Proposal for a Permanent Magnet System with

a Constant Gradient Mechanically Adjustable

in Direction and Strength
€
PETER BLUMLER
Institute of Physics, University of Mainz, 55099 Mainz, Germany

ABSTRACT: A design for a permanent magnet system is proposed that generates spa-
tially homogeneous, constant magnetic field gradients, thus creating conditions suitable
for MRI without gradient coils and amplifiers. This is achieved by superimposing a
weak Halbach quadrupole on a strong Halbach dipole. Rotation of either the quadru-
pole or the entire magnet assembly can be used to generate two-dimensional images
via filtered backprojection. Additionally, the mutual rotation of two quadrupoles can be
used to scale the resulting gradient. If both gradients have identical strength the gradi-
ent can even be made to vanish. The concept is demonstrated by analytical considera-
tions and FEM simulations. However, a demonstration on a working prototype is still
pending. C 2016 Wiley Periodicals, Inc.
V Concepts Magn Reson Part B (Magn Reson Engineer-
ing) 46B: 41–48, 2016

KEY WORDS: MRI; imaging; Halbach; quadrupole; backprojection

INTRODUCTION exact knowledge of the field, mechanical rotation of the

In recent years several permanent magnet systems were
developed for the emerging field of compact or even
mobile NMR (1,2). Many of them were also equipped
with classical gradient coils driven by electrical currents
to allow for MRI applications. A particularly suitable
magnet design for this purpose is the Halbach cylinder
which nowadays can be constructed and shimmed to
produce images of very high quality (3,4) and even
high-resolution NMR-spectra (5). However, shimming
permanent magnets is still a demanding iterative pro-
cess. In 2014 Zimmerman Cooley et al. presented a
method to produce MR-images utilizing the inhomoge-
neous magnetic field of a non-shimmed Halbach (6).
This was done by collecting spatial information from
Received 9 October 2015; revised 21 January 2016;
accepted 22 January 2016
Correspondence to: Peter Bl€
umler; E-mail: bluemler@uni-mainz.de
Concepts in Magnetic Resonance Part B, Vol. 46B(1) 41–48 (2016)
Published online in Wiley Online Library (wileyonlinelibrary.
com). DOI: 10.1002/cmr.b.21320
Ó 2016 Wiley Periodicals, Inc.
41
entire magnet, using an eight channel receiver array and
image reconstruction via special software. On a first
glance this appears like a step back to the very first MRI
methodology where images were reconstructed from
one-dimensional projections along different angles
using filtered Radon-transforms (backprojection). The
direction of the projection was changed by either rotat-
ing the gradient or the sample (7). On the other hand,
getting rid of the very power consuming gradient ampli-
fiers helps to make MRI simpler, silent, more economi-
cal and portable. However, there is a slight problem
with the proposed method in Ref. 6 because the image
reconstruction is limited in regions of rather homogene-
ous fields (e.g. in the center). Signals originating from
this or other ambiguous field regions could nevertheless
be spatially encoded by using multiple receive coils and
channels.
The aim of this contribution is to follow-up on this
inspiring idea and suggests a design that removes the
problems of nonbijective projections. Instead of using
an inhomogeneous field with ambiguities it is proposed
to use constant gradients of high spatial homogeneity
superimposed on a Halbach dipole and hence, allow
42 €
BLUMLER
for scaling their direction and strength, eventually even
to make them disappear, by rotating additional rings of
permanent magnets. Like in Ref. 6 only permanent
magnets are involved, however, deliberately generat-
ing better defined inhomogeneities that are scalable in
amplitude and direction. It is hoped that simple
Fourier-transformation and Radon-transformation (fil-
tered backprojection) will be sufficient to generate
images by rotating either gradient or object between
scans. Of course this can also be combined with multi-
ple receivers to speed up acquisition via parallel MRI
and non-linear image reconstruction schemes to relax
the demands of construction accuracy.
So far the gradients generated by permanent mag-
nets were mainly used to excite thin slices for scanning
profiles either close to surfaces or across thin films (8).
For most of these implementations the stray-field of
unilateral setups were used (9,10), or the design was
tailored for a special application, e.g. the GARfield-
magnet for the study of thin films (11). Taken together
they filled a niche in NMR methodology, on the other
hand they have been a working horse in focusing
beams of charged particles in accelerators and mass-
spectrometers. There the most simple implementation
is a quadrupole which is also proposed in this work in
the form of a Halbach cylinder—superimposed on an
Halbach dipole.
The proposed concept only generates gradients per-
pendicular to the axis of the Halbach cylinder which
allow for two-dimensional (2D), spatial encoding and
obviously for a real MRI system three-dimensional
(3D) resolution will be required. There are various
ideas how this can be achieved. The excitation of a thin
slice in the third (axial) direction could be realized for
instance by truncating the cylinder. A thin cylinder will
have a pronounced field maximum in its center. Alter-
natively, two such cylinders can be mounted coaxially
in a suitable distance to produce a local field minimum.
The latter arrangement would have the advantage that
the field shape can be mechanically controlled by the
width of the gap. Either way the width of such a local
field extremum will be much flatter for wider cylinders
(12) and the resulting slice might become too thick.
Simply reducing the excitation bandwidth might not be
an option due to the general field inhomogeneity over
the region of interest. Nonetheless, the field profile can
be sharpened by adding smaller correction magnets
(13). A more extraordinary idea could be the extension
of the presented concept of dipole and quadrupole Hal-
bach cylinders to Halbach spheres (14). Like cylinders
they can be opened without strong forces to access their
very strong and homogeneous inner field (4) and then
be rotated in 3D around a static sample. Finally and
more ordinarily static B0- or radio-frequency (rf) B1-
Concepts in Magnetic Resonance Part B (Magnetic Resonance Engineering) DOI 10.1002/cmr.b
gradients (15) can be introduced along this direction by
additional coils to spatially encode the third dimension.
This article only introduces the concept by calcu-
lations and FEM-simulations. No such device has
yet been built by the author.
CONCEPT
Klaus Halbach’s original article (16) already introduces
the fundamental idea, because he was particularly inter-
ested in producing magnetic fields with many poles for
e.g. accelerators. The basic construction principle for
this is an annular cylinder of permanent magnet material
where the direction, a, of the magnetization is a multi-
ple of its location angle h [see Fig. 1(a)].
a 5 ðm11Þ h with m 2 Z [1]
The polarity (field polarity of two is corresponding
to a dipole, four to a quadrupole, six to a sextupole,
etc.) of the resulting field is then given by 2m and a
positive sign directs the field inwards (with no stray
field outside, for an infinitely long cylinder) while a
negative sign creates a magnetic field of the same
polarity outside the cylinder (4). For NMR purposes
only the case m 5 11 was used so far, because it pro-
duces a homogeneous dipolar field inside the cylinder
[see Fig. 1(b)]. Halbach ((16); Eq. [21]) also already
provided an analytical expression of the field magni-
*
tude at position r depending on the polarity
 
ri
B 5 BR ln for m 5 1
ro
!m21 "   #
  *
r m ri m21 [2]
*
B r 5 BR 12
ri m21 ro
for m  2
where BR is the remanence of the permanent mag-
netic material and ri and ro are the inner and the
outer radius of the cylinder.
The straightforward approach to obtain spatial reso-
lution inside a spatially homogeneous field is to super-
impose a much weaker gradient field which transforms
the NMR-spectrum into a spatial projection along the
gradient direction. It* is *important to realize that Max-
well’s second law (r  B 5 0) prevents the generation
of a gradient field in one exclusive direction only, and
consequently only that component (dBy/dr) of the gra-
dient field is of relevance which aligns with the direc-
tion of the much stronger homogeneous field (B0 here
along y-direction) needed for polarizing the nuclear
spins. Depending on the strength of the other (so-called
concomitant) gradient components relative to B0 image
distortions will appear. Yablonskiy et al. (17) show
 PROPOSAL FOR PERMANENT MAGNET GRADIENTS 43

Figure 1 Schematic illustration of the Halbach magnet geometries: a) construction principle of Halbach-cylinders of polar-
ity 2m. Magnetic material placed with its center on a circle (dashed line) at an angle h has its magnetization direction (bold
arrow) at an angle a 5 (m 1 1) h. The coordinate system is used for all other graphs. b) Inner dipole (m 5 11) field created
by an annular cylinder of permanent magnet material (dark gray) with continuously changing magnetization direction as indi-
cated by the white arrows. The gray shade in the center illustrates the amplitude of the field, while its direction is specified
by the black arrowed flux lines. c) Inner Halbach quadrupole (m 5 12) in the same representation as b). d) and e) show the
Bx and By component of the flux in c). Here the black arrows indicate strength and direction of the flux (no flux lines). f) A
nested arrangement of the magnets in b) and c) producing a superposition of the homogeneous dipole field with a weaker
gradient field. Here the quadrupole (darker gray) is inside the dipole but, of course, it could also encase the dipole.

that this deviation from linearity can be described by cated by “Q”]. For such a superposition Eq. [2] results
bending lines or planes in the image with a curvature in the following vector field inside the geometry
given by a radius Rc 5 B0/G. If this radius is much big-
ger than the field of view of the image (respectively rf- " *
#
coil or magnet size) they can be neglected. In typical * *
Bx ðr Þ * *
B ðr Þ 5 5B 0 1 G  r
clinical MRI (B0  1 T, G 5 10–20 mT/m) Rc  50 m By ðr Þ
*

and hence too big to cause a noticeable image distor- " # " #
tion. However, in MRI with low or even modestly 0 21 0 *
5 B0 1G r;
strong B0-fields, the concomitant gradients may have 1 0 1
to be considered. " # [3]
 D
A linear field dependence or constant gradient *
x ri
with r 5 ; B0 5 BD
R ln and
field can either be obtained from a current driven y roD
coil or a permanent magnet. If this gradient should !
be normal to the axis of the annular cylinder the Q 1 1
G 5 2BR Q 2 Q
most straight-forward implementation is a field of ri ro
quadrupolar polarity. Such a quadrupole field [e.g.
Fig. 1(c)] consist of two orthogonal constant gradi-
ent fields [see Figs. 1(e, f)] while its magnitude has of which only By(x, y) 5 B0 1 G y is experimentally
a conical shape [see gray shades in Fig. 1(c)]. relevant (for B0  G, obviously the concomitant
Figure 1(f) now shows a dipolar Halbach magnet gradient has always the same strength). Then MRI
with its field B0 along the y-direction [cf. Fig. 1(b) in the experiments can be performed by rotating either
following indicated by superscript “D”] encasing a magnet or sample using the backprojection idea pre-
smaller quadrupolar Halbach magnet [see Fig. 1(c) indi- suming that both cylinders are rotated synchronously.

Concepts in Magnetic Resonance Part B (Magnetic Resonance Engineering) DOI 10.1002/cmr.b

44 €
BLUMLER

Figure 2 By-field component of a Halbach-quadrupole rotated by an angle b in the field of a (Halbach) dipole. For a
clearer view the dipole is omitted. Otherwise it would dominate the graph since it should be much stronger than the quad-
rupole. However, it is necessary because its field direction (here y) selects the field component of the quadrupole, generat-
ing a constant gradient [dBy/dx, dBy/dy] which rotates then with 2b. a) Schematic drawing: The original positive y
direction of the quadrupole is indicated by a black triangle. The arrow inside the cylinder marks the direction of the gradi-
ent whose field change is additionally displayed by a gray scale. b) rotation by b 5 458, c) b 5 908.

If however the quadrupole is turned by an angle cylinder might be considered as a minor advantage over
b relative to the dipole, the rotation matrix, R, has rotating the entire magnet, this concept can be extended
*
to be applied to the gradient matrix, G, and r to change the total strength of the gradient (this is equiv-
alent to changing the total field in amplitude and direc-
* 0 * * * tion by rotating two or more nested dipoles (18,21)).
B ðr Þ 5 B 0 1 RG  ðR21 r Þ
  This is done by placing yet another quadrupole around
cos b sin b the structure of dipole and quadrupole. If this second
with R 5
2sin b cos b [4] quadrupole is now rotated with respect to dipole and first
   
0 2xcos 2b 1 ysin 2b quadrupole [see Fig. 3(a)] the gradient field will rotate
5 B0 1G and change its amplitude. If the first quadrupole is indi-
1 xsin 2b 1 ycos 2b
cated by “Q1” and the second by “Q2” their combined
field can be expressed by Eq. [6] as
of which again only the y-component is experimen-
tally relevant    
* * * sin 2b sin 2/
GR 5 GQ1 1 GQ2 5 GQ1 1 GQ2
B0y ðx; yÞ 5 B0 1 G ðxsin 2b 1 ycos 2bÞ: [5] cos 2b cos 2/
!
1 1
The gradient vector of this rotated field, B0y is with GQi 52BQi
R 2 and i 51; 2
then riQi roQi

23 [7]
dB0y " #
6 dx 7 sin 2b
*
0 6 7 where Q1 is rotated by an angle b and Q2 by / rel-
r By 5 6 0 7 5 G : [6]
4 dBy 5 cos 2b ative to B0. If for simplification only Q1 is turned
dy (/ 5 0) this becomes
   
This means that a rotation of the quadrupole by an * sin 2b 0
GR 5 GQ1 1 GQ2
angle b relative to the dipole causes the resulting field cos 2b 1
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
gradient to rotate by 2b with respect to the direction *
jGR j 5 GQ1 1 2GQ1 GQ2 cos 2b 1 G2Q2
2 [8]
of B0y. This is a direct consequence of the C2-symme-  
try of the quadrupolar field and allows to perform a *
21 GQ1 sin 2b
]GR 5 tan
complete rotation of the gradient direction by turning GQ1 cos 2b 1 GQ2
the quadrupole by 1808 only (as illustrated in Fig. 2).
While the ability to rotate the direction of the gradient *
Hence, the amplitude of the total gradient field,
field by turning e.g. a thinner, lighter, outer quadrupole- jG R j, has a maximal value of jGQ1 1 GQ2 j for b 5 08,

Concepts in Magnetic Resonance Part B (Magnetic Resonance Engineering) DOI 10.1002/cmr.b

 PROPOSAL FOR PERMANENT MAGNET GRADIENTS 45

Figure 3 a) Schematic drawing of a Halbach dipole (light gray, black arrows) concentrically nested in two quadrupoles
(darker grays, white arrows). If only one of the quadrupoles (here the outer is turned while the inner is static, / 5 0) is
turned by an angle b the gradient strength changes as shown in b). b) FEM-simulation with BQ1 Q2
R 5 BR 5 1 T,
riQ1 5 15 cm, roQ1 5 16 cm, riQ2 5 16.5 cm, roQ2 5 17.686 cm. The latter value was chosen so that the simulated values
GQ1 5 GQ2 5 0.801 T/m although Eq. [9] gave roQ2 5 17.72 cm. This small discrepancy stems from the discretization of the
ideal Halbachs for the simulation (see below). Both quadrupoles
*
were simulated in 2D as cylinders with 32 segments. The
dashed line
*
is a plot of Eq. [8]. Note that the direction of G R is rotated by b (and not 2b) relative to y, because Eq. [8]
gives ]G R 5 tan 21 ½sin 2b=ðcos 2b 1 1Þ 5 b for GQ1 5 GQ2. In other words, the effective gradient is formed by the vec-
tor sum of two vectors of same length and hence its angle is halved.

1808 and a minimum of jGQ1 2 GQ2 j at b 5 908, 2708.
In principle the gradient field can be made to disappear
by choosing GQ1 5 GQ2 , e.g. by adjusting the outer
radius of the second quadrupole to
BQ2 Q1 Q1 Q2
R ri ro ri
roQ2 5 Q1
: [9]
BR riQ1 riQ2 2roQ1 riQ2 1 BQ2 Q1 Q1
R ri ro
Equations [8] and [9] were tested and confirmed in
a 2D-FEM simulation (COMSOL Multiphysics 5.1)
as shown in Fig. 3, where it is demonstrated that
quadrupoles of modest size can already produce very
strong gradients (GR > 1.5 T/m). They are certainly
too strong for MRI (for B0  0.5 T, Rc  33 cm) but
maybe interesting for other experiments (e.g. NMR
diffusion measurements). For MRI the gradient
strength can be easily adjusted by reducing size or
remanence of the magnetic material using Eq. [8].
It should be noted that the combination of a Halbach
dipole with a weaker quadrupole can also be realized
by constructing only one cylinder from pieces whose
magnetization is the vector sum of that of dipole and
quadrupole. The gradient of this elaborate design could
still be varied by a second quadrupole, however
strength or distance and direction of the magnetic sub-
units will have to be adjusted individually.
PRACTICAL CONSIDERATIONS
So far only ideal Halbach systems were considered.
However, in reality their continuous magnetization
distribution and infinite length has to be discretized
and truncated (truncating the length of a Halbach-
cylinder has significant influence on the field strength
and homogeneity along the cylinder axis. These
effects are discussed in detail elsewhere (12)) in order
to be constructible from permanent magnetic parts.
This perturbation of ideality causes inhomogeneities
as it can be seen in Figs. 4(a, b). When a continuous
Halbach magnet [cf. Fig. 4(a)] is discretized by cut-
ting it into a number of segments with homogeneous
magnetization [cf. Fig. 4(b)] the strongest deviation
from the ideal field are observed at the mutual interfa-
ces. Nevertheless, the production of such cylindrical
segments with individual magnetization direction at
precise angles is extremely difficult. Therefore, it was
suggested (12,20) to simplify the construction by
using magnets with polygonal or circular cross-
sections and identical magnetization and mount them
such that the magnetization direction fulfills Eq. [1]
[cf. Fig. 4(c)]. Due to this construction principle pro-
duction errors of the magnet segments in magnetiza-
tion direction and strength can be reduced
significantly using different approaches (3,4,12,21)
resulting in Halbach assemblies with inhomogeneities
in the order of ppm.
While all these investigations were done for Hal-
bach dipoles to generate magnetic fields of high
homogeneity, Fig. 4 illustrates that the same concept
allows to produce discrete Halbach quadrupoles with
very homogeneous field gradients. However, there are
small differences in the experimental demands for

Concepts in Magnetic Resonance Part B (Magnetic Resonance Engineering) DOI 10.1002/cmr.b

46 €
BLUMLER

Figure 4 a) Ideal Halbach of magnetic material (BR 5 1.0 T, ri 510 cm, ro 5 12.37 cm) with a continuous distributed
magnetization direction, producing a homogeneous gradient of 3.83 T/cm as given by Eq. [3]. b)–f) 2D FEM-simulation
of discrete Halbach quadrupoles. The grayscale images show the produced gradient dBy/dy of the overlaid magnetic mate-
rial (gray with white arrows indicating the magnetization direction). The range of the gradient strength is displayed on the
right of each image and adjusted to the central gradient value 6 10%. b) Discretized version of a) consisting of 32 seg-
ments producing Gc 5 3.69 T/cm in the center. c–f) Discrete variations using different numbers of magnets of squared
cross-section with 1.58 cm side length. c) 32 magnets with Gc 5 1.76 T/cm (corresponding to f 5 47.7% of magnet mate-
rial relative to the amount used in a and b). d) 16 magnets with Gc 5 0.88 T/cm and f 5 23.9%. e) Eight magnets with
Gc 5 0.44 T/cm and f 5 11.9% and f) four magnets with Gc 5 0.22 T/cm and f 5 5.97%.

such gradients in MRI. Firstly, they do not need to be
as strong as possible. Therefore, the magnet concen-
tration can be reduced, which reduces the gradient
amplitude proportional to the area/volume covered by
magnetic material [cf. Figs. 4(b–f)]. Secondly the
homogeneity requirements for the gradient field are
less demanding in image reconstruction as nicely
demonstrated in Ref. 6 and from Figs. 4(c–f) it can
be estimated that the area of a certain gradient homo-
geneity scales logarithmically with the number of
used magnets. In other words, one can probably work
with inhomogeneous gradient fields as long as they
are bijective.
Hence, a practical design suggestion for an MRI
magnet could be a strong, stationary dipolar Hal-
bach magnet encased by two quadrupole rings
which can both be rotated relative to the dipole [cf.
Fig. 3(a)]. A rotation of both quadrupoles relative to

Concepts in Magnetic Resonance Part B (Magnetic Resonance Engineering) DOI 10.1002/cmr.b

 PROPOSAL FOR PERMANENT MAGNET GRADIENTS 47

Figure 5 2D FEM-simulation of a possible magnet design. a) Design and flux: The inner Halbach dipole consists of 16
octagonal magnets with a side length of 1.96 cm and a remanence of BR 5 1.5 T producing a field of 0.45 T over an
accessible inner volume of 20 cm in diameter. The quadrupole is made of eight cylindrical segments (BR 5 0.3 T,
ri 5 16 cm, ro 5 17 cm) of which four (Q1) are stationary while the others can be turned by an angle b. Each subset of
four magnets produces a gradient of GQ1 5 GQ2 5 0.026 T/cm. The gray scale image *
shows the magnitude of the generated
flux in the range of the gradient field, i.e. 0.44 to 0.448 T. b) Effective gradient jG R j produced by turning Q2 by an angle
b. The inner error bars represent the standard deviation, the outer the minimal/maximal values in an analyzing circle of
10 cm in diameter (black circle in center of a)). The dashed line is a plot of Eq. [8].

each other can be used to adjust the gradient strength CONCLUSION

according to Eq. [8]. If the angle between both quad-
rupoles is then locked and they are both rotated by
908, full k-space can be acquired via projections. If
image reconstruction methods are available which tol-
erate modest gradient inhomogeneities the density of
magnets in both quadrupoles can be greatly reduced
as shown in Fig. 4. Finally these two ideas can be
merged in a magnet design similar to that depicted in
Fig. 5(a), where four magnets (Q1) are meant to be
stationary while four others (Q2) can be rotated on
the same perimeter around a dipole. In this way the
strength of the gradient can be adjusted [see Fig.
5(b)] in a very compact design. In this arbitrarily cho-
sen geometry the gradient can be varied from 44.5 to
53.0 mT/m (for 1H this corresponds to a resolution of
1,890 to 2,250 Hz/mm; or with a sampling bandwidth
of 500 kHz the field of view will be 22 to 26 cm
which are reasonable values) and other values are, of
course, possible by changing shape and/or strength of
Q1 and Q2. While one quadrupole (Q2) is turned to
adjust the gradient strength, the other magnets of Q1
can be moved away from their initial position to
homogenize (shim) the fields (3). Although the gradi-
ent strength can only be changed and not completely
cancelled in this device such a design might be suffi-
cient for most MRI applications where the gradient
strength merely needs to be adjusted to sample size
and spatial resolution.
The concept of producing homogeneous magnetic field
gradients by superimposing Halbach quadrupoles to
Halbach dipoles was introduced together with possibil-
ities to rotate the direction and scale the amplitude of
the resulting gradient by turning one or two quadru-
poles. Although no real prototype has been constructed
yet, the parameters of suggested designs were calcu-
lated and successfully verified by 2D-FEM simulations.
For the construction of a first prototype care has to
be taken, because the gradients produced by Halbach
quadrupoles are very strong. Their concomitant gra-
dients have to be considered even at the relatively
strong B0 fields available from Halbach-dipoles. On
the other hand this has the advantage that small mag-
nets added to the outside of a Halbach-dipole superim-
pose gradients which are already sufficient for MRI.
Having the “gradient”-magnets on the outside of the
“field”-magnet might also allow to remove and add
them depending on the application or to move them to
rotate the gradient. Finally rotating only a subset of
the “gradient”-magnets can be used to scale the gradi-
ent (as shown in a simulated design study in Fig. 5).
The latter idea as well as all discretization and trunca-
tion of ideal magnet structures, which are necessary
for constructing a real device, will of course reduce
the homogeneity of magnetic fields and field gradients
[see Fig. 5(b)]. Furthermore, intrinsic deviations in the

Concepts in Magnetic Resonance Part B (Magnetic Resonance Engineering) DOI 10.1002/cmr.b

48 €
BLUMLER
magnetic properties of the permanent magnet materi-
als will further worsen the homogeneity (typically by
one or two orders of magnitude (4)). However, these
deteriorations will be comparable to the device used
in Ref. 6 from which already impressive MR-images
could be obtained. Adding the proposed homogene-
ous, constant gradients can then make the compact
MRI method proposed in Ref. 6 more simple, robust
and applicable and probably weaken the demands for
multiple receive coils and channels, because nonbijec-
tive projections of object space to encoding space are
reduced or completely avoided.
It is hoped that these ideas will help to construct
simple, robust, and effective MRI systems suitable
for medical use in emerging countries.
ACKNOWLEDGMENTS
The author wants to thank his wife, Friederike
Schmid (Physics Department, University of Mainz),
for carefully checking and commenting the manu-
script. Furthermore, Lukas Winter (Berlin Ultrahigh
Field Facility—B.U.F.F., Max Delbr€uck Center for
Molecular Medicine, Berlin) is acknowledged for
stimulating discussions and his enormous engage-
ment to design, build, and disseminate an open
source MRI system suitable for third world coun-
tries. The author is also grateful to Miss Antonia
Barghoorn for first simulations of the proposed
magnet geometries, and Stefan Hiebel has to be
acknowledged for checking manuscript and equa-
tions. Last but not least two anonymous referees
have to be thanked for their constructive criticism
on the first version of this conceptual work.
REFERENCES
1. Bl€umich B, Haber-Pohlmeier, S, Zia W. 2014. Com-
pact NMR. Berlin: Walter de Gruyter.
2. Johns M, Fridjonsson EO, Vogt S, Haber A., eds.
Mobile NMR and MRI: Developments and Applica-
tions. 2015, Royal Society of Chemistry: Cambridge.
3. Danieli E, Mauler J, Perlo J, Bl€umich B, Casanova F.
2009. “Mobile sensor for high resolution NMR spec-
troscopy and imaging”. J Magn Reson 198:80–87.
4. Bl€umler P, Casanova F. 2015. Hardware develop-
ments: Halbach magnet arrays. In: Johns M, et al.,
eds. Mobile NMR and MRI: Developments and
Applications. Cambridge: Royal Chemical Society.
5. Danieli E, Perlo J, Bl€umich B, Casanova F. 2010. “Small
magnets for portable NMR spectrometers”. Angew Chem
Int Ed 49:4133–4135.
6. Zimmerman Cooley C, Stockmann JP, Armstrong
BD, Sarracanie M, Lev MH, Rosen MS, Wald LL.
2015. “Two-dimensional imaging in a lightweight
Concepts in Magnetic Resonance Part B (Magnetic Resonance Engineering) DOI 10.1002/cmr.b
portable MRI scanner without gradient coils”. Magn
Reson Med 73:872–883.
7. Lauterbur PC. 1973. “Image formation by induced
local interactions - examples employing nuclear mag-
netic resonance”. Nature 242:190–191.
8. Mitchell J, Bl€umler P, McDonald PJ. 2006. “Spatially
resolved nuclear magnetic resonance studies of planar
samples”. Prog Nucl Magn Reson Spectrosc 48:161–
181.
9. Bl€umler P, Casanova F. 2015. Hardware develop-
ments: single sided magnets. In: Johns M, et al., eds.
Mobile NMR and MRI: Developments and Applica-
tions. Cambridge: Royal Chemical Society.
10. Bl€umich B, Perlo J, Casanova F. 2008. “Mobile
single-sided NMR”. Prog Nucl Magn Reson Spec-
trosc 52:197–269.
11. Glover PM, Aptaker PS, Bowler JR, Ciampi E,
McDonald PJ. 1999. “A novel high-gradient perma-
nent magnet for the profiling of planar films and
coatings”. J Magn Reson 139:90–97.
12. Soltner H, Bl€umler P. 2010. “Dipolar Halbach mag-
net stacks made from identically shaped permanent
magnets for magnetic resonance”. Concept Magn
Reson A 36:211–222.
13. Barghoorn A. 2015. Halbach Magnet Array Design
for Low Cost Magnetic Resonance Imaging. Bachelor
Thesis. TU Berlin: Institut f€ur Festk€orperphysik.
14. Leupold HA, Tilak AS, Potenziani II E. 2000.
“Permanent magnet spheres: design, construction, and
application”. J Appl Phys 87:4730–4734.
15. Maffei P, Mutzenhardt P, Retournard A, Diter B,
Raulet R, Brondeau J, Canet D. 1994. “NMR micros-
copy by radiofrequency field gradients”. J Magn
Reson 107:40–49.
16. Halbach K. 1980. “Design of permanent multipole
magnets with oriented rare-earth cobalt material”.
Nucl Instrum Methods 169:1–10.
17. Yablonskiy DA, Sukstanskii AL, Ackermann JJH.
2005. “Image artifacts in very low magnetic field
MRI: the role of concomitant gradients”. J Magn
Reson 174:279–286.
18. Leupold HA, Tilak AS, Potenziani II E. 1993.
“Adjustable multi-tesla permanent magnet field
sources”. IEEE Trans Magn 29:2902–2904.
19. Bauer C, Raich H, Jeschke G, Bl€umler P. 2009.
“Design of a permanent magnet with a mechanical
sweep suitable for variable-temperature continuous-
wave and pulsed EPR spectroscopy”. J Magn Reson
198:222–227.
20. Raich HP, Bl€umler P. 2004. “Design and construction
of a dipolar Halbach array with a homogeneous
field from identical bar magnets: NMR Mandhalas”.
Concept Magn Reson B Magn Reson Eng 23:16–25.
21. Windt CW, Soltner H, van Dusschoten D, Bl€umler P.
2011. “A portable Halbach magnet that can be
opened and closed without force: The NMR-CUFF”.
J Magn Reson 208:27–33.