Samuel Ok. Moore Hello. I’m Samuel Ok. Moore for IEEE Spectrum‘s Fixing the Future podcast. Earlier than we begin, I wish to let you know that you would be able to get the newest protection from a few of Spectrum‘s most vital beats, together with AI, local weather change, and robotics, by signing up for certainly one of our free newsletters. Simply go to spectrum.ieee.org/newsletters to subscribe. The semiconductor trade is within the midst of a serious enlargement pushed by the seemingly insatiable calls for of AI, the addition of extra intelligence in transportation, and nationwide safety issues, amongst many different issues. Governments and the trade itself are beginning to fear what this enlargement may imply for chip-making’s carbon footprint and its sustainability typically. Can we make the whole lot in our world smarter with out worsening local weather change? I’m right here with somebody who’s serving to work out the reply. Lizzie Boakes is a life cycle analyst within the Sustainable Semiconductor Applied sciences and Techniques Program at IMEC, the Belgium-based nanotech analysis group. Welcome, Lizzie.
Lizzie Boakes: Hiya.
Moore: Thanks very a lot for coming to speak with us.
Boakes: You’re welcome. Pleasure to be right here.
Moore: So let’s begin with, simply how massive is the carbon footprint of the semiconductor trade? And is it actually large enough for us to fret about?
Boakes: Yeah. So quantifying the carbon footprint of the semiconductor trade isn’t a straightforward process in any respect, and that’s as a result of semiconductors at the moment are embedded in so many industries. So the obvious trade is the ICT trade, which is estimated to be about roughly 3 % of the worldwide emissions. Nevertheless, semiconductors may also be present in so many different industries, and their embedded nature is growing dramatically. In order that they’re embedded in automotives, they’re embedded in healthcare functions, so far as aerospace and protection functions too. So their enlargement and adoption of semiconductors in all of those completely different industries simply makes it very arduous to quantify.
And the worldwide affect of the semiconductor chip manufacturing itself is predicted to extend as nicely due to the truth that we want an increasing number of of those chips. So the worldwide chip market is projected to have a 7 % compound annual progress price within the subsequent coming years. And allowing for that the manufacturing of the IC chips itself usually accounts for the biggest share of the life cycle local weather affect, particularly for client electronics, for example. This improve in demand for therefore many chips and the demand for the manufacturing of these chips could have a major affect on the local weather affect of the semiconductor trade. So it’s actually essential that we concentrate on this and we establish the challenges and attempt to work in direction of decreasing the affect to attain any of our ambitions at reaching internet zero earlier than 2050.
Moore: Okay. So the way in which you checked out this, it was form of a— it was cradle-to-gate life cycle. Are you able to form of clarify what that entails, what that actually means?
Boakes: Yeah. So cradle to gate right here signifies that we quantify the local weather impacts, not solely of the IC manufacturing processes that happen contained in the semiconductor fab, but additionally we quantify the embedded affect of the entire power and materials flows which can be coming into the fab which can be mandatory for the fab to function. So in different phrases, we attempt to quantify the local weather affect of the worth chain upstream to the fab itself, and that’s the place the cradle begins. So the extraction of the entire supplies that you simply want, the entire power sources. As an example, the extraction of coal for electrical energy manufacturing. That’s the cradle. And the gate refers back to the level the place you cease the evaluation, you cease the quantification of the affect. And in our case, that’s the finish of the processing of the silicon wafer for a particular expertise node.
Moore: Okay. So it stops principally whenever you’ve received the die, nevertheless it hasn’t been packaged and put in a pc.
Boakes: Precisely.
Moore: And so why do you’re feeling like it’s important to take a look at all of the upstream stuff {that a} chip-maker might probably not have any management over, like coal and such like that?
Boakes: So there’s a massive want to investigate your scope via what is named— in greenhouse gasoline protocol, you may have three completely different scopes. Your scope one is your direct emissions. Your scope two is the emissions associated to the electrical energy consumption and the manufacturing of electrical energy that you’ve consumed in your operation. And scope three is principally the whole lot else, and lots of people begin with scope three, all of their upstream supplies. And it does have— it’s clearly the biggest scope as a result of it’s the whole lot else apart from what you’re doing. And I feel it’s essential to coordinate your provide chain so that you simply ensure you’re doing probably the most sustainable answer that you would be able to. So if there are— you may have energy in your buying, you may have energy over the way you select your provide chain. And in case you can manipulate it in a method the place you may have decreased emissions, then that must be accomplished. Usually, scope three is the biggest proportion of the entire affect, A, as a result of it’s one of many largest teams, however B, as a result of there’s quite a lot of supplies and issues coming in. So yeah, it’s mandatory to take a look up there and see how one can greatest cut back your emissions. And yeah, you may have energy in your affect over what you select in the long run, by way of what you’re buying.
Moore: All proper. So in your evaluation, what did you see as form of the most important contributors to the chip fabs carbon output?
Boakes: So with out efficient abatement, the processed gases which can be launched as direct emissions, they’d actually dominate the entire emissions of the IC chip manufacturing. And it’s because the processed gases which can be usually consumed in IC manufacturing, they’ve a really excessive GWP worth. So if you don’t abate them and you don’t destroy them in a small abatement system, then their emissions and contribution to world warming are very massive. Nevertheless, you may drastically cut back that emission already by deploying efficient abatements on particular course of areas, the high-impact course of areas. And in case you try this, then this distribution shifts.
So you then would see that the direct emission– the contribution of the direct emissions would cut back since you’ve decreased your direct emission output. However then the next-biggest contributor could be {the electrical} power. So the scope to the emissions which can be associated to the manufacturing of the electrical energy that you simply’re consuming. And as you may think about, IC manufacturing may be very energy-intensive. So there’s quite a lot of electrical energy coming in, so it’s mandatory then to attempt to begin to decarbonize your electrical energy supplier or cut back your carbon depth of your electrical energy that you simply’re buying.
After which when you try this step, you’d additionally see that once more the distribution adjustments, and your scope three, your upstream supplies, would then be the biggest contributors to the entire affect. And the supplies that we’ve recognized as being probably the most or the biggest contributors to that affect could be, for example, the silicon wafers themselves, the uncooked wafers earlier than you begin processing, in addition to moist chemical compounds. So these are chemical compounds which can be very particular to the semiconductor trade. There’s quite a lot of consumption there, they usually’re very particular and have a excessive GWP worth.
Moore: Okay. So if we may begin with— unpack a number of of these. First off, what are a few of these chemical compounds, and are they typically abated nicely lately? Or is that this form of one thing that’s nonetheless a coming drawback?
Boakes: Yeah. In order that they may very well be from particular photoresists to— there’s a very heavy consumption of primary chemical compounds for neutralization of wastewater, these kinds of issues. So there’s a mix of getting in a excessive embedded GWP worth, which signifies that it takes a really great amount of– or has a really massive affect to supply the chemical itself, otherwise you simply have so much that you simply’re consuming of it. So it might need a low embedded affect, however you’re simply utilizing a lot of it that, in the long run, it’s the upper contributor anyway. So you may have two type of buckets there. And yeah, it might simply be a matter of, it’s important to multiply via the quantities by your embedded emission to see which of them come on prime. However yeah, we see that usually, the wastewater therapy makes use of quite a lot of these chemical compounds only for neutralization and therapy of wastewater on web site, in addition to very particular chemical compounds for the semiconductor trade comparable to photoresists and CMP cleans, these varieties of very particular chemistries which, once more, it’s tough to quantify the embedded affect of as a result of usually there’s a proprietary— you don’t precisely know what goes into it, and it’s quite a lot of problem making an attempt to truly characterize these chemical compounds appropriately. So usually we apply a proxy worth to these. So that is one thing that we would love to enhance sooner or later could be having extra communication with our provide chain and actually understanding what the actual embedded affect of these chemical compounds could be. That is one thing that we actually would want to work on to essentially establish the high-impact chemical compounds and take a look at something we will to cut back them.
Moore: Okay. And what about these direct greenhouse gasoline emission chemical compounds? Are these typically abated, or is that one thing that’s nonetheless being labored on?
Boakes: So there’s fairly, yeah, a considerable quantity of labor going into the abatement system. So we’ve the same old methane combustion of processed gases. There’s additionally now growth in plasma abatement programs. So there are completely different abatement programs being developed, and their effectiveness is kind of excessive. Nevertheless, we don’t have such a very good oversight in the meanwhile on the quantity of abatement that’s being deployed in high-volume manufacturing. This, once more, is kind of a delicate matter to debate from a analysis perspective whenever you don’t have perception into the fab itself. So asking explicit questions on how a lot abatement is deployed on sure instruments isn’t such simple information to return throughout.
So we regularly go together with fashions. So we apply the IPCC Tier 2c mannequin the place, principally, you calculate the direct emissions by how a lot you’ve used. So it’s a mathematical mannequin based mostly on how a lot you’ve consumed. There’s a mannequin that generates the quantities that might be emitted straight into the ambiance. So that is the mannequin that we’ve utilized. And we see that, yeah, it does correlate typically with the top-down reporting that comes from the trade. So yeah, I feel there’s quite a lot of method ahead the place we will begin evaluating top-down reporting to those bottom-up fashions that we’ve been producing from a type of analysis perspective. So yeah, there’s nonetheless quite a lot of work to do to match these.
Moore: Okay. Are there any explicit nasties by way of what these chemical compounds are? I don’t assume individuals are accustomed to actually what comes out of the smokestack of chip fab.
Boakes: So one of many highest GWP gases, for example, could be the sulfur hexafluoride, so SF6. This has a GWP worth of 25,200 kilograms of CO2 equal. So that actually signifies that it has over 25,000 occasions extra damaging results to the local weather in comparison with a CO2, so the equal CO2 molecule. So that is extraordinarily excessive. However there’s additionally others like NF4 that— these even have over 1,000 occasions extra damaging to the local weather than CO2. Nevertheless, they are often abated. So in these abatement programs, you may destroy them they usually’re now not being launched.
There are additionally efforts going into changing excessive GWP gases comparable to these that I’ve talked about to make use of options which have a decrease GWP worth. Nevertheless, that is going to take quite a lot of course of growth and quite a lot of effort to enter altering these course of flows to adapt to those new options. And it will then be a sluggish adoption into the high-volume fabs as a result of, as we all know, this trade is kind of inflexible to any adjustments that you simply counsel. So yeah, it is going to be a sluggish adoption if there are any options. And for the meantime, efficient abatement can destroy rather a lot. However it might actually be having to make use of and actually have these abatement programs on these high-impact course of areas.
Moore: As Moore’s Regulation continues, every step or manufacturing node might need a special carbon footprint. What have been a few of the massive tendencies your analysis revealed concerning that?
Boakes: So in our mannequin, we’ve assumed a relentless fab operation situation, and because of this we’ve assumed the identical abatement programs, the identical electrical carbon intensities, for the entire completely different expertise nodes, which– yeah. So we see that there’s a basic improve in complete emissions beneath these assumptions, and we double in complete local weather affect from N28 to A14. So once we evolve in that expertise node, we do see it doubling between N28 and A14. And this may be attributed to the elevated course of complexity in addition to the elevated variety of steps, in course of steps, in addition to the completely different chemistries getting used, completely different supplies which can be being embedded within the chips. This all contributes to it. So typically, there is a rise due to the method complexities that’s required to essentially attain these aggressive pitches within the extra superior expertise nodes.
Moore: I see. Okay. In order issues are progressing, they’re additionally type of getting worse in some methods. Is there something—?
Boakes: Yeah.
Moore: Is that this inevitable, or is there—?
Boakes: [laughter] Yeah. When you make issues extra sophisticated, it can most likely take extra power and extra supplies to do it. Additionally, whenever you make issues smaller, it’s worthwhile to change your processes and use– yeah, for example, with interconnect metals, we’ve actually reached the bodily limits typically as a result of it’s gotten so small that the bodily limits of actually conventional metals like copper or tungsten has been reached. And now they’re in search of new options like ruthenium, yeah, or platinum. Several types of metals which– once more, if it’s a platinum group steel, in fact it’s going to have a better embedded affect. So once we hit these limits, bodily limits or limits to the present expertise and we have to change it in a method that makes it extra sophisticated, extra energy-intensive— once more, the transfer to EUV. EUV is an especially energy-intensive device in comparison with DUV.
However an fascinating level there on the EUV matter could be that it’s actually vital to maintain this holistic view as a result of despite the fact that transferring from a DUV device to an EUV device, it has a big soar in power depth per kilowatt hour. The facility depth of the device is far larger. Nevertheless, you’re in a position to cut back the variety of complete steps to attain a sure deposition or edge. So that you’re in a position to total cut back your emissions, otherwise you’re in a position to cut back your power depth of the method stream. So despite the fact that we make all these adjustments and we would assume, “Oh, that’s a really highly effective device,” it may go and reduce down on course of steps within the holistic view. So it’s all the time good to maintain a type of life cycle perspective to have the ability to see, “Okay, if I implement this device, it does have a better energy depth, however I can cut back half of the variety of steps to attain the identical consequence. So it’s total higher. So it’s all the time good to maintain that type of holistic view once we’re doing any kind of sustainability evaluation.
Moore: Oh, that’s fascinating. That’s fascinating. So that you additionally checked out— as form of the nodes get extra superior and processes get extra advanced. What did that do to water consumption?
Boakes: Additionally, so once more, the variety of steps in the same sense. When you’re growing your variety of course of steps, there could be a rise within the variety of these moist clear steps as nicely which can be usually the high-water-consumption steps. So in case you have an elevated variety of these explicit course of steps, you then’re going to have a better water consumption in the long run. So it’s simply based mostly on the variety of steps and the complexity of the method as we advance into the extra superior expertise nodes.
Moore: Okay. So it appears like complexity is type of king on this area.
Boakes: Yeah.
Moore: What ought to the trade be specializing in most to attain its carbon objectives going ahead?
Boakes: Yeah. So I feel to begin off, it’s worthwhile to consider the biggest contributors and prioritize these. So in fact, in case you’re trying on the complete affect and we’re a system that doesn’t have efficient abatement, then in fact, direct emissions could be the very first thing that you simply wish to attempt to concentrate on and decreasing, as they’d be the biggest contributors. Nevertheless, when you begin transferring right into a system which already has efficient abatement, then your subsequent goal could be to decarbonize your electrical energy manufacturing, go for a lower-carbon-intensity electrical energy supplier, so that you’re transferring extra in direction of inexperienced power.
And on the similar time, you’d additionally wish to attempt to goal your high-impact worth chain. So your supplies and power which can be coming into the fab, it’s worthwhile to take a look at those which can be probably the most extremely impacting after which attempt to discover a approach to discover a supplier that does a type of decarbonized model of the identical materials or attempt to design a method the place you don’t want that sure materials. So not essentially that it needs to be accomplished in a sequential order. After all, you are able to do all of it in parallel. It will be higher. So it doesn’t should be one, two, three, however the concept and the prioritizing comes from concentrating on the biggest contributors. And that might be direct emissions, decarbonizing your electrical energy manufacturing, after which your provide chain and looking out into these high-impact supplies.
Moore: Okay. And as a researcher, I’m certain there’s information you’d like to have that you simply most likely don’t have. What may trade do higher about offering that type of information to make these fashions work?
Boakes: So for lots of our quite a lot of our scope three, in order that upstream, that cradle-to-fab, let’s name it— these impacts. We’ve had to make use of rather a lot— we needed to rely rather a lot on life cycle evaluation literature or life cycle evaluation databases, which can be found via buying, or typically in case you’re fortunate, you may have a free database. So I might say– and that’s additionally as a result of my function in my analysis group is extra that LCA and upstream supplies and quantifying the environmental affect of that. So from my perspective, I actually assume that this trade must work on offering information via the availability chain, which is standardized in a method that folks can perceive, which is product-specific in order that we will actually allocate embedded affect to a particular product and multiply that via then by our stock, which we’ve information on. So for me, it’s actually having a standardized method of speaking sustainability affect of manufacturing, upstream manufacturing, all through the availability chain. Not solely tier one, however all the way in which as much as the cradle, the start of the worth chain. So that is something– and I do know it’s evolving and it is going to be sluggish, and it does want quite a lot of cooperation. However I do assume that that might be very, very helpful for actually making our work extra lifelike, extra consultant. After which individuals can depend on it higher once they begin utilizing our information of their product carbon footprints, for example.
Moore: Okay. And talking of form of your work, are you able to inform me what imec.netzero is and the way that works?
Boakes: Yeah. It is a internet app that’s been developed in our program, so the SSTS program at IMEC. And this internet app is a method for individuals to work together with the mannequin that we’ve been constructing, the LCA mannequin. So it’s based mostly on life cycle evaluation, and it’s actually what we’ve been speaking about with this cradle-to-gate mannequin of the IC-chip-manufacturing course of. It tries to mannequin a generic fab. So we don’t essentially level to any particular fab or course of stream from a sure firm. However we attempt to make a really generic trade common that folks can use to estimate and get a extra lifelike view on the fashionable IC chip. As a result of we seen that, in literature and what’s out there in LCA databases, the semiconductor information is extraordinarily outdated, and we all know that this trade strikes in a short time. So there’s a large hole between what’s taking place now and what’s going into your telephones and what’s going into the computer systems and the LCA information that’s out there to attempt to quantify that from a sustainability perspective. So imec.netzero, we work with all of— we get pleasure from being related with the trade and now a place in IMEC, and we’ve a view on these extra superior expertise nodes.
So not solely do we’ve fashions for the nodes which can be being generated and produced right this moment, however we additionally predict the long run nodes. And we’ve fashions to foretell what’s going to occur in 5 years’ time, in 10 years’ time. So it’s a extremely highly effective device, and it’s out there publicly. We’ve a public model, which is a limited– it has restricted performance compared to this system companion model. So we work with our program companions who’ve entry to a way more sophisticated and, yeah, deep method of utilizing the net app, in addition to the opposite work that we do in our program. And our program companions additionally contribute information to the mannequin, and we’re continually evolving the mannequin to enhance all the time. In order that’s a little bit of an summary.
Moore: Cool. Cool. Thanks very a lot, Lizzie. I’ve been talking to Lizzie Boakes, a life cycle analyst within the Sustainable Semiconductor Applied sciences and Techniques Program at IMEC, the Belgium-based nanotech analysis group. Thanks once more, Lizzie. This has been implausible.
