I feel truly honored Stan Sykora Slides available at DOI 103247SL6Nmr17001 NMR in XXI Century 30 selected things to do and explore or W hat for Magnetic Resonance sake would I do if I were 50 years ID: 722755
Download Presentation The PPT/PDF document "Many thanks to organizers for inviting m..." is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
Slide1
Many thanks to organizers for inviting me to talk here.
I feel truly honored.
Stan Sykora
Slides available at DOI
10.3247/SL6Nmr17.001Slide2
NMR in XXI Century: 30 selected things to do and explore
or “What
, for Magnetic Resonance sake, would I do if I were 50 years
younger”
Stan S
ýkora
alias Stan’s Hub (www.ebyte.it)
Presented at MMCE 2017, Budapest, Hungary, 8-12 March 2017
Who am I: a physicist, an NMR buff, a retired entrepreneur, programmer, OBSERVERExperiences: academic (11 y), head of Bruker Italiana (6 y), cofounder of Stelar (20 y)I acquired my first NMR spectrum in February 1964, and did NMR ever since.Sold, installed and serviced over 100 instruments (and designed some of them)Currently: since 2000, running Extra Byte and working closely withMestrelab Research on Mnova NMR spectroscopy software
Hobby
Bread & butter
Why am I telling you all
this crap about myself?
Because I
recently started asking
a
provocative
question,
namely,
Is NMR
stuck?Slide3
My
personal perception is that
There indeed is something wrong with NMR
At a recent meeting
(MRPM 13, Bologna, Italy, 4-8 September 2016)
I have discussed why I haveTo some extent lost my once strong enthusiasm about NMR methodology as a whole:
As a technologist, I can not help much with what matters most (that is, marketable Applications)I will just make 30 technical proposals which, if enough ofthem worked, might enable new application areas
Lack of novel ‘killer’ applications, other than Chemical Spectroscopy (~1955) and MRI (~1980)Acceptance of status-quo by current markets (chemists happy with the spectra they produce)Insufficient market pressure for innovation; basically just magnets are being tinkered with
Insufficient academic interest in NMR
(very few groups worldwide still research NMR as such)
Lukewarm industrial R&
D investments (why
produce cheaper if the market remains the same?)
In a sense, this is a continuation of my Bologna MRPM presentation,
but this time along more positive, forward-looking lines.Slide4
About the Proposals
Nature
: they are
subjective musings of an ageing physicist
Originality
: some are original (
#), some are ‘in the air’ (#), and may are already running (#) and not mine at all, I am just convinced that they merit more attention.Traps: I counted that in the 30 Proposals I have fallen into each of the 45 Csaba’s Mental Traps at least once.Primary goal
: stimulate a lively, constructive discussion!
The Bottom Line:View this Talk simply as an incomplete review of open possibilities.Slide5
Data Harvesting and
primary Preprocessing
Early acquisition
and
Original-Data Safeguarding
Fast
AD conversion (more data is always better than less data)Total monitoring: ubiquitous & assiduous (more data is better)Greedy mass storage of raw data with no preprocessing
Group ISlide6
Early Acquisition &
Original Data Safeguarding
... don’t fool around with analog signals! ...
Since digital data are
UNCORRUPTIBLE
,
convert the analog signal to digital form as soon as possible!
Safeguarding the original data:
Modern approach: Smart sensor [Immediate] Storage [Later] processing# 1
Direct detection by smart sensors:
Sensor
Amplifier
ADC, all bundled together!
Obsolete flowchart
:
Dumb Sensor
Analog signal
Analog transfer
Analog preprocessing
A/D Conversion
Digital preprocessing
Data storage
Processing
Grapes preprocessing
(no longer approved by EU)Slide7
Brief history of MR
Data Acquisition ...
~ 1975 - 1995
Sensor
(an inductive coil, C)
+
passive tune & match adapter+ quadrature receiverBefore then, one of the two branches was missing (
single detectionversus quadrature detection)Legend:Red lines ... analog signalsBlue lines ... digital dataGreen lines ... digital/analog controlBlack lines ... rf controlpA ... pre-amplifierhfA ... high-frequency amplifierLO ... local oscillator (master frequency)... phase detector (mixer, lock-in, ...)S ... phase splitter/shifter
lpF ... low-pass filterlfA ... low-frequency (audio) amplifierADC ... analog-to-digital converteraccB ... accumulation buffer (RAM)IntroSlide8
Brief history of NMR
Data Acquisition ... ~ 1980 - 2000
‘Multinuclear
’
instruments:
Sensor
(an inductive coil)+ passive tune & match adapter+ heterodyne
(inter-frequency IF)+ quadrature receiver (u/v) with two ADC’sLegend:Red lines ... analog signalsBlue lines ... digital dataGreen lines ... digital/analog controlBlack lines ... rf controlpA ... pre-amplifieraaF ... anti-aliasing filterhfA ... high-frequency amplifier
MiX ... mixer (down/up converter)LO ... local oscillator (master frequency)IF ... interfrequency... phase detector (mixer, lock-in, ...)S ... phase splitter/shifterlpF ... low-pass filterlfA ... low-frequency (audio) amplifierADC ... analog-to-digital converteraccB ... accumulation buffer (RAM)
Intro
– the culmination of analog complexitySlide9
Brief history of MR
Data Acquisition ...
~ 1990 - today
Sensor
(inductive coil)
+ passive
tune & match adapter+ FPGA ‘digital receiver’Much
hardware is now inside a single-chip FPGA, programmed by a software (HDL)Single-board MR Console is a reality!Legend:Red lines ... analog signals: shrinkingBlue lines ... digital data: expandingGreen lines ... digital/analog controlBlack lines ... rf control: synthesizedpA ... pre-amplifier
hfA ... high-frequency amplifierMiX ... mixer uses synthesized reference aaF ... anti-aliasing filterADC ... earlier and faster conversion... phase detector (mixer, lock-in, ...)Clk ... instrument master clockDDS ... direct digital synthesis of RFlpF ... low-pass filter: digital firmwareaccB ... accumulation buffer (RAM)
Stelar 2003
Takeda 2008
Modern FPGA chip
This could run five NMR’s
Intro
– arrival of FPGA’s Slide10
Brief
history of MR Data Acquisition
... ~ 1995 - today
Sensor
(inductive coil)
+ passive tune
& match adapter+ active RF adapter with IF+ single ADC= DDR
(Chemagnetics/Varian?)Legend:Red lines ... analog signalsBlue lines ... digital dataGreen lines ... digital/analog controlBlack lines ... rf controlpA ... pre-amplifierhfA ... high-frequency amplifierMiX ... mixer
aaF ... anti-aliasing filterADC ... Analog/digital converter... I guess that this was how the DDR of Chemagnetics and Varian/Agilent worked ...
Intro
– maybe this existed (not sure)Slide11
Direct Waveform Sampling (DWS
)
Sensor
(inductive coil)
+ passive tune
&
match adapter+ active base-band adapter (no IF)+ single, fast ADC= DWS receiver
Legend:Red lines ... analog signals: shrinkingBlue lines ... digital data: uncorruptibleGreen lines ... digital/analog controlpA ... pre-amplifierhfA ... high-frequency amplifier
ADC ... Analog/digital converter... what we should have had X years ago ... but it was not to be ...
IntroSlide12
The future
: Smart Sensors
... smart sensor: sensor + active adapter + ADC + COM link ...
Communication link types:
Digital
/ Analog
Wireless
/ GalvanicOptical / ElectricHybrids ...Always bi-directionalfor clock &
commandsWiFi looks optimal =>Issues to ponder carefullyPower supply(noise import to sensor area)Needed: < 10 mm2
#
1Slide13
Fast
Analog to Digital Conversion
... physical signal is just a function of time: digitize it
as finely as
technically possible ...
Avoid naïve questions like
“what data sampling rate is best for my task”!
Fast, precise and very affordable ADC’s are now available => use them!
Raw data are always a bonus: the more of them, the better!If you don’t know yet how to process them efficiently, never mind:processing algorithms keep evolving, raw data don’t.Note about the required ADC resolution (in bits):Fast raw-signal sampling rate As (say, 4 Gs
/s) reduces the needed ADC resolution: Post-processing narrows the band-width to the MR spectral window Sw
(say, 20 kHz), inducing a resolution enhancement by log2(
sqrt(As/Sw)) bits (8.8 in our example).
100 points
1000 points
10000 points
#
2Slide14
Total Monitoring
NMR instruments are complex, and so are pulse sequences. They generate lots of signals.
So far, we tried to keep just the data we understood best, which are a tiny percentage.
It is time to stop such a waste!
Ubiquitous
monitoring
The ideal
: monitor every input/output to/from the probe (and not just from the probe)And yes, monitor the inputs as well: don’t just assume what went in: monitor it!)
Assiduous monitoringThe ideal: never stop harvesting the data, keep acquiring at full ratefrom the moment your sample is loaded, until it is unloaded.It is not strictly necessary but, once the hardware is there,why not use it – the usage is completely free!
The term ‘total monitoring’ was coined in 2009 by an Agilent engineer in Palo Alto while we were discussing a blog entry of mine, but I do not remember his name now. I hope this is not what drove Agilent out of NMR.
#
3Slide15
Console
Full take-away
MR Session Record
Total Monitoring: Ubiquitous
... make the state of the whole instrument a matter of record, integrated with the data ...
What can one monitor during a MR data acquisition session?
All
RF outputs and inputs to the probe
(even when ‘unused’),
... including the
lock RF channel
(field stability)
Field flux-stabilizer
output (when
present
)
Shim currents (stability, spikes)
Sample temperature (an actual record)
Sample spinning (value, regularity)
Power
supplies (spikes, ripple, stability
)
Flow of gases (spinning, cooling)
etc.
The Bottom Line:
The stored raw data should include a log of the full
instrumental environment during an whole data acquisition session.
# 3
Buzzwords:
Avoidance
&
self-correction of artifacts
Objective raw data Quality CertificationSlide16
Total Monitoring: Assiduous
... monitoring everything is fine, but not enough: make sure you do it all the time ...
... even
during recycle delay
(receiver noise level is very informative),
... and
during the whole sequence
: some signals are generated all the time,... and on apparently ‘unused’ channels: there are signals on them as well,etc.
# 3
Ye olde faithful: HSQC with its DEPT preambleSlide17
Total Monitoring: Assiduous
... just think how much information are we throwing away once it gets tough ...
The Bottom Line:
So far, we kept the bare minimum of data.
In so doing, we were throwing out a lot of useful information.
# 3
Copied from
S.Berger
,
S.Braun
,
200 and More NMR Experiments, A Practical Course
, Wiley, 2
nd
Ed. 1998. Slide18
Greedy Mass Storage of
Raw Data with
No Preprocessing
... before we go into it: what do I mean by
‘preprocessing’
...
Striking-out most items, one:
Simplifies hardware Gets better control of data Shifts much of R&D to software! Achieves a huge flexibility boost Enables new, cute algorithmsDo not even accumulate (!),just record
every scan separately:an appropriate software can handlethe data better than any wired stuff!Examples of preprocessing steps:Analog: * Plain amplification (always needed) * Antialiasing filters (some are needed) Analog or Digital (firmware): Down-conversion (with quadrature)
Low-pass low-frequency filters
* Analog to Digital conversionDigital (on-board firmware):
Cartesian to polar conversion (optional)
Firmware accumulation (averaging)
#
4
Intro
The Bottom Line:
Any modification of the signal, even analog, is a processing.
Digital processing is more flexible and, when done properly, usually better. Slide19
Greedy Mass Storage of
Raw Data with
No Preprocessing
When
it comes to memory
usage, we tend to be over-conservative in MR.
This has historic roots and we need to fight it actively!
In 1976 I was personally selling data storage at $ 1.0 /bit.Today, I am buying it at a local shop for $ 10-10 /bit.
The cost of memory was once a justification for nasty hacks.Much of analog preprocessing (down-conversion, filtering, ...) was doneto reduce the data volume. So were, in part, the data sampling strategies, such as sticking to an ill-interpreted Nyquist’s rule DW = 1/SW(this also had to do with the sluggishness of old ADC’s).
The Bottom Line:
Doing something
‘because it saves memory’ is no longer very intelligent!
#
4
Today, there are no such limits, but
we still often stick
to the old habits!
We even teach them to students!
2
TByte flash key
!
Jan 2017!Slide20
Greedy Mass Storage of
Raw Data with
No Preprocessing
You might say: But pushing this game as far as you imagine, we would need
over 1 TB of mass-storage to save an experiment! Is that not shocking?
The Bottom Line:
This is the
BIG DATA ERA and, eventually, MR data will be part of it !
# 4 8:40Shocking is to keep telling kids that the sampling rate
equals
1/SW whenthen truth is that it has to be
at least 1/SW, but the higher, the better!
No, it is not!
One TB costs 50 bucks,
and it is even reusable!
So what’s so shocking
about it?
My answer:Slide21
MR Signal Sensors
Going
broad-band
&
non-tuned Cluster sensing (multiple and composite sensors) Chiral sensing (exploit the circularity of MR signals) Differential sensing (suppress non-local noise and artifacts)
Direct electric field detection (and electro-inductive sensors) Better coupling to the nuclei (the old transformer principle) Magneto-optical detection (using LED transceiver circuits) Interferometric optical detection (beyond classical ODMR)
... equally important: ongoing strife for a better understanding of noise ...
Group II
The Bottom Line:
Dear old
Plain Tuned Coil
, after 72 years, is it not time to retire?Slide22
Going Broad-Band
& Non-Tuned
With RF receivers based on a metal
‘cat whisker’
touching a galena crystal,
the tuned circuit was an
enabling technology - well worth the Nobel prize!
K. F. Braun – The Prime Tuner -
dixit:
< RF receiver ca 1900
Crystal of galena >
History:
tuning & matching started half a century before MR!
Later came other devices: solid-state diodes, vacuum diodes and triodes,
transistors of various types, with improving performance.
But the tuned circuit kept reigning supreme!
#
5
IntroSlide23
Going Broad-Band &
Non-Tuned
Tuned
&
Matched Coil
is to an
MR Specialist* as a nipple is to a newborn baby***
**
Since they got what they wanted, Glenn and the boy are both so happy they look alike(but you can tell them apart by the glasses)# 5 IntroSlide24
Going Broad-Band &
Non-Tuned
But outside MR, communications (especially military) are becoming ever more broad-band.
How is that possible? What has happened?
Let me try and explain it in five graphs.
They show the same RF frequency window on their
linear horizontal scale
, anda signal composed of two peaks, plus noise, on their logarithmic vertical scale.
The horizontal scale covers a frequency window (here in MHz). This is the kind of display common on spectrum analyzers.
‘True’ sensor noise and two Lorentzian-shaped signals.
These might represent the ‘raw’ inputs from an MR sensor.
20 dB/divison;
10 dB in terms of amplitude
S/N = 150
S/N = 100
#
5
IntroSlide25
Going Broad-Band &
Non-Tuned
Simulated acquisition of the raw data (light gray) by a receiver with an
equivalent input noise
(red
line
) larger than the raw data noise (the dots are samples taken at the receiver output).All signals are either lost or very doubtful!
?
# 5 InterludeSlide26
Going Broad-Band &
Non-Tuned
Passing the raw signal through a passive tuned circuit with Q=20,
the raw data within the tuning band get amplified (resonating tank
circuit effect). The Signal/Noise ratio does not change.
Since these are no longer ‘raw’ data,
they are plotted in a different color.
Tuning is a kind of early analog processing with its own artifacts.
Passing the resulting tuned signal through the noisy receiver, the tuned signal is almost completely recovered, with almost no loss in S/N.
Wow! Tuning works! But note that the other signal is still lost.
#
5
InterludeSlide27
Going Broad-Band &
Non-Tuned
The Bottom Line:
since evolution makes receivers less and less noisy, while ‘natural’ signals remain what they are,
tuning becomes less necessary. It was just a hack to overcome a technological weakness.
So why do we tune at all !? Let us try without:
Now we replace the receiver with a less noisy one (the cool, blue noise trace).
We see a perfect recovery of BOTH signals, regardless of the tuning band! Wow!
#
5
InterludeSlide28
Going Broad-Band &
Non-Tuned
Graph 5a
Graph 5b
Advantages for MR:
- Forget tuning and matching
- Use the same receiver device
- over a broad frequency band to
acquire simultaneously all present radio signals and store them all in a single digital record- Imagine having just one probe, fit to carry out any NMR experiment, involving any number of nuclides!
Current uses outside MR:
-
Software-defined radio (SDR)
- Military electronics (air combat)
- Satellite communications,
etc
...
Want to tinker a bit? The items below go for $30 /apiece
#
5
InterludeSlide29
Going Broad-Band &
Non-Tuned
... an example of a real broad-band, non-tuned NMR probe ...
Graph 5a
Graph 5b
Fratila R.M., Gomez M.V., Sykora S, Velders A.H.,
Multinuclear nanoliter one-dimensional and two-dimensional NMR spectroscopy with a single non-resonant microcoil
. Nat. Commun. 5:3025 doi: 10.1038/ncomms4025 (2014).
#
5Slide30
Going Broad-Band &
Non-Tuned
Graph 5a
Graph 5b
... some 2D examples, using the same BBC microcoil all the time ...
32 turns,
OD 2.5 mm
#
5Slide31
Going Broad-Band &
Non-Tuned
... a broader view ...
Graph 5a
Graph 5b
# 5
Basic facts about inductive sensing
A tiny magnetic dipole moment m
0
gyrating with frequency
at the center of a loop with diameter D
induces therein the EMF
U(t) = U cos(
, where
U =
m
0
/D
For a spin-S particle with gyromagnetic
ratio
,
precessing
at Larmor frequency
in a magnetic field B,
this gives
,
and
U = (
/4)k(2S)(
)
2
B/D
For
a single proton, D = 1 cm, B = 1 T, this givesU = 474 yV (yocto = 10-24)A 1 cm sphere of fully polarized water would induce in a fitting single loop almost 16 V.
At T = 300 K, the Boltzmann polarization factorb m0B/kT evaluates to 3.4*10-6, reducing the induced voltage to ~50 V (realistic).Looks good, but that is a 111 molar sample with just one peak, and a nucleus with a very high !
So, will tuned circuits disappear?
50
V is a large signal, but in real life we need to handle sub
nV signals from millimolar samples. Marconi and Braun would never doubt that to handle those, tuned receivers were a must.Yet, the problem is not at all the signal intensity! It is exclusively the intrinsic sensor noise versus the equivalent input noise of the receiver. With today’s electronics it looks unlikely that we might discover sensors with noise low enough to pose a challenge for modern preamps. But who can say?
We know far too little about the noise!With this reservation I would say:Yes, tuned receiver circuits should disappear.Note: With large coils, we still have a problem with broad-band generation of the needed B1 values which will probably push us to abandon the concept of single, time-shared
Tx/Rx coil.But that is another story ...Slide32
Cluster Sensing: Multiple and
Composite Sensors
When the coupling of a sensor to a signal source is weak/null (source unaffected by the presence of sensors), using multiple sensors improves detection and enables new possibilities.
We can average between multiple sensors (better signal)
o
r serve their outputs to different ‘clients’ (this case).
Sep 1943: Franklin D. Roosevelt announcing Italian armistice
#
6 IntroSlide33
Cluster Sensing: Multiple and
Composite Sensors
MR
is a
“near”
phenomenon, not a
“far”, radiative one(no radio waves are leaking out from the sample area).
The Bottom Line:Using multiple sensors in MR is a sensible thing to do!
... an existing example: the phased-array MRI coil assemblies ...It follows that some
sensor-source interaction
is to be expected and, indeed, it seems to exist.But
the coupling to the nuclei is very
weak. We use only a tiny fraction of their energy
, one which we can make arbitrarily small
.
#
6
InterludeSlide34
Cluster Sensing: Multiple and
Composite Sensors
... noise considerations pertinent to sensor clusters (examples of a few selected schemes) ...
One sensor, multiple preamplifiers
Equivalent to one preamp with
e
in
divided by N---------------------------------
S ... Signal from a single sensor ... Single sensor ‘true’ noise s.t.d.ein .. Preamp equivalent input noiseA ... Preamp amplification (linear)
Multiple sensors, one preamplifier
Equivalent to dividing
by
N and
e
in
by N
----------------------------------
S | r ... Signal S versus casual noise r
... Plain summation symbol
N ... Number of repeated elements
Note: problems related to driving the adders are not tackled here.
Multiple smart sensors
Equivalent
to
dividing both
and
e
in
by
N
----------------------------------
Note:
Instead of plain summation, one can store the data separately and do more tricks with them
#
6
IntroSlide35
Cluster Sensing: Multiple and
Composite Sensors
... decomposing
old
coil geometries ... and designing new ones ...
#
6
Wait a momentAny conductive strip is an inductive sensor!It’s fun to design assemblies such as this one
Traditional coil geometries can be split/replicated
.
Note:
Solenoids
are also easy to split into
transversal sections or into coaxial layers
Helmholtz pair
Penta-coil
Striped Crown of Smart SensorsSlide36
What is in this single camera image?
Looks just some rough surface.
Chiral Sensing
... beyond quadrature: full exploitation of
the
chirality
of MR
signals ...
# 7
All MR signals are generated by a precessing magnetization,
while almost none of the noise is precessing!
This is a HUGE
chunk of
a-priori
knowledge to exploit
Wow: pattern recognition using a broader, dynamic view
p
ermits to separate ‘desired signal’ from the ‘noise’Slide37
Chiral Sensing
... sifting the gold from the dirt: hints on data analysis ...
#
7
The Bottom Line:
Functional
filtering algorithms
will become a
buzzword of future NMRN-dimensional chiral data filteringLet the data be digitized at regularly spaced times tk in a way that comprises exactly D samples. Then the true signal obeys the recursionS
n+1
(tk
) = S
n(t
k-D
)
valid for any
n <
N
and
k D.
Since the noise in the individual sensors does not do this, one can write an algorithm which filters most of it out.
The result is a S/N boost of 2
(N-1)/2
With a Helmholtz pair we have N=2 (a boost of 2 over a single coil), and the algorithm is hard-wired: a plain sum of the two coil voltages.
With N sensors, the boost over plain Helmholtz pair could be as high as
2
(N-2)/2
, giving 128 for a 16-segment sensor.
n-th sensor stored data: S
n
(t), n = 1,2, ..., N-1.
Fast digitization (preferably direct).
Since nothing much changes in an NMR FID in 1
s, for many Larmor cycles (e.g 100 at 100 MHz) the
true
signals (but not the noise) look almost identical, just time-shifted a bit, from one sensor to another:
Sn+1(t) = Sn(t-), = (2/)/NSlide38
Differential Sensing
# 8
... the goal: discriminate signals generated in the sample from nonlocal interferences ...
Exemplified on an N=3 striped crown
Basic principles
Sensors dislocated in a way to ‘feel’ the sample in different (but known) ways allow to
discriminate
between the signals arising from the sample and nonlocal interferences.
Functional filtering algorithms again!
I
maging with
multi-sensor assemblies, possibly movable ones
(like in X-ray CT).
In all such approaches, the prerequisite isdata acquisition from all sensor terminals
and massive post-processing.Slide39
Electric-Field Detection
#
9
8:50
From Maxwell equations we know that a time - varying
magnetic field must be associated with an electric field:
The magnetic field of a rotating magnetic dipole therefore produces a rotary electric field (impossible in static situations). This can be detected through-the air (no galvanic connection), using a high-impedance, electrometric potential sensor (no current).
R. J. Prance, A. Aydin, Phys.Rev.Let., 91, 2007, DOI 10.1063/1.2762276.
The upper trace shows a conventional free induction decay signal as a function of time for a sample of glycerine. The lower trace is a preliminary result for an EPS coupled through the wall of the sample tube
.From: R
J. Prance et al, Biological and medical applications of
a new electric field sensor, Proc. ESA Annual Meeting on Electrostatics 2008, Paper
N2.There are also numerous indications that capacitive electric field detection is at play in some on-chip NMR sensors developed by various groups.
Induction coil
Electrometric pick-upSlide40
Classical inductive sensors can be used concurrently with electric field sensors.
This can both boost the sensitivity of the sensor assembly and, in imaging versions, improve the power of spatial discrimination.
Electro-inductive
sensor optimization study is needed to maximize the electric signal.
Again, I assume that all the sensors are individually monitored.
The electric field generated by precessing magnetic moments has a complicated, but computable, spatial distribution.
Electric-Field Detection
#
9
... towards hybrid
electro-inductive
sensors
...
Violet dots: example locations of electric field sensorsSlide41
Enhanced Coupling to the
Nuclei
... the old
transformer
principle ...
#
10
The concept
When an interaction between objects S (source) and D (detector) is too weak, one can insert between them another object R (relay) which couples more strongly with both S and D. This boosts the effective coupling between S and D very significantly: Source ................... Detector S ==== R === D Weak coupling Stronger effective (relayed) couplingIn the case of magnetic induction, the best known example are transformers, with R being the ferromagnetic yoke between the primary coil S and the secondary coil D.
But it is also at play in the
Dimitris Sakellarius’ MAS microcoil assembly
with an intermediate small coil (in this case integrated with the rotor) placed between the sample (nuclear magnetic moments) and the standard, rather remote static coil
(Sakellariou D, Le Goff G, Jacquinot J-F, High-resolution, high-sensitivity NMR of nanolitre anisotropic samples by coil spinning
, Nature 447, 694-697, 7 June 2007).
Note: A number of other papers, both in NMR and MRI, have used inductive coupling to
relay the signal further away, and some of these reported an associated boost in S/N.Slide42
Enhanced Coupling to the
Nuclei
... the ol
d
transformer
principle ... continued ...
# 10
NMR probe with an RF transformer insert between the sample and the coil?We can not place any ferromagnetic material in the sample because that would destroy homogeneity, and also because they saturate and lose eefficiency. On the other hand, non-ferromagnetic materials have too low susceptibilities to be of any help.But we can place a coil between the sample and the main coil, like in the Sakellariou’s arrangement. We could make a cylindrical insert and print on its surface (or volume) many coils, forming a super-inductive meta-material. Using the planar microcoil design, we should be able to make it broad-band as well.
Planar multi-coil
(multi-sensor).
Each coil relays a tiny induced moment.
Proper scaling makes it advantageous (keep constant vertical trace thickness). Math!
Same thing, but wrapped on a cylinder:
An inductive insert.
Bi-layer sheet:
The Bottom Line:
Why not! This is something to pursue.Slide43
MR Spin-Off: 2D and 3D hyper-inductive meta-materials
... realized on various scales, maybe 3D printed ...
It should work somewhat like ferromagnetic sheets and bodies, but with
no residual magnetization, no saturation effects, and at frequencies up to UHF
The Bottom Line:
This points far beyond MR, and sounds soooo fashionable, too!
2D bilayer, axial
3D multilayer, axial
Cylinder, biaxial
V
arious geometries
, some of which
most unusual
Note:
replacing simple loops with
spiral, multi-turn coils
makes such designs
chiral
!
InterludeSlide44
Outline of basic principles:
Many early NMR instruments used a biased diode between the coil caps as a detector (later replaced by triodes, and then transistors).
A LED, electrically, can still do the same job!
But a LED does,
at the same time
, emit a light signal which can be fed into an optical fibre and brought out of the sample & magnet area for a possibly less noisy optical detection.
If this approach works, we might ask for – and likely get – light emitting low-noise transistors suitable as electro-optical preamplifiers.
Wow! One device generatingtwo signals of different kinds!
Combined Electric and Optical Detection... using a LED (light emitting diode) as a detector ...
#
11
less noisy?
Devices are on the way!Slide45
Interferometric Optical Detection
... beyond ODMR ... how might universal FIR
- UV
detectors work ...
#
12
Current optically
Detected EPR is great, but too specific (limited applications). But suppose that there exists a material with a sharp spectral line (just one) that gets affected by external magnetic field, without saturating.Then optical sensors might become feasible! Note: the ‘Pinuts’ played a lot with optical NMR detection at very low fields, but from a different angle and with limited applications:pines.berkeley.edu/research/optical_detection
Note:
this would be compatible with multi-sensor approaches: think about a ring of sensor rods (or capillaries) around a sample, each with its independent optical fibre path.
Potential S/N boost
due to the (unofficial)
frequency-conversion principle:
RF
(IR | VIS | UV)
To look for:
sreferably strongly anisotropic
magneto-optical materialsSlide46
Interferometric Optical Detection
... interferometric arrangements like this might be
feasible
: ...
#
12
The Bottom Line:
Look for molecules whose IR–UV spectra are in some way affected by magnetic fields!Optical paths (fibers) would just pass by the sample, with the detector being outside.Optical phenomena to consider: phase delay, dichroism, plane rotation, ...Slide47
Nuclear Magnetic Polarization
In-situ (optical) polarization
Fully polarized
electron
beams
... boosting sensitivity by fighting Boltzmann (just some thoughts) ...
Group III
Logo of San Miniato Chianti Workshops on MR Nuclear and Electron Relaxation.Permission by courtesy of the late Ivano Bertini.
Dilemma: We want these guys to be either all UP (E
2
) or all DOWN (E
1
).
By pulses, we can only flip (interchange) the UP / DOWN states
Slide48
In - situ polarization
... ? can we avoid mechanical transfer of sample and/or of polarized materials ? ...
#
13
The Bruker
gyrotron
set-up for ex-situ polarization by unpaired electrons which are 658 times more polarized than 1H (and much more than that for many other nuclei.
We also heard a lot about guys hauling around magnetic flasks with pre-polarized
129Xe across labs, buildings, and even distant towns. It stays polarized for hours!
But we should stress approaches which avoid mechanical transport:
Mixing paramagnetics with sample and irradiating in-situ to achieve electron – nuclear cross polarization (pros: no material transfer, cons: contamination, problematic at high fields).
Exploiting just changes in relaxation rates for faster repetition times in the presence of tolerable paramagnetics, such as
pressurized O
2
(a Thesis at the University of Santiago de Compostela,
www.usc.es/ciqus/es/noticias/tesis-ciqus-Luiz-Pinto
)
Optical in-situ optical pumping would be ideal, but it is hard to make it universal!Slide49
Polarized - electron guns
... just an idea - I am not sure whether it might ever work ...
#
14
The Bottom Line:
This might be worth researching, though I have growing doubts that it will work
My doubts
:
Why such devices are not yet on the market ? * Maybe they can not be made to work, or * maybe there is no market for them yet.
Can the polarized free electrons get in contact with molecules fast enough to transfer the polarization before loosing it all ?
How conductive should the samples be?
Generator of fully polarized electrons
Exploitation scheme:
Slide50
MR Signal Excitation
Independent excitation
(back to crossed coils)
Thermal excitation
(monitoring spin noise) Weak excitation modalities (there are many)
... which is apparently less amenable to innovation than signal detection ...
Group IVTo generate 90o RF pulses, we need to generate RF fields of about
0.5 – 10 mT over the whole sample volume.
This is
easy for microsamples (even in broadband mode),
but progressively more difficult for larger volumes
( 0.5–100 cm) where large currents are needed
(1–100 A.turns).Slide51
Independent
Excitation Hardware
...
back
to the future:
crossed
coils again? ...
# 15
The Bottom Line:Return to independent transmitter and receiver gears looks likely (with exceptions)!Technical aspects of signal detection may conflict with the need to generate strong RF fields, especially when one insists on maintaining the concept of a single Tx/Rx coil.
Crossed coils:
1945 - 1980
Single coil:
1975 - today
PROS
Implicit Tx/Rx isolation (40 dB)
Tx can be tuned even while Rx is BB
Separate tuning/matching
Less noisy
Simpler geometry
CONS
Hard Tx/Rx isolation (diodes,
/4)
Electric constraint: same Tx/Rx properties
More noisy: Tx noise reaches RxSlide52
Thermal Excitation
...
the simplest way to solve the transmitter problems is to eliminate the
transmitter
...
#
16
... And it has some most intriguing features:
Signal is proportional to n rather than n (number of nuclei), which is advantageous for micro- and nano-samples.Signal does not saturate and can be acquired continuously.Signal is proportional to 2, not 3, in favor of low nuclei.
Signal does not drop with temperature.
Moreover, I believe that the spin noise signal can be boosted by refining sensors
as discussed before and, in addition, by applying novel evaluation algorithms.
The Bottom Line:
This will hardly become a mainstream, but it definitely merits attention!
Spin noise MR
is
well known
(Sleator et al , PRL 1985), but presently it does not look as a practical option because i
ts low S/N is about four orders of magnitude below ‘normal’.
<=
But it is viable, ... =>
M. Nausner et al, Nuclear magnetic spin noise spectra, 49
th
ENC
N. Müller, A. Jerschow, Nuclear spin noise imaging, PNAS 2006Slide53
Weak Excitation M
odalities
... Fast-sweep CW, Hadamard-like, random ...
#
17
The Bottom Line:
Weak excitation looks as an excellent compromise
Up
to a few Watts, RF power amplifiers are tiny, on-board affairs.Nothing like the large and heavy 1 kW boxes.
They can hardly generate 90
o
pulses (except in microcoils), but they can produce pseudo-stochastic correlations, similar to spin-noise, but orders of magnitude stronger, bringing the sensitivity
almost back to normal
.
Such low-power excitation modalities might yet become common
The
key
is
fast acquisition
and novel
evaluation algorithms
. Slide54
Magnetic Field (in)Homogeneity Shimming
Controlled
displacement
actuators
for passive systems
Controlled percolation with paramagnetic fluids Mechatronic, using controlled swarms of nano-robots Eddy-currents, using rotating disks, plates, spheres, ...
beyond current passive shims and Golay coils
Group V
During the last decade, I have witnessed several
would be table-top NMR spectrometer makers
follow this path:
3 months (ahead of time): develop a magnet for the target field
Getting very optimistic and planning 1
st
spectrum in 6 months
After 3 years, gloomy and depressed, still working on the shims
In some cases, giving up (not all, fortunately)
The Bottom Line:
Shimming the field is a major hurdle, up to 15 T worse than getting the field!Slide55
A word about electric currents shimming
... which I would not know how to improve, not in principle, but I have to mention it
The Grandfather of NMR shim coils (+ many other inventions)
Marcel J.E. Golay 1902-1989
Before 1958, many tricks were used to reduce the inhomogeneity of the field below several Hz:
Matching probe position to match the sweet point.
Separating pole caps from the yoke by a folium.
Exerting asymmetric tensions on the pole caps
Placing ferromagnetic rings and other objects at strategic and carefully adjusted locations.
All these
passive shimming methods, still in use,
do not suffice to reach the desired homogeneity.
Presented with the problem, Marcel Golay, a mathematician well versed in Maxwell equations, immediately saw the solution:
active, current bearing, electric coils
to compensate any magnetic field gradients in a ‘magnetically empty’ inner space.
We still use those in every NMR!
also US Patent 3,569,823
0 references,
hehe
0.15 Hz !
IntroSlide56
Pros and Cons of Golay shim coils
(nothing is ever perfect)
Cons:
Bring outside noise into the sample area
Possible heating problems
Can not correct local gradients present in mechanically complex probes
Correct a modest number of harmonics (8 – 32), not generic field profiles
In general, they still just barely reach the desired homogeneity.
Pros:No mechanical movementsEasy remote, digital control of currentsHuge range (>1:106), smooth settings Permit dynamic & automatic shimmingAdaptable to variously shaped volumes (flat/long cylinder, cuboid, ...)Occupy relatively little spaceCan be pre-computed to about 80%
The Bottom Line:
After 60 years, one would have expected something new to pop up!
Particularly considering that shimming is still such a problem.Slide57
Controlled Displacements of Magnetic Materials
Passive shims revisited (just a couple of untested ideas)
#
18
P
ercolation systems:
Another possibility is a controlled percolation of paramagnetic liquids around the sample area, possibly combined with modern microfluidic techniques (percolation systems).
Electro-mechanical actuators:
There exist electrically driven, magnet friendly actuators based on piezoelectric effect, or on special polymers, which might permit very precise displacements of passive shim objects even during run-time (just like Golay shims)
First of all, let it be said that one does not need to use just
s
trongly invasive ferromagnetic materials to move around.
One can use gentle paramagnetic materials.
This said, I see two possibilities to explore:
The Bottom Line:
There ARE ways to explore, and they overlap with stuff other people are developing! Slide58
Mechatronic Shimming
by controlled
swarms of nano-robots
#
19
9:00
But what if
the dust were made of
nano-robots with ferro/para magnetic insets and tiny feet, whom we could send (by WiFi) to where we want them to go ???The Bottom Line:Ok, let us better keep this for the end of the Century! But think about most dust being paramagnetic - and keep your lab and probes clean!
We can not just blow
paramagnetic dust
into a probe!
It deposits in places where the field is already stronger,
boosting the field inhomogeneity
, for much the same
reason a ‘ferro fluid’ behaves bizarre in magnetic fields.
Nor would a
diamagnetic dust
help!Slide59
E
ddy Currents
... Let Nature do the calculations! ...
IntroSlide60
E
ddy Currents Shimming, the idea
... using
rotating disks, plates, spheres, ...
#
20
So, let us see again: eddy currents try to stop whatever caused them.
The motion of conductive bodies across field gradients, for example.
But what if we do not let them affect the motion?Simple, Watson: They will suppress the gradients themselves!The Bottom Line:Eddy currents in spinning bodies look as a new paradigm in field shimming!
Rotating a non-diamagnetic metal cylinder around the sample & probe assembly should have much the same effect like sample spinning.
Actually better, because it really makes the field more homogeneous and therefore produces no sidebands and has none of the adverse effects on the refocusing spin echoes used in many sequences.Slide61
E
ddy Currents Shimming,
fun with geometries
... evolution of the species: spinning disks, spheres, split cylinders, ...
#
21
Right:
Spinning sphere and asplit spinning cylinderFar right:Spinning cylinder andspinning disks
The Bottom Line:
It appears that there is a lot to explore here
Some bi-axial
examples
(
just to convey the idea
)
Component parts for spinning assemblies:
Cylinders, split cylinders, twin disks, spheres with a bore, spherical caps, ...Slide62
E
ddy Currents Shimming
... evolution of the species: using (again) anisotropic metamaterials, ...
#
21
In MR, we care only about B
z field gradients (3 out of 9).Can we exploit the fact? The answer is YES, of course
The Bottom Line:There appears to be more and more to explore here The sketch on left (bottom) shows a spinning cylinder made from stacked plates of an insulator printed with a dense covering of conductive metal loops.In this way, all induced currents run in planar loops, coaxial with Bz. As an eddy shim, the cylinder therefore responds effective only to the Bz component of the field, but it does so more efficiently than a massive metal body because the 3D electric current flow equation does not interfere with the Lentz’s law (no cross-layer couplings).Slide63
Magnetic field (
in)Stability Management
Frequency tracking
instead of field
lock (a simple control loop) Reference NMR signal, but no lock: just ubiquitous monitoring No lock, nor tracking
... beyond current NMR field lock systems ...
Group VISlide64
Traditional NMR Field Lock
Current instruments use
internal or external NMR lock sensor
, synchronized with a master clock, and a feedback loop to the magnet control, to stabilize the magnetic field strength (top stability requirements go up to 0.01 ppb)
Traditionally, magnetic field is what being controlled,
while the master clock is a reference.
* The Master Clock does not need to be very stable: if it varies, the field varies,
but so does the Rx reference frequency, and the spectrum looks exactly the same.
IntroHere, the magnet control feedback is a crucial piece of electronics with its own noise and many possible weaknesses (a frequent source of problems/ artifacts.The observe data are stable enough to be accumulated in hardware, if so desired.
*
Classical NMR LockSlide65
Frequency Tracking
instead of Field Lock
... a much simpler and more versatile approach ...
#
22
With modern frequency synthesis, there is no need to control the magnet.
Rather, one can use the lock signal to synchronize the Rx reference clock
(the frequency-control loop is easy to implement in an FPGA):Replacing field locking with frequency tracking, one saves a lot
of analog hardware with
NO compromise in performance.
The FID’s look exactly the same with both methods.
*
*
Field Lock
Frequency tracking
firmwareSlide66
Reference NMR Signal, but no Lock, nor Tracking
... just ubiquitous signal acquisition & data post-processing ...
#
23
Magnetic field fluctuations produce phase drifts of acquired NMR signals, but do not affect signal magnitudes. Knowing the time evolution of the lock signal in each scan, one can correct any observe signals to what they would be in a perfectly stable field.
Note:
For this modern approach, monitoring of the lock channel,
and the individual storage of every observe scan, are a MUST !
No lock, nor tracking
Frequency tracking
firmwareSlide67
Neither Lock, nor Tracking ?!
... who said that we really need any lock signal at all !? ...
#
24
NMR lock is a complete NMR spectrometer and, in its internal variety,
it also imposes constraints (deuterated solvent) on sample composition.
Can we do away with it?
With assiduous acquisition
& storage, maybe:Mathematically, any FID can be used as a self-reference. Algorithmically, the a-priori knowledge that it is a reproducible transient is sufficient to correct field fluctuations, provided that a single-scan signal has good S/N.In such a scenario, when many scans (N)
are taken, field fluctuation effects drop with
N. The approach therefore works best a large number of scans
is anyway required.
The Bottom Line:
This is not universal, but it might come handy sometimesSlide68
Novel Phenomena to
Search for and/or to Study
Magnetic molecules and nano-systems
(to be used
in
sensors) Sharp spectral lines sensitive to magnetic fields (for sensors, again) Sources of noise (we do not know enough about it by a far shot!) Field
(in)dependence of chemical shifts and J’s (never properly tested) Spin radiation: passive (MR in astronomy) and active (remote-detection) Molecular spins (persistent currents in a-cyclic molecular fragments)
... which might be useful in MR ...Group VII
Luigi Mussini:
The Triumph of TruthSlide69
Magnetic Molecules and Nano-Systems
... to
be used in
all kinds of magneto-optical couplers and sensors ...
#
25
Single-molecule magnets are today a very hot topic. They are studied for many good reasons, but we should keep an eye on them because they are the most likely ones to be able to couple magnetic fields to a passing light.
Note: we do not need the magnetically extreme structure because they saturate in high fields.
The most interesting ones might be those discarded by the mainstream. Slide70
Sharp Spectral Lines Sensitive to
Magnetic Fields
We should search for them systematically from microwaves to far UV
(People almost never stick
their molecules into strong magnetic
fieldsin order to measure spectra other than NMR or EPR)We might be missing a whole world of phenomena and opportunities!
# 26
The Bottom Line:There are myriads of spectral lines and a few might be right for us!
... again, for their potential use as magneto-optical sensors ...
An almost random pick of a spectrum
exhibiting some very sharp UV lines:
De Sousa et al,
Incorporation of europium III complex into nanoparticles and films
...
,
DOI: 10.1590/S1516-14392010000100015Slide71
Sources of Noise
... because of the ‘N’, S/N theory is still a virgin field ...
#
27
Since years I am trying, for my own peace of mind, to find out what might be the ultimate, theoretical S/N ratio in NMR.
NO SUCCESS SO FAR
, Sorry!
The Bottom Line:One sure thing: S/N ratios are dominated by technology, not by basic physics!Hence, there is a lot to do and there could be orders of magnitude to gain!
Planck’s
noise
power
K.T.
2
Planck’s
noise
amplitude
K.T
½
.
Planck’s
induced noise signal
K.T
½
.
2
If only Planck’s noise existed, S/N should not depend on !
Since h/k
B
T << 1,
Which is not at all born out by practice
we see nothing like “black body” noise !!!
A glimpse of ‘problems’
But!:
Where is any
‘black body’?
And is T that
of sample?
or coil?
or enclosure?
There exist
several tens of potential sources
of EM noise (Johnson, Planck, dielectric, semi-conductive, shot, 1/f, ..., each with
many empirical formulas
.Slide72
Field (
in)Dependence of Chemical Shifts and J’s
... we now have 20 MHz and >1300 MHz spectrometers; it is time to (re)check this ...
#
28
It is taken for granted that chemical shielding is proportional to B
0.However, there is no reason why it should be an absolute rule,except for a heuristic feeling that, since it is very small, any higher-order coefficients should be still smaller, in a similar proportion.This might be true, but it does not need to.The Bottom Line:It is time to check if chemical shifts and J’s are really independent of B
0
If any coefficients of the dependence of chemical shifts on B
0
were
measurable
(at least in some cases), they would be certainly dependent on local molecular structure and therefore could be used for chemical interpretation.
They might even provide a
raison-d’ȇtre
for the ultra-high field instruments Bruker is trying to build and hoping to produce in large series
Think about the aromatic ring currents, for example.
Might not a saturation-like deviation from linearity start showing up?Slide73
Spin Radiation
... ok, remote MR (in astronomy, but not only) is my
old
hang-up ...
#
26
GIDRM 2006, ENC 2009, ???... 2009
I now think I know how to coax spin radiation out of a sample and detect it elsewhere.
But that is another story, yet to come.Slide74
Molecular S
pins
... persistent
currents in
molecular fragments are, instead, my
emergent hang-up ...
# 30
The Bottom Line:We might have new magnetic particles (molecular spins) with low with which to do Molecular Magnetic Resonance (MMR) ... to be continued
ENC 2012Slide75
End of Proposals, start of (gloomy) PerspectivesSlide76
Perspectives (I): the Academic point of view
Today, to pursue a research goal in the Academy, one must formalize it and get it funded. That sounds reasonable, but it has many defects, too.
For example, just about one out of 10 proposals gets funded
On top of it, the proposals I made here look as
‘mere technological R&D’ with little academic appeal. Publishable, but there is no Basic Science.They do not address any of the buzz themes likeBig Bang, Gravitational waves, Origin of Life, ... not even Cure for Cancer, ...Therefore, they address nothing! Just MR!Sorry, try again next yearSlide77
Perspectives (II): the Industrial point of view
You might think that a large manufacturer (is there any left?)
should invest into such proposals.
In theory yes, but they will not do it, for a good reason:
the market does not press them to do it
.Small Companies (there are a few!) do not have the resources,and they need to develop new Applications first.So this leaves very few medium-size Companies!Guys, please, Go For It!Slide78
Thank you
for your
Attention and
Patience
Thanks also to the many people with whom I had the pleasure to discuss these things,
and to Ester Maria Vasini without whom I would have never finished all these slides,nor even started thinking about writing it all down in a planned article..
Next Talk Announcement:The NMR operator in XXI Century:30 selected things to do and exploreLet us stay in touch!
$ 39.99The slides of this talk are available at DOI 10.3247/SL6Nmr17.001