6 ms·

I actually work directly on this problem as part of my PhD. I work in designing simple "closure models" for Hall thruster anomalous transport that can be implem

I actually work directly on this problem as part of my PhD. I work in designing simple "closure models" for Hall thruster anomalous transport that can be implemented in simulations to improve how well they match with experiment. It's a really interesting problem, and like the OP alludes to, it probably comes down to turbulence and plasma instabilities which are really hard to measure and even harder to include self-consistently in a device-scale simulation. The really tricky thing is that the simulations show that basically every aspect of Hall thruster performance is extremely sensitive to the choice of anomalous transport parameter in certain parts of the discharge, but completely insensitive to it in other parts of the discharge. this makes validating models difficult, as many different models could agree in the small region of the discharge where it matters, but completely disagree in other parts of the discharge, and we would have little chance of discriminating between them.

sampo 4y ago
> I work in designing simple "closure models" What is the order of the turbulence closure models used in modern work? 1.5? 2? 2.5?
- archermarks 4y ago
  Depends on what you mean by order i suppose. The simplest are algebraic (zero-order) models where we prescribe the anomalous transport as an explicit function of other plasma parameters. More complex ones are coupled systems of hyperbolic PDEs, which are typically first-order in space and time. If you are referring to the order of accuracy of the simulations, that's typically 1st to 4th order, depending on method used.
  - johnbcoughlin 4y ago
    I answered the GP based on my understanding of "closure" meaning different things for moment models vs turbulence. I think they're referring to energy partition scaling in saturated turbulence.
    - archermarks 4y ago
      Yes, thank you. It's a closure problem in the former sense.
  - sampo 4y ago
    > Depends on what you mean by order i suppose. I mean, in the context of Reynolds-averaged Navier–Stokes equations [1], what sort of turbulence model [2] you are using. But apparently you are not working in this context at all, so maybe my question doesn't make sense. [1] https://en.wikipedia.org/wiki/Reynolds-averaged_Navier%E2%80%93Stokes_equations https://en.wikipedia.org/wiki/Reynolds-averaged_Navier%E2%80... [2] https://en.wikipedia.org/wiki/Turbulence_modeling https://en.wikipedia.org/wiki/Turbulence_modeling
    - archermarks 4y ago
      Appreciate the clarification. I am familiar with RANS modeling to a reasonable extent but not with a lot of the modern developments. We're still kind of at the stage of trying to find something that works as well as Prantdl's mixing length model, or maybe a k-epsilon model. Additionally, while our processes are turbulent, the turbulence differs significantly from classical turbulence. For instance, turbulence in Hall thrusters is likely to be governed by an inverse energy cascade, where small disturbances magnify to larger and larger scales and the dissipation occurs via convection at device-scale instead of by diffusion at kolmogorov scales.
- johnbcoughlin 4y ago
  "closure" in the kinetics to fluid model context refers to a closed expression for the heat flux or anisotropic pressure in terms of the lower order fluid variables. You aren't really looking at a turbulence closure in the sense you're getting at in this context--for these hyperbolic models it's more about the transport than about energy partition.
mrfusion 4y ago
AMA? How much can these thrusters be scaled up? Or does it make more sense to stick more smaller ones on a craft if we need more thrust? Bonus question: how would you design a planet scale thruster? (Asking for a friend)
- belter 4y ago
  Don't tell him! He wants to move us away! https://en.wikipedia.org/wiki/The_Wandering_Earth https://en.wikipedia.org/wiki/The_Wandering_Earth
- vutnlushaetoc 4y ago
  These thrusters need fuel to move. That ends the fun on this line of inquiry. There is no free ride (unless you sail around very slowly, but this is more like sucking Earth's wake in style)
- archermarks 4y ago
  To scale up thrusters you can simply make them bigger, but that leads to a lot of wasted space as the thrust is only produced by a ring-shaped region of the device. To improve thrust density, you can nest these inside each other (see this page on my research group's website https://pepl.engin.umich.edu/project/high-power-hall-thrusters/ https://pepl.engin.umich.edu/project/high-power-hall-thruste...). More recently, we have been looking at running smaller thrusters at higher power densities, which was previously impractical due to the way older thrusters shaped their magnetic fields. Nowadays we can design thrusters such that very little power is lost to the walls so high power density operation is looking more feasible. Per the second question, you would need a very efficient thruster to limit the amount of mass needed as propellant. If you have an entire planet, you might look into using a photon rocket. Alternatively, you can try a massive solar sail or magnetic sail but I'm not an expert on these things.
  - mrfusion 4y ago
    Thanks! Could super conductors help increase efficiency or density? Why or why not? Could this be a viable space to launch a start up? With launch costs coming down there could be a lot of demand for more efficient thrusters.
vutnlushaetoc 4y ago
> extremely sensitive to the choice of anomalous transport parameter in certain parts of the discharge, but completely insensitive to it in other parts of the discharge. When you find it, will it not likely be a secondary electromagnetic condition is affecting the permittivity and permissibility of materials in a way that was not previously anticipated? Like, this problem is "understood", right? You know where to look and likely what it will be? The fact Hall Thrusters work with fuel of very small mass, it doesn't seem to change the nature of the problem: detonation and fluid (plasma) dynamics.
- archermarks 4y ago
  The "material" in this case is vacuum and plasma. Yes, something small and electromagnetic is affecting the conductivity of the plasma, and we have an idea what it is, but that doesn't mean we know how to simulate it or predict its effects on the thruster. The fuel actually has quite high mass compared to a typical rocket engine. We use Xenon (131.293 g/mol) or Krypton (~83 g/mol), compared to air which is (~29 g/mol).
  - vutnlushaetoc 4y ago
    > and we have an idea what it is I mean, it's on the phrase diagram. The answer will be the statistics of stable orbits at heat and pressure. Any gradient or annulus in a phase diagram is going to mean that there will be potential for rapid "phase" changes in all directions. Isn't this all just to say that you need instruments, instruments, instruments? You know where to look, what it is doing and why (you have guesses, you'll 'get it' when you see the final answer). Is this blog post not mystifying it instead of saying that's it's cool and we know where to look to make it better?
    - archermarks 4y ago
      I'm really not following what you're talking about. What is the phase diagram you're referring to?
      - 4y ago
madaxe_again 4y ago
Maybe I’m missing something, but couldn’t models be validated by building the things and seeing how they measure up in reality? Or is cost the limitation? I have no idea how exotic the required materials are.
- archermarks 4y ago
  We build them all the time. The difficulty is measuring the right regions of the plasma with the right precision. The plasma is pretty sensitive to perturbations so you can't always just stick a probe in them. This often restricts us to using noninvasive measurements like laser diagnostics, but because the plasma is quite rarified, it can be difficult to get a good signal to noise ratio. Additionally, the laser diagnostics can be restricted in what frequencies (spatially or temporally) they can measure, depending on the technique employed. Lastly, even if we fully understood the process, modeling it would remain a challenge (for all the same reasons that turbulence modeling is, despite the fact that turbulence is much better understood)
- tel 4y ago
  This tweet addresses your question, at least in part https://twitter.com/lougrims/status/1516013558475235335 https://twitter.com/lougrims/status/1516013558475235335
ncmncm 4y ago
Can you say something about the difference you experience between "plasma fluid dynamics" and "magnetohydrodynamics"? As I understand it, processes you can treat as MHD practically never occur in nature, and are always something that has been carefully arranged by design so the system can be modeled at all. Astrophysicists seem to treat the two terms as interchangeable. Would you say that MHD suffices to model ion propulsion systems? To model anything, anywhere in nature?
- archermarks 4y ago
  MHD is a specific set of a assumptions you can make to reduce the plasma fluid equations to something more tractable. This includes things like assuming infinite conductivity and combining all of your species into a single "fluid". Those are reasonable assumptions for astrophysical plasmas but not at all for Hall thrusters, so we do not use MHD. In my work, we use the Euler equations for ions (assuming they lack viscosity), with electromagnetic and ionization source terms. Electrons are treated as massless and generally steady with respect to the ions. Both of their equations of motion can be derived starting with the Boltzmann equation and applying simplifying assumptions.
Melatonic 4y ago
Seems like this could be a good area to apply the "keep trying shit until we find something that works better" style of research. Would be lots of failures and of course it is quite unsophisticated but sometimes this kind of work is the best way to actually come to a greater understanding of the mechanisms behind somethin?
- archermarks 4y ago
  That's what we do! But that's also what we've been doing for 40 years and it's nice to have some theoretical grounding to why stuff works.
phkahler 4y ago
Is there any chance something is deflecting the rotating particles onto the axis? Similar to the particle jets emitted at the poles of black holes and other large objects - I don't think that process is fully understood either.
- archermarks 4y ago
  Unless I misunderstand your meaning, I guess in a certain sense you could frame it in that way, where you have a rotational drift which provides energy to an axially-propagating wave. I'm not well-versed in black hole plasma physics. It is worth noting that anomalous cross-field transport is a problem that exists in lots of plasmas, including fusion plasmas like tokamaks and probably also in astrophysical plasmas.

sampo 4y ago

> I work in designing simple "closure models" What is the order of the turbulence closure models used in modern work? 1.5? 2? 2.5?

archermarks 4y ago
Depends on what you mean by order i suppose. The simplest are algebraic (zero-order) models where we prescribe the anomalous transport as an explicit function of other plasma parameters. More complex ones are coupled systems of hyperbolic PDEs, which are typically first-order in space and time. If you are referring to the order of accuracy of the simulations, that's typically 1st to 4th order, depending on method used.
- johnbcoughlin 4y ago
  I answered the GP based on my understanding of "closure" meaning different things for moment models vs turbulence. I think they're referring to energy partition scaling in saturated turbulence.
  - archermarks 4y ago
    Yes, thank you. It's a closure problem in the former sense.
- sampo 4y ago
  > Depends on what you mean by order i suppose. I mean, in the context of Reynolds-averaged Navier–Stokes equations [1], what sort of turbulence model [2] you are using. But apparently you are not working in this context at all, so maybe my question doesn't make sense. [1] https://en.wikipedia.org/wiki/Reynolds-averaged_Navier%E2%80%93Stokes_equations https://en.wikipedia.org/wiki/Reynolds-averaged_Navier%E2%80... [2] https://en.wikipedia.org/wiki/Turbulence_modeling https://en.wikipedia.org/wiki/Turbulence_modeling
  - archermarks 4y ago
    Appreciate the clarification. I am familiar with RANS modeling to a reasonable extent but not with a lot of the modern developments. We're still kind of at the stage of trying to find something that works as well as Prantdl's mixing length model, or maybe a k-epsilon model. Additionally, while our processes are turbulent, the turbulence differs significantly from classical turbulence. For instance, turbulence in Hall thrusters is likely to be governed by an inverse energy cascade, where small disturbances magnify to larger and larger scales and the dissipation occurs via convection at device-scale instead of by diffusion at kolmogorov scales.
johnbcoughlin 4y ago
"closure" in the kinetics to fluid model context refers to a closed expression for the heat flux or anisotropic pressure in terms of the lower order fluid variables. You aren't really looking at a turbulence closure in the sense you're getting at in this context--for these hyperbolic models it's more about the transport than about energy partition.

mrfusion 4y ago

AMA? How much can these thrusters be scaled up? Or does it make more sense to stick more smaller ones on a craft if we need more thrust? Bonus question: how would you design a planet scale thruster? (Asking for a friend)

belter 4y ago
Don't tell him! He wants to move us away! https://en.wikipedia.org/wiki/The_Wandering_Earth https://en.wikipedia.org/wiki/The_Wandering_Earth
vutnlushaetoc 4y ago
These thrusters need fuel to move. That ends the fun on this line of inquiry. There is no free ride (unless you sail around very slowly, but this is more like sucking Earth's wake in style)
archermarks 4y ago
To scale up thrusters you can simply make them bigger, but that leads to a lot of wasted space as the thrust is only produced by a ring-shaped region of the device. To improve thrust density, you can nest these inside each other (see this page on my research group's website https://pepl.engin.umich.edu/project/high-power-hall-thrusters/ https://pepl.engin.umich.edu/project/high-power-hall-thruste...). More recently, we have been looking at running smaller thrusters at higher power densities, which was previously impractical due to the way older thrusters shaped their magnetic fields. Nowadays we can design thrusters such that very little power is lost to the walls so high power density operation is looking more feasible. Per the second question, you would need a very efficient thruster to limit the amount of mass needed as propellant. If you have an entire planet, you might look into using a photon rocket. Alternatively, you can try a massive solar sail or magnetic sail but I'm not an expert on these things.
- mrfusion 4y ago
  Thanks! Could super conductors help increase efficiency or density? Why or why not? Could this be a viable space to launch a start up? With launch costs coming down there could be a lot of demand for more efficient thrusters.

vutnlushaetoc 4y ago

> extremely sensitive to the choice of anomalous transport parameter in certain parts of the discharge, but completely insensitive to it in other parts of the discharge. When you find it, will it not likely be a secondary electromagnetic condition is affecting the permittivity and permissibility of materials in a way that was not previously anticipated? Like, this problem is "understood", right? You know where to look and likely what it will be? The fact Hall Thrusters work with fuel of very small mass, it doesn't seem to change the nature of the problem: detonation and fluid (plasma) dynamics.

archermarks 4y ago
The "material" in this case is vacuum and plasma. Yes, something small and electromagnetic is affecting the conductivity of the plasma, and we have an idea what it is, but that doesn't mean we know how to simulate it or predict its effects on the thruster. The fuel actually has quite high mass compared to a typical rocket engine. We use Xenon (131.293 g/mol) or Krypton (~83 g/mol), compared to air which is (~29 g/mol).
- vutnlushaetoc 4y ago
  > and we have an idea what it is I mean, it's on the phrase diagram. The answer will be the statistics of stable orbits at heat and pressure. Any gradient or annulus in a phase diagram is going to mean that there will be potential for rapid "phase" changes in all directions. Isn't this all just to say that you need instruments, instruments, instruments? You know where to look, what it is doing and why (you have guesses, you'll 'get it' when you see the final answer). Is this blog post not mystifying it instead of saying that's it's cool and we know where to look to make it better?
  - archermarks 4y ago
    I'm really not following what you're talking about. What is the phase diagram you're referring to?
    - 4y ago

madaxe_again 4y ago

Maybe I’m missing something, but couldn’t models be validated by building the things and seeing how they measure up in reality? Or is cost the limitation? I have no idea how exotic the required materials are.

archermarks 4y ago
We build them all the time. The difficulty is measuring the right regions of the plasma with the right precision. The plasma is pretty sensitive to perturbations so you can't always just stick a probe in them. This often restricts us to using noninvasive measurements like laser diagnostics, but because the plasma is quite rarified, it can be difficult to get a good signal to noise ratio. Additionally, the laser diagnostics can be restricted in what frequencies (spatially or temporally) they can measure, depending on the technique employed. Lastly, even if we fully understood the process, modeling it would remain a challenge (for all the same reasons that turbulence modeling is, despite the fact that turbulence is much better understood)
tel 4y ago
This tweet addresses your question, at least in part https://twitter.com/lougrims/status/1516013558475235335 https://twitter.com/lougrims/status/1516013558475235335

ncmncm 4y ago

Can you say something about the difference you experience between "plasma fluid dynamics" and "magnetohydrodynamics"? As I understand it, processes you can treat as MHD practically never occur in nature, and are always something that has been carefully arranged by design so the system can be modeled at all. Astrophysicists seem to treat the two terms as interchangeable. Would you say that MHD suffices to model ion propulsion systems? To model anything, anywhere in nature?

archermarks 4y ago
MHD is a specific set of a assumptions you can make to reduce the plasma fluid equations to something more tractable. This includes things like assuming infinite conductivity and combining all of your species into a single "fluid". Those are reasonable assumptions for astrophysical plasmas but not at all for Hall thrusters, so we do not use MHD. In my work, we use the Euler equations for ions (assuming they lack viscosity), with electromagnetic and ionization source terms. Electrons are treated as massless and generally steady with respect to the ions. Both of their equations of motion can be derived starting with the Boltzmann equation and applying simplifying assumptions.

Melatonic 4y ago

Seems like this could be a good area to apply the "keep trying shit until we find something that works better" style of research. Would be lots of failures and of course it is quite unsophisticated but sometimes this kind of work is the best way to actually come to a greater understanding of the mechanisms behind somethin?

archermarks 4y ago
That's what we do! But that's also what we've been doing for 40 years and it's nice to have some theoretical grounding to why stuff works.

phkahler 4y ago

Is there any chance something is deflecting the rotating particles onto the axis? Similar to the particle jets emitted at the poles of black holes and other large objects - I don't think that process is fully understood either.

archermarks 4y ago
Unless I misunderstand your meaning, I guess in a certain sense you could frame it in that way, where you have a rotational drift which provides energy to an axially-propagating wave. I'm not well-versed in black hole plasma physics. It is worth noting that anomalous cross-field transport is a problem that exists in lots of plasmas, including fusion plasmas like tokamaks and probably also in astrophysical plasmas.