Does ultrafiltration remove arsenic?

Ultrafiltration removes only the particulate and adsorbed fraction of arsenic. Arsenic in true solution, which is most of it in many groundwaters, passes through a UF membrane because the dissolved species are far smaller than the pore size. Where arsenic is the target, the usual route is pre-oxidation to convert arsenite to arsenate, followed by adsorption, co-precipitation or ion exchange, with reverse osmosis where the target is very low. In that arrangement UF protects the downstream step rather than performing the removal.

Does ultrafiltration remove iron and manganese?

Ultrafiltration removes iron and manganese once they have been oxidised and precipitated, because the resulting particles are well above the pore size. Dissolved ferrous iron and manganous manganese pass through. The practical consequence is that the oxidation step has to be designed to go to completion and to be matched to the pH and the organic content of the water, otherwise unoxidised metal reaches the membrane and leaves in the permeate.

Does ultrafiltration remove hardness or TDS?

No. Hardness is caused by dissolved calcium and magnesium, and TDS is by definition dissolved. Both pass through an ultrafiltration membrane. Hardness is addressed by chemical softening, by ion exchange or by reverse osmosis; TDS is addressed by reverse osmosis. Ultrafiltration is normally placed upstream of those steps to protect them from the solids and colloids that would otherwise foul them.

Is ultrafiltration a disinfection barrier?

Ultrafiltration is an effective physical barrier for bacteria and, depending on pore size and on the integrity of the module, removes a substantial fraction of viruses. It is not a substitute for a disinfection step where a regulatory disinfection credit is required, and the credit depends on the integrity testing regime rather than on the nominal pore size alone.

Technical note

What ultrafiltration does not remove

Ultrafiltration is a size screen. It holds back particles and colloids larger than its pores, and it lets everything smaller through. That is the whole mechanism — so the useful question is not "is UF good?", but "is the thing I need to remove bigger than the pore?"

A membrane supplier who does not put this page in front of you is not doing you a favour. Most of the disappointment in membrane projects traces back to this one point.

The size line

UF separates between roughly 0.01 and 0.1 micrometres

Above that window, microfiltration and conventional clarification do the work. Below it, nanofiltration and reverse osmosis take over. Ultrafiltration sits in a band that is large enough to catch suspended solids, colloids and micro-organisms, and far too large to catch anything that is genuinely dissolved.

If the contaminant is in solution, a UF membrane does not see it. This is a property of the separation, not a limitation of a particular product.

Process	Approximate pore or cut-off	Removes, broadly
Microfiltration (MF)	~0.1–10 µm	Suspended solids, bacteria, some colloids
Ultrafiltration (UF)	~0.01–0.1 µm	Suspended solids, colloids, bacteria, most viruses, oil droplets, macromolecules
Nanofiltration (NF)	~0.001–0.01 µm	Multivalent ions, hardness, some dissolved organics
Reverse osmosis (RO)	Below ~0.001 µm	Dissolved ions, most dissolved organics, TDS

Indicative ranges for orientation. The cut-off that matters is the one measured on your feed, not the one printed on a curve.

What UF does remove

It is worth being precise in both directions, because the value of UF is real and it is easy to lose sight of it when the focus shifts to what it cannot do.

Retained

Particulate

Suspended solids
Colloids and turbidity-causing fines
Metal hydroxide floc
Some precipitated hardness sludge

Retained

Biological

Bacteria
A substantial fraction of viruses
Protozoa and cysts
Algae and cell debris

Retained

Macromolecular

Emulsified and free oil droplets
Proteins and polysaccharides
High-molecular-weight organics
Some colloidal colour bodies

This is exactly the duty that protects a downstream RO train. Solids, colloids, oil and biological load are what foul RO membranes and drive up cleaning frequency; UF holds them back in a stage that can be cleaned and re-cleaned without damaging the RO elements behind it.

What passes straight through

Contaminant	Does UF remove it?	What removes it instead
Dissolved salts (Na, Cl, SO₄)	No	Reverse osmosis
Hardness (Ca, Mg)	No	Chemical softening, ion exchange, NF, RO
Arsenic in true solution	No — only the particulate fraction	Pre-oxidation plus adsorption, co-precipitation or ion exchange; RO for very low targets
Dissolved iron and manganese	No — only the oxidised precipitate	Oxidation followed by filtration
Reactive silica	No	Lime softening, MgO softening, RO with careful recovery limits
Boron	No	RO at high pH, ion exchange
Nitrate and ammonia	No	Biological treatment, ion exchange, RO
Dissolved organics, micropollutants	Mostly no — only the high-MW fraction	Activated carbon, oxidation or advanced oxidation, RO
Colour in true solution	No — only colloidal colour	Coagulation, adsorption, oxidation, NF or RO

The failure mode this table is meant to catch

A project specifies UF to meet a dissolved-species limit, the plant is built, and the limit is not met — not because the membrane failed, but because the membrane was never able to meet it. Retrofitting an adsorption or RO stage afterwards is far more expensive than designing the train correctly the first time. The table above is the check to run before the specification is frozen.

Three cases worth spelling out

Arsenic

Arsenic is present as arsenite, As(III), or arsenate, As(V), and the two behave differently. In many groundwaters the dissolved fraction dominates, and dissolved arsenic is far smaller than a UF pore — so it passes. UF will remove the fraction that is already particulate or adsorbed onto iron and manganese oxides in the feed, and that fraction varies from site to site.

The usual treatment route is oxidation to convert As(III) to the more readily removed As(V), then adsorption onto a suitable medium, co-precipitation with iron, or ion exchange. Where the target is very low, RO follows. In that train UF has a clear job — protecting the arsenic-removal step from solids — but it is not the arsenic step.

Iron and manganese

Here the position is more favourable, provided the oxidation step is taken seriously. Ferrous iron and manganous manganese in solution pass through UF. Once oxidised and precipitated, the particles are far larger than the pore and are removed well. Manganese is the harder of the two: it needs a higher pH and a stronger oxidant to go to completion, and it is sensitive to the organic content of the water.

If oxidation is incomplete, the unoxidised metal arrives at the membrane and leaves in the permeate. The membrane is then blamed for a failure that belongs to the upstream step — which is why we ask for the oxidant dose, the pH and the contact time rather than only the raw metal concentration.

Hardness and silica

Both are dissolved and both pass through. Hardness is removed by chemical softening, ion exchange or RO. Reactive silica is the more awkward of the two because it limits RO recovery: as the concentrate concentrates, silica approaches saturation and begins to scale. That is the situation TMF softening and STRO / DTRO are designed for, and it is a recovery-limit problem rather than a membrane-selection problem.

How to use this when you specify a train

Write the train from the target backwards

Start from the species that has to be removed to meet the discharge or reuse limit, and ask for each one: is it in solution? If it is, name the unit operation that removes it. Only then decide where UF sits in the line.

In most industrial trains UF ends up upstream of RO, protecting it. That is a strong and defensible role. It is a different role from being the removal step, and a specification that confuses the two will not meet its limit however good the membrane is.

Send us the target species and the limit, not only the solids. We would rather tell you that UF is not the answer than have you find out at commissioning.

Questions we get asked

Does UF remove microplastics?

Largely yes. Microplastic particles are typically well above the UF pore size, so they are held back as particulate material. The dissolved and nanoscale fraction is a separate question and is not well defined by a pore size specification.

Does UF remove PFAS?

Only the fraction associated with particulate or colloidal material. Dissolved PFAS molecules are far smaller than the pore and pass through. Removal relies on activated carbon, ion exchange or RO, with UF again acting as a protective stage rather than the removal step.

Does UF remove colour?

Colloidal colour is removed; colour that is genuinely dissolved is not. In practice a plant sees a mixture of the two, which is why colour removal is normally confirmed by a jar test or a pilot rather than assumed from a membrane specification.

Tell us the target, not just the feed

Send the species that has to be removed and the limit it has to meet. We will tell you what UF contributes, what it does not, and which step has to carry the rest.

Describe your target