What’s the right stack pressure?

A new Nature Energy paper puts a number on it

I’ve written before about cell electrode pressure and how it’s applied and lost across a module’s life, and about the practical business of mapping pressure distribution across a pouch cell in operando. Both of those pieces stopped short of answering the question every pack designer actually wants answered: what pressure should I be aiming for, and what breaks if I get it wrong in either direction?

A new paper paper, led by Cambridge in collaboration with WMG, Oxford and Sydney (Wang et al., Nature Energy, 2026) has a decent go at answering exactly that, for a Gr‖NMC811 pouch cell, and I think it’s worth a proper look because a couple of the numbers land uncomfortably close to home.

Heng Wang, Rui Wang, Christopher A. O’Keefe, Erik Björklund, Daniela Proprentner, Joe C. Stallard, Hwee Jien Tan, Wesley M. Dose, Louis F. J. Piper, Robert S. Weatherup, Angkur J. D. Shaikeea, Clare P. Grey & Michael De Volder, The interplay between stack pressure, mechanical expansion and degradation pathways in lithium-ion batteries. Nat. Energy (2026).

The Headline Result

The team built a custom dilatometer using compliant pneumatic bellows rather than the usual bolted plates or springs, and used it to hold five different stack pressures constant: 1.5, 3, 6.5, 12.5 and 37.5 bar. Using otherwise identical Gr | NMC811 pouch cells (402035 format, ~210 mAh, plain LP57 electrolyte, no additives) through hundreds of charge/discharge cycles.

The result is a clear U-shape. Cycle life to 80% capacity retention was:

• 3 bar (their “LP” condition): 169 cycles
• 12.5 bar (their “OP”, optimal): more than 375 cycles
• 37.5 bar (their “HP” condition): 161 cycles

So there’s a factor of ~2.2 in cycle life between the optimum and either extreme. This is the number that stopped me as that optimum sits at roughly four times the pressure they measured in standard coin cells, and well above the 0.05–0.3 MPa (0.5–3 bar) range that’s typical in production pouch/prismatic modules, the same range I quoted from Mussa et al. and Müller et al. on the Cell Electrode Pressure page. If this generalises, a lot of packs may be running at pressures that curtail cycle life without anyone knowing it.

Two Distinct Failure Modes

What I like about this paper is that they didn’t stop at “too low is bad, too high is bad”: they went and found out why, on each side, using dilatometry, XCT, NMR, XPS, sXRD and ToF-SIMS on post-mortem cells.

1. Too little pressure (3 bar): the cathode cracks
   • With fewer particles in contact, load concentrates on the ones that are, and that drives crack initiation, consistent with force-chain behaviour from granular mechanics. Once cracks start, more particle surface is exposed to electrolyte, and they measured this directly: higher Ni and Mn dissolution by MP-AES, more SEI growth, and, on the anode side, a good fraction of the lithiated graphite had become physically isolated and stuck in a partially lithiated state, confirmed by NMR and X-ray diffraction. That’s capacity lost to disconnected material, not just classic SEI thickening.
2. Too much pressure (37.5 bar): the anode plates
   • The cell compacts for roughly the first 150 cycles as porosity falls, and the point where it starts irreversibly swelling again lines up almost exactly with the onset of accelerated capacity fade. Falling porosity, electrolyte squeezed out of the electrode, and (their inference rather than a direct measurement) possible localised separator pore closure all push up local overpotential until lithium starts plating. They see this directly too: silvery deposits on disassembly, a lithium metal peak in the NMR spectrum, and a strong lithium signal at the top of the electrode by ToF-SIMS.

Two genuinely different results, and knowing which one you’re looking at in a post-mortem tells you which way to move the pressure.

Things to Flag

A few things I want to flag before anyone treats “12.5 bar” as gospel, because I think they matter for how this translates to a real pack:

• It’s a moderate-rate result over a moderate number of cycles
  • The main ageing test runs at C/3, not some gentle C/10 crawl, which is reasonable, but the cycle counts involved (up to ~375–400) are well short of the 1,000+ cycles most of us actually care about for automotive life. There’s a separate, much longer 1,100-cycle/96.2%-retention result in the discussion, but that’s a different cell entirely: lower voltage (4.2 V vs the 4.6 V nominal UCV used in the main study) with FEC/VC/LiBF4 additives, presented to show pressure and additive strategies are complementary, not as evidence the 12.5 bar optimum itself holds out to 1,000+ cycles. Whether the OP condition’s apparent stability past 200 cycles is a genuine plateau or just a slower version of the same eventual failure isn’t shown.
• They didn’t test the pressures most of us actually use
  • The lowest pressure tested was 1.5 bar. Production pouch and prismatic modules routinely start well below that: the 0.05–0.3 MPa figures on the Cell Electrode Pressure page are typical. So the claim that today’s typical pack pressures are suboptimal is an extrapolation of their trend downward, not a directly measured point. Worth being honest about that distinction.
• The 4.6 V upper cutoff is doing more work than the paper lets on
  • They cycle to a nominal 4.6 V, which they say corresponds to an effective cathode voltage of ~4.3 V at C/3 once overpotential is accounted for, and the stated reason is to guarantee every cell reliably crosses the H2-H3 (c-lattice collapse) transition each cycle. Fair enough as a design intent, but the correction from 4.6 V nominal to 4.3 V effective looks like a single fixed number applied across all five pressure conditions, when their own central finding is that pressure changes porosity, electrolyte access and internal resistance, exactly the things that set how much overpotential you get at a given current. If high pressure cells really do have reduced porosity and squeezed-out electrolyte, they should see more overpotential than low pressure cells, not the same 0.3 V, meaning the cathodes in different pressure conditions may not actually be seeing the same effective voltage at all. On top of that, 4.3 V effective is still a hard push on NMC811, well beyond the ~4.2 V full-cell upper cut-off voltage typical in automotive packs, and high voltage independently drives electrolyte oxidation, oxygen loss and transition metal dissolution: the same symptoms they attribute to low stack pressure. So some of what’s being read as a pressure effect could be a voltage effect that simply expresses itself differently depending on how much overpotential each condition happens to have. This is worth flagging alongside the pressure result itself, not just as a footnote.
• The separator-resistance question isn’t actually closed
  • My working concern going in was that high pressure shuts down the separator directly, pore collapse increasing ionic resistance. The paper leans on this as a contributing factor to the plating mechanism, but their own tomography found cathode and separator thickness “similar within the accuracy of the tool” across all pressure conditions, and the pore-closure argument itself is supported by citation to earlier literature rather than a dedicated measurement in this paper. Their rate-performance evidence for “transport limitation” at high pressure doesn’t cleanly separate separator pore closure from reduced anode porosity or electrolyte being squeezed out of the stack. If you want that question nailed down, you need a study measuring through-plane ionic resistance of the separator in isolation as a function of applied pressure, and this isn’t quite that.
• Pressure was actively regulated throughout, and that’s not incidental: it’s the whole point of the method
  • The bellows use pressure control valves to hold fluctuation below 0.8%, against ~20% for a bolted rig and ~6% for spring-loaded, in their own comparison. This isn’t just a formation-cycle nicety: it runs for the full 375+ cycles, correcting for both the cyclic SOC swelling and the slow SOH-driven compaction/expansion they track in Fig. 2e. That means their 12.5 bar optimum was demonstrated under near-ideal constant-pressure conditions. It says nothing about what happens if pressure is allowed to drift by some tolerance band around that point, which is the situation almost every real pack is actually in.

Why this matters for anyone specifying real hardware

This is where it gets uncomfortable, because I went back to the worked example on the Cell Electrode Pressure page for a 12-cell PHEV2 module (148 mm × 91 mm × 26.5 mm prismatic cells): initial end-plate force of ~3 kN works out to 2.23 bar, and end-of-life force growth to ~30 kN works out to 22.3 bar.

When assembling prismatic cells into a module there will be an initial pressure requirement and at end of life there will be a final pressure.

For a typical 12 cell module made using PHEV2 format prismatic cells (148mm x 91mm x 26.5mm) the initial force applied to the end plates is ~3kN.

148mm x 91mm = 13468mm² = 0.013468m²

Pressure = 3000N / 0.013468m² = 222750Nm^-2 = 2.23 bar

At end of life this force can increase to ~30kN, a pressure of 22.3bar.

Lay that against Wang et al.’s pressure map and the picture is not flattering. That module starts almost exactly on top of their LP condition (the cracking-prone regime) and by end of life it’s landed between their OP and HP conditions, well into territory where their mechanism shifts toward plating. In other words, a fixed-spacing, bolted module doesn’t sit at one bad pressure for its whole life: by this calculation it sweeps across nearly the entire range the paper tested, starting near one failure mode and finishing near the other.

That’s a direct consequence of running constant-volume rather than constant-pressure, the same distinction covered in the operando pressure-mapping piece: a fixed end-plate spacing means pressure is dictated by however much the stack swells, not chosen by the designer. Wang et al.’s bellows are the opposite case, actively regulated to stay flat regardless of swelling, and that’s exactly why their result doesn’t transfer straightforwardly to a bolted module.

Obviously the format and chemistry are different (prismatic vs pouch, presumably polycrystalline vs single-crystal NMC811), so the absolute numbers won’t map across directly. But the shape of the problem, a real module’s pressure trajectory crossing the entire span between “too low” and “too high” over its service life, is a genuinely useful way to think about why cycle life outcomes can vary so much between packs that look identical on paper.

Could you actually hold 12.5 bar with a passive gap filler?

Worth a paragraph on its own, because it’s the natural next question. 12.5 bar (≈1.25 MPa) is a serious number: on a typical module face that’s tens of kN of clamping force, well beyond what a simple foam pad and light frame will comfortably sustain without either bowing the end plates or driving the pad into full densification early in life.

A passive pad has to do two conflicting jobs: soak up initial manufacturing tolerance (wants a soft, shallow-slope material) and hold pressure roughly constant as the cell compacts and then irreversibly swells over hundreds of cycles (wants enough remaining travel to track that, without the stiffness rising sharply once the easy compliance is used up). Most gap-filler materials designed for the 0.1–0.3 MPa world (silicone or PORON foams) are already well up their compression curve at 1+ MPa, meaning there’s limited remaining travel to absorb the compaction-then-swelling trajectory the paper measures. Push toward solid elastomer pads or metallic wave-spring/Belleville stacks and you get better load-holding but lose the compliance that spreads pressure evenly across a real cell face, the exact problem the paper’s self-aligning bellows were built to avoid.

So my honest read: a passive pad sized to hit 12.5 bar at beginning of life is unlikely to hold that pressure within a tight band for the life of the pack, unless it’s specifically engineered with a long, controlled compression range spanning both the early compaction and the later swelling, which usually means trading pack volume for pressure stability. True constant pressure at this level really wants active or semi-active regulation, not a fixed pad.

Conclusion

I think this is the most rigorous long-term pressure-vs-degradation dataset I’ve seen for a conventional (non solid-state, non anode-free) pouch cell, and the mechanistic split (cracking below the optimum, plating above it) gives a genuinely useful diagnostic lens for anyone doing post-mortem failure analysis: heavy transition metal crossover and cathode cracks point one way, plating and low anode porosity point the other. Combined with the PHEV2 module numbers, it’s a strong argument for treating pressure as a designed, actively-managed parameter rather than whatever a bolted end-plate spacing happens to deliver.

That said, I don’t think 12.5 bar itself should be read as a universal target. Beyond the format, chemistry and cycle-count caveats above, the 4.6 V cutoff sits close enough to the pressure story’s own mechanisms (porosity, resistance, overpotential) that I’m not fully convinced the five conditions were tested under equivalent electrochemical stress. That doesn’t undermine the qualitative U-shape or the diagnostic value of the two failure modes, but it does mean I’d treat the exact optimum as chemistry-and-protocol-specific rather than something to lift straight into a pack specification.

This post has been built based on the support of and sponsorship from: illumion, Custom Power, Quarto Technical Services, Eatron Technologies and About:Energy

References

1. Heng Wang, Rui Wang, Christopher A. O’Keefe, Erik Björklund, Daniela Proprentner, Joe C. Stallard, Hwee Jien Tan, Wesley M. Dose, Louis F. J. Piper, Robert S. Weatherup, Angkur J. D. Shaikeea, Clare P. Grey & Michael De Volder, The interplay between stack pressure, mechanical expansion and degradation pathways in lithium-ion batteries. Nat. Energy (2026).
2. Mussa, A. S., Klett, M., Lindbergh, G. & Lindström, R. W. Effects of external pressure on the performance and ageing of single-layer lithium-ion pouch cells. J. Power Sources 385, 18–26 (2018).
3. Müller, V., Scurtu, R. G., Memm, M., Danzer, M. A. & Wohlfahrt-Mehrens, M. Study of the influence of mechanical pressure on the performance and aging of Lithium-ion battery cells. J. Power Sources 440, 227148 (2019).
4. Cannarella, J. & Arnold, C. B. Stress evolution and capacity fade in constrained lithium-ion pouch cells. J. Power Sources 245, 745–751 (2014).
5. Peabody, C. & Arnold, C. B. The role of mechanically induced separator creep in lithium-ion battery capacity fade. J. Power Sources 196, 8147–8153 (2011).
6. Cell Electrode Pressure, batterydesign.net
7. Mapping Pouch Cell Pressure Distribution in Operando, RHD, batterydesign.net
8. LFP Cell Clamping Pressure, batterydesign.net
9. Cell Electrode Pressure and Expansion, batterydesign.net

See also: Stack Pressure & Interfacial Resistance for Pack Designers, Compression Pads