Criteria

These are the rules for what belongs on this list. They are the chemistry reading of the rules used for the Millennium Problems for Biology (opens in a new tab), as stated with that project and in its announcement (opens in a new tab).

A problem that fails one of them does not belong, however important it is.

Inclusion

A problem is included only if all of the following are true.

• It is very hard. A careful group with today’s methods should not already be able to clear it by a routine experiment.
• It is easy to grade. A standard chemistry laboratory, without a unique instrument or a national facility, can say whether the attempt passed. The experiment fits in days, or at most a couple of weeks.
• The grade is as objective as the subject allows. Prefer a number, a named control run side by side, and a preregistered target. Where a human still has to look at a spectrum and assign a structure, a second person does it from the same packet, and the rule for agreement is written down.
• A person with a basic chemistry background can understand what is being asked, even if they could not attempt it.
• Solving it would change practice in drug discovery, energy, materials, climate, or synthesis.
• It can serve as a last reasonable test of a proposed method, including a method that uses a model. The method may be new. The measurement may not be.

Exclusion

Leave a problem off the list when any of the following is true.

• Success is a matter of taste, or of whether a story about a mechanism is satisfying. “Explain why this catalyst works” is not a problem in this sense. “Make this product, under these conditions, past this control” is.
• Grading takes a specialized facility, a years-long trial, or a sample that cannot be made twice.
• The statement is only meaningful to a narrow subfield, so that an outsider cannot tell what would count.
• The quantitative bar is a number that nobody can source, presented as if it were a measured record. A pass mark that the problem itself defines is allowed. It must be labeled as a pass mark. A comparison with a named control, measured in the same experiment, is preferred whenever the absolute number is uncertain.
• The problem is already solved on the terms written down. When a later paper clears a bar, the honest act is to revise the bar or retire the problem, not to pretend the paper does not exist.

How a problem is written

Each problem is one Markdown file in problems/. The file name is the id, a hyphen, and the slug: 04-ambient-ammonia.md.

The frontmatter carries:

• id, a two-digit string
• slug
• title
• field, one of Synthesis, Catalysis, Energy, Materials, Spectroscopy, Molecular design, Polymers
• one_liner, a single sentence
• status, one of draft, open, revised, retired
• contributors, optional

The body has these sections, in this order, each as a level-two heading: Claim, Specification, Pass threshold, Controls, Allowed, Not allowed, Partial success, Sources.

The pass threshold says what number is required and whether that number is a pass mark defined here or a figure taken from a cited measurement. The controls name the blank, the reference material, or the isotope experiment. Allowed and not allowed are specific enough that two labs would make the same call. Partial success is a real, lesser outcome, not a compliment.

docs/PROBLEM_TEMPLATE.md is the empty shape.

Grading

The people who grade an attempt did not design it. Targets, substrates, and the analysis plan are preregistered before the attempt sees them, wherever the problem says so. A procedure is run as written. A failed step is a failed attempt, not an invitation to edit the procedure and try again, unless that problem explicitly allows a second written procedure submitted at the same time.

Isotope labels, empty reactors, vehicle wells, and side-by-side reference materials are part of the grade. An impressive yield with a failed control is a failure.

Status

draft means the specification is public and the thresholds are open to revision. open means the wording has settled and the problem is unsolved. revised means a real change landed after discussion. retired means the problem should no longer be attempted on these terms; the file stays so the reason is visible.

The twelve problems in this repository are drafts.