Celgard 2325 vs 2400 vs 2500: Separator Comparison

Use Celgard 2500 for rate and power studies, 2325 when you need a shutdown layer or extra mechanical margin against lithium dendrites, and 2400 only when you are matching older published baselines. All three are 25 µm dry-process polyolefin films, so thickness will not decide it for you — porosity (55%, 41% and 39% respectively) and ionic resistance will. This Celgard separator comparison goes past the datasheet into the measured tortuosity and rate numbers that predict what your cell actually does, plus the handling errors that quietly invalidate the comparison before you even run EIS.

The 30-second answer: which Celgard separator for which job

Match the film to the failure mode you are trying to avoid, not to the highest number on the spec sheet.

If you are…	Use	Why
Running rate capability or power studies	2500	55% porosity and the lowest ionic resistance of the three
Cycling lithium metal or high-loading anodes	2325	Tighter pore structure; 380 gf minimum puncture vs 200 gf
Reproducing pre-2015 literature baselines	2400	The historical default in a large body of older work
Building a pouch headed for abuse testing	2325	The only one of the three with a PE shutdown layer
Screening electrolytes where wetting varies	2500	Roughly 2× electrolyte uptake, so less soak-time noise

 

Celgard 2325 vs 2400 vs 2500 datasheet specs, side by side

All three films are 25 µm thick and made by the same dry-stretch process, so the real differences sit in porosity, air permeability and layer count. The values below are from Celgard's own technical data sheets (October 2021 revisions; 2325 has a February 2025 revision with identical film properties).

Property	Celgard 2325	Celgard 2400	Celgard 2500
Structure	Trilayer PP/PE/PP	Monolayer PP	Monolayer PP
Thickness	25 µm	25 µm	25 µm
Porosity (calculated)	39%	41%	55%
Gurley (JIS)	645 s	620 s	200 s
TD shrinkage, 90 °C / 1 h	0%	0%	0%
MD shrinkage, 90 °C / 1 h	3.0% typ / 5.0% max	2.3% typ / 5.0% max	1.1% typ / 5.0% max
MD tensile (typ / min)	1900 / 1000 kg/cm²	1420 / 700 kg/cm²	1170 / 770 kg/cm²
TD tensile (typ / min)	150 / 80 kg/cm²	140 / 70 kg/cm²	130 / 80 kg/cm²
Puncture (typ / min)	450 / 380 gf	450 / 340 gf	325 / 200 gf
Shutdown layer	Yes, PE at ~130 °C	No	No

The Gurley number trips people up

Gurley is an air-permeability time, and the unit matters more than the number. Celgard publishes JIS Gurley: the seconds needed for 100 cc of air to cross one square inch of film at 4.9 inches of water.

ASTM D726 uses 10 cc at 12.2 inches of water. The same film can legitimately be quoted as 620 s (JIS) or roughly 24 s (ASTM). If a distributor quotes "Gurley 25" for a 25 µm PP film, they are on the ASTM scale — do not put that number in the same column as 645.

Why porosity alone gets this comparison wrong

Porosity tells you how much void is in the film; it does not tell you how twisted the path through that void is, and the twist is what your cell feels. Published tortuosity values for the same Celgard grade vary by more than a factor of two depending on measurement method and applied pressure.

Study and method	2325	2500
EIS, symmetric blocking cell (JES, 2016)	τ ≈ 4 (Bruggeman predicts 1.6)	—
EIS in coin cell (JES, 2019)	τ = 3.92	τ = 2.89
3D X-ray tomography (J. Energy Storage, 2024)	τ = 1.98	τ = 1.70
EIS under 0.05 MPa, 25 °C (JES, 2025)	τ = 4.3	MacMullin number = 3.3

 

Two things follow from that spread:

• Never lift a tortuosity value from a paper into your P2D or Newman model without checking the method and the stack pressure. The method moves the number more than the film choice does.
• The classic Bruggeman relation (τ = ε−0.5) badly underestimates polyolefin separators. For 2325 at ε = 0.39 it predicts τ = 1.6, roughly 40% of the experimentally measured value. Fitted exponents closer to 2.5–3.1 match the data far better.

The number that actually predicts performance: ionic resistance

Ionic resistance is the single measurement that separates these three films in a working cell, and the gap is large enough to change your conclusions.

Parikh and co-workers (Journal of The Electrochemical Society, 2019) built NMC532/Li half cells with a 27 mg/cm² cathode — about 3.9 mAh/cm² areal capacity — and swapped only the separator. They measured:

• Cell resistance of 7.51 Ω with 2500 against 12.58 Ω with 2325, a difference of 2.23 Ω/cm²
• Roughly double the electrolyte uptake for 2500, with a visibly lower contact angle
• 57% higher discharge capacity at 2C and 47% higher at 3C for the 2500 cells

 

Walkthrough 1: measure your own MacMullin number in an afternoon

Stop borrowing tortuosity values and measure your film in your electrolyte — the multi-layer EIS method takes about four hours and needs no specialised hardware.

What you need: two blocking electrodes (stainless steel spacers in a coin cell work), your electrolyte, a potentiostat that reaches 100 kHz, and six punched discs per film.

1. Punch six discs of each film with the same punch, from the same roll orientation.
2. Assemble a symmetric SS | separator | SS cell with one layer. Wet under vacuum for 10 minutes, then rest 2 hours.
3. Run EIS from 100 kHz to 1 Hz at 10 mV amplitude, at open-circuit voltage.
4. Read the high-frequency intercept on the real axis, where −Z″ crosses zero. That is R_HF.
5. Repeat with 2, 3, 4, 5 and 6 stacked layers.
6. Plot R_HF against layer count. The slope is resistance per layer; the intercept is contact and hardware resistance. Discard the intercept — it is your setup, not your film.
7. Convert: N_M = (slope × A × κ) / d, then τ = N_M × ε.

Worked example with real numbers

Using 16 mm discs (A = 2.01 cm²), 1 M LiPF₆ in EC:EMC 3:7 at κ ≈ 8.5 mS/cm [VERIFY against your own conductivity measurement — it shifts with salt batch and temperature], film thickness d = 25 µm = 0.0025 cm, and a measured slope of 1.5 Ω per layer:

N_M = (1.5 × 2.01 × 0.0085) ÷ 0.0025 = 10.3

τ = 10.3 × 0.39 = 4.0

That lands squarely in the 3.9–4.3 band reported for Celgard 2325, which is your sanity check that the setup is sound.

Four mistakes that wreck this measurement

• Using a single layer. Without the multi-layer slope you cannot subtract hardware resistance, and τ comes out 30–50% too high.
• Ignoring stack pressure. The MacMullin number moves measurably with compression. A crimped coin cell applies pressure you did not choose and cannot report.
• Running EIS before the film is wet. PP is hydrophobic and wetting is not instant. A 10-minute reading and a 2-hour reading differ, and 2325 drifts more than 2500.
• Borrowing κ from a paper. Electrolyte conductivity feeds directly into N_M. Measure it or your tortuosity inherits someone else's error.

Walkthrough 2: picking the film for three real builds

The right choice changes completely across three common lab projects, even though the films differ by only 16 percentage points of porosity.

Build A — NMC622/graphite coin cells, 2.5 mAh/cm², rate tested to 5C

Choose 2500. At 5C the 2.23 Ω/cm² penalty is directly visible in your capacity-versus-rate curve, and using 2325 will make a perfectly good electrode look rate-limited.

Trade-off: 2500's minimum puncture strength is 200 gf against 380 gf for 2325. If your cathode discs have burrs from a tired punch, you will see soft shorts in the 2500 arm and misread them as a separator problem.

Build B — Li‖Cu or Li‖LFP with lithium metal at 0.5–1 mA/cm²

Choose 2325. The tighter pore structure and higher puncture strength give more mechanical margin against dendrite penetration.

Be honest about what that buys you: no 25 µm polyolefin stops dendrites. If you short at 100 cycles with 2500 and 250 cycles with 2325, you have bought delay, not protection. For serious lithium metal work, treat all three as the baseline and evaluate coated or composite separators as the actual variable.

Watch the geometry: a 16 mm separator on a 15.6 mm lithium disc leaves 200 µm of overhang. Step up to 19 mm discs and the edge-short failures usually disappear.

Build C — 2 Ah pouch headed for hot-box and nail penetration

Choose 2325, because it is the only one of the three with a shutdown layer. The PE core melts near 130 °C and closes the pores; the PP skins hold dimensional integrity to roughly 165 °C.

That 35 °C window is narrower than it sounds. In a 5 °C/min oven ramp you will see shutdown work. In a nail penetration event, local heating outruns it and the film shrinks before it seals.

Two things the datasheets will not tell you

Celgard's current 2400 datasheet is a drug-delivery document

The 2400 technical data sheet published on celgard.com is titled "Celgard 2400 (Transdermal)" and lists its primary application as a rate-limiting membrane in drug delivery systems, with an FDA Drug Master File listing. The film properties are unchanged: 25 µm, 41% porosity, 620 s JIS Gurley.

Practically, 2400 is still manufactured and still stocked by lab distributors, but Celgard no longer markets it as a battery separator. By contrast, 2325 and 2500 both carry current battery-facing datasheets.

If you are specifying a film for a multi-year study or anything that might scale, 2400 carries the most supply risk of the three. [VERIFY current battery-grade availability and lead time with your distributor before designing it into a long programme.]

Shutdown does almost nothing in a 2032 coin cell

Buying 2325 "for safety" in a coin cell trades away rate performance for a feature that will not engage. Pore closure needs sustained stack pressure to hold the melted PE in place, and a crimped coin cell gives you neither controlled pressure nor a representative thermal mass.

If safety behaviour is the question, test it in the format you intend to ship.

Five handling mistakes that ruin a separator comparison

Most failed separator comparisons are handling failures, not material differences — these five account for the majority.

1. Mixing up 2400 and 2500. Both are 25 µm monolayer PP and visually identical. Label the bag the moment you cut. The only fast field check is Gurley — 620 s against 200 s is unmistakable.
2. Punching in different directions. These films are strongly anisotropic: 2325 is 1900 kg/cm² in MD against 150 kg/cm² in TD, a 12:1 ratio. Mark MD on first unroll and punch every disc the same way.
3. Drying above 60 °C. Shrinkage is specified at 90 °C. Vacuum-drying separators at 100–120 °C alongside electrodes will shrink and curl them. Dry separators at 50–60 °C, separately.
4. Reusing a dull punch. A worn punch leaves a fibrillated edge that becomes a soft-short path. If OCV drifts down slowly over 24 hours, inspect the punch before you blame the electrolyte.
5. Not standardising soak time. 2325 at 39% porosity wets more slowly than 2500 at 55%. Fix one protocol — 10 minutes vacuum plus 2 hours rest works — and apply it identically to every arm.

When this comparison does not apply at all

These three films are lab benchmarks, and there are several situations where none of them is the right starting point.

• Above roughly 4.4 V vs Li/Li⁺. PP oxidation becomes a genuine concern over long cycling. A ceramic-coated separator is the better baseline.
• Sustained operation above 60 °C. All three are polyolefin; creep and shrinkage dominate. Look at coated grades or aramid/polyimide films.
• Solid-state and semi-solid cells. There is usually no porous separator, or the film is only a temporary handling scaffold.
• Sodium-ion and aqueous systems. PP's hydrophobicity works against you. Glass fibre remains the practical lab default for Na-ion half cells.
• Anything heading to production. 25 µm is thick by current standards. Commercial cells use 9–16 µm coated films. Benchmark on 2325/2500, but do not design a product around them.

Sourcing and cost notes

All three are inexpensive relative to the cost of re-running a study with the wrong film, and all three are sold in lab-friendly quantities.

• Celgard sells hand samples — the 2325 hand sample is roughly 10 in × 116 in, about 0.75 m². Allowing for punching waste, that is on the order of 1,500 discs at 19 mm.
• Small rolls and cut sheets are widely available through battery materials distributors. [VERIFY current pricing, roll widths and minimum order quantities — these vary considerably by supplier and quantity.]
• Buy from a supplier who states roll orientation, or mark the machine direction yourself before you cut anything.
• Store flat, sealed, at room temperature and out of direct UV. Polyolefins degrade under UV exposure.

Frequently asked questions

Which Celgard separator has the highest porosity?

Celgard 2500, at 55% calculated porosity, against 41% for 2400 and 39% for 2325. That higher void fraction gives 2500 the lowest air resistance of the three — 200 seconds JIS Gurley against 620 and 645 seconds — and the lowest ionic resistance once the film is wetted with electrolyte.

Can I substitute Celgard 2400 for 2500?

Not without expecting different results. They share thickness, chemistry and appearance, but 2500's 55% porosity against 2400's 41% changes ionic resistance enough to shift measured rate performance. Use one grade consistently across every arm of an experiment, and record which one in your methods section.

Is Celgard 2325 better for lithium metal anodes?

It has more mechanical margin: 380 gf minimum puncture strength against 200 gf for 2500, plus a tighter pore structure. That delays dendrite penetration but does not prevent it. For serious lithium metal work, treat any 25 µm polyolefin film as a baseline rather than a solution.

What does the Gurley number mean on a Celgard datasheet?

It is the time in seconds for a fixed volume of air to pass through the film, so lower means more permeable. Celgard publishes JIS Gurley: 100 cc through one square inch at 4.9 inches of water. ASTM Gurley uses 10 cc at 12.2 inches, so the scales are not interchangeable.

Do I need to dry Celgard separators before cell assembly?

Yes, but gently. Polyolefin films carry surface moisture that reacts with LiPF₆ electrolyte to form HF. Dry under vacuum at 50–60 °C, not the 100–120 °C used for electrodes. Celgard specifies shrinkage at 90 °C, and hotter drying will shrink and curl the film.

Which Celgard separator do most published papers use?

Celgard 2325 and 2500 dominate recent literature, while 2400 appears more often in older work. If comparability matters, state the exact grade, the lot if you know it, and your soak protocol — separator grade alone explains a meaningful share of the spread in reported rate data.

Conclusion

The Celgard separator comparison comes down to three decisions. Pick 2500 when ionic resistance limits your result, 2325 when mechanical margin or shutdown behaviour matters, and 2400 only for literature continuity — while accepting that Celgard has quietly repositioned it away from batteries.

Everything else in that spec table is secondary, and the tortuosity numbers in the literature are too method-dependent to copy blindly.