Much Ado about “Embodied Energy”

(December 2019- updated with the new IVL study results by Dahllof et al)

If you’re interested in electric vehicles (battery EVs) or hybrids, or in grid storage, you’ve no doubt seen headlines like this:

“Hummer Greener than Prius”

“Swedish study calls for smaller EV batteries, finds Tesla more polluting than an 8-year-old car”

“Tesla’s Electric Cars Aren’t as Green as You Might Think”

“Dirty Secrets of Electric Cars”

“Could Lithium Become the New Oil?”

You’ve also likely heard people talking about the “rare earth metals” in electric cars, or warning about dire impending shortages of lithium, cobalt, graphite or other materials used in making EVs. We’ll tackle that issue in a subsequent article. 

Many of the arguments in these pieces rest on the concept of “embodied energy” or “embodied emissions”- how much energy and greenhouse gas (GHG) emissions result from producing one or more components of a more complex system like a battery or a vehicle.

Some of the claims in these articles are just garbage- badly done studies, using careless or sometimes deliberately distorted methodology, that were somehow picked up and promoted in the media. But many of them are myths, which is something very different. Myths generally contain a kernel of truth which give the reports a basis for credibility, even when the overall conclusions are unfounded, distorted or just plain wrong. I’ll tackle some of these in a subsequent article about battery materials, but for now, we’ll tackle embodied energy and especially embodied emissions.

Lifecycle Assessments- a Very Blunt Tool

Right up front, it’s extremely important to realize that all forms of energy production and consumption have environmental impacts. There is no perfect technology: in the real world, we have to select between real alternatives, acknowledging the strengths and weaknesses of each and weighing them in accordance with our values. Too frequently, we are guilty of assuming that the “do nothing” option is artificially low in risk, burdening any innovation with an insurmountable obstacle: competition against the impossible perfect technology.  When we do this, we’re guilty of the Nirvana fallacy:

 https://en.wikipedia.org/wiki/Nirvana_fallacy

Acknowledging that all technologies have strengths and weaknesses is NOT the same as saying that they all have equivalent environmental impacts!  But how do we know whether or not the cool new thing is any better than what we’re already doing? And the even harder question is this: is doing something new sufficiently better to justify making the change?

If we care about both energy efficiency and GHG emissions (and the science says we should- full stop), then it’s very important for us to understand the net amount of energy and GHG emissions associated with all the steps in the production, storage, consumption and end-of-life of a plant, vehicle etc.

Ah, there’s the rub! That’s dead easy to say, but deadly difficult to actually do, in a meaningful and consistent fashion that allows meaningful comparison. And it’s this simple fact that muddies the waters and leads to those bad headlines.

Just How Blunt is This Tool?

One of the principal skills that either is, or should be, taught to a chemical engineer is something called “drawing a system boundary”- the first step in analyzing a problem. When evaluating a unit operation, or a group of them which form a system, we need to establish a boundary by drawing a dashed line around the whole process. We then have to investigate every single thing which crosses that dashed line, either entering or exiting the process. We then use conservation of mass and energy and simple balance equations (input minus output equals accumulation) to help us figure out what’s going on.

The trouble with every single lifecycle assessment, though, is where the heck do you draw the system boundary to make the analysis meaningful?

Let’s take a seemingly easy example: aluminum used to make a Tesla frame or a current collector in a lithium ion battery. That’s just one of many materials, in many product forms, which are necessary to make anything we might care about.

Here are the basic steps:

• Mine bauxite, cryolite, salt and petroleum
• Refine petroleum to make a host of products, leaving pitch (residuum) as a byproduct
• Further process that residuum to more products and coke
• Make electrodes from coke and pitch (we need to include these because they’re not just a tool, they’re a feedstock- they’re consumed by the process)
• Electrolyze the salt to make sodium hydroxide (and sell chlorine, bleach etc.)
• Leach the bauxite with the sodium hydroxide to make aluminum hydroxide
• Roast the aluminum hydroxide to make aluminum oxide (alumina) and CO2
• Melt the alumina with cryolite as flux
• Electrolyze the melt, producing molten aluminum and hydrogen fluoride
• Capture the hydrogen fluoride and recycle it as aluminum fluoride
• Mix in scrap aluminum from recycling
• Alloy the aluminum with other metals
• Produce the final product form (bar, sheet etc.)

We could try to draw a system boundary around that whole mess, but that would be a waste of time. Instead, you’ll rely on someone else having broken down many of the component subsystems (petroleum recovery and refining, sodium hydroxide manufacture etc.) so you can just take an energy content and GHG figure for that component (coke, resid, or even finished electrodes) . That’s just a practical necessity. 

But here’s the real problem: even if you pick just one of those subsystems, where do you draw its system boundary? Even for a single component like sodium hydroxide, that’s not easy. A chlor-alkali plant has many inputs and many products. But even if you were to account for all of those properly, are you done?  What about the embodied energy of the chlor-alkali plant itself? And what the heck do you do with the labour? When we do the calculation, do we include the energy required for all the workers to get to and from work? To produce the food they eat?  And what about all the support labour and services- the lawyers, doctors and engineers, salespeople etc. associated with the business?

The answer you get will depend on how far out you go when drawing that system boundary!

And remember, so far we’ve only really talked about one step: making one of the materials to make a product. A lifecycle analysis requires similar evaluation of the energy and emissions from the use, maintenance and disposal or recycling of the thing itself, all with similar complexities and opportunities for error.

Fortunately, there is at least a little rigour in the field of doing lifecycle analysis for a living, which for full disclosure I am not involved in doing myself, so I freely admit I do not fully understand. Regrettably, people who aren’t competent whatsoever in this field also attempt to do lifecycle analysis themselves, and their reports are picked up by the media, especially when they fulfil a desire for a shocking headline. 

The best example is the source of the first headline: “Hummer Greener than Prius”. This resulted from a 2006 study by a company called “CNW Marketing Research”, titled “Dust to Dust”. This “study” attempted a lifecycle energy and emissions comparison among various types of vehicles, and has been since thoroughly debunked. One of the best de-bunkings is linked below: 

The fundamental flaws in this study were many, including a lot of totally unjustified assumptions, but some of the biggest ones were in drawing the system boundaries in a totally ridiculous way, inconsistent with the established methods used by people who actually do such studies for a living.

Does Embodied Energy Defeat the Benefits of EVs?

So, where does this leave electric vehicles and hybrids?  Does their embodied energy overwhelm the benefits they provide? The answer is yes, it’s possible- but in practice, probably not- in fact, the opposite is more likely, i.e. EVs are expected to give significant net savings in GHG emissions over their projected lifetime, and those benefits are expected to improve as grids get greener (as coal is replaced by natural gas and renewables).

The Swedish Study

Even relatively good studies, like this one from May 2017 into the embodied energy and emissions of EV batteries:

http://www.ivl.se/download/18.5922281715bdaebede95a9/1496136143435/C243.pdf

are misquoted, misinterpreted and mis-used in the media to draw conclusions, often on an ideological basis. That’s exactly what has happened, in my opinion, to this study. It was done by Mia Romare and Lisbeth Dahllof of IVL Swedish Environmental Research Institute Ltd. The study itself is a review and composite of other studies, which differ in methodology and hence give quite a range of results.  EV detractors have used the study to draw incorrect conclusions, and EV proponents have attacked its credibility without reviewing or understanding it in detail. 

DECEMBER 2019 UPDATE: IVL has revised their study, available here:

https://www.ivl.se/download/18.14d7b12e16e3c5c36271070/1574923989017/C444.pdf

The new study depends heavily on one paper for the majority of the new data: Dai et al, Batteries 2019, 5, 48. A link to that paper is given below:

https://www.mdpi.com/2313-0105/5/2/48

This paper, done by the well respected team at the Argonne National Laboratory- the group responsible for the go-to tool for lifecycle cost analyses (LCAs) the GREET model, has done a good job of identifying both the energy and the greenhouse gas (GHG) emission intensities of lithium ion batteries per kWh of battery pack capacity.

Here are the conclusions I’ve drawn from the most recent IVL study, and from its supporting references, that I think are actually reliable:

1. A lithium ion battery pack of 1 kWh capacity takes about 1,125 MJ of energy to make in total (including making all the materials that the battery pack is made out of- that was previously estimated at 350-650 MJ by Dahllof et al) and results in about 70 kg (with an estimate range of 61-106kg) of CO2 equivalent GHG emissions per kWh. That figure was previously estimated at 150-200 kg CO2e in the 2017 report.  

 As I previously indicated, the 150-200 kg figure for CO2e intensity were conservative (by a factor of at least 2), and the NEW figures are likewise still conservative- despite the fact that the original energy intensity estimates seem to have been low. For instance, the new 1125 MJ/kWh figure includes emissions from using a 4-effect evaporator to recover sodium sulphate from the water treatment system. Treating process water by a process as energy intensive as evaporation is quite unusual.

 The new study makes it clear that most of the source energy is heat rather than electricity, which should be obvious. If you were to convert MJ to kWh (confusing work with heat) that's about 311 kWh (up from 97-180 kWh) of energy input to make 1 kWh worth of battery pack. Looking at 61-106kg CO2e per 311 kWh, that's a range of to 196 to 340 g of CO2 per equivalent kWh of input energy- a rather dramatic change from the previous range of 830 to 2060 g CO2/kWh (figures which were, clearly, impossibly bad, since even if 100% of the energy used were electricity from coal, the intensity would only have been about 1000 g CO2/kWh).

2. About 30% of the energy and emissions are from production of the battery cells and pack from the finished battery-grade materials (down from 50% previously)

3. The range of energy sources used to make batteries currently are used in the assessments. These range from grids in Asia which are still heavily fossil-fuel dependent, to grids in Europe which are less so.

4. The result (still) doesn’t depend too strongly on battery chemistry. The three primary cathode chemistries used in vehicle batteries right now, known as NMC (lithium nickel manganese cobalt oxide, used by Chev and Nissan), NCA (lithium nickel cobalt aluminum oxide, used by Tesla) and LMO (lithium manganese oxide, used by Nissan and Renault), all have similar energy and emissions intensities- well within the (poor) accuracy with which those factors are known.

5. Emissions from mining are a comparatively small fraction of the total emissions. Processing mined materials to battery grade materials takes more.

6. Recycling is still not considered, and will be a significant energy/emissions intensity reduction factor in future once there are sufficient dead EV batteries to recycle. EV batteries are better ores for their value metals than any ore currently mined for these metals on earth, so recovering the metals from spent batteries will be less energy and emissions-intensive than mining and refining from raw ores by a significant margin.

7. While it seems clear that used vehicle batteries, at least from fleet use, are substantially being put to use in “2nd life” applications in China as we speak, the energy intensity benefits of 2nd life uses are ignored. I still think this is a good long-term assumption, since most EV batteries from personally owned vehicles will ultimately arrive in crashed or derelict cars at junkyards and hence are much more likely to be recycled rather than re-used.

A figure from Dai et al gives some insight into where the emissions are generated. The majority of the energy is used to make the NMC cathode material, aluminum collector foils, carbon/graphite components and the copper collector foils.

(figure from Dai et al)

Does It Matter? Time for Some Calculations!

First, let’s see if my E-Fire converted Triumph Spitfire has an embodied energy or emissions problem. It has an 18.5 kWh LiFePO4 pack. Let’s take the worst case GHG emissions: 200 kg CO2/kWh, from the Swedish study. That’s 200x 18.5 = 3,700 kg of CO2 emissions to make the pack. 

(Update 2019: that’s now estimated to be 1100 to 2000 kg- down significantly from 3700 kg)

Remembering from my first paper,

https://www.linkedin.com/pulse/so-exactly-how-much-electricity-does-take-produce-gallon-paul-martin/

-gasoline’s well to tank efficiency of about 80% and that combustion of 1L of gasoline produces 2.3 kg of CO2 just from combustion, we end up with a well to tank CO2 emission of 2.87 kg/L. The battery pack is therefore equivalent to about 1290 L (341 US gal) of gasoline from well to tank. Now remember that the E-Fire used 8L/100km (29 miles per US gal) driven prior to the conversion, making it equivalent to a “fleet average” car circa ~2015- it was clearly ahead of its time! 

Assuming we recharged the E-Fire using 100% renewable energy with zero GHG emissions, we’d need to drive the new E-Fire 16,125 km (about 9800 miles) just to pay back the GHG emissions required to make the battery pack.   Using Ontario’s 40 g CO2/kWh grid, that number rises a bit to about 10,100 miles. On a grid which was say 50% fossil fired, that number might climb to 20,000 miles or even higher. Not an insignificant number! 

(Update 2019: that’s now dropped by a factor of 2 to 3.3 as a result of the new embodied emissions figures. The battery pack would pay back its embodied emissions after as little as 3-5,000 miles driven. Not zero, but certainly a TINY fraction of the battery pack’s expected lifetime).

However, we have to look at how many miles we can expect the pack to last before it needs to be replaced, to make a fair comparison. For that, we’ll use 3,000 charge/discharge cycles to 70% depth of discharge (DOD) each time, which is what the manufacturer of my cells indicates in their literature as an expected lifetime for an 80% DOD. The car gets 235 Wh/mile out of the battery, so that’s the figure that matters here. 3000 cycles x 70% x 18,500 Wh/cycle /235 Wh/mile gives about 165,000 miles (266,000 km). Over that distance, had I not converted the car, and assuming that I didn’t need to either rebuild or replace my Leyland engine a few times to go that far (a certainty in my experience…the engine in this car was 1970s garbage through and through!), I would have needed 21,280 L of gasoline (~5,600 gallons), resulting in CO2 emissions of ~61,000 kg of CO2 from source. The electricity to recharge the pack that many times would be 3000 x 70% x 18.5 kWh/90% (charger and battery efficiency)/94% (grid losses), or 45,900 kWh. At 40 g CO2/kWh in Ontario, that’d be about 1,800 kg of CO2.  Net savings over the lifetime of the pack are therefore 61,000 minus 3700 minus 1800 kg, or over 55.5 tonnes of fossil CO2 emissions averted – CO2 which wouldn’t end up in the atmosphere warming the planet. OK, so I can personally continue to feel good about driving the E-Fire to work rather than my Prius!

(Update 2019- we’re now looking at 58-59 tonnes of net CO2 saved over the battery lifetime- a very impressive result, and yet another reminder that EVs are an environmental no-brainer for the 75% of Canadians who have access to a grid which is 40 g CO2/kWh or less)

Now let’s repeat the calculation for a Tesla model S (not that I can afford one!): that car has an energy efficiency of closer to 400 Wh/mile due to its greater mass and frontal area and despite its superior drag coefficient (and also due to the tendency of their drivers to have a heavy foot due to how fun these cars are to drive). That’s a rough number, but not too far off. It also comes with a pack which can range from 60-100 kWh. Let’s choose 70 kWh as a typical pack size. Re-running the calcs in the same way, here are the results:

Pack embodied CO2 emissions, worst case:    14,000 kg (2019- now 4200 to 7400 kg)

Equivalent gasoline from source: 4,880 L (2019- now 1460 to 2580 L)

Equivalent fleet average distance driven to break even (0 g CO2/kWh): 61,000 km (~37,000 miles) (2019- now that figure is down to 18,250 to 32,250 km)

Distance to payback, assuming 450 g CO2/kWh (the US grid average figure): ~138,000 km (~86,000 miles) (2019: this is now around 26,300 to 46,000 miles)

Pack life estimate: 367,500 miles (~592,000 km)

OK, here’s where it gets interesting: how many people drive their cars over 300,000 miles before trading them in or scrapping them? Precious few…So is it fair to give the car credit for the full potential CO2 savings from such a long lifespan? Not really. Then again, it’s not fair to compare the 7-seater model S to the fleet average vehicle, either- it really should be compared against luxury cars- but we’ll keep the 29 mpg car in there just to keep things simple.

So here’s a table, again using the worst case CO2 embodied emissions of 200 kg/kWh:

Distance Driven (miles) Net CO2 Emissions Saved    Net CO2 Emissions Saved (tonnes, 40 g/kWh) (tonnes, 450 g/kWh)        

50,000                                                 3.8                                          -5.8

100,000                                                22                                           2.2

200,000                                                57                                           18

300,000                                                93                                           35

367,500                                                117                                        46          

The new 2019 figures over the 367,500 mile lifetime are a range between 127 tonnes of net CO2 emissions saved using 40 g CO2/kWh electricity and the 61 kg CO2/kWh embodied figure, and 52 tonnes net savings using a 450 g CO2/kWh grid and the 106 kg CO2/kWh embodied figure.  

It is possible to calculate a “break even” grid carbon intensity at which the Model S used in this calculation would start to give net GHG emissions over its lifetime. That would be 750 g CO2/kWh using the high (106 kg/kWh) figure and 770 g CO2/kWh using the low (61 kg/kWh) embodied emissions figure. Grids which have a higher net CO2 intensity than that, will not generate net GHG emissions savings if used to recharge EVs with large packs like this. Does a modern EV save net GHG emissions when recharged from a grid supplied only with coal-fired power plants? No. The emissions there are around 1000 g CO2/kWh from source, giving no net payback from a 70 kWh battery recharged from such a grid over the reasonably expected lifetime of the pack. Note however, that the net reduction in toxic tailpipe emissions is still there in that case. At least a coal plant CAN be fitted with scrubbers and continuous emissions monitoring equipment- nobody is even talking about such measures for ICE vehicles.

Although we haven’t done a full evaluation of the whole EV, everybody agrees that the pack is the biggest issue- the motor and inverter, charger etc. stack up well against the ICE and its subsystems that they replace on an energy and emissions basis. And there’s no comparison when looking at required maintenance, either- no oil, filters or brakes to replace etc. is worth something too.

Some things are made crystal clear in this table:

1)     Big packs take longer to break even on an emissions basis than small ones (obviously!)

2)     The source of grid power matters a lot, both for making the packs and for recharging them

3)     Cycle life matters, but it’s already high enough to permit EVs to NOT need more than one pack over their lifetime. The real risk is if people get bored and throw the car away before its pack is finished.

The column to the right looks worse when comparing against a 52 mpg (4.5 L/100 km) Prius hybrid- its 1.3 kWh pack has far lower embodied energy and yet saves a huge amount of fuel relative to the fleet average ICE vehicle. (Note, a Prius pack is VERY DIFFERENT than a Li ion bulk storage pack in an EV, so its embodied emissions per kWh will be significantly higher as it is a pack designed for power absorption and delivery rather than bulk energy storage)

Does Embodied Energy Matter to EVs? In a Word: YES

…but it doesn’t eliminate the net GHG emissions benefit of EVs!

Obviously, buying EV range or service life that you don’t need is costly- not just for you, but to the planet in emissions terms. Making them bigger still, in a vain attempt to make them equivalent to gasoline powered cars in range per charge terms, is going to make the cars not only more expensive than necessary, but also negate even more of the benefits of the EV over the ICE vehicle. And the figures in both columns are going to improve- a lot- when Tesla opens their solar-powered Gigafactory, which will give them an advantage they can sell relative to their competitors. Remember that around 1/3 of the embodied energy and emissions per the study are from battery production, not raw materials production. 

Let’s not forget another key factor in this discussion: even if the EV were to be GHG neutral relative to an ICE vehicle, we would still get the benefit of having zero tailpipe emissions discharged into the breathing zone of passersby. That alone should be worth something who care about not just the environment in general terms, but also about human health.

Here are some links to other articles I’ve written about batteries and battery materials:

https://www.linkedin.com/pulse/how-much-lithium-li-ion-vehicle-battery-paul-martin/

https://www.linkedin.com/pulse/enough-lithium-cobalt-build-ev-fleet-paul-martin/

https://www.linkedin.com/pulse/part-2-battery-materials-we-dont-need-worry-paul-martin/

https://www.linkedin.com/pulse/part-3-lithium-cobalt-risky-materials-paul-martin/