BRENNEN

Industries / Mining and materials / Chemicals and petrochemicals

Reactions and catalysts

This is the work of turning one set of molecules into another at scale, usually with a catalyst, a material that speeds a reaction without being used up.

5

Milestones achieved, from 1913 to 2020.

1

To watch for: announced, not yet confirmed outside.

2

Open deliverables. Nobody has shipped them.

The road

AchievedWatch forOpen

Achieved

  1. 1913

    Ammonia from the air

    BASF's first ammonia synthesis plant went into operation at Oppau, north of Ludwigshafen, combining nitrogen and hydrogen at high pressure. BASF says this made industrial fertilizer production possible and made it the first company to use high-pressure technology.

    basf.com

    Credit

    Carl Bosch and a BASF team, building on laboratory work by Fritz Haber

  2. 1920

    The first petrochemical

    Standard Oil's plant at Bayway, New Jersey made isopropyl alcohol, which the American Fuel & Petrochemical Manufacturers calls the first commercial petrochemical, meaning a chemical made from oil or natural gas.

    afpm.org

    Credit

    Standard Oil

  3. 1953

    Plastics without high pressure

    Karl Ziegler's team found that metal-organic compounds such as aluminium alkyls could make polyethylene without high pressure and temperature. In 1954 Giulio Natta used such catalysts to make stereoregular polypropylene, whose chains have a uniform shape. The Science History Institute says the work made common plastics such as high-density polyethylene and polypropylene possible.

    sciencehistory.org

    Credit

    Karl Ziegler and Giulio Natta, who shared the 1963 Nobel Prize in Chemistry

  4. 2002

    Runaway reactions counted

    On 17 September 2002 the US Chemical Safety Board closed a two-year study of 167 serious US accidents from 1980 to 2001 caused by uncontrolled chemical reactions, which killed 108 people. More than 90% involved hazards already described in public literature. The board found information on causes for only about 20% of the incidents and called its count almost certainly too low.

    LessonA hazard that is written down can still be missed if nobody finds it before the work starts.

    csb.gov

    Credit

    US Chemical Safety and Hazard Investigation Board

  5. 2020

    Open catalyst data

    Facebook AI Research and Carnegie Mellon University released OC20, a public set of over 1.3 million computed relaxations of molecules on catalyst surfaces. A relaxation finds the resting position of a molecule on a surface, and each can take hours or days of quantum-mechanical computing; the aim was to train AI models that give the answer in seconds.

    ai.meta.com

    Credit

    Facebook AI Research and Zachary Ulissi's group at Carnegie Mellon

Watch for

  1. An AI-run lab screens catalysts

    A preprint reports that an AI-guided high-throughput laboratory screened 2,942 catalysts across 53 material systems, looking for a replacement for scarce iridium and ruthenium in the oxygen-producing half of acid water electrolysis, which splits water into hydrogen and oxygen. Its best material, a palladium oxide with indium and manganese, kept its overpotential (the extra voltage the reaction needs) below 0.5 V for 1,000 hours in 1 M sulfuric acid at 10 mA/cm².

    Achieved whena laboratory not connected to the authors reproduces the 1,000-hour result, or the material runs in a full electrolyser stack with results published by someone else.

    arxiv.org

    Claimed by

    the preprint's 22 authors, September 2026; press reports link the work to Lila Sciences

Open

  1. 01

    Turn catalyst papers into data

    An open table of measured results for one industrial reaction, such as propane dehydrogenation (making propylene from propane), taken from published papers by a document-reading model and checked by people. Each row gives the catalyst, the conditions, how much feed was converted and how much became the wanted product, with a link to the page it came from.

    The brief
    Credit · unclaimed
  2. 02

    Update the runaway-reaction record

    An open, coded list of serious accidents caused by uncontrolled chemical reactions since 2002, built only from public investigation reports, using the same fields as the Chemical Safety Board's 1980 to 2001 dataset, with each field linked to its source page.

    The brief
    Credit · unclaimed

Open 01 · the brief

Turn catalyst papers into data

Credit · unclaimed

The first to publish is credited on this line, and the deliverable moves from Open to Achieved.

What to build

An open table of measured results for one industrial reaction, such as propane dehydrogenation (making propylene from propane), taken from published papers by a document-reading model and checked by people. Each row gives the catalyst, the conditions, how much feed was converted and how much became the wanted product, with a link to the page it came from.

Why it is still open

The open benchmark for experimental catalysis run by the University of Minnesota's Center for Programmable Energy Catalysis, CatTestHub, holds a short, growing list of chemistries, measured by its own members or curated by them from the literature.

cpec.umn.edu

Built with what exists

A model built for extraction and long-document parsing, such as Meta's Muse Ember (Notes: "The jobs agents repeat"); open-access papers; CatTestHub's data layout.

  1. First step

    Pick 20 open-access papers on the reaction, copy their result tables by hand, and use them as the answer key.

  2. Done looks like

    At least 3,000 records from at least 300 papers. A person checks 200 randomly chosen records against the papers and finds at least 95% of fields correct. The table is released under an open licence in CatTestHub's layout and offered to its maintainers.

Open 02 · the brief

Update the runaway-reaction record

Credit · unclaimed

The first to publish is credited on this line, and the deliverable moves from Open to Achieved.

What to build

An open, coded list of serious accidents caused by uncontrolled chemical reactions since 2002, built only from public investigation reports, using the same fields as the Chemical Safety Board's 1980 to 2001 dataset, with each field linked to its source page.

Why it is still open

The board released its dataset in September 2003 with a warning that the sources it drew on were inadequate to identify how often reactive incidents happen and why.

csb.gov

Built with what exists

Chemical Safety Board reports and records from eMARS, the EU's database of major accidents, run since 2026 by the European Environment Agency; an agent built with a kit such as Cursor's Agent SDK that cites the source behind every finding (Notes: "Findings that cite notes").

industry.eea.europa.eu

  1. First step

    Code five Chemical Safety Board reactive-incident reports by hand to fix the field list.

  2. Done looks like

    At least 300 incidents coded. Two people code 50 of them independently, and the tool matches them on at least 90% of fields.