[Enthalpy]

Figures about aluminium coils at 20K for MRI. 1.5T, the rest is arbitrary just to assess the feasibility.

A uniform induction needs more current density at the ends. I take two currents for simplicity. Solving for 1.5T at two points gives 1.81MA and 1.17MA, as illustrated.

Real cases refine that. Dividing more the current distributes the induction better. Choosing some reasonable axial induction distribution, outside the coils too, and deconvolving by the distribution of a current loop using a Kálmán filter would do it too. Surely dozens of methods exist.

Each of the four coils is L=1m R_i=0.87m R_o=1.13m big on the example. Aluminium fills only 70% because other coils need voids to vary the flux locally. So length=6.28m and S=0.182m² for 1 turn.
  nvlpubs.nist.gov (5MB) page 20 - lss.fnal.gov
the resistivity is 6.3pΩ×m at zero field, but magnetoresistivity adds 4.6× at 15kOe (1.5T in air). This extreme value happens at the inner face, so as a mean 2.3× for 21pΩ×m. 0.72nΩ/turn² dissipate then 2.4+1.0+1.0+2.4=6.7kW at 20K. Heat leaks can be much smaller. 30% of the 6.2% Carnot limit towards +50°C=323K let evacuate 360kW in the air. If the machine operates for 4h/day and 250 days/year, the electricity bill is 300k€ in 4 years. I suspect superconducting coils cost more, and they need cooling power too, at 4K.

0.23kg/s = 0.10m³/s gaseous helium from 19K to 21K shall remove the 2.4kW from the outer coils thanks to 10.5kJ/kg and 2.4kg/m³. An axial flow through 10% of the coil section or 0.16m² needs mean 0.6m/s helium speed, so easy.

That's a part of the picture. For instance the mechanical stability of the coils matters. Open MRI apparatus, more generally a vertical main field, saves power with shorter induction lines in air, while cold aluminium coils create a longer observation zone than permanent magnets do.

A thought for Sapo.
Marc Schaefer, aka Enthalpy

[Enthalpy]

Sources of γ rays sterilize objects, measure thicknesses without contact, display the contents of closed containers, etc. But radionuclides like ⁶⁰Co are also a danger. A switchable γ source could be useful. As usual, I didn't check what is already done or abandoned. Nor am I reliable on the topic.

Bombardment by protons or deuterons lets some targets emit γ.

• It often produces β⁺ emitters. When the 511keV photons from positron-electron annihilation are a drawback, shielding them away lets waste many 2MeV photons. Very few nuclides disintegrate by electron capture without β⁺.
• Some reactions create β^- emitters, typically by deuteron absorption and proton emission. The β braked by surrounding matter (Bremsstrahlung) creates photons rarely useful that can be minimized. The disintegration often emits γ rays of single or several energies by internal transition. If the β^- emitter is short-lived, the γ emission stops some time after the protons or deuterons beam. But seconds would be better than minutes.
• A beam could first produce neutrons whose absorption creates β^- emitters. But neutrons take decimetres to brake, and a wide source makes fuzzy γ images. Neutrons tend to activate all materials, possibly for a long time. And they need strong protons but the double conversion is inefficient. I didn't insist.
• A few reactions just absorb a weak proton or deuteron and emit a strong photon, immediately at human scale. Most promising.

========== Example

Among these fourth-type reactions, in this message I consider
  ¹⁹F(p,γ)²⁰Ne. For chemists: ¹⁹F + proton ²⁰Ne + γ
Data is gratefully taken from the Janis books
  protons (8MB) page 62 - deuterons (3MB)

Natural fluorine is pure ¹⁹F. ²⁰Ne is stable. I understand the γ carries 12844keV reaction energy plus 19/20 of the incident proton contribution: an energy not available from radioactivity, twice as penetrating as 1.33MeV from ⁶⁰Co, for instance to measure thicker red-hot steel plates when rolled.

The first competing reaction is on p63 in the Janis book, with similar sections over the energy range
  ¹⁹F(p,α)¹⁶O
The product is stable and the α stops within the source. If emitting no additional γ, this reaction looks harmless and acceptable.

The next competing reaction is on p59
  ¹⁹F(p,n)¹⁹Ne
which is 4MeV endothermal, so a smaller proton energy prevents it and the others.

The useful reaction has a measured section around 40mb from 1.4 to 0.3MeV. Target fluorine can be AlF₃ since the Janis book lists no reaction of Al active at these energies. NaF and MgF₂ seem clean too. ¹³N from (CF₂)_n would emit 1/1000 γ at 511keV with 10min half-life. Very pure ¹¹B is less convenient. Li and Be emit other radiations.

32%wt Al and 68%wt F brake 1.4MeV protons over 44g/m², that is 30g/m² F, as deduced from
  physics.nist.gov
and then 40mb section let 3.8ppm of all protons create a γ, so a 1.4MeV 1.6mA 2.2kW proton beam emits 1Ci. Maybe protons up to 4MeV improve the yield but the Janis book lacks experimental data, and 1.4MeV accepts a small cyclotron, or maybe a linear accelerator with radiofrequency quadrupoles (RFQ).

Aluminium or other, possibly cooled by water, can support AlF₃ of which 10-20µm stop the protons. The two transmutations deplete F in few years of continuous 1Ci operation, faster degradation processes are plausible.

Marc Schaefer, aka Enthalpy

[Enthalpy]

Some more proton-to-γ reactions. Reactions not indicated in the Janis book may ruin the attempts.

Target  Nature  Product  Janis    Q   Avoid  Section  From   To   Yield
           %              page   MeV   MeV      mb     MeV   MeV  Ci/mA
=========================================================================
  ⁷Li    92.50     ⁸Be    007  +17.3   1.64    7?      1.6   0.5  0.5?
  ⁹Be   100       ¹⁰B      13   +6.6   1.85    0.5?    1.8   1.0  0.02
 ⁴⁴Ca     2.09    ⁴⁵Sc    103   +6.9   4.43    1.8     4.0   2.0  0.2
 ⁵¹V     99.75    ⁵²Cr    147  +10.5   1.53    0.2     1.5   1.2  0.02
 ⁵²Cr    83.79    ⁵³Mn    162   +6.6   5.49    0.6     4.0   2.0  0.06
 ⁵⁸Fe     0.28    ⁵⁹Co    194   +7.4   3.09    1.4     3.0   2.0  0.03
 ⁵⁹Co   100       ⁶⁰Ni    201   +9.5   1.86    0.6     1.8   1.6  0.002
 ⁶²Ni     3.63    ⁶³Cu    230   +6.1   4.73    2.4     4.3   3.0  0.06
 ⁶⁴Ni     0.93    ⁶⁵Cu    235   +7.5   2.45    1.1     2.4   1.2  0.02
 ⁶³Cu    69.17    ⁶⁴Zn    243   +7.7   4.15    1.8     4.1   2.8  0.1
 ⁶⁵Cu    30.83    ⁶⁶Zn    250   +8.9   2.13    0.6     2.1   1.8  0.004
 ⁶⁸Zn    18.80    ⁶⁹Ga    274   +6.6   3.70    7       3.6   2.8  0.1
 >Zn                     Lacks γ data
=========================================================================

  ⁷Li: No data about ⁶Li. Does the data distinguish ⁸Be from α+α? How much energy in the photon vs α+α?
⁴⁴Ca: Remove well ⁴²Ca ⁴⁸Ca.
⁵¹V : Leave ⁵⁰V.
⁵²Cr: Remove well ⁵³Cr ⁵⁴Cr. Produced ⁵³Mn has 3.7My half-life.
⁵⁸Fe: Remove well ⁵⁴Fe ⁵⁶Fe ⁵⁷Fe.
⁵⁹Co: No separation.
⁶²Ni: Remove well ⁵⁸Ni ⁶⁰Ni ⁶¹Ni ⁶⁴Ni. Can start below 4.3MeV.
⁶⁴Ni: Remove ⁵⁸Ni ⁶⁰Ni ⁶²Ni. Remove well ⁶¹Ni.
⁶³Cu: Remove ⁶⁵Cu and start below 2.13MeV, or remove well ⁶⁵Cu.
⁶⁵Cu: Remove ⁶³Cu.
⁶⁸Zn: Remove well ⁶⁴Zn ⁶⁶Zn. Remove ⁶⁷Zn and start below 2.41MeV, or remove well ⁶⁷Zn.

⁹Be, ⁵¹V, ⁵⁹Co need no isotopic separation. ⁶³Cu, ⁶⁵Cu accept limited isotopic purity and emit lower energy γ. Insufficient data to tell about ⁷Li. The others need strong isotopic purity for cleanliness.

Energetic protons make more gammas but need a bigger cyclotron. All nuclides accept weaker protons.

Marc Schaefer, aka Enthalpy

[Enthalpy]

A 1.4MeV protons cyclotron for a switchable gamma source is D=0.8m H=0.6m small and can be added to a production facility or an imaging apparatus. Illustration appended. Please remember I didn't consider beam focussing.

0.89T give protons 13.56MHz cyclotron frequency, 6 sectors receive power at the 40.68MHz ISM frequency. If the energy must vary, move the extraction electrode radially?

The DC magnet coils provide 2×14.2kA×turn. 90% fill factor
  chemicalforums
gives them 2.4µΩ/turn² so they consume 2×0.5kW.

1.36T and 2×20.34MHz would shrink the cyclotron to D=0.7m but need 2×1kW in the coils.

6 sectors all fed with slightly over 5kVpk provide 60keV/turn. Each sector has stray 50pF to the iron compensated by a 0.3µH coil each. These coils can consist approximately of 2.5 turns of 6mm gold-plated wire on D=37mm air. They dissipate roughly 6×65W, so oil and even blown air can isolate and cool them.

A linear accelerator for 1.4MeV using radiofrequency quadrupoles (RFQ) is shorter than 2m and sleek. Its beam can be bent, even by permanent magnets. Serious competitor.

Marc Schaefer, aka Enthalpy

[Enthalpy]

More reactions for a switchable γ source, as alternatives to proton capture.

Data is from the Janis book.

========== β⁺ and ε

Proton impact makes naturally proton-rich nuclides that decay by β⁺ emission or by electron capture ε. The impact often emits neutrons too, but surrounding boron can absorb them with little added radioactivity.

The β⁺ decay can emit a useful γ, but the e⁺e^- annihilation emits also two 511keV γ. If the 511keV are usable, many radionuclides and reactions are possible and known. Removing undesired 511keV is difficult and wastes the other γ even at 2MeV. I didn't insist.

A few proton-rich nuclides decay by ε without β⁺, hence without 511keV γ. The emitted neutrino takes some energy, and at least a part of the γ spectrum is continuous, less valued for imagery. Maybe some nuclide exists with useful γ energies and credible production reactions, but I didn't find any in a limited time.

========== β^-

No 511keV γ here, and some nuclides have a clean γ spectrum. Neutron-rich nuclides are harder to produce by impact. The reactions I found get a deuteron and emit a proton. I checked for unwanted reactions only in the Janis book, they exclude for instance ¹⁸O to ¹⁶N and ¹⁸O to ¹⁹O.

Target  Nature  Product  Janis  Avoid  Barn  From   To   t1/2  Gamma   %  Yield
           %              page   MeV          MeV   MeV         MeV       Ci/mA
================================================================================
  ¹⁵N      0.37    ¹⁶N      33     ?     0.7   2.6+  1.2  7.1s  6.13   67    19
  ²⁷Al   100       ²⁸Al     46    4.1    0.2   4.0   2.4  2.2mn 1.78  100    10
================================================================================

The β^- emits few photons by braking, even in Al. At 1.5MeV it's around 2%.

The choice is again narrow and implies diverse drawbacks:

• Isotopic separation, uneasy compounds for ¹⁵N. Yield for pure N!
• Minutes half-life is good for transport and disposal, less for development and use.
• Deuterons need a cyclotron as big as protons of double energy. Linear accelerator!

Maybe some applications prefer these drawbacks to a permanent radioactive source.

Marc Schaefer, aka Enthalpy

[Enthalpy]

I didn't find how to contain gaseous ¹⁵N as a target for 2.6MeV deuterons already stopped by 10µm nickel, so this gamma source needs a nonvolatile nitrogen compound.

The deuterons react with pretty much all light elements like B, C, O, Al, Si... to produce radionuclides that would still radiate after the useful ¹⁶N has decayed. My answer is to combine N with yttrium or heavier, as the 2.6MeV deuterons attain the nucleus rarely, a million times less often than nitrogen. Unless someone wants to shoot at HN₃ or synthesize N₈, of course.

The compounds I found are nitrides: YN, ZrN, NbN, Ta_xN_y, WN, while HfN and others are less known, possibly nonexistant.

Among these, Ta₃N₅ provides the best nitrogen content and reaction probability for ¹⁵N: only /2.7 as compared with pure N. The yield drops from 19 to still excellent 7Ci/mA. It offers decent resistance to heat and water too. Some Nb in Ta is acceptable. The heavy Ta reacts extremely little with 2.6MeV deuterons.

The other compounds reduce the reaction probability by /4 roughly, so they can be considered if more stable for instance.

Marc Schaefer, aka Enthalpy

[Enthalpy]

How big is a cyclotron for 4MeV deuterons used in the reaction with aluminium I proposed here on 14 Aug 2022?

1.78T lets deuterons orbit at 13.56MHz, so the acceleration power can use the 27.12MHz or 40.68MHz ISM frequencies. At 4.0MeV, the path radius is 0.23m with a slightly reduced gap there. 2×28.3kA achieve 1.78T in 40mm gap.

The flux returns between R=0.41m and R=0.51m, where 1/3 of the area is open. The cyclotron takes D=1.02m H=0.7m as sketched and weighs 4.5t.

Smaller is possible. Less γ activity enables less energetic deuterons. Cold coils, optionally superconducting, can produce a stronger induction, but then iron doesn't help.

Lukewarm copper coils can fill 90% of R=0.26m to R=0.40m and H=0.18m each. 430 turns of 0.3mm foil plus 0.025mm spacing resist 165mΩ, so 10.9V and 66A dissipate 715W per coil. Two gamer PC power supplies suffice.

Or use D=1mm aluminium wire at 20K. I imagine yarn wound loosely on the wire so 0.04mm spacing leaves helium through, with a polymerizing resin that impregnates the yarn. At 0.25T=2.5kOe, the magnetoresistance adds 0.7× the zero-field resistivity, but this induction drops linearly with the height and drops also where the return path is removed, so ×0.23 as a mean. 170×131 turns resist 0.87Ω so 1.27A and 1.10V dissipate 1.40W at 20K. 50+ plies of multilayer insulation wrapped around a coil leak 0.3W. The atmosphere presses on the separators, which may need more plies. If 30% as efficient as Carnot's limit, the singe cooler consumes 180W. Square aluminium wire would save 30W.

========== 2.6MeV deuterons

For the reaction with nitrogen, the cyclotron can be 0.81³× as big, or D=0.82m H=0.56m. Consider a linear accelerator with RFQ.

========== Switch on faster

When the γ activity needs less beam current, starting with the full current achieves the sought activity faster.

Marc Schaefer, aka Enthalpy

[Enthalpy]

[MRI apparatus use to be much shorter than depicted here on 31 July 2022. The conclusion remains.]

==========

Open Magnetic Resonance Imaging began with permanent magnets and 1/4T that blurrs the image. One present commercial success is Philips' Panorama High-Field Open that offers 1.6m opening between two pillars and 1.0T. The pillars seem too narrow for an iron return path, so the supposedly superconducting coils must create the induction unhelped. They look like horizontal Helmholtz coils
  wikipedia
Smaller currents added nearer to the axis must even out the induction. More coils create induction gradients, produce and pick the RF fields.

20K aluminium coils seem globally cheaper than superconductors in that use. Cold sensor coils and preamplifiers are good for small signals too.

I compute with plain Helmholtz coils as the corrections need little current and power. R=1m and 1m spacing take 2×569kA×turn. Diagram appended. Data source as previously.

Aluminium wire is square 2mm×2mm here, finer can help varying fields. Band would limit the supply voltage. 15 900 turns fit in D=0.3m. A straight cable would create 0.76T at its surface, but I take 1.0T=10kOe, so the magnetoresistance would add 1.75× the 24pΩ×m zero-field resistivity with RRR=2100, and averaged over the radius it's 1.17×, leading to 52pΩ×m. Each 1.3Ω coil uses 35.8A and safe 46.5V, together 3.32kW at 20K.

Narrow (adhesive) tape or (resin impregnated) yarn wound around the wire insulates and defines the 100µm cooling channels. Interleaved 50µm windings guarantee the channel thickness. 0.13m³ helium at 1atm from 19K to 20K remove the heat. The mean speed is 1.5m/s in the channels, so 1.4mm²/s and 100µm define Re=150 < 2000 and the flow is laminar. Speed curvature 1.74Gm/s/m² in the channels lets drop only 6kPa over 0.3m.

The heat insulation needs vacuum and reflective surfaces, multilayer insulation is optional. Polymer straps can hold the coils, but I didn't put figures. Forces are like 2MN while deformations are undesired, interesting.

The cooler consumes and dumps 180kW if 30% as efficient as Carnot's limit. This is much, but:

• Operation 3h/day and 250d/yr at 0.2€/kWh costs 0.11M€ electricity over 4yr, 0.27M€ over 10yr. That looks cheaper than superconductors.
• 15m³/s air absorb the heat and exit 10K warmer. At 3m/s it takes a D=2.5m fan blowing a h=0.63m exchanger on the roof.
• The big cooler can improve.

Marc Schaefer, aka Enthalpy

[Enthalpy]

Updates to my last message.

The magnetic forces are nearer to 1/2 MN rather.

0.13m³/s helium. The helium pressure drop over straight 0.3m would rather be 2kPa but the path zigzags. 4kPa wouldn't be pleasant, adding 520W heat to ohmic 1600W. A quadrupolar flow improves this. Or increase the 100µm spacing, it works cubed.

Helium data is from BNL, many thanks
  bnl.gov

[Enthalpy]

New update of the protons-to-neutrons conversion, starting from 40MeV.

⁹⁸Mo is needed at the neutron target anyway. Y outperforms Ga and Ta. ¹⁰⁰Mo(p,n+p)⁹⁹Mo uses no intermediate neutrons. kg/m² use the natural isotopic composition. d(d,n+p)d at 20MeV may be more efficient and convenient than p(d,n+p)p at 40MeV but I lack data.

Beam power is for 1.4×10¹⁴n/s as previously.

Nuclide  Page   kg/m²     Barn     Conv    mA      kW
======================================================
  ²H      004     6.7      0.13    5.2%    0.43    17  <<< Cleaner!
 ⁸⁹Y    411-417  21      3×0.30... 2.7%    0.83    33
²³⁸U      784    26    2.3×1.4     2.1%    1.1     43  Fission
 ⁹⁸Mo     473    13      3×0.60... 1.5%    1.5     54
======================================================
¹⁰⁰Mo     482    20        0.18    0.23%  10      400  ⁹⁹Mo directly
======================================================

²H needs a non-volatile compound or a container that waste few protons.

Marc Schaefer, aka Enthalpy

[Enthalpy]

If a switchable source provides intense γ, it needs an strong beam, and the design must cool the target. Examples and illustrations for the reactions of 14 Aug 2022, 14 Aug 2022, 06 Aug 2022, 07 Aug 2022 here.

========== Al target, 4MeV×3mA in 1cm², 30Ci

Deuterons stop in 0.18mm Al, so boiling water cools the target easily. I've already cooled a thin copper tub from an oxygen-acetylene burner that provided >10kW over 1cm², as measured from the water evaporation speed. Rocket engines jackets accept also 5kW/cm² just with "kerosene".

Al creeps and corrodes at heat, so 50µm Ni is deposited on the inner face. 10bar water pressure in D=20mm create 200MPa in warm Ni.

12kW/cm² drop about 200K in Ni and 200K in 0.3mm pure Al, bringing the surface around +500°C. The Al vapour pressure is negligible.

12kW evaporate 5g/s water. Or liquid water from 40 to 70°C would take 100g/s, D3mm suffice at +1bar, I didn't check the transfer.

AlMg5 (alloy AA5083) as the inner face seems less easy than Ni. The uncommon Mo would outperform Ni. Cermets are interesting. Rotating or liquid targets accept more power density, as for X-rays. At melting temperature, the vapour of Al has just 10^-8torr equilibrium pressure, and it condenses locally.

========== ¹⁵N target, 2.6MeV×3mA in 40mm², 21Ci

Supposing here that Ta₃N₅ can be processed, is stable at heat, reasonably resistant to occasional air, contact and moisture. It's a bit new and less known than TaN, which YN, ZrN, NbN may outperform. Heat conductivity matters.

Deuterons stop in <0.1mm Ta₃N₅. 0.3mm CuCr1Zr (>200MPa after 10000h at 350°C and 360W/m/K, wow) supports the layer and conducts the heat to the water with 160K drop. It can be hard-drawn for easier machining.

========== F target, 1.4MeV×3mA in 22mm², 1.8Ci

20µm AlF₃ seems perfect, but possible reactions of Al with 1.4MeV protons would let prefer heavier nuclei. CrF₃ is one candidate, compounds with more F seem more volatile.

CuCr1Zr remains a good substrate. AlMg5 might help processing.

========== ⁷Li target, 1.6MeV×3mA, 1.5Ci

Oxygen seems unreactive to 1.6MeV protons, P Cl Br maybe too, but compounds like Li₃PO₄ waste many protons. More promising is metallic Li, cleaned and covered by semiconductor processes with thin Al, Si, Mo, Ta, W, Rh, Ir, Pt, Au... Something unreactive to 1.6MeV protons and impervious to warm Li. 0.5µm Ir wastes 0.05MeV above 1.6MeV.

Lithium can also be molten once in vacuum. At melting temperature, its vapour pressure is small.

========== Downsize

Not every use needs 30Ci, then the impact spot can shrink. A tilted target too accepts a finer beam, and the γ source seems smaller from some directions. A finer beam allows a bigger curvature for a sturdier or thinner target substrate.

Marc Schaefer, aka Enthalpy

[Enthalpy]

I expect very quick proton-to-γ reactions, so the proton beam can modulate the gamma emission.

The quickly modulated γ must have more uses than I imagine:

• Measure fluorescence times
• Measure detector reaction times
• Make 3D images by fluorescence or diffusion, the propagation delay tells the distance
• Measure the tiny refraction index for gammas?
• Deduce the composition of objects from the refraction index?

Modulated γ should inherit the uses of modulated X-rays, just at thicker items, including the soil.

Cavities modulate naturally the beam at a linear accelerator or a cyclotron or other. If cheap magnetrons for microwave ovens feed them at 2.45GHz, then 10mm depth make lambda/2pi phase shift on the direct-and-return path. Cyclotron sectors fed at 100MHz give rather 1/4m depth resolution, in that first estimate.

Known means can suppress chosen proton bunches, especially before the acceleration. This lets modulate the γ emission more slowly than the acceleration does, to avoid the uncertainty over an integer number of periods. Some pulse sequences with good autocorrelation combine range and resolution, for instance PN sequences. Ask your local signal processing specialist.

Marc Schaefer, aka Enthalpy

[Enthalpy]

Here's a concept of a target for a 250kW beam of 75MeV protons. Meant for the ¹⁰⁰Mo(p,n+p)⁹⁹Mo reaction of 10 Jul 2022 here, it can inspire less difficult designs, like neutron production from a Ta target.

All protons stop in the Mo target to confine the radioactivity, none in the coolant. 0.6mm Mo and tg=1/16 suppose parallel proton paths, keep an eye.

Do lighter Mo isotopes produce undesired nuclides? I didn't check enough. That would demand purer ¹⁰⁰Mo than for yield. After irradiation, chemical separation keeps essentially Mo, of which 100 is the substrate, 99 the product, 98-94 and 92 are stable and natural, so a bit of 98-95 isotopes in the target seem acceptable.

This design targets only 31g Mo of which for instance 1/5000 becomes ⁹⁹Mo. Not as concentrated as in uranium fission products, but better than by neutron irradiation. I hope the moly cow, which separates the desired ^99mTc from Mo, accepts such a concentration.

The imagery offices send back the used but precious ¹⁰⁰Mo to make a new target, for instance by electroforming at the cyclotron site. Some hot compaction may be needed. A D=30mm t=0.6mm cylinder would buckle under 30bar, a flanged cone improves. Conduction through 0.6mm/3 of compact Mo drops 70K.

250kW evaporate 0.11kg/s water. Squirting at the cone may be safer. 50ms without water pierce the target, so something stops the beam if a point of the target overheats. Before it gets too hot, Mo emits light, but so do the p reactions. Distinguish?

A secondary cooling circuit can confine potential radioactivity better. I'd have 1bar in the primary, possibly less in the secondary. Heat pipes?

The sought reaction emits one neutron per ⁹⁹Mo, others contribute. After thermalization, neutrons can vanish in boron. Or they can produce useful radioisotopes, by ⁹⁸Mo(n,g)⁹⁹Mo or other. This would favour heavy water.

Marc Schaefer, aka Enthalpy

[Enthalpy]

Here's a concept for protons to neutrons using deuterium that seemed efficient and rather clean here on 28 Aug 2022.

The deuterium compounds I checked waste many protons and add much radioactivity, so the target is gas at +60 to +100°C and 300bar to stop the 40MeV protons in 0.37m=7kg/m².

I found only beryllium for the proton window.

• 10mm hemispherical radius, 1mm thickness induce 150MPa.
• 1mm uses 2.3MeV more p energy, or 1kW. Spread over 1mm and D=10mm, they drop 42K thanks to 100W/m/K.
• Activation emits "only" 478keV gammas, not worse than the sought neutrons.
• The window emits some neutrons too.

But what reactions are missing in the Janis book, what do the neutrons produce? Oxygen, carbon and more impurities in beryllium get activated too.

Page  Product   Decay     Emits keV
====================================
 11      ⁹B    p2α 1as      Clean
12&16    ⁷Be   ec 53d       γ 478
 13     ¹⁰B    Stable
 14      ⁸Be   2α 0.07fs    Clean
 17      ⁶Li   Stable
====================================

18m/s axial wind evacuates the 18kW to the finned walls and to the water. Transverse 0.4m/s would cool the deuterium but not the window. At the sketched position, the blower is less exposed to the neutrons supposedly emitted forward. The pressure vessel's material should minimize the activation by neutrons and their absorption.

Best Cyclotron Systems, Inc. have a gaseous target to produce neutrons, no idea how similar it is. They boil with heavy water too, after all.

Maybe some part, possibly the vessel, should multiply the neutrons. I suppose the neutron source and the Mo, U or other are all in a big pool.

This converter uses 1.2GeV per neutron if the RF source is 70% efficient, and without neutron multiplication. I didn't compare with lead spallation. If D-T +Li fusion provides 25MeV converted to 10MeV electricity, then such a source can complement only 0.8% to the regeneration of tritium, that is, uninteresting.

Marc Schaefer, aka Enthalpy

[Enthalpy]

I seek 1.4×10¹⁴ neutrons/s for medical ⁹⁹Mo, but cyclotrons can produce 1mol/year neutrons for instance, or 100× more. The one at PSI supposedly exceeds that.

If the incident particle has more energy, the target's electrons squander less of it, while fission and spallation are more frequent and produce more neutrons. I suppose spallation produces more than Janis' maximum 7 neutrons, and less stable elements like Ta might outperform Pb, but I have no data, so I consider the fission of ²³⁸U by 200MeV protons.

Protons brake from 200MeV to 30MeV in 510kg/m² uranium (27mm metal, oxide a bit more)
  Nist
The fission section is 1.6b so 1 proton provokes 0.20 fission, each emitting mean 9 neutrons
  osti.gov p28/69 and p30/69
so 1mol/year neutrons needs 1.7mA beam current, 0.34MW beam power, 0.48MW electricity.

Just 2.5kW beam power would provide 1.4×10¹⁴ neutrons/s, while 40MeV protons hitting deuterium need 17kW
  chemicalforums
but saving 20kW over 10 years, or 0.35M€, intuitively doesn't buy the bigger machine.

A 500MeV cyclotron would triple the neutron flux using 2.5× more electricity, not quite original neither. Increasing the current is doubtful.

Marc Schaefer, aka Enthalpy