Construction Classics

Studies with a Three-Dee Three-Phase Proton Cyclotron

Why this document can be hosted

UCRL-3187 is a Radiation Laboratory report prepared under AEC contract W-7405-eng-48 and printed for the US Atomic Energy Commission. AEC contractor reports are treated as US Government works and carry no copyright; pre-1978 distribution without a copyright notice is an independent public-domain basis. The scan carries the contract number and the government-sponsorship disclaimer, and no copyright notice, publisher imprint, or journal-reprint marking appears anywhere in it. The one Physical Review mention is a citation in the reference list, not a marking on this report. Verified against the scan and the OSTI record; determination made August 2026.

Reading guide

Page references are the hosted PDF’s pages.

Of everything the August 2026 repository search turned up, this is the closest match in size to the machines this site’s readers build. It is an 18-page report on a 20-inch cyclotron — 8.5 inches of usable field radius, protons to 1 MeV — written not as a facility description but as an account of getting an awkward machine to work.

Its first sentence of engineering is the one worth stealing: The machine was designed around a magnet with 20-inch diameter pole pieces and a C-type yoke that was already available (p. 8). Everything downstream follows from that constraint, which is the constraint an amateur builder starts from too. The vacuum tank was built around the magnet, incorporating the pole tips as part of the vacuum wall, and made hexagonal to match the threefold symmetry of the RF: three sides carry the dee-stem tanks, one goes to a 14-inch oil diffusion pump manifold, and the remaining two provide viewing ports and feedthroughs for the ion source and probes.

Three dees, three modes, and why that is hard

The three-dee three-phase arrangement is the machine’s reason for existing. With three 60° dees driven 120° apart, the same frequency and magnetic field accelerate different particles on different RF modes — protons on the forward mode, deuterons on the reverse mode, tritons on the neutral mode (table, p. 6). One machine setting, three species. This is the small sibling of the cloverleaf programme whose RF system is documented in the already-hosted UCRL-3153, and the two should be read together: UCRL-3153 is the RF topology in general, UCRL-3187 is what happened when it was built at 20 inches.

The honest part is the difficulty. Stable operation in the intended mode with balanced dee voltages needed both phase servos and amplifier-efficiency servos, and the dees could not be servoed individually at all until the interdee capacitance was neutralised. That is a warning worth carrying into any multi-dee design: the dees are coupled whether or not the design acknowledges it.

Mechanically the RF is unusually legible. Each dee is a 60° sector on a foreshortened quarter-wave stem, with 2 inches of vertical clearance inside the dee and a 1-inch dee-to-liner gap; the dee tips are demountable so the central geometry can be changed. Zircon insulators on top of the dee-stem tanks do double duty as the main mechanical support and as the vacuum seals. Coarse tuning is a movable ground plane about three feet up the stem; fine tuning is a 100–200 pF air capacitor. When ion loading pulled the system off tune, the crew retuned by exciting each dee-stem tank with a grid-dip oscillator and adjusting for resonance — a bench technique, on a cyclotron.

The result that transfers: move the ion source off centre

The ion-source section (p. 12) is the most directly useful in the report. The source began at the geometric centre, in the conventional position, and was tried with a one-slot hood, a three-slot hood, an unshielded arc, and an open ring-shaped arc with a centre post. The best of those gave 3.2 mA of protons at the 8-inch radius, and caused considerable heating of the dee tips.

Replacing it with an off-centre source injecting azimuthally into one of the dees — hooded-arc type, exit slit 1/8 by 3/4 inch, at a radius of about 1¾ inches, patterned after sources developed at Oak Ridge for the 12-inch and 86-inch machines — roughly doubled the beam, to a routine 6–7 mA. The report calls it a major improvement, and it removed the dee-tip heating problem along with it. Beam current rose linearly with peak RF voltage and with total DC amplifier power, and the beam load raised amplifier plate currents by a factor of two to three over the source-off condition — a useful sanity check that the beam is real.

A measurement cross-check worth copying

The beam was measured two independent ways, and the report says what happened when they disagreed. Against an unshielded probe, the calorimetric method read 90% of the probe value at 1–2 mA and 95% in the 6–8 mA range; the crew judged the probe more reliable. They also checked the probe for secondary-electron error by biasing it: at 8 inches a +450 V bias changed nothing, but inside 6 inches the probe current rose steeply with decreasing radius and responded to the bias — so the probe reading is trustworthy at the outer radius and not at the inner one.

Two independent methods, a stated discrepancy, a stated reason for preferring one, and a bias test that maps where the instrument can be believed. That is the measurement practice this collection exists to surface, and it costs an amateur nothing but patience. Compare the target-plane methods in the hosted target reports, which measure the same beam a different way.

Two closing details. A system of slits let the crew programme the beam for two complete turns and photograph the starting orbits — but the slits ran cool over only a narrow range of RF voltage. And the orbit centre was found displaced 3/4 inch from the geometric centre, with the dee-voltage variation in the orbit figure attributed to imperfect neutralisation. Both are reminders that the central region is where small machines are won or lost.

One caution on the numbers

The beam currents here are internal currents on a probe, in milliamperes, from a programme whose explicit goal was high intensity — the report distinguishes them from microampere readings elsewhere in the same paragraph. They are not extracted beam, and they are not what a first amateur machine will produce. Read the report for its geometry, its source result, and its measurement discipline rather than as a performance target.

Design rules extracted from this document

8 Design Guide rules cite UCRL-3187, most of them ion-source geometry and beam-measurement discipline. Representative examples:

• dg-734 — move the ion source off centre and inject azimuthally into a dee
• dg-735 — cross-check probe current calorimetrically, and expect the discrepancy to grow with current
• dg-736 — use a positive probe bias as a purity test — accept the reading only where it is bias-flat
• dg-741 — do not use amplifier efficiency as a proxy for dee phase
• dg-740 — expect thermal detuning plus "ion lock" after shutting down from high power

Used by the ion sources and beam measurement deep dives.