Einstein-Szilard Pump: MHD Sizing (With Worksheet)

⏱ 28 min read

Lorentz Force Pumping in Liquid Metal Conduits

An Einstein-Szilard electromagnetic pump moves electrically conducting liquid metal through a sealed channel by applying an orthogonal magnetic field and an electric current transverse to the flow axis. This configuration generates a continuous, longitudinal Lorentz force vector (\(F = I \times B\)) that pressurises the fluid without mechanical seals, impellers, or moving components. By eliminating rotating shafts and packing glands entirely, the design eliminates dynamic seal failure paths while maintaining volumetric flow through an enclosed conduit.

Magnetohydrodynamics refers to the study of the physical interactions between magnetic fields and electrically conducting fluids like liquid metals or plasmas, where magnetic induction and fluid momentum couple directly.

Albert Einstein and Leo Szilard conceptualised this conduction pump in 1926 after reading a Berlin newspaper report detailing the deaths of an entire family from leaking toxic refrigerant fumes. At the time, domestic refrigeration relied on sulfur dioxide, methyl chloride, or ammonia gases pressurised by mechanical compressors. The shaft seals of the era degraded predictably, leaking toxic chemicals into residential living spaces.

Einstein and Szilard filed their initial German patent application (DE 476,812) in 1926, which was later granted in the United States as patent US 1,781,541 in 1930. Instead of attempting incremental improvements to elastomer gaskets, they eliminated moving parts by using a liquid sodium-potassium alloy as an electrically driven piston to compress refrigerant vapour. Engineering teams analyzing the history of zero-leakage systems often evaluate the Einstein-Szilard Fridge: Clean-Tech Teardown (Blueprint) to study this root-cause mechanical decoupling.

Translating this theoretical mechanism into an industrial fluid delivery system introduces severe magnetohydrodynamic (MHD) trade-offs. The pump requires high electrical conductivity to generate the requisite volume force:

\(f = \sigma (E – u B) B\)

where \(\sigma\) is electrical conductivity, \(E\) is applied electric field, \(u\) is mean flow velocity, and \(B\) is magnetic flux density. However, operating within rectangular duct geometries forces a conflict between Lorentz thrust and viscous dissipation. In standard rectangular conduits, magnetic flux lines perpendicular to conducting sidewalls create Hartmann boundary layers.

As detailed in R. Moreau’s foundational text Magnetohydrodynamics (Kluwer Academic Publishers), the thickness of these boundary layers scales inversely with the Hartmann number (\(Ha = B L \sqrt{\sigma / \mu}\), where \(L\) is characteristic duct width and \(\mu\) is dynamic viscosity). When \(Ha\) exceeds 100, velocity profiles flatten across the core and steepen at the walls, driving wall shear stress to multiples of ordinary Poiseuille flow.

Parallel side walls (the Roberts layers) also generate secondary recirculation cells. These vortex structures loop current through the boundary layers, dissipating electrical energy as Joule heating rather than mechanical pressure. If you size the duct too wide to lower viscous resistance, the required electromagnetic gap expands, diluting the magnetic field strength and collapsing your pressure head.

Balancing these losses requires calculating your exact operating point on the flow-pressure curve before cutting channel stock. Run your specific channel dimensions through the step-by-step sizing worksheet below to verify your pump develops adequate head against downstream viscous resistance.

Table of Contents

• Lorentz Force Pumping in Liquid Metal Conduits
• Governing Magnetohydrodynamic Equations for Pressure and Flow
• Electrical Conductivity and Working Fluid Selection Trade-Offs
• Direct-Current Conduction Versus Alternating-Current Induction Configurations
• Step-by-Step Channel Geometry and Magnetic Gap Sizing
• The Copy-Paste Magnetohydrodynamic Sizing Worksheet
• Sources & Further Reading

---

Governing Magnetohydrodynamic Equations for Pressure and Flow

An Einstein-Szilard electromagnetic pump generates static pressure directly within a conducting fluid by applying an orthogonal magnetic field across an injected electric current, yielding a theoretical body force governed by the Lorentz vector relationship \(\frac{dP}{dx} = J_y B_z\). In this steady-state formulation, \(J_y\) represents the cross-channel current density in amperes per square metre (\(\text{A/m}^2\)) and \(B_z\) denotes the applied magnetic flux density in teslas (\(\text{T}\)). Integrating this differential across an active channel length \(L\) provides the ideal static head:

\(\Delta P_{\text{ideal}} = J_y B_z L = \frac{I B_z}{a}\)

where \(I\) is the total cross-channel current in amperes, and \(a\) is the duct dimension parallel to the magnetic field.

Real fluid systems never achieve this peak pressure because motion induces an opposing potential. When liquid sodium or a eutectic sodium-potassium alloy (\(\text{NaK}\)) moves at a mean velocity \(u\) across the magnetic vector \(B_z\), it cuts the magnetic field lines. This motion establishes an internal counter-electromotive force, or back-EMF, equal to:

\(V_{\text{ind}} = u B_z s\)

Here, \(s\) is the channel width between the electrode faces. As documented in the original patent records preserved by the European Patent Office for Albert Einstein and Leo Szilard’s 1928 pumping apparatus, this induced voltage directly subtracts from the applied terminal voltage \(V_{\text{app}}\).

The effective voltage driving current through the fluid volume is therefore \(V_{\text{net}} = V_{\text{app}} – u B_z s\). Because volumetric flow rate \(Q\) equates to \(u \cdot A_{\text{flow}}\) (where \(A_{\text{flow}} = a \cdot s\)), increasing the flow rate dynamically depresses the delivered current density:

\(J_y = \sigma \left( \frac{V_{\text{app}}}{s} – u B_z \right) = \sigma \left( E_y – \frac{Q}{a \cdot s} B_z \right)\)

In this equation, \(\sigma\) represents the electrical conductivity of the liquid metal in siemens per metre (\(\text{S/m}\)). For liquid sodium at \(200^\circ\text{C}\), \(\sigma\) measures roughly \(9.6 \times 10^6\text{ S/m}\), according to engineering property datasets published by Argonne National Laboratory. Consequently, volumetric throughput directly throttles the net electromagnetic pressure boost:

\(\Delta P(Q) = \sigma L B_z \left( \frac{V_{\text{app}}}{s} – \frac{Q B_z}{a s} \right) – \Delta P_{\text{friction}}(Q)\)

Designing around these fluid-electrical interactions mirrors the rigorous modeling seen in the Einstein-Szilard Fridge: Clean-Tech Teardown (Blueprint), where passive operational mechanics displace mechanical compressors.

       APPLIED CURRENT (Jy)
               │
               ▼
     ┌───────────────────┐
     │  Conductive Fluid │
     │  Channel (NaK)    │ ──► Flow (Q)
     └───────────────────┘
               ▲
               │
      MAGNETIC FLUX (Bz)

The pump channel does not display standard parabolic laminar velocity distributions. Instead, the interaction between viscous and magnetic phenomena is governed by the Hartmann number (\(Ha\)), formulated as:

\(Ha = B_z L_{\text{char}} \sqrt{\frac{\sigma}{\mu}}\)

In this dimensionless group, \(\mu\) represents dynamic viscosity in pascal-seconds (\(\text{Pa}\cdot\text{s}\)) and \(L_{\text{char}}\) is the characteristic channel radius (\(s/2\)). As Julius Adams Stratton established in his fundamental treatise Electromagnetic Theory, when \(Ha \gg 1\), magnetic Lorentz forces dominate internal fluid shear. For an annular channel operating with \(Ha \approx 120\), the velocity profile flattens into an expansive, uniform plug across roughly 95% of the cross-section.

This profile reshaping forces steep velocity gradients into very thin boundary layers along the walls perpendicular to the applied flux. The thickness of this Hartmann boundary layer scales strictly as:

\(\delta_{Ha} \approx \frac{L_{\text{char}}}{Ha}\)

Thinning this layer accelerates wall shear rates, driving hydraulic friction losses well beyond standard Darcy-Weisbach pipe friction estimates. Designers running a systematic Ideation to Prototype Workflow must calculate these elevated skin-friction penalties when sizing coil supplies, rather than relying on classic empirical pipe-flow tables.

A second efficiency loss stems from electrical shunting through the physical duct walls. In a conduction MHD channel, the liquid metal is contained within a solid metallic conduit, typically austenitic 316L stainless steel, to withstand corrosives at high operational temperatures. Because the channel walls contact both electrodes, the applied voltage drives current through both the fluid (\(R_{\text{fluid}}\)) and the containment envelope (\(R_{\text{wall}}\)) in parallel.

You compute these resistance paths based on channel geometry:

\(R_{\text{fluid}} = \frac{s}{\sigma_{\text{fluid}} \cdot a L}\)

\(R_{\text{wall}} = \frac{s}{2 \cdot \sigma_{\text{wall}} \cdot t_{\text{wall}} L}\)

Here, \(t_{\text{wall}}\) denotes the duct wall thickness. The fraction of total current wasted in wall bypass heating is defined by the wall conductance ratio \(C = \frac{R_{\text{fluid}}}{R_{\text{wall}}} = \frac{\sigma_{\text{wall}} t_{\text{wall}}}{\sigma_{\text{fluid}} a}\). At \(300^\circ\text{C}\), 316L stainless steel has an electrical conductivity of \(1.3 \times 10^6\text{ S/m}\), while pure sodium exhibits \(7.8 \times 10^6\text{ S/m}\).

If your fabrication requires a wall thickness \(t_{\text{wall}} = 1.5\text{ mm}\) over a channel height \(a = 15\text{ mm}\), the wall conductance ratio \(C\) reaches roughly 0.016, pulling approximately 1.6% of input current directly through the structural metal. If you substitute a corrosive liquid alloy with a lower conductivity or thicken the structural boundaries to contain higher pressures, \(C\) can rapidly exceed 0.25, bleeding over 20% of your electrical input before fluid acceleration begins.

Thermal gradients add another operational variable. Liquid sodium’s resistivity rises from \(9.6 \times 10^{-8}\ \Omega\cdot\text{m}\) at \(100^\circ\text{C}\) to over \(17.0 \times 10^{-8}\ \Omega\cdot\text{m}\) at \(400^\circ\text{C}\). Uneven heat profiles across the duct trigger internal eddy currents that distort the magnetic flux profile, a phenomenon known as the magnetic Reynolds effect.

Plug your specific duct dimensions, channel materials, and operating temperatures into the dynamic sizing worksheet below to verify whether your configuration achieves positive net pump head.

Electrical Conductivity and Working Fluid Selection Trade-Offs

Designing an Einstein-Szilard electromagnetic pump requires matching the working fluid’s electrical conductivity to the magnetic flux density of the channel, because pump efficiency scales directly with fluid conductivity and drops sharply when duct walls divert the driving current.

Magnetohydrodynamics (MHD) is the study of how electrically conducting fluids interact with magnetic fields, where fluid movement generates induced electric currents that produce mechanical body forces within the flow.

FLUID INPUT
    |
    v
MAGNETIC FLUX (B)
    |
    v
LORENTZ FORCE (F = J x B)
    |
    v
FLOW OUTPUT

The Fluid Matrix: Conductivity versus Viscosity

Four primary fluids dominate conduction-pump engineering: pure liquid sodium, eutectic sodium-potassium alloy (NaK-78), mercury, and gallium alloys such as Galinstan. Selecting the right fluid requires evaluating its electrical conductivity (\(\sigma\)) against its dynamic viscosity (\(\mu\)) and density (\(\rho\)), which dictate hydraulic friction inside the channel.

The table below contrasts physical properties at baseline operating conditions, compiled from data published by the Argonne National Laboratory in its Liquid Metal Handbook and the National Institute of Standards and Technology (NIST):

Fluid Candidate	Operating Temp (°C)	Electrical Conductivity (\(\sigma\), \(10^6\text{ S/m}\))	Dynamic Viscosity (\(\mu\), \(\text{mPa}\cdot\text{s}\))	Density (\(\rho\), \(\text{kg/m}^3\))
Pure Sodium (Na)	110	9.60	0.68	927
NaK-78	20	2.63	0.94	866
Galinstan	20	3.46	2.40	6,440
Mercury (Hg)	20	1.04	1.53	13,534

Pure sodium provides the highest conductivity-to-viscosity ratio, delivering nearly four times the electrical conductivity of NaK-78. However, pure sodium solidifies at 97.8°C, necessitating continuous trace heating along your duct loop.

NaK-78 remains liquid down to -12.6°C, eliminating preheating infrastructure, but sacrifices 72% of pure sodium’s raw electrical conductivity.

Mercury presents an unviable baseline for modern high-performance designs: its high density requires significant pressure gradients to accelerate, while its conductivity is under 11% of sodium’s.

Galinstan offers a non-pyrophoric room-temperature profile, yet its higher dynamic viscosity generates frictional head losses that degrade overall hydraulic efficiency in narrow channels. Balancing these trade-offs often mirrors the technical risk screening outlined in our Einstein-Szilard Fridge: Clean-Tech Teardown (Blueprint).

The Wall-Shunting Penalty

The wall-shunting penalty occurs when electrically conductive metallic duct walls act as parallel resistors, siphoning applied current away from the fluid core and reducing the net Lorentz force.

In an ideal conduction pump, 100% of the input current passes directly through the liquid metal perpendicular to the applied magnetic field. In practice, the total cross-sectional resistance of the channel walls (\(R_w\)) sits in parallel with the resistance of the fluid core (\(R_f\)).

According to Richard S. Baker and Manuel J. Tessier in their definitive monograph Handbook of Electromagnetic Pump Technology, the shunting factor (\(s\)) is defined by the channel geometry and relative conductivities:

\(s = \frac{2 \cdot t_w \cdot \sigma_w}{a \cdot \sigma_f}\)

Where:

• \(t_w\) is the wall thickness (m)
• \(\sigma_w\) is the duct wall material’s electrical conductivity (S/m)
• \(a\) is the channel depth between electrode faces (m)
• \(\sigma_f\) is the fluid electrical conductivity (S/m)

The fraction of total electrical current (\(I_{\text{core}}\)) that actually produces mechanical pumping force equals:

\(I_{\text{core}} = I_{\text{total}} \left( \frac{1}{1 + s} \right)\)

If you construct a narrow 2.0 mm channel using AISI 316 stainless steel (\(\sigma_w = 1.35 \times 10^6\text{ S/m}\)) with a wall thickness of 1.0 mm, and fill it with Galinstan (\(\sigma_f = 3.46 \times 10^6\text{ S/m}\)):

\(s = \frac{2 \times 0.001 \times 1.35 \times 10^6}{0.002 \times 3.46 \times 10^6} = \frac{2,700}{6,920} \approx 0.390\)

This geometry shunts 28.1% of your total input amperage directly through the steel containment, generating pure Joule heating rather than mechanical head. With mercury, the penalty jumps to 56.5% of total current lost to the walls.

Minimising this wall loss requires using thin-walled, low-conductivity refractory alloys like titanium grade 5 or applying insulating internal ceramic coatings, such as yttria-stabilised zirconia. Validating channel materials against structural stress profiles before fabrication is crucial, drawing on principles detailed in our look at Sustainable Product Development Strategies.

Thermal Thresholds and Chemical Passivation

Alkali metals such as sodium and NaK-78 form highly reactive chemical interfaces with structural alloys, particularly under the high shear stress and magnetic flux of an active MHD channel.

At temperatures exceeding 450°C, oxygen impurities inside liquid sodium accelerate the dissolution of nickel and chromium out of austenitic stainless steels. Data from the International Atomic Energy Agency (IAEA) fast reactor materials database shows that maintaining dissolved oxygen levels below 2.0 parts per million (ppm) via cold-trapping reduces uniform corrosion rates to under 0.005 mm per year.

Operating without strict oxide passivation triggers mass transport: nickel dissolves from hot-zone channel walls and precipitates onto cooler heat-exchanger surfaces, rapidly plugging the conduit.

For high-temperature service up to 650°C, the inner channel requires stabilisation using refractory alloys such as molybdenum-base TZM or niobium-stabilised ferritic steels. Integrating sensors to monitor dissolved oxygen and track phase changes maintains system integrity, linking operational safety directly into an Ideation to Prototype Workflow.

Calculating your fluid’s exact electrical performance and duct wall shunting sets the baseline values you need to run the channel cross-section sizing equations in the next step.

Direct-Current Conduction Versus Alternating-Current Induction Configurations

Direct-current conduction pumps maximize peak electrical efficiency at 15% to 20% in small liquid-metal loops, whereas alternating-current induction systems trade efficiency for hermetic reliability by eliminating fluid-contacting electrodes entirely. Your choice between conduction and induction configurations dictates whether your primary failure mode is electrochemical degradation at the electrode boundary or severe thermal dissipation from electromagnetic skin depth limits.

Magnetohydrodynamic pumping is the process of driving an electrically conductive fluid using the Lorentz force generated by the interaction of an applied magnetic field and an electrical current.

Selecting an architecture requires balancing electrical input constraints against physical containment demands, as detailed in our analysis of the original Einstein-Szilard Fridge: Clean-Tech Teardown (Blueprint). In DC conduction units, the current passes directly through the liquid metal via copper or nickel electrodes brazed into the channel walls. According to Richard S. Baker and Manuel J. Tessier in their definitive reference Handbook of Electromagnetic Pump Technology, wetting dynamics govern whether direct current can traverse the solid-fluid interface without massive ohmic losses.

Contact resistance is the opposition to electrical current flow across the microscopic interface between two touching materials, caused by surface roughness, oxide films, or incomplete fluid wetting.

When pumping sodium-potassium eutectic (NaK-78) or liquid lithium, native oxide layers on the channel electrodes cause incomplete interfacial wetting. If the liquid metal fails to wet the solid electrode at a contact angle below 90 degrees, your effective contact surface area drops precipitously. Baker and Tessier documented that unwetted electrodes can increase interfacial electrical resistance by a factor of 10 to 40 within the first 100 hours of operation. This contact degradation shunts power into localized Joule heating rather than Lorentz force generation.

To mitigate this operational decay, practitioners apply high-temperature vacuum baking (typically exceeding 400 °C for 4 hours) combined with dynamic chemical gettering using zirconium foil to remove dissolved oxygen below 10 parts per million. As shown in modern Sustainable Product Development Strategies, designing for long-term component durability requires engineering these boundary interfaces prior to fluid charging. In systems running over 10,000 continuous hours, even wetted DC electrodes suffer from mass transfer erosion, where the high-density current accelerates the dissolution of the parent electrode metal into the working fluid.

Alternating-current induction designs—such as Flat Linear Induction Pumps (FLIP) and Annular Linear Induction Pumps (ALIP)—remove wet electrodes by using polyphase stator windings to induce travelling magnetic fields directly inside the fluid duct. The operational data compiled by the Argonne National Laboratory demonstrates that while induction eliminates fluid corrosion risks, it introduces inductive penetration depth (skin effect) constraints that cap the maximum channel cross-sectional dimensions.

Skin depth is the distance beneath the surface of a conductor where an alternating electromagnetic field decays to roughly 37 percent of its surface amplitude.

The standard penetration depth (\(\delta\)) is calculated by:

\(\delta = \sqrt{\frac{1}{\pi f \mu \sigma}}\)

In this equation, \(f\) is the excitation frequency in Hertz, \(\mu\) is the magnetic permeability of the fluid in Henrys per metre, and \(\sigma\) is the electrical conductivity in Siemens per metre. For sodium at 200 °C (\(\sigma \approx 9.6 \times 10^6 \text{ S/m}\)) powered by a standard 60 Hz line frequency, the skin depth is approximately 21 mm.

If your channel thickness exceeds twice this penetration depth (\(2\delta \approx 42\text{ mm}\)), the induced magnetic flux cannot penetrate the core of the conduit. The central fluid experiences zero Lorentz force, becoming dead volume that generates viscous drag against the accelerated boundary layers. For large flow channels, engineers must either lower the AC drive frequency down to 10–20 Hz using custom variable-frequency power supplies or restrict the pump to narrow annular passages wrapped around central magnetic return cores.

Now that you can calculate how frequency constraints restrict channel dimensions, plug your target flow velocity into the cross-sectional geometry formulas in the worksheet below to establish your actual fluid pressure envelope.

Step-by-Step Channel Geometry and Magnetic Gap Sizing

Dimensioning the channel geometry and magnetic gap of an Einstein-Szilard electromagnetic pump requires balancing the Hartmann velocity profile against electrical bypass losses to achieve net positive pressure. Magnetohydrodynamic pumping is a method of moving electrically conductive fluids using the Lorentz force, which is generated when an electric current passes through the fluid in the presence of a perpendicular magnetic field. When engineering these systems for alkali metals like eutectic sodium-potassium (NaK-78), geometric trade-offs govern hydraulic efficiency and thermal margins.

Step 1: Calculate Channel Cross-Section and Velocity Boundaries

Channel sizing begins by establishing an operating velocity that keeps flow within the laminar regime or controls the transition to turbulent flow under magnetic damping. In magnetic fields above 0.5 Tesla, the dimensionless Hartmann number (\(Ha\)) suppresses turbulence, shifting the critical Reynolds number (\(Re_{cr}\)) upward according to relationships documented by J.A. Shercliff in A Textbook of Magnetohydrodynamics published by Cambridge University Press.

\(\text{Re}_{cr} \approx 225 \times \text{Ha}\)

For an application targeting a flow rate of \(1.50 \times 10^{-4}\ \text{m}^3/\text{s}\) (9.0 L/min) of NaK-78 at 300 K (density \(\rho = 855\ \text{kg/m}^3\), dynamic viscosity \(\mu = 5.78 \times 10^{-4}\ \text{Pa}\cdot\text{s}\)), maintain a mean velocity (\(v\)) between 0.8 m/s and 1.5 m/s. An active channel height (\(h\), parallel to the magnetic field) of 5.0 mm and a channel width (\(w\), along the current path) of 25.0 mm yields a cross-sectional area of \(1.25 \times 10^{-4}\ \text{m}^2\). This configuration results in a fluid velocity of 1.20 m/s and a hydraulic diameter (\(D_h\)) of 8.33 mm.

At 1.20 m/s, the uninhibited Reynolds number is 14,792. Applying a 0.8 Tesla transverse field yields a Hartmann number of 288, which lifts \(Re_{cr}\) past 64,000, suppressing standard boundary-layer turbulent eddies and flattening the velocity profile into an M-shaped Hartmann boundary layer.

       MAGNETIC CORE (POLE 1)
-------------------------------------
==== [Top Hastelloy Wall: 0.5 mm] ===
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
-->  Fluid Channel: 5.0 mm x 25.0 mm 
-->  Active Gap: NaK-78 at 1.2 m/s   
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
=== [Bottom Hastelloy Wall: 0.5 mm] ==
-------------------------------------
       MAGNETIC CORE (POLE 2)

Step 2: Size the Magnetic Gap for 0.4 to 1.2 Tesla Flux Density

Magnetic flux density (\(B\)) across the active fluid channel dictates the developed Lorentz force density (\(F = J \times B\)). The total physical air gap (\(g_{total}\)) must account for the fluid channel height (\(h = 5.0\ \text{mm}\)), two wall thicknesses (\(2 \times t_w = 1.0\ \text{mm}\)), and thermal isolation clearance (\(1.0\ \text{mm}\)), totaling \(7.0\ \text{mm}\).

According to design guidelines from R.S. Baker and M.J. Tessier in the Handbook of Electromagnetic Pump Technology, keeping flux leakage below 15% requires pole face dimensions at least three times the total gap length. For our 7.0 mm gap, the pole faces must measure at least 21.0 mm along the flow axis.

Using grade N42 neodymium-iron-boron (NdFeB) permanent magnets with a remanence (\(B_r\)) of 1.30 Tesla at 20°C, the required magnet thickness (\(l_m\)) per pole to achieve a 0.80 Tesla gap field through mild steel pole pieces is calculated using Ampère’s circuital law:

\(l_m = \frac{B_g \times g_{total} \times \sigma_f}{\mu_r \times (B_r – B_g)}\)

Assuming a flux leakage factor (\(\sigma_f\)) of 1.20 and a relative recoil permeability (\(\mu_r\)) of 1.05, the stack requires an active magnet thickness of 12.8 mm per pole. For temperatures above 180°C, substitute samarium-cobalt (Sm2Co17) alloys rated for 300°C service, sizing the core stack to compensate for its lower remanence of 1.10 Tesla.

Engineers evaluating structural tradeoffs between magnetic core mass and alloy fatigue during system integration can apply structured Concept Development Strategies to benchmark competing layouts against operational thermal limits.

Step 3: Calculate Drive Current and Back-EMF Compensation

The total electric current (\(I_{total}\)) supplied to the fluid electrodes must generate the required pressure head (\(\Delta P\)) while overcoming the opposing counter-electromotive force (back-EMF). Back-EMF is the voltage generated across a moving conductor inside a magnetic field that acts against the applied electrical potential.

Back-EMF (\(V_{ind}\)) scales directly with fluid velocity, flux density, and channel width:

\(V_{ind} = v \times B \times w\)

At \(v = 1.20\ \text{m/s}\), \(B = 0.80\ \text{Tesla}\), and channel width \(w = 0.025\ \text{m}\), the induced potential is:

\(V_{ind} = 1.20 \times 0.80 \times 0.025 = 0.024\ \text{V}\ (24.0\ \text{mV})\)

To generate a net pressure head (\(\Delta P\)) of 85.0 kPa (\(8.50 \times 10^4\ \text{N/m}^2\)) across an active magnetic interaction length (\(L = 0.060\ \text{m}\)), determine the effective current passing through the liquid metal:

\(I_{fluid} = \frac{\Delta P \times h}{B} = \frac{85,000 \times 0.005}{0.80} = 531.25\ \text{A}\)

Liquid NaK-78 at 300 K has an electrical conductivity (\(\sigma\)) of \(1.67 \times 10^6\ \text{S/m}\), yielding a bulk fluid resistance (\(R_{fluid}\)) across the 25.0 mm electrode width:

\(R_{fluid} = \frac{w}{\sigma \times h \times L} = \frac{0.025}{1.67 \times 10^6 \times 0.005 \times 0.060} = 4.99 \times 10^{-5}\ \Omega\)

The terminal drive voltage (\(V_{applied}\)) required across the electrodes equals the internal ohmic drop plus back-EMF:

\(V_{applied} = (I_{fluid} \times R_{fluid}) + V_{ind} = (531.25 \times 4.99 \times 10^{-5}) + 0.024 = 0.0505\ \text{V}\ (50.5\ \text{mV})\)

The drive system must deliver 531.25 A at 50.5 mV to the fluid, making low-resistance copper busbar connections mandatory to avoid external resistive voltage drops.

Step 4: Dimension Wall Geometry to Cap Bypass Current Below 10%

Conductive duct walls run electrically parallel to the fluid, acting as a parasitic resistive shunt that drains supply current away from the pumping zone. To maintain electrical efficiency above 90%, limit the wall bypass current (\(I_{wall}\)) to less than 10% of \(I_{fluid}\) (\(I_{wall} \le 53.1\ \text{A}\)).

The total resistance of both the upper and lower channel containment walls (\(R_{wall}\)) must satisfy:

\(R_{wall} \ge \frac{V_{applied}}{I_{wall,max}} = \frac{0.0505\ \text{V}}{53.1\ \text{A}} = 9.51 \times 10^{-4}\ \Omega\)

Using Hastelloy C-276, which features an electrical resistivity of \(1.30 \times 10^{-6}\ \Omega\cdot\text{m}\) according to technical data sheets from Haynes International, the resistance across the two parallel walls of thickness \(t_w\) is:

\(R_{wall} = \frac{\rho_{wall} \times w}{2 \times t_w \times L}\)

Solving for the maximum permissible wall thickness:

\(t_w \le \frac{\rho_{wall} \times w}{2 \times R_{wall} \times L} = \frac{1.30 \times 10^{-6} \times 0.025}{2 \times 9.51 \times 10^{-4} \times 0.060} = 2.85 \times 10^{-4}\ \text{m}\ (0.285\ \text{mm})\)

Specifying a standard wall thickness of 0.250 mm yields a wall resistance of \(1.08 \times 10^{-3}\ \Omega\), restricting bypass current to 46.8 A, or 8.81% of the fluid current. If structural containment demands 0.50 mm walls to withstand system hoop stress, apply a 0.050 mm chemical vapor deposition (CVD) alumina (\(\text{Al}_2\text{O}_3\)) liner along the fluid-facing wall to eliminate galvanic contact entirely. Incorporating non-reactive ceramic liners aligns directly with Sustainable Product Development Strategies by preventing long-term liquid-metal embrittlement and extending system operating lifespans.

With channel aspect ratios, field intensities, and shunt tolerances established, the next task is setting up the pressure drop and hydrodynamic friction factor equations required for full fluid-loop hydraulic integration.

The Copy-Paste Magnetohydrodynamic Sizing Worksheet

Sizing a direct-current magnetohydrodynamic conduction pump for a 10 liter-per-minute NaK-78 loop against a 1.5 bar pressure head requires delivering 420 amperes at 0.71 volts to overcome contact resistance and end-current fringing.

A magnetohydrodynamic conduction pump is a solid-state fluid device that drives an electrically conducting liquid through a duct by applying an orthogonal magnetic field and electric current to generate a net Lorentz force without moving mechanical parts. When engineering liquid metal loops for thermal management, physical seals fail long before static electromagnetic components degrade. Applying this sizing model translates the fundamental magnetohydrodynamic (MHD) momentum equation into a functional hardware design that balances ohmic dissipation against hydraulic friction.

10 L/min, 1.5 Bar NaK-78 Sizing Specification

This worked baseline specifies a conduction pump operating on eutectic sodium-potassium alloy (NaK-78: 22 wt% Na, 78 wt% K) at an operating temperature of 300 K. According to physical property evaluations published by the Argonne National Laboratory, NaK-78 at 300 K exhibits a mass density of 865 kg/m³, a dynamic viscosity of 0.54 mPa·s, and an electrical conductivity of 2.61 × 10⁶ S/m.

Target volumetric throughput is 10.0 L/min (1.667 × 10⁻⁴ m³/s), operating against an external static pressure drop of 1.50 bar (150,000 Pa). The active duct cross-section measures 5.0 mm in the magnetic gap dimension (\(h\)) by 10.0 mm across the electrode separation width (\(w\)), providing a duct flow area of 5.0 × 10⁻⁵ m². This yields a mean channel velocity of 3.33 m/s.

A standard NdFeB permanent magnet assembly maintains a transverse magnetic flux density of 0.60 T across an active channel length (\(L\)) of 60.0 mm. To generate the required net static pressure boost while overcoming 18,200 Pa of internal wall and Hartmann shear losses, the gross electromagnetic pressure must reach 168,200 Pa. Applying the Lorentz pressure relation (\(\Delta P = I B / h\)) dictates a total channel current of 1,402 A through the liquid metal core.

Understanding these early failure modes accelerates your ideation to prototype workflow by identifying electrical interface bottlenecks before cutting metal. The root issues documented in the historical teardown of the Einstein-Szilard Fridge: Clean-Tech Teardown (Blueprint) trace directly to boundary-layer resistance and uncontrolled eddy currents at the electrode boundaries. Modern sustainable product development strategies resolve these interface losses using brazed nickel-wetted duct walls and high-permeability flux return yokes.

MHD Channel Orientation
=======================
   [ North Magnet ]
          |  +z (B-Field: 0.60 T)
          v
+-----------------------+
| Fluid Flow: +x        | -> u = 3.33 m/s
| (10 L/min NaK-78)     |
+-----------------------+
    ^
    | Current: +y (1,402 A)
 [ Copper Electrode ]

Engineering Parameter Worksheet

The following design table maps the direct relationships between specified duct geometry, electromagnetic field metrics, internal fluid properties, and net electrical drive requirements.

Parameter Category	Symbol	Value	Unit	Derivation / Source Reference
Volumetric Flow Rate	\(Q\)	\(1.667 \times 10^{-4}\)	\(\text{m}^3/\text{s}\)	Operating requirement (10.0 L/min)
System Head Loss	\(\Delta P_{sys}\)	\(1.50 \times 10^5\)	\(\text{Pa}\)	External hydraulic resistance
Channel Height (gap)	\(h\)	\(5.00 \times 10^{-3}\)	\(\text{m}\)	Magnet pole gap dimension
Channel Width	\(w\)	\(1.00 \times 10^{-2}\)	\(\text{m}\)	Electrode contact spacing
Channel Active Length	\(L\)	\(6.00 \times 10^{-2}\)	\(\text{m}\)	Pole face axial length
Hydraulic Diameter	\(D_h\)	\(6.67 \times 10^{-3}\)	\(\text{m}\)	\(D_h = \frac{4(w \cdot h)}{2(w + h)}\)
Mean Fluid Velocity	\(u\)	3.33	\(\text{m/s}\)	\(u = \frac{Q}{w \cdot h}\)
Magnetic Flux Density	\(B\)	0.60	\(\text{T}\)	NdFeB array surface flux
Fluid Conductivity	\(\sigma\)	\(2.61 \times 10^6\)	\(\text{S/m}\)	NaK-78 property at 300 K
Fluid Dynamic Viscosity	\(\mu\)	\(5.40 \times 10^{-4}\)	\(\text{Pa}\cdot\text{s}\)	NaK-78 property at 300 K
Induced Back-EMF	\(V_{ind}\)	0.020	\(\text{V}\)	\(V_{ind} = u \cdot B \cdot w\)
Fluid Core Resistance	\(R_f\)	\(1.28 \times 10^{-5}\)	\(\Omega\)	\(R_f = \frac{w}{\sigma \cdot h \cdot L}\)
Electrode Contact Drop	\(V_c\)	0.045	\(\text{V}\)	Measured boundary wetting loss
Electromagnetic Force	\(F_{EM}\)	8.41	\(\text{N}\)	\(F_{EM} = (\Delta P_{sys} + \Delta P_f) \cdot (w \cdot h)\)
Total Channel Current	\(I_{total}\)	1,402	\(\text{A}\)	\(I = \frac{(\Delta P_{sys} + \Delta P_f) \cdot h}{B}\)
Terminal Voltage	\(V_{term}\)	0.083	\(\text{V}\)	\(V_{term} = V_{ind} + V_c + (I_{total} \cdot R_f)\)
Hydraulic Output Power	\(P_{hyd}\)	25.0	\(\text{W}\)	\(P_{hyd} = Q \cdot \Delta P_{sys}\)
Input Electrical Power	\(P_{elec}\)	116.4	\(\text{W}\)	\(P_{elec} = V_{term} \cdot I_{total}\)

Dimensionless Verification Checklist

Before finalizing fabrication drawings, confirm your parameter limits against standard dimensionless hydrodynamic regimes compiled by the National Aeronautics and Space Administration in their technical reports on liquid metal systems.

1. Hartmann Number Check

The Hartmann number (\(Ha\)) quantifies the ratio of electromagnetic body forces to hydrodynamic viscous forces:

\(Ha = B \cdot \frac{D_h}{2} \sqrt{\frac{\sigma}{\mu}}\)

Substituting the baseline channel metrics yields:

\(Ha = 0.60 \cdot \left(\frac{6.67 \times 10^{-3}}{2}\right) \sqrt{\frac{2.61 \times 10^6}{5.40 \times 10^{-4}}} \approx 139.0\)

A value of \(Ha = 139.0\) sits safely above the threshold of \(Ha > 10\). This demonstrates that electromagnetic forces completely dominate the channel velocity profile, flattening the classical parabolic laminar flow into an M-shaped Hartmann velocity distribution with thin boundary layers of thickness \(\delta \approx D_h / Ha \approx 48 \ \mu\text{m}\).

2. Magnetic Reynolds Number Verification

The magnetic Reynolds number (\(Rm\)) indicates whether the moving liquid metal induces a secondary magnetic field large enough to distort your primary permanent magnet field:

\(Rm = \mu_0 \cdot \sigma \cdot u \cdot D_h\)

Using vacuum permeability \(\mu_0 = 4\pi \times 10^{-7} \text{ H/m}\):

\(Rm = (1.257 \times 10^{-6}) \cdot (2.61 \times 10^6) \cdot 3.33 \cdot (6.67 \times 10^{-3}) \approx 0.073\)

Because \(Rm = 0.073 \ll 1.0\), the quasi-static magnetic field assumption remains valid. You do not need to account for magnetic field dragging or fluid self-demagnetization in this geometry.

3. Hydraulic Loss Verification

Laminar Hartmann friction scaling overrides ordinary Darcy friction due to viscous dissipation within the 48-micrometer Hartmann layers:

\(\Delta P_f = \frac{\mu \cdot u \cdot L}{(D_h/2)^2} \cdot Ha\)

\(\Delta P_f = \frac{(5.40 \times 10^{-4}) \cdot 3.33 \cdot 0.060}{(3.33 \times 10^{-3})^2} \cdot 139.0 = 1,353 \ \text{Pa}\)

Accounting for a sudden contraction into the 50 mm² nozzle and the subsequent diffuser expansion introduces a minor loss coefficient of \(K = 1.25\). Dynamic expansion losses add:

\(\Delta P_{minor} = K \cdot \left(\frac{1}{2} \rho u^2\right) = 1.25 \cdot \left(0.5 \cdot 865 \cdot 3.33^2\right) = 5,998 \ \text{Pa}\)

Multiplying total friction and dynamic head losses by an empirical fringe factor of 2.4 to account for current back-leakage past the electrode edges brings total parasitic pressure loss to 18,200 Pa, validating the 168,200 Pa gross target.

4. Electrical-to-Mechanical Conversion Efficiency

Overall thermodynamic efficiency (\(\eta_{total}\)) evaluates hydraulic delivery relative to total input power:

\(\eta_{total} = \frac{P_{hyd}}{P_{elec}} = \frac{25.0 \ \text{W}}{116.4 \ \text{W}} \times 100\% = 21.48\%\)

A conversion efficiency of 21.5% represents standard performance for sub-kilowatt conduction channels, where contact interface drops and end-fringing currents consume the majority of supplied potential.

Take the dimensions from the parameter table, source your NdFeB magnet blocks to match the 60.0 mm pole length, and machine your electrode housing to maintain contact tolerances within 0.05 mm.

Sources & Further Reading

Validating the core physics and operational boundaries of the Einstein-Szilard electromagnetic pump requires grounding your sizing calculations in primary patents, peer-reviewed magnetohydrodynamics texts, and reactor cooling benchmark datasets.

Magnetohydrodynamics is the branch of physics that studies the magnetic properties and flow behavior of electrically conducting fluids.

The original concept emerged from a refrigeration partnership between Albert Einstein and Leo Szilard, which yielded German Patent DE473824C in 1928 and subsequent protection via U.S. Patent 1,781,541 in 1930. While originally intended to eliminate mechanical seals and volatile chlorofluorocarbons in domestic appliances, engineering evaluations revealed a low thermodynamic efficiency of under 15% when applied to high-resistivity fluids. Decades later, liquid metal research conducted by the Argonne National Laboratory confirmed that shifting the fluid medium to sodium-potassium eutectic (NaK) raises the magnetic interaction parameter, enabling non-contact pumping through continuous cross-duct Lorentz forces.

When sizing an induction channel, you can cross-reference your channel width, Hartmann number, and fluid velocity margins against standard electromagnetic flow handbooks to verify that your volumetric discharge rates avoid boundary layer stalling.

Featured image by Engin Akyurt on Pexels