Jeffrey LewisKemp: Setting a Goal for Iran Talks

A guest post by R. Scott Kemp.

Friend of Wonk R. Scott Kemp is an assistant professor of nuclear science and engineering at MIT, where he directs the Laboratory for Nuclear Security and Policy. Here, he asks what sort of negotiated outcome with Iran would suffice to address the threat of nuclear breakout. The charts in this post, prepared by the University of Maryland’s Steve Fetter, could help negotiators design an agreement. Scott’s last guest post at ACW was in June 2013.

Well, this is the moment we’ve been waiting for.  Both Iran and the United States have been making positive sounds about the nuclear issue. Obama says he has made Iran a top diplomatic priority for the United States. According to diplomats who met recently with Iranian Foreign Minister Zarif, Iran will consider limits on the number of operating centrifuges and stocks of LEU. The E3+3 and Iran have a meeting scheduled for Geneva on October 15 and 16.

Now it’s time to get down to brass tacks.

While Zarif’s offer of caps is welcome, what really matters is the level of confidence that can be gained from an agreement.  To a large extent, that can only come by increasing the time it would take Iran to make a bomb, which in turn depends on the size of Iran’s nuclear program.

Why does the size of the program matter? There is a fundamental problem with safeguards at enrichment facilities: even if the IAEA could instantaneously detect an attempt to make a bomb, it would take some non-trivial length of time to respond, especially if there is to be a political response and not just a last-ditch military strike. Iran’s “breakout” time needs to be considerably longer than U.S. Central Command’s bare-minimum response time.

The problem is that Iran has already amassed a considerable enrichment capability; by some estimates, it could produce a bomb’s quantity of HEU in less than two months, even if Iran doesn’t use any of its 20%-enriched material.  Merely capping the program at the present size isn’t going to provide enough assurance to restrain the hawks back in the United States. Some rollback of the program, even if not done immediately, is really the only path to confidence and stability.

In this post, I give you the tools to decide just how much rollback is needed by enabling you to calculate the breakout time based on some retained centrifuge-enrichment capacity and the residual stocks of enriched uranium. It’s a Choose Your Own Adventure for nuclear diplomacy.

Where Does Iran Stand Today?

In adding up the numbers, I’m going to set aside the 20%-enriched uranium program and focus exclusively on the 3.5%-enriched uranium program located at FEP, the underground facility in Natanz. I make this simplification because I believe that some closure of the 20% program can be negotiated and the 20%-enriched uranium stockpile can either be converted to fuel elements or removed from the breakout calculation through one of many available technical options.

According to the last IAEA report, at Natanz as of August 24, 2013, Iran had 89 fully installed cascades of IR-1s and was working on 37 more.  Of those 89 IR-1 cascades, 54 were operating. Each IR-1 cascade contains 174 centrifuges and each IR-1 is about 0.9 SWU/year. That gives a capacity of:

IR-1 (operating):  9,396 centrifuges =  8,500 SWU/year

IR-1 (installed): 15,486 centrifuges  =  14,000 SWU/year

The IAEA also reports that Iran has six cascades of IR-2ms installed and is working on 12 more. None have been fed with uranium and we do not know for sure the exact layout of these cascades or how efficient they will be.  However, we could assume the IR-2m cascades also contain 174 machines. We could also assume, consistent with my calculations, that the IR-2m produces about 5 kg-SWU/year/centrifuge. Correcting for some cascade losses, a reasonable net performance would be around 4.7 SWU/year/installed IR-2m centrifuge.  That gives:

IR-2m (potential): 1,044 centrifuges =  4,900 SWU/year  = 5,450 IR-1 equivalents

Notice that this tiny installation of IR-2m centrifuges has more than half the potential of all the currently operating centrifuges at Natanz.  That underscores an obvious but important point: we cannot negotiate simply on the basis of numbers of centrifuges. We must base our computation on the maximum potential separative capacity installed, measured in units of SWU/year. There is some nuance to how one actually goes about accounting for and verifying this number.

In addition, Iran has “stored” separative work in the form of enriched uranium. Thus, from the breakout perspective, the residual stockpile of LEU matters because there is a direct trade-off between the number of centrifuges required and the size of the LEU stockpile available. The IAEA report indicated that Iran had a net accumulation of 5,576 kg of low-enriched uranium (not kg of UF6), which we assume has an enrichment of around 3.5%.

How Long to the Bomb?

Well, let me instead tell you how long to a notional nuclear weapon’s quantity of HEU, rather than to a fully fabricated bomb.  The below charts were produced by Steve Fetter as a result of some discussions we’ve been having.  Steve based these on the standard of one IAEA Significant Quantity (25 kg of uranium-235 contained within HEU), which is reasonable for a first-generation implosion design with manufacturing losses. Some believe Iran would need more than a single bomb; I’m of the view that a single bomb is significant—especially one based on HEU.

Centrifuge performance is based on my estimate of 0.9 SWU/IR-1/year.  We have attempted to include the effects of adjusting the tails to optimize the utilization of the LEU stockpile. However, we have ignored second-order effects such as the time required to adjust the cascades, losses from suboptimal cascade configurations or off-optimal centrifuge operation, and machine crashes associated with working on the cascades.  These problems will extend the breakout time, but the calculations below are the right place to start because they give the most efficient credible case.

These charts show the same thing under different presentations. This first chart shows the trade-off between time and centrifuges under assumptions of different LEU stockpile sizes:


This second chart lets you pick a breakout time and trade off between LEU stockpile size and centrifuge plant capacity:


What Constitutes A Workable End State?

Whatever makes you comfortable.  Policymakers will have to decide how many months of advance warning they can tolerate–or rather, how few.  From one perspective, a relatively short period is acceptable if it allows for swift and decisive military action.  I am of another view: even if military action is swift, it will only be temporarily decisive, resulting in Iran’s withdrawal from the NPT and pursuit of the bomb at a newly constructed clandestine location.  More breathing room is needed to avoid such a precarious balance, preferably closer to a year. That much time would enable reinstitution of sanctions of the harshest type and hopefully lead to a peaceful outcome before resorting to military means.


  1. Nicoullaud (History)

    Thank you, Mr.Kemp. As you rightly say yourself, one has to add to any of these breakout estimated times the time necessary to adjust the cascades to a high enrichment campaign. What is your personal evaluation for this? And of course, at the other end, one has to add also the time necessary to transform the product into uranium metal, to give it the required shape and to put together all the components of an explosive device. How much additionnal time would this process require, assuming that Iran already masters all the technologies of bomb engineering?

    • R. Scott Kemp (History)

      In response to Nicoullaud:

      The time needed to adjust cascades can be very small. One option is to cut out centrifuges rather than fully re-pipe cascades. Under the right conditions, that strategy would increase breakout times by only about 5%. It’s very situation specific, but this is not a major factor.

      I do not believe one should include time for conversion to metal, casting, etc. All that can occur offsite at a secret location that is effectively invulnerable to the military strike meant to disable the centrifuge plant. The requirement is not that Iran posses the bomb before the military strike. The requirement is only that the HEU be safely produced and hidden away so that a bomb can eventually be made.

  2. Paul Stokes (History)

    This is very useful information regarding the time it would take Iran to produce weapon quantities of enriched uranium. It seems to me that limiting the LEU available to an amount that would provide significant lead time to having a Significant Quantity of HEU would be incompatible with a nuclear power program. If so, I wouldn’t expect Iran to accede to such a limit.

    In order to detect a breakout, I would expect there would need to be an inspection regime operating cooperatively with various national intelligence programs that would be designed to detect, not only diversion of enriched uranium from peaceful applications, but also weaponization activities, such as a nuclear explosives test site, explosive test sites that could be used to test implosion of heavy metals, the presence and use of furnaces capable of forming spherical uranium parts, machine shops where machining of uranium parts could occur, etc.

  3. AEL (History)

    Now do a similar analysis calculating the amount of enrichment capacity that Iran would need if it produced a substantial amount of its electricity via nuclear energy.

    Then check to see if there is any overlap.

    I.e. if Iran generates 10% of its electricity via nuclear power (and self creates the fuel to power the needed reactors) would it automatically blow past all of these breakout curves?

    If so, then I can’t see how Iran would agree to permanently reduce their enrichment capacity to a level where civilian nuclear energy would be stuck at “toy” levels.

  4. Anon (History)

    It’s probably incompatible with a domestic fuel fabrication capability. That’s not the same thing as a nuclear power program.

    • SQ (History)

      Yet something has to happen with the enriched material if enrichment continues and if the size of the 3.5%-enriched stockpile is to stay constrained. That would presumably be fuel fabrication.

  5. Bahram Doostdar (History)

    Iran isn’t going to accept an agreement that would be humiliating. To forgo a civilian fuel fabrication capability would be humiliating. If the Euro-Americans want an agreement, they need to be reasonable, compromise, and recognize Iran’s right to a real civilian enrichment program.

    While the sanctions are painful, they will not bring Iran to its knees. Sanctions have a way of becoming porous over time. Think back to the draconian sanctions on Iraq in the 90s.

    • Fred (History)

      There are no limits on enrichment either in the NPT or the Safeguards agreement.

      So if limits are being negotiated then implicitly Iran is being held to a higher standard than that required by law. If so, Iran should either be asked to leave the NPT or leave the NPT of its own volition.

      A better idea is to allow unlimited enrichment but offer carrots to Iran to ratify the AP.

  6. Eric Petersen (History)

    The latest IAEA report put Iran’s inventory of 3.5% at seven-and-a-fraction tons. I’ve read that the initial fuel load at Bushehr weighed 82 tons, but it wasn’t stated whether this number was “uranium only” or included fuel rods, assemblies etc. If memory serves (doubtful!), LWR reactors have to change out half (?) their fuel load every 18-24 months; if so, Iran may soon need an additional 41 tons of LEU. Like the initial fuel load, they could get this material from Russia, but one assumes Iran has built its enrichment capacity to become self sufficient, and has plans for additional nuclear power reactors. Long winded way of asking how much LEU do they need simply to sustain Bushehr? PS: How much 20% do they need for the medical isotope reactor? I’ve seen comments that Iran “now has enough” but no hard data.

  7. Milton Hoenig (History)

    You say, “Each IR-1 cascade contains 174 centrifuges,” but Witt, Migliorini, Albright, and Wood, “Modeling Iran’s Tandem Cascade Configuration,…” state, “Approximatly 30 of Iran’s roughly 54 single cascades at Natanz are 17-stage, 174 machine cascades; the rest are older 15-stage, 164 machine cascades.”

  8. Jeannick (History)

    In a spirit of good fellowship ,
    maybe the Us could build a nuclear power plant
    lifting the embargo on the 20% medical plates for the Teheran university hospital would be nice

  9. Rob Goldston (History)

    Hi Scott!

    Here is a rough answer to the question that seems to be bothering a number of your readers. A modern 1000 MWe fission power plant, operating 90% of the time, requires about 125,000 – 140,000 kg SWUs per year of enrichment capability, depending on various factors. It takes about 4500 – 5000 kg SWUs to make the sort of bomb Steve is referring to, depending on various factors, as you can also see from the left-hand-end of the top curve in the first figure. Thus a facility that is sized to support 1 power plant can instead make the fuel for about 30 bombs per year. Or one every 12 days.

    And if Iran instead starts from 4.0% enriched U that they have on hand, only about 1400 kg SWUs are needed, a bomb about every 4 days.

    This evidently presents a problem if you want many months to re-impose biting sanctions after detecting breakout, and you eventually also want to grant Iran the right to provide fuel for its reactor(s), as a member in good standing of the NPT.

    I imagine that a deal will require a very effective version of the Additional Protocol, renouncement of enrichment above 5%, and on-site monitoring 24/7. The P5+1 response to a verified measurement of >5% enrichment should not be stated explicitly nor even subtly implied, but also should not be expected to take much time.

    BTW, a strengthened AP, enrichment < 5%, and 24/7 monitoring would be good things at all enrichment plants, not just those in Iran.

    – Rob