Nuclear fission releases far more energy per kilogram of fuel than conventional hydrocarbon sources. However, proposed nuclear rocket engines like NERVA (https://en.wikipedia.org/wiki/NERVA), while having potentially 2x the $I_{sp} $ as chemical rockets (800 - 1000 seconds), have never been shown (in testing) to generate a T/W ratio high enough to be practically usable for payload launches to LEO and beyond. NERVA for example was supposed to be around 7.5:1. Why is this?

Compare these statistics to the state-of-the-art chemical rockets: the SpaceX Merlin engines achieve close to a MN of thrust at a T/W ratio of ~150. And yet nuclear rocket engines which take advantage of a much more energetic reaction, and could thus presumably pass much more energy for a given amount of mass into the reaction mass, thereby generating potentially extremely high exhaust velocities, cannot generate high T/W ratios?

I understand that nuclear engines have a lot more complexity to them in the form of the reactor itself (heavy!) and shielding. But, the energy density of a fission reaction is not simply 2-3x higher than a chemical reaction, it's 16,000+x (depending on the enrichment level of the fuel) higher. It seems to me that intuitively, the massive energy gain should be able to compensate for the extra complexity and weight needed. But in practice this has not shown to be the case for proposed nuclear rockets (none have actually flown).

An exception (not the only one) to this fact is the proposed Project Orion, which uses literal nuclear bombs and generates extremely high $I_{sp} $ and MNs of thrust. Why is this design so much better at "tapping" the energy in the fission reaction?

What accounts for these facts?

  • 15
    $\begingroup$ This question talks about nuclear rockets as if they exist. I would suggest re-wording it. That is not just a question of style but also relevant as some of your questions (presumably) are about certain proposed designs and you might want to explain which design you are referring to and where those numbers come from. $\endgroup$ Jan 2, 2020 at 20:30
  • 4
    $\begingroup$ Minor nitpick: the thrust to weight ratio is a dimensionless number. So a "ratio of ~150 T/W" should read "ratio of ~150". $\endgroup$ Jan 2, 2020 at 20:32
  • 2
    $\begingroup$ Thrust is also a function of mass flow rate, and hydrogen isn't very massive. $\endgroup$
    – John Bode
    Jan 3, 2020 at 17:56
  • 1
    $\begingroup$ @user2705196 Dimensionless numbers are still "about" something. We cannot reasonably say "There are 24" (hours/day). Or 168 or 8765 or ... . | Otherwise, the PASSAGE in question reads: "Why is this? Compare these statistics to the state-of-the-art chemical rockets: the SpaceX Merlin engines achieve close to a MN of thrust at a ratio of ~150." $\endgroup$ Jan 4, 2020 at 9:41
  • $\begingroup$ @RussellMcMahon: You're right. And even dimensioned quantities need to be about something: the dimensions of torque and of energy are the same, but they're not really the same thing. The units of torque are 'really' joules per radian, but radians are dimensionless. $\endgroup$
    – user21103
    Jan 4, 2020 at 13:21

3 Answers 3


At their core nuclear rockets working by heating a working fluid and running it out a nozzle are still constrained by the same physics as a chemical rocket where exhaust temperature cannot be much higher than the melting point of nozzle (cooling the nozzle lets you cheat a bit), putting limits on how much energy can go into the fuel.

Nuclear rockets get some advantages from the fact that exhuast is pure H, which boosts ISP, having a single tank structure and less exciting chemistry in the engine (no oxidisers).

Where they lose out on is the method of adding heat to the working fluid. For a chemical rocket this is pretty straight forward, pump in your chemicals and let them react and the heat happens without further engineering.

For a nuclear rocket you need to come up with a method of transferring all that nuclear energy into heat in fuel, and that generally means lots of surface area which translates into mass and therefore lower thrust/weight.

Orion, and some other hypothetical magnetically contained systems bypass the solid nozzle and the related constraints. In the case of Orion the 'nozzle' could be described as the plasma going in all directions other than the pusher plate and pretty much unconstrained by temperature (if somewhat inefficient).

  • $\begingroup$ I think it bears mentioning that gas-core NTR and arcjet engines can have significantly greater Isp than NTRs while still dealing with the nozzle heat limit. $\endgroup$
    – ikrase
    Jan 3, 2020 at 1:30
  • $\begingroup$ @ikrase, gas-core certainly changes a number of the assumptions, by design running large parts of the system not just molten but above boiling and gets pretty much ideal heat exchange with the reaction mass. Also gives a good data point on just how extreme you have to get to produce reasonable thrust out of fission. $\endgroup$ Jan 3, 2020 at 7:32
  • $\begingroup$ @ikrase are there any designs of those which have high thrust though? A closed-cycle GCNTR may as well be magic, but I don't see anyone suggesting that it would have a T/W even as high as 10, and arcjets are much worse in that regard. $\endgroup$ Jan 3, 2020 at 10:23
  • $\begingroup$ There's also the minor detail about shielding to prevent irradiating the entire payload necessarily being rather heavy stuff. The few designs I've seen that do without the shielding perform rather "better" for at least some definitions of the word. $\endgroup$
    – Perkins
    Jan 3, 2020 at 20:39
  • $\begingroup$ @GremlinWranger Your answer is extremely illuminating, and depressing (because the most "efficient" way of getting a lot energy from mass we currently have makes it difficult to get thrust out of it). $\endgroup$
    – user34435
    Jan 4, 2020 at 5:25

The one exception to this fact is Project Orion

Not quite. Project Timberwind was a solid-core NTR using a pebble-bed reactor design that combined high Isp with a moderate T/W of 30. The DUMBO NTR used a quite different core design, and had predicted T/W ratio of 70. Still somewhat shy of a good modern chemical rocket, but with a much better specific impulse. Whether those numbers would have held up in real world production rockets is anybody's guess, but they were certainly beefier designs than NERVA.

The big problem is that not many people really think that lighting up a solid core NTR in a biosphere that they live in is a very good idea. People don't even like the regular kind of nuclear reactors that don't directly vent into the air, and this is something else entirely. As a result, there is/was little political will to develop high T/W nuclear rockets because they'd be mostly useful at taking off from the Earth, and that wasn't a mission anyone was going to be happy with. The Project Rho link for dumbo above suggested that NASA just didn't care about high T/W NTRs as a result; they just concentrated on the NERVA design (which was probably the most mature... real working models were even tested as part of Project Rover) before that ultimately got shelved as well.

  • $\begingroup$ Do designs proposed like NERVA, Timberwind, DUMBO, etc., always vent radioactive waste products into the atmosphere? Or is there a way to design them to be "closed cycle" (like Westinghouse was experimenting with in the ANP program: en.wikipedia.org/wiki/Aircraft_Nuclear_Propulsion)? Shouldn't the heat exchange mechanism (which is slow and thus is a barrier to high thrust - according to @GremlinWranger) prevent radioactive waste products from entering exhaust? $\endgroup$
    – user34435
    Jan 4, 2020 at 5:31
  • 2
    $\begingroup$ @user34435, there are several potential sources of radioactivity in the exhaust but even with perfect containment the working fluid being heated must be close enough to the fission process that it is being struck by high energy particles, so your rocket is exhausting exotic Hydrogen isotopes, even if the rest is contained. This is why nuclear power uses a primary and secondary loops, since the primary gets irradiated over time. $\endgroup$ Jan 4, 2020 at 5:45
  • $\begingroup$ @user34435 the aircraft reactor design has rather different T/W requirements to a rocket, I note. Closed cycle gas core rockets have been proposed, but they're Quite Complex (to say the least). They avoid the heat exchanger inefficiency issues by being really hot, but even then they would have a much, much lower Isp than is predicted of open cycle gas core rockets (2000s vs ~8000s, sort of thing). $\endgroup$ Jan 4, 2020 at 10:11
  • $\begingroup$ @user34435 Ideally, no. Realistically, fuel elements inevitably erode and leak. Ironically, a "nuclear lightbulb" might actually be better at this, but nuclear lightbulbs, let alone ones with thrust/weight over 1, are not really demonstrated. $\endgroup$
    – ikrase
    Jan 9, 2020 at 6:04

The reason is that you have to carry a fission reactor, which is large & heavy, and a heat exchanger which is also probably fairly heavy. This is all especially heavy if you want to avoid the working fluid from ending up radioactive, and also if you want to avoid the payload from getting irradiated.

If you're willing to forego some or all of this protection you can make things a bunch lighter and more practical. For instance you can use a single-loop heat-exchanger (so the working fluid is what cools the reactor directly). This means the working fluid almost certainly ends up radioactive, and this makes such designs at best unpopular if you use them in the atmosphere.

A system like this still has to deal with lifting the fission reactor & rather lighter heat exchanger (and I guess such a system is pretty exciting if something happens to the flow of working fluid). Orion avoided even that: it dumped essentially a whole bunch of small fission reactors (bombs) out of the back of the rocket and used them both as the source of energy and as the working fluid. This was efficient, but meant you were dumping really awful things into the atmosphere as fission bombs are a lot dirtier than fission reactors in general.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.