Semi-Solid State Battery Integration for Aerospace: An Engineer’s Guide to Airworthy Pack Integration
I have sat in enough preliminary design reviews to recognise the exact moment an aerospace battery program starts to go wrong. It is not when someone questions the cell chemistry. It is when the systems engineer asks, “what is the preload on the stack at the end of the qualification vibration run, and who owns the evidence?” and the room goes quiet. At that point you are no longer selling a battery. You are integrating one into a certified aircraft, and the rules change completely.
This guide is about that transition. Over the last few years my team and I have taken semi-solid cells from the benchtop into airframe integration work for UAV, eVTOL and high-altitude platforms, and the pattern has been consistent: the chemistry gives you the energy, but the integration decides whether you get an airworthy article. A semi-solid state battery with a genuinely impressive cell-level number will lose to a well-integrated conventional pack every single time if the mechanical preload, the fault-isolation architecture or the thermal path has not been thought through. Everything below comes from that integration work, including the parts that hurt.
Why Integration, Not Cell Chemistry, Decides Whether an Aerospace Program Flies
Let me define terms first, because “semi-solid” is doing a lot of marketing work right now. In my usage, a semi-solid state battery sits between a conventional liquid-electrolyte lithium ion battery and an all-solid-state cell. It retains a gel or highly viscous electrolyte phase, with free liquid electrolyte reduced from the 15–20 wt% typical of a conventional cell down to roughly 5–10 wt%. That reduction is why the safety numbers move in the right direction: less free electrolyte means less fuel for a vent event, less gas generation, and measurably slower propagation. It is also why the rate capability moves in the wrong direction, because the ionic conductivity of the electrolyte phase drops from roughly 8–12 mS/cm into the 3–7 mS/cm range. A fully solid-state battery goes further still and removes the compliant phase almost entirely, which is why its stack pressure requirement is higher and its mechanical integration is more demanding, not less.
At cell level, the promise is real. Cells in the 300–360 Wh/kg range are being demonstrated against 250–300 Wh/kg for good conventional high-nickel cells, and at pack level that translates into roughly 200–260 Wh/kg against 160–200 Wh/kg. For an eVTOL where the battery is commonly 30–40% of maximum take-off mass, that difference is the difference between a viable mission and a brochure.
It is worth contrasting this with the world most of my readers already know. On a sub-25 kg survey UAV, a drone battery is a removable, plug-in module on an uncertificated airframe. The operator swaps it on condition, the risk is bounded by the aircraft and the pilot, and a wrong design assumption shows up on the second flight rather than in a certification review. As a drone battery manufacturer we ship a lot of custom drone battery assemblies into that world, and almost none of that workflow transfers. We are also flying solid state drone battery evaluation samples into small UAS programs, where the duty cycle is gentler and the airworthiness burden is low, precisely so that we can learn the pressure and interface behaviour before committing the chemistry to a certified installation. Certified aerospace is a different discipline.
But here is the integration problem that nobody puts on the first slide. Most semi-solid and solid-state cell designs need external stack pressure to keep the electrode–electrolyte interface intact, typically in the 0.2–2 MPa range depending on whether the anode is silicon-bearing or lithium-metal. Maintaining that pressure over a ten-year service life, through thermal cycling, vibration and creep, requires structure. Structure has mass. On the programs I have worked on, the preload and containment structure has added between 5% and 12% to the pack mass. If your chemistry advantage is 15–20% at pack level, you have just spent most of it. That is not a reason to abandon semi-solid; it is a reason to do the mechanical integration properly and co-design the structure with the airframe rather than bolting on a box.
Start With the Airworthiness Baseline, Not the Cell Datasheet
The single biggest schedule risk I see is teams starting from the cell datasheet and working outward. In a certified program, you start from the airworthiness baseline and work inward. The document set that governs a rechargeable lithium installation on an aircraft looks roughly like this:
- RTCA DO-311A — Minimum Operational Performance Standards for Rechargeable lithium battery and Battery Systems. This is the core airborne battery standard. Its central requirement is that a single-cell thermal event must be contained: no propagation to adjacent cells, no explosion, no breach of the enclosure, and no hazard to the aircraft or its occupants.
- RTCA DO-160G — environmental conditions and test procedures for airborne equipment. Section 4 (temperature and altitude), Section 5 (temperature variation), Section 6 (humidity), Section 7 (operational shock and crash hazard), Section 8 (vibration), Section 16 (power input), Section 17 (voltage spike), Section 18 (audio frequency conducted susceptibility), Section 20 (RF susceptibility), Section 21 (emission of RF energy), Section 22 (lightning induced transient susceptibility), Section 25 (electrostatic discharge).
- DO-178C / DO-254 / ARP4754A / ARP4761 — software, airborne electronic hardware, system development and safety assessment. If the battery management function is the sole energy source for a flight-critical load, expect a DAL B or DAL A assignment, not DAL D.
- MIL-STD-704F (aircraft electric power characteristics), MIL-STD-461G (EMC), MIL-STD-810H (environmental) and SAE AS50881 (aerospace vehicle wiring) for military and many UAV programs.
- FAA AC 20-184 for lithium battery installations, the applicable FAR/CS certification basis (Part 23, 25, 27, 29 or the powered-lift route under 21.17), EASA Special Condition VTOL for eVTOL, and ISO 21384 for UAS.
- Cell-level transport and safety baselines still apply underneath all of it: UN 38.3 (T1–T8), IATA DGR with the 30% state-of-charge limit for UN3480 air freight, IEC 62619 and IEC 62133-2.
Note what is absent from that list: nothing there cares about your 360 Wh/kg cell. Every one of those documents is about the installed article. If you write your integration plan as a list of DO-160G sections with a test category and an owner against each, you are already ahead of most programs.
Electrical Architecture: Bus Voltage, Fault Isolation, and the Return Path
The first real decision is bus architecture. A small UAV platform will typically sit on a 28 V DC bus, where MIL-STD-704F gives you a steady-state window of roughly 22–29 V with emergency operation down to about 18 V, and an allowable ripple amplitude in the region of 1.5 V peak-to-peak. Larger aircraft and most eVTOL designs use a high-voltage DC bus, commonly 270 V DC for military platforms (250–280 V window under MIL-STD-704F) or 400–800 V DC for electric propulsion.
Above roughly 400 V DC, you enter a regime where an arc does not self-extinguish. That single fact drives most of the protection design. Practical consequences:
- Interruption. A conventional HV DC contactor will not reliably break a few thousand amps. Pyrotechnic disconnects that open in 2–5 ms are the practical answer above about 1000 A prospective fault current, and they need a control path that is independent of the software that might be the reason you are disconnecting.
- Precharge. Size the precharge resistor for 3–5 RC time constants, which for typical aerospace DC-link capacitances lands in the 2–5 second range, and verify the resistor can absorb the full inrush energy without drifting out of tolerance over the maintenance interval.
- Insulation monitoring. I specify the insulation monitoring device to warn at 500 Ω/V and to open the contactors at 100 Ω/V. On a 540 V nominal bus that is roughly a 270 kΩ warning threshold and a 54 kΩ trip. The numbers are the same order of magnitude used in ISO 6469-3 for road vehicles, and they transfer cleanly because the failure physics is identical.
- Clearance and creepage at altitude. This one catches people. A 1 cm air gap that holds roughly 30 kV at sea level drops to about 5 kV at 55,000 ft, where ambient pressure is near 80 Torr. Paschen’s law puts the breakdown minimum near 0.5–1 Torr·cm, which is around 700 V for the same gap at 100,000 ft. Derive your creepage and clearance for the low-pressure case and then add margin; do not reuse a sea-level rule of thumb.
- Cable sizing. Free-air convection falls with density, so a harness that is comfortably rated at sea level needs derating of roughly 20–30% at 15,000 ft and considerably more above that. Bundle derating per SAE AS50881 on top. I still use the 3% main-circuit / 1% sense-line voltage drop targets as a starting point, then let the thermal analysis move them.
Mechanical Integration: Stack Pressure, Preload, and Crash Loads
This is where semi-solid differs from conventional lithium most sharply at the integration level. The interface between a semi-solid electrolyte and the electrode is not self-healing. If the stack pressure falls out of its window, contact resistance rises, the local current density redistributes, and you get accelerated and uneven ageing that no amount of clever battery management algorithm will recover. The cell datasheet will give you a target pressure; your job is to guarantee it for the whole service life.
Three mechanisms work against you. Creep in the cell stack and in any polymer components relaxes the preload over time. Differential thermal expansion between the cell stack, the compression hardware and the airframe structure modulates the pressure across the operating temperature range. Vibration and shock seating settles joints and fasteners. The standard countermeasure I use is a disc-spring (Belleville) stack in the compression path, sized so that the working range sits on the flat part of the spring curve, which holds pressure within a narrow band across several millimetres of relaxation. Then verify it: measure preload before and after the DO-160G Section 8 endurance run, and set an acceptance criterion. I use 20% preload loss as an investigation trigger and 30% as a failure.
Above that sits the airframe load case. Battery attachments have to carry the emergency-landing envelope: on the order of 20 g forward and 15–20 g vertical at the attachment fittings for small aeroplanes and rotorcraft under 14 CFR 23.562 and 27.561, and 9 g forward for transport-category aeroplanes under 25.561. The battery must not detach, and it must not become a projectile. In practice this means the mounting hardware ends up heavier than anyone expected, which loops straight back to the mass penalty I described earlier.
One more mechanical rule that has saved me twice: keep the first natural frequency of the installed pack-and-bracket assembly clear of the airframe excitation orders. As a working rule I target a first mode above 1.5× the highest significant excitation frequency from the rotors or propellers, and I verify it with a tap test on the actual installation hardware, not on the pack alone.
Thermal Integration at Altitude: Where Air Cooling Stops Working
Air density is roughly 57% of sea level at 15,000 ft, about 24% at 35,000 ft and near 10% at 55,000 ft. Convective heat transfer scales with it. Any cooling concept that depends on moving air over a finned surface loses most of its capacity by the time you are in the flight levels, which is why forced-air cooling is largely confined to low-altitude UAV and to ground operations.
For sustained high-power aerospace duty, assume liquid. If you are moving 5 kW of heat with a 5 K coolant rise on a 50/50 glycol-water loop with a specific heat near 3.3 kJ/kg·K, you need roughly 0.30 kg/s, which is about 18 L/min. Cold-plate pressure drop typically lands between 20 and 60 kPa. If the coolant path runs anywhere near HV conductors, use a dielectric fluid and accept the roughly 30–40% penalty in heat capacity, because a conductive coolant leak into an 800 V bus is a far worse problem than a warm cell.
Cold soak is the other half of the problem. DO-160G operating categories run down to -55 °C for equipment in unpressurised areas, and a pack that has been soaked overnight at -40 °C cannot accept charge. My standard charge-enable logic is: inhibit below 0 °C, restore at +5 °C with hysteresis, and derate to 0.2–0.5C between 0 and 15 °C. For platforms that have to launch cold, budget heater power. A 100–500 W heater on a small UAV pack gives you a warm-up rate in the 0.5–2 °C/min range, which is 20–60 minutes of pre-flight. That is an operational constraint, and it belongs in the flight manual, not in a footnote.
Containment is the third element and it is non-negotiable under DO-311A. In practice: an intumescent or ceramic fibre barrier of 2–5 mm between modules, a defined vent path routed overboard rather than into the pressurized volume, and vent gas management so that you never accumulate a flammable concentration anywhere in the installation. Every semi-solid design I have worked on has made this easier, because the reduced free electrolyte cuts both peak heat release rate and total gas volume materially compared with an equivalent conventional cell, but “easier” still means you have to test it.
BMS and Avionics Integration: Talking to the Avionics Without Lying to It
The interface between the battery management system and the avionics is where most integration rework happens. Three themes recur.
Determinism and bus choice
Pick the data bus for the certification path, not for convenience. ARINC 429 is a 12.5 or 100 kbit/s single-source, multi-drop bus with a fixed 32-bit word, and it is the lowest-risk choice for a federated architecture. CANaerospace and plain CAN are fine for UAV and for integrated modular architectures where the battery sits behind a remote data concentrator. ARINC 664 (AFDX) is the right answer for high-data-rate integrated architectures, but it drags the whole network certification story into your battery program. Whichever you choose, define the message set and the failure-flag semantics early, because “what does the flight deck show when a cell temperature sensor is open-circuit” is a certification question and not a software detail.
Independence of the protection path
Whatever design assurance level your BMS software achieves, keep a hardware protection path that does not depend on it. I use a comparator ladder on cell voltage and a discrete over-temperature chain that drives the contactor coil directly. The rationale is straightforward: a DO-178C DAL C software item has an acceptable failure rate of 10-5 per flight hour, and for a function that prevents a fire that is not something I want to be the only barrier.
State estimation accuracy as a dispatch requirement
In aerospace, state of charge is not a dashboard nicety; it is a dispatch criterion. If you are computing a reserve fuel equivalent, a 5–10% SOC error at low temperature and high rate is unacceptable. Extended Kalman filtering with OCV correction and a temperature-dependent resistance map gets you into the 2–3% band, but only if you characterise the cell across the whole operating envelope and keep the characterisation current when the cell lot changes. My standard dispatch gates are: state of health at or above 80% of type-design capacity, direct current resistance at or below 1.3× beginning-of-life, and a static cell-to-cell voltage spread below 50 mV (below 30 mV on a healthy new pack).
EMI/EMC, Bonding and Lightning: The Three That Fail Reviews
Batteries are quiet until they are not. A high-current switched DC bus with a PWM-driven heater or DC-DC converter is a broadband noise source, and it will find your navigation receivers if you let it. The DO-160G sections that matter most here are Section 21 for emissions, Section 20 for RF susceptibility, Section 22 for induced lightning transients, and Sections 17 and 18 for spikes and audio-frequency conducted susceptibility. On military programs the equivalents are MIL-STD-461G CE102, CS101, CS114, CS115, CS116, RE102 and RS103.
Three practical rules have kept me out of trouble. Terminate cable shields with 360° backshells bonded to the connector, never with a pigtail; a pigtail is an inductor at the frequencies that matter. Keep the bonding resistance across every joint in the lightning return path at or below 2.5 mΩ. And separate HV and signal harnesses by at least 150 mm, or cross them orthogonally with both shielded, because parallel runs are how noise gets from a propulsion inverter into a compass.
Qualification: Build the Evidence Pyramid
Qualification evidence in aerospace is a pyramid, and the temptation is always to skip a level. Do not.
- Cell level: UN 38.3 T1–T8, IEC 62619 or IEC 62133-2 as applicable, plus the characterisation matrix (capacity, DCR and OCV across temperature, rate and SOC) that your BMS model depends on.
- Module level: propagation testing with a deliberately initiated single-cell event, in the mechanical configuration you intend to fly.
- Pack level: DO-311A thermal runaway containment, plus DO-160G environmental categories with margin. I qualify at 10–15% beyond the installed environment, because the installed environment will be measured too late to change the design.
- Installation level: this is the level that gets skipped and it is the one that finds the real problems. Test the pack mounted in its actual airframe hardware, with the actual harness, the actual cooling loop and the actual flight software. A pack that passes on a shaker table regularly fails when it is bolted into a bracket that has its own resonance.
Production then needs its own discipline: first article inspection per AS9102, environmental stress screening on a defined sample, and 100% end-of-line testing with the data archived against the serial number.
Documentation, Traceability and Continued Airworthiness
A battery is not done when it ships. It is done when there is a defensible maintenance and retirement story. Practically, that means: an Instructions for Continued Airworthiness document, an Airworthiness Limitations Section if the battery has a life limit, and inspection intervals in flight hours and calendar time. Typical values I have signed off: a visual and torque-check inspection every 100–500 flight hours, a capacity verification every 1,000–2,000 flight hours, and an annual insulation resistance and vent-path check.
Traceability is what makes any of that possible. Every pack needs a record of cell lot, weld parameters (with pull-test results — I hold a minimum of 20–30 N on a nickel tab weld and review the distribution, not just the mean), BMS firmware version, torque values and end-of-line test data. That is also what makes the incoming digital battery passport regime workable: EU Regulation 2023/1542 starts requiring a digital passport for batteries from February 2027. Even if your aircraft battery sits outside its direct scope, your ground support fleet and your European customers will not, so build the data model now rather than retrofitting it.
Finally, write down the retirement criteria and put them in the maintenance manual. Mine are unchanged across programs: retire at 80% of rated capacity, at 1.3× beginning-of-life DC resistance, at a static cell spread above 50 mV that does not resolve after a controlled balance cycle, or on physical damage and any vent event. A pack that has vented is a removed pack, regardless of what the capacity test says.
Where This Leaves the Semi-Solid Decision
I am bullish on semi-solid for aerospace, with one condition: you have to budget honestly for the integration. The energy density is real, the safety margin gain from cutting free electrolyte from 15–20 wt% down to 5–10 wt% is real, and for a platform where every kilogram of battery is a kilogram of payload you will not fly without, that is decisive. But the preload structure, the containment hardware, the liquid cooling loop and the certification evidence all have mass and cost, and they will consume a meaningful share of the chemistry advantage. The programs that succeed are the ones where the lithium battery manufacturer and the airframe integrator do the mechanical and thermal design together, from the first sketch, rather than throwing a box over a wall. On our side that means a custom battery solution where the battery pack design, the BMS solution and the compression structure are iterated as one assembly against the aircraft load case, not procured as three separate line items and reconciled at the critical design review.
If you are at the start of that process, start with the airworthiness baseline and the stack pressure question. Everything else is downstream of those two conversations.
Frequently Asked Questions
Does a semi-solid state battery still need UN 38.3 and IATA DGR compliance for air transport?
Yes. There is no exemption for semi-solid or solid-state lithium cells at present. A semi-solid cell with a lithium-bearing anode and cathode is still a lithium-ion cell for transport purposes: it ships as UN3480 (or UN3481 when packed with or contained in equipment), it needs the full UN 38.3 T1–T8 test series with a valid test summary, and standalone cells and batteries move by air at no more than 30% state of charge under the IATA Dangerous Goods Regulations. Plan your incoming logistics around that, because a 30% SoC restriction on cell procurement changes your production sequencing.
How much stack pressure does a semi-solid cell actually need, and how is it maintained in service?
Typical targets run from 0.2 MPa for silicon-blended anodes up to 1–2 MPa for lithium-metal designs, and the number is specific to the cell, so take it from the cell supplier’s integration manual rather than from a competitor’s datasheet. Maintenance of that pressure is a mechanical design problem: use a Belleville disc-spring stack in the compression path, size it so the working range sits on the flat portion of the spring curve, and verify preload retention at the end of DO-160G Section 8 vibration endurance. I treat a 20% preload loss as an investigation trigger and 30% as a failure.
What design assurance level does an aerospace BMS normally get?
It depends entirely on the consequence of the battery failing. If the battery feeds a non-essential load with an independent backup, DAL D or C is achievable. If it is the sole energy source for a flight-critical function, which is the case for most eVTOL propulsion architectures, expect DAL B or DAL A, which brings DO-178C objectives for the software and DO-254 for any FPGA or programmable logic in the BMS. That is a large cost and schedule driver, and it is also why I always keep an independent hardware protection path that opens the contactors without any software involvement.
Can I use air cooling on an aerospace semi-solid pack?
Only at low altitude and low continuous power. Air density is about 57% of sea level value at 15,000 ft and near 10% at 55,000 ft, and convective heat transfer follows it. For sustained high-rate duty above roughly 25,000 ft, plan on a liquid cold plate. If the coolant path is anywhere near HV conductors, use a dielectric fluid and accept the 30–40% penalty in specific heat capacity compared with a 50/50 glycol-water mix.
Why does clearance and creepage need re-deriving at altitude?
Because breakdown voltage in air is a function of pressure times gap distance, not of gap distance alone. Paschen’s law puts the minimum breakdown near 0.5–1 Torr·cm, at roughly 327 V. In practical terms, a 1 cm gap that holds about 30 kV at sea level falls to roughly 5 kV at 55,000 ft, where ambient pressure is near 80 Torr, and to roughly 700 V at 100,000 ft. Any creepage rule of thumb you carried over from a ground vehicle product is wrong at altitude.
What is the single most important requirement in RTCA DO-311A?
Thermal runaway containment. DO-311A expects that a single-cell thermal event is contained within the battery: no propagation to adjacent cells, no explosion, no breach of the enclosure, and no hazard to the aircraft or occupants. The practical design response is inter-module barriers of 2–5 mm intumescent or ceramic fibre, a defined vent path routed overboard, and a verified, tested demonstration — not an analysis. Semi-solid chemistry helps here because the reduced free electrolyte cuts both peak heat release rate and total gas generation, but you still have to run the test in the configuration you intend to certify.
How is the airworthiness path different for an eVTOL compared with a conventional aircraft?
eVTOL has no pre-existing prescriptive battery rule in most jurisdictions, so the type certification basis is assembled from the existing parts plus special conditions: EASA’s Special Condition VTOL for Europe, and the powered-lift route with issue papers under FAA 21.17. In practice this means your DO-311A and DO-160G evidence is necessary but not sufficient, and you will be negotiating means of compliance for things like propulsion battery redundancy, emergency energy reserve and dispatch criteria. Budget calendar time for that negotiation; it is not a paperwork exercise.
What retirement criteria should go into the maintenance manual?
The four I use on every program: state of health below 80% of rated capacity, direct current resistance above 1.3× beginning-of-life value, a static cell-to-cell voltage spread above 50 mV that does not resolve after a controlled balance cycle, and any physical damage or vent event. Pair these with inspection intervals — typically a visual and torque check every 100–500 flight hours and a capacity verification every 1,000–2,000 flight hours — and make sure every pack has the serialised build and test record needed to evaluate them.
