Ball Valves in Direct Lithium Extraction: PH 1 Eluate, Gypsum Scale And 438,000 Cycles in Twenty-Five Years

Three thousand tonnes of lithium carbonate went back to the re-dissolve tank because a certificate read 18 parts per million iron where the purchase order allowed 15. The plant had been running for eight months. Six weeks of sampling narrowed the source to a four inch ball valve in the eluate transfer line, a carbon steel body chosen in the basic design because the line ran at six bar and the engineer's filing category for it was water. The fluid was pH 1.4, 55 degrees C and 120 grams per litre of chloride. When the valve came out, the body wall had lost three millimetres and the ball was pitted like a practice golf ball.

Most of the iron that valve released left the plant in the residue stream and never came near the product. Enough of it did. That is the shape of the problem in direct lithium extraction: the corrosion is not dramatic, the leak is not visible, and the consequence arrives two process steps later as an off-spec batch and a customer audit. A ball valve that would be considered entirely normal in a water plant is a contamination source in a battery-grade lithium plant, and the reason is not that the duties are extreme in pressure or temperature. It is that this process moves between pH 1 and pH 12 several times a day, at chloride levels that make every stainless steel choice a compromise, and does it four hundred thousand times over the design life.

1. Direct lithium extraction is a pH swing, not a brine problem

The brine itself is the easy part. A salar brine at 200 to 3,000 milligrams per litre lithium, 200 to 350 grams per litre total dissolved solids and 30 to 35 degrees C is a chloride environment that 2507, 6Mo or titanium handles without difficulty, and the industry has decades of experience with it from seawater and potash duty.

The difficulty starts when the lithium has to be taken out of that brine and put into a much smaller volume of water. Every commercial direct extraction process does this by trapping lithium selectively on a solid and then releasing it into a strip solution whose chemistry is deliberately hostile to the equipment: dilute hydrochloric acid at pH 1 to 2, 40 to 60 degrees C, one to three grams per litre lithium, 50 to 150 grams per litre chloride. That stream is the reason a lithium plant cannot be equipped from a water treatment catalogue. It is also the reason the valve count in the acid loop, not the brine loop, drives the maintenance budget.

After elution the liquor is neutralised to pH 10 or 11, which precipitates magnesium and calcium, and then carbonated with soda ash at 85 to 95 degrees C to crystallise lithium carbonate. So the same train also contains hot caustic, hot carbonate slurry and an abrasive crystal product. Four different corrosion problems and one abrasion problem, in a line-up of perhaps sixty valves.

2. The train, in the order the valves see it

It helps to stop thinking of the plant as one service and to walk the streams one at a time, because the correct valve for each is decided by a different line of the same table.

Two of those six rows are ordinary metallurgy. Four are not, and each of the four fails in a way that a pressure test at the factory cannot detect.

3. pH 1 at 55 degrees C removes most of the stainless list

Hydrochloric acid is a reducing acid, and that single fact invalidates most of the reasoning that works elsewhere in a chloride plant. In oxidising chloride, such as seawater or hypochlorite, the passive film on a stainless steel is continually repaired by oxygen and the alloy competes on chromium, molybdenum and nitrogen content, which is what the pitting resistance equivalent number ranks. In reducing acid there is nothing to repair the film with, and the metal dissolves uniformly at a rate set by the acid concentration and the temperature.

Published isocorrosion data puts 316L at about 0.1 millimetres per year in 0.5 percent hydrochloric acid at room temperature, and at several millimetres per year in the same acid at 60 degrees C. The elution loop runs at one to two percent acid, which is two to four times that concentration, at the top of that temperature range. A 316L body in that service has a wall life measured in months, and its corrosion product is iron, which is the one impurity the product specification is least willing to forgive.

Duplex grades are not the answer, and this is where a lot of specifications go wrong. The 22 percent chromium in 2205 buys resistance to chloride pitting and to chloride stress corrosion, but the grade carries only 3 percent molybdenum and the ferrite phase is the first thing attacked in a reducing acid. Super duplex 2507 does better and is still not an acid alloy. The reasoning that selects duplex for a hot chloride stream is correct in that stream and wrong in this one, and a plant built with one material philosophy across both loops will have a corrosion problem in exactly the loop where nobody expects it.

What actually works is a short list. Hastelloy C-276, with 16 percent molybdenum and 4 percent tungsten, handles dilute hydrochloric acid at 60 degrees C comfortably and is the reference alloy for the duty; it also costs eight to fifteen times a 316L valve of the same size. Titanium is discussed below and is not on the list. The practical industrial answer, and the one that most operating plants converge on, is not a metal at all: a steel body with a PFA or PTFE lining, or a solid thermoplastic body in PVDF or PP where the pressure class allows it.

4. Titanium is the right answer to the wrong question

Titanium grade 2 is the default material for oxidising chloride service, and engineers who have spent time in a bleach plant or a seawater reverse osmosis train reach for it instinctively when they see 120 grams per litre of chloride. In hydrochloric acid it is the wrong choice, because titanium depends on a titanium hydride-free oxide film that reducing acids destabilise. Grade 2 is normally limited to a few tenths of a percent hydrochloric acid at ambient temperature and is not usable at one to two percent and 55 degrees C. Grade 7, with 0.12 to 0.25 percent palladium, extends that range but does not make it an elution loop material at commercial cost.

There is also a hydrogen story. In acid service with any galvanic coupling, or with cathodic protection on an adjacent structure, titanium absorbs hydrogen and embrittles, and the failure is a brittle crack rather than a wastage allowance. A titanium valve in an acid loop is the kind of specification that looks rigorous on paper and is quietly replaced with a lined valve within two years of start-up.

5. The iron budget is 0.16 parts per million

The reason a small amount of corrosion matters here is the concentration step. An eluate carrying two grams per litre of lithium becomes about 10.6 grams of lithium carbonate per litre of solution, because 13.9 grams of lithium makes 73.9 grams of carbonate and the ratio is 5.32 to one. Iron that enters with the eluate is carried through that step almost quantitatively. Working backwards, one milligram per litre of iron in the eluate arrives in the product at roughly 94 parts per million, and the specification is 15. So the whole iron budget for the process, downstream of the adsorbent, is about 0.16 milligrams per litre.

That number explains a set of rules that otherwise look like housekeeping. Nothing downstream of the columns may be carbon steel, not a body, not a pipe spool, not a strainer basket. Galvanised hardware is banned because zinc has its own limit and because a galvanised bolt in a lined flange is an invitation. Cast iron and ductile iron are banned for the same reason as carbon steel, with the added problem that graphite corrosion leaves a spongy surface that releases iron continuously. Pickling and passivation to ASTM A967 is routine, not cosmetic, because a heat tint left on a weld is a depleted zone that will corrode preferentially and feed the product stream for years.

The same arithmetic applies to the other impurities, and it is worth putting on one sheet because it converts a product specification written in percent into a process limit written in milligrams per litre, which is the form a valve and piping specification can actually use.

6. Scale grows where the flow does not go

Calcium sulphate is the first scaling salt in most salar brines, and it has two properties that matter to a valve. Its solubility is only about 2.4 grams per litre at 25 degrees C, and it is retrograde, meaning it becomes less soluble as the temperature rises above about 40 degrees C. A heater in a gypsum-saturated brine is therefore a scaling device, and so is any surface that is warmer than the bulk fluid, including a steam-traced valve body and a bonnet that sits in still air.

Calcium carbonate behaves differently and is pH-driven. At pH 10.5 the carbonate equilibrium shifts and calcium carbonate, with a solubility of about 14 milligrams per litre, comes out of solution on any surface it can reach, including the inside of a seat cavity that has not been swept since the last stroke. Magnesium hydroxide precipitates in the same step. The result is a plant that is chemically clean when it is designed and progressively fouled in the places where there is no flow to keep the surface clear.

The valve cavity is the worst of those places, and a ball valve has the largest one in the industry. A cavity of a few tens of cubic centimetres, filled with scaling liquor and isolated for six hours between strokes, can grow a two millimetre scale bridge across the seat annulus. Two millimetres is enough to hold the ball off its seat and enough to double the breakaway torque. Scale that spalls off the cavity wall becomes hard debris in the seat contact band, and a particle of calcium carbonate at 90 degrees C is soft enough to embed and hard enough to score the ball over a few thousand cycles.

Two design responses work. The first is a cavity that is continuously swept, either by a seat design that permits flow through the cavity or by a flush connection on the body. The second is to accept the cavity but give the seat enough travel and enough scraping geometry to break the bridge on the first few degrees of rotation, which in practice means a seat with a defined wiping edge rather than a flat pad. What does not work is a standard cavity with a standard seat and a purchasing argument about price.

7. Four hundred thousand cycles is a seat specification, not a sales claim

Adsorbent columns in a direct extraction plant are not batch vessels; they are sequencing machines. A loading step of 30 to 60 minutes, an elution step of similar length, and a wash and repressurisation between them means each column changes state between 24 and 48 times a day. Isolation valves on the column headers see that full count, because a lead-lag pair with block and bypass valving multiplies it.

Now put that against seat life. A PTFE seat in clean liquid service is commonly rated for 50,000 to 100,000 full cycles before the sealing load relaxes enough to show a leak, and that rating assumes a clean fluid at moderate temperature. Introduce 0.5 to 2 percent abrasive adsorbent fines, alternate that with acid, and the number moves down rather than up. A plant designed on a 30 minute cycle is therefore specifying a seat replacement interval of three to five years as a normal maintenance item, and a plant that has not written that into the operating budget will discover it as an unplanned shutdown in year four.

There are three ways to design around the count. Use a metal seated valve with a hard-faced ball where a Class C or Class D leak rate is tolerable, and accept that tight shutoff degrades rather than fails. Use a soft seat with a replaceable cartridge so that the replacement is a two-hour job with the valve in line. Or reduce the number of cycles per valve by using a double block and bleed arrangement that lets the isolation valve stay put while a smaller sequencing valve does the work. The third is the cheapest over twenty-five years and the least likely to be proposed, because it adds valves to the bill of materials.

8. What abrasive beads do to a ball and a seat

Adsorbents are not fine powders. Granular titanium oxide and aluminium-based layered double hydroxide beads run from about 0.3 to 0.8 millimetres, and they are hard enough to embed in a polymer seat and to abrade a stainless ball. In the loading direction the bed is fluidised or packed, and the valve upstream of the column sees a slurry at one to three metres per second with a few percent of solids. That is not a slurry valve duty in the mining sense, but it is enough to round the leading edge of a soft seat over tens of thousands of cycles.

Hard facing buys life here in a way that is easy to quantify and hard to argue with. A Stellite 6 overlay at 400 to 450 HV, or a tungsten carbide cobalt coating applied by high velocity oxy-fuel at 1,000 to 1,200 HV, moves the wear from the ball to the particle. The trade is leakage class: a metal seated valve seals on a lapped contact band and will hold ISO 5208 Rate A when new, then Rate B or C after a few thousand cycles, where a soft seat holds Rate A until it suddenly does not. For a column isolation valve the choice usually falls to soft seat with a cartridge, and the metal seat is reserved for the effluent and slurry ends of the train where a small leak is tolerable.

9. Torque, air supply, and the fail position nobody specifies

Actuator sizing on these valves usually fails for a reason the valve itself has nothing to do with. The nameplate torque is measured on a clean, unseated valve at ambient temperature, and the service torque is measured on a scaled valve full of beads at 60 degrees C with a seat that has taken a set. The ratio between the two is routinely 1.5 to 2.5. A ten inch Class 150 valve that turns at 900 newton metres when new can require 2,400 after eight weeks in a scaling stream, and an actuator sized at 1.2 times nameplate will stall in the middle of a stroke, which is worse than failing to open because the flow path is then partly open and partly blocked.

Sizing has to be done at the minimum air pressure the plant can actually deliver on a Friday afternoon, and at the minimum ambient temperature of the site, not at the compressor's rated pressure. A spring return actuator has its weakest point at the end of the return stroke, where the spring is least compressed, so the spring has to be selected against the seated breakout torque rather than the running torque.

The fail position is the clause that gets left blank most often and matters most. On an elution loop, the safe failure is closed, because an open valve drains acid into a brine header and the consequence is a chloride and pH excursion through the whole adsorption train that takes a day to rinse out. On a spent brine reinjection line the safe failure is also closed, for a different reason: an open valve with the injection pump stopped allows the well to backflow into the plant. On the column inlet, where a closed valve would dead-head a running feed pump, the safe failure is open and the protection belongs on the pump. Three valves on the same skid, three different answers, and a single "fail closed" note on the datasheet will be wrong for one of them.

10. The specification clauses that decide whether it works

A datasheet for this service that lists size, pressure class and body material has described perhaps a third of what determines the outcome. The clauses that carry the rest are these. The complete wetted path, including the cavity and the flange faces, must be lined or in the nominated alloy, with a written statement of how the lining is anchored and whether the shell is vented so that permeation cannot pressurise it. The seat and seal materials must be named with their temperature limits, not as "PTFE". Passivation must cite ASTM A967 and require a post-pickle iron test. The cavity relief arrangement must be drawn, because a self-relieving seat lets process fluid into the cavity on the high pressure side and that is a cross-contamination path between acid and brine. Leakage class must be stated to ISO 5208 with the test medium, since a Rate A claim tested on water says nothing about a gas-tight requirement on hydrogen chloride vapour. And the actuator torque basis must be stated as a multiple of breakaway, at a named air pressure and a named temperature.

Where hydrogen chloride vapour is present in the gland area, a fugitive emission class to ISO 15848-1 is worth writing, not for the environment but for the maintenance technician who has to break a flange on a line that weeps acid. Fire safety does not apply to the acid loop itself, but it does apply where solvent extraction with kerosene or a similar diluent is used downstream, and that loop should carry an API 607 qualification.

The works behind these valves builds to order rather than to stock: bores from 15 millimetres, which is the half inch size, up to 1,400 millimetres or 56 inch; bodies rated from ASME 150 up to 2500; and a temperature envelope that opens at minus 196 and closes at 650 degrees C. Each unit is proof tested before dispatch, at one and a half times cold working pressure on the shell and one point one times across the seat, and a priced quotation backed by a general arrangement drawing comes back inside three working hours once the datasheet is complete.

11. What a buyer should ask before the purchase order

Four questions separate a valve that will run for a decade from one that will be cut out in eighteen months. What is the pH range and the chloride level at the valve, not at the battery limit of the plant. What is the switching interval and therefore the design cycle count. What is the scaling salt and at what temperature does it come out of solution. And what is the impurity budget downstream of this valve, because that number, not the pressure class, sets the material.

The answers are usually available in the process package and rarely reach the valve supplier, because the datasheet is written by an instrument or piping engineer from a template that was last revised for a water plant. Ten minutes spent adding pH, chloride, cycle count and impurity limit to that template would prevent most of the failures described in this article, and would make the difference between a valve that is merely compliant and one that is correct.

12. Frequently Asked Questions

Can 316L be used in the acid elution loop?

No, at pH 1 to 2 and 50 to 60 degrees C with 50 to 150 grams per litre chloride. Hydrochloric acid is reducing, so the passive film is not repaired and the alloy dissolves uniformly rather than pitting locally. Published isocorrosion data puts 316L at several millimetres per year under those conditions, which is a wall life of months, and the iron released goes straight into a product with a 15 part per million limit. The workable choices are a PFA or PTFE lined body, a solid PVDF or PP body at low pressure class, or Hastelloy C-276 where metal is mandatory.

Why does a lined valve fail when the lining looks perfect?

Because the failure is permeation, not corrosion. Hydrochloric acid and water vapour pass slowly through a fluoropolymer liner, reach the steel shell and corrode it from behind, and the first evidence is a blister or a pinhole on the outside of the body rather than a leak into the bore. Vented body designs, bonded liners with a defined anchor pattern, and a shell material that can tolerate a little acid all address it. A liner that has been thermally cycled also creeps at the flange face, which is why lined valves need retorquing after the first heat-up.

How often should seats be replaced in a direct extraction plant?

On cycles, not on calendar time. A 30 minute switching interval gives 17,520 cycles a year, so a 50,000 to 100,000 cycle seat life is three to five years. The number drops with abrasive adsorbent fines and with scale. The practical answer is a replaceable seat cartridge and a planned replacement at 70 percent of the rated cycle count, which also gives the maintenance team a chance to inspect the ball for scoring before it becomes a leak.

Does the acid loop need fire-safe certification?

Not for the acid itself, which is not flammable. It does need it wherever a solvent extraction step with kerosene, or an alcohol based process, is installed downstream, because those streams are combustible and a seat that fails in a fire turns a small fire into a line rupture. The clause belongs on the solvent loop and on any valve separating it from an ignition source, and it is a common omission because the plant is classified as a chemical plant rather than an oil and gas facility.

Metal seat or soft seat for the column isolation valves?

Soft seat with a replaceable cartridge for the loading and elution isolation valves, because tight shutoff protects the lithium balance and the leak path between acid and brine. Metal seat with a hard faced ball for the effluent, spent brine and slurry ends of the train, where a Class C leak is acceptable and abrasion would destroy a polymer seat in a year. Mixing the two policies across the plant, rather than applying one rule, is usually the difference between a maintenance budget that holds and one that does not.