There has to be a better way to make titanium

On paper, titanium should be the world's #1 structural metal. There isn’t another metal with such great overall properties– incredible toughness and specific strength, light weight, exceptional corrosion resistance, and performance even at extreme temperatures. Titanium is also abundant, the 9th most common element in the earth's crust—cheap enough as ore to use in paint pigment and plastic filler^[1]. And making titanium metal should take only half as much energy as making aluminum^[2]. It should be cheap and ubiquitous.

Titanium was born for war—and in particular, to make war machines go as fast as possible. Its first major use was as the main structural material for the world's fastest planes (A-12 and SR-71) and submarines (Papa class). For more on the early wartime birth of the Ti industry check out Brian Potter’s excellent Story of Titanium, which inspired us to look more closely at Ti.

But in reality, titanium is a rounding error in the world metals market – produced 5000x less than iron and almost 200x less than aluminum. Half a century after the US and Soviet governments conjured a commercial titanium industry from scratch to feed their cold war machines, we’ve seen virtually no progress in making the metal cheap or abundant. Learning curves that manifested quickly for aluminum and stainless steel simply didn’t appear for Ti— its “idiot index”, the cost of Ti parts as a multiple of the cost of ore and embodied energy, remains >10x that of steel. At $25-$50/kg, titanium is just too expensive to be widely used.

The total lack of progress keeps titanium in a reverse-Goldilocks zone where it loses to steel and aluminum on cost, and to composites on weight^[3], and so is used only where its lightness and toughness are absolutely essential. Outside of aerospace, defense, corrosion-resistant process equipment, artificial joints and premium sporting goods, titanium is practically irrelevant.

‍

Why is titanium so expensive?

Here’s how we get Ti today:

• We mine TiO₂ as ilmenite ore, which is widespread and fairly concentrated at 40-65% TiO₂. (~$1/kg Ti)
• The ilmenite is purified to synthetic rutile (90-95% TiO₂) (~$2/kg Ti) then carbo-chlorinated to TiCl₄ which can be distilled for very high purities. ($3-4/kg Ti)
• Next comes the main act: the Kroll process, reducing TiCl₄ with magnesium metal TiCl₄+2Mg -> 2MgCl₂ + Ti. This is a touchy operation in which TiCl₄ is meticulously metered into molten magnesium – too fast and the heat liberated in the reaction would vaporize the Mg metal. Since we have to remove the Mg and MgCl as gases, and the process happens well under Ti’s melting temperature, the metal we get initially is in porous and hard-to-use “sponge” form --full of bubbles like lava rock (~$6-8/kg Ti).
• The porous sponge is crushed, ground, mixed with alloying metals (useful Ti most often shows up as Ti₉₀Al₆V₄), pressed, welded into 10-15 ton “consumable electrodes”, and homogenized in a vacuum arc remelter (VAR). (~$15/kg Ti)
• The ingots are then whittled down to saleable parts with sundry processes and techniques, which themselves are painstaking for many of the same reasons that Ti is so awesome as a structural metal – it’s hard enough to destroy tooling, forms a thick oxide film in O₂, melts only at a staggering 1670ºC etc. … (~$25-60/kg Ti)

The cost of titanium roughly doubles with each of these steps^[4] – and there are lots of steps!

Cost and embodied energy for titanium mill products and intermediaries. In the refining stage Ti ore is cleaned of metallic (Fe, Al, Mg, Ca, V etc.) and semi-metallic (Si) impurities. The reduction stage provides the 4 electrons required to produce titanium metal from TiCl₄. The homogenization removes porosity left over from the reduction stage to reach full-strength Ti. The shaping stage increases the surface area to volume ratio of the ingot into shapes which can be machined into useful parts.

Why are there so many steps here? Part of it is that titanium is both electropositive and fairly refractory—you need the uncomfortable combination of very high temperatures and a very reducing environment to get it into a useful form. But the bigger reason is that titanium is reactive—ferociously so. Molten Ti dissolves nearly every other metal. It readily forms oxides, nitrides, hydrides, and carbides. Titanium powders will ignite in carbon dioxide, and even in helium with as little as 9% oxygen. Sometimes this reactivity can be annoying, requiring an inert environment, or causing embrittlement or loss of material. Other times it can be deadly^[5].

(But wait – isn’t titanium meant to be corrosion resistant? Shouldn’t it behave more like a staid transition metal than an electron-shooter like sodium? Like aluminum, titanium forms a passivating oxide on its surface that imparts its famous corrosion resistance, until about 600ºC. At that point, the titanium below the oxide skin begins to draw in oxygen at a meaningful rate, embrittling the part and exposing it to attack from whatever’s around.)

Titanium’s reactivity is especially punishing when you try to process it into useful shapes, which requires high temperatures where oxygen and anything else around gets in and ruins it. Consider for example just the ingot breakdown step. In order to make the ingots more pliable, producers heat them up to high temperatures and forge them in open air. As the surface area of the ingot increases, so does oxygen pickup, forming a brittle hide you have to shave off. That’s up to 50% of the metal you made, discarded useless on the mill room floor—just to forge an ingot! Frictions like this add up at every step on the value chain. Since Ti is only transformed incrementally at each step in this process, it all adds up to significant capital, time, labor and material loss tied into each step.

‍

It's not a matter of optimization or learning rates

As bad as things seem, there’s good reason to believe that the way we make titanium today is pretty close to its cost floor.

The TiCl₄ precursor to Kroll is a high-volume commodity chemical (>7 MMT/yr) that costs ~$4/kg Ti. Making TiCl₄ is a mature process that’s substantially optimized already. Kroll also requires magnesium metal which is recycled on-site for ~$1.2/kg Ti^[6] -- only a small multiple of the energy costs involved. Throw in a 15% profit margin, and you have arrived at the VAT excluded price for the cheapest Ti sponge quoted on the Shanghai Metal Market: ~$6/kg. So Kroll costs hardly more than its material inputs. 60 years of Ti production have whittled most of the inefficiencies away. The real cost and complexity comes later, in trying to work Ti sponge into useful parts.

Clearly, what titanium needs is reimagining of the whole process that reduces the number of steps in the value-add chain. Since cost roughly doubles with each step, eliminating one or two steps could easily save more than improving each one by 10%.

This fact is widely recognized—and has been for half a century. As Brian Potter writes in his excellent history of the Ti industry, even as the Kroll process was first scaling in the 1950s, newer and better ways of making titanium were widely expected to materialize. William Kroll himself expected that an electrolytic process similar to Hall-Heroult for aluminum would come to replace his method.

But the sad truth is that there have been many serious and well-funded attempts to reinvent Ti production over the last 50 years, and all have failed to meaningfully displace the Kroll-centric approach we use now.  Let’s look through what has been tried, and see what we can learn:

‍

The cruel lessons from 60 years of failing to reinvent titanium

People have tried all sorts of methods: electrolysis, plasma, alternative reductants, additive manufacturing. None ever produced full-density Ti cheaper than Kroll:

1960s – Dow-Howmet/TIMET/RMI electrolysis – Electrolysis of TiCl₄ to porous Ti.

1994 – Hydrogen plasma reduction – Thermal plasma for reduction of TiCl₄ to Ti powder.

2000 – Armstrong process – Na reduction of TiCl₄ to Ti powder.

2000 – Farthing-Fray-Chen (FFC) Cambridge & Ono-Suzuki – Electrolysis of TiO₂ to produce porous Ti.

2003 – Cardarelli-Rio Tinto electrolysis – Direct electrolysis of molten TiO₂ to molten Ti.

2003 – Ginatta electrolysis – Electrolysis of TiCl₄ in fluoride/chloride salt to make molten Ti.

2004 – MER Corp electrolysis – Electrolysis of Ti oxycarbide to porous Ti.

2012 – ADMA process – Hybrid H₂/Mg reduction of TiCl₄ to TiH₂ powder.

2013 – Hydrogen Assisted Magnesiothermic Reduction (HAMR) – Hybrid H₂/Mg reduction of TiO₂ to Ti powder.

Here’s our longer review of all the attempts to reinvent Ti, and the lessons we’ve drawn from their failure.

We take away four main lessons:

• Start with TiCl₄. Trying to start from TiO₂ makes things much harder, and can’t meaningfully reduce costs.
• Any new process will have to make dense Ti directly if it’s going to be useful. Kroll already makes porous Ti efficiently. The problem is that porous Ti simply isn’t a valuable product.
• Beware additive techniques. These are the most promising of the alternative approaches, but signs indicate that they won’t reach the scales we’d need to revolutionize Ti.
• Beware electrolysis. It’s the most frequently attempted and least successful of the alternative approaches. Ti has several cruel chemical features that make it especially challenging to manipulate in solution.

1) Trying to start with TiO₂ makes things much harder

Most of the attempts reinvent Ti since the early 2000s have tried to substitute TiO₂ for Kroll’s TiCl₄ starting material. That's a way to save a few bucks on feedstock costs-- and the only possible route to getting Ti as cheap as Al or Mg. But unfortunately, today none of those projects produce titanium at commercially relevant tonnages. We think that much of the problem has to do with the hassle of purifying TiO₂.

The vast majority of titanium used on earth is sold not as Ti metal, but as white TiO₂ pigment. TiO₂’s whiteness depends strongly on its purity. So the TiO₂ pigment industry has come up with an excellent way of making it as pure as possible: carbo-chlorinating synthetic rutile, fractionally distilling the various volatile metal chlorides until just the TiCl₄ remains, and then burning the purified TiCl₄ in oxygen to get back to pure TiO₂.

As complicated as this sounds, the chloride distillation route is just about as good a way as is thermodynamically possible for purifying a metal as chemically “sticky” as Ti. Since Ti metal has to meet strict purity standards (~0.1wt% for most elements), trying to make metal directly from TiO₂ saddles you with a serious purity problem right at the start.

TiCl₄ also serves a valuable function in keeping oxygen out of your reaction mixture. Oxygen is fiendishly soluble in Ti metal—so much so that its presence makes metallothermic reduction almost impossible.

The thermodynamic nightmare of trying to separate Ti from O in a reaction mixture. Commercially pure titanium requires <1000 ppm O which is accessible only by TiO₂ reduction with calcium and yttrium (the lines for magnesium and lanthanum appear to intersect that of Ti-1000 ppm O at a low enough temperature, but in practice those reactions are starved for activation energy and bogged down by solid-state diffusion)

Just say no to TiO₂. TiO₂ that is pure enough to be a feedstock for Ti metal is more expensive than TiCl₄ anyway, and makes the reduction of Ti vastly harder. We want to delete steps, but not at the cost of pure titanium.

2) Porous Ti is not a valuable product

Most of the inefficiencies in the Kroll process have been whittled away over the last 60 years. At this point, Ti sponge sells at a price that is very nearly the sum of the cost of TiCl₄, a high-volume commodity chemical, and the cost of producing magnesium with a well tamed electrolytic process. What exactly would you improve on?

Imagine you did invent a magical new process for making porous Ti sponge. Say it’s half the capital cost of Kroll and a quarter of the energy cost—just above the thermodynamic minimum for making Ti. You buy the cheapest pure feedstock that gets you the purity you need, TiCl₄ at ~$4/kg. If you want a 10% profit margin, that leaves you only ~$2.3/kg Ti to finance, operate, and maintain your process if you’re going to compete with ~$7/kg Kroll sponge. Kick in $1.35/kg Ti for capex (half of Kroll’s capex of $27/kg Ti per year amortized at 8% for 20 years)^[7], $0.5/kg Ti for electricity costs, and ~$0.5/kg Ti for labor and maintenance—and even though your process is magical, it’ll still be borderline unfinanceable if all you can sell is Kroll-equivalent sponge!

Titanium sponge just isn’t where the value is. Saving a few tens of cents on $7/kg Ti that will later be turned into mill products that cost as much as $50/kg Ti won’t move the needle.

3) Additive manufacturing is unlikely to give titanium the volume it needs

So if sponge isn’t where the value is, maybe we need an all-new way to turn sponge Ti into finished parts?

A lot of people have seen it this way in recent years, and so there’s been a lot of excitement about additive manufacturing for Ti parts. It makes sense in some ways. After all, subtractive manufacturing sucks for titanium – its hardness and low thermal conductivity chew through tooling^[8], and its reactivity costs you material in high temp processing. Both Ti and additive manufacturing are mostly used at the top of the market where expense hardly matters. And a lot of the newfangled ways you might make Ti give you spongy or powdered metal that needs additive post processing to get to full density anyway.

But while additive manufacturing with Ti powder does some specific things very well, one thing it won’t do is make Ti cheaper than it is today. Additive manufacturing tools are generally less capital intensive but have lower throughput, while conventional methods get economies of vertical scale. This has played out in the market for a few years now and cost analyses show that there’s a breakeven manufacturing volume above which conventional Ti production is more cost-effective. The exact value varies but this paper suggests that conventional Ti casting beats additive manufacturing at scales above a few hundred parts (valve bodies made with hot isostatic pressing vs. conventional forming), and this analysis of a landing gear reports that high pressure die casting is more cost effective than selective laser sintering above ~40 parts produced.

There are also lingering concerns that the main approaches to additive Ti leave micro-pores in the final metal product that concentrate stress and make weaker overall parts, especially under cyclic loading. Fear of microporosity pushes additive manufacturing to use high-cost ($70-200/kg) ultra-spherical powders in order to minimize pore volumes^[9]. But concerns over part strength have persisted especially among the usual sectors for early adoption—defense, aerospace, and medical equipment– where failures can be catastrophic and the cost for Ti is borne easily.

Additive manufacturing does shine for low production volumes of highly complicated parts that conventional manufacturing struggles to make at all (see SpaceX’s SuperDraco). But an abundant Ti future needs a new method that can make Ti at high volumes and low cost and full strength, and that method probably won’t be additive manufacturing.

There are some long-lasting concerns about the durability of additively manufactured Ti parts, which can contain thousands of stress-concentrating voids per cc of material. This graph shows applied stress vs. number of cycles for parts made of wrought Ti-64 alloy versus for hot isostatically pressed parts made with Ti-64 powder. Adapted from Cheng et al. (2022) and Rao and Stanford (2021).

4) Electrolysis isn’t out of the question but it comes with severe challenges

Electrolysis seems like a logical approach to making titanium. After all, the development of an electrolytic process for aluminum took it from a precious metal used only in jewelry to a ubiquitous structural material in just a few decades. Could it unlock similar cost-savings for titanium?

Probably not, we think:

Ti has too many valence states in solution. Titanium, unlike aluminum, adopts multiple valence states which engage in a phenomenon called redox cycling in an electrolyzer. Ti²⁺ can flow across the cell in one direction, drop off an electron, and flow backwards as Ti³⁺. So current flows through the cell for no reason, generating no titanium metal and producing nothing but joule heat! Electrolytic magnesium produced for the Kroll process at ~30% voltage efficiency and >90% current efficiency is just a more efficient way of packing electrons into a titanium feedstock.

Ti melts at too high a temperature, but lowering the temperature makes things worse. Part of the magic of the Hall-Heroult process for aluminum is cryolite, which lowers the melting temperature of Al₂O₃. Since Al metal also has a relatively low melting temperature (660ºC), a Hall-Heroult cell can operate at reasonable temperatures and still pour molten aluminum into shapes that can be sold directly to customers.

If only that were possible for a Ti electrolyzer, it could delete the costly post-processing steps in making Ti parts, and make finding compatible electrolyte and electrolyzer materials that much easier. But Ti’s 1670ºC melting point makes that impossible.

You could make building your electrolyzer much easier by running it below titanium’s  melting point. Some people have tried this. But at meaningful current densities, the product will then be dendritic titanium – tree-like metal crystals embedded in frozen electrolyte. What would it take to make this titanium useful? First, you must clean the titanium of the salt and then melt it down to make homogenous titanium stock – the same steps that make Kroll’s sponge titanium expensive. A non-starter.

Overall, electrolysis is the most-frequently attempted and least successful of the alternative approaches to Ti. People will keep trying—but we will be surprised if anyone gets it to work.

‍

Tempering our expectations

People who love titanium tend to compare it to steel and aluminum since Ti’s base properties are so much better. They’ll also point out that the ore is cheap and abundant and that the metal costs vastly more than it should, a staggering 60x over its theoretical cost floor (the cost of ore plus energy). That kind of argument makes it easy to convince yourself that Ti should be a main structural material if not the main structural material of the future—if only we could find a better way to make it.

But it’s worth offering a moderating opinion, both on how cheap titanium can really get, and on how useful it is.

Why you can’t say that titanium beats aluminum or steel categorically. Consider the bicycles used in the Tour de France, ultra-premium machines made of the best possible materials. From 1904 to 1994 the Tour de France was won on steel bikes. From 1995 to 1998 it was won on aluminum bikes. And then from 1999 until now it’s been won exclusively on carbon fiber bikes. Why was Ti passed over completely?

Titanium’s huge advantages in specific strength over aluminum and steel are muted when you look at the structural metrics that actually matter. The bar plots compare these metrics for an aluminum alloy (Al 2024), a titanium alloy (Ti-64), stainless steel (SS 316), and a mild steel (A36), all metrics normalized to Ti-64. For cost parity in a tie (tension member) titanium ties need to be <2X the cost of aluminum, <10X the cost of stainless steel, and <4X the cost of mild steel. The multipliers are even lower for panels and beams.

Strength-to-weight ratio isn’t everything. Larger-diameter tubes in Al bike frames are stiffer than Ti and can even be stronger—not based on material properties, but based on geometry. In a bending tube in a bike frame, the critical buckling load scales with the cube of the outer diameter, but only linearly with stiffness. If you want a stiff and strong frame, or one that’s easy to shape into aerodynamic forms, aluminum with its lower density and better formability can end up as the winning choice.

This type of dynamic repeats all across industry. Properties like density end up being just as important as specific strength, and lesser metals that are easier to form into the right geometry end up competing with and even beating titanium.

Why titanium will always be pricier than Al or steel. Hopefully we’ve landed this point by now: Ti is extremely hard to work with. The reasons why are written in its atomic structure. It’s semi-refractory and a sponge for nearly everything around it. The purity of the feedstock and reducing agent, inertness of the reactor volumes, and metal handling at elevated temperatures are crucial aspects that will increase capital and labor costs and decrease material utilization, no matter how clever your process is. Unlike titanium, aluminum and steel smelters operate in atmospheric conditions, the molten metals can be poured in open air to form shapes that brokers can store on shelves, and they are both compatible with common high temperature materials like alumina, silica, and graphite^[10]. None of that works for titanium.

‍

Where does this all leave us?

Back when it was young and promising, ARPA-E ran a program called METALS which sought to reinvent many of the light metals for the modern era. They estimated substitutional elasticities for different structural metals to come up with some cost targets, how cheap these metals would have to get to really change the world. For Ti it was $10/kg at the ingot level. That’s the cost where they thought Ti could start to meaningfully displace steel and aluminum parts.

Is that possible? Despite all the harsh lessons over the last few pages, we think so.

All this reflection on past failed attempts to reinvent titanium production makes us think that the straightest line to cheaper titanium is to delete steps wherever possible and to not shy away from energy intensity. Paying extra energy for purification and reduction could get you to fewer total steps and direct castability. Fewer steps is key since titanium is a refractory metal that can soak in nearly everything around it, leaving it worse off. Every step wastes material and makes it harder to meet purity standards.

At this intermediate stage in our brainstorm on the topic, what looks best to us is a single-step metallothermic reduction of titanium chloride that produces molten titanium directly. That’s like a more energy-intense version of Kroll. With 150% of Kroll’s capex and opex you could make ingots that hit the ARPA-E target. Better yet, using more energy in the reduction step could leave you with molten Ti so you could skip the ingot step entirely, getting you customer-ready bars and sheets for less than ingot prices. If somebody could do that, we think they could beat Kroll Ti by wide margins.

It’d be hugely energy intensive—but that could be OK. People are already making metals off-grid with dirt cheap electricity. We’ve said before why we shouldn’t be shy of energy intensity in materials production— falling energy costs have always been a main driver for lower material costs, and energy-intensity has always gone hand in hand with building the world out of much cooler materials ^[11].

High energy intensity, direct castability, deleting steps. If you’re also thinking about titanium along similar lines, please get in touch.