LiFePO4 abuse tolerance & overall safety.

Been doing some reading on LiFePO4 battery safety & abuse tolerance. All I have to say is this is really incredible technology.

They withstand incredible degrees of overcharge without significant damage (some maufacturers have tested to 10 volts on a single cell without significant damage other than somewhat shorter cycle life).

They can in some cases withstand overdischarge with minimal damage i.e. shorter cycle life.

Withstand short circuits without safety issues

Withstand being punctured without serious safty issues.

Withstand high temperatures up to 300 degrees F without issue

Able to supply very high power without excessive heating. Temperatures remaining at levels that can be touched by human hands without being burned & for the most part remaining comfortable. (110degrees F)

Due to ability to withstand overcharge they are also self balancing so they can be charged in a serial connection without charge imbalance as long as you fully charge the pack.

These batteries are safe for the environment.

Other Than capacity issues which are improving these look to be the battery of the future. Thier safety & abuse profiles appear to exceed that of all other batteries in high abuse situations, even the stallwarts of lead acid (explosion potential during charging due to the generation of hydrogen & oxygen gasses, also environmentally hazardous) nickle metal hydride (excess heating under load may cause leakage & nasty burns when handled after subjection to high loads that significantly exceed max disharge or charge rates) & Nicad (same as Ni-MH but less pronounced but with far greater enviromental hazard)

germanium said:

They withstand incredible degrees of overcharge without significant damage (some maufacturers have tested to 10 volts on a single cell without significant damage other than somewhat shorter cycle life).

Click to expand...

As do properly made LiMn2O4 (aka "spinel", LMO, IMR, etc...). Both cathode materials are completely "delithiated" at top of charge, so that additional charging only serves to increase the voltage. This will degrade the electrolyte and increase the internal resistance, but it will not make the cell unsafe.

germanium said:

They can in some cases withstand overdischarge with minimal damage i.e. shorter cycle life.

Click to expand...

This is a common misconception. The damage from overdischarging a lithium-ion cell occurs on the negative electrode (usually referred to as the anode). The negative electrode in ALL lithium-ion cells is graphite or hard carbon coated on a copper foil current collector. When you overdischarge a lithium-ion cell, the copper current collector can dissolve and re-plate in the separator, resulting in internal short circuits. In addition, the passivation film on the graphite (commonly referred to as the "solid electrolyte interface" or "SEI") dissolves and gas is generated when the cell is recharged. This effect is most often seen in pouch cells, which are known to swell up when charged after an overcharge event. Cells made with LiFePO4 positive electrodes use the same negative electrode, and therefore are neither more or less tolerant to overdischarge.

germanium said:

Withstand short circuits without safety issues

Withstand being punctured without serious safty issues.

Click to expand...

Cells with LiFePO4 cathodes have approximately 30-40% of the same storage energy as cells made with "conventional" cathode materials. If you charge a "conventional" lithium-ion cell to the same storage energy as a cell with LiFePO4, you will find that it has similar safety characteristics.

germanium said:

Withstand high temperatures up to 300 degrees F without issue

Click to expand...

When you heat a lithium-ion cell, the first thing to degrade is the negative electrode (graphite or hard carbon). The lithium dissolved in the graphite or carbon begins to react with the electrolyte, resulting in a permanent decrease in capacity and a permanent increase in internal resistance. This reaction starts between 70-90 degC (158-176 degF). Since LiFePO4 cells use the same negative electrode as other lithium-ion cells, they also have the same problem with respect to temperature stability.

germanium said:

Able to supply very high power without excessive heating. Temperatures remaining at levels that can be touched by human hands without being burned & for the most part remaining comfortable. (110degrees F)

Click to expand...

High power cells made by Sanyo and E-moli (among others) have better discharge power capability than LiFePO4 cells.

germanium said:

Due to ability to withstand overcharge they are also self balancing so they can be charged in a serial connection without charge imbalance as long as you fully charge the pack.

Click to expand...

Whoa, be very, very careful with that statement. There is NO lithium-ion cell that has self-balancing characteristics. That being said, very well made, high quality cells made with LiFePO4 (or LiMn2O4) can be charged in series without active balancing if they are initially well-balanced. However, this will inevitably result in one cell being charged to a higher voltage, and that cell will age faster. These are not like NiMH cells, that can be slowly charged and rebalanced due to oxygen recombination.

germanium said:

These batteries are safe for the environment.

Click to expand...

LiFePO4 cells use the same electrolyte, negative electrode and current collectors as conventional lithium-ion cells. Since they have less than half the energy density, you need to use 2X the amount of all of these materials and container materials to achieve the same number of Watt-hours. I would say that LiFePO4 are no worse than other lithium-ion chemistries, but I certainly would not say that they are better. Many people argue that they are worse.

germanium said:

Other Than capacity issues which are improving these look to be the battery of the future.

Click to expand...

More like the battery of the past. LiFePO4 has reached its peak from a performance standpoint. The low energy density is a killer issue for this chemistry. However, many companies are trying to get LiMnPO4 to work, which has a higher voltage and therefore higher energy density compared to LiFePO4.

germanium said:

Thier safety & abuse profiles appear to exceed that of all other batteries in high abuse situations,

Click to expand...

Not when you normalize the results to total energy stored. When you do this, all lithium-ion cells perform in a similar manner.

germanium said:

even the stallwarts of lead acid (explosion potential during charging due to the generation of hydrogen & oxygen gasses, also environmentally hazardous) nickle metal hydride (excess heating under load may cause leakage & nasty burns when handled after subjection to high loads that significantly exceed max disharge or charge rates) & Nicad (same as Ni-MH but less pronounced but with far greater enviromental hazard)

Click to expand...

It is difficult to compare safety characteristics between cell chemistries. However, it is worthwhile to note that LiFePO4 cells have lower energy density than NiMH cells.

Cheers,
BG

Radio controlled Aircraft hobbyists like LiFePO4 cells for their durability and safety.

$10 3S A123 charger

A123 racing / B&D VPX 1100 mAh CBA Graphs

All the statements I made were confirmed in manufacturers testing & as to capacity I have had in my own testing K2 energy's LFP123 batteries outlast all the regulated lithium ion rcr123 batteries in led flashlights by 50% without the excess heating of the battery compartment caused by said regulators as the regulators are not needed with this chemistry so capacity is slowly becoming less of an issue. Even LiMn2O4 cells would have to be regulated in many of our lights that use more than 1 cell due to excess voltage compared to standard lithium primaries. http://www.hipowergroup.com/Advantages/. for overcharge & cell balancing http://www.gebattery.com.cn/geb/EN/ProductList.asp?sortID=138&Sortpath=0,133,138, Again showing cell balacing not required http://www.yesa.com.hk/pages.asp?id=19 According to K2 energy thier batteries can be over discharged down to 1 volt with the only issue being shorter cycle life. One of mine in testing got down to .9 & recharged fine However I never said you could leave them in an overdicharged state for long as that is where the real damage is done even with standard lithium ion batteries. They must be recharged promptly to avoid serious damage. I have even over dicharged standard litium ion batteries that were unprotected without issue provided they were charged immediately upon noticing the light reaching a state of dimming.There seemed to be no performance loss short term from this. This was with the original unregulated unprotected juice battery RCR123 batteries modern ones are regulated & protected. The copper dendrites that cause battery shorting are formed during long term overdischage IE leaving them that way for days, months or years, not hours.

germanium said:

All the statements I made were confirmed in manufacturers testing

Click to expand...

Word to the wise: the lithium-ion battery industry in Asia is like the wild west. The Chinese government is pumping huge amounts of money into the industry, especially for LiFePO4. Take manufacturers' claims with a grain of salt, and remember that these cells have only 30-40% of the energy of lithium-ion cells made with LiCoO2 and its variants.

germanium said:

& as to capacity I have had in my own testing K2 energy's LFP123 batteries outlast all the regulated lithium ion rcr123 batteries in led flashlights by 50% without the excess heating of the battery compartment caused by said regulators as the regulators are not needed with this chemistry so capacity is slowly becoming less of an issue. Even LiMn2O4 cells would have to be regulated in many of our lights that use more than 1 cell due to excess voltage compared to standard lithium primaries.

Click to expand...

Do not confuse differences in cell design with differences in chemistry. My comments in the previous post refer to the chemistry. I would take a high quality LiFePO4 cell like those made by A123 over a low quality ****fire LiCoO2 any day of the week.

And I agree that LiFePO4 has some advantages because of the really flat discharge voltage. It is great for hotwire flashlights.

germanium said:

http://www.hipowergroup.com/Advantages/. for overcharge & cell balancing http://www.gebattery.com.cn/geb/EN/ProductList.asp?sortID=138&Sortpath=0,133,138, Again showing cell balacing not required http://http://www.yesa.com.hk/pages.asp?id=19

Click to expand...

Again, take what the manufacturer says with a grain of salt. As someone who designs and tests lithium-ion cells, I can tell you that it is possible to charge both LiFePO4 and LiMn2O4 cells in series strings without balancing leads. BUT, the cells need to be very high quality and care should be taken to rebalance them on a regular basis. Even then, it is almost certain that at least one cell in the string is going to take a beating and age faster than the rest. Even the DeWalt powertool packs with A123 cells have active balancing.

germanium said:

According to K2 energy thier batteries can be over discharged down to 1 volt with the only issue being shorter cycle life. One of mine in testing got down to .9 & recharged fine However I never said you could leave them in an overdicharged state for long as that is where the real damage is done even with standard lithium ion batteries. They must be recharged promptly to avoid serious damage. I have even over dicharged standard litium ion batteries that were unprotected without issue provided they were charged immediately upon noticing the light reaching a state of dimming.There seemed to be no performance loss short term from this. This was with the original unregulated unprotected juice battery RCR123 batteries modern ones are regulated & protected. The copper dendrites that cause battery shorting are formed during long term overdischage IE leaving them that way for days, months or years, not hours.

Click to expand...

This is no different from any lithium-ion cell. You can discharge a lithium-ion cell under load to ~1 V without causing much damage, as you said. This is true for all lithium-ion cells, and has nothing to do with the cathode. The real damage comes when the cell is slowly discharged below 2 V and sits there for a long time. ANY lithium-ion cell will be hurt by this kind of treatment. However, LiFePO4 cells tend to fail in a more benign manner than other lithium-ion cells because they have such a low energy density.

Don't buy into the hype. LiFePO4 has its uses and will always be around as an option that works well in some applications, but it is definitely not the future of lithium-ion.

Cheers,
BG

On my K2 LFP123 batteries I only had 2 batteries so I shimmed my my Inova T-5 to test these as the T-5 takes 3 batteries normally. This means there is 33% less voltage available causing the switch mode buck regulator to make up for the loss of voltage with 50% extra current Even under these dire circumstances of having to provide 50% extra current the K2 LFP123 batteries lasted as long as the best regulated rcr123 batteries which in my testing was the newer ultralast rcr123 batteries using 3 batteries. By the way the original Ultralast RCR123 batteries appeard to be LiFePO4 batteries but were woefully deficient in capactity. About 1/3 of the K2 energy LFP123 capacity & strangly they were not overcharge tolerant as I tried that to see if I could get longer run time but no dice & the battery was ruined after only one overcharge to 4.2 volts whereas these acording to the K2 energy thiers are safe to charge up to 4.2 volts though there would be no extra capacity from doing so. Cycle life would be reduced but not below what you get from standard lithium ion types.

By the way K2 energy is a USA based company, not chinese. They seem to have products that live up to thier claims.

so i guess what battery guy said (as is often suspected)
is , it is relative

the list of all the safe features sounds more like the "safety data sheet" as opposed to the "recommended operations and specs sheet" :shrug:

a well built properly done li-ion of the "less safe" variety , has a "safety data sheet" that has similar statements, but you would never see that be used for the "Sales sheet"

As things progress, seems often (or is it also relative) that they will use sales that are "misleading" (to say it nicely) more and more. AKA the chemistry hasnt changed much but the lies about it have especially if you need grant money to continue research

and still that thing about quality, makes soo much differences.
thanks for the total discussion. thanks OP for your experiences.

how many times have we been told that some battery was . .. . . and it was till i started using it good thing there are rare gems of quality made products out there too.

germanium said:

Due to ability to withstand overcharge they are also self balancing so they can be charged in a serial connection without charge imbalance as long as you fully charge the pack.

Click to expand...

bump for an old thread ... BS's guys prediction that LiFePO4 batteries will not be significant turns out not to be true.

LiFePO4 batteries found their first sweet spot in Motorcycle batteries. Lots of motorcycle riders are willing to pay a premium to save the 5-15lb over lead acid battery. whereas saving 50lb or so in a car makes little to no difference for most folks.

due to popularity of using A123 26650 cells in 4s configuration, stacked in parallel to create higher amp hour batteries.

the question has come up does LiFePO4 batteries actually self balance? some say yes... some say no way..

what say you battery folks on CPF?

I have a build planned right now using two "D" LiFePO4 cells in series for the reasons stated above, including fast charge, light weight, less danger, etc.

From browsing the last few days, there does indeed seem to be a huge market for these in retrofitting electric bicycles, motorcycles, and light cars.

LEDAdd1ct said:

I have a build planned right now using two "D" LiFePO4 cells in series for the reasons stated above, including fast charge, light weight, less danger, etc.

From browsing the last few days, there does indeed seem to be a huge market for these in retrofitting electric bicycles, motorcycles, and light cars.

Click to expand...

yup ... but the biggest potential market for LiFePO4 batteries are Motorcycles. lots of folks paying $$$ for carbon fiber farkles to save a few oz on their high tech motorcycles. think in terms of cell voltage ... lithium cobalt nominal volt of 3.7v doesn't match 12v charging systems or 3.5v to 4.2v x 4 cell = 14v to 16.8v

LiFePO4 nominal volt 3.3v or for 4x cell... 12.7v to 14.6v fully charged or matches perfect to a normal 12v charging system of 13.8v to 14.2v

back to original question of ... does LifePO4 cells self balance in 4s configuration?

2S - 8S RC prismatic LiFeP04 battery packs come with a balance wire harness to be used when charging so I guess they do not self-balance.

BVH said:

2S - 8S RC prismatic LiFeP04 battery packs come with a balance wire harness to be used when charging so I guess they do not self-balance.

Click to expand...

some motorcycle LiFePO4 batteries contain up to 16x 26650 A123 cells with no balance circuit of any kind ... some come with internal BMS, some with external balance ports

what happened to battery guy?

cy said:

what happened to battery guy?

Click to expand...

I still lurk, and check my inbox from time-to-time, but work and life commitments have kept me from jumping into the fray on CPF.

Cheers,
BG

Battery Guy said:

I still lurk, and check my inbox from time-to-time, but work and life commitments have kept me from jumping into the fray on CPF.

Cheers,
BG

Click to expand...

thanks .. my thread on Advriders has quickly morphed into the most technical information on LiFePO4 motorcycle batteries on the WWW. with several major LiFePO4/charger mfg supporting the effort.

please consider visiting here and leave a comment on what you think ..

like several of us old timers on CPF .. we were the guinea pigs for consumer lithium batteries .. go back to 2002 era when Surefire was among the first to use lithium cells to power flashlights using CR123.

Arc flashlights was first to introduce a production Luxeon flashlight right here on CPF... the first offerings was limited to 100 .. registered for that first run, but for some reason didn't happen. anyways .. Arc LS First Run was a hit with folks on CPF responsible.

didn't take long after that for CPF'er to start experimenting on li-ion batteries for flashlights. our own JS Burley basically hocked his house to finance development of world's first protected li-ion battery (R123) .. capacity was a miserable 150 milliamp hour if you were lucky. but he delivered as promised in that shaky group buy.

then the Chinese basically grabbed technologies that JS Burley paid for and ran with it. rest is history on how the Chinese came to dominate li-ion batteries. it started right here on CPF ..

cy said:

bump for an old thread ... BS's guys prediction that LiFePO4 batteries will not be significant turns out not to be true.


LiFePO4 batteries found their first sweet spot in Motorcycle batteries. Lots of motorcycle riders are willing to pay a premium to save the 5-15lb over lead acid battery. whereas saving 50lb or so in a car makes little to no difference for most folks.

Click to expand...

Although weight is less of an issue in cars, there are still other reasons to go LiFePO4 for automotive use. Better cycle life is one. It also won't leak and destroy your engine bay, trunk, or other battery storage area. And, of course, it produces no gasses and requires no maintenance. But larger scale adoption of LiFePO4 for automotive use depends on cost. From what I have seen, LiFePO4 is actually quite cost competitive with AGM (although not flooded batteries). So it is certainly a good option here. And as far as I know, Porsche even offers it is an option in some vehicles (although at VERY high cost).


But besides automotive/motorcycle use, I wouldn't call LiFePO4 an abject failure as far as battery technology. For one, it actually finds good use in cordless power tools. LiFePO4 excels here, since it can supply high current and because its very constant voltage output means that you don't lose power as the battery discharges (or need fancy, power-consuming circuitry to keep the motor supplied with constant voltage from a very non-constant source like IMR batteries). Furthermore, the ease with which a 12V battery can be built (as well as good safety and cycle life) also makes LiFePO4 a good choice for solar applications. Like automotive applications, it's still more expensive than flooded lead acid. But it remains a good choice in situations where low maintenance and long life are paramount. It is also commonly used for solar garden lights. The long life, lack of a need for charge control and protection circuitry, ability to direct drive an LED with a single cell (unlike Nicad) and overall ruggedness makes LiFePO4 the best choice for garden lights.


The issue of capacity remains problematic. It is hard to say whether the technology is 'maxed' (as has been said here) or whether R&D is simply ignored because LiCo represents a workable solution that's available now (despite it's many warts). On the other hand, even with currently available cells, the capacity advantage of LiCo DOES tend to diminish as cells become larger. So while we see a big advantage in terms of capacity if we compare a Panasonic 3400mAH LiCo 18650 to an 1100mAH A123 LiFePO4 18650, this narrows considerably if we step it up to a 26650 (4000-4500mAH for LiCo vs 2500mAH-3300mAH for LiFePO4). And if we compare Feilong's LiFePO4 32650 to their LiCo 32650, the increase drops to 20% (5000mAH vs 6000mAH). Indeed, large form factors seem to be where LiFePO4 really shines even today.


Interestingly, someone before mentioned LiMnPO4. This is a new one to me. And I haven't heard of it in the two years since it was apparently posted here. Not sure if this is still in development, or whether it has been canned because things didn't pan out. But a safe Li-Ion battery that produces 4.1V would be a REAL winner, ESPECIALLY if it retains LiFePO4's constant voltage characteristics.

StorminMatt said:

Although weight is less of an issue in cars, there are still other reasons to go LiFePO4 for automotive use. Better cycle life is one. It also won't leak and destroy your engine bay, trunk, or other battery storage area. And, of course, it produces no gasses. But much of this depends on cost. From what I have seen, LiFePO4 is actually quite cost competitive with AGM (although not flooded batteries). So it is certainly a good option here. And as far as I know, Porsche even offers it is an option in some vehicles (although at VERY high cost).


But besides automotive/motorcycle use, I wouldn't call LiFePO4 an abject failure as far as battery technology. For one, it actually finds good use in cordless power tools. LiFePO4 excels here, since it can supply high current and because its very constant voltage output means that you don't lose power as the battery discharges (or need fancy, power-consuming circuitry to keep the motor supplied with constant voltage from a very non-constant source like IMR batteries). Furthermore, the ease with which a 12V battery can be built (as well as good safety and cycle life) also makes LiFePO4 a good choice for solar applications. Like automotive applications, it's still more expensive than flooded lead acid. But it remains a good choice in situations where low maintenance is paramount.


The issue of capacity remains problematic. It is hard to say whether the technology is 'maxed' (as has been said here) or whether R&D is simply ignored because LiCo represents a workable solution that's available now (despite it's many warts). On the other hand, eve with currently available cells, the capacity advantage of LiCo DOES tend to disappear as cells become larger. So while we see a big advantage in terms of capacity if we compare a Panasonic 3400mAH LiCo 18650 to an 1100mAH A123 LiFePO4 18650, this narrows considerably if we step it up to a 26650 (4000-4500mAH for LiCo vs 2500mAH-3300mAH for LiFePO4). And if we compare Feilong's LiFePO4 32650 to their LiCo 32650, the increase drops to 20% (5000mAH vs 6000mAH). Indeed, large form factors seem to be where LiFePO4 really shines even today.


Interestingly, someone before mentioned LiMnPO4. This is a new one to me. And I haven't heard of it in the two years since it was apparently posted here. Not sure if this is still in development, or whether it has been canned because things didn't pan out. But a safe Li-Ion battery that produces 4.1V would be a REAL winner, ESPECIALLY if it retains LiFePO4's constant voltage characteristics.

Click to expand...

interesting about possibility of a safe Li-ion battery putting out 4.1v. but that would be a liability in terms of widest adoption. 12v charging systems are the most common in the world. nothing else remotely comes close. ALL charging system designed to support 12v PB will support 12v LiFePO4 .. voltages matches up nicely. with zero mods to charging system to run 12v LiFePO4.

actual voltage for 12v LiFePO4 ranges from 14.6v fully charged to 12.85v (20%) but at first amp draw voltage quickly drops from 14.6v to 13.3v (90%) .. then discharge curve is almost flat to 12.85v .. about 90% of available power occurs within 1/2 volt.

ALL 12v LiFePO4 uses four cells (3.65v full) in series .. with cylindrical call batteries using series/parallel to achieve higher AH batteries vs prismatic pouch LiFePO4 cells still use four cells, but pouch size increases for larger AH batteries.

lithium cobalt at 4.2v fully x 3 = 12.6v or not a match for normal charging voltage of 14.4v range. 4.2v x 4 = 16.8v or still not a match for PB charging system's 14.4v. lithium cobalt simply doesn't match 12v charging systems without complicated buck/boost circuits. vs an almost perfect match for 4x 3.65v = 14.6v which immediately drops to 13.3v or perfect for any 12v PB charging system.

aside from special applications like Porsch's $1,200+ LiFePO4 battery for racing 911's. cost to mfg LiFePO4 large enough for a normal car is simply too high. so what you save say 50lbs BFD in a 4,000 to 5,000lb car .. vs saving 15lb in a 400lb motorcycle means a LOT ..some of the lowest costs lbs saved from a motorcycle already trimmed down.

LiFePo4 motorcycle batteries has been a smash success!!! catching major player by surprise at speed of adoption. several new LiFePO4 mfg hitting the market place with lots more to come. it's the world first successful adoption of stand alone LiFePO4 batteries. where extra costs are justified by weight/performance alone.

cy said:

interesting about possibility of a safe Li-ion battery putting out 4.1v. but that would be a liability in terms of widest adoption. 12v charging systems are the most common in the world. nothing else remotely comes close. ALL charging system designed to support 12v PB will support 12v LiFePO4 .. voltages matches up nicely. with zero mods to charging system to run 12v LiFePO4.

Click to expand...

Clearly, a safe 4.1V Li-Ion battery wouldn't be a replacement for LiFePO4 in 12V applications (such as automotive, motorcycle, or solar). Rather, it would be a replacement for ICR and IMR (if it proves to be superior to IMR). This would especially be the case if LiMnPO4 has a voltage curve like LiFePO4, and can hold a fairly constant 3.8-4.0V throughout its discharge. Not only could regulation circuitry be simpler and more efficient. But with a higher sustained voltage throughout the discharge, stored energy would be higher than if voltage started at the same point and dropped drastically (like ICR and IMR).

StorminMatt said:

The issue of capacity remains problematic. It is hard to say whether the technology is 'maxed' (as has been said here) or whether R&D is simply ignored because LiCo represents a workable solution that's available now (despite it's many warts). On the other hand, even with currently available cells, the capacity advantage of LiCo DOES tend to diminish as cells become larger. So while we see a big advantage in terms of capacity if we compare a Panasonic 3400mAH LiCo 18650 to an 1100mAH A123 LiFePO4 18650, this narrows considerably if we step it up to a 26650 (4000-4500mAH for LiCo vs 2500mAH-3300mAH for LiFePO4). And if we compare Feilong's LiFePO4 32650 to their LiCo 32650, the increase drops to 20% (5000mAH vs 6000mAH). Indeed, large form factors seem to be where LiFePO4 really shines even today.

Click to expand...

I wonder if the reason for this is because LiFePO4 needs a more robust separator compared to LiCo? I don't know if it does or doesn't, but I'm speculating on a possible reason why the disparity between LiFePO4 and LiCo decreases as size increases. If so, this would cause less capacity loss in larger sizes. In any case, I'm personally more than willing to exchange capacity for safety and cycle life. The latter is especially important. A quality LiFePO4 cell can be recharged thousands of times, and could last well over a decade in service (perhaps even much longer but the jury is still out). NiCd (and possibly LSD NiMH) are about the only other chemistries which can approach or match this but LiFePO4 beats them hands down in capacity in the larger sizes.

BTW, I don't think LiFePO4 is being maxed out yet. I recall when A123 started shipping their 26650s that there was a mention of future capacity increases to 4000 mAh in one of the press releases. This was probably a practical theoretical maximum which they hoped to realize but never did. I suspect if we threw some R&D at the problem we could get capacities close to that. Maybe given the other technologies like IMR it's not considered worthwhile pouring in R&D money just to get at best a ~20% capacity increase, perhaps with greatly decreased cell reliability.

I'm not sure why the disparity in capacity disappears at larger sizes. A thicker separator could be a reason. But safety could be another. Specifically, as the size of a cell increases, the thermal path to the outside also increases. This means that a 32650 is more prone to overheating than a 26650, and a 26650 is more prone to overheating than an 18650. It may be necessary to pack larger cells less densely with LiCo in order to reduce the chance of thermal runaway. But with a MUCH higher threshold for thermal runaway, LiFePO4 may not require the same precautions. Therefore, capacity can increase in a more linear fashion with volume.

I'm also unsure of why capacity hasn't increases with LiFePO4, but believe it can at least somewhat. Like you mentioned, it may be impractical to do so. But there could be other reasons. As I said before, manufacturers could figure that LiCo is 'good enough' right now. So increases in LiFePO4 capacity may not be necessary. At the same time, this emphasis on LiCo could be hurting companies like A123 Systems. So while it may be possible to increase capacity, they may be in no financial shape to do the necessary R&D. In any case, a 4000mAH LiFePO4 26650 doesn't sound too far fetched. Right now, A123 is at 2300, which is over halfway there. And 3300mAH LiFePO4 26650 cells already exist. Admittedly, they are not of the high current variety like A123 cells. But that's not too far off. Maybe it will happen someday. But it might nit quite be prime time for LiFePO4.