Photolithography is the process that prints circuit patterns onto silicon wafers, and it is the single most critical step in manufacturing every modern computer chip. The technique works somewhat like a projector shining light through a stencil, shrinking the image down and transferring it onto a light-sensitive coating on the wafer’s surface. Over the past five decades, engineers have pushed photolithography from printing features measured in micrometers to features just a few nanometers across, and that relentless shrinking is the engine behind the exponential growth in computing power most people take for granted.
How the Process Works
Chip manufacturing involves dozens of sequential steps, with photolithography, material deposition, etching, and polishing repeated in cycles to build up the layered structures of a finished processor or memory chip.1Applied System Innovation. Review of Applications of Experimental Designs in Wafer Manufacturing Photolithography itself fits within each of those cycles as the patterning step. A thin layer of light-sensitive material called photoresist is spun onto the wafer. Light passes through a patterned mask (also called a reticle), and a set of lenses projects a shrunken image of that mask pattern onto the coated wafer. Where the light hits the resist, chemical changes occur that make those areas either soluble or insoluble in a developer solution, depending on whether the resist is “positive” or “negative.” The wafer is then dipped or sprayed with developer to wash away the unwanted portions, leaving behind a precise resist pattern. That pattern acts as a stencil for the next step: etching trenches into the silicon, depositing metal lines, or implanting atoms to alter the electrical properties of specific regions. Once the pattern has served its purpose, the remaining resist is stripped away, and the cycle starts again for the next layer. A modern processor can require 80 or more of these lithography cycles stacked on top of one another.
Why Resolution Matters So Much
The smallest feature a lithography system can print is governed by the wavelength of light it uses and the optical properties of the lens. This relationship traces back to a principle from 19th-century optics that became the semiconductor industry’s guiding framework. The resolution formula, derived from what’s known as the Rayleigh criterion, tells engineers that finer patterns require shorter wavelengths of light, a higher numerical aperture in the lens (meaning the lens gathers light from a wider angle), or clever tricks that reduce a process-dependent factor known as k1.2Annals of Business Administrative Science. Rayleigh Criterion: The Paradigm of Photolithography Equipment For decades, each new generation of lithography equipment tackled at least one of those three levers. Lenses grew larger and more complex, light sources moved to shorter wavelengths, and computational tricks squeezed more out of the optics. The Rayleigh criterion acted almost like a roadmap, predicting what seemed impossible and then daring engineers to find workarounds.
This framework also made the industry chronically pessimistic about how far optical lithography could go. At several points over the past 30 years, experts declared that the physical limits of light would force a switch to an entirely different patterning technology. Each time, a new combination of wavelength, lens design, and process innovation proved the predictions wrong, at least for another generation.
The Shift to Shorter Wavelengths
Early photolithography systems used mercury arc lamps, filtering out specific spectral lines to get a usable wavelength. The g-line (436 nm) and i-line (365 nm) served the industry well through the 1980s and into the 1990s. But as chips demanded features below one micrometer, those wavelengths were no longer short enough. The industry turned to excimer lasers, which produce intense pulses of deep ultraviolet (DUV) light at much shorter wavelengths.
Krypton fluoride (KrF) excimer lasers emit at 248 nm, and argon fluoride (ArF) lasers push further down to 193 nm.3Conference on Lasers and Electro-Optics. Excimer lasers for DUV lithography This transition to DUV lithography emerged as the technology capable of routinely producing features as small as 0.5 micrometers and beyond.4Microelectronic Engineering. Experience with deep UV excimer laser lithography KrF systems dominated through the 130 nm and 90 nm technology nodes, and ArF systems, especially once paired with immersion lithography (where water is placed between the final lens and the wafer to boost the effective numerical aperture), carried the industry all the way down to about 7 nm node features using a technique called multiple patterning. Multiple patterning prints the same layer in two or more exposures, effectively halving the pitch each time, but at the cost of more steps, more time, and more potential for alignment errors.
Extreme Ultraviolet Lithography
The jump from DUV to extreme ultraviolet (EUV) lithography was one of the most ambitious engineering projects in industrial history, spanning roughly two decades of development. EUV systems use light at 13.5 nm, a wavelength so short it is absorbed by virtually everything, including air and glass. That single fact forced engineers to rethink every component of the lithography tool.
The light source is perhaps the most extraordinary part. A high-power carbon dioxide laser fires pulses at tiny droplets of molten tin, each about 25 micrometers across, traveling through a vacuum chamber at high speed. When the laser pulse hits a tin droplet, it vaporizes the metal into a plasma hot enough to emit EUV radiation at 13.5 nm.5Plasma Sources Science and Technology. Physics of laser-driven tin plasma sources of EUV radiation for nanolithography This laser-produced plasma approach became the preferred design for high-volume manufacturing, with source power targets exceeding 200 watts of in-band EUV light.6Physics Research International. Development of Laser-Produced Tin Plasma-Based EUV Light Source Technology for HVM EUV Lithography Getting to those power levels was one of the biggest bottlenecks. Each droplet yields only a tiny flash of EUV, so the system has to fire the laser and hit a fresh droplet roughly 50,000 times per second.
Mirrors Instead of Lenses
Because EUV wavelengths are absorbed by glass, conventional refractive lenses are out of the question. Instead, the entire optical path of an EUV system uses mirrors. These are not ordinary mirrors. Each one is a multilayer structure made of alternating ultra-thin films of molybdenum and silicon, deposited with atomic-level precision. The layers are tuned so that each interface reflects a small fraction of the incoming EUV light, and the reflections add up constructively, similar to how a thin film of oil on water creates vivid colors, but engineered to work at a single target wavelength.
Early experiments with molybdenum-silicon multilayers demonstrated peak reflectivities reaching as high as roughly 78% at near-normal incidence for EUV-range wavelengths, showing that traditional optics approaches could be extended into the EUV regime.7Applied Optics. Molybdenum-silicon multilayer mirrors for the extreme ultraviolet Even so, each mirror absorbs a significant fraction of the light. A typical EUV projection system has around 10 to 12 mirrors between the source and the wafer, and the cumulative losses mean that only a small percentage of the original EUV light actually reaches the resist. This is why the source has to be so powerful.
Keeping those mirrors in working condition is its own challenge. Under intense EUV exposure, the molybdenum-silicon interfaces can degrade. Research has shown that adding carbon barrier layers between the molybdenum and silicon films can boost the single-shot damage threshold by about 46%, a meaningful improvement for extending mirror lifetime under the harsh conditions inside an EUV scanner.8Nano Letters. Enhanced Damage Resistance of Mo/Si Multilayer Mirror with Carbon Barrier Layers under Intense Nanosecond EUV Irradiation
Protecting the Mask
In older DUV systems, a thin transparent membrane called a pellicle is stretched over the mask to keep dust particles from landing on the pattern surface. Any speck of dust on the mask would print as a defect on every single wafer exposed through it, so the pellicle acts like a protective window. The concept is simple, but translating it to EUV has been extremely difficult. At 13.5 nm, there is no material that is both transparent and mechanically robust the way glass or polymer films are for visible or DUV light.
EUV pellicles must be nanometer-scale thin membranes that transmit as much EUV light as possible while surviving high temperatures and exposure to hydrogen plasma inside the tool. One approach uses metal silicide films. Recent work has demonstrated full-size pellicles (110 mm by 144 mm) with over 90% EUV transmittance, less than 0.04% reflectance, and sufficient mechanical strength to withstand the rigors of high-power EUV exposure. In durability testing, these pellicles survived the equivalent of 20,000 wafer exposures at 400 watts of EUV power.9Nanotechnology. Development and optimization of metal silicide EUV pellicle for 400W EUV lithography
Another line of research explores membranes made from single-walled carbon nanotubes. These freestanding films have shown EUV transmittance around 74%, and researchers demonstrated that the membranes can be renewed after degradation, restoring their original transmittance, which could help extend their useful life in a production environment.10Carbon. Renewable single-walled carbon nanotube membranes for extreme ultraviolet pellicle applications Whether nanotube-based pellicles or metal silicide membranes ultimately win out in high-volume fabs remains an open question, and some manufacturers have simply opted to run EUV tools without pellicles, relying instead on extremely clean mask-handling procedures and accepting the defect risk.
The Stochastic Problem
As features shrink to just a few nanometers, the statistical nature of light itself starts to matter. At 13.5 nm, each photon carries far more energy than a DUV photon, which means fewer photons land in any given area of resist for the same total exposure dose. When you are trying to define the edge of a 5 nm feature, the randomness of where individual photons happen to land and where chemical reactions happen to occur creates visible roughness along pattern edges. This phenomenon, called line-edge roughness, is one of the most stubborn challenges in EUV lithography.
Computational modeling of EUV processes for 5 nm-class patterns has shown that photon shot noise and a blurring effect within the resist are key drivers of line-edge roughness.11PubMed Central. Line-Edge Roughness from Extreme Ultraviolet Lithography to Fin-Field-Effect-Transistor: Computational Study The practical consequence is that transistor performance varies slightly from one device to the next across a chip, since the roughness translates directly into variations in the physical dimensions of each transistor. At older, larger nodes this variation was swamped by other factors and didn’t matter much. At 5 nm and below, it starts to eat into performance margins. The resist community is actively developing new materials that react more efficiently to EUV photons, aiming to get sharper edges with the same or lower dose, but it is fundamentally a tradeoff: higher sensitivity to light tends to come with worse roughness, and vice versa.
Complementary Patterning Approaches
Photolithography doesn’t have to do everything alone. Several complementary techniques are being developed to work alongside it, either to push resolution further or to reduce costs for certain pattern types.
Directed self-assembly (DSA) takes advantage of materials called block copolymers, which naturally organize themselves into regular nanoscale patterns when given the right conditions. By using a lithography-defined guide pattern as a template, engineers can coax these materials into forming features with tighter pitch than the lithography tool could produce on its own. Combining DSA with EUV, DUV, electron-beam, or nanoimprint lithography can enhance the resolution of target patterns and increase device density.12PubMed Central. Review of Directed Self-Assembly Material, Processing, and Application in Advanced Lithography and Patterning Early demonstrations showed that block copolymers in thin films can be directed to form nearly the complete set of dense and isolated patterns that the semiconductor industry uses, from regular line-and-space gratings to the more complex contact holes and bends needed in real circuit layouts.13PubMed. Directed self-assembly of block copolymers for nanolithography: fabrication of isolated features and essential integrated circuit geometries
Electron-beam direct-write lithography is another approach. Instead of projecting an image through a mask, an electron beam traces the pattern directly onto the resist, point by point. The resolution is excellent, potentially below 10 nm, but throughput has always been the bottleneck. Writing a full wafer one pixel at a time is fantastically slow compared to exposing an entire field at once with light. To address this, researchers have developed massively parallel systems that generate arrays of a million or more individual electron beamlets simultaneously, targeting throughput levels above 100 twelve-inch wafers per hour.14Nature. Development of massively parallel electron beam direct write lithography using active-matrix nanocrystalline-silicon electron emitter arrays Whether multi-beam e-beam systems can ever compete with EUV on throughput for leading-edge logic and memory production remains to be seen, but they fill a valuable niche for mask-making, prototyping, and certain specialty applications where the flexibility of maskless patterning outweighs the speed penalty.
The Environmental Footprint
A less visible side of photolithography involves the chemicals used in the process. Photoresists and the solutions used to develop, strip, and clean them contain a range of industrial chemicals, and some of those chemicals have come under increasing environmental scrutiny. Per- and polyfluoroalkyl substances (PFAS), the so-called “forever chemicals” that resist breakdown in the environment, are used in certain photolithography-relevant formulations as surfactants and additives.
Recent toxicology research has examined PFAS compounds relevant to photolithography and found meaningful differences in their biological effects. Among the compounds tested, some well-known PFAS like PFOS showed significant toxicity in developmental assays, while several less-studied compounds used in semiconductor manufacturing showed lower acute toxicity at the concentrations tested.15Environmental Science & Technology. Developmental Toxicity of Photolithography-Relevant Per- and Polyfluoroalkyl Substances (PFAS) Reveals Concerns for Less-Studied Functional Groups The researchers noted that less-studied functional groups in these chemicals still warrant concern, even when their acute lethality is lower, because developmental and sublethal effects can occur at much lower concentrations than lethal ones. The semiconductor industry has begun exploring PFAS-free alternatives for some process steps, driven by both regulatory pressure in the EU and parts of the United States and by the industry’s own sustainability goals. But replacing PFAS in photolithography is not straightforward, because the chemicals serve specific functions related to wetting, coating uniformity, and defect prevention that are hard to replicate with other materials.
Where Things Stand and What Comes Next
EUV lithography is now the workhorse for the most advanced chip layers at the leading fabs, used in production for nodes marketed as 7 nm, 5 nm, and 3 nm. The next major hardware step is high-numerical-aperture (high-NA) EUV, which roughly doubles the resolving power of current EUV systems by using a larger, more complex set of mirrors. The first high-NA tools have been delivered to research fabs, and early results are being characterized. High-NA EUV is expected to extend single-exposure patterning to features that would otherwise require double patterning on current EUV systems, simplifying the process and potentially reducing cost per layer even though the tools themselves are more expensive.
Further out, there is speculation about what might come after EUV. Some researchers are investigating even shorter wavelengths, while others think the future involves hybrid approaches that combine optical lithography with self-assembly, direct-write, or nanoimprint techniques to handle different parts of a chip’s pattern stack. The history of photolithography is a history of defying predicted limits, and the industry’s track record suggests the technology still has room to evolve in ways that are hard to anticipate from the current vantage point. But the problems ahead are genuinely harder than the problems behind. Stochastic effects, mirror degradation, pellicle fragility, resist chemistry constraints, and the sheer cost of the tools all represent challenges where the physics pushes back harder with each generation. The era of relatively straightforward wavelength-and-lens improvements is over; what comes next will require more creativity from more disciplines working together.