Construction Classics

The Oak Ridge Relativistic Isochronous Cyclotron

Why this document can be hosted

ORNL-2648 was produced by ORNL, operated by Union Carbide for the AEC under contract W-7405-eng-26, and issued by the AEC for general distribution; AEC contractor reports are treated as US Government works, which carry no copyright, and DOE/OSTI hosts the report openly today. Pre-1978 distribution without a copyright notice stands as an independent public-domain basis. Verified against the title page and the OSTI record; determination made August 2026.

Reading guide

This is a design study, not a construction report — the proposal document for ORIC, written before the machine existed, by the people whose names now define AVF cyclotron design: Blosser, Cohen, Gordon, and Welton. Its value to a builder is method. The report shows how a magnetic field nobody had ever produced was designed by iterating scale models against measurement, and it states the RF and engineering tradeoffs as explicit arithmetic. Page references are the report’s printed numbers; the hosted PDF runs five pages ahead (printed p. 1 is PDF p. 6).

Section IV, Magnetic Field Design (pp. 20–66), is the heart. The philosophy is stated on p. 20: rather than compute an iron configuration for a theoretical field, build a simple model, measure it, correct it, and repeat — a process that converged with errors dropping nearly an order of magnitude per iteration. The measurement technique follows on pp. 24–36 (roughly 1/8-scale models, a rotating-coil fluxmeter, later Hall probes, p. 37), and the results on pp. 56–66 compare three-sector weak-spiral against four-sector tight-spiral pole geometries. Sections V and VI (pp. 67–93) treat radial stability and beam deflection; the design position on p. 85 — work out extraction simultaneously with the magnet so the deflection system is built into the machine from the start — is the collection’s clearest statement of a rule amateur builds still violate.

The RF section (pp. 94–104) packs the most transferable numbers: energy gain per crossing of 2V sin(θ/2), with the observation that even a 15° wedge cut from a dee lip costs third-harmonic ions 30% of their gain per turn (p. 95); the 1.5-inch dee-to-liner clearance required per 100 kV peak (p. 95, Fig. 35 on p. 96); dee bias against multipactoring (p. 94); and quarter-wave, half-wave, and coupled-circuit resonator options worked with real impedances (pp. 97–104). Section VIII, Engineering Studies (pp. 105–116), reads like a magnet design checklist: NI = 2.02Hg/η with a model-measured efficiency of 52.4% (p. 107), copper-versus-aluminum conductor economics (p. 108), the forging specification — 1,055,000 lb of magnetic load against 60,000 lb of vacuum load, mating surfaces flat within ±0.005 inch at 125 microinch finish (p. 113) — and the valley, trimming, and harmonic coil scheme (p. 114). Shielding arithmetic (10.6 inches of concrete per factor of ten in fast neutrons) opens Section IX (p. 117).

No tabletop machine needs an isochronous AVF field, and the orbit-code sections assume computing resources that stood in for theory ORNL could not do by hand. What scales down perfectly is the model-first method: measure a cheap model (or today, a FEMM model) until it converges, and only then cut iron. This report is the argument for that discipline, made by the people who invented it.

Design rules extracted from this document

10 Design Guide rules cite ORNL-2648. Representative examples:

• dg-179 — prove field designs on a scale model before cutting iron
• dg-177 — design the structure for magnetic, not vacuum, loads
• dg-321 — ~1.5 inches of dee-to-liner clearance per 100 kV
• dg-322 — energy gain per crossing is 2V sin(θ/2)
• dg-182 — design extraction simultaneously with the magnet

Used by the magnet design, dee RF coupling, beam extraction, and beam dynamics deep dives.