Industries / Technology / Electronics and semiconductors / Semiconductor equipment
Deposition
Deposition machines add thin films of metal, insulator or crystal to a wafer, often only a few atoms thick, by reacting gases on the surface (chemical vapour deposition, CVD, and atomic layer deposition, ALD), by knocking atoms off a solid target (physical vapour deposition, PVD, also called sputtering) or by plating metal out of a liquid (electroplating).
6
Milestones achieved, from 1974 to 2012.
3
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1974
Atomic layer epitaxy patented
Tuomo Suntola's team at the Finnish instrument maker Instrumentarium fed the ingredients of a compound to a surface one at a time; each pulse reacts until the surface is covered and then stops by itself, so the film grows one layer per cycle. The Finnish patent application, naming Suntola and Jorma Antson, was filed on 29 November 1974, and pilot production of thin-film electroluminescent displays (flat screens whose film glows under voltage) began at Lohja in 1983. The method was later renamed atomic layer deposition (ALD), and in 1999 Suntola's company Microchemistry was sold, with its staff and patents, to the Dutch toolmaker ASM.
LessonSuntola thinks nobody infringed the patent because the method was seen as too slow for practical use, yet it entered chip production decades later. Riikka Puurunen's history notes that the principle was also found in the Soviet Union in the 1960s as "molecular layering".
Credit
Tuomo Suntola and Jorma Antson; Instrumentarium, Lohja, Microchemistry and ASM
1998
Copper plating gets a machine
IBM's move to copper wiring needed copper laid into trenches cut in the insulator. Novellus launched SABRE, an automated machine that electroplates copper onto wafers from a liquid bath, after a joint development programme with IBM's Microelectronics Division; Wired reported a starting price of $3 million. Novellus said it filled trenches narrower than 0.15 micron without voids and kept the bath off the back of the wafer.
Lessonthe machine for a new process was built alongside the first chipmaker to use it.
Credit
Novellus Systems, with IBM Microelectronics
2003
Spin-on insulator loses out
Faster chips need insulators with a low dielectric constant ("low-k") between their wires, which cuts signal delay. IBM had shown manufacturing success with Dow Chemical's SiLK, a polymer spun onto the wafer like paint, but in September 2003 it chose a low-k film made by CVD for its 90 nm chips, saying the economics meant "a new technology needs broad support to make it viable".
Lessona material that worked in one company's fab lost because the rest of the industry backed another method, and the job stayed on deposition machines.
Credit
IBM Microelectronics; Dow Chemical
2007
ALD enters chip production
Intel's 45 nm processors went on sale in November 2007 with a gate insulator made with hafnium; the gate is the switch of each transistor, and the insulator is the thin film beneath it. The teardown firm Chipworks cut open an Intel Xeon E5410 and found about 1.1 nm of silicon dioxide plus 1.2 to 2.0 nm of hafnium-based insulator, and Physics Today quoted Raj Jammy crediting ALD as critical to such gates. ASM said in July 2008 that it had delivered its 100th Pulsar ALD module, which it said was running in volume production for these gates in over 30 fabs.
Credit
Intel; ALD toolmakers including ASM
2010
Solar factory lines withdrawn
Applied Materials had sold SunFab, complete production lines for thin-film solar panels built on its deposition know-how. In July 2010 it stopped selling SunFab lines to new customers, offered single CVD and PVD machines instead and expected charges of about $375 million to $425 million, saying the thin-film market had been hurt by several factors, among them competition from crystalline silicon panels.
Lessonselling whole factories tied the toolmaker to the fate of one panel technology.
Credit
Applied Materials
2012
Deposition and etch combine
Lam Research, a maker of etch and wafer-cleaning machines, completed its merger with Novellus, a maker of deposition machines, on 4 June 2012. Novellus holders received 1.125 Lam shares for each share and ended up owning about 43% of the combined company.
Credit
Lam Research; Novellus Systems
Watch for
Molybdenum transistor contacts
Contacts are the tiny metal plugs that join each transistor to the wiring above it. Applied Materials says its Centris Spectral Mo ALD system lays down single-crystal molybdenum only where contacts go, replacing tungsten, that it lowers contact resistance by up to 15% against Applied's own tungsten tool, and that leading logic chipmakers are adopting it at 2 nm and below.
Achieved whena teardown of a chip on sale finds molybdenum contacts, or a named chipmaker says it uses them in production.
Claimed by
Applied Materials, February 2026
Films that grow selectively
Area-selective deposition grows a film only on chosen surfaces, which saves patterning and etching steps. ASM says molybdenum ALD and area-selective deposition have entered high-volume production for the 2 nm generation; in April 2026 Applied announced a selective silicon nitride tool, and an analyst at Futurum wrote that "the commercial execution will determine the rest".
Achieved whena teardown or a chipmaker's paper confirms a selectively deposited layer in a chip on sale.
Claimed by
ASM International, March 2026; Applied Materials, April 2026
Thick insulator for chip stacks
Lam says its VECTOR TEOS 3D fills the gaps between stacked dies with insulating film up to 60 µm thick, that it is the first tool to lay down crack-free films over 30 µm in a single pass, that it is nearly 70% faster than Lam's earlier tools, and that it is installed at unnamed logic and memory fabs.
Achieved whena named packaging customer or an outside lab confirms crack-free single-pass films over 30 µm in production.
Claimed by
Lam Research, September 2025
Open
01
ALD growth-per-cycle table
An open table of the 30 or so ALD processes most used in chipmaking, such as aluminium oxide from trimethylaluminium and water, hafnium oxide, titanium nitride, silicon oxide and nitride, molybdenum and ruthenium. Each row gives growth per cycle (nm per cycle), the temperature window in which growth stays steady (°C) and the dose needed to cover the surface, tied to a paper's DOI and figure or table number, with a short script to query it.
The briefCredit · unclaimed02
Open precursor vapour pressures
Measured vapour-pressure curves (raw points, fitted equations, valid temperature range and measuring method) for about 40 common ALD and CVD precursors, the chemicals that carry each element into the reactor, plus a Python calculator that estimates how much precursor each pulse delivers from source temperature, carrier-gas flow and pulse time. This is data and software only; no chemicals are handled.
The briefCredit · unclaimed
Open 01 · the brief
ALD growth-per-cycle table
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
An open table of the 30 or so ALD processes most used in chipmaking, such as aluminium oxide from trimethylaluminium and water, hafnium oxide, titanium nitride, silicon oxide and nitride, molybdenum and ruthenium. Each row gives growth per cycle (nm per cycle), the temperature window in which growth stays steady (°C) and the dose needed to cover the surface, tied to a paper's DOI and figure or table number, with a short script to query it.
Why it is still open
A January 2026 paper by ALD researchers and the Open Research Knowledge Graph team says review data are "locked in narrative text and static PDF tables", and that even asking which processes fit a temperature window and growth-per-cycle range takes "substantial manual effort"; they converted 18 tables from nine reviews, three of them on ALD. The crowdsourced AtomicLimits ALD database, started in 2019, records materials, reactants and papers, and its first-year review does not say it stores growth values.
Built with what exists
Python and pandas; WebPlotDigitizer to read values off published graphs; Zotero; the Open Research Knowledge Graph for publishing; Zenodo for a DOI; an AI model to draft rows from papers for a person to check.
First step
For aluminium oxide from trimethylaluminium and water, extract growth per cycle against temperature from 10 papers linked in the AtomicLimits database and post the CSV.
Done looks like
at least 30 processes, each with values from at least three independent papers; a second person re-extracts a random 10% and agrees within 10%; released under CC BY 4.0 with a DOI.
Open 02 · the brief
Open precursor vapour pressures
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
Measured vapour-pressure curves (raw points, fitted equations, valid temperature range and measuring method) for about 40 common ALD and CVD precursors, the chemicals that carry each element into the reactor, plus a Python calculator that estimates how much precursor each pulse delivers from source temperature, carrier-gas flow and pulse time. This is data and software only; no chemicals are handled.
Why it is still open
The software firm Schrödinger says datasets for predicting precursor volatility are "sparse and incomplete" and that "commercial sensitivity sometimes means that data are partially withheld"; it built its own set from the literature, and its page does not offer that set publicly. NIST's ThermoML archive holds scattered machine-readable records, such as a 2006 study of metal-organic precursors by Fulem and co-authors.
Built with what exists
NIST ThermoML files; Python with NumPy and SciPy for curve fitting; WebPlotDigitizer; Zenodo.
First step
Download the ThermoML records for metal-organic precursors and fit the Fulem data.
Done looks like
40 precursors, each with at least one measured dataset (not a supplier brochure); every fit within 5% of its source's points over the stated range; the calculator matches hand calculations for three worked cases.