ELECTROMAGNETIC TRAPS FOR CHARGED

AND NEUTRAL PARTICLES

Nobel Lecture, December 8, 1989
by
WO L F G A N G P A U L

Physikalisches Institut der Universitt Bonn, Nussallee 12, D-5300 Bonn,

F.R.G.

Experimental physics is the art of observing the structure of matter and of

detecting the dynamic processes within it. But in order to understand the
extremely complicated behaviour of natural processes as an interplay of a
few constituents governed by as few as possible fundamental forces and
laws, one has to measure the properties of the relevant constituents and
their interaction as precisely as possible. And as all processes in nature are
interwoven one must separate and study them individually. It is the skill of
the experimentalist to carry out clear experiments in order to get answers to
his questions undisturbed by undesired effects and it is his ingenuity to
improve the art of measuring to ever higher precision. There are many
examples in physics showing that higher precision revealed new phenomena, inspired new ideas or confirmed or dethroned well established theories. On the other hand new experimental techniques conceived to answer
special questions in one field of physics became very fruitful in other fields
too, be it in chemistry, biology or engineering. In awarding the Nobel prize
to my colleagues Norman Ramsey, Hans Dehmelt and me for new experimental methods the Swedish Academy indicates her appreciation for the
aphorism the Gttingen physicist Georg Christoph Lichtenberg wrote two
hundred years ago in his notebook one has to do something new in order
to see something new. On the same page Lichtenberg said: I think it is a
sad situation in all our chemistry that we are unable to suspend the constituents of matter free.
Today the subject of my lecture will be the suspension of such constituents of matter or in other words, about traps for free charged and neutral
particles without material walls. Such traps permit the observation of isolated particles, even of a single one, over a long period of time and therefore
according to Heisenbergs uncertainty principle enable us to measure their
properties with extremely high accuracy.
In particular, the possibility to observe individual trapped particles opens
up a new dimension in atomic measurements. Until few years ago all
measurements were performed on an ensemble of particles. Therefore, the
measured value - for example, the transition probability between two
eigenstates of an atom - is a value averaged over many particles. Tacitly

602

Physics 1989

one assumes that all atoms show exactly the same statistical behaviour if one
attributes the result to the single atom. On a trapped single atom, however,
one can observe its interaction with a radiation field and its own statistical
behaviour alone.
The idea of building traps grew out of molecular beam physics, mass
spectrometry and particle accelerator physics I was involved in during the
first decade of my career as a physicist more than 30 years ago. In these
years (1950 - 55) we had learned that plane electric and magnetic multipole
fields are able to focus particles in two dimensions acting on the magnetic or
electric dipole moment of the particles. Lenses for atomic and molecular
beams [1,2,3] were conceived and realized improving considerably the
molecular beam method for spectroscopy or for state selection. The lenses
found application as well to the ammonia as to the hydrogen maser [4].
The question What happens if one injects charged particles, ions or
electrons, in such multipole fields led to the development of the linear
quadrupole mass spectrometer. It employs not only the focusing and defocusing forces of a high frequency electric quadrupole field acting on ions
but also exploits the stability properties of their equations of motion in
analogy to the principle of strong focusing for accelerators which had just
been conceived.
If one extends the rules of two-dimensional focusing to three dimensions
one posseses all ingredients for particle traps.
As already mentioned the physics or the particle dynamics in such focusing devices is very closely related to that of accelerators or storage rings for
nuclear or particle physics. In fact, multipole fields were used in molecular
beam physics first. But the two fields have complementary goals: the storage
of particles, even of a single one, of extremely low energy down to the microelectron volt region on the one side, and of as many as possible of extremely high energy on the other. Today we will deal with the low energy part. At
first I will talk about the physics of dynamic stabilization of ions in two- and
three-dimensional radio frequency quadrupole fields, the quadrupole mass
spectrometer and the ion trap. In a second part I shall report on trapping of
neutral particles with emphasis on an experiment with magnetically stored
neutrons.
As in most cases in physics, especially in experimental physics, the
achievements are not the achievements of a single person, even if he
contributed in posing the problems and the basic ideas in solving them. All
the experiments I am awarded for were done together with research students or young colleagues in mutual inspiration. In particular, I have to
mention H. Friedburg and H. G. Bennewitz, C.H. Schlier and P. Toschek in
the field of molecular beam physics, and in conceiving and realizing the
linear quadrupole spectrometer and the r.f. ion trap H. Steinwedel, O.
Osberghaus and especially the late Erhard Fischer. Later H.P. Reinhard, U.
v. Zahn and F. v. Busch played an important role in developing this field.

W.Paul

603

Focusing and Trapping of particles

What are the principles of focusing and trapping particles? Particles are
elastically bound to an axis or a coordinate in space if a binding force acts
on them which increases linearly with their distance r
In other words if they move in a parabolic potential

The tools appropriate to generate such fields of force to bind charged

particles or neutrals with a dipole moment are electric or magnetic multipole fields. In such configurations the field strength, or the potential
respectively increases according to a power law and shows the desired
symmetry. Generally if m is the number of poles or the order of symmetry
the potential is given by

For a quadrupole m = 4 it gives @ ~ r2 cos 2, and for a sextupole m = 6 one

gets CD ~ r cos 3~ corresponding to a field strength increasing with r and r2
respectively.
3

Trapping of charged particles in 2- and J-dimensional quadrupole fields

In the electric quadrupole field the potential is quadratic in the Cartesian

coordinates.

The Laplace condition A@ = 0 imposes the condition + + = 0

There are two simple ways to satisfy this condition.
a) a = 1 = , =0 results in the two-dimensional field
(2)
b) a = = 1 , = -2 generates the three-dimensional configuration, in
cylindrical coordinates

The two-dimensional quadrapole or the mass filter [5,6]

Configuration a) is generated by 4 hyperbolically shaped electrodes linearly

extended in the y-direction as is shown in Fig. 1. The potential on the
electrodes is 0/2 if one applies the voltage a0 between the electrode
pairs. The field strength is given by

604

Physics 1989

Figure I. a) Equipotential lines for a plane quadrupole fild, b) the electrodes Structure for the
mass filter.

If one injects ions in the y-direction it is obvious that for a constant voltage
<Do the ions will perform harmonic oscillations in the the x-y-plane but due
to the opposite sign in the field E their amplitude in the z-direction will
increase exponentially. The particles are defocused and will be lost by
hitting the electrodes.
This behaviour can be avoided if the applied voltage is periodic. Due to
the periodic change of the sign of the electric force one gets focusing and
defocusing in both the x- and z-directions alternating in time. If the applied
voltage is given by a dc voltage U plus an r.f. voltage V with the driving
frequency
z

the equations of motion are

At first sight one expects that the time-dependent term of the force
cancels out in the time average. But this would be true only in a homogenous field. In a periodic inhomogenous field, like the quadrupole field there
is a small average force left, which is always in the direction of the lower
field, in our case toward the center. Therefore, certain conditions exist that
enable the ions to traverse the quadrupole field without hitting the electrodes, i.e. their motion around the y-axis is stable with limited amplitudes
in x- and z-directions. We learned these rules from the theory of the Mathieu
equations, as this type of differential equation is called.
In dimensionless parameters these equations are written

(5)

W.Paul

605

By comparison with equation (4) one gets

63)
The Mathieu equation has two types of solution.
1. stable motion: the particles oscillate in the x-z-plane with limited amplitudes.
They pass the quadrupole field in y-direction without hitting the electrodes.

Figure 2. The overall stability diagram for the two-dimensional quadrupole field.

0.3
0.23

0.2
t

a
0.1

Figure 3. The lowest region for simultaneous stability in x-and z-direction. All ion masses lie on the
operation line, m1 > m1.

606

Physics 1989

2. unstable motion: the amplitudes grow exponentialy in x, z or in both

directions. The particles will be lost.
Whether stability exists depends only on the parameters a and q and not
on the initial parameters of the ion motion, e.g. their velocity. Therefore, in
an a-q-map there are regions of stability and instability (Fig.2). Only the
overlapping region for x and z stability is of interest for our problem. The
most relevant region 0 < a, q < 1 is plotted in Fig. 3. The motion is stable in
x and z only within the triangle.
For fixed values r0 , U and V all ions with the same m/e have the same
operating point in the stability diagram. Since a/q is equal to 2U/V and
does not depend on m, all masses lie along the operating line a/q = const.
On the q axis (a = 0, no d.c. voltage) one has stability from 0 < q < q =
0.92 with the consequence that all masses between > m > m have
stable orbits. In this case the quadrupole field works as a high pass mass
filter. The mass range m becomes narrower with increasing dc voltage U
i.e. with a steeper operating line and approaches m = 0, if the line goes
through the tip of the stability region. The bandwidth in this case is given
only by the fluctuation of the field parameters. If one changes U and V
simultaneously and proportionally in such a way that a/q remains sonstant,
one brings the ions of the various masses successively in the stability region
scanning through the mass spectrum in this way. Thus the quadrupole
works as a mass spectrometer.
A schematic view of such a mass spectrometer is given in Fig. 4. In Figs.
5a,b. the first mass spectra obtained in 1954 are shown [6]. Clearly one sees
the influence of the d.c. voltage U on the resolving power.
In quite a number of theses the performance and application of such
max

min

W.Paul

607

Figure 5. a) Very first mass spectrum of Rubidium. Mass scanning was achieved by periodic
variation of the driving frequency v. Parameter: u = $, at u = 0.164 85 Rb and 87 Rb are fully
resolved. b) Mass doublet 83Kr - C 6H 11. . Resolving power m/m = 6500 [9].

608

Physics 1989

instruments was investigated at Bonn University [7,8,9]. We studied the

influence of geometrical and electrical imperfections giving rise to higher
multipole terms in the field. A very long instrument (l = 6 m) for high
precision mass measurements was built achieving an accuracy of 2 * 10-7 in
determining mass ratios at a resolving power 2 = 16 000. Very small ones
were used in rockets to measure atomic abundances in the high atmosphere.
In another experiment we succeeded in separating isotopes in amounts of
milligrams using a resonance method to shake single masses out of an
intense ion beam guided in the quadrupole.
In recent decades the r.f. quadrupole whether as mass spectrometer or
beam guide due to its versatility and technical simplicity has found broad
applications in many fields of science and technology. It became a kind of
standard instrument and its properties were. treated extensively in the
literature [10].
The Ion Trap

Already at the very beginning of our thinking about dynamic stabilization of

ions we were aware of the possibility using it for trapping ions in a threedimensional field. We called such a device Ionenkfig[11,12,13]. Nowadays the word ion trap is preferred.
The potential configuration in the ion trap has been given in eq. (3).
This configuration is generated by an hyperbolically shaped ring and two
hyperbolic rotationally symmetric caps as it is shown schematically in Fig.
6a. Fig. 6b gives the view of the first realized trap in 1954.

Figure 6. a) Schematic view of the ion trap. b) Cross section of the first trap (1955).

W. Paul

609

If one brings ions into the trap, which is easily achieved by ionizing inside
a low pressure gas by electrons passing through the volume, they perform
the same forced motions as in the two-dimensional case. The only difference
is that the field in z-direction is stronger by a factor 2. Again a periodic field
is needed for the stabilization of the orbits. If the voltage Q0 = U + Vcos mt
is applied between the caps and the ring electrode the equations of motion
are represented by the same Mathieu functions of eq.(5). The relevant
parameters for the r motion correspond to those in the x-direction in the
plane field case. Only the z parameters are changed by a factor 2.
Accordingly, the region of stability in the a-q-map for the trap has a
different shape as is shown in Fig. 7. Again the mass range of the storable
ions (i.e. ions in the stable region) can be chosen by the slope of the
operation line a/q = 2U/V. Starting with operating parameters in the tip of
the stable region one can trap ions of a single mass number. By lowering the
d.c. voltage one brings the ions near the q-axis where their motions are
much more stable.
For many applications it is necessary to know the frequency spectrum of
the oscillating ions. From mathematics we learn that the motion of the ions
can be described as a slow (secular) oscillation with the fundamental fre-

Figure 7. The lowest region for stability in the ion trap. On the lines inside the stability region /$
and B, resp. are constant.

610

quencies o,,, = PI,L . /2 modulated with a micromotion, a much faster

oscillation of the driving frequency if one neglects higher harmonics. The
frequency determining factor is a function only of the Mathieu parameters a and q and therefore mass dependent. Its value varies between 0 and
1; lines of equal are drawn in Fig. 7.
Due to the stronger field the frequency O, of the secular motion becomes
twice 0,. The ratio O/O, is a criterion for the stability. Ratios of 10: 1 are
easily achieved and therefore the displacement by the micromotion averages
out over a period of the secular motion.
The dynamic stabilization in the trap can easily be demonstrated in a
mechanical analogue device. In the trap the equipotential lines form a
saddle surface as is shown in Fig. 8. We have machined such a surface on a
round disc. If one puts a small steel ball on it, then it will roll down: its
position is unstable. But if one let the disk rotate with the right frequency

Figure 8. Mechanical analogue model for the ion trap with steelball as particle

w. Paul

611

appropriate to the potential parameters and the mass of the ball (in our case
a few turns/s) the ball becomes stable, makes small oscillations and can be
kept in position over a long time. Even if one adds a second or a third ball
they stay near the center of the disc. The only condition is that the related
Mathieu parameter q be in the permitted range. I brought the device with
me. It is made out of Plexiglas which allows demonstration of the particle
motions with the overhead projector.
This behaviour gives us a hint of the physics of the dynamic stabilization.
The ions oscillating in the r- and z-directions to first approximation harmonically, behave as if they are moving in a pseudo potential well quadratic
in the coordinates. From their frequencies O, and O, we can calculate the
depth of this well for both directions. It is related to the amplitude V of the
driving voltage and to the parameters a and q. Without any d.c. voltage the
depth is given by Dz = (q/8) V, in the r-direction it is half of this. As in
practice V amounts to a few hundred volts the potential depth is of the
order of 10 Volts. The width of the well is given by the geometric dimensions. The resulting configuration of the pseudo potential [14] is therefore
given by

Cooling process

As mentioned, the depth of the relevant pseudopotential in the trap is of

the order of a few volts. Accordingly the permitted kinetic energy of the
stored ions is of the same magnitude and the amplitude of the oscillations
can reach the geometrical dimensions of the trap. But for many applications
one needs particles of much lower energy well concentrated in the center of
the trap. Especially for precise spectroscopic measurements it is desirable
to have extremely low velocities to get rid of the Doppler effect and an
eventual Stark effect, caused by the electric field. It becomes necessary to
cool the ions. Relatively rough methods of cooling are the use of a cold
buffer gas or the damping of the oscillations by an external electric circuit.
The most effective method is the laser induced sideband fluorescence
developed by Wineland and Dehmelt [15].
In 1959 Wuerker et al. [16] performed an experiment trapping small
charged Aluminium particles ( ~ mm) in the quadrupole trap. The
necessary driving frequency was around 50 Hz accordingly. They studied all
the eigenfrequencies and took photographs of the particle orbits; see Figs.
9a, b. After they have damped the motion with a buffer gas they observed
that the randomly moving particles arranged themselves in a regular pattern. They formed a crystal.
In recent years one has succeeded in observing optically single trapped
ions by laser resonance fluorescence [17]. Walther et al., using a high
resolution image intensifier observed the pseudo-crystallization of ions in
the trap after cooling the ions with laser light. The ions are moving to such

Figure 9. a) Photomicrograph of a Lissajous orbit in the r-z-plane of a single charged particle of

Aluminium powder. The micro motion is visible. b) Pattern of condensed Al particles [16].

positions where the repulsive Coulomb force is compensated by the focusing forces in the trap and the energy of the ensemble has a minimum. Figs.
10a, b show such a pattern with 7 ions. Their distance is of the order of a
few micrometers. These observations opened a new field of research [18].
The Ion Trap as Muss Spectrometer

As mentioned the ions perform oscillations in the trap with frequencies O,

and o, which at fixed field parameters are determined by the mass of the
ion. This enables a mass selective detection of the stored ions. If one
connects the cap electrodes with an active r.f. circuit with the eigenfrequency , in the case of resonance = o,, the amplitude of the oscillations
increases linearly with time. The ions hit the cap or leave the field through a
bore hole and can easily be detected by an electron multiplier device. By
modulating the ion frequency determining voltage V in a sawtooth mode
one brings the ions of the various masses one after the other into resonance,
scanning the mass spectrum. Fig. 11 shows the first spectrum of this kind
achieved by Rettinghaus [19].
The same effect with a faster increase of the amplitude is achieved if one
inserts a small band of instability into the stability diagram . It can be
generated by superimposing on the driving voltage V cos t a small additional
rf voltage, e.g. with frequency /2, or by adding a higher multipole term
to the potential configuration [5b,20].
In summary the ion trap works as ion source and mass spectrometer at
the same time. It became the most sensitive mass analyzer available as only a
few ions are necessary for detection. Its theory and performance is reviewed
in detail by R.E. March [21].

613

W. Paul

Figure 10. a) Pseudo crystal of 7 magnesium ions. Particle distance 23 m. b) The same trapped
particles at higher temperature. The crystal has melted [18].

The Penning Trap

If one applies to the quadrupole trap only a d.c. voltage in such a polarity
that the ions perform stable oscillations in the z-direction with the frequency CO: = , Ix
, ; the ions are unstable in the x-y-plane, since the field is
directed outwards. Applying a magnetic field in the axial direction, the
z-motion remains unchanged but the ions perform a cyclotron motion
in the x-y-plane. It is generated by the Lorentz force F directed towards
the center. This force is partially compensated by the radial electric force
F, = & r. As long as the magnetic force is much larger than the electric one,
stability exists in the r-y-plane as well. No r.f. field is needed. The resulting
rotation frequency calculates to
L

Figure Il. First mass spectrum achieved with the ion trap. Gas: air at 2 . 10 -9 torr [19]

It is slightly smaller than the undisturbed cyclotron frequency eB/m. The

difference is due to the magnetron frequency

which is independent of the particle mass.

The Penning trap [22], as this device is called, is of advantage if magnetic
properties of particles have to be measured, as for example Zeeman transitions in spectroscopic experiments, or cyclotron frequencies for a very
precise comparison of masses as are performed e.g. by G. Werth. The most
spectacular application the trap has found in the experiments of G. Grff
[23] and H. Dehmelt for measuring the anomalous magnetic moment of the
electron. It was brought by Dehmelt [24] to an admirable precision by
observing only a single electron stored for many months.
Traps for neutral particles

In the last examination I had to pass as a young man I was asked if it would
be possible to confine neutrons in a bottle in order to prove if they are
radioactive. This question, at that time only to be answered with no,
pursued me for many years until I could have had replied: Yes, by means of
a magnetic bottle. It took 30 years until by the development of superconducting magnets its realization became feasible.
Using the example of such a bottle I would like to demonstrate the
principle of confining neutral particles. Again the basis is our early work on
focusing neutral atoms and molecules having a dipole moment by means of
multipole fields making use of their Zeeman or Stark effect to first and
second order [1,2,3]. Both effects can be used for trapping. Until now only
magnetic traps were realized for atoms and neutrons. Particularly, B. Martin, U. Trinks, and K. J. Kgler contributed to their development with great
enthusiasm.

W. Paul

615

The principle of magnetic bottles

The potential energy U of a particle with a permanent magnetic moment
in a magnetic field is given by U = -B. If the field is inhomogenous it
corresponds to a force F = grad(B). In the case of the neutron with its spin
h/2 only two spin directions relative to the field are permitted. Therefore,
its magnetic moment can be oriented only parallel or antiparallel to B. In
the parallel position the particles are drawn into the field and in the
opposite orientation they are repelled. This permits their confinement to a
volume with magnetic walls.
The appropriate field configuration to bind the particles harmonically is
in this case a magnetic sextupole field. As I have pointed out such a field B
increases with r2, B = 3 * P and the gradient $ with r respectively.
In such a field neutrons with orientation p tt B satisfy the confining
condition as their potential energy U = + B ~ r2 and the restoring force
gradB = --Cr is always oriented towards the center. They oscillate in the
field with the frequency 2 = q. P articles with ,u t 1 B are defocused and
leave the field. This is valid only as long as the spin orientation is conserved.
Of course, in the sextupole the direction of the magnetic field changes with
the azimuth but as long as the particle motion is not too fast the spin follows
the field direction adiabatically conserving the magnetic quantum state.
This behaviour permits the use of a magnetic field constant in time in
contrast to the charged particle in an ion trap.
An ideal linear sextupole in the x-z-plane is generated by six hyperbolically
shaped magnetic poles of alternating polarity extended in y-direction, as
shown in Figs. 12a, b. It might be approximated by six straight current leads

Figure 12. a) Ideal sextupole field. Dashed: magnetic field lines, dotted: lines of equal magnetic
potential, B = const. b) Linear sextupole made of 6 straight current leads with alternating
current direction.

616

Figure 17. Sextupole sphere

with alternating current directions arranged in a hexagon. Such a configuration works as a lense for particles moving along the y-axis.
There are two possibilities to achieve a closed storage volume: a sextupole
sphere and a sextupole torus. We have realized and studied both.
The spherically symmetric field is generated by three ring currents in an
arrangement shown in Fig. 13. The field B increases in all directions with r 2
and has its maximum value B0 at the radius r0 of the sphere. Using superconducting current leads we achieved B0 = 3T in a sphere with a radius of 5 cm.
But due to the low magnetic moment of the neutron = 6 * 10-8 eV/T the
potential depth B o is only 1.8 * 10-7 eV and hence the highest velocity of
storable neutrons is only v max = 6m/s . Due to their stronger moment for Na
atoms these values are 2.2 * 10-4 eV and 37 m/s, respectively.
The main problem with such a closed configuration is the filling process,
especially the cooling inside. However, in 1975 in a test experiment we
succeeded in observing a storage time of 3 s for sodium atoms evaporated
inside the bottle with its Helium cooled walls [25]. But the breakthrough in
confining atoms was achieved by W. D. Phillip and H.J. Metcalf using the
modern technique of Laser cooling [26].
The problem of storing neutrons becomes easier if one uses a linear
sextupole field bent to a closed torus with a radius R as is shown in Fig. 14.
The magnetic field in the torus volume is unchanged B = $Q * ? and has
no component in azimuthal direction. The neutrons move in a circular orbit
with radius Rs if the centrifugal force is compensated by the magnetic force

w. Paul

617

Figure 14. Sextupole torus. Rs orbit of circulating neutrons.

In such a ring the permitted neutron energy is limited by

It is increased by a factor ($ + 1) compared to the case of the sextupole

sphere. As the neutrons have not only an azimuthal velocity but also
components in r and z directions they are oscillating around the circular
orbit.
But this toroidal configuration has not only the advantage of accepting
higher neutron velocities, it also permits an easy injection of the neutrons in
the ring from the inside. The neutrons are not only moving in the magnetic
potential well but they also experience the centrifugal barrier. Accordingly,
one can lower the magnetic wall on the inside by omitting the two inward
current leads. The resulting superposition of the magnetic and the centrifugal potential still provides a potential well with its minimum at the beam
orbit. But there is no barrier for the inflected neutrons.
It is obvious, that the toroidal trap in principle works analogous to the
storage rings for high energy charged particles. In many respects the same
problems of instabilities of the particle orbits by resonance phenomena
exist, causing the loss of the particles. But also new problems arise like, e.g.
undesired spin flips or the influence of the gravitational force. In accelerator physics one has learned to overcome such problems by shaping the
magnetic field by employing the proper multipole components. This technique is also appropriate in case of the neutron storage ring. The use of the
magnetic force gradB instead of the Lorentz force being proportional to
B just requires multipole terms of one order higher. Quadrupoles for

Physics 1989

focusing have to be replaced by sextupoles and e.g. octupoles for stabilization of the orbits by decapoles.
In the seventies we have designed and constructed such a magnetic
storage ring with a diameter of the orbits of 1 m. The achieved usable field
of 3.5 T permits the confinement of neutrons in the velocity range of 5 - 20
m/s corresponding to a kinetic energy up to 2 * 10-6 eV. The neutrons are
injected tangentially into the ring by a neutron guide with totally reflecting
walls. The inflector can be moved mechanically into the storage volume and
shortly afterwards be withdrawn.
The experimental set up is shown in Fig. 15. A detailed description of the

W. Paul

619

storage ring, its theory and performance is given in [27]. In 1978 in a first
experiment we have tested the instrument at the Grenoble high flux reactor. We could observe neutrons stored up to 20 min after injection by
moving a neutron counter through the confined beam after a preset time.
As by the detection process the neutrons are lost, one has to refill the ring
starting a new measurement. But due to the relatively low flux of neutrons
in the acceptable velocity range, their number was too low to make relevant
measurements with it.
In a recent experiment [28] at a new neutron beam with a flux improved
by a factor 40 we could observe neutrons up to 90 min, i.e. roughly 6 times
the decay time of the neutron due to radioactive decay. Fig. 16 shows the
measured profile of the neutron beam circulating inside the magnetic gap.
Measuring carefully the number of stored neutrons as a function of time we
could determine the lifetime to = 877 10 s (Fig. 17).
The analysis of our measurements lets us conclude that the intrinsic
storage time of the ring for neutrons is at least about one day. It shows that
we had understood the relevant problems in its design.

Figure 16. Beam profile of the stored neutrons inside the magnet gap 400 sec after injection.

620

Physics 1989

Figure 17. Logarithmic decrease of the number of stored neutrons with time.

The storage ring as a balance

This very reproducible performance permitted another interesting experiment. As I explained the neutrons are elastically bound to the symmetry
plane of the magnetic field. Due to their low magnetic moment the restoring force is of the order of the gravitational force. Hence it follows that the
weight of the neutron stretches the magnetic spring the particle is hanging
on; the equilibrium center of the oscillating neutrons is shifted downwards.
The shift z0. is given by the balance mg = gradB. One needs a gradient %
= 173 Gauss/cm for compensating the weight. As the gradient in the ring
increases with z and is proportional to the magnetic current Zone calculates
the shift z0 to
It amounts in our case to z0 = 1.2 mm at the highest magnet current Z = 200
A and 4.8 mm at 50 A accordingly.
By moving a thin neutron counter through the storage volume we could
measure the profile of the circulating neutron beam and its position in the
magnet. Driving alternating the counter downwards and upwards in many
measuring runs we determined z0 as a function of the magnet current.
The result is shown in Fig. 18. The measured data taken with different
experimental parameters are following the predicted line. A detailed analysis gives for the gravitational mass of the neutron the value

It agrees within 4 % with the well known inertial mass.

Thus the magnetic storage ring represents a balance with a sensitivity of
10 -25 g. It is only achieved because the much higher electric forces play no
role at all.

w. Paul

621

I am convinced that the magnetic bottles developed in our laboratory as

described will be useful and fruitful instruments for many other experiments in the future as the Ion Trap has already proved.

Figure 18. Downward shift of the equilibrium center of the neutron orbits due to the weight of
the neutron as function of the magnetic current.

622

Physics 1989

REFERENCES
[1]
[2]
[3]
[4]
[5]

H. Friedburg and W. Paul, Naturwissenschaft 38, 159 (1951).

H. G. Bennewitz and W. Paul, Z. f. Physik 139,489 (1954).
H. G. Bennewitz and W. Paul, Z. f. Physik, 141, 6 (1955).
C. H. Townes, Proc. Nat. Acad. of Science, 80, 7679 (1983).
a) W. Paul and H. Steinwedel, Z. f. Naturforschung 8a, 448 (1953); b) German
Patent Nr. 944 900; USA Patent 2939958.
[6] W. Paul and M. Raether, Z. f. Physik, 140, 262 (1955).
[7] W. Paul, H. P. Reinhardt, and U. v. Zahn, Z. f. Physik 152, 143 (1958).
[8] F. v. Busch and W. Paul, Z. f. Physik, 164, 581 (1961).
[9] U. v. Zahn, Z. f. Physik, 168, 129 (1962).
[10] P. H. Dawson: Quadrupole Muss Spectrometry and its Application, Elsevier, Amsterdam 1976.
[11] W. Paul, 0. Osberghaus, and E. Fischer, Forsch.Berichte des Wirtschaftsministeriums Nordrhein-Westfalen Nr. 4 15 (1958).
[12] K. Berkling, Thesis Bonn 1956.
[13] E. Fischer, Zeitschrift f. Physik 156, 1 (1959).
[14] H. Dehmelt, Adv. in Atom and Molec. Phys., Vol. 3 (1967).
[15] D. J. Wineland and H. Dehmelt, Bull. Am. Phys. Soc., 20, 637 (1975).
[16] R. F. Wuerker and R.V. Langmuir, Appl. Phys. 30, 342 (1959).
[17] W. Neuhauser, M. Hohenstett, P. Toschek and A. Dehmelt, Phys. Rev. A22,
1137 (1980).
[18] F. Dietrich, E. Chen, J. W. Quint and H. Walter, Phys. Rev. Lett. 59, 2931
(1987).
[19] G. Rettinghaus, Zeitschrift Angew. Physik, 22, 321 (1967).
[20] F. v. Busch and W. Paul, Z. f. Physik, 165, 580 (1961).
[21] R. E. March and R. J. Hughes, Quadrupole Storage Mass Spectrometry, -John
Wiley, New York 1989.
[22] F. M. Penning, Physica 3, 873 (1936).
[23] G. Grff, E. Klempt and G. Werth, Zeitschrift f. Physik 222, 201 (1969).
[24] R. S. van Dyck, P. B. Schwinberg, H. G. Dehmelt, Phys. Lett. 38, 310 (1977).
[25] B. Martin, Thesis Bonn University 1975.
[26] A. L. Migdal, J. Prodan, W. D. Phillips, Th. H. Bergmann, and H.J. Metcalf,
Phys. Rev. I.ett. 54, 2596 (1985).
[27] K. J. Kgler, W. Paul, and U. Trinks, Nucl. Instrument. Methods A 228, 240
(1985).
[28] W. Paul, F. Anton, L. Paul, S. Paul, and W. Mampe, Z. f. Physik C 45, 25
(1989).