See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/231144094

Tracing Planck's constant to the kilogram by electromechanical methods

Article  in  Metrologia · November 2003

DOI: 10.1088/0026-1394/40/6/007

CITATIONS READS

97 362

3 authors:

Ali L. Eichenberger Beat Jeckelmann

Federal Institute of Metrology (METAS) Federal Institute of Metrology (METAS)
50 PUBLICATIONS   640 CITATIONS    104 PUBLICATIONS   1,471 CITATIONS   

SEE PROFILE SEE PROFILE

Philippe Richard
Federal Institute of Metrology (METAS)
39 PUBLICATIONS   563 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Gravity: metrology, data processing, tides View project

METAS Watt balance View project

All content following this page was uploaded by Beat Jeckelmann on 12 September 2015.

The user has requested enhancement of the downloaded file.

INSTITUTE OF PHYSICS PUBLISHING METROLOGIA
Metrologia 40 (2003) 356–365 PII: S0026-1394(03)71275-4

Tracing Planck’s constant to the kilogram

by electromechanical methods
A Eichenberger, B Jeckelmann and P Richard
Swiss Federal Office of Metrology and Accreditation, Lindenweg 50, 3003 Bern-Wabern,
Switzerland

Received 10 June 2003

Published 27 November 2003
Online at stacks.iop.org/Met/40/356
Abstract
Among the priority tasks in the further development of the International
System of Units is the redefinition of the kilogram based on fundamental
constants. One of the strategies pursued today is to relate mass to Planck’s
constant h using the equivalence between mechanical and electrical
energies. In this paper, possible experimental approaches in this direction
are described. The approach which promises to reach the required
uncertainty at the earliest is the concept of the moving-coil watt balance.
The status of the different watt balance experiments is reviewed in detail.

1. Introduction system for the increasing needs of science and technology, a

replacement of the kilogram based on fundamental constants is
The kilogram is the last remaining material artefact in the needed. There is general agreement among metrologists that a
International System of Units (SI). The kilogram is defined new definition should be considered if the experiments relating
as the mass of the international prototype of the kilogram mass and fundamental constants reach a relative uncertainty of
kept at the BIPM. One of the major disadvantages of this less than one part in 108 .
definition is the fact that the amount of material constituting Two strategies are being pursued to reach this goal [4, 5].
the international prototype is subject to possible changes in The first one is to relate the mass of a fundamental particle,
time. Long-term studies of the differences between selected
such as the mass of the proton, to the kilogram by counting a
1 kg prototypes and the international prototype [1, 2] indicate
large number of identical atoms [6,7]. The second strategy is to
that the long-term variation of the kilogram could be as much
relate the Planck constant to the kilogram using the equivalence
as five parts in 109 per year. This conclusion is based on the
between mechanical and electrical energies.
observation of the relative mass drift between the international
prototype and its copies. The aim of this paper is to review the experiments
A drifting mass unit would also influence the electrical following the second strategy. Among the different experi-
units, since they are linked to the kilogram through the mental ideas, the most promising approach is the concept of
definition of the ampere. In 1990, international electrical the moving coil watt balance proposed by Kibble in 1976 [8].
reference standards based on the Josephson and the quantum The experiments based on this approach are described in detail.
Hall effects were introduced [3] by defining conventional
values KJ-90 and RK-90 for the Josephson constant and the
von Klitzing constant, respectively. All electrical quantities 2. Relating mass to the Planck constant via
can be measured in terms of these two conventional values. electromechanical experiments
As a consequence, the worldwide uniformity and consistency
of the electrical measurements has improved by almost two 2.1. Superconducting magnetic levitation
orders of magnitude. However, due to the uncertainty of the
values of the constants KJ and RK of several parts in 107 , results The experiment makes use of the force acting on a body with
expressed in the 1990 practical system of electrical units may diamagnetic properties in a non-uniform magnetic field. The
differ from the results in the SI by this amount. Moreover, the idea was first brought up by Sullivan and Frederich [9] as a
difference may change with time because of the possible drift possibility to realize the ampere. When a superconductor in
of the kilogram. the Meissner state is introduced into the field of a coil with
This inconsistency is not yet a problem in most of the decreasing magnetic flux density in the vertical direction, a
practical applications of the SI. However, to prepare the stable levitation of the body can be obtained (see figure 1).

0026-1394/03/060356+10$30.00 © 2003 BIPM and IOP Publishing Ltd Printed in the UK 356
 Tracing Planck’s constant to the kilogram

Timer 9 GHz
Id source φ m
Floating
body F
JJ m
JJ s
driver
Coil U

Is z
Laser- Figure 2. Principle of the voltage balance.
R SQUID interferometer

experiment can be operated in a virtual work mode following

U
the suggestion of Kibble [10].
Figure 1. Principle of the superconducting magnetic levitation The approach of superconducting magnetic levitation
experiment. A superconducting body is floating in a magnetic flux, has been pursued at the National Research Laboratory of
produced by a superconducting coil. Metrology (NRLM, now NMIJ) in Japan [11, 12] and the
D I Mendeleyev Research Institute of Metrology (VNIIM)
The energy equation of the system can then be written as in Russia [13]. The NRLM group has reached a resolution
of one part in 106 in its experiment [14]. A new set-up that
I U dt = I dφ = dA + dW, (1) should reduce some of the systematic errors was proposed in
2001 [15]. In the same year, a design study for a magnetic
where A is the work done by the field to increase the
levitation system was presented by the Centre for Metrology
gravitational energy of the body and W the magnetic field
and Accreditation (MIKES) in Finland [16]. The MIKES
energy. If the levitated body has ideal diamagnetic properties
and the coil circuit is superconducting as well, the energy terms group is developing a cryogenic calorimeter to measure the
are given by energy losses due to the non-ideal diamagnetic properties of
the levitated body.
A = mgz, W = 21 φI. (2)
2.2. Voltage balance
Considering two equilibrium positions with heights zl and zh ,
where the subscripts h and l denote high and low positions The principle of this approach is illustrated in figure 2. The
respectively, the energy difference takes the form electrostatic force acting between the plates of capacitance C is

φh compared with the weight mg of the test mass m, where g is the
I dφ = 21 (φh Ih − φl Il ) + mg(zh − zl ). (3) gravitational acceleration. The movable plate of the capacitor
φl is suspended from the balance. In the equilibrium position of
the balance, the forces are connected by the relationship
In practice, the experiment can be realized as shown in
figure 1. The superconducting coil is driven by a supply 1 2 ∂C
circuit which is controlled by a SQUID ammeter so that the mg = U , (4)
2 ∂z
drive current Id corresponds to the coil current Is . The drive
current is determined by measuring the voltage drop across a where U is the voltage across the capacitor and ∂C/∂z
resistance standard calibrated in terms of the quantized Hall the capacitance gradient in the vertical direction. The
resistance. When the Josephson junction in the coil circuit is measurement of the voltage is performed against a Josephson
biased on the first step for a time interval t, the flux in the voltage standard and the capacitance change is expressed in
coil is increased by φ = fJ tφ0 , where fJ is the Josephson terms of the quantized Hall resistance. In this way, a link
frequency and φ0 = h/2e the magnetic flux quantum. If φ is between the test mass and the Planck constant is established.
large enough, the superconducting body of mass m is levitated In a typical set-up (see [17] for a review), the voltage needed
and reaches the equilibrium position zl , which is measured is around 10 kV and the test mass is a few grams.
by laser interferometer. By repeating the process, a series Using this approach, Funck and Sienknecht from the
of equilibrium positions can be obtained where equation (3) Physikalisch-Technische Bundesanstalt (PTB) [18] reached a
describes the energy between any two positions. Since the relative standard uncertainty of 6.3×10−7 in the determination
flux change can be expressed in units of φ0 and the coil current of h [19]. The experiment was also carried out at the University
can be measured using the Josephson and the quantum Hall of Zagreb [20]. A relative uncertainty of 3.5 × 10−7 in the
effects, the experiment relates the mass of the floating body to determination of the volt was obtained in 1987–1988 [3].
the Planck constant. Subsequently, several systematic errors were found by Bego
The experiment has some major metrological difficulties et al in the set-up which led to improvements [20] and the
to overcome. The most important problems are: all unwanted proposition for a new 100 kV voltage balance [21]. To our
energy expenditure, e.g. due to horizontal force components knowledge, however, this work is not being continued, at least
on the trajectory of the levitated object or distortion of the at present.
object under the force of levitation, have to be avoided. The With the present techniques, the voltage balance approach
floating body has to be a perfect diamagnet and its mass has to does not promise to reach an uncertainty below one part
be known in the low-temperature environment. Note that this in 107 . The main problems are the high voltage required in

Metrologia, 40 (2003) 356–365 357

A Eichenberger et al

the experiment, the voltage and frequency dependence of the the voltage of the Josephson standard, C1 is a calibration
capacitance and its mechanical imperfections. As suggested constant, and FJ = ifJ denotes the product of the Josephson
by Kibble, the issues of the capacitance measurement can be frequency fJ and the quantum number i of the voltage step.
eliminated if ∂C/∂z is determined in a separate experimental Similarly, the current I can be determined from the voltage
phase [20]. drop U
∝ FJ
across a resistance calibrated against a quantum
Hall resistance standard. In this case, the resistance can be
2.3. The principle of the watt balance experiment expressed as R = C2 RH = C2 h/ne2 . C2 is again a numerical
factor determined from the calibration process, RH denotes
The concept of the moving coil watt balance was proposed by the quantized Hall resistance and n is an integer. Finally,
Kibble in 1976 [8]. The experiment is performed in two parts equation (7) can be rewritten as
(see figure 3). Consider a coil suspended from one arm of a
balance. The coil is immersed in a horizontal magnetic flux. FJ FJ

m=C h, (8)
The current I in the coil exerts a vertical force on the conductor gv
that is balanced against the weight (Fm ) of the test mass m and
we have where C represents the different calibration constants. The watt
∂ balance thus allows us to express the test mass in terms of the
Fm = mg = −I , (5)
∂z metre, the second and the Planck constant. One of the major
where g is the local acceleration due to gravity. In the second advantages of the experiment is that neither the geometry of the
part of the experiment, the coil is moved at a constant velocity coil nor the magnetic flux has to be known. Moreover, virtual
v in the vertical direction through the flux and the voltage U electrical and mechanical energies are related. This means
induced across the coil is measured, being that in contrast to the superconducting magnetic levitation
described above, real energy dissipation does not enter into
∂ ∂ ∂z ∂ the basic equation of the experiment.
U =− =− =− v (6)
∂t ∂z ∂t ∂z
2.4. A possible new definition of the kilogram based on the
at the location of the weighing. The elimination of the flux
Planck constant
gradient, γ = −∂/∂z, from equations (5) and (6) then
leads to The experiments described above establish a link between the
U I = mgv. (7) kilogram and the Planck constant. As the Planck constant
plays a unique role among the fundamental constants, both as
The experiment thus allows the virtual comparison of
quantum of action and as a factor of proportionality in many
the watt realized electrically (left-hand side of the equation)
equations, it would be a natural choice to fix the value of
to the watt realized mechanically. The voltage U can be
h and to link the kilogram to this value using experiments
measured against a Josephson voltage standard. It can thus
like the watt balance. To ensure the continuity within the SI,
be expressed as U = C1 UJ = C1 (h/2e)FJ , where UJ is
the conventionally fixed value of h would be the generally
accepted experimental value at the time of the new definition.
(a) According to propositions of Taylor and Mohr [22, 23], the
m new definition of the kilogram could read as follows: ‘The
kilogram is the mass of a body at rest whose equivalent
energy equals the energy of a collection of photons whose
I
frequencies sum to 135 639 274 × 1042 Hz’, or in other words:
R ‘One kilogram is a mass such that the Planck constant h is
exactly 6.626 068 76 × 10−34 J s’. These definitions are based
F on the well-known Einstein relation E = mc2 , where c is the
speed of light fixed by the definition of the metre, and the
B
relation E = hν valid for the energy of photons. Note that
(b) it is also possible to go directly to defining the mass of an
atomic particle, for example carbon-12, without a significant
reduction in accuracy (see equation (9)).

v 3. The watt balance experiments

3.1. The system design

B The watt balance experiment involves the quasi-simultaneous
determination of many different electrical and mechanical
Figure 3. Principle of the watt balance. (a) Static mode: the quantities at the highest attainable accuracy. The electrical
electromagnetic force acting on the current carrying coil is balanced
against the weight of the test mass. (b) Dynamic mode: the coil is quantities are the current I and the induced voltage U .
moved in the vertical direction through the magnetic field and the The mechanical quantities are the translation velocity v, the
induced voltage is measured. gravitational acceleration g and the mass m. Neither the

358 Metrologia, 40 (2003) 356–365

 Tracing Planck’s constant to the kilogram

dimensions of the coil nor the flux density and distribution of

v
the magnetic flux need to be measured and the ratio between Fe z
mechanical and electrical watt is obtained by direct comparison
of γ extracted from both phases of the measurement sequence ϕ
(U/v and Fm /I , respectively).
ψ λ
3.1.1. The electromechanical design, alignment and other α
constraints. Because a variety of high accuracy measure-
ments have to be performed with one single experimental δx ρx x

set-up, no single separate key element can be optimally Center of mass

designed. The difficulty is to make the right compromises. Electrical center
Optical center
One of the key elements of the apparatus is the magnetic Fm = m g
circuit. There are basically two possibilities to produce the
magnetic flux: either with permanent magnets combined with a
yoke or with a suitable electromagnet. The required magnitude Figure 4. The laser beam (λ), the velocity (v), and the force (Fe )
of the flux limits the choice of materials for the permanent generated by the current in the coil must ideally be aligned with
magnets and almost implies the use of superconducting respect to the vertical direction (α = ϕ = ψ = 0) defined by the
gravitational acceleration g. Moreover, the centre of mass of the coil
and its suspension (⊗), the electrical centre (◦) and the optical
solenoids in the second alternative. This last option involves
a cryogenic environment leading to a considerable increase in centre (•) must coincide in the x–y plane (δx = ρx = 0). Any
complexity and size of the whole apparatus whereas permanent misalignment can result in systematic errors.
magnets have the major disadvantage of a large temperature
dependence. The magnet design is strongly coupled to the
the system. If the optical and electrical centres are different,
coil used for both phases of the measurement sequence. This
a non-zero angular component in the velocity of the coil (about
pair of crucial elements is often referred to as the magnet–coil
x or y axis) during the motion induces a signal in the velocity
assembly. The coil geometry and structure must be optimized
that has no corresponding voltage and therefore causes an error
in terms of ampere-turns, orientation of wires, size, rigidity
in the measured watt value (often called Abbe error). The
and weight. The final design is also strongly affected by
sensitivity of the system to these alignment errors strongly
the size of the test mass. Indeed, a 1 kg test mass is not
depends on the practical realization of the experiment and the
necessarily required since submultiple mass comparisons can
ability to control and check the alignment of the system. An
be performed at the level of one part in 108 . A smaller test
extensive description of alignment methods and uncertainties
mass means a weaker constraint on the magnet–coil assembly.
can be found in [24, 25].
Either the magnitude of the field and ampere-turns of the coil or
Last but not the least, to achieve the required level of
its size can be reduced. In principle, the weighing current has
accuracy (one part in 108 ), the experiment must be run in
to remain small (a few milliamperes) in order to avoid thermal
vacuum. This avoids air buoyancy and index of refraction
problems caused by power dissipation during the static phase.
corrections that would necessitate a thorough knowledge of
The way the coil moves in the magnetic field has also to be
the surrounding air composition, pressure and temperature.
carefully considered and directly depends on the mechanical
realization of the dynamic phase. This motion is controlled
and measured with laser interferometry. 3.1.2. The test mass. The material used for the test mass is
Both the velocity vector v and the laser beam direction a critical point since a large number of specifications must be
have to be aligned with respect to the vertical direction simultaneously fulfilled. A magnetic susceptibility as weak as
defined by the gravitational acceleration g (identified as the possible is highly desirable to reduce the parasitic forces due to
z direction, see figure 4). In principle, a non-verticality of the proximity of strong magnetic fields in the experiment [26].
the laser interferometer (α = 0 in figure 4) can be measured A very good surface stability is required to reach the best
and corrected. However, any horizontal motion of the coil possible mass stability in air and in vacuum. The material
(ϕ = 0) can produce an additional induced voltage (due itself, the surface quality and the cleanliness of the surface
to inhomogeneities or fringe fields) that contributes to the may have a strong influence on this stability. The test mass
signal measured in the dynamic phase but not in the static must also have a high density to reduce the active surface in
one, and therefore produces a systematic error. If the force is contact with the atmosphere. To guarantee long-term stability
not perfectly vertical (ψ = 0), only the vertical component of its mass, the test mass should have a very good homogeneity,
contributes to the measurement and causes an error in the no porosity, cavity or occluded gases. In addition, the ideal test
determination of the mechanical watt. In general, the coil and mass should be hard enough to facilitate machining, polishing
its suspension mechanism have a centre of mass, an optical and reduce wear during manipulations. A good electrical
centre (where the interferometer laser beam hits the coil and conductivity to avoid parasitic forces due to the accumulation
the velocity is measured) and an electrical centre defined by of electrostatic charges and a good thermal conductivity to
the origin of the force produced by the weighing current. If the minimize temperature effects when weighing in air are also
projections of these centres in the x–y plane do not coincide, an desirable. Some of these requirements can be relaxed if
error can result. For example, if the centre of mass is different suitable precautions are taken. For example, a conductive
from the electrical centre, the force applies a torque to the coating or the use of an additional air/vacuum transfer mass
coil and its suspension and could cause an undesirable tilt in releases the constraints on the electrical properties of the

Metrologia, 40 (2003) 356–365 359

A Eichenberger et al

material or does not necessitate the direct transfer of the test to experimentally overcome the discontinuity of the mass
mass to air. scale between air and vacuum were proposed by Kibble and
Robinson [40, 41].
3.1.3. The dynamic mode. In the dynamic mode of the experi- Finally, the accurate determination of the absolute value
ment, the voltage U induced in the coil by a vertical translation of the gravitational (or free-fall) acceleration g next to the
at a velocity v is measured against a voltage reference. This experiment and synchronously to the static mode measurement
is most conveniently done using a programmable Josephson is required to get the expression of the mechanical force
voltage standard (PJVS) that can be rapidly set to any chosen F = mg. The measured gravitational acceleration only needs
value between −1 V and +1 V [27–29]. to be slightly corrected to take into account the horizontal and
The velocity signal comes from a carefully designed vertical gradients between the absolute gravity site where the
gravimeter is located and the exact position of the test mass
interferometer [30, 31] and either this signal or the induced
inside the experiment. The influence of the mass distribution
voltage can be used in a regulation loop to control the
of the experiment itself must also be considered.
motion. Since the sign of the induced voltage is reversed
The ratio between the weight of the test mass and
when the direction of the motion is inverted, voltage offsets
the weighing current is the second determination of γ at
in the electrical circuit can be removed when up and down
wpos . The comparison of the two values of γ (wpos ) is a
measurements are averaged. In principle, instantaneous values
virtual comparison between mechanical and electrical watt.
for U and v are required as the coil passes through the weighing
Because in both phases the sense of the voltage measured
position wpos . In practice, the magnetic flux can be shaped and of the voltage standard are reversed, any constant (at the
in a way that U/v is almost constant and an average over a timescale of the measurement sequence) parasitic voltage
finite distance is sufficient. Since the instantaneous values of sources (like thermoelectric emf or offset of the integrating
voltage and velocity are proportional, if the mean voltage and voltmeter) are eliminated by taking the mean of the forward and
velocity over the same time interval are observed, their ratio reverse measurements. A proper timing of the measurement
will be independent of their fluctuations to a great extent (due sequence can also take care of a linear drift of the experimental
to vibrations, for example) [32]. The mean velocity during the parameters (like temperature, voltage offsets, etc).
integration time of the voltage measurement is extracted from
the interferometer signal and the value of the ratio U/v at wpos
3.2. The existing projects and their main characteristics
is then interpolated and gives a first measurement of γ (wpos ).
Watt balance experiments are currently pursued at three
3.1.4. The static mode. In the static mode of the experiment, different institutes.
a current I flowing in the coil generates a Lorentz force Fe to Shortly after Kibble’s proposal in 1976 [8], a moving coil
balance the mechanical force Fm produced by the test mass in apparatus was developed at the National Physical Laboratory
the gravitational field. In practice, the balance is underloaded (NPL) in the UK. The magnetic flux was generated by a
by half of the value of the test mass. A first weighing with permanent magnet and the moving coil consisted of a flat
a current producing the force needed to balance the system 8-shaped coil. The final result of this set-up with a relative
without the test mass is followed by a second one, where the standard uncertainty of 2×10−7 was published in 1990 [32,42].
sign of the current is reversed and the test mass placed on the An improved apparatus, described in the next sections, was
presented during the same year [43, 44]. In the mean time,
balance. These currents are controlled to keep the balance
this experiment has reached a short-term reproducibility in the
at the equilibrium position and the values are measured with
order of one part in 108 [45], and a new result for the Planck
the help of a standard resistor, periodically calibrated against
constant may be expected in the near future.
the quantum Hall resistance standard, and a voltage reference
Studies for a watt balance also began very early at the
(PJVS, for example) [33].
National Institute of Standards and Technology (NIST) in
The mass m of the test mass is determined in air by
the USA. Shortly after the introduction of the concept, an
the classical methods of mass metrology using a prototype apparatus with a magnetic flux generated by an electromagnet
balance or a mass comparator [34, 35]. The test mass is then was constructed [46]. The first results published in 1989
directly traceable to national prototypes of the kilogram. As [47, 48] had a relative standard uncertainty of 1.3 × 10−6 .
watt balances are operated under vacuum, the mass of the test After completion of this first phase, NIST researchers replaced
mass should also be known under vacuum. The immediate the electromagnet by a superconducting magnet and realized
advantage of the vacuum measurement is the suppression of various further improvements [49–51], which led to the result
the air buoyancy correction. The main disadvantage is the reported in 1998 with a relative standard uncertainty of 9×10−8
discontinuity of the mass scale between air and vacuum. This [52]. Currently, the NIST group continues to work with a
discontinuity is due to the physical and chemical adsorbed further improved version of the experiment.
layers at the surface of the test mass [36]. Due to their different In 1997, a third institute, the Swiss Federal Office of
binding energies, the structure and the thickness of the different Metrology and Accreditation (METAS), started to build a
adsorbed layers are not identical in air and in vacuum. Many watt balance [53]. The apparatus incorporates some new
studies have already been conducted to evaluate this effect on approaches to the difficulties of the measurement (see later).
different materials such as stainless steel [37, 38] and gold- It is fully assembled and the testing and evaluation phase has
plated copper [39]. If the adsorption phenomenon is well started [53]. First results can be expected within the next few
characterized and the surface area of the mass standard is years.
known, the relative uncertainty component due to sorption The main characteristics of the three existing set-ups are
effects is far below one part in 108 [39]. Different methods summarized in table 1.

360 Metrologia, 40 (2003) 356–365

 Tracing Planck’s constant to the kilogram

Table 1. Summary of main parameters in the existing watt balance experiments.

NPL NIST METAS
Test mass, nominal value (kg) 1 (0.5) 1 0.1 (0.05)
Test mass, material Gold-plated Steel, Au, Pt–Ir Steel, Au, Au–Pt
OFHC copper
Vacuum working level (Pa) 100–0.01 0.5 10−4
Overall size of the experiment (m) 2×1×2 3.7 × 4.5 × 7 1 × 1 × 1.5
Magnet type SmCo Superconducting SmCo
Magnet weight (kg) 900 350 4
Magnitude of the field at coil position (T) 0.4 0.1 0.5
Magnetic field geometry Radial Radial Parallel homogeneous
Field magnitude variation over eff. 50 ×10−6 , parabolic 150 ×10−6 , M-shape 400 ×10−6 , parabolic
travel range, shape
Coil ampere turns 11 25.2 20
Coil geometry Circular Circular 8-shape
Coil dimension (cm) ∅ 34 ∅ 70 8×8
Coil wire insulation Enamel HTAIH wire insulation, Modified polyurethane
fibreglass weave interlayer spacer
Coil material Pyrex former G10 spool form, ceramic Macor
reinforcing outer form
AC rejection method Two coils connected Dual fixed coils 8-shape
in opposition
Interferometer type Michelson Michelson, heterodyne (3 on coil, Fabry–Perot
1 on spider support)
Static mode
Typical weighing current (mA) 16 10.18 5
Power dissipated in the coil (mW) 13 75 50
Balance type Beam (l = 1.2 m) wheel (∅ 0.62 m) Mass comparator
Total mass suspended
(beam, coil, tare, etc), net load (kg) 70 63 2.2
Absolute gravimeter type FG5 FG5 FG5
Dynamic mode
Typical velocity (mm s−1 ) 1.3 2 3
Typical induced voltage (V) 0.4 1.018 0.3
Regulation source Velocity Induced voltage Velocity
Regulation stability (velocity 50 ×10−6 /1.75 s 1000 ×10−6 /67 ms 500 ×10−6 /0.2 s
or induced voltage)/corresponding
averaging time
Total travel length (mm) 40 89 (95 max) 50
Effective travel length (where (mm) 20 79 30
measurements take place)
Typical duration of one (min) 120 60 50
complete cycle

3.2.1. The electromechanical designs. NPL opted for a external perturbations in the surroundings of the experiment.
design with a standard beam balance on a knife edge pivot. To that end, a non-magnetic building has been built and special
The beam system is simple, rigid and convenient to couple to care has been taken to avoid or reduce any electromagnetic
the rest of the experiment. It has, however, the disadvantage of perturbation close to the experiment (additional fixed coils,
producing a non-vertical motion following an arc of a circle. To for example). The resulting size of this design is quite large
minimize this horizontal motion, the length of the balance arm and the vacuum chamber geometry complicated. A schematic
must be as long as possible (1.2 m in this case). A permanent of the NIST set-up is shown in figure 6.
magnet produces a radial field (0.4 T) between two cylindrical In the METAS design, the main idea was to separate
surfaces where the circular main coil (∅ 34 cm) is placed. A the static from the dynamic phase, allowing the use of a
picture of the present design is shown in figure 5. mass comparator for the residual force measurement and
In the NIST design, the beam has been replaced by avoiding most of the hysteretic problems at the balance
a wheel (∅ 62 cm), reducing the horizontal motion to the suspension [54]. Special care was also taken to keep the size
degree of the non-perfect circularity of the wheel. The major of the whole experiment as small as possible to minimize
difference of the NIST design is undoubtedly the option of vibration problems. With that in mind, a 100 g test mass
two superconductive solenoids wired in series opposition to was chosen instead of a 1 kg test mass. The mechanism to
produce a radial magnetic field of 0.1 T (at the coil position). induce the vertical motion is a mixture between a beam and
A large circular coil (∅ 70 cm) is placed around the liquid a wheel balance. It consists of two arms and two vertical
helium dewar. In this case, no yoke or magnetic materials boxes assembled together with BeCu strips and forming a
are used to confine the field lines and special care should be parallelogram structure. The top arm is suspended from a
taken to avoid any noticeable distortion of the field lines due to central structure and the arm ends are machined at a radius

Metrologia, 40 (2003) 356–365 361

A Eichenberger et al

Mass
comparator

Test
mass

Auxiliary
coil

Permanent
magnets
Figure 5. Picture of the present NPL design. Note the beam of the Parallelogram
balance (partly visible) and the thermally stabilized vacuum structure
enclosure containing the permanent magnet (bottom left). Main
coil

0.3 m
Balance
wheel

Vacuum
chamber

Laser
interferometer
Auxiliary Test
drive coil mass
Figure 7. Schematics of the METAS design. Note the mass
comparator on top of the set-up and the seesaw mechanism used to
produce the vertical motion of the coil in the dynamic phase.

3.2.2. The test masses. NPL uses 1 kg gold-plated oxygen-

free high purity (OFHC) copper test masses [30]. The surface
1m

mirror finishes of the test masses are similar to those of the

most recent Pt/Ir kilogram prototypes. The NPL test masses
Induction coils: are always kept in vacuum except for the short time needed
fixed
movable
to transfer them to a vacuum container or to a vacuum mass
fixed comparator for the mass calibration [30].
NIST used two pure gold cylinders totalling 1 kg. After
Superconducting noticing clear signs of wear following intensive measurement
Laser interferometer
magnet
(1 of 3) campaigns, it was decided to manufacture a special Pt–Ir
Figure 6. Schematics of the NIST design. Note the balance wheel mass for the experiment [55]. Stainless steel masses are also
and the superconducting magnet. available for test purposes.
At METAS, 100 g pure gold test masses were
characterized. Gold/platinum alloys are now considered for
equal to half of the arm length. The whole seesaw mechanism
their similar low magnetic susceptibility and much higher
can induce an up/down motion of the two vertical parts literally
hardness [26]. For tests and preliminary measurements,
rolling over the arm ends. A horizontal (parallel homogeneous)
stainless steel cylinders of 50 g and 100 g have been used [56].
magnetic field is produced by a permanent magnet coupled
to a yoke. The coil is wired in 8-shape to reduce coupling
to external electromagnetic perturbation. A considerable 3.2.3. Characteristics of dynamic modes. Opposite to the
technical challenge of the METAS design is the necessity measurement side (where the main coil is suspended) is the
to transfer the coil from the parallelogram structure (in the so-called servo side where an auxiliary coil and an additional
dynamic phase) to the comparator frame (in the static phase). magnetic circuit are used to produce the required force to
Since the coil is carrying the interferometer moving mirror, its move the coil in the dynamic phase. In the NIST experiment,
repositioning must be controlled at the microradian level. This the induced voltage is usually kept constant by regulating
ability was achieved with mechanical lifters actuated with dc the current in the auxiliary coil. In the NPL and METAS
motors from the outside of the vacuum chamber. A schematic experiments, the scrolling rate of the interferometer fringes is
of the METAS design can be found in figure 7. used to adjust the servo current to maintain a constant velocity.

362 Metrologia, 40 (2003) 356–365

 Tracing Planck’s constant to the kilogram

The choice of a Fabry–Perot interferometer (instead of beam balance and power dissipation during the static phase
the regular Michelson set-up) in the METAS experiment was (only 4.25 mW). In the dynamic phase, the coil and the mass
motivated by the much sharper shape of the fringes and comparator will be moved at a constant velocity (2 mm s−1 )
its extreme sensitivity to misalignment [31]. In the NIST producing a 1 V induced voltage at the coil terminals. The
experiment, the use of three interferometers (placed 120˚ apart vertical displacement is guided by a set of six flexure strips
on the coil circumference) allows one to monitor the plane of and controlled with a heterodyne laser interferometer. An
the coil and to detect tilts. The motion of the optical centre impressive velocity stability is announced and should reach
is then calculated from the information collected from these one part in 109 [59]. The French group also plans to develop
three points. a cold atom absolute gravimeter for the determination of
the gravitational acceleration. The final experiment will be
3.2.4. Characteristics of static modes. In the NPL or NIST installed in a specially designed new building.
experiment, the balance arm or wheel is maintained at a fixed In 2002, the BIPM presented a proposal for a cryogenic
position. The weighing current is regulated using the signal watt balance experiment [60]. The BIPM planned to dedicate
from a laser interferometer to keep the balance from moving. at least one year to the consideration and analysis of different
In the NPL system, this signal comes from the same laser designs and of realistic technical solutions.
system as the one used for the dynamic phase whereas in the
NIST experiment, it comes from a fourth interferometer that 4. The Planck constant today
hits the centre of the spider suspension of the mass. In both
experiments, hysteresis problems in the knife edge suspension The CODATA recommended value for the Planck constant
of the balance are one source of systematic error [54]. h [19] is mainly based on the results of the watt balance
The main difference between the METAS set-up and the experiments. In figure 8, all results with a relative standard
two others is the use of a mass comparator. In this case, the uncertainty below 1 × 10−6 are shown.
current in the main coil is kept constant and a mass comparator With the exception of the results deduced from the
takes care of the residual force measurement. This device Avogadro experiments, the values are taken from [19]. Due to
is a modified commercially available piece of equipment improvements in the analysis, they may, in some cases, differ
optimized for this application [53]. from the data first published by the experimenters. The values
The three watt balance experiments are using the same labelled ‘Avogadro 2001’ and ‘Avogadro 2003’ are determined
type of commercial free-fall absolute gravimeter (FG5) [57]. from the latest published values of the Avogadro constant
In such an instrument, a laser interferometer and an atomic [61–63]. The Planck and Avogadro constants are related by
clock are used to determine the free-fall trajectory of an optical
cAr (e)M0 α 2
element within an evacuated vacuum chamber. The free-fall h= , (9)
acceleration is then calculated by fitting the time–position data 2R∞ NA
to the equation of motion. The three absolute gravimeters used where Ar (e) is the relative atomic mass of the electron, α is
in each watt balance experiments were recently successfully the fine structure constant and R∞ is the Rydberg constant.
compared to 14 other absolute gravimeters of different type The combined uncertainty of this group of constants is below
at BIPM [58]. The results of this comparison show that one part in 108 . The value of the molar mass constant is
the uncertainty component due to the determination of the M0 = 10−3 kg mol−1 exactly.
gravitational acceleration is not a limiting factor for a watt The relative differences between h calculated using
balance experiment. equation (9) and the CODATA value are (1.3 ± 0.5) × 10−6
and (1.1 ± 0.3) × 10−6 , respectively. This may point to an
3.3. Future projects unresolved systematic error in one of the experiments.
The largest diversity in the different watt balance
One of the challenges at the start of a new watt balance experiments around the world is to be encouraged. This
project is the choice of an original set-up, considering as is the only way to check in every detail each possible
much as possible the advantages and drawbacks of the previous source of systematic errors. An agreement between the
experiments.
Recently, two new projects were announced. In 2000, the
Bureau National de Métrologie (BNM) in France has chosen voltage balance, PTB 91
the watt balance approach. The French set-up will differ Hg electrometer, NML 89
substantially from the three already existing ones. The design
Avogadro, 2001
phase is about to be finished and the individual components
are undergoing tests in different laboratories. In the static Avogadro, 2003
phase, the weight of the 500 g gold platinum alloy test mass watt balance, NPL 90
will be balanced by a 5 mA current flowing in a circular coil
watt balance, NIST 98
(∅ 27 cm) immersed in a 1 T radial magnetic field generated
by permanent magnets (SmCo). The yoke is an assembly CODATA-98
of pure iron and iron–cobalt core with a circular geometry -10 -5 0 5 10 15 20
(total weight ∼400 kg). A flexure strip mass comparator will (h - hCODATA)/hCODATA × 10
7

be specially developed for this experiment. The design of the

coil (2.9 ampere turns) is optimized in terms of load on the Figure 8. Experimental values for the Planck constant.

Metrologia, 40 (2003) 356–365 363

A Eichenberger et al

values of the Planck constant obtained from these various [18] Funck T and Sienknecht V 1991 Determination of the volt
experiments would definitely provide a convincing argument with the improved PTB voltage balance IEEE Trans.
Instrum. Meas. 40 158–61
to metrologists working on a redefinition of the mass unit based [19] Mohr P J and Taylor B N 2000 CODATA recommended values
on fundamental constants. of the fundamental physical constants: 1998 Rev. Mod.
Phys. 72 351–495
[20] Bego V, Butorac J and Poljančić K 1995 Voltage balance for
Acknowledgments replacing the kilogram IEEE Trans. Instrum. Meas. 44
579–82
The authors would like to thank Ian Robinson, Richard Steiner, [21] Bego V, Butorac J and Ilić D 1999 Realization of the kilogram
Edwin Williams and Gérard Genevès for providing figures by measuring at 100 kV with the voltage balance ETF IEEE
Trans. Instrum. Meas. 48 212–15
and sharing information about their experiments. The authors [22] Taylor B N and Mohr P J 1999 On the redefinition of the
are grateful to Blaise Jeanneret for the critical reading of the kilogram Metrologia 36 63–4
manuscript. [23] Taylor B N and Mohr P J 2001 The role of fundamental
constants in the international system of units (SI): present
and future IEEE Trans. Instrum. Meas. 50 563–7
References [24] Stenbakken G N, Steiner R L, Olsen P T and Williams E R
1996 Methods for aligning the NIST watt-balance IEEE
[1] Quinn T J 1991 The kilogram: the present state of our Trans. Instrum. Meas. 45 372–7
knowledge IEEE Trans. Instrum. Meas. 40 81–5 [25] Gillespie A D, Fujii K, Newell D B, Olsen P T, Picard A,
[2] Girard G 1994 The third periodic verification of national Steiner R L, Stenbakken G N and Williams E R 1997
prototypes of the kilogram (1988–1992) Metrologia 31 Alignment uncertainties of the NIST watt experiment IEEE
317–36 Trans. Instrum. Meas. 46 605–8
[3] Taylor B N and Witt T J 1989 New international electric [26] Silvestri Z, Davis R S, Genevès G, Gosset A, Madec T, Pinot P
reference standards based on the Josephson and quantum and Richard P 2003 Volume magnetic susceptibility of
Hall effects Metrologia 26 47–62 gold–platinum alloys: possible materials to make mass
[4] Kestenbaum D 1998 Recipe for a kilogram Science 280 standards for the watt balance experiment Metrologia 40
823–4 172–6
[5] Göbel E O 2001 Wer gewinnt den Wettlauf um das Kilogramm [27] Benz S P, Hamilton C A, Burroughs C J, Harvey T E and
Physikalische Blätter 57 35–41 Christian L A 1997 Stable 1 volt programmable voltage
[6] Becker P 2001 History and progress in the accurate standard Appl. Phys. Lett. 71 1866–8
determination of the Avogadro constant Rep. Prog. Phys. 64 [28] Burroughs C, Benz S P, Harvey T E and Hamilton C A 1999
1945–2008 1 volt DC programmable Josephson voltage standard IEEE
[7] Ratschko D, Knolle D, Finke E and Gläser M 2002 Trans. Appl. Supercond. 9 4145–9
Accumulation of decelerated gold ions Nucl. Instrum. [29] Jeanneret B, Rüfenacht A and Burroughs C J 2001 High
Methods Phys. Res. B 190 217–21 precision comparison between SNS and SIS Josephson
[8] Kibble B P 1976 A measurement of the gyromagnetic ratio of voltage standards IEEE Trans. Instrum. Meas. 50
the proton by the strong field method Atomic Masses and 188–91
Fundamental Constants 5 ed J H Sanders and A H Wapstra [30] Robinson I A and Kibble B P 1997 The NPL moving-coil
(New York: Plenum) pp 545–51 apparatus for measuring Planck’s constant and monitoring
[9] Sullivan D B and Frederich N V 1977 Can superconductivity the kilogram IEEE Trans. Instrum. Meas. 46 596–600
contribute to the determination of the absolute ampere? [31] Courteville A, Salvadé Y and Dändliker R 2000
IEEE Trans. Magn. 13 396–9 High-precision velocimetry: optimization of a Fabry–Perot
[10] Kibble B P 1983 Realizing the ampere by levitating a interferometer Appl. Opt. 39 1521–6
superconducting mass—a suggested procedure IEEE Trans. [32] Kibble B P, Robinson I A and Belliss J H 1990 A realization of
Instrum. Meas. 32 144 the SI watt by the NPL moving-coil balance Metrologia 27
[11] Shiota F and Hara K 1987 A study of a superconducting 173–92
magnetic levitation system for an absolute determination of [33] Beer W, Eichenberger A L, Jeanneret B, Jeckelmann B,
the magnetic flux quantum IEEE Trans. Instrum. Meas. 36 Richard P, Schneiter H, Pourzand A R, Courteville A and
271–4 Dändliker R 2001 The OFMET watt balance: progress
[12] Shiota F, Miki Y, Namba A, Nezu Y, Sakamoto Y, report IEEE Trans. Instrum. Meas. 50 583–6
Morokuma T and Hara K 1995 Absolute determination of [34] Schwartz R 2000 Mass determination with balances
the magnetic flux quantum using superconducting magnetic Comprehensive Mass Metrology ed M Kochsiek and
levitation IEEE Trans. Instrum. Meas. 44 583–6 M Gläser (New York: Wiley-VCH) pp 232–95
[13] Frantsuz E T, Gorchakov Y D and Khavinson V M 1992 [35] Davis R S 2001 Proc. International School of Physics
Measurements of the magnetic flux quantum, Planck Enrico Fermi, Course CXLVI ed T J Quinn et al,
constant, and elementary charge at VNIIM IEEE Trans. pp 241–61
Instrum. Meas. 41 482–5 [36] Kochsiek M 1982 Measurement of water adsorption layers on
[14] Shiota F, Miki Y, Fujii Y, Morokuma T and Nezu Y 2000 metal surfaces Metrologia 18 153–9
Evaluation of equilibrium trajectory of superconducting [37] Schwartz R 1994 Precision determination of adsorption layers
magnetic levitation system for the future kg unit of mass on stainless steel mass standards by mass comparison and
IEEE Trans. Instrum. Meas. 49 1117–21 ellipsometry. Part I: Adsorption isotherms in air Metrologia
[15] Fujii Y, Miki Y, Shiota F and Morokuma T 2001 Mechanism 31 117–28
for levitated superconductor experiment IEEE Trans. [38] Schwartz R 1994 Precision determination of adsorption layers
Instrum. Meas. 50 580–2 on stainless steel mass standards by mass comparison and
[16] Riski K, Heikkinen P, Kajastie H, Manninen J, Rossi H, ellipsometry. Part II: Sorption phenomena in vacuum
Nummila K, Frantsuz E and Khavinson V 2001 Design of a Metrologia 31 129–36
superconductiong magnetic levitation system Proc. IMEKO [39] Beer W, Fasel W, Moll E, Richard P, Schneiter U, Thalmann R
TC3 2001 pp 239–46 and Egger J 2002 The METAS 1 kg vacuum mass
[17] Bego V 1988 Determination of the volt by means of voltage comparator—adsorption layer measurement on gold-coated
balances Metrologia 25 127–33 copper buoyancy artefacts Metrologia 39 263–8

364 Metrologia, 40 (2003) 356–365

 Tracing Planck’s constant to the kilogram

[40] Kibble B P 1990 Comparing a mass in vacuum with another in The NIST watt balance: progress toward the monitoring of
air by conventional weighing Metrologia 27 157–8 the kilogram IEEE Trans. Instrum. Meas. 46 601–4
[41] Robinson I A 1990 Comparing in-air and in-vacuum mass [52] Williams E R, Steiner R L, Newell D B and Olsen P T 1998
standards without buoyancy corrections via in-vacuum Accurate measurement of the Planck constant Phys. Rev.
weighing Metrologia 27 159 Lett. 81 2404–07
[42] Davis R S 1989 The stability of the SI unit of mass as [53] Beer W, Jeanneret B, Jeckelmann B, Richard P, Courteville A,
determined from electrical measurements Metrologia 26 Salvadé Y and Dändliker R 1999 A proposal for a new
75–6 moving-coil experiment IEEE Trans. Instrum. Meas. 48
[43] Kibble B P, Robinson I A and Belliss J H 1990 Re-defining the 192–95
kilogram via a moving-coil apparatus Conf. on Precision [54] Schwarz J P, Liu R, Newell D B, Steiner R L, Williams E R
Electromagnetic Measurements, CPEM, Conference Digest and Smith D 2001 Hysteresis and related error mechanisms
pp 178–9 in the NIST watt balance experiment J. Res. Natl Inst.
[44] Kibble B P, Robinson I A and Belliss J H 1992 The new NPL Stand. Technol. 106 627–40
moving-coil watt balance—a progress report Conf. on [55] Newell D B, Schwartz J P and Williams E R 1999 Reference
Precision Electromagnetic Measurements, CPEM, standard, uncertainties, and the future of the NIST
Conference Digest pp 8–9 electronic kilogram Proc. NCSL Workshop and Symp. 1999
[45] Robinson I A and Kibble B P 2002 Progress in relating the pp 319–26
kilogram to Planck’s constant with the NPL watt balance [56] Beer W, Eichenberger A L, Jeanneret B, Jeckelmann B,
Conf. on Precision Electromagnetic Measurements, CPEM, Pourzand A R, Richard P and Schwarz J P 2003 Status of
Conference Digest pp 574–5 the METAS watt balance experiment IEEE Trans. Instrum.
[46] Olsen P T, Bower V E, Phillips W D, Williams E R and Meas. 52 626–30
Jones G R 1985 The NBS absolute ampere experiment [57] Niebauer T M, Sasagawa G S, Faller J E, Hilt R and
IEEE Trans. Instrum. Meas. 34 175–81 Klopping F 1995 A new generation of absolute gravimeters
[47] Olsen P T, Elmquist R E, Philips W D, Williams E R, Metrologia 32 159–80
Jones G R and Bower V E 1989 A measurement of the NBS [58] Vitushkin L et al 2002 Results of the sixth international
electrical watt in SI units IEEE Trans. Instrum. Meas. 38 comparison of absolute gravimeters, ICAG-2001
238–44 Metrologia 39 407–24
[48] Cage M E et al 1989 NBS determination of the fine-structure [59] Genevès G 2003 private communication
constant, and of the quantized Hall resistance and [60] Quinn T J, Davis R S, Witt T J, Delahaye F, Picard A and
Josephson frequency-to-voltage quotient in SI units IEEE Fang H 2002 Outline proposal for a cryogenic watt balance
Trans. Instrum. Meas. 38 284–9 experiment Document CCEM/02-09 Submitted to 23rd
[49] Olsen P T, Tew W L, Williams E R, Elmquist R E and meeting of the CCEM
Sasaki H 1991 Monitoring the mass standard via the [61] de Bièvre P et al 2001 A reassessment of the molar volume of
comparison of mechanical to electrical power IEEE Trans. silicon and the Avogadro constant IEEE Trans. Instrum.
Instrum. Meas. 40 115–20 Meas. 50 593–7
[50] Olsen P T, Steiner R L, Jones G R and Williams E R 1994 The [62] Becker P 2001 The molar volume of single-crystal silicon
present watt balance: a measure of its resolution Conf. on Metrologia 38 85–6
Precision Electromagnetic Measurements, CPEM, [63] Becker P, Bettin H, Danzenbrink H-U, Gläser M, Kuetgens U,
Conference Digest pp 528–9 Nicolaus A, Schiel D, de Bièvre P, Valkiers S and Taylor P
[51] Steiner R L, Gillespie A G, Fujii K, Williams E R, 2003 Determination of the Avogadro constant via the
Newell D B, Picard A, Stenbakken G N and Olsen P T 1997 silicon route Metrologia 40 271–87

Metrologia, 40 (2003) 356–365 365

View publication stats