Azuma 2015 Metrologia 52 360

Metrologia
PAPER • OPEN ACCESS Related content

- Realization of the kilogram by the XRCD
Improved measurement results for the Avogadro method
Kenichi Fujii, Horst Bettin, Peter Becker et
constant using a 28Si-enriched crystal al.
- A new 28Si single crystal: counting the

atoms for the new kilogram definition
To cite this article: Y Azuma et al 2015 Metrologia 52 360 G Bartl, P Becker, B Beckhoff et al.
- Counting the atoms in a 28Si crystal for a

new kilogram definition
B Andreas, Y Azuma, G Bartl et al.
View the article online for updates and enhancements.
Recent citations
- The Elimination of the Effect of the
External Field in the Joule Balance
Jinxin Xu et al
- Some aspects on the uncertainty

calculation in Mueller ellipsometry
Matthias Wurm et al
- Trustworthy data underpin reproducible

research
Martin J. T. Milton and Antonio Possolo
This content was downloaded from IP address 93.139.8.60 on 14/03/2020 at 23:06

Bureau International des Poids et Mesures Metrologia
Metrologia 52 (2015) 360–375 doi:10.1088/0026-1394/52/2/360
Improved measurement results for

the Avogadro constant using a
28
Si-enriched crystal
Y Azuma1, P Barat2, G Bartl3, H Bettin3, M Borys3, I Busch3, L Cibik3,
G D’Agostino4, K Fujii1, H Fujimoto1, A Hioki1, M Krumrey3, U Kuetgens3,
N Kuramoto1, G Mana4, E Massa4, R Meeß3, S Mizushima1, T Narukawa1,
A Nicolaus3, A Pramann3, S A Rabb5, O Rienitz3, C Sasso4, M Stock2,
R D Vocke Jr5, A Waseda1, S Wundrack3 and S Zakel3
1
National Metrology Institute of Japan NMIJ, 1-1-1 Umezono, Tsukuba, Ibaraki 305-8563, Japan
2
Bureau International des Poids et Mesures BIPM, Pavillon de Breteuil, 92312 Sèvres Cedex, France
3
Physikalisch-Technische Bundesanstalt PTB, Bundesallee 100, 38116 Braunschweig, Germany
4
Istituto Nazionale di Ricerca Metrologica INRIM, str. delle cacce 91, 10135 Torino, Italy
5
National Institute of Standards and Technology NIST, 100 Bureau Drive, Gaithersburg, MD 20899, USA
E-mail: [email protected]
Received 31 December 2014, revised 16 February 2015

Accepted for publication 16 February 2015
Published 25 March 2015
Abstract
New results are reported from an ongoing international research effort to accurately determine
the Avogadro constant by counting the atoms in an isotopically enriched silicon crystal. The
surfaces of two 28Si-enriched spheres were decontaminated and reworked in order to produce
an outer surface without metal contamination and improved sphericity. New measurements
were then made on these two reconditioned spheres using improved methods and apparatuses.
When combined with other recently refined parameter measurements, the Avogadro constant
derived from these new results has a value of NA = 6.022 140 76(12) × 1023 mol−1. The x-ray
crystal density method has thus achieved the target relative standard uncertainty of 2.0 × 10−8
necessary for the realization of the definition of the new kilogram.
Keywords: Avogadro constant, kilogram, Planck constant, enriched silicon, molar mass, lattice
parameter, interferometry
(Some figures may appear in colour only in the online journal)
1. Introduction the Planck constant h. Additionally, an accurate NA constant

is important because its value will be used to define the mole.
In 2011, the International Avogadro Coordination (IAC) pub- One of the principal issues with the 2011 determination
lished a comprehensive survey of the different measurements of NA [1] was the fact that the Si spheres used for the volume
contributing to the determination of the Avogadro constant and mass determinations were covered by a thin layer of
NA by counting the atoms in a 28Si-enriched single crystal [1, metallic contaminant, composed of Ni, Cu and Zn atoms. This
2]. This approach, called the x-ray-crystal-density (XRCD) contamination probably occurred during the polishing pro-
method, is one of the candidate methods for the realization of cedure due to a contamination of the slurry. The layer was
the new kilogram definition that is based on fixing the value of localized in metal silicide islands ‘floating’ on the silicon
core matrix [3]. As the optical constants of this layer were
Content from this work may be used under the terms of the
Creative Commons Attribution 3.0 licence. Any further
unknown, it was later removed by a FreckleTM etch with a
distribution of this work must maintain attribution to the author(s) and the title selectively high etching rate for silicides [4]. After etching,
of the work, journal citation and DOI. the sphere AVO28-S8b (‘b’ for the status after etching) was
0026-1394/15/020360+16$33.00 360 © 2015 BIPM & IOP Publishing Ltd Printed in the UK
Metrologia 52 (2015) 360 Y Azuma et al
remeasured at PTB and NA was redetermined. The consist- defects on the density ρ of the crystal is considered, the sim-
ency of the new result with the formerly obtained NA value plest way to implement measurement equation (1) is to cal-
was excellent [5]. Because the etching had degraded the shape culate the mass of an equivalent sphere, having the same core
of the spheres, they were then reworked at PTB, using a new volume and lattice parameter a measured by combined x-ray/
procedure to improve their sphericity [6]. In June 2013, the optical interferometry, but having Si atoms at all regular sites.
surface of the repolished sphere AVO28-S5c (‘c’ designating The concentrations of carbon, oxygen, boron and
the sphere after repolishing) was checked by x-ray fluores- vacancy-related defects have already been reported [1]. The
cence (XRF) spectrometry, revealing negligible amounts of concentration of nitrogen was additionally measured at PTB.
foreign metals. Moreover, no subsurface damage to the crystal INRIM also developed a method based on instrumental neu-
could be detected by high-resolution x-ray diffractometry [7] tron activation analysis (INAA) to evaluate the concentra-
when compared with strain free etched reference crystal sur- tions of various impurity elements. These results are given
faces. The roughness of the surface was below 0.2 nm, near in section 2.1.
the detection limit of the measurement. The topography of The amount-of-substance fractions of the Si isotopes in
the sphere was measured interferometrically, establishing that the crystal were measured independently by PTB, NMIJ and
the shape of the sphere was defined only by the orientation NIST using isotope dilution and a multicollector inductively
of the crystallographic axes. Maximum peak-to-valley (p-v) coupled plasma mass spectrometer (MC ICP-MS). Instead
deviations of the diameter were below 70 nm (see section 2.5). of using NaOH as solvent and diluent, tetramethylammo-
The second sphere, AVO28-S8, was then repolished using a nium hydroxide (TMAH) was used by these three institutes
slightly different polishing process in order to achieve a better to reduce the baseline level of the ion current measurements
roundness which produced a p-v value below 40 nm for the during the mass spectrometry. These results are given in
sphere diameters. section 2.2.
A new measurement campaign was then initiated using To measure the lattice parameter, INRIM upgraded a com-
these repolished spheres and these new results are presented bined x-ray/optical interferometer to further reduce the uncer-
in this paper together with the refinements to the measure- tainty. To demonstrate crystal homogeneity, NMIJ evaluated
ment methods and instrumentation that have occurred since the crystal perfection using strain topography, carried out by
the 2011 review. All of these improvements were necessary means of a self-referenced x-ray diffractometer at the Photon
to reach a total relative standard uncertainty of 20 × 10−9 for Factory of the High Energy Accelerator Research Organization
NA, an essential goal for the realization of the new kilogram (KEK, Japan). Detailed results are given in section 2.3.
definition. Sections 2.4–2.6 describe the measurement of the densities
of the two 28Si-enriched spheres. PTB and NMIJ character-
ized the composition, mass and thickness of the sphere sur-
2. Determination of the crystal quantities
face layers by XRF, x-ray reflectometry (XRR), and optical
spectral ellipsometry (SE). These results are given in sec-
The measurement of the Avogadro constant NA, using a sil-
tion 2.4. The sphere volumes were determined via diameter
icon crystal, is based upon the following equation,
measurements. NMIJ measured about a thousand diameters
NA = nM /(ρa3),
(1) for each sphere using an improved optical interferometer
with a flat etalon. PTB used a spherical Fizeau interferometer
where n is the number of atoms (8) per unit cell of a silicon which allowed about 105 diameters to be measured, resulting
crystal and ρ, M and a are its density, molar mass and lat- a complete topographical mapping of the spheres. Details are
tice parameter, respectively. Details on counting the atoms in given in section 2.5. Mass comparisons of the two spheres
a silicon crystal are given in [1, 2]. A 28Si-enriched silicon with Pt–Ir kilogram standards were carried out both in air
single crystal was grown primarily to reduce the uncertainty and under vacuum by the BIPM, NMIJ and PTB. In order to
of measuring the molar mass M. Two 1 kg spheres were manu- provide a better traceability to the international prototype of
factured from the crystal, and the density ρ of each sphere the kilogram (IPK), the BIPM revised the mass values for the
was determined from its mass and volume measurements. The BIPM calibrations following the Extraordinary Calibration
sphere surfaces were covered with oxide layers having a total Campaign against the IPK conducted in 2014 [8]. NMIJ and
thickness around 2 nm. In order to determine the density of the PTB also used the revised mass values of their Pt–Ir kilogram
crystal at the highest levels of accuracy, the surface of each standards that were reported by the BIPM in December 2014.
sphere needed to be chemically and physically characterized Details are given in section 2.6.
at an atomic scale so that the density of the crystal could be The final NA values obtained from these measurements
determined from the mass and volume data, excluding these are given in section 3 together with their uncertainty budget.
oxide layers. In this paper, these two parameters are desig- In section 4, the XRCD final result for NA is compared with
nated as ‘core mass’ and ‘core volume’, respectively. those from the watt balance experiments.
In a real crystal, the lattice spacing and density are affected
by impurity atoms and vacancies. For example, interstitial 2.1. Point defects
oxygen expands the lattice spacing and increases the unit cell
mass, and substitutional carbon shrinks the lattice spacing and The infrared (IR) absorption measurements of dissolved
decreases the unit cell mass. When the effect of these point carbon, oxygen and boron within the silicon crystal have
361
Table 1. A summary of the molar mass and amount-of-substance fraction determinations of the AVO28 crystal material. The uncertainties
(k = 1) in parentheses apply to the last respective digits. Note that measurements prior to 2013 were carried out using solutions of aqueous
NaOH while all subsequent measurements used aqueous TMAH.
NMI Diluent M/(g/mol) x(28Si)/(mol/mol) x(29Si)/(mol/mol) x(30Si)/(mol/mol) Ref.
PTB 2011 NaOH 27.976 970 27 (23) 0.999 957 50 (17) 0.000 041 21 (15) 0.000 001 29 (4) [12, 13]
NRC 2012 NaOH 27.976 968 39 (24) 0.999 958 79 (19) 0.000 040 54 (14) 0.000 000 67 (6) [14]
PTB 2014 TMAH 27.976 970 22 (17) 0.999 957 26 (17) 0.000 041 62 (17) 0.000 001 12 (6) [15]
NMIJ 2014 TMAH 27.976 970 09 (14) 0.999 957 63 (3) 0.000 041 20 (7) 0.000 001 18 (3) [16]
NIST 2014 TMAH 27.976 969 880 (41) 0.999 957 701 (41) 0.000 041 223 (41) 0.000 001 076 (88) [17]
PTB 2015 TMAH 27.976 970 13 (12) 0.999 957 50 (12) 0.000 041 38 (12) 0.000 001 121 (14) [18, this paper]
already been reported in [1]. The gradients in the impurity solutions for inductively coupled plasma mass spectrometry
concentrations are caused by the float zone technique used to (ICP-MS). The studies all noted that aqueous NaOH could
purify and grow the single crystal. Additionally, the nitrogen create subtle but significant biases to Si isotope amount ratio
present in the AVO28 crystal was determined by infrared measurements. For example, the presence of aqueous NaOH
measurements using the method developed by Itoh et al as sample diluent causes ion scattering in the detector region
[9]. The average content of nitrogen in the spheres amounts of an MC ICP-MS. This charged background elevates the 30Si
to 0.17(10) × 1014 cm−3 and 1.38(30) × 1014 cm−3 for the baseline in particular [22]. Normally, this is not a problem
AVO28-S5 and AVO28-S8 spheres, respectively. This yields when the Si isotope sample signals are more than a few hun-
a mass deficit for the spheres of −0.2(1) µg and −1.4(3) µg, dred mV. However, when the sample signals for 30Si+ and
29 +
respectively (see section 2.6). Si are very low (≤1 mV), such elevated baseline signals
INRIM has also developed a method based on INAA giving could lead to an overcorrection of measured sample signals.
direct evidence of the crystal purity with respect to a very This is especially problematic for the 30Si+ sample signal in
large number of elements. Test measurements carried out at the AVO28 crystals and could lead to a significant biasing
the TRIGA Mark II reactor at the University of Pavia (with a of the x(29Si)/x(30Si) ratio. This effect is amplified when the
thermal neutron flux of 6 × 1012 cm−2 s−1) included fifty-nine NaOH concentration of the matrix solution is increased [22].
elements and achieved a detection limit of less than 1 ng g−1 Careful analysis of the published data using NaOH as
for thirty-five elements [10, 11]. Two samples were cut from solvent and matrix diluent suggests that this effect, together
the AVO28 boule and another purity check is planned in with possible memory carry-over, were contributing factors
early 2015, using the OPAL reactor of the Australian Nuclear that gave rise to the discrepancy between the 2011 molar
Science and Technology Organisation (with a thermal neu- mass measurements reported by PTB [12, 13] and the 2012
tron flux of 20 × 1013 cm−2 s−1). This analysis is expected to results published by NRC [14] (see table 1). When measuring
include sixty-four elements and should also reach a detection the AVO28 samples, PTB reported using a mass fraction of
limit of less than 1 ng g−1 for more than forty elements. 1 mg g−1 NaOH as sample diluent while NRC reported using
24 mg g−1 NaOH. The extremely high levels of NaOH in the
NRC samples magnified the 30Si signal of the blank, causing
2.2. Molar Mass
the very low 30Si signals of the AVO28 material (which were
Following the 2011 report on the status of molar mass mea- typically ≤1 mV) to be seriously over-corrected. The net effect
surements for the Avogadro constant [12, 13], a number of of this bias was to produce the artificially high x(29Si)/x(30Si)
additional investigations have been published, providing new ratios reported in [14]. A more detailed analysis of the causes
molar mass data on additional crystals from the AVO28 boule of the biasing of the NRC x(29Si)/x(30Si) ratios is underway
[14–17]. In addition, a detailed molar mass and amount-of- [20].
substance fraction homogeneity study undertaken by the PTB The NRC had also postulated that the PTB samples (and
produced data on an additional 14 crystals. The details of this by implication, the NMIJ and NIST samples) had undergone
homogeneity study will be published elsewhere [18]; how- direct contamination by natural silicon [14]. This possibility
ever, the average molar mass result for the 14 new crystals was ruled out when the absolute mass fraction of silicon in
is reported in table 1 while the individual molar masses are the NaOH material used for the PTB analyses was found to
reported in table 2. Table 1 also lists all the average molar be a factor of ten lower than the concentration necessary to
mass and amount-of-substance fraction results on the AVO28 explain the PTB-NRC molar mass differences. Given that
crystal material published by national metrology institutes these points of contention involve the use of NaOH as solvent
(NMIs) to date. and diluent and have not yet been fully resolved, this study
Subsequent to the original development and application of has excluded all molar mass data reported using this particular
the virtual-element isotope dilution-mass spectrometric ((VE) solvent. Instead, the abundant molar mass data acquired using
IDMS) approach to determining an accurate molar mass of TMAH as solvent and diluent are used to deduce the molar
the highly 28Si-enriched AVO28 boule [19], several studies mass results presented in this paper (table 2).
[20, 21] have examined the analytical problems of Si isotope The use of aqueous TMAH solutions, as proposed and car-
amount ratio measurements when aqueous sodium hydroxide ried out for the first time at NIST [17], has many advantages.
(NaOH) is used as the diluent or solvent to prepare silicon The extreme enrichment of the 28Si isotope in the AVO28
362
Table 2. Compilation of the molar masses measured on 24 indi- accuracy of these measurements was provided by a recent
vidual crystal samples that used TMAH as solvent and diluent (see internal study at PTB. A small disc of AVO28 material was
text for explanation). Uncertainties (k = 1) in parentheses apply to
analyzed by glow discharge mass spectrometry (GDMS) [21].
the last two digits. Samples are given in the order of their longitudi-
nal position in the original crystal. Sample association with sphere This complementary study produced numerical values for the
AVO28-S5 or sphere AVO28-S8 is indicated by (S5) and (S8) in the molar mass and the amount-of-substance fractions in agree-
Sample ID column. Note that the 5 crystals from part 7 (PTB-7-1 ment with those reported by PTB, NMIJ and NIST within the
(S5, S8) to PTB-7-5 (S5, S8)) are included in both calculations of limits of uncertainty, although the associated uncertainties of
the averages for spheres S5 and S8 because this part lies between
this measurement were larger.
the two spheres. The part 7 values have therefore not been arbitrar-
ily associated with one or the other sphere. All molar mass calcula- INRIM proposed to measure the amount-of-substance frac-
tions used atomic mass values reported in the AME2012 atomic tion of 30Si by neutron activation [23]. A sample of the AVO28
mass evaluation [25]. crystal was analyzed using the TRIGA Mark II reactor. The
NMIa Sample ID M / (g/mol) result, x(30Si) = 0.000 001 043(19) mol mol−1 [24], further sup-
ports the amount-of-substance fraction findings from the PTB,
PTB 2015 PTB-4-1 (S5) 27.976 970 29(14) NMIJ and NIST measurements. This measurement is being
PTB 2015 PTB-4-2 (S5) 27.976 970 25(13)
PTB 2015 PTB-4-3 (S5) 27.976 970 00(11) repeated in Australia using the OPAL reactor.
PTB 2015 PTB-4-4 (S5) 27.976 970 28(13) In 2014, with the specific backing of the IAC, PTB initi-
PTB 2015 PTB-4-5 (S5) 27.976 970 13(13) ated an investigation of the variability of the molar mass and
PTB 2014 PTB-5 (S5) 27.976 970 24(17) the isotopic composition across the AVO28 boule. The experi-
NIST 2014 NIST-5-1 (S5) 27.976 969 842(93)
NIST 2014 NIST-5-2 (S5) 27.976 970 141(71) mental design for this study called for 5 adjacent radial samples
NMIJ 2014 NMIJ-5-1 (S5) 27.976 970 10(22) (each with a mass of approximately 300 mg) to be taken from
NMIJ 2014 NMIJ-5-2 (S5) 27.976 970 05(8) three distinct longitudinal positions (parts 4, 7, and 9) of the
PTB 2015 PTB-7-1 (S5, S8) 27.976 970 06(12)
PTB 2015 PTB-7-2 (S5, S8) 27.976 970 09(13) original crystal ingot (figure 1 in [12]) and analyzed for their
PTB 2015 PTB-7-3 (S5, S8) 27.976 969 94(12) molar mass. A detailed description of this study will be pub-
PTB 2015 PTB-7-4 (S5, S8) 27.976 969 96(12) lished elsewhere [18]. However, given that this homogeneity
PTB 2015 PTB-7-5 (S5, S8) 27.976 970 00(12)
PTB 2014 PTB-8 (S8) 27.976 970 20(17) study incorporated most of the state-of-the art improvements
NIST 2014 NIST-8-1 (S8) 27.976 969 745(57) in the measurement of Si isotope-amount-ratios developed
NIST 2014 NIST-8-2 (S8) 27.976 969 797(90) since 2011, the average and individual molar masses as well
NMIJ 2014 NMIJ-8-1 (S8) 27.976 970 14(21)
NMIJ 2014 NMIJ-8-2 (S8) 27.976 970 08(21)
as the Si isotope amount fractions derived from these data are
PTB 2015 PTB-9-1 (S8) 27.976 970 08(11) presented in tables 1 and 2. The (VE) IDMS approach [19]
PTB 2015 PTB-9-2 (S8) 27.976 970 26(11) was used for measuring the molar mass, with the main experi-
PTB 2015 PTB-9-3 (S8) 27.976 970 20(11) mental details given in [13, 15]. All measurements have been
PTB 2015 PTB-9-5 (S8) 27.976 970 33(11)
made on a commercial MC ICP-MS (Neptune™, Thermo
a
NIST 2014 [17]; NMIJ 2014 [16]; PTB 2014 [15]; PTB 2015 [18, this Fisher Scientific). The calibration and sample runs were
paper]. separated as originally suggested by colleagues from NRC,
saving time and material without any significant loss in accu-
material leaves very little 29Si and almost no 30Si to be ionized racy. TMAH was used as the solvent and matrix diluent. The
and detected. With TMAH as the matrix diluent, the large flux marked increase of the Si+ ion signals resulting from the use
of Na+ ions was no longer present in the plasma source. This of TMAH enabled all the Si isotope data to be measured with
had the positive effect of increasing the intensity of the very Faraday detectors (1011 Ω resistors). To date, five crystals from
small 29Si+ and 30Si+ signals by nearly an order of magnitude. part 4, five crystals from part 7 and four crystals from part 9
Ancillary negative effects, like the ion scattering produced of the Avogadro boule have been measured (table 2). Because
when NaOH was used, were no longer observed. Additionally, the crystals from parts 4 and 7 bracket part 5 (the source loca-
the progressive clogging of the skimmer and sampler cone ori- tion of sphere AVO28-S5), and parts 7 and 9 bracket part 8
fices during sample analyses was strongly reduced or absent (the source location of sphere AVO28-S8), the new PTB data
when using TMAH. This has led to stable silicon ion beam listed in tables 1 and 2 are fully representative of the spheres
intensities lasting several days or more. The TMAH blanks AVO28-S5 and AVO28-S8 [18]. When combined together
measured during the analysis of AVO28 materials also showed with the data given in [15], these measurements represent the
a more natural silicon isotopic composition when compared PTB contribution that is combined with the NMIJ and NIST
with similar NaOH blanks. This attribute also suggests that data for the calculation of a new NA.
there has been a real time mitigation of any memory-carryover NMIJ also carried out molar mass measurements of
issues [17]. All of these factors arising from the use of TMAH AVO28 crystals using a commercial MC ICP-MS (Neptune™,
have served to greatly improve the quality of the molar mass Thermo Fisher Scientific) with TMAH as the solvent. To cor-
measurements by decreasing the possibility of measurement rect for the mass bias arising in the plasma and sample source
biasing, particularly from signal suppression causing an over- areas, three different blends, as described by NRC [14] were
correction of the detected 30Si and 29Si sample signals. employed. Four AVO28 samples, identified as 5B1.2.2.1,
While only molar mass data taken with TMAH as solvent 5B1.2.2.2, 8B3.2.2.1 and 8B3.2.2.5 were analyzed. All of
and diluent are used in this study to compute the average molar the AVO28 crystals had been cut from the single crystal pro-
mass of the AVO28 boule, an additional blunder check on the duced by the IAC specifically to determine the Avogadro
363
constant. The axial positions of the 5B1 series and the 8B3 Table 3. Summary of the measured molar masses of spheres
series are respectively 275 mm and 414 mm from the outer AVO28-S5 and AVO28-S8 as well as the average of all molar mass-
es combined. Uncertainties (k = 1) in parentheses apply to the last
surface of the boule; their radial distance from the center of
two digits. The final column lists the number of crystals (n) used
the 28Si-enriched crystal ingot is approximately 40 mm. The to compute the respective averages. Note that the average molar
average molar mass of the four AVO28 crystals was deter- masses for S5 and S8 each include the 5 molar mass measurements
mined to be 27.976 970 09(14) g mol−1, with a relative standard made on the 5 crystals from part 7, as noted in table 2. The average
uncertainty of 5.2 × 10−9 (table 1). molar mass for the AVO28 boule is the arithmetic mean of all 24
molar mass measurements listed in table 2.
NIST analyzed 4 different crystals of the AVO28 silicon,
two proximal to AVO28-S5 (5B2.1.1.3, 5B1.1.1.1) and two Sample M/(g/mol) urel/10–9 n
taken near AVO28-S8 (8A4.1.1.3, 8B1.1.1.1) [17]. All sil-
Average S5 27.976 970 09(09) 3.1 15
icon samples were dissolved and diluted using TMAH and Average S8 27.976 970 06(15) 5.4 14
also analyzed on a commercial MC ICP-MS (Neptune™, AVO28 boule 27.976 970 09(15) 5.4 24
Thermo Fisher Scientific) in high resolution mode. To cor-
rect for measurement mass bias, two independent sets of
calibration solutions were created. The AVO28 samples may come from either the external reproducibility of the
were always run as pairs, with one from part 5 and the measurement due to tiny yet uncontrolled blank variations,
other from part 8. These samples were always run together the signal detection itself, the reproducibility of the sample
with the calibration solutions. This particular experimental preparation, a tiny but detectable variability in the Si isotopic
design optimized the detection of any small but measurable composition, or an as-yet-unknown additional influence. In
heterogeneities in the silicon isotope amount ratios of the order to render these data internally consistent, an additional
different samples. The calibration corrections followed the uncertainty contribution with an expectation value of zero was
approach originally developed by PTB [19]. One sample set added to the overall uncertainty of the mean, by adapting the
was measured using only one calibration solution, while the recommendations from [28]. The criterion for determining the
other was analyzed using both solutions. The average molar value of this additional element of uncertainty, udisp, is the nor-
mass of the four AVO28 crystals measured by NIST was malized error [29]. Following [30], udisp was adjusted so that
27.976 969 880(41) g mol−1, with a relative standard uncer- all normalized errors were equal to or less than 1.
tainty of 1.5 × 10−9 [17] (table 1). Combining the 24 results from PTB, NMIJ, and NIST
In summary, the basis for the average molar mass of the (table 2), the average molar mass for AVO28 is calculated
AVO28 boule, with its associated uncertainty, was derived to be 27.976 970 09(15) g mol−1, with an associated relative
from molar mass measurements using TMAH as solvent and standard uncertainty of 5.4 × 10−9. These data clearly suggest
diluent. The individual molar masses of five crystal samples that, at the present level of measurement accuracy and preci-
from part 4 (PTB [18, this paper]), one crystal from part 5 sion, there are no longitudinal or axial molar mass gradients
(PTB [15]), two crystals from part 5 (NMIJ [16]), two crys- within the AVO28 boule. The individual molar masses of the
tals from part 5 (NIST [17]), five crystals from part 7 (PTB spheres AVO28-S5 and AVO28-S8 were calculated from the
[18, this paper]), one crystal from part 8 (PTB [15]), two crys- PTB, NMIJ, and NIST results listed in table 2 and labelled
tals from part 8 (NMIJ [16]), two crystals from part 8 (NIST ‘S5’ and ‘S8’, respectively. Table 3 summarizes the molar
[17]) and four crystals from part 9 (PTB [18, this paper]) were mass data for each separate sphere, as well as the combined
used. These data allow a robust molar mass to be calculated data. The average individual molar mass data for the two
for updating the Avogadro constant that is based on 24 dif- spheres are indistinguishable, thus the average AVO28 boule
ferent sample crystals spread longitudinally across the AVO28 value derived by combining all results is the number that is
boule and measured by three different NMIs, each employing used in the calculation of the new NA reported in this study.
a different experimental approach (table 2).
The uncertainty weighted mean (UWM) of all 24 results
2.3. Lattice parameter
was determined to be 27.976 970 030(38) g mol−1 with an asso-
ciated uncertainty expanded by the Birge ratio (σB ≈ 1.70) cal- INRIM’s combined x-ray/optical interferometer, used to
culated according to [26]. The arithmetic mean including all determine the {2 20 } lattice-plane spacing of the enriched
24 results was 27.976 970 09 g mol−1 with a standard deviation silicon crystals, was upgraded and measurements repeated to
of the mean of 0.000 000 03 g mol−1. The UWM would nor- either confirm the previous result [31] and its uncertainty or to
mally be the estimator of choice for combining these data, as identify possible errors.
they show a relatively large spread in their associated uncer- First, a 532 nm frequency-doubled Nd:YAG laser was sub-
tainties. However, a data consistency check (chi-squared test) stituted for the 633 nm diode laser. The pressure in the vacuum
recommended in [27] was carried out, yielding a Χobs 2 ≈ 66
chamber was also reduced to less than 0.04 Pa. This made any
which is larger than the 95th percentile of Χ0.05,23 ≈ 35, with
2 correction for the refractive index of the residual gas essen-
23 degrees of freedom. This data set is therefore not entirely tially inconsequential and ensured a calibration of the optical
internally consistent. This suggests that one or more of the interferometer with negligible uncertainty.
contributions to the overall uncertainty have not been con- Next, the delivery, collimation, phase-modulation, and
sidered fully. Possible sources for the observed inconsistency pointing systems of the laser beam were rebuilt to conform
364
to the new wavelength. The beam divergence was reduced, Accelerator Research Organization (KEK, Japan) [34, 35].
thereby halving the correction for diffraction effects. To have The analyser crystal of the x-ray interferometer and a sample,
real-time control of the beam pointing, a home-made tel- identified by the 4.R1 code and cut from the seed end of the
escope was placed at the interferometer output port; to ensure crystal, show a smooth and homogeneous distribution of lat-
stability, it was clamped on the same base plate of the x-ray/ tice spacing values. The standard deviation of the observed
optical interferometer. variations is 4.9 × 10−9 for the 4.R1 crystal. This value is
Then, a plate beam-splitter was substituted for the previously consistent with what is observed by x-ray interferometry,
used cube beam-splitter. This ensured that the length difference 1.5 × 10−9 d220, with a strain smoothing over (2 × 4) mm2
of the transmitted and reflected light paths was insensitive to areas. By way of contrast, a tail-end sample, identified by
any beam translations and rotations. The fixed components the 9.R1 code, shows a 2D swirl-like pattern and a greater
of the interferometer—beam splitter, quarter-wave plates, and variability in its lattice spacing values. This observation is
fixed mirror—were replaced and assembled anew. consistent with the segregation of impurities into the tail of a
In order to produce parallel interfering beams, the inter- crystal purified by the float-zone process. Therefore, the tail
ferometer fixed-components were cemented onto a glass sample can be expected to be more contaminated and to dis-
plate supported by three piezoelectric actuators. A new power play significant variations in its lattice spacing [35].
supply, producing sub part-per-million noise and stability, was The mean lattice parameter of each sphere,
designed and built to eliminate instabilities between the x-ray
and optical interferometers. (3) i (
a(S ) = 1+ ∑ βi ΔNi a(XINT), )
PTB also found that the surfaces of the x-ray interferom-
eter crystals were contaminated with Cu, Fe, Zn, Pb, and Ca, was calculated by taking account of the different point-defect
caused by the wet etch used to remove any residual surface concentrations in the spheres and the interferometer. In equa-
stress after crystal machining. These contaminants were tion (3), S is sphere AVO28-S5 or AVO28-S8, and XINT is
removed by cleaning the crystals in aqueous solutions of HF the x-ray interferometer [67]. The subscript i refers to point
and (NH4)2S2O8. defects, where βi is the strain coefficient [12, 36] and ΔNi is
The final upgrade focussed on more accurate temperature the concentration difference of the point-defect i between the
measurements. The fixed point cells of INRIM and PTB were sphere and the interferometer. Unlike the previous determi-
compared to establish their temperature difference and to link nation, where only carbon, oxygen, and boron contamination
the INRIM and PTB extrapolations of the lattice parameter were considered, the newly measured gradient of nitrogen
and sphere volume to 20 °C. It was not yet possible to verify concentration was also taken into account.
if the thermometer readings at 20 °C were identical to within
the same uncertainty of the fixed-point cell temperatures; this
non-uniqueness error was cautiously set to 0.1 mK [32]. 2.4. Surface
To make a reassessment of the measured value and its uncer- The surface layer of the Si sphere must be accurately charac-
tainty, all the systematic effects were scrutinized and reevalu- terized and measured to refine the correction values required
ated with a view to reducing the overall uncertainty and to for the mass and volume determinations of the spheres. The
confirm that the intended goals could be met. Seven surveys of basic requirements and details of the methods used for surface
the lattice spacing (with the interferometer crystals aligned as characterisation are described in [3]. The next section will out-
they were originally in the boule and in a reversed arrangement) line the characterisation techniques, highlighting any changes
were carried out from February to June 2014. These surveys to the procedures described in [3].
were made over 0.95 mm crystal segments centered in 48 dif- To completely analyze the surface of a Si sphere, a rapid
ferent positions. At each position, x-ray counts were recorded in measurement method is required. An ideal technique for
eight different pixels of the interference pattern (11.2 mm high) investigating a SiO2 layer on a Si substrate is SE which
and then processed using linear regression to obtain 48 values combines fast measurements (a thickness measurement at a
along a line that was the continuation of the laser beam and, single point in less than 10 s) and high precision (repeatability
therefore, unaffected by Abbe errors. The final average is around 10 pm). Using SE, the automatic measurement of a
d220(XINT) = a(XINT)/ √8 = 192.014 711 98(34) pm (2) spherical surface with 2600 data points can be completed in
12 h. Unfortunately, SE has two shortcomings: its accuracy
where XINT is the x-ray interferometer. At tITS-90 = 20 °C and (approximately 1 nm) is insufficient for the present applica-
p = 0 Pa, equation (2) expresses the mean lattice spacing value tion and, as an inverse method, a model of the surface layers
along a strip 46 mm long, orthogonal to the crystal axis, and must be generated, which then becomes part of the data refine-
at a distance of 306 mm from the seed. Details about the mea- ment process. To overcome these limitations, a calibration
surements and the data analysis together with a discussion of of the ellipsometric measurement process must be carried
the full error budget are in [33]. out. Surface characterisation thus becomes a two-step pro-
As regards the crystal perfection, the NMIJ carried out cess. First, the ellipsometer is calibrated based on reference
topographic measurements of the lattice strain in several sam- methods such as XRR at NMIJ and the combination of XRR
ples from the AVO28 crystal by means of a self-referenced and XRF analysis at PTB. The second step, the mapping of the
lattice comparator at the Photon Factory of the High Energy surface using ellipsometry, can then proceed.
365
Figure 1. The surface layers model used for the characterisation of the AVO28 spheres.
Subsequent to the last determination of NA, the 28Si total layer thickness was well below 2 nm, no oscillations with
spheres, AVO28-S5 and AVO28-S8, were repolished. The multiple periods could be observed and the thickness deter-
surface layer was therefore modified and, more importantly, mination was therefore based on previously obtained optical
simplified when compared to the measurements made in 2009 constants, which lead to increased uncertainties.
and 2010. The surficial metallic contaminants (Cu, Ni and Zn) For the calibration of the ellipsometric measurements at
were removed and were therefore no longer part of the surface PTB, the OL thickness of a few well defined points on the
layer model. For the new NA determination, the surface layer sphere were determined using the XRR-based XRF measure-
model illustrated in figure 1 is now applicable. ments. These points could be located using three different
The measurements for the surface layer characterisa- markings (cross, ‘T’ and triangle) to an accuracy better than
tions were performed at NMIJ and PTB. NMIJ has recently 0.5°. The calibration points were then used for the calibra-
installed a new spectral ellipsometer, which enables the auto- tion of the PTB spectral ellipsometer during the mapping
matic measurement of an entire sphere surface, accumulating of the sphere surface (in effect, an ‘in-vivo’ calibration). By
a large number of data points. PTB used the identical equip- adjusting the alignment of the sphere, a calibration point could
ment as for the 2010 Avogadro constant determination, which be included in each measurement of a great circle. This meant
is also capable of automatic surface coverage producing sev- that a short term stability of the instrument of less than 30 min
eral thousand data points. was required, which could be expressed as an uncertainty con-
NMIJ and PTB used different approaches for the calibra- tribution of less than 20 pm. With this ‘in-vivo’ calibration,
tion of their spectral ellipsometers. The general methodology requirements for the simulation model for the refinement of
has already been described in detail in [3]. At the synchrotron the ellipsometric data were dramatically simplified because
radiation laboratory of PTB at BESSY II [37], direct measure- the influence of the CWL and CL are inherently included
ments of the oxide layer (OL) and carbonaceous layer (CL) in the calibration constant C. Only the linearity of the ellip-
thicknesses were made using XRF. The calibration of the XRF someter and the homogeneity of the CWL and the CL were
thickness measurement was done using the ratio of the O-K required for the simulation model.
peak to the Si-L peak of SiO2 reference samples, whose thick- The calibration measurements of the spheres at the PTB
ness had been determined by XRR at different photon ener- synchrotron radiation laboratory were carried out in November
gies in the vicinity of the oxygen absorption edge at 529 eV. 2013 and January 2014 (AVO28-S5c) and in January 2014 and
Because these measurements were made in a vacuum, only July 2014 (AVO28-S8c), respectively. The ellipsometric map-
the chemisorbed water layer (CWL) was present. For the OL ping of sphere AVO28-S5c was done with 5184 data points,
thickness determination, it was assumed that all measured O while sphere AVO28-S8c had 15 552 data points. The results
atoms were in the SiO2 layer. To correct for the CWL, the of these measurements are listed in tables 4(a) and (b).
measured thicknesses were adjusted using the mass deposi- NMIJ used a different approach for the calibration of their
tion of the chemisorbed water given in [38] to derive the OL ellipsometer. The oxide layer thickness on the two spheres was
thickness. measured by a spectroscopic ellipsometer at NMIJ. The reli-
For the determination of the CL, the ratio of the C-K peak ability of the spectroscopic ellipsometer was checked by using
to the Si-L peak was used and compared to a reference carbon SiO2 layers on Si wafers with thicknesses certified by XRR
layer whose thickness was again determined by XRR. To [3]. In the previous paper [3], the oxide layer thickness was
derive a thickness for the CL, a mass density of 1.1 g cm−3 measured at only 20 points on the surface of the spheres. To
was assumed, this being about half of the carbon bulk den- increase the number of measurement points for improving the
sity. This thickness was required for the NMIJ ellipsometric reliability of the measurement, a new spectroscopic ellipsom-
measurements as input data for the surface layer model used eter equipped with an automatic sphere rotation system was
in their data refinement. This thickness was also used in the developed [40]. This new instrument was a spectral ellipsom-
volume determinations. Since the chemical structure of the eter with a rotating polarizer. Its spectral bandwidth ranged
CL is highly unpredictable and unclarified, a generous uncer- from 250 nm to 990 nm. The Si sphere was placed on two
tainty was applied to this measurement. rollers made of PEEK (Polyether ether ketone) and could be
In addition to XRF, XRR was performed directly on the rotated by the automatic sphere rotation system around both
spheres in the vicinity of the O-K edge as described in [3, 39]. its vertical and horizontal axes. The sphere rotation system
The thicknesses determined for the involved layers were in was integrated into the spectroscopic ellipsometer, thereby
good agreement with the values from XRF. However, as the enabling the automatic mapping of the oxide layer thickness
366
Table 4. (a) The thickness of the surface layer and its constituent sub-layer components, dCL, dCWL, dPWL and dOL. (b) Mass of the surface
layer and its constituent sub-layers. See figure 1 for the key to the subscript abbreviations.
(a) Thickness
Date of
Sphere Lab dCL/nma dCWL/nmb dPWL/nmc dOL/nm measurement (OL) dSL/nmd
AVO28-S5c PTB 0.60 (18) 0.28 (8) — 0.91 (14) Jan. 2014 1.79 (24)
NMIJ 0.60 (18) 0.28 (8) 0.39 (9) 0.76 (27) July and Sep. 2014 1.64 (33)
average 0.60 (18) 0.28 (8) 0.88 (12) 1.76 (23)
AVO28-S8c PTB 0.49 (16) 0.28 (8) — 1.17 (13) July 2014 1.94 (22)
NMIJ 0.49 (16) 0.28 (8) 0.43 (9) 0.64 (25) June 2014 1.41 (31)
average 0.49 (16) 0.28 (8) 1.06 (22)e 1.83 (28)
(b) Mass
Date of
Sphere Lab mCL/µg mCWL/µg dPWL/nm mOL/µg measurement (OL) mSL/µgf
AVO28-S5c PTB 16.6 (5.7) 7.7 (2.2) — 55.2 (8.9) Jan. 2014 79.5 (10.9)
NMIJ 16.6 (5.7) 7.7 (2.2) 10.8 (2.5) 46.1 (16.5) July and Sep. 2014 70.4 (17.7)
average 16.6 (5.7) 7.7 (2.2) 53.4 (7.7) 77.7 (10.0)
AVO28-S8c PTB 13.5 (5.2) 7.7 (2.2) — 71.0 (8.5) July 2014 92.2 (10.2)
NMIJ 13.5 (5.2) 7.7 (2.2) 11.9 (2.7) 38.9 (15.3) June 2014 60.0 (16.3)
average 13.5 (5.2) 7.7 (2.2) 64.3 (13.7) 85.5 (14.8)
a
The thickness of the CL measured by XRF at PTB was based on the assumption that the density of the CL was 1.1 g cm−3. The uncertainty of this thickness
was estimated using the surface analysis results from the previous measurement [3].
b
The dCWL was calculated from data reported by Mizushima [43].
c
The data for the dPWL came from comparison weighings of the two spheres in nitrogen gas, at a pressure of ca.1200 Pa, and in water vapour, at a pressure of
ca. 1200 Pa [42]. The density of the PWL was assumed to be 1.0 g cm−3.
d
This value does not include the thickness of the PWL.
e
The Birge ratio of the thickness values of the oxide layer of AVO28-S8c is 1.8. Therefore, the uncertainty of the weighted mean was multiplied by 1.8. A
possible reason for the difference in the oxide layer determinations may be that NMIJ used the CL thickness value of PTB and the surface cleaning status
were not identical at PTB and NMIJ.
f
The mass of the PWL was not included in this value.
over the entire surface of the sphere. The measurements were estimated to be 0.39(9) nm and 0.43(9) nm for AVO28-S5c
based on 2594 points, regularly distributed over the entire and AVO28-S8c, respectively. These results were obtained
sphere surface. The sphere surface was subdivided into small at NMIJ from the comparison weighings in nitrogen gas of
cells of equal area and the measurement points were distrib- about 1200 Pa and in water vapour of about 1200 Pa [42] for
uted uniformly to each cell. the two spheres using a stainless steel weight as a reference.
The oxide layer thickness of AVO28-S8c was measured in The amount of PWL on the stainless steel weight was deter-
June 2014 and that of AVO28-S5c was measured in July and mined in advance by comparison weighings using artifacts
September 2014. Although the 2594 points were distributed with large surface area difference. The thickness of the CWL
almost uniformly over the sphere surface, the uniformity was was estimated from the published value of Mizushima [43] to
not perfect. To estimate the effect of this non-uniform distri- be 0.28(8) nm. The thickness of the CL was measured by XRF
bution of the measurement points on the average value of the at PTB to be 0.60(18) nm and 0.49(16) nm for AVO28-S5c and
oxide layer thickness, the measurements at the 2594 points AVO28-S8c, respectively, on the assumption that the density
were repeated 3 times. Between each set of measurements, of the CL is 1.1 g cm−3.
the sphere was rotated to distribute 7782 (=2594 × 3) points To evaluate the OL thickness, the measured ellipsometric
as uniformly as possible. Before each set of measurements, parameters were fitted by the aforementioned four-layer
the spheres were washed using the same procedure used for model, fixing all the sub-layer parameters, except the thick-
the international comparisons of the mass and diameter of a ness of the OL. The evaluated OL thicknesses were 0.76
Si sphere in the International Avogadro Coordination project (27) nm and 0.64(25) nm for AVO28-S5c and AVO28-S8c,
[38, 41]. The standard deviation of the mean OL thickness for respectively. The thickness of each layer is summarised in
3 sets of the measurements was less than 0.1 nm, showing the table 4(a). The dominant uncertainty source for the OL thick-
uniformity of the distribution of the 2594 points and the reli- ness determination is the thickness of the CL.
ability of the measurement system.
The ellipsometric data were analyzed at NMIJ based on
2.5. Volume
the surface model with four layers to evaluate the SiO2 thick-
ness. Since the ellipsometric measurement was performed Optical interferometers were utilised to determine the volume
in air, the model consists of a carbonaceous layer (CL), a of the two spheres by measuring the diameters of the spheres
physisorbed water layer (PWL), a chemisorbed water layer and calculating their volumes. Although the elementary
(CWL) and an oxide layer (OL). The thickness of PWL was dimensional measurements were based on the same principle,
367
different types of interferometers with different optical set- Table 5. Uncertainty budgets for apparent volume measurements of
ups were used. A precise measurement typically takes advan- the 28Si spheres at NMIJ [44].
tage of a differential approach. Thus, the measurement of (u(V)/V) /10–9
a sphere proceeds in two steps: first is the measurement of
Avo28-S5c Avo28-S8c
the dimensions of a stable optical etalon, D. The second step
then involves the insertion of a sphere into the etalon and the Interferogram analysis 10.3 10.3
measurement of the gaps between the sphere and the etalon, Temperature 4.8 4.8
Diffraction effect 16 16
d1 and d2. The diameter of the sphere is calculated from the Standard deviation of 3.3 9.6
difference of these measurements d = D − d1 − d2. To fully the mean volume
characterise a sphere, the diameter is measured in many dif- Total 20 22
ferent directions.
NMIJ measured 1450 and 870 diameters of AVO28-S5c
and AVO28-S8c, respectively, using an improved optical for AVO28-S5c and AVO28-S8c, respectively. The rela-
interferometer with a flat etalon [40]. The 28Si sphere was tive standard uncertainties of the volume measurement are
placed between the two flat etalon plates, and d1, d2 and D 2.0 × 10−8 and 2.2 × 10−8 for AVO28-S5c and AVO28-S8c,
were measured by phase shifting interferometry with optical respectively. Table 5 shows the uncertainty budget for the
frequency tuning. The main improvements from the previous determination of the apparent volumes. In the previous work,
work [44] are summarized below. the volumes of the two spheres were determined by NMIJ with
Firstly, the geometrical shapes of the optical components relative standard uncertainties of 5.0 × 10−8 and 4.4 × 10−8
were optimized. The largest uncertainty source in the previous for AVO28-S5 and AVO28-S8, respectively [44]. Because of
volume measurement at NMIJ was the analysis of the interfer- the improvements described above, the uncertainty contribu-
ence fringes [44]. An analysis using the ray-tracing method tions from the interferogram analysis and the random com-
showed that a possible cause for the uncertainty in the analysis ponent were decreased, resulting in the significant reduction
of the interference fringes is the multiple reflection of the beam in the uncertainty of the volume measurement. The dominant
between the tilted surface of the etalon and the sphere surface. uncertainty source at present is the correction of the diffrac-
An increased tilt of the etalon surface was therefore expected to tion effect on the diameter measurement [45]. The value of
reduce the effect of the multiple reflections. Based on this anal- this correction was estimated to be 0.45(50) nm.
ysis, a new etalon with a larger wedge angle was installed [40]. The PTB interferometer was based optically on a com-
Secondly, the uniformity of the diameter measurement pletely spherical geometry [46]. This means that the refer-
directions was improved. One of the major uncertainty ence faces were spherical, so that the etalon forms opposing
sources in the previous volume measurement was the experi- segments (caps) which surround the sphere. Furthermore,
mental standard deviation of the mean volume. The diameter the illuminating light wave was converted by a set of objec-
measurement directions were based on 70 directions and the tives into a focused beam, so that these conical rays would
non-uniform distribution of these 70 directions was estimated hit the reference face and the sphere perpendicularly. Thus,
to increase the experimental standard deviation of the mean the relationship d = D − d1 − d2 would be true for all points
volume [44]. To decrease this uncertainty, a strategy to dis- θ, φ within the field of view (covering 60°). This enabled
tribute the measurement directions as uniformly as possible the acquisition of high resolution topographies of the sphere.
was developed [40]. The new distribution consists of 145 Each sphere was provided with three different marks (fol-
directions. The sphere surface is subdivided into small cells lowing the orientation of the crystallographic axes) so that a
of equal area and the measurement points are distributed uni- sphere could be initially oriented by one mark with the help
formly to each cell. of the interferometer camera. Subsequently it was oriented
Thirdly, the optical frequency standard was upgraded. by means of the high-resolution encoder equipment of the
An iodine-stabilized He-Ne (HeNe/I2) laser was used as the sphere positioning motors. Each measurement, at all orien-
optical frequency standard in the NMIJ interferometer [44]. tations of the sphere, could then be related to its absolute
However, the HeNe/I2 laser was highly sensitive to acoustic position on the sphere [47]. The volume of this ‘sphere’ (see
noise. An optical frequency comb was therefore employed as figure 2) was calculated by considering the areal weighting
the standard to obtain more precise and reliable volume meas- of each measuring point [48]. To eliminate any possible sta-
urements [40]. This light source system is much more robust bility problems of the etalon and to monitor the stability of
than the HeNe/I2 laser and can be operated for a long period the interferometer, PTB always alternated measurements
of time such as 10 d. The relative uncertainty of the frequency between sphere and empty etalon.
of the laser is estimated to be approximately 1 × 10−11 at 1 s The dimensional measurement is traced back to the laser
averaging time. This uncertainty is limited by the statistical wavelength of a 633 nm He–Ne laser recommended by BIPM.
frequency fluctuation of the offset laser and can be ignored for The stability of the stabilised He-Ne laser, the auxiliary
the volume measurement. unmodulated He-Ne laser and the extended-cavity-diode-laser
The measured diameter is the ‘apparent’ diameter, (ECDL) for the wavelength-tuning was at the 10−12 level and
which is not corrected for the phase shift due to the sur- therefore did not contribute to the volume uncertainty budget.
face layer. The mean apparent diameters at 20.000 °C and For the measurements used in this paper, the interferom-
0 Pa are 93 710.811 11(62) µm and 93 701.526 29(68) µm eter was improved by stabilizing the irradiance of the laser
368
Figure 2. Diameter topographies of the 28Si spheres AVO28-S5c (left, (p-v)diameter = 69 nm) and AVO28-S8c (right, (p-v)diameter = 38 nm).
light. The light from the ECDL was split into two parts and Table 6. Uncertainty budgets for the apparent volume measure-
guided to the two arms of the interferometer by multi-mode ments of the 28Si spheres at PTB [48].
fibres. Two aspects of these measurements were taken into (u(V)/V) /10–9
account: a mode scrambler provided a uniform intensity dis-
AVO28-S5c AVO28-S8c
tribution within the field of view, and a monitor photo-diode
combined with a ‘noise eater’ (i.e. a servo control with a fast Interferometry 4 4
liquid crystal light modulator) stabilised the intensity of the Temperature 6 6
Wavefront distortions 25 19
interferometer input [49]. Parasitic interferences 0.2 0.2
Sphere AVO28-S5c was measured in August 2013. It was Volume 6 6
washed, following the suggestions of the IAC, with distilled Total 27 21
water and deconex OP162, a pH neutral, salt-free cleaning
concentrate for precision optical components. After exten- Table 7. The optical constants (refractive index n and the absorp-
sive purging with distilled water, the sphere was rinsed with tion index k) of the surface layers (SiO2, H2O and CmHn) and the
alcohol (p.A., pro analysi, analytically pure). The measure- silicon core (Si). The Abbr column refers to the abbreviations of the
model layers in figure 1.
ment followed a 50%-overlapping procedure: with a field of
view of 60°, the sphere was rotated for the next measurement Layer Abbr n k
by only 30°. The sphere was measured from both sides, so
Si Si Core 3.881 (1) 0.019 (1)
that a complete set of diameters measured through arm 1 and SiO2 OL 1.457 (10) 0
a complete set measured through arm 2 of the interferometer H2O CWL, PWL 1.332 (10) 1.54 (1.00) × 10–8
were derived. For each single measurement, the temperature CmHn CL 1.45 (10) 0 (0.1)
was corrected by a recently calibrated Pt25 resistance ther-
mometer as well as with a set of thermocouples [50]. roundness (decreased p-v values) for the repolished spheres,
Sphere AVO28-S8c was measured in a like manner in the influence of the wave front aberrations were also pre-
September 2014. In this case, the data came from two sets sumably reduced (table 6), but such distortions remain the
of completely overlapping measurements, each taken in a dif- principal uncertainty contribution for the present sphere
ferent sequence and with different rotation steps. Nevertheless, interferometer at PTB. A new interferometer, with a set of
the volume characterization was based on 7 × 105 diameter objectives with considerably reduced wave front aberra-
values. tions, is currently being tested. Results for this new instru-
Compared to the initial spherical state produced by the ment together with optical simulation calculations will be
Australian polishing, the PTB polishing removed about 12 µm reported soon.
from the diameter of sphere AVO28-S5. The p-v value of To take into account the optical behaviour of the surface
the diameter was reduced from 98 nm to 69 nm, illustrating layers, a layer model was created [48] and is illustrated in
nicely the effects of anisotropy of the modulus of elas- figure 1. The optical constants of the different surface layers
ticity (E-modulus) of a silicon single crystal [51, 52]. For are given in table 7.
AVO28-S8, two previous measurement cycles, AVO28-S8a Table 8 lists the diameter values and the phase shift cor-
and AVO28-S8b, have already been reported [5, 12, 48]. The rections ΔΦ due to the optically applicable layers for both
diameter of the sphere AVO28-S8 has now been reduced by participating institutes (different types of interferometers).
20 µm and its p-v value decreased from 90 nm to 38 nm. Calculations using the layer model suggested that the influ-
The uncertainties follow the considerations and calcula- ence of the surface layers on the optical diameter measure-
tions presented in [48]. Due to the smaller deviations from ments was quite small. The value of the phase shift showed
369
28
Table 8. Diameter and volume of Si spheres (tITS-90 = 20 °C, vacuum).
Mean apparent Mean diameter of Volume V of Si
Sphere Lab. diameter/nm Date of measurement ΔΦ/nma Si core/nm core/cm3
AVO28-S5c PTB 93 710 811.38 (83) August 2013 0.000 (17) 93 710 811.38 (83) 430.891 291 (12)
AVO28-S5c NMIJ 93 710 811.11 (62) June–July 2014 −0.001 (25) 93 710 811.11 (62) 430.891 288 (9)
AVO28-S5c weighted mean 93 710 811.21 (50) 93 710 811.21 (50) 430.891 289 (7)
AVO28-S8c PTB 93 701 526.24 (66) September 2014 0.010 (32) 93 701 526.26 (66) 430.763 222 (9)
AVO28-S8c NMIJ 93 701 526.29 (68) Nov.–Dec. 2014 −0.009 (28) 93 701 526.27 (68) 430.763 223 (9)
AVO28-S8c weighted mean 93 701 526.26 (47) 93 701 526.26 (47) 430.763 223 (7)
a
ΔΦ = phase shift corrections.
Table 9. Masses of the 28Si spheres as measured in air (without a correction for the reversible water layer).
Sphere Laboratory Mass/kg Mass unc./µg Date of measurement

AVO28-S5c PTB 0.999 698 447 12 Nov./Dec. 2013
BIPM 0.999 698 423 13 Feb./March 2014
NMIJ 0.999 698 437 16 May/June 2014
AVO28-S8c BIPM 0.999 401 303 13 Feb./March 2014

NMIJ 0.999 401 328 13 Apr./May 2014
PTB 0.999 401 325 12 Oct./Nov. 2014
Table 10. Masses of the 28Si spheres as measured in vacuum.
Sphere Laboratory Mass/kg Mass unc./µg Date of measurement

AVO28-S5c PTB 0.999 698 438 5 6.9 Nov./Dec. 2013
BIPM 0.999 698 430 1 4.4 Feb./March 2014
NMIJ 0.999 698 437 3 7.6 May/June 2014
Weighted mean 0.999 698 433 2 3.5
AVO28-S8c BIPM 0.999 401 309 5 4.3 Feb./March 2014

NMIJ 0.999 401 320 9 7.8 Apr./May 2014
PTB 0.999 401 316 0 7.3 Oct./Nov. 2014
Weighted mean 0.999 401 312 8 3.5
Table 11. The traceability path between the reference masses used for the mass determination of the 28Si spheres at PTB and the prototypes
of the kilogram at the BIPM.
Reference masses
Last
Last calibration by calibration by
Sphere Date Name Last calibration by PTB Name PTB Name BIPM
AVO28-S5c Nov./Dec. 2013 PtSk-Z Nov. 2013 (against No. 70) No. 70 Jun. 2013
AVO28-S8c Oct./Nov. 2014 PtSk-Z Oct. 2014 (against No. 70)
No. 70 Oct. 2014 (against No. 52) No. 52 Nov. 2013 (against No. 70 Jun. 2013
No. 70)
a zero-crossing for thicknesses in the range of 1.5 nm to two spheres were cleaned three times, using the cleaning pro-
2 nm. As regards the structure of the Si-SiO2 interface, cedure recommended by the National Metrology Institute of
density-functional calculations of the lattice strain sug- Australia, NMIA [53]. A set of stainless steel air buoyancy
gested that the effect of surface relaxation and reconstruc- artefacts, consisting of a tube (Cp) and a hollow cylinder (Cc),
tion is an order of magnitude smaller than the present was used to determine the air density for the buoyancy cor-
uncertainty of the volume measurement [68]. rection. A set of Pt-Ir sorption artefacts, consisting of a cyl-
inder (A0) and a stack of 8 disks (A18), was used to establish
the link between masses in air and vacuum. The masses in air
2.6. Mass
were measured three times, with weighings in vacuum occur-
The masses of the spheres AVO28-S5c and AVO28-S8c were ring in-between. After each change of condition, a stabiliza-
determined in air and under vacuum by the BIPM, NMIJ and tion period of at least 3 d was observed before starting the next
PTB. set of measurements under the new conditions.
All measurements at the BIPM were carried out using their The weighings of the two spheres were carried out directly
Sartorius CCL 1007 mass comparator. Before weighing, the after the first phase of the Extraordinary Calibrations against
370
the IPK [69]. The spheres were compared in air with BIPM The traceability of NMIJ’s mass value to the IPK was
working standard No. 77, which had itself been weighed achieved by using the mass value 1 kg + 0.360 mg for the Pt-Ir
against working standards Nos. 91 and 650. The two latter prototype No. 94 (BIPM Certificate No. 59, 2009) measured
working standards had already been compared directly with at the BIPM in August 2009 and also by applying a correction
the IPK, in air. Tables 9 and 10 give the masses of both of −0.0301 mg, which corresponded to the value during this
spheres, in air and in vacuum, with respect to the mass of the time period recommended by the BIPM in December 2014
IPK. During the Extraordinary Calibrations, it was observed [8]. In addition, the mass value 1 kg + 0.176 mg for the Pt–Ir
that the BIPM ‘as-maintained’ mass unit was 35 µg smaller prototype No. 6 (BIPM Certificate No. 8, 1993) at the third
than the mass of the IPK. The standard uncertainty of this periodic verification of national prototypes of the kilogram
mass difference is estimated as 3 µg. Therefore, the masses was used to determine the long-term drift rate of NMIJ’s mass
of AVO28-S5c and AVO28-S8c, expressed in the mass unit value. The results for the weighings are shown in tables 9 and
maintained formerly at the BIPM, are 35 µg larger than the 10. The NMIJ measurement results are the average obtained
values given in tables 9 and 10. by two measurement cycles.
A study was undertaken at the BIPM to determine the mass At PTB, the mass determinations of the 28Si spheres
of the CWL present on the surface of a natural silicon sphere. AVO28-S5c and AVO28-S8c were performed in air and under
Two methods were used: baking the sphere under vacuum and vacuum during the periods November/December 2013 and
immersing the sphere in doubly distilled water [54]. For both October/November 2014, respectively. The procedure agreed
methods, the chemical adsorption coefficient was obtained by to in [53] for the Avogadro project was used for cleaning the
determining the mass difference under vacuum conditions (to spheres. After cleaning, the measurements were made on a
reduce the uncertainty) prior to and after placing the sphere Mettler-Toledo M_one mass comparator using the prototype
in air, in order to reintroduce the chemisorbed water to the of the kilogram No. 70 and the Pt-Ir mass standard PtSk-Z
surface of the sphere. The mean chemical adsorption coeffi- as reference masses. Table 11 shows the traceability path
cient thus obtained was 0.026 µg cm−2 with a standard uncer- between the reference masses used for the mass determina-
tainty (k = 1) of 0.012 µg cm−2. The BIPM results confirmed tion of the 28Si spheres and the prototypes of the kilogram
those obtained by NMIJ/AIST (Japan) which had measured at the BIPM. The last calibration of prototype No. 70 at the
the adsorption isotherms on SiO2/Si(1 0 0) plane surfaces [43]. BIPM was performed in June 2013 (Certificate No. 70, BIPM
NMIJ conducted mass measurements using an early ver- 2013). The BIPM determined the mass of this prototype to be
sion of the Mettler-Toledo M_one mass comparator [55]. 1 kg − 0.207(7) mg. Mass comparisons with other prototypes
The measurements under vacuum showed a higher reproduc- of the kilogram were made at PTB before and after the pro-
ibility than those in air, where air buoyancy and convection totype was hand-carried between PTB and BIPM. The results
effects affected the weighing stability. The standard uncer- showed a mass loss of 5 µg. Because it is unknown whether
tainty of the mass difference measurement under vacuum this mass change occurred before or after the calibration of
was about 5 µg. the prototype, this mass difference had to be considered in the
The silicon spheres were washed manually by rubbing uncertainty budget as an uncertainty contribution due to the
their surfaces with nitride rubber gloves using a neutral deter- instability of the mass of the prototype.
gent for optical components for about 10 min, followed by The Pt-Ir cylinder PtSk-Z is one of two Pt-Ir artefacts used
a rinse with 30 dm3 of pure water and 1 dm3 of ethanol. The as sorption artefacts for the determination of the sorption cor-
washing procedure is basically the same as that described in rection and as a link between the mass of the silicon spheres in
the ‘protocol for the international mass comparison on the two vacuum and the prototype of the kilogram No. 70 in air [53].
28
Si spheres’ distributed to the participants in April 2008 [53]. The surface area difference between the artefacts amounts to
The protocol does not clearly specify the number of wash- 183 cm2. In order to apply a buoyancy correction in air with
ings. Therefore, NMIJ investigated the washing effects on the lowest uncertainty, the air density was measured using
reproducibility by washing AVO28-S5c three times and by buoyancy artefacts [53].
washing AVO28-S8c twice. The repeated washings, with sub- For the sphere AVO28-S5c, a mass of 0.999 698 483(13) kg
sequent mass measurements at NMIJ, showed a mass decrease and 0.999 698 4743(93) kg was determined in air (without cor-
between 3 µg and 8 µg after each washing. This phenomenon rection of reversible sorption) and in vacuum, respectively.
could be interpreted as a decrease of a carbonaceous layer For sphere AVO28-S8c, a mass of 0.999 401 362(14) kg was
(CL) on surfaces by the washing procedure. The result of the determined in air (again without correction of reversible sorp-
washing experiment suggests the number of washings speci- tion) and 0.999 401 3534(94) kg in vacuum. For both spheres,
fied in the protocol should not be less than three to achieve the mass difference between the measurements in air and in
a stable mass value at the low microgram level. It should be vacuum amounted to about 8.5 µg. This value corresponds to a
also noted that the washing effect on the mass determination change in the sorption coefficient between vacuum and air of
would be compensated, to a great extent, by the surface char- 30 ng cm−2, which is in good agreement with the values pub-
acterisation described in section 2.4 if we specify the washing lished in [56] and [57].
procedure rigorously. For now, NMIJ estimated the standard In December 2014, PTB was informed by the BIPM
uncertainty of the reproducibility of the washing procedure about revised mass values for BIPM calibrations following
specified in the aforementioned protocol to be 4 µg, assuming the Extraordinary Calibration Campaign against the IPK
uniform distribution with a half width of 7 µg. [8, 69]. Consequently, revised mass values and revised drift
371
Table 12. NA determination. The lattice parameter, volume, and Table 13. The uncertainty budget for the new NA determination us-
density were measured at tITS-90 = 20.0 °C and p = 0 Pa. ing AVO28-S5c. The percent contributions to the total uncertainty
are the relevant variance fractions ratioed to the total variance. The
Quantity Unit AVO28-S5c AVO28-S8c principal uncertainty contributions are, at present, due to surface
characterization and the volume determination.
M g mol−1 27.976 970 09 (15) 27.976 970 09 (15)
a pm 543.099 6219 (10) 543.099 6168 (11) Relative
V cm3 430.891 2891 (69) 430.763 2225 (65) Quantity uncertainty/10−9 Contribution/%
m g 999.698 359 (11) 999.401 250 (16)
ρ = m/V kg m−3 2320.070 943 (46) 2320.070 976 (51) Molar mass (M) 5 6
NA 1023 mol−1 6.022 140 72 (13) 6.022 140 80 (14) Lattice parameter (a) 5 6
Surface 10 23
characterization (mSL)
corrections of the involved prototypes, No. 70 and No. 52, had Sphere volume (V) 16 59
to be considered. Regarding the calibration of prototype No. Sphere mass (m) 4 4
70 in June 2013, a corrected mass value of 1 kg − 0.242(3) Point defects 3 2
Total 21 100
mg was assigned by the BIPM. The difference between the
original and the corrected mass value amounts to 35 µg.
In due consideration of the traceability path given in table 11, Table 14. Value of the Avogadro constant based upon the repol-
the mass values determined for the 28Si spheres AVO28-S5c ished 28Si-enriched silicon spheres.
and AVO28-S8c at PTB were revised correspondingly (tables Sphere NA/1023 mol−1 ur /10−9
9 and 10).
The weighted mean of the masses measured in vacuum was AVO28-S5c 6.022 140 72 (13) 21
AVO28-S8c 6.022 140 80 (14) 23
used for the determination of the Avogadro constant (table Mean value 6.022 140 76 (12) 20
10). The effect of the correlation arising from the common
traceability of all masses to the IPK was taken into account for
the weighted mean and its uncertainty, although its magnitude and express the measurement uncertainties, the approach rec-
is nearly negligible. ommended by the Guide to the Expression of Uncertainty in
To determine the mass of the silicon core, the mass of the Measurement [58] was applied using the GUMWorkbench
surface layers was subtracted from the mass of the sphere. In software [59]; covariances were calculated and folded into
addition, owing to point defects, there is a difference between the uncertainty analysis. The quantities dominating the total
the mass of a sphere having Si atoms occupying all regular uncertainty of the revised Avogadro constant NA were the
sites and the measured mass value. apparent diameter, Dm, of the spheres and the mass of the sur-
face layer, mSL (see table 13).
m deficit = V Σi(m28 − mi )Ni
(4) The new NA determinations, based on a careful reanalysis
In equation (4), m28 and mi are the masses of a 28Si atom of the two AVO28 spheres, are summarized in table 14; they
and of the point defect named i, respectively (a vacancy mass differ by only 13(20) × 10–9 NA. Averaging these two values,
is zero.) Oxygen was associated with an interstitial lattice site, the final value for the Avogadro constant becomes
so that mO is the sum of the oxygen and 28Si masses. The same
NA = 6.022 140 76(12) × 10 23 mol−1,
(5)
applies to nitrogen impurities. V is the sphere volume and Ni is
the concentration of the point defect i (see section 2.1). with a relative standard uncertainty of 20 × 10−9.
The etching did not completely remove the metals from the
AVO28-S5 sphere. Approximately 5% of the original metallic
4. Conclusions
contaminant remained (corresponding to 4 µg) and had prob-
ably diffused into the sphere during the thermal oxidation The value NA = 6.022 140 82(18) × 1023 mol−1 given in a pre-
that was performed before repolishing. After repolishing the vious paper [12] must also be updated due to the recalibra-
spheres, no metals could be detected by XRF at the sphere’s tion of the mass standards following the recent Extraordinary
surface. Therefore, a mass correction of −4(3) µg was added Calibration Campaign against the IPK. A provisional cor-
to the mass deficit. A mass deficit of 3.8(3.8) µg was then rected value would be NA = 6.022 140 99(18) × 1023 mol−1.
calculated for the AVO28-S5c sphere and a mass deficit of Some considerations of the potential correlations between
22.7(3.5) µg for the AVO28-S8c sphere. the different measurement components contributing to the
After the surface-layer mass (mSL, table 4(b) in section 2.4) new and previously reported NA values should be addressed.
was subtracted and the mass deficit (mdeficit) was added, the Si The molar mass measurements are uncorrelated; they were
core mass m = msphere − mSL + mdeficit could be calculated for amply repeated in three different laboratories using different
both spheres (table 12). calibration and measurement approaches as well as TMAH
instead of NaOH as matrix diluent. The measurements of the
3. Avogadro constant lattice parameters differed in significant respects: the optical
interferometer was rebuilt to accommodate a different wave-
Table 12 gives the results of the new and refined measurements length, the temperature measurements relied on new equip-
of the molar mass, lattice parameter, volume, mass, and den- ment and new calibrations, and the apparatus was completely
sity of the cores of the enriched silicon spheres. To evaluate disassembled and realigned. However, because the same
372
6.0221418
6.0221416
6.0221414
10-23NA / mol-1
IAC 2010
revised
6.0221412
NIST-3
revised and NRC 2014
6.0221410
combined revised
CODATA 2010 this paper
not revised
6.0221408
6.0221406
Figure 3. The most accurate Avogadro constant determinations available at present. The value labelled CODATA 2010 is not corrected for
the recalibration of the mass standards following the Extraordinary Calibration Campaign against the IPK. The vertical error bars indicate
the standard uncertainty of each value.
interferometer crystal was used for both measurements, and instrumentation. Therefore, provided the source material is
the lattice parameter values relied on measurements from a chemically and physically well characterized with respect to
single laboratory, the past and present results may be corre- the mass fractions of the minor isotopes and impurities, the
lated by up to 50%. Because the AVO28-S5 and AVO28-S8 value of the lattice parameter, and the crystallographic per-
spheres were fully repolished, the contributions of their sur- fection, material realizations of the kilogram and its submul-
face properties to the error budget are not correlated. The tiples in the form of crystal artefacts require only volume
same applies to the volume measurements; however, diffrac- measurements and surface characterizations. These same two
tion effects—which are extremely difficult to model or inves- parameters would also be the only two quantities necessary to
tigate—may be correlated between the old and new values. monitor the secular stability of the artefacts.
Mass measurements were newly calibrated and traceable back
to the IPK. Therefore, their correlation with the past measure- Acknowledgments
ments is considered to be negligible. Finally, the point defect
contributions to the error budget are basically from the same The authors would like to thank all colleagues involved in this
source and are thus 100% correlated. Details will be published project, in particular A Picard, M Hämpke, M Firlus, F Scholz
in a separate paper [60]. and A Felgner.
The forthcoming definition of the kilogram [61] will be based This work was jointly funded by the European Metrology
on fixing the value of the Planck constant. To compare this new Research Programme (EMRP) participating countries within
NA value with the most recent Planck constant results by the the European Association of National Metrology Institutes
watt balance (WB) experiments [62, 63], we can convert them to (EURAMET) and the European Union.
an equivalent Avogadro constant value using the molar Planck It was also supported in part by the Grant-in-Aid for
constant (NAh = 3.990 312 7176(28) × 10−10 Js mol−1), which Scientific Research (B) (KAKENHI 24360037) from the
has a relative standard uncertainty of only 0.7 × 10−9 [64]. The Japan Society for the Promotion of Science and the Italian
NRC WB value was updated for the recalibration of mass stand- ministry of education, university, and research (awarded pro-
ards following the Extraordinary Calibration Campaign against ject P6-2013, implementation of the new SI).
the IPK, h(NRC) = 6.626 070 11(12) × 10−34 Js [65]. The NIST Certain commercial equipment, instruments, or materials
updated their former h values and published a combined h value are identified in this paper in order to specify the experimental
for the NIST-3 WB, h(NIST-3) = 6.626 069 36(37) × 10−34 Js procedure adequately. Such identification is not intended
[66]. Note however that we have been unable to update the to imply recommendation or endorsement by the National
CODATA 2010 NA value. These results are compared in Institute of Standards and Technology, nor is it intended to
figure 3. The accuracy and uncertainty of the new determination imply that the materials or equipment identified are neces-
of NA (equation (5)) is within the targeted relative uncertainty sarily the best available for the purpose.
so as to make the kilogram redefinition possible [61]; therefore,
this new measurement result demonstrates a successful mise en
pratique of a definition based upon a fixed value of the Planck References
constant.
Because clear-cut effects of crystal imperfections have [1] Massa E and Nicolaus A 2011 Special issue international
not yet been detected, the uncertainty of equation (5) still determination of the Avogadro constant Metrologia
appears to be limited by the performance of the measurement 48 S1–119
373
[2] Andreas B et al 2011 Determination of the Avogadro constant estimation of the achievable uncertainty Metrologia
by counting the Atoms in a 28Si crystal Phys. Rev. Lett. 51 354–60
106 030801 (4 pp) [24] D’Agostino G, Mana G, Oddone M, Prata M, Di Luzio M and
[3] Busch I, Azuma Y, Bettin H, Cibik L, Fuchs P, Fujii K, Pramann A 2015 The 30Si mole fraction of silicon
Krumrey M, Kuetgens U, Kuramoto N and Mizushima S materials highly enriched in 28Si determined by
2011 Surface layer determination for the Si spheres of the instrumental neutron activation analysis Anal. Chem.
Avogadro project Metrologia 48 S62–82 (submitted)
[4] Qin M, Poon M C and Yue C Y 2000 A study of nickel silicide [25] Wang M, Audi G, Wapstra A H, Kondev F G, MacCormick M,
film as a mechanical material Sensors Actuators 87 90 Xu X and Pfeiffer B 2012 The AME2012 atomic mass
[5] Nicolaus R A et al 2013 State of Avogadro 28Si sphere S8 evaluation (II). Tables, graphs and references Chin. Phys. C
IEEE Trans. Instrum. Meas. 62 1499–505 36 1603–2014
[6] Meeß R, Felgner A, Hinzmann G and Lück A 2014 Surface [26] Bodnar O and Elster C 2014 On the adjustment of inconsistent
topography of superpolished silicon spheres Proc. Int. Conf. data using the Birge ratio Metrologia 51 516–21
Eur. Soc. Precis. Eng. Nanotechnol. 2 99–102 [27] CCQM/13-22 2013 CCQM Guidance note: Estimation of a
[7] Bowen D K and Tanner B K 1998 High Resolution x-Ray consensus KCRV and associated Degrees of Equivalence,
Diffractometry and Topography (London: Taylor and KCWG, BIPM (www.bipm.org/cc/AllowedDocuments.
Francis) jsp?cc=CCQM)
[8] Stock M 2014 Revised mass values for BIPM calibrations [28] Kessel R, Berglund M and Wellum R 2008 Appplication
following the extraordinary calibration campaign against of consistency checking to evaluation of uncertainty in
the IPK correspondence BIPM December 2014 multiple replicate measurements Accred. Qual. Assur.
[9] Itoh Y, Nozaki T, Masui T and Abe T 1985 Calibration curve 13 293–8
for infrared spectrophotometry of nitrogen in silicon Appl. [29] Steele A G and Douglas R J 2006 Extending En for
Phys. Lett. 47 488–9 measurement science Metrologia 43 S235–43
[10] D’Agostino G, Bergamaschi L, Giordani L, Mana G, Massa E [30] Robouch P, Younes N and Vermaercke P 2003 The ‘Naji Plot’,
and Oddone M 2012 Elemental characterization of the a simple graphical tool for the evaluation of interlaboratory
Avogadro silicon crystal WASO04 by neutron activation comparisons Data Analysis of Key Comparisons ed D
analysis Metrologia 49 696–701 Richter et al (Bremerhaven: Wirtschaftsverlag NW)
[11] D’Agostino G, Bergamaschi L, Giordani L, Mana G and [31] Massa E, Mana G, Kuetgens U and Ferroglio L 2011
Oddone M 2014 Instrumental neutron activation analysis of Measurement of the {2 2 0} lattice plane spacing of a 28Si
an enriched 28Si single-crystal J. Radioanal. Nucl. Chem. x-ray interferometer Metrologia 48 S37–43
299 277–82 [32] White D et al 2008 Uncertainties in the realisation of the
[12] Andreas B et al 2011 Counting the atoms in a 28Si crystal for a SPRT subranges of the ITS-90 Report of the CCT-WG3
new kilogram definition Metrologia 48 S1–13 on Uncertainties in Contact Thermometry CCT/08-19/rev
[13] Pramann A, Rienitz O, Schiel D, Schlote J, Güttler B and (Sèvres: Bureau International des Poids et Mesures)
Valkiers S 2011 Molar mass of silicon highly enriched in [33] Massa E, Sasso C, Mana G and Palmisano C 2015 A more
28
Si determined by IDMS Metrologia 48 S20–5 accurate measurement of the 28Si lattice parameter J. Phys.
[14] Yang L, Mester Z, Sturgeon R E and Mejia J 2012 Chem. Ref. Data accepted
Determination of the atomic weight of 28Si-enriched silicon [34] Fujimoto H, Waseda A and Zhang X W 2011 Homogeneity
for a revised estimate of the Avogadro constant Anal. Chem. characterization of lattice spacing of silicon single crystals
84 2321–7 by the self-referenced lattice comparator Metrologia
[15] Pramann A, Rienitz O, Noordmann J, Güttler B and Schiel D 48 S55–61
2014 A more accurate molar mass of silicon via high [35] Waseda A, Fujimoto H, Zhang X W, Kuramoto N and Fujii K
resolution MC-ICP-mass spectrometry Z. Phys. Chem. 2015 Homogeneity characterization of lattice spacing of
228 405–19 silicon single crystals IEEE Trans. Instrum. Meas. accepted
[16] Narukawa T, Hioki A, Kuramoto N and Fujii K 2014 Molar- [36] Becker P et al 2003 Determination of the Avogadro constant
mass measurement of a 28Si-enriched silicon crystal for via the silicon route Metrologia 40 271–87
determination of the Avogadro constant Metrologia [37] Beckhoff B, Gottwald A, Klein R, Krumrey M, Müller R,
51 161–8 Richter M, Scholze F, Thornagel R and Ulm G 2009 A
[17] Vocke R D Jr, Rabb S A and Turk G C 2014 Absolute silicon quarter-century of metrology using synchrotron radiation by
molar mass measurements, the Avogadro constant and the PTB in Berlin Phys. Status Solidi B 246 1415–34
redefinition of the kilogram Metrologia 51 361–75 [38] Picard A, Bignell N, Downes S and Mizushima S 2009 Mass
[18] Pramann A, Lee K-S, Noordmann J and Rienitz O 2015 Probing comparison of the 1 kg silicon sphere Metrologia
the homogeneity of the isotopic composition and molar mass 46 1–10
of the ‘Avogadro-28’-crystal Metrologia in preparation [39] Krumrey M, Gleber G, Scholze F and Wernecke J 2011
[19] Rienitz O, Pramann A and Schiel D 2010 Novel concept for Synchrotron radiation-based x-ray reflection and scattering
the mass spectrometric determination of absolute isotopic techniques for dimensional nanometrology Meas. Sci.
abundances with improved measurement uncertainty: part Technol. 22 094032
1—theoretical derivation and feasibility study Int. J. Mass [40] Kuramoto N, Azuma Y, Inaba H, Hong F-H and Fujii K 2015
Spectrom. 289 47–53 Improvements of the volume measurement for 28Si spheres
[20] Vocke R D Jr et al 2015 in preparation to determine the Avogadro constant IEEE Trans. Instrum.
[21] Rienitz O, Pramann A, Kaltenbach A, Hinrichs J, Barnhart D Meas. accepted
and Churchill G 2015 in preparation [41] Kuramoto N, Fujii K, Nicolaus A, Bartl G, Gray M, Manson P
[22] Pramann A, Rienitz O and Schiel D 2012 Silicon isotope and Giardini W 2011 Diameter comparison of a silicon
ratios affected by sodium-induced broadband interference sphere for the international Avogadro coordination IEEE
in high resolution multicollector-ICPMS Anal. Chem. Trans. Instrum. Meas. 60 2615–20
84 10175–9 [42] Mizushima S, Kuramoto N, Ueda K and Fujii K 2015 Mass
[23] D’Agostino G, Mana G, Oddone M, Prata M, Bergamaschi L measurement of 1 kg silicon spheres for the determination
and Giordani L 2014 Neutron activation analysis of the of the Avogadro and Planck constants IEEE Trans. Instrum.
30
Si content of highly enriched 28Si: proof of concept and Meas. accepted
374
[43] Mizushima S 2004 Determination of the amount of gas [56] Picard A and Fang H 2004 Methods to determine water vapour
adsorption on SiO2/Si(1 0 0) surfaces to realize precise mass sorption on mass standards Metrologia 41 333–9
measurements Metrologia 41 137–44 [57] Picard A 2006 Mass determination of a 1 kg silicon sphere for
[44] Kuramoto N, Fujii K and Yamazwa K 2011 Volume the Avogadro project Metrologia 43 46–52
measurements of 28Si spheres using an interferometer with [58] ISO 1993 Guide to the Expression of Uncertainty in
a flat etalon to determine the Avogadro constant Metrologia Measurement 1st edn (Geneva: International Organisation
48 S83–95 for Standardisation) (corrected and reprinted 1995)
[45] Andreas B, Ferroglio L, Fujii K, Kuramoto N and Mana G [59] GUM Workbench Version 1.2.11.56 ‘96-’99 by Metrodata
2011 Phase corrections in the optical interferometer for Si GmbH
sphere volume measurements at NMIJ Metrologia [60] Mana G et al 2015 An analysis of the correlation of the
48 S104–11 Avogadro constant determinations by counting 28Si atoms
[46] Nicolaus R A and Bönsch G 2005 Absolute volume J. Phys. Chem. Ref. Data submitted
determination of a silicon sphere with the spherical [61] Resolution 1 of the CGPM 2011 On the possible future
interferometer of PTB Metrologia 42 24–31 revision of the International System of Units, the SI (www.
[47] Bartl G, Krystek M and Nicolaus A 2014 PTB’s enhanced bipm.org/en/CGPM/db/24/1)
stitching approach for the high-accuracy interferometric [62] Sanchez C A, Wood B M, Green R G, Liard J O and Inglis D
form error characterization of spheres Meas. Sci. Technol. 2014 A determination of Planck’s constant using the NRC
25 064002 (8 pp) watt balance Metrologia 51 S5–14
[48] Bartl G, Bettin H, Krystek M, Mai T, Nicolaus A and Peter A [63] Schlamminger S, Haddad D, Seifert F, Chao L S,
2011 Volume determination of the Avogadro spheres of Newell D B, Liu R, Steiner R L and Pratt J R 2014
highly enriched 28Si with a spherical Fizeau interferometer Determination of the Planck constant using a watt balance
Metrologia 48 S96–103 with a superconducting magnet system at the National
[49] Spolaczyk R, Nicolaus A and Peter A 2014 Improvements in Institute of Standards and Technology Metrologia 51
illumination of the Avogadro sphere interferometer 29th S15–24
Conf. on Precision Electromagnetic Measurements (Rio [64] Mohr P J, Taylor B N and Newell D B 2012 CODATA
de Janeiro, 25–29 Auguest 2014) pp 394–5 (Conference recommended values of the fundamental physical constants:
Digest) 2010 Rev. Mod. Phys. 84 1527–605
[50] Nicolaus R A and Geckeler R D 2007 Improving the [65] Wood B 2015 private communication
measurement of the diameter of Si-spheres IEEE Trans. [66] Schlamminger S, Steiner R L, Haddad D, Newell D B,
Instrum. Meas. 56 517–22 Seifert F, Chao L S, Liu R, Williams E R and Pratt J R 2015
[51] Collins J G, Giardini W J, Leistner A J and Kenny M J 1997 A summary of the Planck constant measurements using
The influence of Young’s modulus on roundness in silicon a watt balance with a superconducting solenoid at NIST
sphere fabrication IEEE Trans. Instrum Meas. 46 572–5 Metrologia 52 L5–8
[52] Boyd E J and Uttamchandani D 2012 Measurement of the [67] Massa E, Mana G, Ferroglio L, Kessler E G, Schiel D and
anisotropy of Young’s modulus in single-crystal silicon Zakel S 2011 The lattice parameter of the 28Si spheres in
J. Microelectromech. Syst. 21 243–9 the determination of the Avogadro constant Metrologia 48
[53] Picard A, Barat P, Borys M, Firlus M and Mizushima S 2011 S44–9
State-of-the art mass determination of 28Si spheres for the [68] Melis C, Colombo L and Mana G 2015 Lattice strain at c-Si
Avogadro project Metrologia 48 S112–9 surfaces: a density functional theory calculation Metrologia
[54] Barat P 2015 Mémoire d’ingénieur CNAM, Projet 52 214–21
international Avogadro: Détermination de la masse de la [69] Stock M, Barat P, Davis R S, Picard A and Milton M J T
couche d’eau chimisorbée à la surface d’une sphère en 2015 Calibration campaign against the international
silicium naturel prototype of the kilogram in anticipation of the
[55] Mizushima S, Ueki M and Fujii K 2004 Mass measurement redefinition of the kilogram part 1: comparison of the
of 1 kg silicon spheres to establish a density standard international prototype with its official copies Metrologia
Metrologia 41 S68–74 52 310–6
375

Azuma 2015 Metrologia 52 360

Uploaded by

Copyright:

Available Formats

Azuma 2015 Metrologia 52 360

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Azuma 2015 Metrologia 52 360

Uploaded by

Copyright:

Available Formats

Metrologia

PAPER • OPEN ACCESS Related content

- A new 28Si single crystal: counting the

- Counting the atoms in a 28Si crystal for a

- Some aspects on the uncertainty

- Trustworthy data underpin reproducible

This content was downloaded from IP address 93.139.8.60 on 14/03/2020 at 23:06

Metrologia 52 (2015) 360–375 doi:10.1088/0026-1394/52/2/360

Improved measurement results for

Received 31 December 2014, revised 16 February 2015

(Some figures may appear in colour only in the online journal)

1. Introduction the Planck constant h. Additionally, an accurate NA constant

Sphere Laboratory Mass/kg Mass unc./µg Date of measurement

AVO28-S8c BIPM 0.999 401 303 13 Feb./March 2014

Table 10. Masses of the 28Si spheres as measured in vacuum.

Sphere Laboratory Mass/kg Mass unc./µg Date of measurement

AVO28-S8c BIPM 0.999 401 309 5 4.3 Feb./March 2014

You might also like