PFAS treatment bench

How EPA’s four Best Available Technologies remove PFAS. Pick a technology, the compounds in the water and the operating conditions, then move through time to see which compounds break through first and what that does to run length and residuals.

Educational modelNot an engineering design or compliance determination. Treatment selection requires site-specific pilot testing. Limits as of 2026-10-18.

Loading the treatment bench…

What this shows

On GAC, HFPO-DA and the short-chain compounds break through earlier than PFOA and PFOS. PFOS lasts longest of the compounds EPA modelled.
Background organic carbon competes for carbon adsorption sites. Each 1 mg/L of TOC takes about 38,000 bed volumes off every compound’s GAC run in EPA’s equations.
Single-use PFAS-selective resin treats three to four times as many bed volumes as GAC on the sulfonates (PFOS, PFHxS, PFBS), and holds sulfonates longer than carboxylates.
RO and NF remove PFAS broadly without a breakthrough front, but move them into a concentrate stream that has to be managed.
Flow sets vessel size and residual volumes. At a fixed contact time it does not change how long the media lasts.

What it doesn’t

It is not a design tool. EPA’s equations are pooled central-tendency fits (adjusted R² 0.62 for GAC, 0.71 for AIX) and EPA says they should not replace site-specific analysis or pilot testing.
It models one vessel with no lead-lag credit, holds bed volumes constant across contact time, and ignores sulfate, nitrate and organic fouling effects on resin.
It leaves out pairs with no sourced value: PFNA on carbon and resin and HFPO-DA and PFBA on resin (no EPA bed-life equation), and HFPO-DA on NF (no Treatability Database removal value). PFBA on GAC is an illustrative value.
It shows no costs. Cost depends on inputs this tool does not collect and is out of scope.
It makes no health claims. For health information see ATSDR’s PFAS pages ↗.

Federal reference limits

As of October 18, 2026. The 2024 rule is in force as written: initial monitoring by April 26, 2027, MCL compliance by April 26, 2029. EPA’s May 2026 proposals to move the PFOA/PFOS date and to rescind the other limits are not final, and the D.C. Circuit challenge (No. 24-1188) was argued September 18, 2026 with no ruling as of this date.

Compound	Gauge reference	Status
PFOA	4 ng/L	MCL 4.0 ng/L in force. Compliance date April 26, 2029; EPA proposed moving it to 2031 (not final).
PFOS	4 ng/L	MCL 4.0 ng/L in force. Compliance date April 26, 2029; EPA proposed moving it to 2031 (not final).
PFHxS	10 ng/L	MCL 10 ng/L and Hazard Index in force. EPA proposed rescinding both in May 2026 (not final).
PFNA	10 ng/L	MCL 10 ng/L and Hazard Index in force. EPA proposed rescinding both in May 2026 (not final).
HFPO-DA (GenX)	10 ng/L	MCL 10 ng/L and Hazard Index in force. EPA proposed rescinding both in May 2026 (not final).
PFBS	2,000 ng/L	No individual MCL. Counts only in the Hazard Index (health-based water concentration 2,000 ng/L), which EPA proposed rescinding.
PFBA	None	No federal MCL. Shown as a short-chain teaching example.

4 ppt is about 4 drops in 20 Olympic swimming pools.

Assumptions: one pool is 50 m × 25 m × 2 m = 2,500 m³ (2.5 million litres), so 20 pools hold 50 million litres; one drop is 0.05 mL. Four drops are 0.2 mL = 2 × 10⁻⁴ L, and 2 × 10⁻⁴ ÷ 5 × 10⁷ = 4 × 10⁻¹², four parts per trillion. The comparison is by volume; drinking-water limits are mass per volume (4 ng/L), which is equivalent for water.

Go further

EPA’s four Best Available Technologies, in plain language
Funding rounds by state for treatment projects
Funded PFAS treatment projects utilities have already won
Look up a water system, then open it in the bench from its page

Sources, parameters and assumptions

Model

GAC and resin use EPA’s pooled bed-life regressions, BV = A·(TOC or total PFAS) + A_R·%R + B, evaluated at 50% removal. A log-logistic curve, C/C₀ = 1 / (1 + (BV50/BV)ⁿ), passes through that point. Its steepness n is set so the midpoint slope equals EPA’s linear slope (n = 4·BV50/|A_R|), then bounded so that BV50/BV10 stays within the range observed by Chow et al. (GAC 1.2–2.7, resin 1.4–4.9). Days in service = BV × EBCT ÷ 1,440.

Parameter	Value	Basis
GAC A_TOC	-37,932 BV per mg/L	EPA T&C Table 2-6 (10-min EBCT; valid to TOC 3.2 mg/L)
GAC A_R	-36,309	EPA T&C Table 2-6
GAC B, HFPO-DA	113,034 BV	EPA T&C Table 2-6
GAC B, PFBS	129,357 BV	EPA T&C Table 2-6
GAC B, PFHxS	129,357 BV	EPA T&C Table 2-6
GAC B, PFOA	139,862 BV	EPA T&C Table 2-6
GAC B, PFOS	143,731 BV	EPA T&C Table 2-6
GAC BV50, PFBA	5,000 BV	Illustrative. No EPA equation. Bounded by full-scale complete breakthrough near 7,100 BV (Bertanza et al. 2020). No TOC effect applied.
AIX A_PFAS	-6.04 BV per ng/L	EPA T&C Table 3-5 (simulated 3.3-min EBCT; valid to 7,044 ng/L)
AIX A_R	-198,242	EPA T&C Table 3-5
AIX B, PFOA	390,787 BV	EPA T&C Table 3-5
AIX B, PFBS	439,515 BV	EPA T&C Table 3-5
AIX B, PFHxS	439,515 BV	EPA T&C Table 3-5
AIX B, PFOS	439,515 BV	EPA T&C Table 3-5
Not modelled	—	PFNA on GAC and resin; HFPO-DA and PFBA on resin (EPA could not develop a breakthrough relationship). HFPO-DA on NF (no Treatability Database removal value).
TOC range	0.5–2 mg/L	EPA’s cost-analysis bracket. Above about 2.4 mg/L the linear equation gives HFPO-DA a negative bed life, so the bench stops at 2.0.
EBCT, GAC	7.5–26 min	Full-scale PFAS plants, EPA T&C §2.6.1. Bed volumes held constant across EBCT (simplification).
EBCT, resin	2–6 min	Vendor range 3–6 min, EPA T&C §3.6.1
GAC apparent density	0.54 g/mL	Within 0.49–0.62 g/mL vendor datasheet range (Liu et al. 2019, Table 1). Resin density not sourced; no resin mass shown.
RO/NF removal	95%	EPA’s planning assumption for concentrate estimates (T&C §4.2.2). Applied only where the Treatability Database reports a removal for that compound and membrane (maxima 98–99.9% for the compounds here). HFPO-DA on RO has only a lower bound (>64.2%, effluent below detection), so 95% is marked illustrative; HFPO-DA on NF has no value and is not modelled.
RO/NF recovery	70–92%	EPA cost model 80% groundwater, 85% surface water, 70–75% small systems (T&C §7.4.1.4); full-scale plants 85–92% (§4.2.2)
Concentrate strength	(1 − r(1 − R)) ÷ (1 − r)	Mass balance. At 95% removal and 85–92% recovery this gives EPA’s “6 to 12 times” influent.
Change-out trigger	First compound to reach its reference	Single vessel, no lead-lag credit, as in EPA’s cost model (T&C §2.6.2)
Badge thresholds	70% / 100%	Display convention used across PFAS Comply. Not regulatory.
Scenario influents	—	Illustrative teaching values, not real systems. “Load my system” uses the highest reported result per compound.

Technical sources

Regulatory sources (as of October 18, 2026)

What this shows

On GAC, HFPO-DA and the short-chain compounds break through earlier than PFOA and PFOS. PFOS lasts longest of the compounds EPA modelled.

Background organic carbon competes for carbon adsorption sites. Each 1 mg/L of TOC takes about 38,000 bed volumes off every compound’s GAC run in EPA’s equations.

Single-use PFAS-selective resin treats three to four times as many bed volumes as GAC on the sulfonates (PFOS, PFHxS, PFBS), and holds sulfonates longer than carboxylates.

RO and NF remove PFAS broadly without a breakthrough front, but move them into a concentrate stream that has to be managed.

Flow sets vessel size and residual volumes. At a fixed contact time it does not change how long the media lasts.

What it doesn’t

It is not a design tool. EPA’s equations are pooled central-tendency fits (adjusted R² 0.62 for GAC, 0.71 for AIX) and EPA says they should not replace site-specific analysis or pilot testing.

It models one vessel with no lead-lag credit, holds bed volumes constant across contact time, and ignores sulfate, nitrate and organic fouling effects on resin.

It leaves out pairs with no sourced value: PFNA on carbon and resin and HFPO-DA and PFBA on resin (no EPA bed-life equation), and HFPO-DA on NF (no Treatability Database removal value). PFBA on GAC is an illustrative value.

It shows no costs. Cost depends on inputs this tool does not collect and is out of scope.

It makes no health claims. For health information see ATSDR’s PFAS pages ↗.

Federal reference limits

Compound	Gauge reference	Status
PFOA	4 ng/L	MCL 4.0 ng/L in force. Compliance date April 26, 2029; EPA proposed moving it to 2031 (not final).
PFOS	4 ng/L	MCL 4.0 ng/L in force. Compliance date April 26, 2029; EPA proposed moving it to 2031 (not final).
PFHxS	10 ng/L	MCL 10 ng/L and Hazard Index in force. EPA proposed rescinding both in May 2026 (not final).
PFNA	10 ng/L	MCL 10 ng/L and Hazard Index in force. EPA proposed rescinding both in May 2026 (not final).
HFPO-DA (GenX)	10 ng/L	MCL 10 ng/L and Hazard Index in force. EPA proposed rescinding both in May 2026 (not final).
PFBS	2,000 ng/L	No individual MCL. Counts only in the Hazard Index (health-based water concentration 2,000 ng/L), which EPA proposed rescinding.
PFBA	None	No federal MCL. Shown as a short-chain teaching example.

Parameter

Value

Basis

GAC A_TOC

-37,932 BV per mg/L

EPA T&C Table 2-6 (10-min EBCT; valid to TOC 3.2 mg/L)

GAC A_R

-36,309

EPA T&C Table 2-6

GAC B, HFPO-DA

113,034 BV

EPA T&C Table 2-6

GAC B, PFBS

129,357 BV

EPA T&C Table 2-6

GAC B, PFHxS

129,357 BV

EPA T&C Table 2-6

GAC B, PFOA

139,862 BV

EPA T&C Table 2-6

GAC B, PFOS

143,731 BV

EPA T&C Table 2-6

GAC BV50, PFBA

5,000 BV

Illustrative. No EPA equation. Bounded by full-scale complete breakthrough near 7,100 BV (Bertanza et al. 2020). No TOC effect applied.

AIX A_PFAS

-6.04 BV per ng/L

EPA T&C Table 3-5 (simulated 3.3-min EBCT; valid to 7,044 ng/L)

AIX A_R

-198,242

EPA T&C Table 3-5

AIX B, PFOA

390,787 BV

EPA T&C Table 3-5

AIX B, PFBS

439,515 BV

EPA T&C Table 3-5

AIX B, PFHxS

439,515 BV

EPA T&C Table 3-5

AIX B, PFOS

439,515 BV

EPA T&C Table 3-5

Not modelled

—

PFNA on GAC and resin; HFPO-DA and PFBA on resin (EPA could not develop a breakthrough relationship). HFPO-DA on NF (no Treatability Database removal value).

TOC range

0.5–2 mg/L

EPA’s cost-analysis bracket. Above about 2.4 mg/L the linear equation gives HFPO-DA a negative bed life, so the bench stops at 2.0.

EBCT, GAC

7.5–26 min

Full-scale PFAS plants, EPA T&C §2.6.1. Bed volumes held constant across EBCT (simplification).

EBCT, resin

2–6 min

Vendor range 3–6 min, EPA T&C §3.6.1

GAC apparent density

0.54 g/mL

Within 0.49–0.62 g/mL vendor datasheet range (Liu et al. 2019, Table 1). Resin density not sourced; no resin mass shown.

RO/NF removal

95%

EPA’s planning assumption for concentrate estimates (T&C §4.2.2). Applied only where the Treatability Database reports a removal for that compound and membrane (maxima 98–99.9% for the compounds here). HFPO-DA on RO has only a lower bound (>64.2%, effluent below detection), so 95% is marked illustrative; HFPO-DA on NF has no value and is not modelled.

RO/NF recovery

70–92%

EPA cost model 80% groundwater, 85% surface water, 70–75% small systems (T&C §7.4.1.4); full-scale plants 85–92% (§4.2.2)

Concentrate strength

(1 − r(1 − R)) ÷ (1 − r)

Mass balance. At 95% removal and 85–92% recovery this gives EPA’s “6 to 12 times” influent.

Change-out trigger

First compound to reach its reference

Single vessel, no lead-lag credit, as in EPA’s cost model (T&C §2.6.2)

Badge thresholds

70% / 100%

Display convention used across PFAS Comply. Not regulatory.

Scenario influents

—

Illustrative teaching values, not real systems. “Load my system” uses the highest reported result per compound.