E-beam lithography writes nanoscale patterns directly with a focused electron beam instead of projecting light through a mask, and it remains the default way research groups fabricate sub-200 nm photonic, quantum and wide-bandgap prototype devices, while being too slow and costly to be a volume production route.
Summary
Electron-beam lithography (EBL) patterns a resist-coated wafer by scanning a focused electron beam across it point by point. Because the de Broglie wavelength of accelerated electrons is far below optical wavelengths, resolution is not set by diffraction in the way it is for projection optics, and no mask is required: the pattern is written directly from the layout file. That makes EBL a maskless, per-design tool rather than a per-wafer tool, which is exactly why it dominates prototyping and is largely absent from high-volume manufacturing, where the economics favour projecting one mask onto many wafers. The mainstream production counterpart is deep-ultraviolet and extreme-ultraviolet projection, where a 13.5 nm source and numerical apertures of 0.55 and above are used to keep shrinking logic features ref.
The parameters that decide EBL are resolution, beam placement accuracy, throughput and cost. Resolution is rarely the binding constraint in the 2026 literature; placement is. A GaAs photonic-crystal split cavity reported in August 2026 achieved average quality factors of 20,000 or more, enough for the strong-coupling regime, and the authors state explicitly that Q was limited primarily by the precision and accuracy of the e-beam lithography tool rather than by sidewall scattering. That is a useful marker: for high-Q nanophotonics, the writing tool itself is now the error budget.
What EBL buys is arbitrary geometry at nanometre scale without a mask set. Recent uses span deterministic ion placement for optical data storage, where EBL plus ion implantation gave controlled ion number and spatial distribution with multi-bit grayscale encoding across millimetre-scale areas; scaled 200 nm multi-fin channels in a vertical gallium oxide transistor with a threshold voltage of 2 V and on-off ratio above 1e7; and silicon photonic inverse design flows that treat EBL and photolithography as two distinct fabrication platforms with different design rules across the 1,500 to 1,600 nm band.
What EBL costs is money and time, and this is visible in the same literature as a push to avoid it. A graphene nanoribbon FET paper argues that high fabrication cost and complex processing hinder integration of GNR devices, and offers electrochemical AFM lithography with AC bias as a relatively low-cost direct-write route to sub-10 nm features explicitly positioned against photo- and e-beam lithography. A tantalum pentoxide microring paper reports loaded Q of 2.74e6 in the telecom band and intrinsic Q above 4e6 using photolithography-assisted chemo-mechanical etching, and states the result was obtained without employing expensive electron-beam lithography. EBL is therefore both the incumbent benchmark and the thing new nanofabrication methods advertise themselves against.
Evidence base
- 4 Aug 2026: GaAs photonic crystal split cavity reached average Q of 20,000 or more, with Q limited primarily by the precision and accuracy of the e-beam lithography tool rather than sidewall scattering.
- 14 Aug 2026: vertical gallium oxide ISEFET used EBL to define 200 nm multi-fin channels etched to ~1.2 um trench depth, giving enhancement-mode operation at 2 V threshold and on-off ratio above 1e7.
- 29 May 2026: EBL combined with ion implantation gave deterministic control of ion number and spatial distribution for multi-bit grayscale and wavelength division multiplexed optical storage over millimetre areas.
- 14 May 2026: graphene nanoribbon FET work cites high fabrication cost and complex processing of conventional photo- and e-beam lithography, offering AC-bias electrochemical AFM lithography for sub-10 nm features instead.
- 7 Jun 2026: monolithic Ta2O5 microrings reached loaded Q of 2.74e6 in the telecom band and intrinsic Q above 4e6 without employing expensive electron-beam lithography.
- 7 Jun 2026: DRC-compliant inverse design treated electron-beam lithography and photolithography as separate fabrication platforms for splitters, duplexers and mode converters across 1,500 to 1,600 nm, with over 5-fold lower computational cost than prior methods.
- 27 Jul 2023: mainstream leading-edge patterning is framed around 13.5 nm EUV light and numerical apertures of 0.55 and above, not direct-write methods ref.
Open questions
- How much of the beam placement error that capped split-cavity Q at around 20,000 is fixable by better tooling versus intrinsic to serial writing?
- Can electrochemical AFM direct write scale beyond single sub-10 nm graphene devices to multi-layer, registered process flows, or does it remain a single-layer alternative?
- For which photonic geometries does chemo-mechanical etching or photolithography match EBL performance, given that Ta2O5 microrings reached intrinsic Q above 4e6 without it?
- Do inverse-lithography and DRC-compliant design flows reduce the number of EBL fabrication iterations enough to change the cost calculus for photonics groups?
Assessment drafted 2026-08-31 from up to 15 KB sources using the technology-scorecard framework; scores are a draft read pending review.