Cobalt blues: The Li-ion supply chain and the shackles it creates.

In recent weeks, two articles in the Washington Post laid bare the sorry state of the supply chain for Li-ion batteries; batteries that have made our phones into a device we use for everything (except calling people!); is changing the way we drive; and promises to change the way we generate and use energy. 

The first one on cobalt mining in Congo describes dangerous conditions, child labor, and exposure to toxic metals.  Cobalt is essential for cathodes for Li-ion batteries.  The second article on graphite, used as an anode in Li-ion batteries, digs into the pollution, environmental destruction, and health effects from mines in China.  Both stories are very well researched; connecting the dots from the mines, to the supply chain, all the way to the companies that use them in smart phones and electric cars.  The articles are also depressing.  They reveal a side of these devices none of us want to see.  It is like watching Food Inc.

One question that I have gotten since these articles came out is:  Can we move away from cobalt and natural graphite for Li-ion batteries?  This blog post delves into this topic. 

Why did we start using cobalt anyway?

Marca Doeff’s blog post does a fantastic job of walking us through the history of Li-ion cathode materials.  Lithium cobalt oxide, which John Goodenough discovered, started the whole Li-ion revolution.  Cobalt oxide was (and is) expensive, so variations have been found that have no cobalt in them. 

Lithium manganese oxide came along soon after, was eventually commercialized, and reached the market for power tools and in plug-in electric cars. Manganese oxide is significantly safer than cobalt oxide, but it also has far less energy density.  And consumer electronics and electric cars need highly energy dense materials, so manganese oxide is not that useful for these applications. 

Another variant is lithium iron phosphate, which has a complicated IP history.  Many of you probably heard of this chemistry when A123 first commercialized this for power tools.  It has since become pretty popular in Asia for transportation applications.  The chemistry is very safe and has fantastic cycle life.  So there is hope that we can use it as the stationary storage market evolves.  But the wonderful safety and cycle life comes at the expensive of energy density, which is pretty low compared to cobalt oxide.  Ergo, limited use for high-energy applications.

In the late 90’s the nickel variant of cobalt oxide, lithium nickel oxide appeared to be gaining traction in the consumer electronics battery world.  But nickel oxide has some safety issues and this was a cause for concern.  Instead, what became more successful were a class of cathodes where the cobalt content was lowered instead of being eliminated.  These include nickel-cobalt-aluminum (NCA) and nickel-manganese-cobalt (NMC).  There was a period in the early/mid-2000’s when these variants seemed poised to displace cobalt oxide completely, partially driven by, what at that time appeared to be, the limit of cobalt oxide. 

And then, cobalt oxide got better. 

It was long thought that cobalt-based cathodes could not be used beyond a cell voltage of 4.2 V because of reactions between the cathode and the electrolyte that lead to degradation.  Today, with surface coatings to isolate the cathode from the electrolyte, combined with larger particles, cells charge to 4.4 V or more.  Coupled with the ability to achieve high tap densities (defining tap density is a blog post in itself!), cobalt oxide has again become the highest energy density battery available.  For smartphones, laptops etc., where the battery cost is relatively small, the energy density advantage is key.  So cobalt has come back to rule the Li-ion world.

But cobalt is still expensive.  So for larger batteries such as in electric cars and for stationary storage, the lower cobalt content materials such as NCA and NMC are used.  And it is not obvious to me that the high-voltage cobalt oxide cells (operating at 4.4 V) are going to be long-life batteries.  These are probably great for the 2-year change cycle for phones; not the 10-year change cycle for cars.

Although, to digress, I have a 2 year old phone that has only lost 7% of its capacity.  How, you ask?  Because I follow these wonderful battery rules.

I summarize all this in the figure below, including the options for the anode and the electrolyte (well... there really aren't many options for electrolytes!)

 Can we move away from cobalt altogether?

Not clear.  As I said before, we do have options that have no cobalt, but these options are not very high in energy density, except for the nickel oxide material.  The nickel oxide material works well, although the synthesis requires some care.  However the safety of the material is questionable.  And as the recent incidents with the Note 7 have highlighted, one should be designing safety into every component in the battery. 

In the meantime, lowering the cobalt content is probably the logical pathway.  There is a trend in the research stage to minimize the cobalt content, in some cases to as much as 10% of the transition metal content; without scarifying energy density. Some of these materials, referred to as the lithium-rich, manganese-rich materials, show enormous promise.  And they offer the hope of eliminating cobalt.  But they also appear to have some fundamental issues that need to be solved. Work continues, at universities, national labs, and in companies to bring these to market.

In addition, one can also imagine playing the same tricks on the NMC formulation that helped cobalt oxide operate at higher voltages.  This will help increase the energy density of this material, beyond what is possible with cobalt oxide.  But these are all still in the research stage. 

In batteries, the time from lab breakthrough (real ones, not the ones that are alluded to here) to market impact takes a decade or more.  So, we may be stuck with cobalt oxide for a while.  But there is hope that we can move away from it with less cobalt and maybe, one day, with no cobalt.

But, something tells me that even if that happens, we will find out that children can still be exploited and the environment destroyed in the search for the raw materials.  The issue here is probably not cobalt.  It is something far beyond what this blog was meant to address.  

Case in point is the graphite part of the Washington Post story.  Next week I will list our options for moving away from graphite anodes. 

Venkat

The Hero with four faces: Part 1

A few years ago, when I visited a battery company with a few colleagues, I saw data from the company that was pretty impressive.  Slide 1 showed the rate capability of the battery, which was better than the state-of-the-art; Slide 2 was a calculation of the energy density, which was better than the baseline; slide 3 was cycle life and this too looked impressive; slide 4 was…

You get the drill. 

Suitably impressed, I came out wondering what I needed to polish in my resume to land a job at the company before they went public… when my travelling companion wondered aloud if each slide was from a different battery!

 My colleague had a point.  Most (not all) battery chemistries can be made to perform well for a particular metric.  The trick is getting all the metrics to work out for the same battery. 

Make the electrodes thin to get power; make them thick to increase energy (but at the loss of power).  Increase the voltage to increase the energy, but at the loss of cycle life.  The endless games one can play.  This dependence of battery design, cycling conditions, voltage of operation, etc. on performance is the reason battery companies have been getting away with obfuscation. 

But is it really obfuscation?  A proof-of-concept can be a useful learning tool.  If going to high voltages yields more energy but kills the cycle life, while we know that the voltages being accessed should be possible, then it gives us hope that the problem is not fundamental and that a solution exists.  Given time, maybe we will find it. 

Then again, we have been looking for a cure for baldness for ages.  There is proof that baldness is not fundamental (after all, some are oh so lucky), but that does not mean we will find the answer to that problem either.    

But there is merit in learning about what the best possible system is and understanding why it is the best.  To this end, I thought I would list out the best-known (at least to me) battery chemistry for each of four metrics of importance:  energy, charge time, life, and cost.  Each would be a “Hero” battery; a term borrowed from other technology areas to denote a proof-on-concept that something amazing is possible for that metric.  While each metric has a different “Hero” battery there are probably lessons we can learn from them.  These lessons, I summarize in the end.

The energy-density Hero:  All seven readers of my blog know that Li-ion batteries are the highest energy density secondary batteries one can buy, with energy density in the range of 250 Wh/kg.  But the Hero is actually the Li-thionyl chloride battery, which has an energy density of 550 Wh/kg and is, more importantly, commercially available.  The catch is that it is a primary battery (i.e., not rechargeable). 

What gives?

Typical Li-ion cells have a graphite anode.  Moving from graphite to Li metal (which holds ten times the charge and results in a slightly higher voltage) bumps the energy by 50% without changing anything.  On top of that, the thionly chloride cathode has a capacity of 450 mAh/g compared to 180 mAh/g for typical Li-ion cathodes.  Combine the two and you get something much better than Li-ion.

Imagine a battery with 2x the energy density.  Maybe we double the range of a Nissan Leaf, making it a car that is actually useful (joking, joking, all you Leaf lovers) for the same cost.  Or cut the cost of the Tesla battery by half without sacrificing range, bringing it tantalizing closer to the point where everyone can continue to not be able to afford it. 

Unfortunately, all these will not come to pass because the Li anode does not recharge gracefully while the thionly chloride cathode is not at all rechargeable.  If you want to know why, you have to read my blog post titled “A Brief History of Batteries- Part 1” and “A Brief History of Batteries- Part 2”.  Frankly, everything you need to know about batteries is probably hidden somewhere in these pages. 

To make it rechargeable, you just need to change the anode, the cathode, and the electrolyte (i.e., all the three components that make a battery).  When you do that, the battery now cycles a lot better, but at the loss of capacity. 

So is 550 Wh/kg the ceiling?  No.  There is a lab demonstration of a 750 Wh/kg Li-air primary battery.  This would be a record for energy density.  But as I have alluded to in the past, a one-off lab demonstration does not a product make.  So, for now, we shall call 550 Wh/kg the Hero for energy density.  As an aside, the Li-air battery also does not cycle. 

Reversibility restricts us to certain materials; a constraint when removed, allows high energy density batteries to be made. 

An interesting question to ponder: Is it possible to get batteries that are rechargeable and at least as high in energy as these Hero’s?  

The fast-charge Hero:  One of my funnier blog post was titled  “I’ll be back…in 8 hours”.  That post had nothing to do with charging times (then again, this blog is like Seinfeld; its about nothing, as far as I can tell) but the title articulates the problem:  We tend to want to charge our batteries very slowly.

Charge them too fast and we have unwanted side reactions (lithium plating and electrolyte breakdown) that can degrade the battery.  1h charge is doable, 30 mins makes it degrade a bit; 10 mins a lot more; 1 min would basically kill the battery (and kill you from the fire that is creates). 

This is true for most batteries, but it is not fundamental.  Meaning, there is no law that says that we cannot charge a battery fast.  One can design the battery for fast charge.  But the compromise is loss of energy. 

How fast do we really need to charge?  For an electric car, a really smart person (who shall not be named) told me that we needed to charge within the time frame of a restroom break.  Another smart person told me that we needed the battery to charge and discharge fully within the time frame of clouds covering a solar panel.  After a few experiments timing myself on restroom breaks and watching clouds move (I plan to watch grass grow next) we can approximate the charging time needs as 5 minutes.  What can I say: I was having an Austin-Powers-just-got-out-of-hibernation moment. 

The question of the grid actually handling this kind of electricity load is a whole other area of debate, but let us focus on the battery for a second.

Electrochemical capacitors can easily charge this fast.  But they have no energy.  Question is: Is there a battery chemistry that can mimic a capacitor’s charge rate?

There was one system that was kind-of-sort of commercialized that I would consider the Hero in this regard.  This was the Toshiba Super Charge battery, which is rated to charge to 80% capacity in 6 minutes and more than 95% in 10 mins!

Here the anode (lithium titanate) operates at a higher voltage than the typically used graphite anode. This helps because the potential of the anode is far away from the lithium plating potential.  This makes it much easier to charge fast and not worry about plating lithium, shorting, and the ensuing degradation and possible fires.   

But the downside is that the higher anode voltage decreases the overall cell voltage, which in-turn deceases the energy.  The battery has a third of the energy density of a typical Li-ion cell (so a third the driving range).  At 90 Wh/kg and 177 Wh/l at the cell level, it is far, far lower than most Li-ion batteries.  Cost of these devices scales with the energy:  this battery will probably cost three times a typical Li-ion cell!

So… an ideal EV battery for the super-rich-with-overactive-bladder demographic?  

Question is: can we get both high energy and fast charge?

If we want to get there I believe we cannot use thick porous electrodes.   While they are great to spread the current, it seems impossible to get away from the electrolyte losses of these highly-resistive organic electrolytes.  Which means that we cannot have thick bulky current collectors and separators.  But then how do we collect the current (especially if the currents are high, which seems likely with fast charge batteries)?

Should we move to electrodes that are not porous?  This limits the useable energy, unless we use electrodes that undergo deposition. One can continue to deposit the metal on top of itself and so “build” capacity, without the added losses from the porous structure.  This may be an avenue. 

My suspicion is that even if we find ways for the electrodes to accept the charge at that rate, the electrolytes in lithium-based batteries will not have the ability to move ions from one side to the other.  Unless we move toward much thinner separators.   This topic requires some careful thinking.

But let us revisit the question: how fast do we really need to charge? 

If we can ensure we have a 400-mile range battery, this should translate to an approximately 5-6 hour driving time (at 70-80 mph).  Then we may be willing to wait a half hour to charge the battery as we make a beeline to our favorite artery-clogging fast food joint. 

For the grid, if we can use the battery to not just take care of the intermittency, but also to time-shift from peak to off-peak, we can size the battery for the time shift and use the (big) battery for handling the small 5-min intermittency.  Each 5-min charge and discharge would only require the battery to swing by a few percent: easily possible with most systems. 

The catch: both of these would require us to pay for the bigger battery!  But hey, batteries are getting so cheap, and companies are going to give us money when they hand us the battery anyway. 

The cycle/calendar-life Hero:  I know what you are thinking: there is no such thing as a cycle-life Hero!  After all, every battery we own seems to last all of 1-2 years before they crap out.   You must be thinking that our Hero must be a 3-year life battery.

What if I told you there are batteries that last 20, even 30+ years and they cycle 20,000 times (no error there, really meant to have four zero’s)?  And what if I told you that these batteries are not the hybrid car batteries that cycle 3-5% per cycle but are cycled deep, greater than 50% of the capacity per cycle?  And that they are (at least they were) being used day in and day out? 

Imagine batteries that last as long as a solar panel.  Imagine being able to cycle them once a day and make them last those full 20 years with no maintenance.

Intrigued?  Come back next week and you shall learn more.

 Venkat

p.s. the title of this post is inspired by the Joseph Campbell book “The hero with a thousand faces.”  Campbell was trying to point out that all the world’s myths, across religions, shared a Hero figure with similar characteristics.  We will see next week that in batteries there is no one Hero.  We basically have four Hero’s with four faces.  

If your cat has nine lives, shouldn't your battery have at least two?

Let’s see if you can answer this pop quiz: How much capacity fade can you tolerate in your battery before you consider it “dead”?

The answer (as usual) depends on the application.

If you have a laptop and it is anything like mine, then “dead” is when you get 10 mins of run time and you decide to bite the bullet and get a new battery. With some laptops (the one that comes to mind is made by a company whose name starts with a D and ends with a L and has four letters) getting one year from your battery is considered very good.

But laptop batteries are $50. And most customers are probably going to go back and get their next laptop based on price anyway, so there is no long-term effect of giving someone a bad battery. And if your laptop battery dies you can use it plugged in (i.e., as a desktop).

But if you had an EV which run’s 200 miles and if in, say, 3 years you start to get 160 miles, you can be confident that you will sitting in the dealership demanding that they change the battery for you. At $10-$30k a pop you are not about to pay for another battery.

And running your car with a long cord connected to an outlet can be a bit problematic.

But, as definitions go, most manufacturers will consider a battery to be at the end-of-life if it has lost 20-25% of its capacity. In some applications (e.g., HEVs) 20% loss of power is considered end-of-life. The less expensive the battery, the higher the loss that is tolerated and vice versa.

What this means is that when your EV battery is considered dead, it still has 80% of its capacity left.

Despite the fact that you are reading this blog, I’m going to go out on a limb here and assume that you are a smart person. Right about now, you must be thinking to yourself that it is stupid to call something dead when there is 80% of it left.

Do you get new brakes or tires when they have depleted by 20%?

No! You wait till your tires are bald and you skid into your auto repair shop and stop the car by crashing into something soft.

By now you are, hopefully, having this epiphany that there may be a hidden business idea here. One that involves taking these “dead” EV/PHEV batteries and using it someplace else.

Maybe ship it to some other country where they may not be as picky about the number of miles you can get?

Or maybe if you can find an application where energy density were not important, you can repurpose these batteries to sell it for these applications.

While you chew on that, let me change subjects.

Did you know that as states like California decide to enact renewable standards, utilities companies are going to have a problem? California promises that by 2020, 33% of the state’s electricity will come from renewable sources. So let’s say that to satisfy the mandate, the utilities decide to build a giant solar farm in the middle-of-nowhere (Tracy?). But even in the middle-of-nowhere the sun does not shine at night.

So enter storage as the savior. You charge the battery during the day and discharge it to power the cities at night. Transmission allowed us to space-shift the electricity (generate in point A, use it at Point B); storage allows us to time-shift. Its the DVR (or TiVo) of the grid.

And how much electricity are we talking about storing? California uses electricity in the range of giga watt hours (GWh). So a lot!

Some of you may have heard of AB2514. Its a legislation in the state of California that requires utilities to incorporate storage in the grid. The utilities will need to store 2.25% of the daytime peak power by 2014 and 5% by 2020. This is probably the first explicit mandate calling for storage to be a part of the grid.

For all you veteran battery folks who thought the community did not know how to lobby, take heart. We may be nowhere close to what the hydrogen guys have done, but it is a start!

But let’s back to the storage issue.

So storage is the DVR of the grid. And because your DVR is not moving, it can be big (and ugly if it is like my DVR).

But DVRs cost extra money and they die every few years. Batteries are a lot more like your DVR than you can imagine!

Meaning, the energy density of the battery is not that important for this application. The great thing about middle-of-nowhere is that there is a lot of it out there. What is important, however, is the cost.

Hence the interest in taking “used” vehicle batteries and using it in grid applications. A second life, if you will.

Think of the possibilities: If your battery lasts say 15 years instead of 7 years, then you just doubled the time to amortize the upfront cost. You can ask the utilities to buy the car batteries and lease it to the consumer. Once the battery is “dead”, the utility moves it to grid applications and starts its second life. The consumer does not have to worry about paying $15,000 extra for the battery or worry about it dying on them. Someone else owns it; all I need to worry about is making sure I don’t get into a crash and ruin my warranty.

There are other themes on this business plan, but you get the idea.

All this is sounding so great that I’m contemplating sending an email to my bosses telling them what I really think about my job.

But wait a second. This whole blog survives only because we hate our energy storage devices. How many examples can we think of for our batteries lasted more than a few years? My car battery, which died after 7 years, has been the only battery that has done me proud. And I was lucky to get anything close to that!

We all spend an inordinate amount of time babying our batteries and asking how we can extend the life, and here I’m claiming that we can get a second life from our batteries. There must be a reason why my post “pull the plug, your battery will thank you” has the highest hit among all the posts.

What exactly are we missing?

Turns out that we are not missing anything. The second life concept is being pushed by business-types. For a biz. dev. guy this makes perfect sense. But the real question is: Does this make sense for the tech guy in the back actually doing the testing?

Let’s look at this in detail.

If you plot the capacity of a battery versus cycle number you will see that different batteries fade differently. Some batteries fade rapidly in the 1st few cycles and then the capacity stabilizes. Others increase in capacity in the 1st few cycles then level off and then starts a linear fade.

And in some batteries, as you keep cycling them there is a point where the curve starts to “roll over”. In other words, what began as a linear fade starts to accelerate. When this occurs, you are a few cycles away from a complete dead battery (i.e., you can’t even power your watch with it!).

Lets talk a bit more technical for a minute. Those not interested, move a few paragraphs over.

Let's take a Li-ion batteries with a NCM cathode with 14% 1st cycle irreversible loss and a graphite anode with 8% 1st cycle loss. After formation, the battery discharge is dictated by the voltage of the cathode when the battery reaching the cutoff voltage. There is still lithium in the anode. Think of this as an anode-discharge buffer.

But then let us assume that the cathode works very well. The anode, as anodes tend to do, still has a small side reaction because the SEI (which is a passive layer on the anode expected to protect it) is not quite as protecting as we hoped for.

As each cycle proceeds, the lithium comes out of the cathode and a small part goes into making a new SEI and does not intercalate into the anode. As you discharge, you start to slowly deplete this anode buffer. Each cycle slowly pushes the anode potential higher and higher. At some point, all the buffer is fully depleted.

Before this buffer was depleted, if you looked at the graph of capacity vs cycle number you would not have seen any capacity fade. When you hit this point, you will start to see the fade and what you observe is like a roll over. The slope of the capacity fade curve changed.

In some cases, this change starts to accelerate the fade because of the placement of the cutoff potential in the battery.

Ergo... battery fade is nonlinear and just because you have a particular kind of fade in the first five hundred cycles does not mean the fade can be predicted over the next five hundred cycles.

All right, so you think the battery you are making is different. It has a cathode that has less irreversible loss compared to the cathode. So you don’t have the same chemical problem.

But maybe you have a mechanical degradation problem.

All batteries “breath” as they charge and discharge. As each cycle proceeds, you are slowly swinging the volume of the battery. The whole electrode is under stress (compressive in one direction and tensile on the other). Fatigue starts to set in. After repeated cycling, at some point cracking and breaking start to become a problem. Now we have a definite capacity fade.

As some of the particles break and stop participating in the reaction, the rest of the particles have to take the load and these particles are stressed further and this accelerates their fade.

And these effects depend on the kind of cycling that has been conducted. A 3 hour discharge (like that in a EV) will have a different fade characteristic compared to a 1/2 hour discharge (like in a 10 mile PHEV).

Let us try a third case. Let’s say you have a battery pack consisting of 100’s of cells (or 8000 cells if you are at Tesla). Now, its kind of hard to make all the batteries exactly the same. When you manufacture batteries the yield is not that great and companies typically wait three to four weeks before binning (sorting) the batteries. But batteries are not binned for consistent life, because there is no way to do that. Instead they are binned for self-discharge, which does not tell us much about the life.

So let’s say that as the pack is cycled, you start having a few cells fading at a faster rate. Then the cells that are good are going to work that much harder and so they will start to fade more. This will also result in an acceleration of the fade. And because its not possible to predict a priori at what rate each cells will be fading, its impossible to predict how long the whole pack will last.

If that sounded complicated it was a desperate attempt to hide my insecurities by trying to convince you that I know something about batteries.

The short summary: Battery fade is complicated and difficult to predict.

Now, not all batteries are going to do this. Some battery chemistries are better than others. And some companies (that can make consistent cells) will be better than others. But the question remains: How do we predict what is going to happen in 15 years, when the weather report seems so far off for the next day?

But why predict. Why not just cycle these batteries; wait for them to die; and use this data to find out how to cost the battery and amortize it; and then sit back and watch the moolah pile up.

Because it takes.... 15 years to get this information and we don’t have another 15 years for this business plan to come into effect. To be fair, we have been testing these batteries for a few years now and have some data. But this is nowhere close to the number of years of data needed and so we still can’t say for sure how many more cycles/years the batteries will last.

So why not do accelerated testing, you ask? Because one is not sure if the method used to accelerate the fade (e.g. increasing the temperature) results in the different fade mechanism becoming dominant. Chemistry has this nasty habit of being hard!

So all the biz dev types have a problem: If you can’t predict how many cycles/years the battery will last how do you price the batteries today?

And the tech guy in the back of the room doing all the testing is sweating because he/she knows the complexity and knows that it is hard to predict how the battery is going to behave so far in the future.

Personally I think the concept is great and in time we will know how to make this a profitable business plan.

But for now, I’m not going to send that email to my bosses. I’m thinking about buying a house and I need the paycheck!

Venkat