DIY NMR: Best way of driving a swept magnetic field between two solenoids?

[ChristofferB]

It's NMR/EPR time again...

I've posted about this before a few years ago, but I've lost or decided against most of my original progress.

Here's the plan so far: Design an analog-ish base for NMR and EPR experiments, that is modular enough to allow a variety of probe coils, modulation coils and sweep coils for both NMR (nuclear magnetic resonance, principally ¹H (~42 MHz/T), but ¹³C (~10 MHz/T) would be a fun challenge as well. And EPR (electron resonance, at ~ 28 GHz/T

It should be a CW instrument (fixed RF freq, swept magnetic field), capable of both working as normal RF absorption/reflection spectrometer and with lock-in detection.

Field sweep to be provided by a DC voltage ramp over time, likely slow sweep speeds (minutes), data output to either XY graph recorder or data logger.


Attached is the base magnet: large ferrite discs with a fixed distance (adjustable with different matched standoffs).

With 40mm standoffs I get a field strenght B₀ = 550 Gauss, this corresponds to a ¹H resonance at 2.34 MHz - perfect!

So now I want a supply that can sweep the field up from that point. An air coil of to be determined size, but smaller than the pole faces will be made in the cavity, the coil split as two in series on each pole face.

This sweep coil needs to be driven bu a voltage controlled current source.

Ideally I would like the following characteristics:

-Sweep: 200 ppm of the field strenght = rougly 0.1 Gauss
-Variable offset to tune the field to resonance, maybe 0-5.5 Gauss
-Sweep input something like 0-5V


So I guess really a voltage standard for the offset, and then a voltage controlled current supply to drive the coils.
The power required for less than 0.1 Gauss is pretty small.

Just for test I made a single 20 loop coil of 0.2mm copper wire with a dia. of 6.4 cm. This swept the field approx 5 Gauss from 0-1A with a bench supply.

Here are some considerations I'd like your input on:

>Should the "field bias" coil and supply be a separate one from the "sweep" coil/supply? This would minimize temperature drift in the sweep.

>Is there a go-to voltage controlled current supply that's ideal for this application?

>Is it better to go with many-turn high impedance coils or few turn low-impedance coils?

Thanks! I hope you found this interesting!

[Dunckx]

Hi Christoffer,

I've a couple of friends who ran the NMR service at a couple of universities and are now retired so I can ask for their views.  Meanwhile, they provided me years ago with information about building an earth's field nmr system.  The benefit of this is that the nuclear precession is determined by the earth's field which is very uniform and the field of the polarizing solenoid has no uniformity requirements and no influence on the results.  Field uniformity requirements scale with frequency/field strength.  I started collecting bits for it but then as with other projects it got sidelined awaiting an opportunity...

The relevant publication plus the hardware details and arduino code are attached.  This may give you ideas even if it isn't something you'd want to build.  Apparently earth's field nmr is used commercially in food and drink analysis, so it isn't just a curiosity or a proton magnetometer.  I believe the author of the article was intending to extend this to an nmr imaging system but I don't know if he ever published the details.

One of my friends suggested using Overhauser sensitivity enhancement by dissolving a stable free radical in the solution and irradiating at the esr frequency to get polarisation transfer to the protons and boost the proton signal by ~100 times.

Meanwhile I will forward your enquiry to them and see what comes back.  It may take a day or two!
HTH

[doktor pyta]

I'm definitely not an expert, but if You make sweep winding as a Helmholtz coils You will benefit from better field uniformity near center and higher NMR signal.

[Dunckx]

I'm definitely not an expert, but if You make sweep winding as a Helmholtz coils You will benefit in better field uniformity near center and higher NMR signal.

I believe this is how it was done in commercial spectrometers.   It should work well as the variation in field required to sweep across all proton signals is very small, parts per million.  Hence magnet stability and homogeneity requirements.  Commercial spectrometers have a "lock" channel to keep the system in tune.  The solvent is deuterated - deuterochloroform or deuterium oxide (or even more expensive) and there is a second channel in the spectrometer which monitors the deuterium resonance and continually tweaks the magnet field to keep the D resonance centred.  One of my friends tells me that the lock system on his old 90MHz proton spectrometer had a gain of >10^9 - and that was the early 70s!  Chopper amp of course.

There are also some useful videos on YouTube from a New Zealand company https://www.youtube.com/channel/UC59E0ilufyYnOXewHbyXi5Q

Ferrite magnets have a significant temperature coefficient of field, I don't know if it's quite as bad as neodymium but it may not make much difference at low field.  I suspect though at 550gauss all the proton lines from e.g. ethanol will probably appear as one broad hump, so don't be disappointed if you can't see sharp lines or proton-proton coupling.  The Magritec vids (link above) should give you some feel for what to expect.  There's a fair few v. poor nmr vids on YouTube but the Magritec ones are the real deal.

If you live near a university ( chemistry PhD student) it may be worth your time to ask if they have any benchtop nmr systems they are getting rid of.  It doesn't happen often, but it does happen.  These are often used for relaxation time measurements rather than spectra and contain a small (but heavy!) stable permanent magnet.

The esr experiments on http://physicsopenlab.org/ may help with ideas and hardware.

[Kleinstein]

I would not make the number of turns very low (e.g. < 20), so that the wires to the coil would not have very much effect. Otherwise the number of turns does not change much, as the efficiency of a electromagent only depends on the copper volume, not the turns.
So I would go for convenient voltage range, like 10-20 V for the max. current.

For the design of the current driver it may be easier to have only current of one polarity. It's likely easier to have only 1 coil, to avoid complications from coupling tha may effect the drivers.
One could consider to have the coarse shilft in filed at the joke, away from the sample, so kind of helping the permanent magents a bit.

The idea of a helmholz coil is good, but this is calculated for free field. With the joke and paermananet magents, the flield would be slightly different. The tendency is to get a relatively constant field, with less effect of the actial coild separation.

[Dunckx]

I forgot a very useful program I used years ago, FEMM by Dr. David Meeker:

https://www.femm.info/wiki/HomePage

Free finite element software!  You will have fun with this 

I've attached a couple of input files as it's easier to have something to tweak than to have to start from a blank page.  I just re-ran them to be sure they work out of the box.  You'll have to delete the txt file extension and replace it by FEM.  You could begin by tweaking the material of the NdFeB magnet via the Properties menu - select materials and edit - you could choose a ferrite instead of neodymium and see what effect it has.  It isn't that intuitive at the start but the author has a good tutorial for you to follow.  The second file is the polarising coil I designed for the earth's field nmr project.  You can vary the turns and current and see what happens.  The answer file can have a colour gradient applied, and this is very pretty, see attached.

With this you should be able to tweak your scan coil even in the presence of a yoke/magnet.

One of my NMR friends has come back to me in response to your original question.  He writes:

"I'll try reawakening the brain cells, have a think and get back to you.

For a linear sweep I would consider an integrator driving a current
source. I see many hurdles to overcome before implementation of linear
field sweeping."

HTH

[Dunckx]

OK, I've just had a phone call from the friend who rebuilt a cw instrument into a pulse FFT instrument many, many years ago...

His first comment was that without a yoke around the "outside" you've lost a lot of field by having far more reluctance in the magnetic circuit than necessary.  The air gap between the poles should be the only air gap present and the "free ends" of the magnets need connecting by a yoke.  The field sweep was not accomplished by Helmholtz coils but by Golay coils, see:

https://duckduckgo.com/?t=ffnt&q=Golay+coil&atb=v225-1&ia=web

Unfortunately, these don't permit simulation in FEMM without a huge amount of effort as they are not capable of symmetry-based simplification like a Helmholtz coil is - they are 3D structures as you can see from the search results above.  I think you can still run simulations based on Helmholtz coils to get a feel for the magnitude of the sweep you can get.

My suspicions regarding temperature fluctuations of field are not of huge significance and you need the strongest field you can get without crushing your fingers as sensitivity scales with field.  He suggested a magnetron magnet out of a microwave oven.  However you do it, the region of relatively homogeneous field is going to be very small in comparison to the pole size.  This will limit your sample size and hence sensitivity.  You might try using phase sensitive detection and record a derivative spectrum as is usually done for esr spectrometers.  He said that before considering linear sweep it would be a tough enough challenge just to find the proton resonance at fixed field.

He has promised to have a further think about it and I'll let you know what comes of it.  It seems you have set yourself quite a challenge!

[ChristofferB]

Thanks for the earth's field NMR references, and all the input!

Golay coils, which are also called saddle coils i believe needs to be axial to the magnetic field - which means a parallel pole type magnet would need to have the sample tube axial with the magnet faces - and be very short. They're commonly used on modern superconducting solenoid shaped magnets though.

The trick is the sample coil must be normal to the external field to avoid the field sweep affecting resonance, and avoiding modulating the external field with RF.

I think the field tune coil I mentioned would be easiest to just hook up to a lab PSU trough a power potentiometer.

The sweep coils (a sweep is typically 10-100 ppm of the external field, so 5-50 mGauss) may just be something like this:





I'm not too worried about getting a stronger field, I've attempted an iron yoke in the past but I simply don't have the machining capabilities to ensure parallelity.

In the current magnet setup I can shim magnets with slivers of tin foil or other thin metal shim stock.

The probe will be an RF bridge circuit of the type shown in the attachment. This has been used in the past and is supposed to give a null output when the sample is not absorbing //out of resonance. I can't really get my head around how that works, but I guess I'll just try and build one and mess around with it.

[coppercone2]

that circuit will work and be stable, the better option is a op amp supply / programmable power supply, because its built nice with the control loops figured out well. Usually they are impressive when you actually take them apart with good thermal design, beefy components, etc. They don't look like they are wroth the money from the outside but when you take it apart the nice heat sink alone is usually more then enough convincing, when you consider the nice form factor they got it in. They also usually have good mounting for the transistors, replaceable transistors (sockets), good control switches, BNC connectors, noise control etc.

They are particularly heavy instruments (usually means its a good buy). The modern cheap solution is arcadian (for a modest budget and high power levels) but thats a stand alone wire it yourself module.

I have a feeling if you build this one yourself its going to occupy alot of time and resources and not be so fine. They never quite cheapened up so much because they are still very useful.

[Rod]

Attached is the base magnet: large ferrite discs with a fixed distance (adjustable with different matched standoffs).

With 40mm standoffs I get a field strenght B₀ = 550 Gauss...

It may be helpful to realize that changing the separation distance (standoff length) changes not only the strength, but also the shape (and thus uniformity or homogeneity), of the magnetic field. 

It may also be helpful to be aware that sintered magnets (ferrites and rare earths) consist of particles having nonuniform sizes, shapes and alignments, so have nonuniform density and magnetization.  Their fields vary (between two magnets, or cross the surface of a disk magnet) by between 5% and 10% in strength and 5 and 10 degrees in direction (and inexpensive magnets available on Ebay etc., made in workshops in rural China with poor quality control, may be worse than that).  This can be mapped using a Hall probe (see figure 2 in the paper linked below).

For low permeability (rare earth) disk magnets, the magnetic field is identical to that produced by a pair of Helmholtz coils wound on the outer cylindrical surface of the disks.  To produce the most nearly uniform field, the distance between their centers should be equal to their radius (the same spacing as a Helmholtz pair). 

For high permeability (ferrite) disk magnets, the field uniformity should continue to increase as the gap is made smaller than the Helmholtz spacing.  However, if their magnetization is not uniform, it'll instead get worse.
 
It's likely possible to achieve a homogeneity of 1% across a 1 cm3 volume, but unrealistic to expect 0.1% = 1 ppt, let alone 200 ppm.  Realize the the field (in-)homogeneity across the entire sample determines the magnetic resonance linewidth.  It will be very broad, and correspondingly difficult to detect (impossibly so, if not anticipated and designed accordingly).

For these reasons, magnets designed to create high field homogeneity (i.e. for NMR/ESR) rely on iron pole pieces to shape the field, whether driven by an electromagnet or by a permanent ferromagnet.   (To increase the field, most also use iron yokes to create a low-reluctance field return path (i.e. magnetic circuit), or taper the pole pieces to smaller diameter, or both.)   Slightly concave pole pieces are ideal, but even simple flat pole pieces help make the field more uniform.  See Chonlathep et al., JMR (2016) https://ur.booksc.eu/book/77436329/d24ca5 , which is based on Eiichi Fukushima's earlier design (ref. 9 therein).  A similar magnet design is shown in Figures 2 and 3 of Cooley et al. JMR (2020) https://tabletop.martinos.org/images/a/a6/1-s2.0-S1090780719302642-main.pdf  Hope these are helpful.

[RoGeorge]

Tried once to observe NMR using two ferrite magnets from a cheap commercial welding clamp, and failed.  The only successfull step was measuring the magnets.    (https://www.eevblog.com/forum/projects/measure-a-magnets-b-field-with-a-rigol-ds1054z-oscilloscope-and-a-piece-of-wire/)

Curious to see how the project will go.   

[Rod]

Should the "field bias" coil and supply be a separate one from the "sweep" coil/supply?

Conventionally, separate Helmholtz pairs are used for sweep and modulation, because their requirements differ.

The field sweep is slow (so induces negligible eddy currents, and its inductance can be large) but its field needs to be as uniform as the magnet.  So it can be wound right onto the magnet or its pole pieces. 

The modulation field is fast, but as its a small field, doesn't need to be as uniform.  So a smaller Helmholtz or saddle coil pair with smaller inductance is more convenient.  It's typically bipolar, and driven by a small audio amplifier.  Note that it's an inductive load, so it may be easier to drive if a series RC "snubber" is placed in parallel with it.  The RC values would be calculated to form a critically damped RLC (resistive) load at or above the highest anticipated modulation frequency.  This acts as a low pass filter, so also eliminates any problem with its self-resonance.

I do hope you persevere!  This is a neat project.