Experiments by Energy Band

The case against building a cyclotron says there is not much to do with one afterward. This page makes that claim precise — band by band, from 5 keV to 1 MeV — and then makes the most of what survives the arithmetic. For each band: which interactions are physically open, which experiments a nanoampere beam can actually detect (with the count rates worked out), what detection hardware each needs, and where the safety and legal onsets sit. Every reaction threshold and Q-value below comes from the NNDC Q-value calculator, every range from NIST PSTAR, and every level energy and half-life from ENSDF — all retrieved August 2026.

Energy opens doors; current pays admission

Two numbers decide what a beam can do, and they play different roles. Kinetic energy sets the menu: a nuclear reaction with a threshold above the beam energy happens never — not rarely — and even reactions with no threshold at all hide behind the Coulomb barrier, which a low-energy proton crosses only by tunneling. Current sets the bill: every count rate below is strictly proportional to it. The trouble is that the two scale differently. Doubling current doubles a count rate. Doubling energy, in the sub-MeV regime, can multiply it by orders of magnitude — the fitted ¹¹B(p,α) yield coefficient grows by a factor of about 240 between 150 and 650 keV (Spraker et al. 2012, p. 360; dg-500). One well-documented fusor log points the same direction from the other side — D–D yield up two orders of magnitude as drive rose from 22 to 33 kV (Hull; dg-504) — one machine's record rather than a law, since fusor yield also moves with current, pressure, and geometry.

The historical benchmark for this page is Oliphant and Rutherford's 1933 Cavendish apparatus: a discharge tube delivering analysed proton beams of about 100 μA at up to 250 kV, which disintegrated lithium and boron targets at energies every machine in the build census can reach (Oliphant & Rutherford 1933). The energies are amateur energies. The current is the difference: 100 μA is one hundred thousand times the nanoampere a typical first amateur beam delivers (beam measurement). That factor of 10⁵ is what makes most of the bands below read as sparse — and what makes the few experiments that survive nanoampere arithmetic worth taking seriously.

What energy a given machine reaches is set by field and radius — the energy calculator covers that arithmetic. The census machines with measured beams span 24 keV to 800 keV, so the four bands below bracket the entire documented amateur record.

One line of arithmetic prices every experiment

Every rate estimate on this page is the same product:

R = (I/e) · n_t · (dσ/dΩ) · ΔΩ · ε

with I/e the particle rate — 1 nA of protons is 6.24 × 10⁹ per second — n_t the target areal density in atoms/cm², dσ/dΩ the differential cross-section, ΔΩ the detector solid angle, and ε the detection efficiency (near 1 for charged particles landing on silicon). The standing values used below, all from the professional low-energy ¹¹B(p,α) experiment the design guide harvests: a thin target is ~56 μg/cm² — 3.1 × 10^{18 11}B atoms/cm² (dg-542); a conservative spectroscopy geometry is 2.5 × 10⁻⁴ sr per detector at 16.5 cm (dg-541); a close-in geometry of the kind Oliphant used reaches ~0.7 sr (dg-509). A practical amateur middle ground — a 1 cm² silicon diode at 10 cm — is 0.01 sr.

Two housekeeping rules travel with the formula. The current in it must be honest: an unsuppressed target reads falsely high, so the collector carries a bias of about +9 V against secondary electrons (Houghton; dg-524), with technique per beam measurement. And yields are normalized by integrated charge, with detector solid angles calibrated against a known α-source and low-energy Rutherford scattering on gold — the professionals' own procedure (Spraker; dg-520).

The map: 5 keV to beyond 1 MeV

5–25 keV: surface physics — and Rutherford's experiment, for real

A proton in this band never reaches a nucleus' interior and barely enters the target: NIST PSTAR puts the projected range of a 25 keV proton in gold at about 0.1 μm (retrieved August 2026). Everything the beam does, it does at and just under a surface: elastic (screened Rutherford) scattering, secondary-electron ejection — copious enough to corrupt current readings, which is why the Faraday collector carries bias (dg-524) — sputtering, and light or film exposure. Nuclear reactions are absent in practice: even at fifty microamperes, Oliphant and Rutherford saw nothing from a thick lithium target below ~30 kV and nothing from boron below 60 kV (dg-493) — both onsets above this band's 25 keV ceiling. Sputtering, for its part, is real but cosmetic at these currents: even granting an implausibly generous one atom removed per proton, a 1 nA beam erodes roughly 20 μg of copper per year of continuous operation.

What survives is a genuinely good experiment. Rutherford scattering — the 1/sin⁴(θ/2) angular law on which nuclear physics was founded — is enormous at these energies, because the cross-section grows as 1/E². Worked at the top of the band: 25 keV protons on a thin gold foil (20 μg/cm², 6.1 × 10¹⁶ atoms/cm² — thin enough that PSTAR's stopping power says the foil costs only ~1.4 keV), scattered to 150°: the Rutherford formula gives 1.5 × 10⁴ barns/sr, so a 1 nA beam sends about 57,000 protons per second into a 0.01 sr detector. A caveat belongs next to that number: at 150° a 25 keV proton approaches the gold nucleus to within about 5 pm, not far inside the ~10 pm scale on which atomic electrons screen the nuclear charge, so the measured rate falls below the bare formula. The margin is three to four orders of magnitude; the experiment survives the correction. Verifying the angular law, or the 1/E² scaling by stepping the dee voltage, is a real measurement of the real thing — and calibrating detector geometry with Rutherford scattering on gold is exactly what professional groups do (dg-520).

Detection is the real difficulty, and it is instructive rather than fatal. A 25 keV proton deposits a pulse only a few noise-widths above a silicon detector's floor — reference PIPS certificates warrant 11 keV FWHM α resolution with electronic noise near 5.5 keV (dg-512) — so single-particle counting at the bottom of this band is marginal, and the rates are high enough that it is also unnecessary. The scattered flux can be treated as a current: 57,000 protons per second is 9 fA into 0.01 sr, and a close-in collector at Oliphant-style solid angle (~0.7 sr; dg-509) intercepts ~4 × 10⁶ protons per second — 0.6 pA, electrometer territory, with all the technique that implies (beam measurement). Film integrates the same signal with zero electronics, the trick Berkeley used for weak deflected beams (dg-690). By the top of the next band, silicon counts the scattered protons cleanly.

Safety in this band is machine safety, not beam-product safety. X-rays come from the RF structure, not the ion beam — amateur experience puts practical detectability near 18–20 kV of electrode voltage, but the conservative rule is the safety page's: assume X-rays whenever RF or high voltage is applied, and survey. Neutron production is energetically closed for the light targets above, and the thresholdless capture channels discussed under Beyond 1 MeV, while never strictly zero, are suppressed by the Coulomb barrier to yields with no measurable consequence at these energies — the survey practice still applies as a discipline, not because a hazard is expected here.

25–100 keV: transmutation arrives — barely countable, and only via lithium

This band contains one of the great facts of experimental physics: nuclear transmutation is observable below 100 keV. Oliphant and Rutherford, bombarding thick targets, saw the first lithium disintegration α-particles at about 30 kV and the first boron ones at 60–70 kV, with yields "increas[ing] more rapidly with increase of bombarding energy than with the lithium film" (Oliphant & Rutherford 1933, pp. 266–270; dg-493). The COLUMBUS (24–48 keV) and Niell II (~70 keV) machines run in or at the edge of this band. The temptation to conclude that a school cyclotron can repeat the Cavendish transmutation experiments is strong, and the arithmetic cuts it down to one narrow exception.

First, the scale of the 1933 result itself: Oliphant's geometry collected ~0.7 sr at ~50 μA (dg-493, dg-509). A nanoampere machine with a generous 0.1 sr detector runs at (1 nA / 50 μA) × (0.1 / 0.7) — a factor of 1/350,000. An event Oliphant logged once a minute arrives, in that amateur geometry, about once every eight months of continuous running. That is the current-versus-energy distinction in one number. The 1933 result was not delicate; it was rich in current, and current is the one thing a first cyclotron does not have. (The fusor community illustrates the same trade from the other side: amateur D–D fusion at 15–30 kV is detectable at home partly because a fusor passes milliamperes; dg-543, dg-503.)

Worked from cross-sections rather than history, though, lithium is not quite closed. The IAEA's evaluated-data handbook for light-nuclide reactions (INDC(CCP)-326, §4.17 — library entry lib-097, retrieved August 2026) gives a low-energy extrapolation for ⁷Li(p,α)⁴He that yields σ ≈ 68 μb at 100 keV and 15 μb at 70 keV (lab). Folding that curve through PSTAR's stopping power for a thick LiF target (retrieved August 2026): a 1 nA beam at 100 keV drives about 0.3 reactions per second — roughly 130 of the unmistakable ~8.7 MeV α-particles per hour into a close-in 0.7 sr detector, falling to ~20/hr at 70 keV (Niell II's energy) and ~2/hr at 48 keV (COLUMBUS's). The handbook itself documents a factor-of-two discrepancy between the underlying measurement groups (§4.17), so read these as order-of-magnitude figures. The verdict: transmutation at the top of this band is countable in principle — via lithium only, at hours-per-data-point patience, with a silicon detector inside the vacuum. Boron stays shut at nanoamperes: Oliphant's 60–70 kV boron observations took his 50 μA (dg-493).

What is actually performable here is what band I offered, with better margins: Rutherford scattering still yields thousands of counts per second into modest solid angles (the cross-section at 100 keV is a sixteenth of its 25 keV value — still ~10³ barns/sr at back angles), and scattered protons near 100 keV now deposit enough energy in silicon to count as clean pulses. The other real experiment in this band is the machine itself: mapping cyclotron resonance against field and frequency, species by species, is exactly the measurement Niell II published at ~70 keV — resonance peaks for H⁺ and He⁺ matching prediction within ~2% (census) — and it needs only the Faraday cup and electrometer chain covered in beam measurement. Safety: the lithium disintegrations this band permits are individually countable events with stable products (two α-particles) — no neutron channel is open for these targets, and X-ray discipline continues as in band I. The material-scoped activation caveats under Beyond 1 MeV apply in principle here too, at yields far below measurability.

100–500 keV: nuclear counting becomes routine

What band II permitted as hours-per-count marginalia becomes a practical counting experiment here. The exothermic reaction ¹¹B(p,α)⁸Be — Q = +8.590 MeV, no threshold (NNDC QCalc, retrieved August 2026) — turns countable at nanoamperes with a thin target and spectroscopy-grade geometry. The professional measurement is the proof: Spraker et al. took their sub-resonance data at 0.5 to 10 nA on target (dg-518) — amateur current, professional patience. Worked with this page's standing values, thin target and 2.5 × 10⁻⁴ sr:

• 150 keV, 0.91 mb/sr (dg-500): ~16 α counts/hour — a weekend integration per data point;
• 300 keV, 20.8 mb/sr: ~6 counts/minute;
• the same rates times 40 in the 0.01 sr close geometry — at 300 keV, ~4 per second.

Hours-long counts at single-digit rates are feasible only if background is crushed, and the reaction obliges: near 200 keV the two main α-particles emerge 150–180° apart, so two silicon detectors in back-to-back coincidence make a nearly background-free signature (dg-539). The spectra also carry a warning — elastically scattered protons pile up below ~1 MeV and dominate everything (dg-541) — so the α discrimination window sits above them. The detection stack this band demands is the full one: PIPS or surface-barrier silicon, charge-sensitive preamplifier, shaper, counter (library entries lib-022, lib-026, lib-065, lib-066), operated by the design guide's detector rules — vacuum better than 10 Pa at spectroscopy spacing (dg-439), bias only after pumpdown (dg-528), pulser benchmarks throughout (dg-512). The vacuum calculator covers the mean-free-path side.

At the very top of the band a second door opens: the ⁷Li(p,γ)⁸Be capture resonance. The capture reaction itself is exothermic and thresholdless; the resonance is where its yield concentrates. Its lab energy follows from two live-fetched numbers — the 17.640 ± 0.001 MeV level in ⁸Be (ENSDF, retrieved August 2026) minus the reaction Q-value of 17.254 MeV (NNDC QCalc), times the kinematic factor 8/7 — giving E_p = 441 keV, and the capture gammas carry ~17.6 MeV: energy enough to photodisintegrate beryllium (γ,n threshold 1.665 MeV) and deuterium (2.224 MeV; both NNDC QCalc, retrieved August 2026), so a machine tuned to this resonance should keep beryllium and deuterated materials out of the gamma field or add a neutron check to the survey. This is the physics behind the classic proof-of-acceleration trick in the Los Alamos small-cyclotron recommendations: fuse LiF onto the probe tip and look for prompt gammas (Wouters, p. 11; dg-517). It needs a scintillator rather than silicon, and a machine that actually reaches ~440 keV — which the census says is rare: of the measured amateur beams, only Rutgers (800 keV) clears it, with Houghton's 317 keV (capability thesis, lib-005) the closest miss.

One caution on transferring this band's arithmetic to real machines: it is priced at 1 nA on target. Houghton's documented internal beams are picoamperes (census) — divide every rate above by a thousand. The rates are honest; the current column of the ledger still has to be paid. Safety and legal: for the light targets on this page, the (p,n) neutron thresholds remain far away — ⁷Li(p,n)⁷Be at 1.881 MeV is the lowest among them (NNDC QCalc; consistent with the safety page's table) — but "no neutrons" is not "no radioactivity." A natural-boron target feeds ¹⁰B(p,α)⁷Be — Q = +1.146 MeV, no threshold — and ⁷Be is radioactive (electron capture, half-life 53.22 d; NNDC QCalc and ENSDF, retrieved August 2026). At nanoamperes the quantities are trace-level, and the professionals' isotopically pure ¹¹B target (dg-542) sidesteps most of it, but a bombarded target counts as potentially activated material: survey it before handling. A machine running (p,γ) on a lithium target is deliberately producing penetrating radiation — the amount scales with current, target, and tuning — and belongs under the safety page's zero-measurable-exposure survey discipline. State registration (legal survey) does not wait for nuclear physics; it typically attaches to the accelerator as such.

500–1000 keV: the 675 keV resonance is the destination

Everything in band III improves by a factor of ~240 at the top of this one. The ¹¹B(p,α) yield coefficient reaches 218 mb/sr at the 675 keV resonance (dg-500): at 1 nA on a thin target that is about one α per second even in the conservative 2.5 × 10⁻⁴ sr geometry, and ~40 per second at 0.01 sr — real statistics in minutes, an excitation curve in an afternoon. The resonance is nearly isotropic, so detector placement is uncritical (dg-540); counts convert to yields per the luminosity convention, since ~2.1 of the three αs land in the main peak (dg-519). This single experiment — a nuclear excitation function, on the reaction now studied as an aneutronic-fusion candidate — is the strongest scientific argument any sub-MeV machine can make for itself.

Two more channels reach useful yield — both exothermic and present at lower energies in principle, both practical here. ⁷Li(p,α)⁴He (Q = +17.346 MeV, NNDC QCalc), band II's hours-per-count experiment, now runs orders of magnitude faster and yields near-back-to-back α-particles sharing 17.3 MeV — about 8.7 MeV each, shifted by the beam momentum with emission angle. Almost nothing in nature sits that high: among natural α-emitters only ²¹²Po, at 8.785 MeV (ENSDF decay data, retrieved August 2026), reaches the neighborhood, and an in-vacuum silicon spectrum distinguishes a beam-correlated line from a thorium-chain contaminant by switching the beam off. ¹⁹F(p,αγ)¹⁶O (Q = +8.114 MeV) feeds excited ¹⁶O states whose de-excitation gammas — 6.13 MeV dominant, 6.92 and 7.12 MeV beside it (ENSDF level energies, retrieved August 2026) — penetrate the chamber wall to an external scintillator. Its yield is concentrated in narrow resonances from a few hundred keV upward (individual energies not quoted here), which is why the Wouters LiF-probe check (dg-517) works so well in this band: one salt on the probe tip, two elements, two unmistakable gamma signatures. Among census machines, only the Rutgers 12-inch — 800 keV, 200 nA by 2006, tens of μA after its 2012 ion-source upgrade — lives here, and at 10 μA the band's arithmetic scales by 10⁴: resonance αs at tens of thousands per second. That is why it functions as a working teaching laboratory (census).

What does not open, even here, is the vacuum window. PSTAR's ranges price extraction into air: a 500 keV proton stops inside 5.6 μm of aluminum, so any window it survives must be thinner still, and whatever emerges has 8.5 mm of range left in air; a 1 MeV proton crossing a 6 μm aluminum window pays at least ~280 keV for the passage and stops within ~2 cm of air beyond it (all values NIST PSTAR, retrieved August 2026). External work through ultrathin windows or differential pumping is specialist territory with millimetres of reach; in practice sub-MeV experiments happen inside the chamber, so plan the target and detector geometry there. Safety: still below the light-target (p,n) thresholds — no neutron production from the materials discussed here — while the thresholdless capture channels (band III's ⁷Be note, and the list under Beyond 1 MeV) leave trace radionuclides in bombarded targets, X-ray production from the RF structure scales with dee voltage, and gamma-producing targets carry the same survey obligations as band III (safety).

Beyond 1 MeV: where real nuclear physics opens

The character of the machine changes above 1 MeV, and the neutron thresholds for this page's target materials say where (all NNDC QCalc, retrieved August 2026, agreeing with the safety page): ⁷Li(p,n)⁷Be at 1.881 MeV, ⁹Be(p,n)⁹B at 2.057, ⁶⁵Cu(p,n)⁶⁵Zn at 2.167, ¹⁸O(p,n)¹⁸F — the workhorse route to PET's fluorine-18 — at 2.575, ¹¹B(p,n)¹¹C at 3.017, ¹³C(p,n)¹³N at 3.236, ⁶³Cu(p,n)⁶³Zn at 4.215, ⁵⁶Fe(p,n)⁵⁶Co at 5.445, and ²⁷Al(p,n)²⁷Si at 5.804 MeV. Two scope notes keep that list honest. The thresholds are properties of nuclides, not of matter in general: mid- and heavy-Z nuclei can sit far lower — ¹¹⁵In(p,n)¹¹⁵Sn is energetically open from 287 keV, ⁵⁵Mn(p,n)⁵⁵Fe from 1.032 MeV, ⁵¹V(p,n)⁵¹Cr from 1.565 MeV (NNDC QCalc, retrieved August 2026) — though the Coulomb barrier keeps yields near such thresholds vanishingly small; the real boundary is set by the materials actually in the beam path. And neutron thresholds are not the whole of activation: capture and charged-particle channels with no threshold at all make radionuclides at any beam energy — ¹⁰B(p,α)⁷Be (Q +1.146 MeV; ⁷Be, 53.22 d), ⁶Li(p,γ)⁷Be (Q +5.607 MeV), ¹²C(p,γ)¹³N (Q +1.944 MeV; ¹³N, 9.96 min), and ¹⁶O(p,γ)¹⁷F (Q +0.600 MeV; ¹⁷F, 64.5 s) among them (Q-values NNDC QCalc, half-lives ENSDF, all retrieved August 2026). Below 1 MeV these produce trace quantities in bombarded targets, which is why this page's per-band language is material-scoped and its survey advice unconditional. Crossing the first (p,n) threshold of a material you actually use turns the machine into a neutron and bulk-activation source, with the safety and regulatory weight that carries. Characteristic X-ray analysis also matures here: PIXE as practiced uses protons of 1–3 MeV (IAEA-TECDOC-1190, 2000) to excite lines like Cu Kα at 8.048 keV (NIST X-ray transition energies, retrieved August 2026); sub-MeV protons excite the same lines with cross-sections suppressed by the velocity mismatch — a proton matches a copper K-electron's speed only near 16 MeV.

Deuterons deserve a plain sentence: D–D fusion — ²H(d,n)³He, Q = +3.269 MeV; ²H(d,p)³H, Q = +4.033 MeV (NNDC QCalc) — has no threshold, so a deuteron cyclotron with a deuterated target is a potential neutron source at any energy, current permitting. Houghton has run deuterons (census). A builder considering that path inherits the fusor community's neutron-monitoring obligations (safety; dg-546).

The record matches the arithmetic. The first small cyclotron "capable of useful nuclear reactions" built by an undergraduate group ran at 1.5 MeV — Iowa State, first beam 1957 (library, lib-037). The recurring amateur design goal of 2 MeV — MIT's undergraduate project, the Cyclotron Kids (census) — sits exactly where the (p,n) doors begin to open; neither machine reached beam. Above 1 MeV the physics gets real, and so does everything else: activation surveys and the full weight of registration law. That trade is the subject of the page this one is an appendix to.

The scorecard

Read as a whole, the table is this site's answer to "what will it do?": one band of excellent classical physics, one band of machine physics with transmutation at its upper edge, one nuclear reaction that becomes several near a megavolt — every rate bought with energy, every energy bought with magnet and RF. The census machines cluster at 24–800 keV because that is what garage-scale iron and kilovolts buy; the experiments above are what those keV can honestly deliver. To turn a band into a machine — design point, closest documented builds, reading path — see Choosing Your Machine.

Sources

• NNDC (Brookhaven) Q-value calculator — all Q-values and (p,n) thresholds on this page, retrieved August 2026.
• NIST PSTAR — all proton stopping powers and ranges (gold, aluminum, air, copper, lithium fluoride), retrieved August 2026.
• ENSDF evaluated data (level energies ⁸Be 17.640 MeV and ¹⁶O 6.130, 6.917, 7.117 MeV; half-lives ⁷Be 53.22 d, ¹³N 9.96 min, ¹⁷F 64.5 s; ²¹²Po α 8.785 MeV), retrieved August 2026 — basis, with the QCalc Q-values, of the 441 keV resonance energy, the (p,αγ) gamma lines, and the activation notes above.
• INDC(CCP)-326, Nuclear Physics Constants for Thermonuclear Fusion: A Reference Handbook (IAEA, 1991), §4.17 — ⁷Li(p,α)⁴He low-energy cross-section extrapolation used for the band-II thick-target estimate (library entry lib-097), retrieved August 2026.
• M. L. E. Oliphant and Lord Rutherford, "Experiments on the Transmutation of Elements by Protons," Proc. R. Soc. A 141, 259 (1933) — observation energies, currents, and geometry, quoted via design-guide rules dg-493 and dg-509.
• M. C. Spraker et al., "The ¹¹B(p,α)⁸Be → α + α Reaction at Energies Below 5.4 MeV," J. Fusion Energy 31 (2012) — yield coefficients, beam currents, target and detector geometry, via dg-500, dg-518, dg-519, dg-520, dg-539, dg-540, dg-541, dg-542.
• L. F. Wouters, General Recommendations for the Design of Small Cyclotrons — LiF probe-tip gamma check, via dg-517.
• Canberra PIPS manual and detector certificates — resolution and operating discipline, via dg-439, dg-512, dg-528.
• IAEA-TECDOC-1190, Instrumentation for PIXE and RBS (IAEA, Vienna, 2000) — PIXE practice energies (1–3 MeV protons).
• NIST X-ray Transition Energies Database — Cu Kα₁ = 8047.82 eV, retrieved August 2026.
• This site's build census — machine energies and currents plotted in Figure 1 and cited throughout; Houghton capability figures also from the college's theses (library, lib-005).
• Amateur fusor sources (Hull; Kovalchick) — voltage-scaling and D–D detectability context, via dg-503, dg-504, dg-543.