Yinsen Fusion Reactor Concept

• Yinsen is a fusion reactor concept employing a low-power-density HTS tokamak design that prioritizes the longevity of primary in-vessel structures.
• It deliberately limits fusion power density to around 0.7 MW/m² to reduce neutron wall loading and material degradation over a 20-year plant lifetime.
• The design targets niche off-grid applications such as maritime propulsion and remote industrial energy, emphasizing autonomy and simplified maintenance.

Searching arXiv for the cited Yinsen paper and closely related tokamak/HTS reactor context. Yinsen is a first-of-a-kind fusion reactor concept centered on a low-power-density high-temperature-superconducting tokamak architecture for off-grid applications, including maritime propulsion, remote power, industrial energy, and off-grid electricity for data centers and critical infrastructure. Its defining design choice is to trade blanket-area-normalized power density for survivability of primary in-vessel solid structures over the full plant lifetime, rather than pursuing grid-scale power density. In the published concept, the design is anchored to a materials-limited fusion power density of $P_f/S_b = 0.7~\mathrm{MW/m^2}$, derived from a 35 DPA structural limit, a 20-year plant lifetime, 40% utilization, and a geometric damage-peaking correction. The resulting baseline is a shaped, high-field HTS tokamak with $B_0 = 9.29~\mathrm{T}$, $I_p = 9.67~\mathrm{MA}$, a minimum useful fusion power of about $130~\mathrm{MW}$, and net electric output above $25~\mathrm{MWe}$ (Cohen et al., 5 May 2026).

1. Concept definition and application domain

Yinsen is explicitly framed as a low-power-density HTS tokamak for mission classes in which autonomy, dispatchability, and avoidance of fuel logistics have disproportionate value. The cited applications are maritime propulsion and ship-to-shore power, remote industrial energy and high-grade heat, and off-grid electricity for data centers and critical infrastructure. In these markets, performance and autonomy are treated as more important than levelized cost of electricity, and the concept emphasizes avoidance of oil-price volatility, weather-dependent generation, and large thermal cycling of prime movers (Cohen et al., 5 May 2026).

The concept is also defined by what it does not optimize. Rather than maximizing $P_f/S_b$, Yinsen sizes the plant around a materials-limited power density intended to preserve the vacuum vessel and primary blanket structures for the full plant life. The stated purpose is to avoid routine open-heart “vessel swap” maintenance and thereby simplify outage planning, remote handling, cryogenic-magnet disassembly, tritium and activated-waste logistics, and compact packaging. The same design choice is presented as reducing three major engineering burdens of compact high-field tokamaks: divertor power handling, extreme neutron wall loading, and rapid structural-irradiation burn-down (Cohen et al., 5 May 2026).

A plausible implication is that Yinsen is not proposed as a general template for grid-scale fusion economics, but as a first deployment path for a narrower application space in which tens of megawatts of low-carbon, dispatchable power are already valuable. The paper itself characterizes this as a route for relevant FOAK reactors before full grid-scale economics are addressed (Cohen et al., 5 May 2026).

2. Materials-limited power density and lifetime logic

The central quantitative constraint is written as

$$0.8\cdot(P_f/S_b)\cdot F\cdot f_{pk}\cdot U\cdot T_{lifetime} \le L_{dpa},$$

so that

$$P_f/S_b \le \frac{L_{dpa}}{0.8\cdot F\cdot f_{pk}\cdot U\cdot T_{lifetime}}.$$

Here, the factor $0.8$ accounts for the D–T neutron energy fraction of fusion power; $F$ is the energy-fluence-to-damage conversion for the limiting structural material; $B_0 = 9.29~\mathrm{T}$0 is a peak-to-average damage factor; $B_0 = 9.29~\mathrm{T}$1 is lifetime-average utilization; $B_0 = 9.29~\mathrm{T}$2 is plant lifetime; and $B_0 = 9.29~\mathrm{T}$3 is the allowable cumulative structural-damage limit (Cohen et al., 5 May 2026).

The geometric peaking factor is taken from the $B_0 = 9.29~\mathrm{T}$4 variation of toroidal area at fixed poloidal length:

$B_0 = 9.29~\mathrm{T}$5

with $B_0 = 9.29~\mathrm{T}$6. For the adopted aspect ratio $B_0 = 9.29~\mathrm{T}$7, the paper gives $B_0 = 9.29~\mathrm{T}$8 (Cohen et al., 5 May 2026).

Yinsen adopts V–4Cr–4Ti as the vacuum-vessel structural material, with $B_0 = 9.29~\mathrm{T}$9, an allowable life limit $I_p = 9.67~\mathrm{MA}$0, utilization $I_p = 9.67~\mathrm{MA}$1, plant lifetime $I_p = 9.67~\mathrm{MA}$2, and $I_p = 9.67~\mathrm{MA}$3. Substituting these values gives

$I_p = 9.67~\mathrm{MA}$4

which is rounded to a conservative FOAK ceiling of $I_p = 9.67~\mathrm{MA}$5 (Cohen et al., 5 May 2026).

At this limit, the corresponding neutron wall loading is $I_p = 9.67~\mathrm{MA}$6 in the baseline Case A. Detailed OpenMC transport maps this to an inner-vacuum-vessel damage rate of about $I_p = 9.67~\mathrm{MA}$7 at the $I_p = 9.67~\mathrm{MA}$8 baseline and 40% utilization, reaching 35 DPA in about 20 calendar years. The equivalent lifetime-integrated fusion-energy throughput is $I_p = 9.67~\mathrm{MA}$9, equal to $130~\mathrm{MW}$0 (Cohen et al., 5 May 2026).

This lifetime logic is the main differentiator of the concept. High-field compact tokamaks ordinarily benefit from strong confinement scaling, but the Yinsen design argues that very high $130~\mathrm{MW}$1 forces simultaneous resolution of divertor loads, shield thickness, downstream nuclear heating, irradiation damage, and frequent major replacement campaigns. Yinsen’s response is to lower $130~\mathrm{MW}$2 to a level the authors regard as FOAK-credible with presently known materials (Cohen et al., 5 May 2026).

3. Baseline configuration, plasma regime, and magnet system

The minimum useful fusion power class is stated as approximately $130~\mathrm{MW}$3, the smallest scale judged able to support meaningful net electric export after realistic recirculating loads. In Case A at $130~\mathrm{MW}$4, the plant has thermal power $130~\mathrm{MW}$5, gross electric power $130~\mathrm{MW}$6, and net electric power $130~\mathrm{MW}$7, exceeding the stated $130~\mathrm{MW}$8 target (Cohen et al., 5 May 2026).

The baseline geometry at $130~\mathrm{MW}$9 is summarized below.

Parameter	Value
Major radius $25~\mathrm{MWe}$0	$25~\mathrm{MWe}$1
Minor radius $25~\mathrm{MWe}$2	$25~\mathrm{MWe}$3
Inverse aspect ratio $25~\mathrm{MWe}$4	$25~\mathrm{MWe}$5
Elongation $25~\mathrm{MWe}$6	$25~\mathrm{MWe}$7
Triangularity $25~\mathrm{MWe}$8	$25~\mathrm{MWe}$9
Plasma volume $P_f/S_b$0	$P_f/S_b$1
Blanket-facing area $P_f/S_b$2	$P_f/S_b$3

The plasma is a high-field, strongly shaped H-mode configuration that favors high bootstrap fraction without requiring steady-state noninductive operation. The baseline on-axis toroidal field is $P_f/S_b$4, the peak field on TF conductor is about $P_f/S_b$5, and the plasma current is $P_f/S_b$6. The flattop duration is about $P_f/S_b$7 in inductive pulses (Cohen et al., 5 May 2026).

FUSE systems modeling defines the baseline operating point, while ASTRA+TGLF transport runs are used as corroboration. The regime parameters include $P_f/S_b$8 in Case A and $P_f/S_b$9 in Case B, with $$0.8\cdot(P_f/S_b)\cdot F\cdot f_{pk}\cdot U\cdot T_{lifetime} \le L_{dpa},$$0–$$0.8\cdot(P_f/S_b)\cdot F\cdot f_{pk}\cdot U\cdot T_{lifetime} \le L_{dpa},$$1 and a monotonic $$0.8\cdot(P_f/S_b)\cdot F\cdot f_{pk}\cdot U\cdot T_{lifetime} \le L_{dpa},$$2-profile. FUSE gives confinement enhancement $$0.8\cdot(P_f/S_b)\cdot F\cdot f_{pk}\cdot U\cdot T_{lifetime} \le L_{dpa},$$3–$$0.8\cdot(P_f/S_b)\cdot F\cdot f_{pk}\cdot U\cdot T_{lifetime} \le L_{dpa},$$4, whereas ASTRA+TGLF indicates somewhat higher confinement, $$0.8\cdot(P_f/S_b)\cdot F\cdot f_{pk}\cdot U\cdot T_{lifetime} \le L_{dpa},$$5–$$0.8\cdot(P_f/S_b)\cdot F\cdot f_{pk}\cdot U\cdot T_{lifetime} \le L_{dpa},$$6, within $$0.8\cdot(P_f/S_b)\cdot F\cdot f_{pk}\cdot U\cdot T_{lifetime} \le L_{dpa},$$7 pedestal-boundary uncertainty. Plasma gain is $$0.8\cdot(P_f/S_b)\cdot F\cdot f_{pk}\cdot U\cdot T_{lifetime} \le L_{dpa},$$8 at $$0.8\cdot(P_f/S_b)\cdot F\cdot f_{pk}\cdot U\cdot T_{lifetime} \le L_{dpa},$$9 and $$P_f/S_b \le \frac{L_{dpa}}{0.8\cdot F\cdot f_{pk}\cdot U\cdot T_{lifetime}}.$$0 at $$P_f/S_b \le \frac{L_{dpa}}{0.8\cdot F\cdot f_{pk}\cdot U\cdot T_{lifetime}}.$$1; electric gain is $$P_f/S_b \le \frac{L_{dpa}}{0.8\cdot F\cdot f_{pk}\cdot U\cdot T_{lifetime}}.$$2 in Case A and $$P_f/S_b \le \frac{L_{dpa}}{0.8\cdot F\cdot f_{pk}\cdot U\cdot T_{lifetime}}.$$3 in Case B (Cohen et al., 5 May 2026).

For Case A, the reported plasma state has $$P_f/S_b \le \frac{L_{dpa}}{0.8\cdot F\cdot f_{pk}\cdot U\cdot T_{lifetime}}.$$4 with $$P_f/S_b \le \frac{L_{dpa}}{0.8\cdot F\cdot f_{pk}\cdot U\cdot T_{lifetime}}.$$5, $$P_f/S_b \le \frac{L_{dpa}}{0.8\cdot F\cdot f_{pk}\cdot U\cdot T_{lifetime}}.$$6, $$P_f/S_b \le \frac{L_{dpa}}{0.8\cdot F\cdot f_{pk}\cdot U\cdot T_{lifetime}}.$$7, $$P_f/S_b \le \frac{L_{dpa}}{0.8\cdot F\cdot f_{pk}\cdot U\cdot T_{lifetime}}.$$8, $$P_f/S_b \le \frac{L_{dpa}}{0.8\cdot F\cdot f_{pk}\cdot U\cdot T_{lifetime}}.$$9, and $0.8$0. Pedestal values include $0.8$1–$0.8$2, $0.8$3, and $0.8$4–$0.8$5 for Case A, with higher-density Case B giving $0.8$6–$0.8$7 (Cohen et al., 5 May 2026).

Heating and current drive are supplied by $0.8$8 ICRH at $0.8$9 in an H-minority/$F$0 scenario, together with alpha heating; ohmic power is about $F$1 at flattop, and no NBI is included. Conservative ITPA L–H threshold scaling with low-density correction suggests $F$2 at the minimum-threshold density. The operational strategy described is a clean L-mode ramp, double-null positioning, and impurity seeding only after H-mode establishment, with optional additional pulsed ICRH of about $F$3–$F$4 kept as a design reserve for access margin (Cohen et al., 5 May 2026).

The TF magnet system comprises 18 D-shaped wedged inboard coils, with 6 CS coils and 8 PF coils, all HTS. The conductor is REBCO stacked-tape CroCo (“SHIELD”) operating at 20 K. Each TF coil has 188 turns, terminal current $F$5, single-coil inductance about $F$6, and stored energy about $F$7, giving a full TF stored energy of about $F$8. The TF case material is SS316LN with peak von Mises stress about $F$9 under centering load, compared with allowable yield about $B_0 = 9.29~\mathrm{T}$00 at cryogenic temperature. The TF case radial thickness is about $B_0 = 9.29~\mathrm{T}$01 and winding-pack current density about $B_0 = 9.29~\mathrm{T}$02 (Cohen et al., 5 May 2026).

4. Heating physics, transport corroboration, and edge operating space

ASTRA+TGLF corroboration is used to test the robustness of the FUSE baseline. For the baseline edge boundary, ASTRA+TGLF predicts fusion power of about $B_0 = 9.29~\mathrm{T}$03 in Case A and about $B_0 = 9.29~\mathrm{T}$04 in Case B, slightly above FUSE but within uncertainty. The paper emphasizes sensitivity to edge conditions: $B_0 = 9.29~\mathrm{T}$05 pedestal-top variations produce large fusion-power swings, from about $B_0 = 9.29~\mathrm{T}$06 to $B_0 = 9.29~\mathrm{T}$07 in Case A and about $B_0 = 9.29~\mathrm{T}$08 to $B_0 = 9.29~\mathrm{T}$09 in Case B. This is interpreted as expected behavior in a stiff core transport regime (Cohen et al., 5 May 2026).

The same analysis finds that heating localization, with the ICRH peak moved over $B_0 = 9.29~\mathrm{T}$10 to $B_0 = 9.29~\mathrm{T}$11, and $B_0 = 9.29~\mathrm{T}$12 coupling changes have minor effect. The stated conclusion is that edge fueling and D–T ratio are better actuators for $B_0 = 9.29~\mathrm{T}$13 control than modest core-heating adjustments. Core impurity scans over $B_0 = 9.29~\mathrm{T}$14–$B_0 = 9.29~\mathrm{T}$15 with fixed edge content leave $B_0 = 9.29~\mathrm{T}$16 nearly constant, while edge radiation dominates dissipation in detached regimes (Cohen et al., 5 May 2026).

ICRH physics is analyzed with CARDS. The selected frequency is $B_0 = 9.29~\mathrm{T}$17, arranged so that H-minority fundamental heating overlaps with deuterium second harmonic at the magnetic axis. First-pass absorption is reported as $B_0 = 9.29~\mathrm{T}$18 at reactor temperatures with H-minority around 1%; complete first-pass absorption also persists for ion temperatures in the $B_0 = 9.29~\mathrm{T}$19–$B_0 = 9.29~\mathrm{T}$20 range if H-minority near 1% is maintained at low $B_0 = 9.29~\mathrm{T}$21, while at about $B_0 = 9.29~\mathrm{T}$22 even 0% minority can suffice through second-harmonic D damping. Electron Landau damping broadens the deposition profile, minority fraction can tune the radial width of ion absorption, fast $B_0 = 9.29~\mathrm{T}$23 alpha absorption is small at about 7.7% in the nominal case, and third-harmonic tritium absorption is neglected as expected to be small (Cohen et al., 5 May 2026).

The edge operating window is interpreted with SepOS. Empirical scalings give a very narrow scrape-off-layer width, $B_0 = 9.29~\mathrm{T}$24–$B_0 = 9.29~\mathrm{T}$25 at the midplane. The paper cites the Eich 2020 and Brunner 2018 scalings:

• $B_0 = 9.29~\mathrm{T}$26,
• $B_0 = 9.29~\mathrm{T}$27.

Power-exhaust severity is screened with $B_0 = 9.29~\mathrm{T}$28, which is moderated by the deliberately modest $B_0 = 9.29~\mathrm{T}$29–$B_0 = 9.29~\mathrm{T}$30, yielding values around $B_0 = 9.29~\mathrm{T}$31–$B_0 = 9.29~\mathrm{T}$32. A second dissipation metric, $B_0 = 9.29~\mathrm{T}$33, is described as more challenging because H-mode edge densities are constrained, motivating impurity-seeded detachment and potentially operation in Quasi-Continuous Exhaust windows at higher $B_0 = 9.29~\mathrm{T}$34 (Cohen et al., 5 May 2026).

SepOS further indicates that for shaped plasmas like Yinsen, QCE is accessible at higher $B_0 = 9.29~\mathrm{T}$35 and moderate $B_0 = 9.29~\mathrm{T}$36 if the normalized edge-pressure gradient parameter $B_0 = 9.29~\mathrm{T}$37 within the H-mode window. This is presented as consistent with a bias toward higher-$B_0 = 9.29~\mathrm{T}$38 operation to improve compatibility with stable detachment and ELM mitigation (Cohen et al., 5 May 2026).

5. Divertor detachment and thermal-hydraulic feasibility

The divertor analysis begins from the severity of the unmitigated case. A 2D Braginskii fluid model in UEDGE, with diffusive neutrals and fixed-fraction impurity radiation, gives for the base case $B_0 = 9.29~\mathrm{T}$39 and $B_0 = 9.29~\mathrm{T}$40 an unmitigated peak outer-target heat flux of about $B_0 = 9.29~\mathrm{T}$41 (Cohen et al., 5 May 2026).

Neon-seeded detached operation is then shown to be accessible. At $B_0 = 9.29~\mathrm{T}$42, about 2.3% Ne yields deep detachment with peak target heat flux $B_0 = 9.29~\mathrm{T}$43 and peak plate electron temperature below $B_0 = 9.29~\mathrm{T}$44, with neutrals dominating near the target. At $B_0 = 9.29~\mathrm{T}$45, only about $B_0 = 9.29~\mathrm{T}$46–$B_0 = 9.29~\mathrm{T}$47 Ne is needed to reach $B_0 = 9.29~\mathrm{T}$48, and often below $B_0 = 9.29~\mathrm{T}$49–$B_0 = 9.29~\mathrm{T}$50. The operational window identified is $B_0 = 9.29~\mathrm{T}$51–$B_0 = 9.29~\mathrm{T}$52 with Ne about $B_0 = 9.29~\mathrm{T}$53–$B_0 = 9.29~\mathrm{T}$54 and $B_0 = 9.29~\mathrm{T}$55–$B_0 = 9.29~\mathrm{T}$56, keeping peak target heat flux well below $B_0 = 9.29~\mathrm{T}$57 and often near $B_0 = 9.29~\mathrm{T}$58–$B_0 = 9.29~\mathrm{T}$59 (Cohen et al., 5 May 2026).

The target concept is an ITER-inspired tungsten monoblock with FLiBe cooling through a smooth circular tube and with no swirl enhancement credited. The stated geometry is 28 mm W armor, a centered 14 mm FLiBe channel with Cu/CuCrZr annuli, minimum tungsten thickness to coolant about 5 mm, and a 35 mm V–4Cr–4Ti backing wall with bulk FLiBe tank. Uniform volumetric nuclear heating is taken as $B_0 = 9.29~\mathrm{T}$60 in W and $B_0 = 9.29~\mathrm{T}$61 in the vacuum vessel, with an added $B_0 = 9.29~\mathrm{T}$62 from core radiation (Cohen et al., 5 May 2026).

For the specific detached case with $B_0 = 9.29~\mathrm{T}$63, $B_0 = 9.29~\mathrm{T}$64, and Ne = 2.0%, the thermal-hydraulic model gives, at FLiBe velocity $B_0 = 9.29~\mathrm{T}$65 and diameter $B_0 = 9.29~\mathrm{T}$66, an effective smooth-tube heat-transfer coefficient $B_0 = 9.29~\mathrm{T}$67, pressure drop per unit length $B_0 = 9.29~\mathrm{T}$68, and peak tungsten surface temperature about $B_0 = 9.29~\mathrm{T}$69. A parameter sweep over $B_0 = 9.29~\mathrm{T}$70–$B_0 = 9.29~\mathrm{T}$71 and $B_0 = 9.29~\mathrm{T}$72–$B_0 = 9.29~\mathrm{T}$73 gives peak tungsten surface temperatures of about $B_0 = 9.29~\mathrm{T}$74–$B_0 = 9.29~\mathrm{T}$75 across the detached-profile envelope (Cohen et al., 5 May 2026).

Across four representative detached UEDGE cases, peak tungsten surface temperatures are at or below about $B_0 = 9.29~\mathrm{T}$76. The most demanding case, $B_0 = 9.29~\mathrm{T}$77, $B_0 = 9.29~\mathrm{T}$78, Ne = 0.8%, gives about $B_0 = 9.29~\mathrm{T}$79, slightly above a conservative recrystallization reference of about $B_0 = 9.29~\mathrm{T}$80 in the no-swirl model; increasing Ne to 0.9% lowers the peak to about $B_0 = 9.29~\mathrm{T}$81. The conclusion stated in the paper is that impurity-seeded detachment can reduce peak target heat flux well below $B_0 = 9.29~\mathrm{T}$82, often to roughly $B_0 = 9.29~\mathrm{T}$83–$B_0 = 9.29~\mathrm{T}$84, making a FLiBe-cooled tungsten monoblock feasible, while explicit swirl enhancement would lower temperatures further but is not credited in the present model (Cohen et al., 5 May 2026).

A common misconception would be to read the narrow SOL width alone as dispositive against a compact high-field divertor. The Yinsen analysis instead argues that the relevant question is whether the operating point can be moved into a detached regime with compatible $B_0 = 9.29~\mathrm{T}$85 and impurity fraction. The concept’s exhaust feasibility therefore depends less on raw $B_0 = 9.29~\mathrm{T}$86 and more on the coupled edge-density, seeding, and detachment window identified in UEDGE and SepOS (Cohen et al., 5 May 2026).

6. Neutronics, breeding, activation, and maintenance strategy

OpenMC neutronics with a detailed radial build including tungsten first wall, FLiBe channels and tank, V–4Cr–4Ti structures, WC/W$B_0 = 9.29~\mathrm{T}$87B$B_0 = 9.29~\mathrm{T}$88 shields, thermal shield, and magnets provides the main neutronics results. The design attains a tritium breeding ratio of about 1.10 with approximately 30% $B_0 = 9.29~\mathrm{T}$89 enrichment in FLiBe and no dedicated multiplier layer, with blanket energy multiplier $B_0 = 9.29~\mathrm{T}$90. The paper restates the definition

$B_0 = 9.29~\mathrm{T}$91

and notes that the target $B_0 = 9.29~\mathrm{T}$92 provides margin for processing losses and inventory growth (Cohen et al., 5 May 2026).

The vacuum vessel is the lifetime-limiting solid structure. The peak inner-VV DPA rate is $B_0 = 9.29~\mathrm{T}$93, corresponding at 130 MW and 40% utilization to about $B_0 = 9.29~\mathrm{T}$94 calendar, hence 35 DPA in 20 years and $B_0 = 9.29~\mathrm{T}$95. Helium production is described as a few appm/DPA, with an example inner-VV peak of $B_0 = 9.29~\mathrm{T}$96, remaining below regimes dominated by He-induced embrittlement (Cohen et al., 5 May 2026).

The HTS magnets are characterized as lifetime components rather than near-term replacement items. The fast-neutron flux at the TF coils is $B_0 = 9.29~\mathrm{T}$97 per MW$B_0 = 9.29~\mathrm{T}$98, giving a TF fast-flux lifetime of about $B_0 = 9.29~\mathrm{T}$99, or roughly sixteen vacuum-vessel lifetimes. Total TF nuclear heating is about $I_p = 9.67~\mathrm{MA}$00, which becomes about $I_p = 9.67~\mathrm{MA}$01 at $I_p = 9.67~\mathrm{MA}$02 for a 20 K cold mass; even at $I_p = 9.67~\mathrm{MA}$03 it is about $I_p = 9.67~\mathrm{MA}$04 (Cohen et al., 5 May 2026).

Neutron attenuation through the shield is substantial, with flux reduction by almost four orders of magnitude between first wall, around $I_p = 9.67~\mathrm{MA}$05, and TF conductor, around $I_p = 9.67~\mathrm{MA}$06, at 130 MW. Shield optimization studies report that thinning the low-field-side W$I_p = 9.67~\mathrm{MA}$07B$I_p = 9.67~\mathrm{MA}$08 shield while holding an overall TF nuclear-heating limit can reduce W$I_p = 9.67~\mathrm{MA}$09B$I_p = 9.67~\mathrm{MA}$10 mass from about 741 t to about 563 t without compromising TF lifetime, and magnet-adjacent 10 wt% Hf doping cuts TF fast flux by about 9–11% at the TF case and thermal shield (Cohen et al., 5 May 2026).

Activation and shielding analyses feed directly into maintenance strategy. End-of-life activation at $I_p = 9.67~\mathrm{MA}$11 produces near-constant high dose for about 24 h and then decays over decades; major structural components fall below the cited recycling threshold of about $I_p = 9.67~\mathrm{MA}$12 at roughly 50 years. A 4 m ordinary-concrete bioshield around the cryostat reduces shutdown dose to below about $I_p = 9.67~\mathrm{MA}$13–$I_p = 9.67~\mathrm{MA}$14 at the outer surface over a wide cooldown range (Cohen et al., 5 May 2026).

The maintenance concept therefore distinguishes routine and non-routine interventions. Routine work proceeds through ports with JET-like remote handling for tiles, antennae, and diagnostics. If emergency vessel replacement were ever required, the proposed “cut-in-half/slide/rotate-out” extraction of the vacuum-vessel halves through the TF cage is to be performed in a flooded maintenance hall, because water flooding collapses the high-dose zone by orders of magnitude (Cohen et al., 5 May 2026).

The paper also addresses proliferation resistance through deliberate fertile-salt doping with UF$I_p = 9.67~\mathrm{MA}$15/ThF$I_p = 9.67~\mathrm{MA}$16 at 0.5–10 wt%. The resulting fissile accumulations are stated to be tiny, with the example of about 1.5 kg Pu-239 or U-233 after about 0.4 FPY at 5 wt%, well below the 8 kg IAEA significant quantity. Industrial pyrochemical processing in shielded hot cells would be required, which the paper argues is beyond plausible clandestine operation (Cohen et al., 5 May 2026).

7. Power conversion, pulsed operation, and FOAK significance

The balance of plant uses FLiBe primary loops and a regenerated supercritical CO$I_p = 9.67~\mathrm{MA}$17 Brayton cycle. The FLiBe operating window is about 800–925 K, with 925 K blanket-tank outlet and 800 K return into the heat exchanger. Five primary loops are identified: blanket tank, inboard first-wall channel, outboard first-wall channel, upper divertor, and lower divertor (Cohen et al., 5 May 2026).

For Case B at $I_p = 9.67~\mathrm{MA}$18, the representative loop loads and flows are given as about 91.5 MW and 315 kg/s for the blanket tank, 26.1 MW and 89 kg/s for the inboard first wall, 49.8 MW and 170 kg/s for the outboard first wall, and about 19.2 MW and 59–61 kg/s for each divertor. The sCO$I_p = 9.67~\mathrm{MA}$19 secondary operates from 8 to 25 MPa with compressor inlet around 305–310 K, compact turbomachinery, and no intercooling because the complexity is judged unjustified by the small efficiency gain. Integrated cycle efficiency is about 47.5–47.7% over 130–480 MW$I_p = 9.67~\mathrm{MA}$20, while plant efficiency rises from about 42.6% to about 50.4% as fixed parasitics are diluted at higher power (Cohen et al., 5 May 2026).

Pulse transients are specified as 30 s rise, 900 s flattop, 30 s fall, and 60 s dwell. The FLiBe loops track pulsed loads, CO$I_p = 9.67~\mathrm{MA}$21 mass flow adjusts, and the turbine remains productive in dwell due to thermal inertia, reportedly sufficient to cover house loads without violating margins (Cohen et al., 5 May 2026).

The pulsed-power architecture uses a 34 kV AC medium-voltage backbone and a shared 1.5 kV DC magnet bus via 3.3 kV SiC solid-state transformer cells, 17 in series per phase. Magnet supplies are modular 5 kA, 1.5 kV H-bridge units in parallel; CS/PF systems have crowbar and quench protection; the TF supply is tailored for 15-minute charge/discharge; and fast discharge uses hybrid interrupters and dump resistors. Six grouped fast-discharge units each protect three TF coils, about 4.5 GJ per branch (Cohen et al., 5 May 2026).

Energy storage is a hybrid battery/capacitor system used for ride-through and pulse buffering. The baseline pulse usable energy requirement is about $I_p = 9.67~\mathrm{MA}$22. The stated cost optimum is about 79% capacitor share by pulse energy on a lifetime-equivalent basis, while nameplate capacity is dominated by battery, with about 8.75 MWh battery and about 0.66 MWh capacitor, corresponding to about 31.5 GJ and 2.38 GJ nameplate respectively, for a total screening-basis cost of about 6.7 MUSD. With recirculating loads excluding RF of about 10 MWe, battery-only ride-through is about 50 min from full charge, and a 10 MWe diesel backup can carry house loads for controlled standby or shutdown (Cohen et al., 5 May 2026).

Pulse waveform analysis using the present TokaMaker pulse gives post-breakdown loop voltage falling to about 0.04 V in flattop. Flux consumption is about 20 Wb in ramp-up plus about 40 Wb in flattop, for about 60 Wb per pulse. The CS available swing is about $I_p = 9.67~\mathrm{MA}$23 per polarity, or about 90 Wb total, supporting pulses longer than 15 minutes with continued optimization. The paper further suggests that ramp-down to a cold, low-current plasma between pulses may avoid repeated breakdown-flux penalties, though this is left as future control work (Cohen et al., 5 May 2026).

The broader significance of the concept is defined comparatively. Yinsen argues that high-field compact tokamaks can achieve strong confinement but worsen divertor power density, neutron wall load, irradiation damage rate, downstream nuclear heating, and maintenance burden when pushed toward grid-competitive power density. By instead setting $I_p = 9.67~\mathrm{MA}$24, the concept claims to preserve the vacuum vessel and primary structures for the full plant life, keep cryogenic nuclear heating low, maintain long magnet lifetime, operate with a manageable detached divertor, and still achieve $I_p = 9.67~\mathrm{MA}$25 and useful net electric output in long inductive pulses (Cohen et al., 5 May 2026).

This suggests that Yinsen is best understood not as an optimized endpoint reactor, but as a deliberately constrained FOAK operating point. The paper’s closing perspective is that such a reactor can bridge the gap between $I_p = 9.67~\mathrm{MA}$26 demonstrations and later grid-scale systems by targeting application classes where autonomy, fuel security, and dispatchability dominate value. Upgrade paths are explicitly contemplated, including higher $I_p = 9.67~\mathrm{MA}$27 for shorter structural life or higher utilization as materials and operations mature (Cohen et al., 5 May 2026).