Glass Fiber Separator Grades: GF/A to GF/F Compared

Glass fiber separator grades differ in two specs that actually matter: thickness (260–675 µm) and particle retention rating (0.7–2.7 µm). Porosity does not meaningfully change — every grade sits between roughly 90% and 92%. So choosing between GF/A and GF/D is a thickness decision, not a porosity decision, and it quietly changes the electrolyte volume in your coin cell by a factor of two or three. Below is the full comparison, the arithmetic behind the porosity claim, two procedures you can run this week, and the cases where glass fiber is the wrong separator entirely.

The complete glass fiber separator grades comparison

Here are the five binder-free glass microfiber grades used as battery separators, with retention, thickness, basis weight, and porosity calculated from those published specs.

 

Grade	Particle retention (µm)	Thickness (µm)	Basis weight (g/m²)	Porosity (calculated)	Pore volume, 16 mm disc
GF/A	1.6	260	53	≈91%	≈48 µL
GF/B	1.0	675	143	≈90%	≈123 µL
GF/C	1.2	260	53	≈91%	≈48 µL
GF/D	2.7	675	121	≈92%	≈125 µL
GF/F	0.7	420	75	≈92%	≈78 µL

 

Retention, thickness and basis weight are manufacturer specifications. [VERIFY against the current Cytiva Whatman Grade GF datasheet before publishing — figures are periodically restated.] The porosity and pore-volume columns are derived, not quoted; the method is in the next section.

Three things people routinely get wrong:

• There is no GF/E. The binder-free range is A, B, C, D and F. Papers citing "GF/E" are almost always typos.
• GF/A and GF/C are dimensionally identical. Same 260 µm thickness, same 53 g/m² basis weight. Only the retention rating differs, and in a cell they behave nearly the same.
• GF/D is the coarsest grade in the range at 2.7 µm — yet it is by far the most common lab separator. Popularity tracks thickness and mechanical robustness, not fineness.

What "particle retention" actually means

Retention is a filtration test result, not a pore size. A grade is rated by the smallest particle it captures at a defined efficiency in a standardised liquid challenge — nobody measured the pores to get that number.

The real pore-size distribution in a glass microfiber mat is broad, with plenty of channels far larger than the rating — so "2.7 µm pore size" for GF/D is a reviewer magnet. If you need a defensible number, run capillary flow porometry or mercury intrusion porosimetry on your own lot.

Why porosity barely changes across glass fiber separator grades

Because every grade is spun from the same borosilicate microfiber at a similar packing density. You can derive porosity straight from the datasheet:

Porosity = 1 − [ basis weight ÷ (thickness × fiber density) ]

Borosilicate glass density is approximately 2.23 g/cm³.

Worked example — GF/D:

• Basis weight 121 g/m² = 0.0121 g/cm²
• Solid volume = 0.0121 ÷ 2.23 = 0.00543 cm³ per cm²
• Geometric volume = 0.0675 cm × 1 cm² = 0.0675 cm³
• Porosity = 1 − (0.00543 ÷ 0.0675) = 0.92, i.e. 92%

Run the same arithmetic on GF/A and you get 91%. GF/B gives 90%. The spread across the entire family is about two percentage points — inside the measurement tolerance of most bench micrometers.

The practical consequence: if you are switching grades "for higher porosity," you are chasing a variable that does not move. Pick on thickness.

Walkthrough 1: what a grade change does to your electrolyte volume

A grade change moves electrolyte volume more than almost any other decision in a coin cell build. Quantify it before you assemble, not after the data looks strange.

1. Calculate punch area. A 16 mm disc is 2.01 cm² (A = πr², r = 0.8 cm).
2. Calculate geometric volume. Multiply by thickness in cm. GF/D: 2.01 × 0.0675 = 0.136 cm³.
3. Calculate pore volume. Multiply by porosity: 0.136 × 0.92 = 0.125 cm³ = 125 µL.
4. Repeat for the candidate grade. GF/A: 2.01 × 0.026 × 0.91 = 48 µL.
5. Compare against your dosing volume. Pipette 60 µL into a GF/D cell and the separator is 48% saturated. Keep 60 µL but move to GF/A and it is flooded, with roughly 12 µL squeezing out into the case and onto the gasket.

For scale: a 25 µm polyolefin separator at ~40% porosity holds about 2 µL in the same 16 mm footprint. Glass fiber holds 25 to 60 times more electrolyte. That is precisely why glass fiber cells cycle so cleanly — and why those cycle numbers do not transfer to a pouch cell.

Common mistake: benchmarking your Li–S or sodium-ion retention against a paper that used a different grade. If they ran GF/D and you ran GF/A, your electrolyte-to-active-material ratio differs by about 2.6× even with identical dosing. Always report grade, punch diameter and dosed volume together.

Trade-off to acknowledge: those figures are uncompressed. Crimping a CR2032 with a spacer and wave spring compresses glass fiber noticeably, so treat them as an upper bound and measure your own stack with a micrometer under load.

Walkthrough 2: measure the separator's real ionic resistance

The number worth measuring is the MacMullin number — how much the separator slows ion transport relative to free electrolyte. It is dimensionless and always ≥ 1.

1. Punch the separator, dry it (protocol below), and soak in your working electrolyte inside the glovebox.
2. Build a symmetric blocking cell: stainless steel disc | separator | stainless steel disc, in a CR2032.
3. Run EIS at open circuit, 1 MHz to 1 Hz, 5–10 mV amplitude.
4. Take the high-frequency real-axis intercept as bulk resistance Rb (Ω).
5. Compute NM = (Rb × A × κ) ÷ L, where A is separator area in cm², κ is bulk electrolyte conductivity in S/cm, and L is compressed thickness in cm.
6. Repeat with two and three stacked discs. Plot resistance against stack thickness: the slope gives resistance per unit thickness, and the intercept exposes contact resistance you would otherwise misread as separator resistance.

Worked example: Rb = 5.0 Ω, A = 2.01 cm², κ = 0.010 S/cm, L = 0.050 cm compressed → NM = (5.0 × 2.01 × 0.010) ÷ 0.050 ≈ 2.0. Low single-digit values are what a ~90% porosity, low-tortuosity mat should give; dense polyolefin separators typically report considerably higher. [VERIFY typical literature ranges for your specific electrolyte before quoting them.]

What goes wrong in practice: an unsoaked edge, a trapped bubble under the disc, or a stray fiber bridging the two electrodes. Each shifts the intercept and invalidates the calculation. Run three cells minimum and discard outliers with a stated rule.

How to choose a grade

Choose on thickness first, retention second — in that order, every time.

• GF/D (675 µm, 2.7 µm) — the default. Thick enough to survive alkali metal dendrites and rough electrode edges. Use for lithium metal, sodium metal and zinc anode half-cells where a short would end the experiment. Trade-off: it holds ~125 µL, so any energy-density figure you calculate from that cell is fiction.
• GF/A or GF/C (260 µm) — when thickness is the problem. Use when stack height is tight in a CR2032 with a thick electrode and spacer, or when you are deliberately cutting electrolyte volume toward a realistic E/C ratio. Accept a higher short risk with metal anodes.
• GF/F (420 µm, 0.7 µm) — the compromise. Finest retention with mid-range thickness. Sensible for particulate crossover work and for double-layer stacks where GF/D is simply too tall.
• GF/B (675 µm, 143 g/m²) — rarely correct. Same thickness as GF/D but denser and less permeable, with no compensating benefit in a cell. If you reach for it, have a reason.

Stacking two thin discs is common with metal anodes, but it doubles thickness, doubles electrolyte and adds a contact interface that shows up in your impedance. One thicker disc beats two thin ones.

Handling mistakes that quietly ruin glass fiber separator data

Most glass fiber problems are handling problems, not material problems.

• Cutting with scissors. Sheds fibers, frays the edge and gives an out-of-round disc. Use a hardened punch on a cutting mat and replace it when the edge dulls — a dull punch tears rather than cuts.
• Skipping the dry. Borosilicate microfiber picks up moisture fast. Dry punched discs at 120 °C under vacuum overnight and transfer them warm into the antechamber. Binder-free glass microfiber tolerates roughly 500 °C, so hotter bakes are possible — but never assume that for a binder-containing grade.
• Static cling. Bone-dry discs jump off metal tweezers and stick to nitrile gloves. Ceramic or PTFE-tipped tweezers solve it.
• Uncontrolled compression. Crimper pressure changes final thickness and therefore resistance. Fix the pressure at the start of a study and log it.
• Sharing the punch with metal foils. Embedded swarf causes soft shorts that look exactly like poor coulombic efficiency.

Cost check: a 47 mm circle yields four to six clean 16 mm discs after edge waste, so a 100-pack gives roughly 400–600 separators. 20 × 20 cm sheets are cheaper per disc if you punch in volume.

When glass fiber is the wrong separator

Glass fiber is a research tool, not a battery component. Do not use it when:

• You are reporting cell-level energy density. A 675 µm separator plus 125 µL of electrolyte swamps the stack; the number will be wrong by a wide margin.
• You are validating a formulation for scale-up. Glass fiber cannot be wound, has poor tensile strength and offers no thermal shutdown behaviour.
• You are studying separator safety or thermal runaway. Its behaviour is nothing like a polyolefin membrane, so conclusions do not carry over.
• You need tight cross-lab reproducibility. Thickness, compression and dosing vary enough that glass fiber results are genuinely hard to replicate elsewhere.

The sensible workflow: use glass fiber to isolate electrode or electrolyte behaviour, then repeat your best candidates on a commercial polyolefin separator before drawing conclusions.

Frequently asked questions

What is the difference between GF/A and GF/D separators?

Thickness and retention. GF/D is 675 µm with a 2.7 µm retention rating; GF/A is 260 µm at 1.6 µm. Porosity is nearly identical at 91–92%, so the real difference is that GF/D holds roughly 2.6 times more electrolyte and gives far better protection against dendrite shorts.

Which glass fiber separator grade is best for lithium metal batteries?

GF/D is the usual choice. Its 675 µm thickness resists dendrite penetration better than thinner grades, which matters most in long cycling tests. The trade-off is a large electrolyte reservoir — around 125 µL in a 16 mm disc — that flatters cycling performance relative to a commercial separator.

Is there a GF/E grade?

No. The binder-free glass microfiber range runs GF/A, GF/B, GF/C, GF/D and GF/F. References to "GF/E" in papers or listings are almost always typos for GF/F or GF/D. Confirm the retention and thickness figures rather than trusting the letter code alone.

How thick is a glass fiber separator compared with a polyolefin one?

Ten to twenty-seven times thicker. Glass fiber grades run 260–675 µm; typical polyolefin battery separators are 20–25 µm. That gap is why glass fiber cells cannot be used to estimate energy density, and why their capacity retention rarely transfers to pouch cells.

Do glass fiber separators need drying before use?

Yes. Borosilicate microfiber adsorbs moisture quickly, and residual water degrades carbonate electrolytes and metal anodes. Dry punched discs at 120 °C under vacuum overnight, then move them into the glovebox antechamber while still warm. Binder-free grades tolerate much higher bake temperatures if needed.

Which grade should I use for polysulfide shuttle work?

GF/F at 0.7 µm is the finest available, but it will not block dissolved polysulfides — those are molecular species, orders of magnitude smaller than any glass fiber pore. Use GF/F for particulate crossover only; chemical shuttle suppression needs a coated or functional separator.

Conclusion

Across GF/A to GF/F, porosity is effectively constant at 90–92%. What changes is thickness (260–675 µm) and retention rating (0.7–2.7 µm) — and thickness is what drives electrolyte volume, ionic path length and short protection.

Run the four-line pore-volume calculation for the grade sitting in your glovebox right now, compare it against the electrolyte you actually pipette, and put both numbers in your experimental section. If they are badly mismatched, you have found a variable worth fixing before you run another cell.