-intensity-driven nonuniformity (Saxena et al., 2014).

3. Ultrafast selective wet etching in fused silica and related glasses

In fused silica, the most developed LAE regime is post-laser selective wet etching. One important bottleneck in the conventional femtosecond regime is strong polarization sensitivity: self-organized nanogratings form perpendicular to the laser polarization and govern etchant transport, so channel etchability depends on the angle between writing direction and polarization (Li et al., 2018). A systematic study with a $_3$0 source tunable from $_3$1 to $_3$2, focused $_3$3 below the surface of Corning 7980 fused silica, showed that etching in $_3$4 KOH at $_3$5 under ultrasonic agitation changes qualitatively across the pulse-duration axis (Li et al., 2018). At $_3$6 average power, the reported rates were strongly anisotropic at short duration—e.g., at $_3$7, $_3$8, $_3$9, and $3$0—but became nearly polarization-insensitive in the $3$1–$3$2 range. At $3$3 the rate ratio was approximately $3$4, with $3$5, $3$6, and $3$7 (Li et al., 2018).

The SEM-based interpretation is mechanistic rather than merely empirical. In the $3$8–$3$9 interval, nanogratings dominate or coexist with random nanocracks; above about $5\ \Omega\cdot\text{cm}$0, randomly oriented interconnected nanocracks prevail. The conventional femtosecond LAE picture—ordered nanogratings as anisotropic transport pathways—thus gives way to a nanocrack-dominated regime that preserves high etchability while removing most polarization dependence (Li et al., 2018). This directly explains why triangular-loop microchannels etched uniformly in the $5\ \Omega\cdot\text{cm}$1–$5\ \Omega\cdot\text{cm}$2 regime but not in the shorter-pulse regime.

A second fused-silica study shifted attention from pulse duration to net exposure dose and found that the highest etching efficiency need not coincide with fully developed nanogratings (Casamenti et al., 2021). Using a $5\ \Omega\cdot\text{cm}$3, $5\ \Omega\cdot\text{cm}$4, $5\ \Omega\cdot\text{cm}$5 Yb-fiber amplifier focused with $5\ \Omega\cdot\text{cm}$6 to a measured waist $5\ \Omega\cdot\text{cm}$7 inside Corning 7980 0F fused silica, the dose was defined as

$5\ \Omega\cdot\text{cm}$8

With pulse energies $5\ \Omega\cdot\text{cm}$9–$40\%$0 and scan speeds $40\%$1–$40\%$2, the study identified a low-dose optimum around $40\%$3, corresponding to only a few overlapping pulses—around ten—in which etching in hydroxides was strongest even though clear nanogratings were absent (Casamenti et al., 2021). Under these conditions, NaOH at $40\%$4 and $40\%$5 gave a modified-zone etch rate $40\%$6, approximately $40\%$7 KOH and $40\%$8 HF, while pristine fused silica etched at only $40\%$9. Tunnel aspect ratio approached $2.41\ \text{eV}$0 in NaOH, compared with less than $2.41\ \text{eV}$1 in KOH and less than $2.41\ \text{eV}$2 in HF (Casamenti et al., 2021).

Mechanistically, that work attributes the low-dose etch maximum to precursor defects in the glass matrix, notably non-bridging oxygen hole centers, oxygen deficiency centers, and E' centers, rather than to mature nanogratings. Annealing at $2.41\ \text{eV}$3 nearly eliminated the low-dose peak in KOH and NaOH, whereas high-dose behavior persisted to much higher annealing temperatures, supporting a defect-dominated low-dose regime and a porosity/nanograting-dominated high-dose regime (Casamenti et al., 2021). This suggests that “more dose” is not a universal route to better LAE; in fused silica, the optimal chemically privileged state can precede the familiar nanograting morphology.

4. Optical shaping, precision structuring, and device-level glass implementations

LAE in glass has moved well beyond proof-of-principle porosification or simple channel writing. In micro-optics, laser-assisted wet etching followed by light $2.41\ \text{eV}$4-laser polishing has been used to fabricate $2.41\ \text{eV}$5-diameter fused-silica micro-axicons with apex angle $2.41\ \text{eV}$6 and wedge angle near $2.41\ \text{eV}$7 (Skora et al., 2021). The contour was written with a $2.41\ \text{eV}$8 Yb:YAG femtosecond source at $2.41\ \text{eV}$9 and $1.76\ \text{kW/cm}^2$0 through a $1.76\ \text{kW/cm}^2$1 objective, using $1.76\ \text{kW/cm}^2$2 line spacing and access structures to feed the etchant. Wet etching in $1.76\ \text{kW/cm}^2$3 KOH at $1.76\ \text{kW/cm}^2$4 for about $1.76\ \text{kW/cm}^2$5 released the structure; the as-etched representative component had $1.76\ \text{kW/cm}^2$6, tip radius $1.76\ \text{kW/cm}^2$7, and roughness $1.76\ \text{kW/cm}^2$8. After $1.76\ \text{kW/cm}^2$9 polishing, $\sim 100\ \mu\text{m}$0 fell to $\sim 100\ \mu\text{m}$1, the mean tip radius was nearly $\sim 100\ \mu\text{m}$2, and the resulting quasi-Bessel beam had about $\sim 100\ \mu\text{m}$3 diameter over almost $\sim 100\ \mu\text{m}$4 (Skora et al., 2021). The process is significant because LAE generates the shape with much lower initial roughness than ablation, so the thermal finishing step can remain shallow enough to preserve the conical singularity comparatively well.

At larger scale, femtosecond SLE has been transferred to Corning ULE glass for optomechanical manufacturing (Casamenti et al., 2024). Although the paper is intentionally sparse on laser and etch parameters, it demonstrates that ULE supports the same broad SLE logic used in fused silica: buried ultrafast-laser modification followed by preferential removal of the modified volume. The reported devices include a $\sim 100\ \mu\text{m}$5-thick ferrule with $\sim 100\ \mu\text{m}$6 vertical holes on a honeycomb pitch of $\sim 100\ \mu\text{m}$7, hole diameter centered at $\sim 100\ \mu\text{m}$8 with Gaussian standard deviation $\sim 100\ \mu\text{m}$9 and tolerance $45\ \text{min}$0, and hole-position errors following a Rayleigh law with scale parameter $45\ \text{min}$1 so that about $45\ \text{min}$2 of offsets are below $45\ \text{min}$3 over a $45\ \text{min}$4 array length (Casamenti et al., 2024). A $45\ \text{min}$5-groove V-groove array gave pitch $45\ \text{min}$6 and angle $45\ \text{min}$7, while a passive optical alignment substrate maintained deviations below $45\ \text{min}$8 over lengths up to $45\ \text{min}$9 (Casamenti et al., 2024). This establishes LAE not only as a microchannel process but as a route to thermally stable monolithic hardware.

A different precision strategy appears in fused-silica micro-hole machining via spherical-aberration-assisted filamentation combined with LIDE (Bang et al., 21 Oct 2025). Here the femtosecond laser does not directly excavate the final hole; rather, a deliberately aberrated focus writes a millimeter-scale, micrometer-thin internal modified path that is later developed in hot KOH. The machining setup used a $_2$00 fiber-amplified femtosecond laser with a reported pulse-duration inconsistency—$_2$01 in the abstract and figure overview, $_2$02 in the detailed methods—at $_2$03 and $_2$04, with a $_2$05 objective and a $_2$06 plano-convex lens introducing longitudinal spherical aberration of approximately $_2$07 (Bang et al., 21 Oct 2025). Etching in $_2$08 KOH at $_2$09 for $_2$10 under $_2$11, $_2$12 ultrasonics yielded through-holes in $_2$13 fused silica with final diameters approximately $_2$14, $_2$15, $_2$16, $_2$17, and $_2$18, an etch bias of about $_2$19–$_2$20 relative to scan diameter, taper below the SEM detection limit corresponding to effectively $_2$21, etched-sidewall roughness $_2$22 and $_2$23, and about $_2$24 hole-area variation across $_2$25 holes with $_2$26 failure rate (Bang et al., 21 Oct 2025). This suggests that optical aberration engineering can be used as a first-order control variable for downstream etch fidelity.

Display-glass separation extends the same principle from micro-features to edge engineering (Wimmer et al., 2024). Ultrashort laser pulses from a TRUMPF TruMicro Series 2000 in burst mode were shaped by an LCoS-based holographic 3D beam splitter and focused through an $_2$27 long-working-distance objective into Corning Gorilla glass, creating type III modifications arranged as “pearls on a string.” By feed superposition these discrete modified volumes became a full 3D etchable boundary, later opened in $_2$28 KOH in a heated bath (Wimmer et al., 2024). The etch model used a base-glass rate of approximately $_2$29 and selectivity $_2$30 for modified regions. For a $_2$31 sample with double chamfers, measured chamfer angles were $_2$32 and $_2$33, surface roughness was $_2$34, and the standard deviation between simulation and experiment in the central region was about $_2$35 (Wimmer et al., 2024). The paper’s practical contribution is that the optical trajectory alone is insufficient: auxiliary assist lines are mandatory when reagent access and part release would otherwise be blocked.

5. Silicon post-laser selective wet etching and emerging material systems

Silicon also appears in a post-laser selective-etch variant distinct from electrodeless LIE. In this approach, a nanosecond near-IR laser first modifies crystalline Si subsurface regions, and a subsequent wet etch removes the modified material much faster than pristine Si (Borra et al., 2023). The reported laser conditions were $_2$36, approximately $_2$37, $_2$38, pulse energy $_2$39 in most experiments, and $_2$40 focusing. The key fabrication advance is a chromium-free Cu(NO$_2$41)$_2$42-based etchant optimized from MEMC chemistry:

$_2$43

with $_2$44, all at room temperature (Borra et al., 2023). The paper reports selectivity greater than $_2$45 for laser-modified Si relative to unmodified Si and etch-pit sizes of about $_2$46–$_2$47 for the champion formulation. It further demonstrates micro-pillar arrays with lateral sizes down to about $_2$48, depth up to $_2$49, and aspect ratio greater than $_2$50, while the modified region itself was characterized as a mixture of amorphous and locally ordered polycrystalline or nanocrystalline domains (Borra et al., 2023). This is a different selectivity paradigm from porous-silicon LIE: the laser is not the in-situ source of reaction carriers but the writer of a buried chemically privileged phase mixture.

A true simultaneous laser-assisted wet-etch process appears in amorphous Sb$_2$51S$_2$52 thin films (Dey et al., 22 Apr 2025). The film, about $_2$53 thick, was prepared from an Sb-BDCA precursor and then patterned while immersed in BDCA, which served both as etchant and immersion medium. A Nanoscribe Photonic Professional GT2 with a $_2$54, $_2$55–$_2$56, $_2$57 erbium-doped fiber laser and a $_2$58, $_2$59 objective was scanned directly over the film in liquid, producing positive-tone removal of the exposed regions (Dey et al., 22 Apr 2025). On $_2$60-Sb$_2$61S$_2$62/SiO$_2$63, the smallest linewidth fabricated was $_2$64, about $_2$65 of the stated Abbe limit of roughly $_2$66; full penetration occurred at $_2$67 on SiO$_2$68 and at $_2$69 on Au (Dey et al., 22 Apr 2025). Finite-difference time-domain calculations explained that difference through substrate-dependent standing-wave and interfacial-field effects: gold increases the interfacial intensity and lowers threshold but broadens the effective width relative to quartz. This is one of the clearest examples in the supplied literature of a genuine simultaneous laser-plus-liquid etch process outside silicon.

The same broad family extends into harder semiconductor and thin-film systems, although with different degrees of chemical participation. In bulk $_2$70-Ga$_2$71O$_2$72, a near-IR laser with spot size $_2$73 was used to form deep backside microchannels at etch rates as high as $_2$74 through a process explicitly attributed to multi-photon absorption rather than to in-situ etchant chemistry (Shuvro et al., 2023). The most reliable geometry values in the paper are trench opening $_2$75 and depth $_2$76, and the resulting embedded cooling reduced surface temperature from $_2$77 to $_2$78 at $_2$79 and $_2$80 water flow (Shuvro et al., 2023). In few-layer graphene on Cu, repeated Raman exposures with a $_2$81 He-Cd laser at $_2$82 for $_2$83 on a $_2$84 spot caused a progressive narrowing and redshift of the $_2$85 band and a decrease in $_2$86, interpreted as local thinning of the upper graphene layers, while the $_2$87 peak increased, indicating that the process is thinning with defect generation rather than benign layer-count refinement (Piazzi et al., 2012). These examples sit at the edge of strict LAE terminology but are relevant because they show how laser-enabled selectivity can be created in materials that are otherwise difficult to etch or thin controllably.

6. Mechanistic themes, recurrent trade-offs, and common misconceptions

Across these studies, the laser does not play a single invariant role. In n-Si/HF LIE it provides the photogenerated holes required for dissolution and spatially confines the reaction to the illuminated spot (Saxena et al., 2014). In fused silica and ULE it writes a modified topology—defects, nanogratings, nanocracks, or type III damage—that later biases wet chemistry by orders of magnitude (Li et al., 2018, Casamenti et al., 2021, Casamenti et al., 2024, Wimmer et al., 2024). In Sb$_2$88S$_2$89 it acts simultaneously with the wet chemistry, and the final linewidth is controlled by multilayer optical interference as much as by nominal diffraction-limited focusing (Dey et al., 22 Apr 2025). A common misconception is therefore that LAE always means “laser drilling in a liquid.” Much of the field instead concerns latent-volume writing followed by selective development.

Another recurring theme is that optical field structure imprints itself directly onto the etched morphology. The Gaussian Ar-ion beam in porous-Si LIE produces broader, nonuniform pores through center-to-edge etch-rate variation (Saxena et al., 2014). In fused silica, nanograting alignment makes etching strongly polarization dependent in the femtosecond regime, but chirped picosecond pulses can move the system into a nanocrack-dominated regime with nearly polarization-insensitive etch rates (Li et al., 2018). In thin-film Sb$_2$90S$_2$91, substrate reflectivity changes the standing-wave pattern at the reactive interface, lowering threshold on Au while worsening linewidth relative to SiO$_2$92 (Dey et al., 22 Apr 2025). This suggests that LAE optimization is often fundamentally electromagnetic before it is chemical.

Dose dependence is equally nontrivial. Another common misconception is that higher deposited dose monotonically improves selectivity. The fused-silica few-pulse study reports the opposite in a central regime: the most efficient hydroxide etching occurred near $_2$93, an order of magnitude below conventional doses, before clear nanogratings formed (Casamenti et al., 2021). The display-glass work reaches a similar systems-level conclusion from a different direction: the best written trajectory is not merely the desired final shape, but the shape plus reagent-access pathways, because etchant transport and release mechanics determine whether the modified region can actually separate the part (Wimmer et al., 2024).

The limitations are likewise consistent across materials. LAE often trades large-area uniformity for spatial selectivity, as seen in the $_2$94 LIE area in Si and the inability to obtain cross-sectional SEM for those samples (Saxena et al., 2014). High selectivity may come at the cost of anisotropy or roughness, as in the strong polarization sensitivity of low-dose NaOH etching in fused silica and the roughness penalty of hydroxides relative to HF (Casamenti et al., 2021). Process-transfer details may remain incomplete even in otherwise strong demonstrations, as shown by the pulse-duration inconsistency in spherical-aberration-assisted LIDE, the absence of disclosed etch recipes in ULE SLE, and the frequent lack of full selectivity or roughness tables in application-driven papers (Bang et al., 21 Oct 2025, Casamenti et al., 2024). The field is therefore mature enough to show reliable device fabrication, but many implementations remain process-specific rather than universally parameterized.

Taken together, these results define LAE as a selectivity-engineering discipline. Depending on the material system, the privileged etched volume may be created by photogenerated carriers, defect-rich glass, nanograting porosity, nanocrack networks, spherical-aberration-shaped filaments, or chemically activated thin-film exposure. The unifying feature is not a single chemistry or pulse regime, but the controlled conversion of optical localization into differential etchability.