Document [original]

SPECIAL ISSUE RESEARCH ARTICLE

Contributions towards variable temperature shielding for

compact NMR instruments

Martin Bornemann-Pfeiffer

1,2

| Klas Meyer

| Jeremy Lademann

Matthias Kraume

| Michael Maiwald

Bundesanstalt für Materialforschung und

-prüfung, Berlin, Germany

Chair of Chemical and Process

Engineering, Technical University Berlin,

Berlin, Germany

Correspondence

Michael Maiwald, Bundesanstalt für

Materialforschung und -prüfung (BAM),

Richard-Willstätter-Straße 11, Berlin

12489, Germany.

Email: michael.maiwald@bam.de

Abstract

The application of compact NMR instruments to hot flowing samples or exo-

thermically reacting mixtures is limited by the temperature sensitivity of per-

manent magnets. Typically, such temperature effects directly influence the

achievable magnetic field homogeneity and hence measurement quality. The

internal-temperature control loop of the magnet and instruments is not

designed for such temperature compensation. Passive insulation is restricted

by the small dimensions within the magnet borehole. Here, we present a

design approach for active heat shielding with the aim of variable temperature

control of NMR samples for benchtop NMR instruments using a compressed

airstream which is variable in flow and temperature. Based on the system iden-

tification and surface temperature measurements through thermography, a

model predictive control was set up to minimise any disturbance effect on the

permanent magnet from the probe or sample temperature. This methodology

will facilitate the application of variable-temperature shielding and, therefore,

extend the application of compact NMR instruments to flowing sample tem-

peratures that differ from the magnet temperature.

KEYWORDS

benchtop NMR, continuous processes, inline analytics, model predictive control, process

analytical technology, temperature control

1|INTRODUCTION

Compact NMR instruments have extensive

applications,

1–3

including reaction and process

monitoring.

4–7

Unlike high-field NMR, compact NMR

instruments are not restricted to laboratory applications

because process analytical technology analysers based on

these instruments have been demonstrated at industrial

scales.

8,9

The availability of rare-earth permanent

magnetic materials in combination with magnet array

designs

was an important foundation for the develop-

ment of compact NMR instruments with acceptable

weight and sufficient field strength and homogeneity

without the need for cryogens.

A disadvantage of using permanent magnets is the

dependence of the magnetic field strength and field

homogeneity on even the smallest changes in magnet

temperature.

12–14

As shown in Figure 1, magnet

Received: 21 January 2023 Revised: 23 June 2023 Accepted: 28 June 2023

DOI: 10.1002/mrc.5379

This is an open access article under the terms of the Creative Commons Attribution-NonCommercial-NoDerivs License, which permits use and distribution in any

medium, provided the original work is properly cited, the use is non-commercial and no modifications or adaptations are made.

Magn Reson Chem. 2023;1–10. wileyonlinelibrary.com/journal/mrc 1

temperature deviations of less than ΔT=0.01 K can

result in distorted NMR spectra. Relatively slow changes

or negligible deviations in the magnet temperature occur-

ring as a consequence of daily fluctuations in room tem-

perature or insertion of NMR tube samples with varying

temperatures can be compensated for by using an inter-

nal temperature control loop. The precise implementa-

tion of this control loop is the proprietary design of each

manufacturer. To the best of our knowledge, these sys-

tems comprise multiple temperature sensors and corre-

sponding Peltier elements for active temperature control

of the magnet, electronic components, and instrument

housing. For standard applications, such as sample analy-

sis in NMR tubes, and most flow applications around

room temperature, the internal controller can appropri-

ately maintain a stable operation of the magnet without

severe, temperature-induced fluctuations in magnetic

field homogeneity. When 5 mm NMR tubes are placed in

the magnet, the amount of heat transferred between the

NMR magnet and the NMR sample tube at room temper-

ature is sufficiently small. Slight deviations from the mag-

net temperature in flowing samples can be tolerated

when flow-through cells are used. However, when study-

ing chemical reactions or monitoring processes, it may be

necessary for the reaction to take place in the NMR

instrument itself, or the sample stream may need to be

maintained at a temperature much higher or lower than

the magnet temperature.

In such cases, the heat flux

between sample volume and NMR magnet is rapidly

increasing and may no longer be compensated by the

internal temperature control loop as shown in Figure 1

carrying out a continuous chemical acetylation reaction:

Because the magnet temperature increases with the liq-

uid sample temperature, (i) the internal variables of the

instrument (Power1 and Power2) controlling the magnet

temperature decrease (Figure 1b, blue symbols), leading

to a decreasing magnet temperature (Figure 1a, black

curve) at first. As the manipulated variables reach their

lower boundary (at 20 min), the magnet temperature

rises again until the chemical reaction is stopped

(Figure 1a at 28 min, red curve). The highest magnet

temperature (ii) is reached shortly afterwards with a

delay. This effect can be tolerated for a short time but has

deteriorating effects on magnetic field homogeneity, as

demonstrated in Figure 1c.

A heat source within an NMR magnet is by far not

new. Measurements at varying probe temperatures are

commonly performed in high-field NMR experiments;

commercial and custom-made variable-temperature

probes are available for this purpose.

16–18

These probes

adjust the sample temperature through either an electric

heater in combination with air or gaseous nitrogen flow

or a fluorinated liquid. Despite the nonnegligible heat

flux between the thermostated sample and the magnet,

the homogeneity of the magnetic field is less impaired

compared with the case of permanent magnets in com-

pact NMR instruments, as the magnetic field homogene-

ity of the electromagnet is less sensitive to temperature

changes.

The inline analysis of exothermic reaction mixtures

or mixtures with a significantly higher temperature than

that of the NMR magnet is a particular case of applica-

tion. To the best of our knowledge, none of the compact

NMR instrument manufacturers proved a commercially

available integral active temperature control or sufficient

probe temperature shielding and insulation for a longer

experiment or continuous operation at a deviating sam-

ple temperature. Some of the available instruments can

heat up the whole magnet. Still, they focus on maintain-

ing a constant probe temperature within a sample tube

for offline applications (e.g., for mimicking fixed reaction

conditions). However, reshimming the instrument after

changing the magnet temperature is time-consuming and

cumbersome, as shim sets vary widely from each other.

Current publications have addressed the issue of temper-

ature effects of NMR measurement quality

2,13,19

without

FIGURE 1 Continuous chemical acetylation within the NMR

magnet demonstrating overheating: (a) course of the reaction

mixture temperature and corresponding NMR magnet temperature,

with coloured indication of the time the spectra was acquired

(vertical arrows from a to c) and local maxima of the magnet

temperature (i/ii). (b) Mass flow of the reactant benzyl alcohol with

triethylamine and acetyl chloride and two internal power control

variables of the NMR instrument (Power1 and Power2 are arbitrary

internal control variables driving the actuators). (c) Superimposed

spectra with an indication of the time of measurement, which

deteriorate with time.

2BORNEMANN-PFEIFFER ET AL.

1097458xa, 0, Downloaded from https://analyticalsciencejournals.onlinelibrary.wiley.com/doi/10.1002/mrc.5379 by Technische Universitaet Berlin, Wiley Online Library on [20/02/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License

providing potential solutions for active shielding or

passive insulation.

As discussed later, developing an efficient tempera-

ture control is impeded by two boundary conditions.

First, the internal instrument control loop works

independently of a newly developed temperature control

system. To avoid the mutual interference of both control

loops, fluctuation in the heat flux from the flowing

sample to the NMR magnet must be reduced. Second,

comprehensive access to relevant temperature sensing

(e.g., surface temperatures within the instrument or the

probe) can only be achieved with major hardware

modifications. Therefore, a substitute system (introduced

in Section 3.1) is developed in this study, which allows

access to the surface temperature and later application to

different instruments. Based on surface temperature

sensing, a controller for active temperature control is

developed (Section 3.2) and successfully validated

(Section 3.3).

2|CONCEPT AND STRATEGY

The aim of this work is to contribute to the thermal

decoupling of the sample feedthrough from the perma-

nent magnet of a benchtop NMR spectrometer. Prelimi-

nary investigations with passive insulation did not prove

to be effective enough, although there was access to state-

of-the-art materials, such as modified fumed silica

(aerosil

Thankfully, we had a wide bore compact NMR proto-

type with a 12 mm inner diameter at our disposal

(Nanalysis Corp., Calgary, Alberta, Canada), which con-

ceptually can also serve as a basis for typical sample port

designs. Initially, a double-flow design was adopted, with

one airstream regulating the NMR sample temperature

and the other airstream thermally separating the sample

from the surrounding NMR magnet. However, this

design is complex due to its many filigree concentric

tubes and low wall strengths. As a result, the preliminary

experiments could not achieve effective sample tempera-

ture control and simultaneous thermal decoupling of the

NMR magnet. Hence, the control target is solely the

thermal decoupling of the NMR magnet.

2.1 |Variables affecting measurement

quality

The magnet used in compact NMR instruments is an

array of small and, in most cases, prismatic, rare-earth

permanent magnets. The arrangement proposed by

Halbach

remains superior and is used by most

manufacturers. The available energy difference in the

various nuclei states and, hence, the achievable signal

sensitivity depends linearly on the magnetic flux density

of the instrument. Increasing the field strength by a com-

pact design limits the free space available around the

sample head. The overall magnetic field homogeneity

achievable at the measurement volume depends on the

structure of the Halbach array (e.g., number and position

of the single magnets and length-to-radius ratio of the

magnet array

), as well as the perturbations of the single

magnets themselves.

Existing inhomogeneities of the

magnetic field can be overcome by adjusting the position

of movable permanent magnet blocks and the current of

auxiliary electric coils, known as shim coils, during a

shim procedure.

Another important factor is the filling

factor, defined as the volume of the liquid sample within

the receiver coil divided by the overall volume of the

receiver coil. A filling factor close to its maximum value

results in a high signal-to-noise ratio, which can be

further improved using optimised coil volumes

(e.g., micro coils

23,24

or an inner inductively coupled

coil

). A good filling factor also limits the available space

in the immediate vicinity of the sample cell, for example,

for temperature insulation.

2.2 |Temperature shielding strategies

Strategies to minimise the heat flux between sample

probe and magnet are generally distinguished as passive

(hereinafter, insulation) or active (hereinafter, tempera-

ture shielding). The first choice of insulation is a

Dewar, which is a typical silver-mirrored double-walled

vessel containing a vacuum and was discarded due to

the absorption of radio frequency of the coating. A

Dewar without mirror coating containing polytetrafluor-

oethylene (PTFE) tubing was tested but did not prove

sufficient insulation leading to an intolerable increase

of magnet temperature (results are shown in

Figure S2). Another approach is to minimise heat con-

duction by introducing a separation with low thermal

conductivity. Thermal resistance increases with insula-

tion layer thickness, usually resulting in thick layers.

As mentioned earlier, this approach is limited because

of the low filling factor (sensitivity). A larger borehole

in the magnet leads to an exponentially increasing

requirement in the magnet design to obtain the same

field strength.

In addition, classical insulation only

reduces the heat flux and slows down the temperature

compensation process. In the case of long-lasting heat

input, the magnet heats up as well (see Figure S2). The

shielding strategies examined in this study are shown

in Figure 2.

BORNEMANN-PFEIFFER ET AL.3

In contrast, a more suitable choice of temperature

shielding using a fluid (hereinafter, Fluid

iso

) to dissipate

the heat flux was developed. Controlled mass flow and

temperature of the Fluid

iso

enable a flexible drain of

energy to compensate for constant and even periodically

occurring heat of reaction. A usual choice of the Fluid

iso

is a liquid cooling agent, such as (tri)ethylene glycol,

ethanol, or water. However, in our case, the choice is

limited, as the Fluid

iso

flows within the NMR coil

(Figure 3); hence, any protonated liquid cooling

agent would be observed on the measured NMR spectra.

An alternative is the use of liquid perfluorinated

hydrocarbons (e.g., perfluoro decalin or perfluoro tributy-

lamine), as they do not interfere with the actual

H spec-

tra acquisition

but are discarded due to their persistent

environmental impact and greenhouse potential. There-

fore, pressurised air was selected as a cheap and available

alternative in this study. Its low heat capacity compared

with that of liquid agents was not a severe drawback, as

the overall thermal energy which needs to be dissipated

was quite small (e.g., Q

loss

=3.84 W) for the previously

shown experiment (Figure 1and Supporting Information

S1, Preliminary Studies).

A potential controller uses the flow and temperature

of the Fluid

iso

as manipulated variables to fulfil its regula-

tory objectives. The main objective is to maintain a con-

stant magnet temperature, which is already the objective

of the internal instrument control loop. As mentioned in

Section 1, this internal control loop is designed to cope

with slow deviations in the magnet temperature

(e.g., daily fluctuations in the laboratory room tempera-

ture) using actuators like Peltier elements for subsequent

cooling or heating. We have therefore based our control

concept on shielding all temperature fluctuations within

the sample cell as far as possible from the area of the

magnet. As the conducted experiments have shown, the

main danger is to disturb the sensitive magnetic control

circuit of the device, provoking protracted thermoregula-

tion activities accompanied by temporary changes in

magnetic field homogeneity. Access to the control param-

eters of the device and magnet temperature control is

currently not provided by any manufacturer for confiden-

tiality reasons.

Therefore, the main objective of the proposed

approach is to reduce fluctuation in the heat flux from

the probe to the magnet to minimise the influence of the

FIGURE 3 (a) Physical set-up within the NMR instrument. Two concentric pipes within the magnet borehole: outer glass cylinder as

the separation between the insulation flow stream and instrument. Inner PTFE tubing containing the liquid fluid to be measured.

Connection flanges on top and bottom containing the fluid connectors are custom made of polyoxymethylene. (b) NMR substitute box with

thermography camera and temperature control unit.

FIGURE 2 Simplified insulation/temperature shielding

concepts for a potential 4 mm fluid tubing carrying the reaction

mixture: (a) temperature shielding only and no additional passive

insulation; (b) temperature shielding and additional ‘inner’

insulation; and (c) temperature shielding and additional ‘outer’

insulation.

4BORNEMANN-PFEIFFER ET AL.

internal control loop on the reaction mixture. The heat

flux itself is, in general, not directly measured but is

accessible through the relevant temperature difference in

this case (between the magnet [T

mag

] and the adjacent

glass capillary surface holding the measurement coil

sur

]).

3|RESULTS AND DISCUSSION

3.1 |Design of the experimental set-up

In an NMR instrument, the surface temperature cannot

be accessed entirely. Therefore, an NMR substitute box

(see Figure 3b for the simplified scheme) was designed to

mimic the conditions in an NMR instrument and access

the temperature control set-up through optical thermog-

raphy. The optical thermography enables the measure-

ment of the surface temperature T

sur

at a sufficiently

high measurement rate (0.5 s

1

) and a resolution of

0.4 mm

per pixel. Further information on the thermog-

raphy can be found within Supporting Information S1,

Thermographic Measurements. In the simplest case,

variable-temperature shielding comprises two concentric

pipes. While the inner tube contains the sample fluid,

which is to be measured, the annulus contains the insula-

tion fluid (in our case, pressurised air), which is required

to remove heat energy from the system. Additional pas-

sive insulation, as shown in Figure 2b,c, was examined in

the preliminary studies (results are shown in Supporting

Information S1, Passive Insulation) and discarded as it

could not significantly reduce T

sur

. The additional passive

insulation connected to the outer cylinder (Figure 2c)

introduced an additional delay time for the control

system, impeding the control task. In contrast, additional

passive insulation connected to the fluid tubing helps

improve the overall shielding performance but was not

necessary for the examined reaction (Figure 2b).

The substitute box was later replaced with the actual

wide bore NMR prototype instrument (Figure 3a), but

the surrounding piping and instrumentation (Figure 4)

remained the same. Changes occurred in the surface tem-

perature measurement (replacement of the IR thermogra-

phy camera with thin-film thermocouples and that of the

substitute box with the NMR instrument). More details

and photographs of the set-up are shown in Figures S5–

S10.

3.2 |Temperature control model

The dependency of T

sur

on the substitute box temperature

box

and the ingoing and outgoing fluid streams can be

derived from the energy balances of the set-up. Although

the underlying physics of heat transfer within a system of

concentric pipes or tubes are well understood, deviations

between the model and physical reality are unavoidable.

Several methods available can help improve the model fit

(e.g., simplifying the models through parameter lumping,

parameter fixing or parameter omitting,

and using

empirical correction terms). As the model will be later

used for synthesising the controller and would need fur-

ther transformation and improvement, first principle

models are discarded here and replaced with an empiri-

cal, state-space model (see Equations 1and 2) providing

a better fit to the experimental data.

According to Ljung,

the model development or sys-

tem identification involves two steps: first, the choice of a

FIGURE 4 Scheme of the

temperature-regulated NMR

substitute box containing the box

temperature control unit (ventilation

and heat exchanger), probe piping

with mass flow controller and pump

and thermography device.

BORNEMANN-PFEIFFER ET AL.5

mathematical model of the smallest dimension, which is

described by a set of parameters, and second, identifica-

tion of these parameters through the experimental data

by a fitting routine. A widely spread model structure is

the so-called state-space form related to the system

input(s), u*, and output(s), y, through first-order differen-

tial equations (indicated by a dot above the variable)

using auxiliary state variable(s), x.

x¼Ax þBuwith u¼u



,ð1Þ

y¼Cx þDu:ð2Þ

Here, A,B,C, and Dare—depending on the number

of states, inputs, and outputs—variables, vectors, or

matrices. Due to their determined structure, on the one

hand, and the flexible number of states, x, on the other

hand, state-space representations are well suited to model

the dynamic behaviour of many systems with single or

multiple inputs and outputs. This approach is widely

used in control engineering, and many controller synthe-

sis procedures are based on state-space forms. For sim-

plicity, we adopt a linear time-invariant structure which

produces a sufficient model match in many cases. The

model parameters are identified by minimising of the

prediction error using the measured data and a numeri-

cal optimisation routine.

The system output T

sur

is physically dependent on

many other dynamic variables. Therefore, a model struc-

ture involving multiple inputs and a single output

variable is chosen. The input variables u* are further

divided (Table 1) into manipulated variables u, which are

the actuators of the controller, and the measured distur-

bances zconsidered within the model. A meaningful

choice of the considered measured disturbances is

achieved by performing a sensitivity analysis using the

minimum redundancy maximum relevance algorithm,

discarding the less relevant inputs.

To achieve a high similarity between the computa-

tional model and physical system, meaningful and, there-

fore, representative data are required. If applicable, as in

the given case, a design of experiment, including all

manipulable system inputs, will yield the best results.

Following standard textbooks, inputs should be per-

turbed simultaneously and uncorrelated.

28,31

Commonly

used inputs are pseudo-random binary or multisinusoi-

dal

sequences. In this study, a combination of multisi-

nusoidal inputs and the course of the manipulated

variables based on the controller decisions of an earlier

experimental run (Figure 5, yellow dashed boxes) yielded

the best results. During single excitation periods, the

uncorrelated response of the system output to the single

system inputs was examined, while the common excita-

tion period revealed potential cross-correlations. The

additional system input from the earlier experimental

run revealed system dynamics under actual operation

conditions.

The influence of single disturbances and manipulated

variables (especially cooled insulation flow and fluid tem-

perature) on the surface temperature T

sur

can be clearly

seen throughout the experimental identification run. The

state-space model was estimated through state-of-the-art

TABLE 1 Variable allocation of the finally identified model containing control variables, manipulated variables, and measured

disturbances.

Variable name Unit Short Allocation

sur

Cy(1) Control variable y

Viso L min

1

u(1) Manipulated variable u

Viso,cooled L min

1

u(2) Manipulated variable u

Mliquid g min

1

z(1) Measured disturbance z

liq

Cz(2) Measured disturbance z

box

mag

Cz(3) Measured disturbance z

I,in

Cz(4) Measured disturbance z

I,out

Cz(5) Measured disturbance z

C,in

Cz(6) Measured disturbance z

C,out

Cz(7) Measured disturbance z

Power1a.u.z(8) Measured disturbance z

Power2a.u.z(9) Measured disturbance z

6BORNEMANN-PFEIFFER ET AL.

numerical prediction error minimisation. Additional

details can be found in Supporting Information S1,

Controller.

There are many potential types of controllers for

multiple input single output systems (e.g., linear–

quadratic–Gaussian, H-infinity, and model predictive

control [MPC]).

Here, a linear MPC approach was

selected, as many disturbance variables are accessible

through direct measurement and hence can be consid-

ered in the future controller action calculation. An MPC

solves a quadratic programming optimisation problem

within each control interval, considering a cost function

and given constraints. The standard cost function con-

sists of several terms with individual weights, allowing

for a specific controller tuning. The following terms

were part of the cost function: control variable reference

tracking, manipulated variable tracking, manipulated

variable movement suppression, and constraint viola-

tion. The tuning of the controller is, despite available

performance measures

and tuning guidelines,

iterative process. The finally applied parameter can be

found in Table S2.

3.3 |Results

For the demonstration of the positive effect of tempera-

ture control, three experiments each were performed on

the NMR substitute box and the real NMR wide bore pro-

totype: (a) no insulation air stream at all, (b) constant

insulation air stream, and (c) fully MPC controlled insu-

lation air stream. Experiments with the NMR substitute

box are described only in Supporting Information S1 for

clarity (see Figures S11 and S12).

After successfully evaluating the variable-temperature

shielding within the NMR substitute box through optical

thermography, the set-up was retransferred into the com-

pact NMR wide bore prototype instrument for validation

purposes. During experimentation within the NMR sub-

stitute box, no significant geometric surface temperature

gradient was revealed; hence, the IR camera was replaced

with three equally distanced thin-film thermocouples.

The average temperature value now representing T

sur

.It

should be noted again that the available compact wide

bore NMR device with 60 MHz proton frequency is a pro-

totype device with an increased coil diameter of 12 mm.

This device was modified by Nanalysis for our specific

application. Consequently, due to the reduced filling fac-

tor, the achievable linewidths are by no means compara-

ble with products or solutions available on the market.

The aim of this project was solely to create a suitable

design basis that can be adapted to specific applications

and not to demonstrate the highest achievable line

widths.

The following experimental runs were conducted

with the NMR substitute box (see Supporting

FIGURE 5 System

identification experiment consisting

of (a) single multisinusoidal

excitations of model inputs with one

common excitation period, followed

by a former experimental run with a

controller. (b) Development of

relevant temperatures within the

system.

Heating wire voltage as an

auxiliary variable raising the liquid

temperature T

liq

BORNEMANN-PFEIFFER ET AL.7

Information S1, NMR Substitute Box Results) and within

the NMR wide bore prototype (shown below):

aðÞconstant insulation air stream,u¼const:

const:



bðÞMPC controlled insulation air stream,u¼fðw,y,z

fðw,y,z



cðÞno insulation air stream at all,u¼0



The results of variants (a) and (b) are shown in

Figure 6. Further information as well as the results of

variant (c) can be found in Figures S17 and S18. During

the uncontrolled run (a), the course of the magnet tem-

perature as well as the internal temperature control vari-

ables (z(8) and z(9)) of the instrument is strongly

fluctuating, leading to a maximum deviation of the mag-

net temperature, max (ΔT

mag

)=±0.01 K. Furthermore,

it can be seen that the internal control variable z(8) is

reaching its minimal restriction during experimentation

indicating the limit of temperature compensation of the

compact NMR instrument. The effect on the recorded

spectra is clearly visible through the change of the peak

shapes and line widths during the experiment (Figure 6a,

bottom). These changes even prevent characterisation of

the spectra quality through full width at half maximum

measurement over the full period of the experimentation

because the two signals of the NMR spectrum merge.

In comparison with that, the MPC controlled run

(Figure 6b) results into a by factor 7:1 reduced magnet

temperature fluctuation of max (ΔT

mag

)=±0.0014 K

while ensuring an almost steady course of z(8) and z(9).

As can be seen in the acquired spectra (Figure 6b,

bottom), the shape and signal line widths remain almost

constant, a prerequisite for, for example, evaluation of

overlapping substance peaks. The internal control vari-

ables not only indicate an existing potential for even

higher fluid temperatures but also remain almost con-

stant under changing temperature loads. Hence, a stable

FIGURE 6 Two experimental runs within the compact NMR wide bore prototype instrument distinguishable as variants: (a) ‘Constant

Iso Flow’, that is, steady isolation insulation flow, and (b) ‘MPC’, that is, MPC controlled (w=0). Opposed in each case are as follows: (top)

The course of the disturbance variables z(1) and z(2) as well as manipulated variables u(1) and u(2). Both manipulated variables are set up

with an upper limit of 40 L min

1

. (middle) Course of the NMR magnet temperature, z(3), and temperature control parameters of the

instrument, z(8) and z(9). (bottom) Plot of the stacked spectra during experimentation which were recorded with. Legends apply for both

horizontal panels.

8BORNEMANN-PFEIFFER ET AL.

magnetic field homogeneity for consistent spectra quality

could be ensured.

Qliq ¼_

mliq cP,liq Tliq,in Tliq,out



:ð3Þ

The thermal energy introduced by the sample was

determined according to the simplified energy balance

(Equation 3), assuming quasi-steady-state and isobaric

conditions. During the relevant MPC controlled run

(Figure 6), an average amount of 0.4 W thermal energy

was introduced over the course of the experiment

(Figure 7). Much more meaningful, however, is the con-

sideration of the maximum heat flow Q

max

=2.2 W and

the largest rate of change of approximately dT/

dt=0.05 K s

1

which is the critical parameter as shown

in the prior results. However, due to the different charac-

teristics of set-ups and compact NMR instrument, these

parameters can only serve as a guide value for other set-

ups.

4|CONCLUSION

In this study, a controlled variable-temperature shielding

set-up was realised, which applies to compact NMR

instruments. A heat load of up to 2.2 W, in this particular

case, was counterbalanced using thermostated air flows

without additional insulation. For confidentiality rea-

sons, the controller was set up independently of the NMR

internal magnet temperature control system. Although

this is not as desirable as a holistic approach, it allows for

easy transferability to other instruments. To gain access

to the surface temperature of the outer glass tube, by sep-

arating the borehole from the magnet, the set-up was

transferred to an NMR substitute box for preliminary

experiments allowing easier access to surface

temperatures.

Two major findings were obtained from this study.

First, additional passive insulation was not beneficial, as

it decelerated the internal system dynamics and thus

impeded the overall controllability of the test case stud-

ied. Second, the combination of two different manipu-

lated variables (insulation air flows of different

temperature) showed an excellent controller performance

regarding rapid changes in thermal load. Additionally,

the set-up was reinserted into the NMR instrument as a

proof of concept. The experimental findings suggested an

efficient temperature shielding. The minimisation of

interference between the inner instrument and external

temperature shielding control loop was crucial for the

successful implementation within the instrument.

It should be noted that the space requirement for the

presented variable-temperature shielding competes with

the demand for low fill factors and thus small spectral

linewidths. Commercially available instruments are not

compatible with the presented method without prior

hardware modifications, as they come standard with

5 mm coils aimed at the highest spectral resolution with-

out adequate control of the thermal stress that occurs in

exothermic continuous chemical reactions. The ability to

modify NMR instruments for larger coil diameters is gen-

erally available on all instruments. However, a holistic

control approach that incorporates the temperature con-

trol loops internal to the instrument and a substantial

improvement in spectrum quality will require the cooper-

ation of the manufacturers in all cases. We hope that this

paper will provide a solid basis for future design

guidelines.

ACKNOWLEDGEMENTS

First and foremost, we would like to thank Nanalysis

Corp., Calgary, Alberta, Canada, for providing the wide

bore NMR prototype and valuable assistance in using this

instrument. In particular, the support of Juan Araneda

and Neal Gallagher for discussion and support as well as

Martin Berberov and Greg McFeetors for their help

regarding out-of-spec magnet shimming (all Nanalysis

Corp., Calgary, Alberta, Canada) is gratefully acknowl-

edged. Furthermore, we thank Simon Kern (S-Pact

GmbH) for the discussion and initial ideas. We thank

Christoph Naese (BAM) for his fast and reliable support

in designing and building customised construction parts.

We thank our colleagues from the BAM Division 8.7

“Thermographic Methods”for their support. Finally, we

acknowledge the support from Björn Hagedorn and Para-

stoo Semnani, whose experimental work in preliminary

studies produced the first foundation to build this work.

FIGURE 7 Model predictive control (MPC) controlled

experimental run (same as shown in Figure 6b) within the compact

NMR wide bore prototype instrument: calculated heat introduced

into the system by the liquid probe and corresponding

temperatures. The negative heat flow is caused by the lower mass

flow of the liquid probe compared with the insulation air flow.

BORNEMANN-PFEIFFER ET AL.9

CONFLICT OF INTEREST STATEMENT

The authors would like to mention that the wide bore

prototype was acquired on a paid loan basis from

Nanalysis Corp. of Calgary, Alberta, Canada, for most of

the experimental work presented here.

PEER REVIEW

The peer review history for this article is available at

https://www.webofscience.com/api/gateway/wos/peer-

review/10.1002/mrc.5379.

ORCID

Martin Bornemann-Pfeiffer https://orcid.org/0000-

0003-1758-4594

REFERENCES

[1] M. Grootveld, B. Percival, M. Gibson, Y. Osman, M. Edgar, M.

Molinari, M. L. Mather, F. Casanova, P. B. Wilson, Anal.

Chim. Acta 2019,1067, 11.

[2] T. A. van Beek, Phytochem. Anal. 2021,32, 24.

[3] T. Castaing-Cordier, D. Bouillaud, J. Farjon, P. Giraudeau, in

Annual Reports on NMR Spectroscopy, (Ed: G. A. Webb) Vol.

103, Academic Press, Burlington House, Piccadilly, London,

United Kingdom 2021, 191.

[4] F. Dalitz, M. Cudaj, M. Maiwald, G. Guthausen, Prog. Nucl.

Magn. Reson. Spectrosc. 2012,60, 52.

[5] A. Friebel, E. von Harbou, K. Münnemann, H. Hasse, Ind.

Eng. Chem. Res. 2019,58, 18125.

[6] K. Meyer, S. Kern, N. Zientek, G. Guthausen, M. Maiwald,

TrAC, Trends Anal. Chem. 2016,83, 39.

[7] J. Giberson, J. Scicluna, N. Legge, J. Longstaffe, in Annual

Reports on NMR Spectroscopy, (Ed: G. A. Webb) Vol. 102,

Academic Press, Burlington House, Piccadilly, London,

United Kingdom 2021, 153.

[8] S. Kern, L. Wander, K. Meyer, S. Guhl, A. R. G. Mukkula, M.

Holtkamp, M. Salge, C. Fleischer, N. Weber, R. King, S.

Engell, A. Paul, M. P. Remelhe, M. Maiwald, Anal. Bioanal.

Chem. 2019,411, 3037.

[9] W. G. Lee, M. T. Zell, T. Ouchi, M. J. Milton, Magn. Reson.

Chem. 2020,58, 1193.

[10] K. Halbach, Nuclear Instruments and Methods 1980,169,1.

[11] E. Danieli, J. Mauler, J. Perlo, B. Blümich, F. Casanova,

J. Magn. Reson. 2009,198, 80.

[12] N. Zientek, K. Meyer, S. Kern, M. Maiwald, Chem. Ing. Tech.

2016,88, 698.

[13] N. Nestle, Z. J. Lim, T. Böhringer, S. Abtmeyer, S. Arenz, F. C.

Leinweber, T. Weiß, E. von Harbou, Magn. Reson. Chem. 2020,

58, 1213.

[14] K. Meyer, J. Ruiken, M. Illner, A. Paul, D. Müller, E. Esche, G.

Wozny, M. Maiwald, Meas. Sci. Technol. 2017,28, 035501.

[15] M. Bornemann, S. Kern, N. Jurtz, T. Thiede, M. Kraume, M.

Maiwald, Ind. Eng. Chem. Res. 2019,58, 19562.

[16] A. M. R. Hall, T. A. A. Cartlidge, G. Pileio, J. Magn. Reson.

2020,317, 106778.

[17] A. Brächer, S. Hoch, K. Albert, H. J. Kost, B. Werner, E. von

Harbou, H. Hasse, J. Magn. Reson. 2014,242, 155.

[18] B. Wouters, P. Miggiels, R. Bezemer, E. A. W. van der

Cruijsen, E. van Leeuwen, J. Gauvin, K. Houben, K. B. S. S.

Gupta, P. Zuijdwijk, A. Harms, A. C. de Souza, T.

Hankemeier, Anal. Chem. 2022,94, 15350.

[19] M. V. Silva Elipe, R. R. Milburn, Magn. Reson. Chem. 2016,

54, 437.

[20] Evonik Industries AG, (n.d), The Chemical All-Rounder the

Fumed Silica AEROSIL,https://history.evonik.com/en/

inventions/aerosil, accessed December 30, 2022.

[21] S. Tewari, T. O'Reilly, A. Webb, J. Magn. Reson. 2021,324,

106923.

[22] K. Halbach, Nuclear Instruments and Methods in Physics

Research 1982,198, 213.

[23] R. M. Fratila, A. H. Velders, Annual Review of Analytical

Chemistry 2011,4, 227.

[24] S. S. Zalesskiy, E. Danieli, B. Blümich, V. P. Ananikov, Chem.

Rev. 2014,114, 5641.

[25] J. A. Tang, A. Jerschow, Magn. Reson. Chem. 2010,48, 763.

[26] K. Halbach, Nuclear Instruments and Methods in Physics

Research 1980,169,1.

[27] K. A. McLean, K. B. McAuley, The Canadian Journal of Chem-

ical Engineering 2012,90, 351.

[28] L. Ljung, System Identification - Theory for the user 2nd edition,

Prentice Hall PTR, Upper Saddle River, NJ 1999.

[29] T. McKelvey, Identification of State-Space Models From Time

and Frequency Data (PhD thesis, Linköping University,

Sweden) 1995.

[30] C. Ding, H. Peng, Journal of Bioinformatics and Computational

Biology 2005,3, 185.

[31] R. Isermann, M. Münchhof, Identification of Dynamic Systems:

An Introduction With Applications, Vol. 85, Springer Berlin,

Heidelberg 2011.

[32] K. E. Häggblom, IFAC-PapersOnLine 2016,49, 308.

[33] K. R. Muske, J. B. Rawlings, AIChE J. 1993,39, 262.

[34] D. H. Olesen, Technical University of Denmark (M. Sc. thesis,

IMM-M. Sc.-2012–69, Kongens Lyngby, Denmark 2012.

[35] M. Alhajeri, M. Soroush, Ind. Eng. Chem. Res. 2020,59, 4177.

SUPPORTING INFORMATION

Additional supporting information can be found online

in the Supporting Information section at the end of this

article.

How to cite this article: M. Bornemann-Pfeiffer,

K. Meyer, J. Lademann, M. Kraume, M. Maiwald,

Magn Reson Chem 2023,1.https://doi.org/10.1002/

mrc.5379

10 BORNEMANN-PFEIFFER ET AL.