---
title: Laser Subtractive Manufacturing
url: https://www.emergentmind.com/topics/laser-subtractive-manufacturing
type: topic
---

# Laser Subtractive Manufacturing

Laser subtractive manufacturing denotes laser-based fabrication in which geometry is defined by removing material rather than depositing it. In the cited literature, that removal appears as selective thin-film ablation, thermal evaporation, vaporization-driven lift-off, drilling, line scribing, trench formation in sacrificial layers, direct micromachining of free-standing membranes, and broader process classes such as cutting, drilling, ablation, and surface structuring. Its technical appeal lies in mask-free or direct-write operation, chemical-free or lower-chemical workflows, and the possibility of high spatial selectivity; its central constraint is that the same laser–matter interaction that enables removal can also generate heat-affected zones, substrate damage, recast, redeposition, pore formation, or morphology-dependent functional degradation if temporal, spatial, and materials parameters are not tightly controlled [2008.08505] [1206.2399] [1109.5047] [2509.22666].

## 1. Conceptual scope and taxonomy

Laser subtractive manufacturing is most clearly defined by intent: the laser removes material to produce the negative space of a pattern, cavity, trench, hole, or released structure. In transparent-electrode micromachining, for example, the laser selectively ablates ITO from chosen regions, leaving the remaining material as the electrode pattern; this route is described as mask-free, direct-write, applicable in ambient air, scalable, and comparatively low-cost [2510.00200]. In selective thin-film processing, the goal is the removal of only the intended metal layer while preserving neighboring features and the underlying substrate [2008.08505]. In silica micro-optics, a single focused CO\(_2\) pulse removes material to form a concave depression while a transient melt layer smooths the resulting surface [1109.5047].

The category is not identical to laser microfabrication as a whole. A recurring misconception is to treat any laser-written micro/nanostructure as subtractive. The literature provides explicit counterexamples. Partition Laser Assembling is described as additive rather than subtractive, because structure formation occurs through two-photon polymerization driven by structured-light “light stamps,” not by ablation [2504.15554]. Conversely, the PAA/TPL/Ni\(_{80}\)Fe\(_{20}\) route is explicitly hybrid: femtosecond laser machining removes a sacrificial polymer layer, two-photon lithography writes a 3D scaffold into the cleared regions, and later deposition and lift-off yield a functional magnetic nanostructure [2005.14479]. The boundary of the term therefore depends on whether laser exposure primarily removes material, adds material, or couples both operations in a single workflow.

A second classification axis is process regime. The cited work spans sub-picosecond ablation of multilayer metal films, femtosecond trenching of sacrificial polymers, femtosecond sequential-hole release of SiN membranes, microsecond CO\(_2\)-laser micromachining of glass, and continuous-wave drilling and machining with dynamic focal fields. This range matters because the dominant removal physics shifts from ultrafast electron–lattice nonequilibrium to evaporation-driven melt reshaping to hydrodynamic melt transport, even when all of these are discussed under the umbrella of subtractive manufacturing.

## 2. Removal physics across pulse-duration and beam-shaping regimes

A central result in ultrashort-pulse subtractive processing is that selectivity can be engineered through nonequilibrium heat transfer rather than through brute-force melting. In Au/Ti/glass and related multilayers, the proposed sequence is: laser energy is absorbed by free electrons in the metal, redistributed by electron–electron coupling, transferred to the lattice by electron–phonon coupling, and then concentrated in a transition-metal interlayer whose electron–phonon coupling is much stronger than that of the noble-metal top film. For a 100 nm Au film with a 5 nm Ti interlayer on aluminosilicate glass, irradiated with 0.2 ps pulses at 1035 nm, clean selective removal was reported at about \(3.2\text{–}3.9\ \mathrm{J/cm^2}\), whereas damage became unavoidable at \(5.2\ \mathrm{J/cm^2}\); longer pulses led to overcut and substrate damage, including glass overcut at \(84.6\ \mathrm{ps}\) and failure of rapid Ti vaporization in the 100 ps simulation [2008.08505]. The analytical description uses a two-temperature model,
\[
C_e \frac{\partial T_e}{\partial t} = \nabla \cdot (k_e \nabla T_e) - G_{e-p}(T_e - T_l) + S,
\]
\[
\rho\left[c_p + L_m \delta(T_l - T_m) + L_v \delta(T_l - T_v)\right] \frac{\partial T_l}{\partial t} = \nabla \cdot (k_l \nabla T_l) + G_{e-p}(T_e - T_l),
\]
with interface resistances and ballistic hot-electron transport included. The key physical result is that Ti reaches vaporization before Au reaches melting, producing a lift-off effect while limiting heat flow into glass [2008.08505].

At the opposite end of the timescale spectrum, CO\(_2\)-laser machining of silica relies on thermal evaporation plus self-smoothing in a thin melt layer. Fused silica strongly absorbs \(10.6\,\mu\mathrm{m}\) radiation in the first few micrometers, allowing localized heating, melt formation, evaporation-driven material removal, and surface-tension-induced suppression of short-wavelength roughness. Across several hundred fabricated structures, variation of average laser power from 300 mW to 2 W, pulse train duration from 4 to 120 ms, and beam waist from 21 to 93 \(\mu\)m produced depressions with depths \(t = 0.01\text{–}4\,\mu\mathrm{m}\), diameters \(d = 10\text{–}60\,\mu\mathrm{m}\), and radii of curvature \(R = 20\text{–}2000\,\mu\mathrm{m}\), with surface roughness near \(0.2\ \mathrm{nm\ rms}\) [1109.5047]. The temperature field is modeled from a Gaussian intensity
\[
I(r) = I_0 \exp\!\left(-\frac{2r^2}{w^2}\right),
\]
and the evaporation front velocity follows
\[
v(r,\tau)=v_0 \exp\!\left(-\frac{U}{k_B T(r,\tau)}\right),
\]
which explains the strong sensitivity of depth and diameter to absorbed power and focus position [1109.5047].

Dynamic beam shaping introduces a third mechanism in which the focal field itself becomes a time-dependent processing tool. Drill-bit-inspired beams generated by cylindrical vector beams and rotational vectorial polarization filtering create a near diffraction limited two-lobe Hermite-Gaussian focus that spins about the optical axis. The governing scan parameter is
\[
l=\frac{u}{f},
\]
with \(u\) the scan speed and \(f\) the spin frequency [2607.11553]. In continuous-wave drilling of 316L stainless steel at 10 W for 2 s, the spinning beam produced holes of about \(70\text{–}108\,\mu\mathrm{m}\), while the conventional Gaussian/LG beam produced about \(9\text{–}21\,\mu\mathrm{m}\); the spinning beam also yielded about four times higher drilling efficiency, measured as drilled area per same power [2607.11553]. The reported interpretation is explicitly hydrodynamic: the rotating two-lobe profile sweeps around an annular path, creates a moving thermal gradient, shears the melt azimuthally, drives molten material outward, and suppresses recast plugging.

## 3. Process architectures, sensing, and control

Precision in laser subtractive manufacturing is often determined as much by control architecture as by absorption physics. A clear case is feedback-controlled CO\(_2\)-laser fabrication of concave micromirror templates on silica and fluoride glass. There, short single focused pulses of 40–200 \(\mu\)s and energy about \(100\,\mu\mathrm{J}\) were used to produce concave features with diameters of \(20\text{–}100\,\mu\mathrm{m}\) and average central RMS microroughness of less than \(0.2\ \mathrm{nm}\). The crucial addition was a fast feedback control loop based on white-light emission from the sample. A Si photodetector generated a voltage \(V_{\rm det}\), a comparator tested
\[
V_{\rm det} > V_{\rm ref},
\]
and when that threshold was exceeded the FPGA switched off the laser. The overall feedback bandwidth was about 20 kHz, and the measured RMS diameter fluctuation improved from \(\Delta D_{\rm RMS}/\bar D \approx 50\%\) in open loop to \(\approx 4\%\) in closed loop [1206.2399]. This establishes emitted-light feedback as a practical in situ observable for crater-size reproducibility when the physical machining event occurs on a timescale shorter than or comparable to the laser pulse duration.

A different control architecture appears in femtosecond micromachining of free-standing SiN membranes. Rather than tracing a continuous cut, the laser releases the structure through a sequence of overlapping microholes distributed around the polygon perimeter so that stress concentration remains limited. The open-source Stress-Controlled Laser Micromachining Toolset converts GDSII layouts into machine instructions through five modules: Layout File Reader, Layout Aligner, Polygon Hole Sequence Generator, Layout Hole Sequence Assembler, and Numerical Control File Writer [2507.18599]. With 1 \(\mu\)m hole diameter and a final target spacing \(s_{f,\mathrm{target}} = 0.5\,\mu\mathrm{m}\), adjacent holes overlap to ensure release; alignment uncertainty and stage error are handled by adding 1 \(\mu\)m padding, leading to a maximum possible overhang of about 2 \(\mu\)m [2507.18599]. Resonators were fabricated in 20 minutes to 1 hour each, demonstrating that geometry-flexible subtractive release can be automated from standard semiconductor layout files.

Both cases show that subtractive laser processing is no longer confined to open-loop point exposure. Closed-loop emission monitoring, adaptive layout alignment, and stress-aware sequence generation all convert ablation from a local material-removal event into a digitally orchestrated manufacturing operation.

## 4. Modeling, simulation, and manufacturability analysis

Theoretical treatment of laser subtractive manufacturing ranges from reduced nonequilibrium heat-transfer models to fully coupled continuum mechanics. For ultrashort thin-film removal, the two-temperature model separates electron and lattice temperatures and includes ballistic transport and interface thermal resistance to explain why a Ti interlayer can vaporize before a top Au film fully melts [2008.08505]. At a broader scale, the universal numerical simulation model formulates laser processing as a compressible multiphase continuum problem that does not assume a priori whether a given parameter set yields welding, drilling, or ablation [2509.22666]. In that framework, the mixture mass and phase-wise mass equations are
\[
\frac{\partial \rho}{\partial t} + \nabla \cdot (\boldsymbol{u}\rho)=0,\qquad
\frac{\partial \rho_i}{\partial t}+\nabla \cdot (\boldsymbol{u}\rho_i)=S_{M,i},
\]
with momentum, energy, capillarity, Marangoni effects, mushy-zone drag, phase change, and laser absorption solved in a unified finite-volume formulation implemented in OpenFOAM [2509.22666]. The electron subsystem for ultrashort pulses is represented by
\[
\frac{\partial (\rho e)}{\partial t} + \nabla \cdot (\boldsymbol{u}\rho e)
= \nabla \cdot (\lambda_e \nabla T_e) + Q_{abs} - G(T_e-T),
\]
and the model explicitly states that there is no imposed “remove cell” criterion: material removal emerges from phase change, density change, vapor pressure build-up, free-surface motion, and outflow from the computational domain [2509.22666].

This continuum framework was validated against keyhole drilling and collapse, copper welding, and single-pulse 120 fs copper ablation; for ultrafast copper ablation it achieved good agreement with measured ablation depth up to about \(45\ \mathrm{J/cm^2}\), with discrepancies at the highest fluence attributed to uncertain high-temperature properties and equation-of-state limitations near critical or supercritical conditions [2509.22666]. For process design, the significance is that focus position, beam profile, pulse duration, and material-specific constitutive data can be optimized within a single multiphysics structure rather than through process-specific heuristics.

Manufacturability analysis introduces a complementary abstraction. DeepMill frames subtractive accessibility as prediction of inaccessible and occlusion regions under varying cutter parameters using a cutter-aware dual-head octree-based CNN. It reports 94.7% accuracy for inaccessible regions, 88.7% accuracy for occlusion regions, and average processing time of 0.04 s on complex geometries [2502.06093]. The formal accessibility condition is written as
\[
\forall s_i \in S,\ \forall d_k \in D,\ s_i \cap C(d_k)=\varnothing,
\]
with inaccessible points defined by collision in every sampled direction [2502.06093]. The paper explicitly notes that the same logic is relevant to laser subtractive manufacturing when a laser-head, nozzle, beam delivery geometry, or associated tool assembly has constrained approach directions. This suggests a pathway from pure process simulation toward DFM-oriented reachability analysis for laser systems whose effective manufacturability is limited not only by fluence and heat flow but also by geometric access.

## 5. Materials systems and application domains

The literature shows that laser subtractive manufacturing is not tied to a single material class or functional sector. Instead, process physics is strongly conditioned by optical absorption, electron–phonon coupling, melt viscosity, stress state, and the relevance of conductivity, optical roughness, or mechanical dissipation in the final device.

| Material system | Subtractive operation | Reported use |
|---|---|---|
| Au, Ag, or Cu / Ti or Cr / glass, SiN, PET | Selective thin-film removal by ultrashort-pulse lift-off | 3-D patterning of optoelectronic devices, localized repair of integrated circuits, micro- and nano-scale line/dot patterning, flexible electronics, transparent conductive structures |
| Fused silica and fluoride glass | CO\(_2\)-laser ablation with local reflow or thermal evaporation | Concave micromirror templates, Fabry–Perot mirrors, negative lenses on fiber tips |
| ITO on soda-lime glass | Femtosecond micromachining of electrode patterns | Transparent electrodes for electrochemistry, optoelectronics, microfluidics, energy harvesting |
| Free-standing silicon-rich SiN membranes | Sequential hole ablation for crack-free release | Nanomechanical resonators for radiation sensing |
| PAA on glass/ITO, followed by TPL and Ni\(_{80}\)Fe\(_{20}\) deposition | Femtosecond ablation of sacrificial layer | Isolated 3D magnetic nanowires and 3D artificial spin-ice structures |

In noble-metal thin films, ultrashort-pulse removal enabled microdots, 2 \(\mu\)m-wide line scribing, integrated-circuit repair on Cu/Ti/SiN structures, large-area periodic gratings, and honeycomb metal grids on PET [2008.08505]. In micro-optics, the subtractive CO\(_2\) route yielded concave spherical templates with RMS microroughness below \(0.2\ \mathrm{nm}\); after coating, these supported cavities with finesse \(F \approx 80\) using gold coating at 850 nm and \(F > 58{,}000\) using dielectric coating at 940 nm [1206.2399]. In transparent-conductor processing, ITO tracks with widths from about 10 to \(1000\,\mu\mathrm{m}\) were fabricated and electrically characterized under 515 nm and 343 nm femtosecond machining [2510.00200]. In nanomechanics, direct femtosecond ablation of 100 nm SiN membranes produced resonators with measured \(Q\) factors up to \(3.7\times 10^6\) and extracted material quality factors above 3700 [2507.18599]. In hybrid nanofabrication, femtosecond trenching of a sacrificial PAA layer created anchored windows for later two-photon-written 3D polymer scaffolds and final Ni\(_{80}\)Fe\(_{20}\) magnetic structures, including \(100\,\mu\mathrm{m}\)-long, 300 nm wide nanowires raised approximately \(3\,\mu\mathrm{m}\) above the substrate [2005.14479].

The application spread indicates that “laser subtractive manufacturing” is less a single process than a family of removal strategies adapted to different figures of merit: minimal substrate damage in multilayer electronics, sub-nanometer roughness in optics, conductivity preservation in transparent electrodes, crack-free release in prestressed membranes, or background-free isolation in functional 3D nanostructures.

## 6. Defects, process limits, and evolving research directions

The principal limitation across the cited work is that removal and collateral modification are tightly coupled. In thin-film ablation, ordinary laser processing can induce heat-affected zones, melt surrounding patterns, and damage the substrate underneath; even in the ultrashort-pulse Ti-assisted regime, damage became unavoidable at \(5.2\ \mathrm{J/cm^2}\) [2008.08505]. In ITO micromachining, the Gaussian energy distribution creates a graded edge region comprising pristine ITO, dense LIPSS, isolated LIPSS, and bare glass rather than a perfectly abrupt boundary. At 515 nm, LIPSS perpendicular to the track exhibit higher resistance by a factor just above two relative to parallel LIPSS, with fitted factors \(a \approx 3.5\) and \(a \approx 1.55\), respectively; at 343 nm, the transition zone is narrower and the contribution of the laser-affected edge can be neglected to within about 5% even for \(20\text{–}25\,\mu\mathrm{m}\) tracks [2510.00200]. Here, a nominally geometric byproduct becomes a functional defect because the structured edge alters percolation paths and effective electrical width.

Material selectivity remains equally important. In feedback-controlled CO\(_2\) micromirror fabrication, good quality templates were obtained only on fused silica and fluoride glass; float glass, BK7, and ceramic Y\(_3\)Al\(_5\)O\(_{12}\) tended to crack or produce poor surfaces [1206.2399]. In free-standing SiN micromachining, overhang was limited to about \(2\,\mu\mathrm{m}\) rather than eliminated entirely, small-feature machining could cause redeposition debris, and more complex or higher-aspect-ratio designs could require slower interleaved machining [2507.18599]. In continuum simulation, predictive quality depends strongly on temperature-dependent density, complex refractive index, viscosity, surface tension, latent heats, and equation of state; discrepancies at high fluence in ultrafast copper ablation were explicitly linked to uncertain high-temperature properties and equation-of-state limitations [2509.22666].

The recent direction of the field is toward tighter coupling of beam shaping, sensing, and digital process intelligence. Dynamic focal-profile control shows that time-dependent intensity distributions can increase drilling efficiency, widen the effective processing zone, and generate programmable surface textures from simple scans [2607.11553]. Emission-based feedback demonstrates that single-shot ablation can be terminated by an in situ optical threshold rather than by fixed pulse duration alone [1206.2399]. Stress-aware and layout-aware hole sequencing shows that arbitrary GDSII geometries can be released directly from prestressed membranes with high \(Q\) retained [2507.18599]. A plausible implication is that future laser subtractive manufacturing will increasingly combine dynamic focal fields, closed-loop observables, multiphysics simulation, and accessibility-aware design tools in a single process chain.

A final misconception is that higher geometric fidelity is obtained only by replacing subtractive routes with additive ones. The evidence is more specific. Some problems are indeed better addressed additively, as in PLA [2504.15554]. Others require hybridization, as in sacrificial-layer removal followed by two-photon lithography and metal lift-off [2005.14479]. But the cited subtractive results show that laser removal itself can yield flat crater bottoms, well-defined edges, RMS microroughness below \(0.2\ \mathrm{nm}\), under-1-hour rapid prototyping of complex SiN resonators, or transparent-electrode isolation with wavelength-dependent control of edge damage. The governing issue is therefore not whether subtraction is intrinsically inferior, but which removal mechanism, control strategy, and material stack are matched to the target functionality.

Source: https://www.emergentmind.com/topics/laser-subtractive-manufacturing