Construction of a permanent magnet NMR spectroscope

[ChristofferB]

Hi!

I've been taking some courses on NMR spectroscopy, and I wonder- Is it possible to build an NMR machine yourself?
Obviously not wanting to deal with liquid helium and superconducting magnets, that'd be super dangerous, but I'm thinking a 1960's style permanent magnet machine.

I'm just unsure what's involved in terms of electronic interfacing machinery - Is there any references on older NMR generators/aquisition gear? Magnet aside, as I see it, you'd only need a probe with the coils for RF pulse and FID signal pickup - then you'd need an RF pulse generator, and some manner of gear for collecting/processing the data, maybe even just a spectrum analyzer.

It's more of a "can this be done" than a thing I'm actually starting, and it'd never rival anything scientific that has been used in the last 46 years, but it would be neat to see working.

What are your thoughts? There has been a lot of development in tiny, tabletop NMR apparatus these last few years.

--Christoffer

[T3sl4co1l]

Shimming is hard, no matter how you do it, so there's that...

Tempco is the hardest thing you'll have to deal with.  Permanent magnets have a pretty good negative tempco, and may not be reproducible (i.e., exhibit hysteresis).  Which means your center frequency is always changing, and if it's changing significantly (<1ppm!) during an acquisition alone, you're kind of screwed, as far as getting good data.

Temperature regulation helps, obviously!  But any residual (convection, drafts?) manifests as 1/f noise, which is hard to deal with.

The first ones, I think, were electromagnet because it's easier to control, and more stable(!), thanks to a nice large lead-acid battery.  The original articles by Purcell and Pound spoke of controlling the magnetic field by varying the position of the window (no doubt actuated by a grad student ).

FYI, it might be tempting to regulate magnetic field with feedback -- that's a good idea too, but mind its limitations.  If your field sensor is quite noisy (as is the case of a cheap Hall effect sensor!), your results will only be worse.  Perhaps considerably so!

It would be nice to have a sample of a pure substance, usually H2O (? ppm) or TMS (0 ppm, by convention), and do a PLL servo using that as the field sensor and frequency standard.  But that's tricky because the sensor has to be in the same place as your sample, physically speaking, and the extra reference signal probably means a poor SNR against whatever your test sample is.  Producing, and shimming, a larger volume of magnetic field, is quite energy intensive, too.

Mind that your solution doesn't need to be exclusively one or the other.  You could use permanent magnets for the bulk of the field (a Halbach array at, say, 1-2T), and put electromagnets around it to tweak the XYZ fields (using perpendicular Helmholtz coils), and their gradients (by applying differential current to the Helmholtz coil pairs).  This definitely saves DC power!

Tim

[max_torque]

This sounds like on of those interesting "Projects" where, by the time you get it working, you've either forgotten why you needed it, or it's morphed into an entirely different thing 

[ChristofferB]

Shimming is hard, no matter how you do it, so there's that...

Tempco is the hardest thing you'll have to deal with.  Permanent magnets have a pretty good negative tempco, and may not be reproducible (i.e., exhibit hysteresis).  Which means your center frequency is always changing, and if it's changing significantly (<1ppm!) during an acquisition alone, you're kind of screwed, as far as getting good data.

Temperature regulation helps, obviously!  But any residual (convection, drafts?) manifests as 1/f noise, which is hard to deal with.

The first ones, I think, were electromagnet because it's easier to control, and more stable(!), thanks to a nice large lead-acid battery.  The original articles by Purcell and Pound spoke of controlling the magnetic field by varying the position of the window (no doubt actuated by a grad student ).

FYI, it might be tempting to regulate magnetic field with feedback -- that's a good idea too, but mind its limitations.  If your field sensor is quite noisy (as is the case of a cheap Hall effect sensor!), your results will only be worse.  Perhaps considerably so!

It would be nice to have a sample of a pure substance, usually H2O (? ppm) or TMS (0 ppm, by convention), and do a PLL servo using that as the field sensor and frequency standard.  But that's tricky because the sensor has to be in the same place as your sample, physically speaking, and the extra reference signal probably means a poor SNR against whatever your test sample is.  Producing, and shimming, a larger volume of magnetic field, is quite energy intensive, too.

Mind that your solution doesn't need to be exclusively one or the other.  You could use permanent magnets for the bulk of the field (a Halbach array at, say, 1-2T), and put electromagnets around it to tweak the XYZ fields (using perpendicular Helmholtz coils), and their gradients (by applying differential current to the Helmholtz coil pairs).  This definitely saves DC power!

Tim

Thanks for that rundown! Yeah, a electromagnet-tweakable permanent magnet field might be the best course of action.

The original oldies had the magnetic field varying to scan across the bands, while the newer just bursts an RF pulse that covers the entire band - that might be more efficient (and easier to implement in practice), the issue there is generating VERY short RF pulses...

But if I read you correctly, you're saying getting a stable magnetic field is by far the biggest challenge.

The good thing about having a short aquisition period is that temp. drift in magnets (and anywhere else) is minimized.

This sounds like on of those interesting "Projects" where, by the time you get it working, you've either forgotten why you needed it, or it's morphed into an entirely different thing 

But those are the best ones!

[ChristofferB]

Obviously there'd be a lot of issues involved, but if goal #1 is to just SEE a free induction decay in a test reference, it doesn't sound completely unachievable.

For an initial test magnet array, could ordinary commercial rare earth metal magnets be used, perchance? I think it must be either that, or see if an old NMR magnet is obtainable from a university or maybe Ebay.. Getting a custom made permanent magnet might make it more pricey than fun. Probes are relatively easy to come by too.

[xygor]

There was an article in Scientific American from the '50s or '60s on building one using a magnetron magnet.  It was later republished in a book.  If you use the correct RF frequency, it should be possible using the earth's magnetic field.  The signal would be noisy though.

Edit:
Book:
The Scientific American book of projects for the amateur scientist.
Author:    Clair L Stong
Publisher:    New York, Simon and Schuster, 1960.

Article:
http://www.scientificamerican.com/article/the-amateur-scientist-1959-04/

[ChristofferB]

There was an article in Scientific American from the '50s or '60s on building one using a magnetron magnet.  It was later republished in a book.  If you use the correct RF frequency, it should be possible using the earth's magnetic field.  The signal would be noisy though.

Edit:
Book:
The Scientific American book of projects for the amateur scientist.
Author:    Clair L Stong
Publisher:    New York, Simon and Schuster, 1960.

Article:
http://www.scientificamerican.com/article/the-amateur-scientist-1959-04/

Magnetron magnet! not a bad idea! Thanks a bunch, that looks interesting.

[langwadt]

Shimming is hard, no matter how you do it, so there's that...

Tempco is the hardest thing you'll have to deal with.  Permanent magnets have a pretty good negative tempco, and may not be reproducible (i.e., exhibit hysteresis).  Which means your center frequency is always changing, and if it's changing significantly (<1ppm!) during an acquisition alone, you're kind of screwed, as far as getting good data.

Temperature regulation helps, obviously!  But any residual (convection, drafts?) manifests as 1/f noise, which is hard to deal with.

The first ones, I think, were electromagnet because it's easier to control, and more stable(!), thanks to a nice large lead-acid battery.  The original articles by Purcell and Pound spoke of controlling the magnetic field by varying the position of the window (no doubt actuated by a grad student ).

FYI, it might be tempting to regulate magnetic field with feedback -- that's a good idea too, but mind its limitations.  If your field sensor is quite noisy (as is the case of a cheap Hall effect sensor!), your results will only be worse.  Perhaps considerably so!

It would be nice to have a sample of a pure substance, usually H2O (? ppm) or TMS (0 ppm, by convention), and do a PLL servo using that as the field sensor and frequency standard.  But that's tricky because the sensor has to be in the same place as your sample, physically speaking, and the extra reference signal probably means a poor SNR against whatever your test sample is.  Producing, and shimming, a larger volume of magnetic field, is quite energy intensive, too.

Mind that your solution doesn't need to be exclusively one or the other.  You could use permanent magnets for the bulk of the field (a Halbach array at, say, 1-2T), and put electromagnets around it to tweak the XYZ fields (using perpendicular Helmholtz coils), and their gradients (by applying differential current to the Helmholtz coil pairs).  This definitely saves DC power!

Tim

whether you need 1ppm depends on what you want to do.

electro magnets are weak,  and SNR is proportional to field^(3/2) so you want the highest field you can get

it might just be possible to make a 2T halbach array but even with high temperature magnetic material it
starts to permanently change at ~40'C

[CD4007UB]

We have a 1960s' vintage Newport pulsed NMR spectrometer that we use in our UG lab for teaching NMR.

It uses a permanent 'Watson' magnet and the NMR frequency is approx 2.3 MHz. The pulse generator is based on an astable at 4.6MHz (so as not to saturate the RF receiver), divided down by 2 with TTL logic during the excitation pulses. (Nowadays, of course, one could use an MCU for the pulse generation.) The RF receiver is a FET cascode followed by a video amp and a Ge diode detector to show FID and spin echoes, which can be recorded on a DSO.

To make a Watson magnet, you need two approximately identical rectangular permanent magnets (magnetized parallel to a short edge). The magnets are sandwiched between two soft-iron sheets (a few mm or more thick), which act as pole pieces (N & S). The NMR sample and coils go inside the narrow cavity formed by the magnets and pole pieces. The field is fairly uniform over ~1cc because the soft iron acts as a magnetic equipotential - so, the pole pieces form the analogue of a parallel-plate capacitor.

We operate with a vertical gap, so that the B field is horizontal. The RF pickup coil is then just a solenoid wound on a vertical cylindrical former. The excitation coil is slightly trickier, as it has to produce a uniform horizontal field (to rotate all the spins by the same amount), perpendicular to both the static B and the pickup coil. The solution to this problem is to use vertical windings on the cylindrical former to produce a current density that varies as aproximately cos(theta), where theta is the azimuthal angle. This produces the uniform horizontal RF field inside the coil former.

With this setup, you can just about see the FID by inserting your finger into the excitation coil (but that's not really recommended). Glycerol is a better sample (fairly short lifetime, so no problems with saturation). It can be diluted with water to explore the amazing phenomenon of motional narrowing. Pure water does not work well (long lifetime, so tends to saturate). But a little bit of copper sulphate reduces the lifetime. All described in a classic paper by Bloembergen, Purcell & Pound.

Our local hospital built a clone of our apparatus some years ago for training staff in the early years of MRI. We haven't replicated the setup with home DIY, but it would be possible if you can find the material to make a Watson magnet - a fairly uniform static B field is essential to see the FID (which decays as T2*, determined by field inhomogeneity). The electronics side should be fairly straightforward.

Hope that helps.

Neil.

[ChristofferB]

We have a 1960s' vintage Newport pulsed NMR spectrometer that we use in our UG lab for teaching NMR.

It uses a permanent 'Watson' magnet and the NMR frequency is approx 2.3 MHz. The pulse generator is based on an astable at 4.6MHz (so as not to saturate the RF receiver), divided down by 2 with TTL logic during the excitation pulses. (Nowadays, of course, one could use an MCU for the pulse generation.) The RF receiver is a FET cascode followed by a video amp and a Ge diode detector to show FID and spin echoes, which can be recorded on a DSO.

To make a Watson magnet, you need two approximately identical rectangular permanent magnets (magnetized parallel to a short edge). The magnets are sandwiched between two soft-iron sheets (a few mm or more thick), which act as pole pieces (N & S). The NMR sample and coils go inside the narrow cavity formed by the magnets and pole pieces. The field is fairly uniform over ~1cc because the soft iron acts as a magnetic equipotential - so, the pole pieces form the analogue of a parallel-plate capacitor.

We operate with a vertical gap, so that the B field is horizontal. The RF pickup coil is then just a solenoid wound on a vertical cylindrical former. The excitation coil is slightly trickier, as it has to produce a uniform horizontal field (to rotate all the spins by the same amount), perpendicular to both the static B and the pickup coil. The solution to this problem is to use vertical windings on the cylindrical former to produce a current density that varies as aproximately cos(theta), where theta is the azimuthal angle. This produces the uniform horizontal RF field inside the coil former.

With this setup, you can just about see the FID by inserting your finger into the excitation coil (but that's not really recommended). Glycerol is a better sample (fairly short lifetime, so no problems with saturation). It can be diluted with water to explore the amazing phenomenon of motional narrowing. Pure water does not work well (long lifetime, so tends to saturate). But a little bit of copper sulphate reduces the lifetime. All described in a classic paper by Bloembergen, Purcell & Pound.

Our local hospital built a clone of our apparatus some years ago for training staff in the early years of MRI. We haven't replicated the setup with home DIY, but it would be possible if you can find the material to make a Watson magnet - a fairly uniform static B field is essential to see the FID (which decays as T2*, determined by field inhomogeneity). The electronics side should be fairly straightforward.

Hope that helps.

Neil.

That is certainly interesting! And having RF in the 1MHz-10MHz range solves a lot of trivial issues concerning my 80's instrumentation.
Is it still the case with this type of magnet, that you have two magnets with a spacing of some centimeters?

[ChristofferB]

I also found this: http://www.conspiracyoflight.com/NMR/NMR.html which has a lot of details on the electronic side of things.
It might be good to develop an NMR receiving coil simulator (basically just a function gen.) to calibrate the 'tronics.

[CD4007UB]

That is certainly interesting! And having RF in the 1MHz-10MHz range solves a lot of trivial issues concerning my 80's instrumentation.
Is it still the case with this type of magnet, that you have two magnets with a spacing of some centimeters?

Yes, the permanent magnets are about 2cm from pole to pole, so there's an approx 2cm gap between the soft-iron pole pieces, which are about 10cm square. The coild former is mounted on a valve socket that fits inside the gap, and we insert a test-tube containing ~1cc of liquid as the sample, placed near the centre of the B field, where it's presumably most uniform.

With this setup, you need field uniformity of ~1 part in a thousand (not ppm), so no shims are required with the Watson magnet.

Neil

P.S. An alternative DIY NMR approach is the proton-precession magnetometer described at http://www.gellerlabs.com/PMAG%20Articles.htm. This uses a strong electromagnet to polarize the sample (H2O). When that is removed, the protons precess in the earth's field - but this has to be done outside, away from anything magnetic.

[CD4007UB]

I also found this: http://www.conspiracyoflight.com/NMR/NMR.html which has a lot of details on the electronic side of things.
It might be good to develop an NMR receiving coil simulator (basically just a function gen.) to calibrate the 'tronics.

Working at 2.3MHz or so, you don't need to use double-balanced mixers or anything fancy. You could generate the pulses from an MCU (e.g., Arduino) and drive the excitation coil with a single-transisitor common-emitter amp (with the coil in a tuned LC circuit). Our RF receiver is a tuned FET cascode followed by a wideband op-amp (which only needs a few MHz bandwidth at ~40dB gain - you can't use a 741, but wideband op-amps are not hard to get hold of).

We inject a test signal using a small coil in series with a 1k resistor soldered at the end of a piece of co-ax and driven by a signal generator. (An MCU could be used instead.) This is useful for initial tuning of the RF receiver. One can also beat a low-level injected signal against the FID. Observing the beats on a scope allows very precise measurement of the FID frequency, and hence precise determination of the B field.

Neil.

[ChristofferB]

That does sound good. The largest issue with the article I linked is the 100W RF power amplifier - Seems like that'd be less desirable.

Keeping it at a couple of MHz might be a wise plan as of first. I assume you'd always be able to ramp up the freq. and thus resolution with more advanced parts, reusing magnet and probe.

Now finding a proper magnet / materials to make one..

[CD4007UB]

I assume you'd always be able to ramp up the freq. and thus resolution with more advanced parts, reusing magnet and probe.

Since the magnetic field determines the NMR frequency, with a permanent magnet you have to stick with more or less the same frequency - the hospital system that we worked on had a small coil wound on one magnet so that the field could be tuned, but only by a tiny amount.

As NMR frequencies for research have increased over the years, our university has scrapped some large high-quality permanent magnets (unfortunately, much too large and heavy for home use). So, when you up the frequency, you need to buy a lot of new kit. For teaching purposes, there's never been much point in us going above the original 2.3MHz.

[ChristofferB]

Hm. That's true, thinking about it. Is there a relatively painless way of determining the appropriate frequency for a given magnet?

[CD4007UB]

I assume you'd always be able to ramp up the freq. and thus resolution with more advanced parts, reusing magnet and probe.

Since the magnetic field determines the NMR frequency, with a permanent magnet you have to stick with more or less the same frequency

Of course, this assumes that you stick with proton NMR. In principle, you can change to a different nucleus and hence work at a different frequency with the same magnet. I don't know if this is actually practical with our simple equipment, as it relies in part on the high concentration of protons in H2O etc to get strong enough signals.

[langwadt]

I assume you'd always be able to ramp up the freq. and thus resolution with more advanced parts, reusing magnet and probe.

Since the magnetic field determines the NMR frequency, with a permanent magnet you have to stick with more or less the same frequency

Of course, this assumes that you stick with proton NMR. In principle, you can change to a different nucleus and hence work at a different frequency with the same magnet. I don't know if this is actually practical with our simple equipment, as it relies in part on the high concentration of protons in H2O etc to get strong enough signals.

there are lots of other nucleus but they all have lower sensitivity and you have a probe tuned for the
frequency

[CD4007UB]

Hm. That's true, thinking about it. Is there a relatively painless way of determining the appropriate frequency for a given magnet?

The B-field, determines the Larmor frequency for the nuclear precession. For protons, this comes out as about 42.58 MHz/tesla.

There's a table here http://kodu.ut.ee/~laurit/AK2/NMR_tables_Bruker2012.pdf from Bruker, who make NMR spectrometers (starting at a field of about 7 teslas!). You can see how modest our 2.3MHz setup is compared to modern equipment - but it's still great for learning about NMR.

Neil.

[ChristofferB]

I assume you'd always be able to ramp up the freq. and thus resolution with more advanced parts, reusing magnet and probe.

Since the magnetic field determines the NMR frequency, with a permanent magnet you have to stick with more or less the same frequency

Of course, this assumes that you stick with proton NMR. In principle, you can change to a different nucleus and hence work at a different frequency with the same magnet. I don't know if this is actually practical with our simple equipment, as it relies in part on the high concentration of protons in H2O etc to get strong enough signals.

But you can get combination probes, with outputs for multiple nuclei. I've seen a 1H, 2H, and 13C combi probe once, i think.
I actually assumed 13C would be more efficient, since they're usually taken at lower frequencies.

With the proton concentration in H2O, you could use another solvent to get a nice 13C peak. CDCl3 is sometimes used, TMS would be good too.

[zappedy]

This might be stupid, but can't you just capture a s**tload of data and then simply chose to look at a sub sample where the temperature and other parameters happened to be stable? They couldn't do that in the 60s, perhaps that could be a way to "cheat" using modern consumer computer tech.

[ChristofferB]

That's not a bad idea, and I did plan on using computer-aided data aquisition. The issue with NMR is that you sometimes need to average the spectra over a long period of time (hence why modern NMR experiments can take up to 48Hrs!).

Definately an option over building an analog system to do FT and then 2d pen-plot the spectrum.

[CD4007UB]

That's not a bad idea, and I did plan on using computer-aided data aquisition. The issue with NMR is that you sometimes need to average the spectra over a long period of time (hence why modern NMR experiments can take up to 48Hrs!).

Definately an option over building an analog system to do FT and then 2d pen-plot the spectrum.

This is possible with research equipment, but way beyond what's possible with the setup that we use. For FT-NMR spectroscopy, you're looking at the chemical shift between protons (or other nuclei) in different chemical environments. That's ~ppm, so I think you need corresponding field homogeneity of ~ppm. With our setup, the field homogenity is ~parts in 10^3, so you can't distinguish between -CH2, -CH3 groups etc. A research chemist would therefore not regard our pulsed NMR apparatus as a useful 'spectrometer'. But it is, nontheless, useful for looking at the FID, spin echoes, measuring T1, T2, T2* and exploring things like motional narrowing and paramagnetic relaxation - the basic physics of NMR, rather than its use in chemistry.

[ChristofferB]

True. You probably wouldn't get anything sensible out of FT'ing a 2,something mhz spectrum.
Even if you had the fieldwise precision, it'd likely be all muttled.

[zappedy]

Sample a lot of parameters and put an FPGA genetic algorithm / neural network AI on the stabilization job?