Tungsten Residue Filtration and Washing: Cut WO3 Loss

Last updated: 2026-11-03

In sodium tungstate flowsheets, digestion is rarely what sets tungsten recovery — the wash train after tungsten residue filtration is. Alkaline pressure digestion typically dissolves 96-99% of feed WO3 (feed-dependent), yet unwashed leach residue can still carry 1-5% of feed WO3 out of the circuit in trapped liquor. Wash design, not autoclave severity, usually decides the final number.

Key Takeaways

• Digestion converts tungsten to sodium tungstate; recovery is lost later, in the solid-liquid separation and washing train, not the autoclave.
• The residue is fine and gelatinous: expect low filtration throughput, cloth blinding, and 25-35% cake moisture (typical).
• Dilution after digestion cuts free alkali and viscosity — but over-dilution inflates the downstream evaporation load.
• Three to five countercurrent wash stages at 1.5-3 m3 of water per tonne of residue hold WO3 loss below 2%.
• Hold filtrate above roughly 60-70 °C, or sodium salts crystallize in cloths, lines, and pumps.

Where Tungsten Residue Filtration Sits in the Flowsheet

Tungsten enters the hydrometallurgical route as one of two minerals: wolframite, a ferberite-huebnerite series (Fe,Mn)WO4, or scheelite, CaWO4. Both are attacked in alkaline pressure digestion — NaOH for wolframite alkaline leaching, NaOH or mixed NaOH-Na2CO3 (soda-assisted) for scheelite — at roughly 160-225 °C and 0.8-2.5 MPa for 1.5-3 hours, with NaOH dosed at 1.3-2.5 times stoichiometric (typical ranges, to be confirmed by test work). Tungsten leaves as soluble sodium tungstate, Na2WO4; iron and manganese report to the solids as hydroxides and oxides, calcium as carbonate or hydroxide, and part of the silicate gangue dissolves as sodium silicate.

Tungsten residue filtration is a solid-liquid separation duty that divides the diluted digester discharge into a pregnant sodium tungstate solution and a solid leach residue, and it is the pivot of every tungsten APT flowsheet. From there the sequence is fixed: dilution of the discharge, filtration, countercurrent washing of the residue, purification of the combined liquor, then crystallization of ammonium paratungstate (APT). Silica removal and molybdenum sulfide precipitation precede ion exchange (IX) or solvent extraction (SX), and the purified ammonium tungstate solution finishes in APT crystallization.

The mass-balance logic is what makes washing — not the filter itself — the recovery lever. After digestion, undissolved scheelite or wolframite is usually a minor WO3 carrier; most WO3 found in residue is dissolved tungsten sitting in the interstitial liquor the cake retains. A filter’s job is bulk liquid removal, and it does that well. But whatever solution stays in the cake at cake-moisture concentration leaves the circuit. Recovery loss therefore equals retained solution volume times WO3 concentration in that solution — a washing problem. The lever works in both directions: cutting cake moisture from 35% to 26% at unchanged tenor removes roughly a quarter of the soluble loss before any wash water is applied, which is why moisture and wash design must be optimized together, not in sequence. A wash ratio is a design figure that expresses wash water applied per tonne of residue, and countercurrent washing is a staging arrangement that moves fresh water against the solids flow so the last contact sees the cleanest liquor. Get those two right and residue losses collapse; get them wrong and no filter upgrade will recover the metal.

Digestion and Dilution Conditions That Set the Tungsten Residue Filtration Duty

Digestion severity and the dilution step immediately downstream jointly fix how hard the filtration duty will be. Higher free caustic and higher tungsten tenor raise liquor viscosity and density; more sodium silicate in solution means more gel on dilution. Plants therefore dilute the discharge — with wash water returning from the wash train where possible — to cut free alkali and viscosity enough for cloth filtration while keeping WO3 tenor high, because every cubic meter of added water must later be evaporated in the APT loop. A filtrate of 100-180 g/L WO3 (indicative) is the usual compromise band.

Treat every row as a typical range to be confirmed on your own feed: mineralogy, silicate content, and grind size move all of them. Two rows deserve special attention. Free alkali is the quiet cost driver — it attacks unsuitable cloths and plate materials and it must be neutralized or recycled later — and cake moisture is the direct multiplier on soluble loss, which is why we treat filter cake moisture as a monitored process number, not a vendor brochure figure. Dilution discipline matters too: enough water to protect the cloth, not so much that the evaporation train becomes the plant’s largest energy consumer.

Why Is Tungsten Residue So Hard to Filter?

Three material properties conspire. First, the solids are fine and gelatinous. Iron and manganese hydroxides from wolframite digestion, and calcium carbonate or hydroxide from soda-assisted scheelite digestion, precipitate as sub-micron, high-surface-area particles. Dissolved silica leaves solution as sodium silicate and, once the pulp is diluted and cooled, polymerizes into silica gel that coats every particle. The result is a residue with the dewatering character of a sludge, not a sand: low specific throughput, high specific resistance, and slow desaturation.

Second, the cake is compressible. Under pressure the gel matrix collapses, permeability drops as the pressure differential rises, and pushing harder yields diminishing returns. This is the classic argument for membrane squeezes — mechanically compacting the cake after a low-pressure formation phase — and for resisting the instinct to raise pump pressure when throughput sags.

Third, the liquor itself works against the equipment. Caustic sodium tungstate solution is viscous, and viscosity falls steeply with temperature. Below roughly 60-70 °C, sodium salts — sodium tungstate hydrates, carbonate, and residual caustic species — approach saturation, and any local cooling, evaporation at a pump gland, or flash across a valve crystallizes them inside cloth pores, pipe walls, and pump internals. Cloth blinding then looks chemical rather than mechanical: the blinding mechanisms in caustic tungsten service are usually scale plus gel, not particles alone. The engineering judgement that follows is easy to state and expensive to ignore: keep the filtrate and wash circuit above roughly 60-70 °C from the filter to the purification feed tank. That single constraint eliminates most low-temperature pump styles, cheap seals, and several cloth polymers, and it argues for short, hot, insulated filtrate runs rather than long transfer lines. It also reframes the whole task: tungsten residue filtration is a heat-management problem as much as a mechanical one, because permeability, viscosity, and salt saturation all move with temperature. Cloth and gasket selection must be confirmed in caustic slurry service at temperature, not extrapolated from acid-plant experience, and leaf tests run cold will flatter a cloth that fails hot.

How Do You Design the Washing Train? A 7-Step Method

Design the wash train from data, in this order:

1. Measure the residue wash curve in the lab. Repulping or percolation tests on actual digester discharge establish how quickly soluble WO3 leaves the cake per wash contact. A bench-scale CCD wash unit gives stage-by-stage data in days; skip this and every later number is a guess.
2. Choose the stage number from the wash curve and the WO3 loss target. Match the cumulative removal at your chosen efficiency to the target — commonly below 2% of feed WO3 in residue, best practice approaching 1% (indicative).
3. Set the wash water ratio. Typical duty sits at 1.5-3 m3/t of residue across the train; the wash ratio method converts the curve into a per-stage figure.
4. Size each thickener or filter per stage. Worked example: 60 t/d of residue at 0.8% WO3 carries 0.48 t/d of WO3; the 0.2% target allows 0.12 t/d, so 75% of soluble WO3 must be removed. With retained liquor of about 0.40 t per tonne of dry residue (about 30% moisture wet basis, indicative) and 0.6 t/t wash per stage, the ideal per-stage residual fraction is 0.40/(0.40+0.60) = 0.40. At a realistic 70% stage efficiency, the residual per stage is 1 – 0.7 x 0.60 = 0.58, and three stages leave 0.583, or about 0.20 — landing residue WO3 near 0.16%, inside target. Two stages would not; the arithmetic decides, not preference.
5. Decide the countercurrent configuration. Fresh water enters the last stage only; intermediate wash filtrate returns one stage upstream. Where a filter press does the primary duty, decide between cake-wash passes in the press and repulping between presses — the wash curve answers which is adequate.
6. Check water balance and evaporation load. Every extra cubic meter of wash water reappears as evaporation duty in the APT loop. Trade wash water against steam cost explicitly, at the margin.
7. Verify with a plant mass balance. After commissioning, close the WO3 balance across digestion, filtration, and washing over at least two weeks of assays before declaring the loss number real.

One habit ties the seven steps together: keep the division of labor explicit. Tungsten residue filtration removes bulk liquor, the wash train removes what the cake retains, and the evaporator pays for any water you borrowed to make either job easier. Designs that blur those three roles — extra dilution to flatter the filter, extra wash water to flatter the wash curve — simply move the cost downstream where it compounds hourly.

Which Filter Fits Tungsten Residue?

No single machine wins; the duty decides. The comparison below assumes caustic sodium tungstate service at temperature, gelatinous solids, and typical feed from the table above — all indicative.

For hot caustic duty where the lowest soluble loss matters, a membrane filter press is preferred because the membrane squeeze cuts cake moisture by several points, and moisture is the direct multiplier on WO3 loss. For plants with large residue tonnages and tolerant loss targets, a belt filter is preferred because continuous operation removes batch labour and gives easy multistage cake washing on one machine. For polishing duty on very fine, gelatinous solids where suspended solids in filtrate must be near zero, a precoat drum is preferred because the precoat layer presents a fresh filtration surface continuously — at the price of precoat consumption. For feed with coarse, fast-settling solids and loose moisture targets, a decanter is preferred because it needs no cloth at all in the most aggressive liquor. For capital-limited mid-size plants, a recessed plate press is preferred because it is the cheapest machine that still delivers a washable cake. Cloth selection and squeeze behaviour should be confirmed on your own pulp, and pilot-scale runs of plate-and-frame filter presses sized for this purpose remain the cheapest insurance in this segment — the same logic set out for filter presses in mining duty generally.

What Goes Wrong, and How Do You Fix It?

1. Silica gel blinding. Throughput decays within weeks of start-up. Cause: silicate-rich feed polymerizing to gel on dilution, coating particles and cloth. Action: confirm silica in feed and filtrate, adjust dilution temperature and rate, add a desilication hold upstream, and shorten cloth wash cycles.
2. Cloth scaling by sodium salts. Cloth stiffens and flux falls despite clean solids. Cause: filtrate cooled below the saturation line, crystallizing sodium tungstate or carbonate in the weave. Action: keep filtrate above roughly 60-70 °C, insulate lines, and hot-caustic wash cloths on a schedule.
3. Fine solids passing into filtrate. Turbid filtrate fouls IX or SX feed. Cause: cloth mesh too open, gel fines squeezing through, or a cloth damaged by caustic attack. Action: run a cloth trial at the real temperature, verify pore rating against particle size data, and inspect cloths on a fixed interval.
4. Cake cracking and uneven washing. Wash water short-circuits, and residue WO3 assays stay high while wash water rises. Cause: cake shrinkage and cracking as it dewaters, letting wash follow the cracks. Action: reduce formation pressure differential, add a membrane squeeze before washing, or repulp-wash instead of in-press washing.
5. High cake moisture. Moisture sits above 35% and soluble losses track it. Cause: compressible gel cake, insufficient squeeze pressure, or blinding reducing effective drainage. Action: restore cloth permeability first, then verify squeeze pressure and time; moisture responds to mechanics before chemistry.
6. Corrosion and caustic attack on cloth and plates. Cloth weakens, plates pit or craze. Cause: free caustic above design, hot spots, or incompatible gasket and polymer grades. Action: audit free alkali against design, upgrade to caustic-rated polymers and gaskets, and log cloth life to catch drift early.
7. Filtrate dilution from wash water breakthrough. Filtrate WO3 tenor sags and evaporation load climbs. Cause: wash water applied too early in the cycle, cracked cakes, or misplaced wash headers. Action: sequence wash after the cake is formed and squeezed, meter wash water, and trend filtrate tenor per cycle.

Cost and WO3 Loss Economics

Express the loss arithmetic in tonnes of WO3 per year, because that is what the recovery ledger sees. APT carries roughly 88.5% WO3 (stoichiometric, dry basis). A small plant producing 500 t/y of APT therefore fixes about 440 t/y of WO3 in product; a 3,000 t/y APT plant fixes about 2,650 t/y. Against feed WO3, moving residue loss from 5% to 2% of feed returns roughly 14 t/y of WO3 at the small scale and roughly 82 t/y at the large scale (indicative; exact figures depend on the plant’s actual balance) — before any metal price is applied, and the same tonnes recur every year.

Read that table as an order-of-magnitude planning tool, not a design number: the loss percentages apply to feed WO3, and feed WO3 exceeds product WO3 by exactly the loss, which is why the two columns do not scale linearly with the 440 and 2,650 figures.

Washing costs are modest against that. Residue tonnage runs at roughly 0.4-0.9 t per tonne of feed WO3 depending on mineral and gangue (indicative), so wash water at 1.5-3 m3/t means a few cubic meters per tonne of WO3. The dominant recurring items are cloth replacement in caustic hot service (cloths typically last 2-6 months, indicative), batch labour on presses — one operator per shift per press line is a fair planning figure — and heat: wash water that ends in the APT loop must be evaporated, a thermal duty of tens of kWh per cubic meter (typical, plant-specific). Over-dilution to make filtration easy is the most common false economy: it buys throughput at the evaporator and pays for it every operating hour.

Field Data: Three Washing Configurations Compared

The following anonymised figures come from plant surveys and commissioning balances (indicative, dry-basis residue assays):

• Plant A — wolframite feed, recessed plate press with two repulp washes. Washed residue 0.9% WO3, cake moisture 33%, cloth life about 10 weeks, wash water 1.8 m3/t. Basis: operating survey over one quarter. Residue loss sat near 2.5% of feed WO3 — workable, not best practice.
• Plant B — scheelite feed, soda-assisted digestion, membrane press with three-stage wash. Residue 0.4% WO3, moisture 26%, cloth life about 14 weeks, wash water 2.2 m3/t. Basis: commissioning mass balance. The membrane squeeze did most of the work; wash water did the rest.
• Plant C — wolframite feed, belt filter, four countercurrent stages. Residue 0.7% WO3, moisture 36%, wash water 2.8 m3/t, highest labour of the three. Basis: plant test run. Continuous washing recovered the loss the wetter cake created — but only with more water.

The pattern repeats across surveys: where moisture is low, fewer stages suffice; where the machine leaves a wet cake, water and stages must compensate. Further washing case notes are collected in our resources library.

Frequently Asked Questions

What is tungsten residue filtration in an APT plant?

It is the solid-liquid separation step that divides diluted digester discharge from alkaline tungsten digestion into a pregnant sodium tungstate solution and a solid leach residue. The residue carries iron, manganese, calcium, and gangue hydroxides plus trapped liquor containing dissolved WO3, so the step — and the washing that follows it — sets how much tungsten actually reaches APT crystallization.

Why does washing decide recovery rather than digestion?

Digestion is normally efficient: 96-99% of feed WO3 dissolves (typical, feed-dependent). What leaves the circuit afterwards is dissolved WO3 trapped in the liquor held by the filter cake at 25-35% moisture. Unwashed residue can carry 1-5% of feed WO3; a properly designed countercurrent wash holds that below 2%. The filter removes bulk liquid; washing removes the last of it.

How many wash stages does a tungsten residue circuit need?

Three to five countercurrent stages is the typical band. The correct number comes from a laboratory wash curve and your WO3 loss target, combined with a realistic stage efficiency — often near 70% (indicative). High cake moisture or a stringent loss target pushes toward the top of the band; a membrane press with low-moisture cake can achieve the target with fewer stages.

How hot must the filtrate and wash circuit be kept?

Above roughly 60-70 °C (plant-specific saturation limit, to be confirmed from your liquor analysis). Below that, sodium tungstate hydrates, carbonate, and caustic salts can crystallize in cloth pores, pipes, and pumps, causing scale-type blinding that hot caustic washing must then undo. The constraint also dictates pump styles, seals, gaskets, and insulated, short filtrate runs.

Can wash liquor be recycled back into the process?

Yes, and it should be. Final wash filtrate is dilute, so it is best used as dilution water for the digester discharge or as first-stage wash water, keeping dissolved WO3 inside the circuit. What must not happen is sending wash liquor forward uncontrolled: it carries free caustic and fine solids into purification, where both interfere with silica removal and IX or SX operation.

Which filter cloth suits caustic tungsten residue service?

Caustic-resistant polymer cloths — typically polypropylene or suitable polyamide weaves with a pore rating matched to the residue particle size — are the usual candidates. The honest answer is that cloth selection is empirical: run trial cloths on your own hot caustic pulp, track flux, filtrate clarity, and life, and re-check whenever digestion chemistry or temperature changes. Acid-plant cloth experience does not transfer.

What test work should precede filter and wash train selection?

At minimum: residue particle size and settling data, filtration leaf tests at operating temperature and pressure, a full wash curve on real digester discharge, cloth compatibility trials in hot caustic liquor, and a closed water-balance calculation. Selection made from vendor throughput tables alone is the most common cause of undersized tungsten residue filtration installations we see in retrofit enquiries.

Conclusion

In a tungsten APT flowsheet, the autoclave usually does its job; the washing train is where tungsten is quietly won or lost. Size tungsten residue filtration for a hot, caustic, gelatinous cake, then design the wash train from a measured wash curve — stage count, wash ratio, and configuration chosen by arithmetic, not habit — and hold residue WO3 loss below 2% of feed. Audit cake moisture and filtrate temperature weekly, because both drift before the assay does. If you want this method applied to your own digester discharge, with cloth trials and a bench wash program scoped and priced, request a quote and our process team will respond with a test plan.

References

• Lassner, E. and Schubert, W.-D., Tungsten: Properties, Chemistry, Technology of the Element, Alloys, and Chemical Compounds, Springer.
• Gaur, R.P.S., “Modern hydrometallurgical production methods for tungsten,” JOM.
• SME Mineral Processing and Extractive Metallurgy Handbook, Society for Mining, Metallurgy and Exploration, Englewood, CO.
• U.S. Geological Survey, Mineral Commodity Summaries, Tungsten chapter (annual).
• Habashi, F., A Textbook of Hydrometallurgy, Métallurgie Extractive Québec.
• Gupta, C.K. and Mukherjee, T.K., Hydrometallurgy in Extraction Processes, CRC Press.

Changelog

• 2026-11-03: First publication.