Why do lithium batteries die and how to improve them?

[TomT]

Some can be programmed to go to 100%. Personally, I think that should all be programmed to 80-90% since the last 10 to 20% takes so long and is the hardest on the battery...

I guess I'd add a caveat about not trying to take DCQC to full SOC.

Is this even possible? I thought all DCQCs taper off to Level 2 rates at about 80% SOC. I don't think one has ever taken me past 10 (out of 12) bars before shutting off automatically.

 

[DaveEV]

Avoid charging at all when it is really hot. His measurements show that ChaDeMo charging when hot will likely be less detrimental than L2 charging because it take less time.

If I lived in Phoenix, I might just look into the "home 20k chademo" that some of the forum members are developing. Heck it might even be good to use around here in August.

Where does he say that? From what I understand he says that normalized for test time, there isn't much difference between accelerated testing when charging at slower vs faster rates.

And coulombic efficiency definitely goes down at higher charge rates, this directly implies that you are accelerating pack degradation.

And that ignores the fact that the pack will heat up more during a fast charge, resulting in higher battery pack temps.

 

[palmermd]

Where does he say that? From what I understand he says that normalized for test time, there isn't much difference between accelerated testing when charging at slower vs faster rates.
And coulombic efficiency definitely goes down at higher charge rates, this directly implies that you are accelerating pack degradation.
And that ignores the fact that the pack will heat up more during a fast charge, resulting in higher battery pack temps.

Right at the beginning. Starting around 3:35. At 5:00 more detail and at 5:30 he shows that slow charging and discharging at temperature kills cells. He continues to about 7:20.

He basically showed that on the same cells at the same temperature, if you went up and down at 1.5c/-2.5c rates, you get very little capacity loss (about 5% in 150 cycles), and the exact same setup with c/56 charge and discharge you can kill the cell (60% loss in 150 cycles).

now we don't charge at c/56 with the 3kW charger, it is more like c/7, but his data does show that slow charging at those temperatures is damaging. The problem is that he tested at 60c. It would be really cool to see what it would do at something close to what we have on an Arizona Leaf. 1c discharge, c/7 charging and 30c-40c temperature. Anyhow the two numbers we do have from him are straddling these real world numbers, and the cars in Arizona were also seeing results that fall right between them (about 27% loss in 250 cycles as a guestimate)

 

[DaveEV]

He basically showed that on the same cells at the same temperature, if you went up and down at 1.5c/-2.5c rates, you get very little capacity loss (about 5% in 150 cycles), and the exact same setup with c/56 charge and discharge you can kill the cell (60% loss in 150 cycles).

You are mis-interpreting his statements. It's not the slow charging that kills the cells, it's the the time it takes to run the test. Once you normalize the results for length of time, the amount of capacity loss is similar, and supports his entire argument that accelerated testing does not give you real-life results.

It's the same thing as taking the DC Fast Charge Effects on Battery life study results at 40,000 miles (ignore the L2 data for the time being) and saying "Look you can fast charge only a LEAF in Phoenix for 40,000 miles and only lose 25% capacity!"

Never mind that they ran through those 40k miles in about 1 year and if you then assume that a regular driver who drives 10k miles / year will make it 4 years before losing 25% capacity in Phoenix, you're going to be sorely disappointed when you find the car down 25% capacity after only 2 years and 20k miles (sound familiar?)

Note that this study directly compares the effects of QC vs 3.3 kW charging and it's clear that QC is worse, especially as it gets hot (miles 20-40k were done during warm months, miles 0-20k were done during cooler months), contradicting your conclusions.

 

[palmermd]

You are mis-interpreting his statements. It's not the slow charging that kills the cells, it's the the time it takes to run the test. Once you normalize the results for length of time, the amount of capacity loss is similar, and supports his entire argument that accelerated testing does not give you real-life results.

It's the same thing as taking the DC Fast Charge Effects on Battery life study results at 40,000 miles (ignore the L2 data for the time being) and saying "Look you can fast charge only a LEAF in Phoenix for 40,000 miles and only lose 25% capacity!"

Never mind that they ran through those 40k miles in about 1 year and if you then assume that a regular driver who drives 10k miles / year will make it 4 years before losing 25% capacity in Phoenix, you're going to be sorely disappointed when you find the car down 25% capacity after only 2 years and 20k miles (sound familiar?)

Note that this study directly compares the effects of QC vs 3.3 kW charging and it's clear that QC is worse, especially as it gets hot (miles 20-40k were done during warm months, miles 0-20k were done during cooler months), contradicting your conclusions.

ok. I see what your saying and thanks for the backup documentation. I'll have to re-watch with this in mind.

 

[RegGuheert]

Awesome presentation. Thanks for finding and sharing it. On the positive note it's great to see that the tools to make faster progress are coming together.

Yes! Thanks to drees for posting this! At the beginning of the video I was worried that it would be really boring, but the professor did not disappoint.

Being able to evaluate the *real* life of new battery chemistries in weeks rather than years (and often in the customer application) is an incredible achievement. It really should accelerate Li-ion battery development.

On the negative side, the existence of the "other" failure mode, the catastrophic one was news to me. The degradation we're all talking about for the LEAF is the gradual one. What about the catastrophic one, does anyone know what should be our expectation about that?

I have been pointing out for over a year now about the appearance of these catastrophic failures in the published literature on calendar effects. Also, I have stressed that the breakdown of the electrolyte is a linear degradation effect and that any linear degradation will eventually swamp out the effects of the exponential decay associated with other calendar and cycling effects that occur. As was pointed out in the presentation, a large number of cycles can ONLY be achieved by cycling quickly and therefore using the battery before the calendar effects have a chance to ruin it.

Yesterday I just posted about recent concerns I have concerning a significant drop in capacity seen in several of the cells in our LEAF this winter. The apparent signature is that the resistance of these cells may have increased significantly versus the other cells in the pack. As such, I have been wondering if our capacity may be starting to roll over a cliff. I really am beginning to wonder if ANY LEAFs will make it past the five-year period without warranty claims.

Since this type of fast life testing is very new, I doubt even Nissan knows the answer to that question.

 

[DaveEV]

The apparent signature is that the resistance of these cells may have increased significantly versus the other cells in the pack. As such, I have been wondering if our capacity may be starting to roll over a cliff. I really am beginning to wonder if ANY LEAFs will make it past the five-year period without warranty claims.

If that were the case, all cells would be affected similarly (unless Nissan has some poor quality control) and cars in hot climates would be seeing the effects sooner than you.

So far, no rapid drops of capacity have been reported unless it's been due to a single fault module and that is extremely rare.

 

[RegGuheert]

So far, no rapid drops of capacity have been reported unless it's been due to a single fault module and that is extremely rare.

I think any degradation that was extra-linear would fit this condition. (Note the much softer roll-off in some of the slides shown.). I think I have documented at least one case of extra-linear degradation here. Let me see if I can find it...

 

[Weatherman]

I looked at the entire lecture, and I thought it was said there was no way to predict, from accelerated tests, when the catastrophic failure would occur. Some look real good at the beginning, then fail quickly. Others look a little less good in the beginning, then last quite a while.

In any case, what I got out of the lecture was that it appears a method has been developed to more accurately predict battery degradation without having to run a test for eight years. That's good. It means a lot of different electrolyte additives can be tried and tested and there's a good chance they will behave in the real world as predicted in the accelerated tests.

If Nissan is using these methods, it bodes well for their ability to solve the current, severe degradation problem they have with the LEAF batteries.

 

[edatoakrun]

="drees"... Once you normalize the results for length of time, the amount of capacity loss is similar, and supports his entire argument that accelerated testing does not give you real-life results.

It's the same thing as taking the DC Fast Charge Effects on Battery life study results at 40,000 miles (ignore the L2 data for the time being) and saying "Look you can fast charge only a LEAF in Phoenix for 40,000 miles and only lose 25% capacity!"

Never mind that they ran through those 40k miles in about 1 year and if you then assume that a regular driver who drives 10k miles / year will make it 4 years before losing 25% capacity in Phoenix, you're going to be sorely disappointed when you find the car down 25% capacity after only 2 years and 20k miles...

Fewer miles driven per year should result in more capacity loss over the longer time period, but the lower battery temperatures over that longer period of time also should (to an unknown extent) offset that factor.

Thread on-topic and direct link to initial presentation of the study below:

...First notice that this study is not designed to determine LEAF capacity loss in typical use.

It is intended to measure the variable loss increase induced by DC charging in a atypical driving pattern, most unusually the (140-150 miles per driving day?) necessary to rack up 40,000 miles in 12 to 13 months.

Even the L2 cars, since they are spending such a large time heating their batteries by repeated charge/discharge cycles, are introducing an additional factor in capacity loss that does not exist for Phoenix LEAF drivers with more typical miles driven per day or year.

So we can't see the extent to which the beneficial effects (from cooler batteries) by driving "normal" miles per day/year, and the detrimental effects longer-term calendar capacity loss, may offset each other, from these reports...

http://www.mynissanleaf.com/viewtopic.php?f=27&t=14271&start=10" onclick="window.open(this.href);return false;

http://www4.eere.energy.gov/vehiclesandfuels/resources/merit-review/sites/default/files/vss113_francfort_2013_o.pdf" onclick="window.open(this.href);return false;

 

[DaveEV]

I looked at the entire lecture, and I thought it was said there was no way to predict, from accelerated tests, when the catastrophic failure would occur. Some look real good at the beginning, then fail quickly. Others look a little less good in the beginning, then last quite a while.

I believe he said they had a way to predict it, but typical accelerated testing did not. I don't recall exactly how, will have to re-review the video.

 

[GregH]

It would be interesting to know how the "time at temperature" component swamps all others..
For example at different SOCs/voltages (how much worse is "full" than 80%? Or 80% than 50%? Or 50% than 10%? Or 10% than 50%(!))
Also with respect to speed of charge (or discharge) or for that matter total Ah throughput..
If you trickle charge 150 cycles (C/56) vs 1.5C/2.5C but with enough rest time between each of the 150 cycles to reach the same total time at temperature...?
Ugh, then of course at what voltage do you leave it during those rest periods? So many factors!

 

[ericsf]

I looked at the entire lecture, and I thought it was said there was no way to predict, from accelerated tests, when the catastrophic failure would occur. Some look real good at the beginning, then fail quickly. Others look a little less good in the beginning, then last quite a while.

I believe he said they had a way to predict it, but typical accelerated testing did not. I don't recall exactly how, will have to re-review the video.

I believe he answers this at the Q and A at the very end. He said that their method can't predict how and when a cell will fail in real world usage. However what they can do is compare cells. Therefore someone who has data about how a type of cell has actually performed, they'll be able to say if it's life expectancy is better or worse and roughly by how much. The key is they can do that quickly.

My takeaway is that manufacturers who got in early will have an edge on latecomers to develop the best batteries in the future. Having the data is key. Also, it's interesting to see how Tesla took the path of making the best performance pack regardless at high cost and how Nissan took the opposite of going for the cheaper one. Neither way is wrong and each has its own challenges. Tesla's got to cost reduce and Nissan's got to improve peformance without increasing cost. Very interesting times ahead. To quote Dhan : "Everybody's excited!"

 

[RegGuheert]

So far, no rapid drops of capacity have been reported unless it's been due to a single fault module and that is extremely rare.

I think any degradation that was extra-linear would fit this condition. (Note the much softer roll-off in some of the slides shown.). I think I have documented at least one case of extra-linear degradation here. Let me see if I can find it...

Here it is:

As an example, consider member "cyellen" who experienced the loss of bars one and two during the summer of 2012 and bars three and four during the summer of 2013:
Bar 1 to 2: 73 days and 1400 miles
Bar 3 to 4: 64 days and 1500 miles

Just one data point (since we don't have much data), but it makes it clear that capacity loss does not slow down with degradation. Does it accelerate? It's not clear, but perhaps. Perhaps someone will hold onto their LEAF beyond four bars of loss to see how degradation progresses beyond that point.

 

[ericsf]

Another thing that intrigued me in the presentation was the comparison of the batteries in the Model S, Leaf, Volt and Fisker at this point: http://www.youtube.com/watch?v=pxP0Cu00sZs#t=1516

He mentions 2 different type of chemistry for the LEAF and the Volt. A good one and a bad one when it comes to resistance to heat. He says it's a 50/50 blend. Li(NiMnCo)O2 and LiMnO2.

Does anybody have any info on this? Does Nissan blend the 2 type of chemistry in each pack or do they produce 50% with one type and other 50% with the other. Or could he be referring to the new "hot climate" battery chemistry? But in that case why would he say it's 50/50 instead of old/new and why did he throw the Volt in the same basket?

 

[RegGuheert]

He mentions 2 different type of chemistry for the LEAF and the Volt. A good one and a bad one when it comes to resistance to heat. He says it's a 50/50 blend. Li(NiMnCo)O2 and LiMnO2.

Does anybody have any info on this? Does Nissan blend the 2 type of chemistry in each pack or do they produce 50% with one type and other 50% with the other. Or could he be referring to the new "hot climate" battery chemistry? But in that case why would he say it's 50/50 instead of old/new and why did he throw the Volt in the same basket?

I'm pretty sure he was talking about the chemistry of the anode and the chemistry of the cathode when he said that. He is discussing the combination of those two chemistries in general, but likely without any knowledge of the specific additives used in the electrolytes of either the batteries made by Nissan or LG Chem (for the Chevy Volt).

 

[dhanson865]

Another thing that intrigued me in the presentation was the comparison of the batteries in the Model S, Leaf, Volt and Fisker at this point: http://www.youtube.com/watch?v=pxP0Cu00sZs#t=1516

He mentions 2 different type of chemistry for the LEAF and the Volt. A good one and a bad one when it comes to resistance to heat. He says it's a 50/50 blend. Li(NiMnCo)O2 and LiMnO2.

Does anybody have any info on this? Does Nissan blend the 2 type of chemistry in each pack or do they produce 50% with one type and other 50% with the other. Or could he be referring to the new "hot climate" battery chemistry? But in that case why would he say it's 50/50 instead of old/new and why did he throw the Volt in the same basket?

it's a blend in each cell not just pack, every cell has to be blended or your pack is useless when half the cells die.

 

[ericsf]

Another thing that intrigued me in the presentation was the comparison of the batteries in the Model S, Leaf, Volt and Fisker at this point: http://www.youtube.com/watch?v=pxP0Cu00sZs#t=1516

He mentions 2 different type of chemistry for the LEAF and the Volt. A good one and a bad one when it comes to resistance to heat. He says it's a 50/50 blend. Li(NiMnCo)O2 and LiMnO2.

Does anybody have any info on this? Does Nissan blend the 2 type of chemistry in each pack or do they produce 50% with one type and other 50% with the other. Or could he be referring to the new "hot climate" battery chemistry? But in that case why would he say it's 50/50 instead of old/new and why did he throw the Volt in the same basket?

it's a blend in each cell not just pack, every cell has to be blended or your pack is useless when half the cells die.

Indeed, mixing different types of cells in a pack wouldn't make sense. It never occured to me that one could blend chemistry "types" like Li(NiMnCo)O2 and LiMnO2 within the same cell. I thought the "mixing" was only about the additives. It's even more complicated than I imagined. Thanks!

 

[Nubo]

I believe he answers this at the Q and A at the very end. He said that their method can't predict how and when a cell will fail in real world usage. However what they can do is compare cells. Therefore someone who has data about how a type of cell has actually performed, they'll be able to say if it's life expectancy is better or worse and roughly by how much. The key is they can do that quickly.

My takeaway is that manufacturers who got in early will have an edge on latecomers to develop the best batteries in the future. Having the data is key. Also, it's interesting to see how Tesla took the path of making the best performance pack regardless at high cost and how Nissan took the opposite of going for the cheaper one. Neither way is wrong and each has its own challenges. Tesla's got to cost reduce and Nissan's got to improve peformance without increasing cost. Very interesting times ahead. To quote Dhan : "Everybody's excited!"

As I understood it, their model ascribes cell degradation and death to over-accumulation of the solid electrolyte interphase (passivation) layer. Their test can predict the rate at which the passivation layer accumulates. Cell death accelerates at the time when the layer becomes thick enough to have blocked most of the pores. Then the Lithium can't get to its destination , begins plating over the anode and cell dies quickly.

So the test method predicts how quickly the layer accumulates, but the amount of time it takes to actually clog the pores depends on the physical characteristics of the anode (pore size). The quick test method can't determine the latter. But for any given anode, the slower the layer accumulates, the longer the cell lives. They select the chemistry that promises the longest life, and then the long-term test validates and quantifies that life.

 

[ericsf]

I believe he answers this at the Q and A at the very end. He said that their method can't predict how and when a cell will fail in real world usage. However what they can do is compare cells. Therefore someone who has data about how a type of cell has actually performed, they'll be able to say if it's life expectancy is better or worse and roughly by how much. The key is they can do that quickly.

My takeaway is that manufacturers who got in early will have an edge on latecomers to develop the best batteries in the future. Having the data is key. Also, it's interesting to see how Tesla took the path of making the best performance pack regardless at high cost and how Nissan took the opposite of going for the cheaper one. Neither way is wrong and each has its own challenges. Tesla's got to cost reduce and Nissan's got to improve peformance without increasing cost. Very interesting times ahead. To quote Dhan : "Everybody's excited!"

As I understood it, their model ascribes cell degradation and death to over-accumulation of the solid electrolyte interphase (passivation) layer. Their test can predict the rate at which the passivation layer accumulates. Cell death accelerates at the time when the layer becomes thick enough to have blocked most of the pores. Then the Lithium can't get to its destination , begins plating over the anode and cell dies quickly.

So the test method predicts how quickly the layer accumulates, but the amount of time it takes to actually clog the pores depends on the physical characteristics of the anode (pore size). The quick test method can't determine the latter. But for any given anode, the slower the layer accumulates, the longer the cell lives. They select the chemistry that promises the longest life, and then the long-term test validates and quantifies that life.

What I think the beauty of their method is, is that it doesn't assume anything about what's going on in the cells. Their model is that any energy missing from an ideal cell is due to parasitic reactions and is bad for the cell's durability. He showed that it worked for both electrolyte degradation and pore blocking. I didn't see any reason why their method would not catch any kind of degradation, even those we don't know exist. The part I didn't fully get is where the calorimetric measurement fits in their method. Is it just a different way to measure the parasitic reactions or does it provide additional data they could not get with coulomb measurement?