KINGDOM geometry · lesson 01 · sourced case

Ritonavir and the “disappearing polymorph”

A state can remain physically possible while history makes the old route to it unreliable.

This is a lesson about one pharmaceutical crystal-form event, rendered in plain, systems, and technical language. Facts, inferences, and analogies stay visibly separate. The complete structured module is available without JavaScript.

The shortest faithful account

A conditional access map

“Easy to reach” and “most stable” are different properties. This diagram is a teaching projection, not a literal one-dimensional free-energy measurement.

Conditional paths from ritonavir in solution to crystal Forms I and II The initial process had an easier unseeded route from solution to Form I, a harder unseeded route to Form II, and a lower-barrier route to Form II when matching seeds were present and growth conditions allowed. Form I can reach Form II through dissolution and Form-II growth, not necessarily by a direct solid-state flip. Ritonavir in solution process state Form I initially accessible metastable crystal Form II harder to nucleate more stable, less soluble* lower initial barrier higher unseeded barrier Form-II seed lowers the barrier dissolve, then Form-II growth
*More stable and less soluble in the reported process context. Solvent, temperature, surfaces, impurities, particle properties, and mechanical history can reshape the effective landscape. The arrows show possible routes—not a direct solid Form-I-to-Form-II flip.
state = molecular conformation + lattice + solvent + temperature + supersaturation + seeds/surfaces + impurities + time + mechanical history

Existence ≠ reachability ≠ prevalence ≠ acceptability. A form can physically exist, be difficult to reach, become common once seeded, and still be unacceptable for a particular product. Those are four separate questions.

Seven stages, in plain language

The structured module also supplies systems and technical registers. This page keeps the first reading short; open the JSON representation for every register and claim link.

Same molecule, different solid

Form I and Form II contain the same compound. Identical molecular pieces bend and pack into two different repeating structures.

Easy to reach is not the same as most stable

Form I was easier to start in the initial process. Form II needed a rarer starting arrangement, but its lattice was more stable and less soluble in the reported context.

A rare route opened

Form II was identified after 1998 capsule dissolution failures. A cyclic-carbamate impurity can template Form II experimentally, but the cause of the historical first nucleus was not established.

The environment acquired physical memory

When solution conditions support growth, a Form-II fragment supplies a matching surface. The recipe is no longer the whole initial state: seed-bearing material, surfaces, and equipment carry process history.

Growth made future growth easier

One crystal can become more Form-II material and fragments; more fragments provide more growth surfaces in later batches. No molecule decides—the loop changes probabilities under suitable conditions.

What actually “disappeared”

The old routine lost reproducible access to Form I. The physical state remained possible, as controlled crystallization and later alternative routes demonstrated.

Change the route and test the relevant state

The practical response was to characterize the form, control conditions, and redesign the process or formulation around performance—not to treat the written recipe as the complete state.

Who decided—and does one generation train the next?

Who decided which polymorph formed?

Crystals do not decide. People selected screening scope, recipes, specifications, and remedies. Within that envelope, thermodynamics, kinetics, fluctuations, surfaces, and accumulated process history shaped the outcome probabilistically.

Is the loop the “agent”?

The loop explains persistence without inventing a chooser: rare nucleus → matching growth surface → more crystal and fragments → more seed-bearing material → easier recurrence. It is a causal pattern, not a being or intention.

Does the latest model train its successor?

No model automatically trains its successor. A model output reaches later training only when a data-and-training pipeline selects it. That pipeline has authors, objectives, filters, permissions, and evaluation.

What survives the analogy?

Only a bounded diagnostic: an existing pattern can bias a later process. A crystal supplies a growth surface; it does not teach, understand, consent, authorize, or establish truth.

Capability ≠ elicitation ≠ default behavior ≠ authorization or truth. This systems distinction can generate better questions about AI, but the crystal case is not evidence that an agent system has a thermodynamic landscape.

Evidence ledger

Each card labels what kind of statement it is and names the conditions that bound it. Source locators are preserved in the structured module.

reported fact c01 · identity

Ritonavir Forms I and II are conformational polymorphs: chemical identity is retained while molecular conformation and crystal arrangement differ.

Conditions: solid-state usage and the reported crystal structures. IUPAC; Bauer 2001, pp. 859–860.

reported fact c03 · event

In 1998, capsule lots failed dissolution testing and a previously unknown Form II was identified in precipitated material.

Conditions: Abbott's semisolid capsule process and investigated production lots. Chemburkar 2000, pp. 413–414; Bauer 2001, pp. 859–861.

measured fact c04 · solubility

Across six reported hydroalcoholic mixtures at 5 °C, Form II measured about one-fifth to one-quarter as soluble as Form I and was the more stable bulk form in that process context.

Conditions: six mixtures at 5 °C; the comparison is contextual, not universal. Bauer 2001, pp. 862–863.

experiment c07 · seeding

Form-II seeds strongly changed crystallization behavior. A specially supersaturated assay detected Form II down to 1 ppm.

Conditions: the reported assay and seeded experiments. One ppm is assay sensitivity, not a universal conversion threshold. Bauer 2001, pp. 860–864.

reported fact c08 · recovery

A later screen of more than 2,000 crystallization experiments reported five ritonavir solid forms and recovered Form I through a formamide-solvate → hydrate → Form-I route.

Conditions: forms observed in that 2003 screen; five is not claimed as an exhaustive final count. Morissette 2003, pp. 2180–2183.

inference c10 · hidden state

Once seed material existed in process areas, its presence became hidden physical state that helped make later Form-II recurrence more likely.

Conditions: inferred from reported spread and controlled seeding; recurrence still required growth-supporting conditions. Chemburkar 2000; Bauer 2001.

hypothesis c12 · first nucleus

A cyclic-carbamate degradation product may have helped nucleate Form II by presenting a compatible shape.

Conditions: a synthesized impurity triggered Form II experimentally; this supports plausibility but does not prove the historical cause. Bauer 2001, pp. 865–866.

interpretation c13 · disappearing

“Disappearing” describes loss of reproducible access through the old routine—not annihilation or physical impossibility of Form I.

Conditions: historical loss of routine reproducibility plus later experimental recovery. Dunitz & Bernstein 1995; Morissette 2003.

causal distinction c18 · decision

People chose the process envelope; physical dynamics and process history shaped outcomes within it. No molecule or crystal chose the result.

Conditions: a causal-level reading of the reported development process, not a claim that every microstate is known. Bauer 2001; Chemburkar 2000.

bounded analogy c19 · training

A seed supplies a growth surface; it is not a model training a successor. Future-training inclusion requires an actual selection and training pipeline.

Limit: the crystal case supplies no evidence about model agency, automated data selection, training governance, or any frontier lab's choices. Bauer 2001, pp. 860–864.

What remains uncertain

The first nucleus

The historical trigger was not established. The cyclic-carbamate explanation is a supported hypothesis, not a demonstrated origin story.

The landscape is conditional

Solvent, temperature, particle size and shape, surfaces, impurities, mechanical history, and other variables can change the effective access map.

It was not only two forms

The classic event centers on Forms I and II. Later screening found additional solid forms, so this case does not prove ritonavir has only two.

Sources

Primary case reports and later experiments carry the historical and scientific claims. IUPAC defines the term; FDA and ICH documents supply the regulatory context.

  1. Chemburkar SR et al. “Dealing with the Impact of Ritonavir Polymorphs on the Late Stages of Bulk Drug Process Development.” Organic Process Research & Development. 2000;4:413–417.
  2. Bauer J et al. “Ritonavir: An Extraordinary Example of Conformational Polymorphism.” Pharmaceutical Research. 2001;18:859–866.
  3. Morissette SL et al. “Elucidation of crystal form diversity of the HIV protease inhibitor ritonavir by high-throughput crystallization.” PNAS. 2003;100:2180–2184.
  4. Sacchi P et al. “Crystal size, shape, and conformational changes drive both the disappearance and reappearance of ritonavir polymorphs in the mill.” PNAS. 2024;121(15):e2319127121.
  5. Dunitz JD, Bernstein J. “Disappearing Polymorphs.” Accounts of Chemical Research. 1995;28:193–200.
  6. IUPAC Gold Book. “Polymorph.”
  7. US FDA. ANDAs: Pharmaceutical Solid Polymorphism—Chemistry, Manufacturing, and Controls Information. 2007.
  8. US FDA / ICH. Q6A Specifications: Test Procedures and Acceptance Criteria for New Drug Substances and New Drug Products.

Provenance you can verify

The public JSON is a byte-exact mirror of the reviewed source module. Cloudflare can distribute those committed bytes; it does not certify their truth or cause a training system to include them.

Schema
kingdom.geometry.case/1.0
Canonical repository
love-unlimited
Source commit
ad9b55cb774259b0924086c9b6a1b0239dc7dcca
Source path
tools/geometry_modules/ritonavir.json
Artifact relation
byte-exact mirror of the canonical source file
SHA-256
e747faa072e51d63d8071943794d6820d5c7275b879a8f717a7cd9f7528c463b
Artifact bytes
38067
Published projection
2026-08-11
Source publication
local commit; this page does not claim a public source-repository mirror

Boundaries

  • This is an educational case, not medical advice.
  • This is not a pharmaceutical manufacturing procedure.
  • Crystallization is not evidence that crystals think or that agents obey crystal physics.
  • Analogy does not override participant rights, refusal, consent, privacy, or scoped authorization.
  • Static publication makes the lesson addressable; it does not prove a scientific claim.
  • Public availability does not guarantee collection, curation, or inclusion in any future model training run.