KNOWLEDGE NETWORK ← Index
Chemistry · Optics · Making

Photolithography

Printing with light needs three things: a film that changes when light hits it, a pattern that decides where the light goes, and a way to wash away one of the two states. That's the whole idea. The same three ingredients made the oldest surviving photograph in 1827, make every circuit board, etch the nameplate on a machine, and print the 8-nanometre features of today's chips. What changes from one to the next is only the chemistry and the instruments. This page takes you through all of it, with five simulators that compute live.

01 — The Idea

A stencil made by light

Lithography means "stone writing": it began as printing from flat stone. Photolithography replaces the hand-drawn image with one written by light. You coat a surface with a thin photoresist, shine light through a mask (or project an image), and the light changes the resist's chemistry wherever it lands. A developer then dissolves one of the two states, leaving the pattern as a thin stencil sitting on the surface.

The stencil is the point. Whatever you do next only touches the bare places: etch them, plate metal into them, deposit a coating, or sandblast them. Then you strip the resist and the pattern is permanently in the material. The resist is just a temporary, extremely precise masking tape that light cuts for you.

Joseph Nicéphore Niépce's 1827 heliograph, the oldest surviving camera photograph, was made exactly this way. He dissolved light-sensitive bitumen in oil of lavender, coated a polished pewter plate, and exposed it in a camera for several days [1]. Where light fell the bitumen hardened; the unhardened parts were then washed away with solvent, leaving the image as bitumen on bare metal [2]. In today's terms, bitumen was a negative photoresist.

02 — Resists

Positive or negative: which part survives

A positive resist becomes more soluble where it is exposed: the developer removes the lit areas, so the resist keeps a copy of the mask's dark parts. The classic positive resist is a novolak resin mixed with diazonaphthoquinone (DNQ). DNQ blocks the resin from dissolving, and light converts it into a compound that no longer blocks, so the exposed areas dissolve in an alkaline developer [3].

A negative resist does the opposite: light (often followed by a bake) cross-links it into an insoluble network, so the lit areas stay [3]. Niépce's bitumen and the dry-film resists used on circuit boards are negative. Which one to use depends on what is easier to draw and what the next step needs: lift-off processes, for example, prefer the overhanging sidewalls negative resists can give.

Each resist responds to a band of wavelengths. Mercury lamps emit three strong lines that gave the industry its vocabulary: g-line 435 nm, h-line 405 nm and i-line 365 nm [3]. Cheap 405 nm and 365 nm LEDs now make the same exposures possible on a workbench. Chips moved to deep-UV lasers at 248 and 193 nm, and then to extreme-UV at 13.5 nm [8].

03 — The Process

From bare surface to pattern, step by step

Step through a complete cycle below and switch between a positive and a negative resist. The steps are the same whether the substrate is a silicon wafer, a copper-clad board or a sheet of steel. Only the chemicals and times change.

04 — Spin Coating

Spin faster, coat thinner

Resist is applied as a liquid and spun: centrifugal force flings off the excess while the solvent evaporates, leaving a film that is remarkably even across the surface. Its final thickness falls roughly with the inverse square root of the spin speed. Double the speed and the film gets about 1.4 times thinner [4][5]. Thickness matters: thicker resist survives longer etches but blurs fine detail, because the light has farther to travel through it, and in contact printing it widens the gap the light must cross (Chapter 5 uses the thickness you set here).

Model t = t₄₀₀₀·√(4000 / ω), the reciprocal-square-root rule from [4][5]. Reference thicknesses follow the convention that a resist's last two digits give its film thickness at 4000 rpm in 100 nm units [4] (e.g. "…18" ≈ 1.8 µm). Real curves bend at very low and very high speeds.
05 — Exposure & Resolution

How small can light draw?

Light is a wave, so it spreads after passing through an opening in the mask. In contact or proximity printing the mask sits on or just above the resist, and the smallest feature you can draw is about d = √(λ·g), where g is the distance the light travels from the mask to the resist [3]. Even perfect contact has an effective gap: the resist's own thickness. A speck of dust lifting the mask by 20 µm already costs you a lot.

In projection printing a lens images the mask onto the resist, and the limit becomes R = k₁·λ/NA, where NA is the lens's numerical aperture and k₁ a process factor [6]. ASML's numbers show what this means in practice: its 0.33-NA EUV scanners print 13 nm features at λ = 13.5 nm, and the 0.55-NA "High-NA" generation prints 8 nm [8]. Work out k₁ from those figures and you get about 0.32 for both, close to the single-exposure floor of 0.25 [6].

d ≈ √(λ·g)  (contact / proximity)  ·  R = k₁·λ / NA  (projection) ASML: 13 nm × 0.33 / 13.5 nm = k₁ 0.318  ·  8 nm × 0.55 / 13.5 nm = k₁ 0.326 [8]

The simulator prints a row of equal lines and spaces. It blurs the mask by the optics you choose, multiplies by the exposure dose, and develops wherever the dose passes the resist's threshold. Change the dose, and lines grow or shrink. Make them too small and the light no longer knows where the lines are.

Simplified model: the mask is blurred by a Gaussian whose width is √(λ·g) for contact/proximity [3] (g includes the resist thickness from Chapter 4) or the Airy-disc width 0.51·λ/NA for projection [6]. Dose is in multiples of the clearing dose of an open area. Real lithography software solves the full partially-coherent imaging equations, and real fabs add tricks (off-axis illumination, mask corrections) that print below what this model allows.
06 — Shrinking With a Microscope

The projector, run backwards

A microscope makes small things look big. Run it backwards — put a pattern where your eye would be and the resist where the specimen would be — and it makes big things small. Love, Wolfe, Jacobs and Whitesides at Harvard used exactly this, an ordinary microscope shrinking a printed transparency, to make masters with micron-scale features for soft lithography, no cleanroom needed [7]. Swap the transparency for a projector's pixel array and you have maskless lithography: the "mask" is whatever image the computer sends.

Two limits compete. Demagnification shrinks the printed dots of the mask by the objective's magnification M. But the objective can't resolve detail finer than about 0.61·λ/NA (Chapter 5). Whichever is larger sets your smallest feature, and paying for a finer mask than the optics can use buys you nothing.

Mask dot = 25.4 mm ÷ dpi. Optical limit 0.61·λ/NA (Rayleigh) [6]. The field per exposure assumes a 20 mm field number, typical of microscope optics; your objective's own value may differ. Objective markings like "10×/0.25" give M and NA directly.
07 — Any Picture as a Matrix

Grey made of black and white

Resist is close to binary: after development a spot is either there or gone. So how do you print a photograph? The same way newspapers do: turn each grey into a pattern of on/off dots whose density your eye averages back into tone. That is the core of your question: any picture can become a matrix of binary features, and the matrix doesn't care about scale. The same grid can be printed with 50 µm dots on a metal plate or 10 nm dots on silicon. Only the chemistry and the instruments change.

Choose a method below. Threshold loses all midtones. Ordered dithering (a Bayer matrix) uses a fixed, regular pattern. Error diffusion (Floyd–Steinberg, 1976 [9]) passes each pixel's rounding error to its neighbours, which keeps the finest detail. Halftone screen grows round dots on a tilted grid, as printing presses do. Then set the smallest feature your process can draw and the physical size of the print: the readout tells you how many features that is.

Source (grey)
Printed matrix (binary)
Your image stays in your browser: nothing is uploaded. The preview is capped at 256 × 256 dots; the readouts give the true count for the size you chose.
08 — Any Surface, Any Size

The same three ingredients, everywhere

Wherever a resist can be coated and a developer can reach, the process works. That is why photolithography quietly runs so much of manufacturing. The table pairs each job with the version of the three ingredients it uses.

ApplicationSurfaceResist & lightWhat the stencil is used for
MicrochipsSilicon waferDUV (248/193 nm) and EUV (13.5 nm) resists, projection scanners [8]Etch, implant, deposit, layer after layer
Circuit boardsCopper-clad laminateNegative dry film, UV through a film or direct imagingEtch away the copper the film doesn't cover
Photochemical machiningThin steel, brass, nickelResist on both sides, UV contact exposureEtch right through: stencils, meshes, springs, nameplates
Screen-printing stencilsFine meshPhoto emulsion, UV through a film positiveInk passes only where emulsion washed out
MicrofluidicsSilicon or glassThick negative resist master, contact or microscope projection [7]Mould for casting silicone channels
The first photographPewterBitumen of Judea, sunlight, camera obscura [1][2]The image itself

Scale runs over six orders of magnitude, from about 10 nm on a chip to dots a millimetre wide on a screen-printed poster, and the principle never changes.

09 — Safety

Read this before you try any of it

This page explains how photolithography works. It is not a protocol. Every chemical below has a safety data sheet; read it, use the protective equipment it names, and follow your local rules for disposal.

⚠ Hazards that matter

  • UV light. 365–405 nm exposure lamps and LEDs can injure eyes and skin, and 365 nm light looks dim or invisible, so a harmful exposure can feel harmless. Never look at the source; shield it and wear UV-blocking eyewear. Exposure limits are set by ICNIRP [12].
  • Developers are strong bases. Sodium and potassium hydroxide solutions burn skin and eyes. The metal-free developer TMAH (tetramethylammonium hydroxide) is far worse: skin contact has killed workers. Taiwan's poison centre recorded three deaths after exposure to 25% TMAH, and a worker splashed with the common 2.38% developer needed two days on a ventilator [11]. Gloves, goggles and an apron, always.
  • Hydrofluoric acid (HF), used to etch glass and silicon oxide, penetrates skin and damages tissue and bone, sometimes without immediate pain. NIOSH lists 30 ppm as immediately dangerous to life [10]. It does not belong in a home workshop.
  • Etchants and solvents. Copper etchants (ferric chloride, persulfates) are corrosive and the spent solution contains dissolved copper, so it is hazardous waste. Acetone and resist thinners are highly flammable; work with ventilation and no open flame.
10 — History

Two hundred years of printing with light

1827
Niépce's heliograph: bitumen on pewter, several days in a camera obscura. It is the oldest surviving camera photograph, and its process was photolithography [1].
1976
Floyd and Steinberg publish error diffusion, still the default way to turn grey into binary dots [9].
1978
Meyerhofer explains why spin-coated film thickness falls with the square root of spin speed [5].
2001
Microscope projection photolithography: micron-scale masters with an ordinary microscope and a printed mask [7].
2010s
EUV at 13.5 nm enters chip production; ASML's NXE scanners (NA 0.33) print 13 nm features [8].
2025–26
High-NA EUV (NA 0.55, 8 nm) is planned to support high-volume chip manufacturing [8].
References

Sources

  1. Harry Ransom Center, The University of Texas at Austin — The Niépce Heliograph. Made in 1827; bitumen dissolved in oil of lavender on a polished pewter plate; several days' exposure in a camera obscura.
  2. View from the Window at Le Gras — Wikipedia (secondary source with primary citations). The bitumen hardened where lit; unhardened bitumen was washed away with oil of lavender and white petroleum.
  3. MicroChemicals GmbH — Exposure of Photoresists (technical document). DNQ/novolak positive resists and cross-linked negative resists; g/h/i-line (435/405/365 nm); proximity resolution d = (λ·g)1/2.
  4. MicroChemicals GmbH — Spin-coating of Photoresists (technical document). Thickness ∝ 1/√(spin speed); doubling the speed thins the film by ≈ 1.4×; product names encode thickness at 4000 rpm in 100 nm units. See also their development document (NaOH, KOH and 2.38% TMAH developers).
  5. Meyerhofer, D. (1978), "Characteristics of resist films produced by spinning", Journal of Applied Physics 49, 3993–3997. doi:10.1063/1.325357
  6. Mack, C. (2007), Fundamental Principles of Optical Lithography, Wiley. doi:10.1002/9780470723876 — resolution R = k₁λ/NA, the Airy disc, and the imaging models behind Chapter 5.
  7. Love, J. C., Wolfe, D. B., Jacobs, H. O. & Whitesides, G. M. (2001), "Microscope Projection Photolithography for Rapid Prototyping of Masters with Micron-Scale Features for Use in Soft Lithography", Langmuir 17, 6005–6012. doi:10.1021/la010655t
  8. ASML — EUV lithography systems. 13.5 nm light; NXE systems NA 0.33, 13 nm resolution; EXE "High-NA" systems NA 0.55, 8 nm resolution, supporting high-volume manufacturing in 2025–2026.
  9. Floyd, R. W. & Steinberg, L. (1976), "An adaptive algorithm for spatial greyscale", Proceedings of the Society for Information Display 17(2), 75–77.
  10. NIOSH Pocket Guide to Chemical Hazards — Hydrogen fluoride. IDLH 30 ppm; skin absorption; eye and skin burns; bone effects.
  11. Lin, C.-C., Yang, C.-C., Ger, J. & Deng, J.-F. (2010), "Tetramethylammonium hydroxide poisoning", Clinical Toxicology 48, 213–217. doi:10.3109/15563651003627777 — 13 cases in Taiwan, three deaths after 25% TMAH; severe effects after 2.38% TMAH.
  12. ICNIRP (2004), "Guidelines on limits of exposure to ultraviolet radiation of wavelengths between 180 nm and 400 nm (incoherent optical radiation)", Health Physics 87, 171–186. doi:10.1097/00004032-200408000-00006