Nuclear Instruments and Methods in Physics Research A 876 (2017) 50–53

Contents lists available at ScienceDirect

Nuclear Instruments and Methods in Physics Research A

journal homepage: www.elsevier.com/locate/nima

From the speed of sound to the speed of light: Ultrasonic Cherenkov MARK
refractometry
G.D. Hallewell
Centre de Physique des Particules de Marseille, 163 Avenue de Lumniny Case 907, 13288 Marseille Cedex 09, France

A R T I C L E I N F O A BS T RAC T

Keywords: Despite its success in the SLD CRID at the SLAC Linear Collider, ultrasonic measurement of Cherenkov radiator
Refractivity refractive index has been less fully exploited in more recent Cherenkov detectors employing gaseous radiators.
Cherenkov radiator This is surprising, since it is ideally suited to monitoring hydrostatic variations in refractive index as well as its
Ultrasonic binary gas mixture analysis evolution during the replacement of a light radiator passivation gas (e.g. N2, CO2) with a heavier fluorocarbon
Ultrasonic flowmetry
(e.g. C4F10[CF4]; mol. wt. 188[88]). The technique exploits the dependence of sound velocity on the molar
concentrations of the two components at known temperature and pressure. The SLD barrel CRID used an 87%
C5F12/13%N2 blend, mixed before injection into the radiator vessel: blend control based on ultrasonic mixture
analysis maintained the β=1 Cherenkov ring angle to a long term variation better than ± 0.3%, with refractivity
monitored ultrasonically at multiple points within the radiator vessel.
Recent advances using microcontroller-based electronics have led to ultrasonic instruments capable of
simultaneously measuring gas flow and binary mixture composition in the fluorocarbon evaporative cooling
systems of the ATLAS Inner Detector. Sound transit times are measured with multi-MHz transit time clocks in
opposite directions in flowing gas for simultaneous measurement of flow rate and sound velocity. Gas
composition is evaluated in real-time by comparison with a sound velocity/composition database.
Such instruments could be incorporated into new and upgraded gas Cherenkov detectors for radiator gas
mixture (and corresponding refractive index) measurement to a precision better than 10−3. They have other
applications in binary gas analysis - including in Xenon-based anaesthesia. These possibilities are discussed.

1. Introduction
Despite its success in the SLD CRID at the SLAC Linear Collider [1–
3], continuous on-line ultrasonic measurement of Cherenkov radiator
refractive index has been less fully exploited in more recent Cherenkov
detectors employing gaseous radiators. This is surprising, since it is
ideally suited to monitoring hydrostatic variations in refractive index as
well as its evolution during the replacement of a passivation gas (e.g.
N2) with a radiator gas. It can also provide a simultaneous flow
measurement in the same instrument. The technique exploits the
dependence of sound velocity on the molar concentrations of the two
components of differing molecular weight at known temperature and
pressure.
Look-up tables of concentration vs. sound velocity are set up using
the general formalism expressing the sound velocity, vs, in terms of the
thermodynamic properties of pairs of component gases:
γm ∙R∙T
vs =
Mm
where γm is the adiabatic index for the mixture, R is the molar gas
constant (8.3145 J mol−1 K−1), T is the absolute temperature in Kelvin
and Mm is the combined molar mass of the mixture.
Mm and γm are given respectively by
∑i ωi CPi
Mm = ∑ ωi Mi andγm =
i ∑i ωi CVi (2)
where ωi is the molar fraction of component i in the mixture while CPi
and CVi are its molar specific heats at constant pressure and volume
respectively.
Fig. 1 illustrates the temperature and pressure variation of Cp and
Cv in pure C4F10 and pure N2, calculated using the NIST REFPROP
package [4]. This gas combination is relevant, for example, to the
choice of radiator and passivation gas in the COMPASS RICH [5] and
LHC RICH-1 [6]. The temperatures shown cover a wide operating
range while the absolute pressure range from 950 to 1050 mbarabs
assumes a radiator vessel referenced to varying atmospheric pressure.
The sound velocity, vs, in the radiator gas is measured from the
transit times of ultrasound pulses over a known acoustic path length.
(1)

E-mail address: gregh@cppm.in2p3.fr.

http://dx.doi.org/10.1016/j.nima.2017.01.014
Received 5 November 2016; Accepted 7 January 2017
Available online 22 January 2017
0168-9002/ © 2017 Elsevier B.V. All rights reserved.
G.D. Hallewell
Fig. 1. Top: Molar specific heat at constant pressure vs. temperature for N2 (left axis)
and C3F8 (right axis) at a range of temperatures and pressures Bottom: Molar specific
heats at constant volume.
Most of the instruments used in previous and current RICH detectors
[3,5,7,8] have used the 50 kHz capacitative ultrasonic transducer [9]
originally developed for Polaroid autofocus cameras. The relatively
high bias voltages (up to 350 V) that can be applied to these
transducers results in reduced jitter compared with many piezoelectric
transducers. Their disadvantage lies in an operating frequency close to
the maximum of the CO2 ultrasound absorption peak running from 10
to 100 kHz (FWHM), which makes them unsuitable for use in high
concentrations of CO2. In these gases piezoelectric transducers with
frequencies of 200 kHz or more can be operated.
Fig. 2 shows Cherenkov thresholds for particle species vs. measured
sound velocity in a radiator gas combining C4F10 with residual N2
radiator passivation gas at 25 °C. The Cherenkov threshold for the
particle species is shown on the left axis, while the C4F10 concentration
is shown on the right axis.
In the LHCb RICH-2 a few percent of CO2 is added to the
tetrafluromethane (CF4) radiator to suppress its scintillation light
[10]. Since CO2 presents advantages of easier thermodynamic separa-
tion from CF4 than a N2 passivation gas, ultrasonic refractometry is
worth considering. Fig. 3 illustrates Cherenkov particle thresholds and
CF4 concentration.
For a pair of gases of interest a database may be created as a series
of parameters to polynomial fits of component concentration vs. sound
velocity, covering the expected range of temperature and pressure.
These fit parameters can be thought of as being vertically-suspended
above the intersections of a 2-D grid of temperature and pressure
points. In practice the temperature and pressure (measured simulta-
51
Nuclear Instruments and Methods in Physics Research A 876 (2017) 50–53
Fig. 2. Cherenkov thresholds (left axis) for particle species vs. measured sound velocity
in a radiator gas combining C4F10 with residual N2 radiator passivation gas at 25 °C.;
right axis, C4F10 concentration.
Fig. 3. Cherenkov thresholds (left axis) for particle species vs. measured sound velocity
in a radiator gas combining CF4 with residual CO2 radiator passivation gas at 25 °C: right
axis, CF4 concentration.
neously with sound transit times) will fall between the grid: sound
velocity vs. concentration fit parameters can first be interpolated to
intermediate values corresponding to the measurement temperature at
grid pressures immediately above and below the measurement pres-
sure. A second interpolation between the intermediate fit parameters
can be made along the orthogonal (pressure) direction to calculate the
final parameters corresponding to both the measured temperature and
pressure. From these parameters the relative concentrations of the two
components are calculated.
The precision of mixture determination, δ(wi), (or 1-δ(wi)) is then
given by;
δ(wi) = m.δVs (3)
where m is the local slope of the concentration vs. sound velocity
dependence.
The sound velocity precision, δVs, of modern versions of the
instrument developed for monitoring C3F8 coolant leaks from the
ATLAS silicon tracker into external N2-purged anti-humidity envelopes
is better than 0.05 m s−1 [11].
Returning to the example combination of N2 and C4F10 from Fig. 2,
the slope at a sound velocity of 150 ms−1 is 0.72% per ms−1. A sound
velocity measurement uncertainty of 0.05 ms−1 would result in a
mixture precision of 3.6x10−4.
From the molar concentrations ωi of the two components the
G.D. Hallewell Nuclear Instruments and Methods in Physics Research A 876 (2017) 50–53

Table 1
Molecular weight, refractivity, Cherenkov threshold and sound velocity in various
radiator and passivation gases.

Gas Mol. Wt. (n-1) 0.10−6 (@7 eV [12]), γth Sound velocity (m s−1 @
20 ⁰C, 1 barabs) 20⁰C, 1 barabs) [5]

CF4 88 488 32 178.74

C2F6 138 793 25 137.22
C5F12 238 1750 (40 ⁰C) 17 92.31 (40 ⁰C)
C4F10 188 1500 18 99.78
Ne 20 67 87 448.92
N2 28 334 40 349.10
CO2 44 400 35 266.57

refractivity of the radiator gas mixture, (n-1)rad, can be calculated

according to the Lorenz-Lorentz equation:

(n − 1)rad = ∑i ωi (n − 1)i (4)

where (n-1)i are the refractivities of the components.

Table 1 shows the molecular weights and refractivities for common Fig. 5. β=1 Cherenkov angle comparison (1995-6 runs) between reconstructed ring data
(mainly fluorocarbon) radiator gases and for gases commonly used for (□) in the SLD barrel CRID and angles from sonar-deduced refractive index corrected for
passivation (and occasionally as radiators). atmospheric pressure (•); after [1,2] (a): without sonar control of radiator blend (b): with
control.

2. Active control of radiator gas refractivity

In the SLD barrel CRID application ultrasonic binary gas analysis
also provided feedback to dynamically mix two gases being injected
into the radiator vessel (Fig. 4). The SLD barrel CRID, used a radiator
blend of 87% C5F12 (perfluoro-pentane) and 13% N2. This recirculator
was unique in its use of continuous thermodynamic (condensation/
evaporation) fluorocarbon circulation (rather than condensation dur-
ing recovery only). C5F12 from the blend leaving the radiator vessel was
condensed through heat exchange with cold N2 gas, evolving from the
cooling loops of a liquid argon calorimeter, and held in a tank at
−80 °C, where its saturated vapour pressure was a few mbar, while the
N2 was allowed to escape.
Blend control based on ultrasonic mixture analysis maintained the
β=1 Cherenkov ring angle to long-term variation better than ± 0.3%
(Fig. 5). The ultrasonic gas analyzer simultaneously acted as a
flowmeter in a “pinched axial” geometry. The radiator gas composition
and hence the refractive index was also monitored ultrasonically at
multiple points at different heights in the radiator vessel. This system
revealed significant hydrostatic stratification in the early stages of
commissioning when N2 passivation gas was being replaced by the
active radiator. This stratification was reduced by repositioning the
supply and recovery ports on the radiator vessel.
3. Recent improvements in simultaneous binary gas analysis
and flowmetry
Recent advances using microcontroller-based electronics have led
to ultrasonic instruments capable of simultaneously measuring very
high gas flow rates and binary mixture composition for the fluorocar-
bon evaporative cooling systems of the ATLAS Inner Detector [11,13].
Sound transit times, tup and tdown, are measured in opposite
directions in flowing gas and used with geometric parameters to
calculate the flow rate. For this it is necessary to know the lengths L
and L′ of the acoustic path in moving and static gas respectively. Eqs.
(5) and (6) illustrate the example of a high-flow, low impedance
flowmeter with an acoustic path crossing gas flowing in a tube of
diameter DMain at an angle α. The acoustic segment, L, in flowing gas
is then DMain/sin a. The gas flow velocity, vg (m s−1), can be
calculated using the measured sound velocity, vs, as;
DMain
vs ( sinα
+ L′ − vs tdown )
vg =
cosα (vs tdown − L′) (5)

and
DMain
vs (vs tup − sinα
− L′)
vg =
cosα (vs tup − L′) (6)

Fig. 4. SLD barrel CRID multiple sonar installation for mixing C5F12 and N2.before
injection into the gas radiator, and for measurement of refractive index in the radiator
vessel. [1,2].
The algebra of Eqs. (5) and (6) has been integrated into an angled
flowmeter to measure C3F8 vapour flow up to ~1.2 kg s−1) with a
precision of around 2% F.S. and can also monitor C3F8/C2F6 blend
composition to better than ± 3×10−3. The algebra of equations (5) and
(6) is adaptable to ultrasonic flowmeters in the common geometries
illustrated in Fig. 6. In each case the total acoustic path length can be
expressed as distances, L, in moving gas and L′ in static (or closed
circulating vortices of) gas. In some cases these lengths are best found
from Computational Fluid Dynamics simulations [11].
Geometry (c) of Fig. 6 corresponds to that of a compact ultrasonic
flowmeter used in anaesthesia. In a related application a gas analysis
capability is being added to the instrument, incorporating a stored
database of concentration vs. sound velocity for a xenon-oxygen gas

52
G.D. Hallewell
Fig. 6. Illustration of 3 geometries of ultrasonic flowmeters with illustration of the
acoustic path segment, L, in uniformly moving gas.
combination for xenon-based anaesthesia. A mixture precision require-
ment of 10−3 results in a required sound velocity uncertainty of around
0.14 m s−1 for a sound velocity range between 175 and 330 m s−1 at
25 ⁰C, which should be attainable.
53
Nuclear Instruments and Methods in Physics Research A 876 (2017) 50–53
4. Conclusions
The ultrasonic binary gas mixture analysis techniques used for
Cherenkov radiator refractometry, initially at the SLD CRID, have been
recently refined with new microcontroller technology and algorithms to
combine simultaneous flowmetry in a variety of ultrasonic cell geome-
tries. The technique can measure the concentration of light residual
passivation gas in a radiator gas to a precision better than 10−3. These
duel-purpose instruments can be incorporated into new and upgraded
gas Cherenkov detectors, and other applications – including Xenon-
based anaesthesia.
References
[1] K. Abe, et al., IEEE Trans. Nucl. Sci. 45 (1998). http://dx.doi.org/10.1109/
23.682601.
[2] J. Vavra, Presentation at the retirement symposium, for Prof. David Leith: SLAC,
December 2014 〈https://portal.slac.stanford.edu/sites/conf_public/leith/
presentations/Vavra.pdf〉
[3] G. Hallewell, et al., Nucl. Instr. Methods. Phys. Res. A 264 (1988) 219.
[4] E. Lemmon, M. Huber, M. McLinden, REFPROP Standard Reference Database 23
(version 9.0), U.S. National Institute of Standards & Technology, Boulder, CO
80305, USA, 2010.
[5] E. Allbrecht, et al., Nucl. Instr. Methods. Phys. Res. A 502 (2001) 266.
[6] A. Papanestis, Nucl. Instr. Methods Phys. Res. A766 (2014) 14.
[7] M. Andrieux, et al., Nucl. Instr. Methods Phys. Res. A371 (1996) 259.
[8] F. Fontanelli, The detector control system of the LHCb RICH detectors conference
record, in: Proceedings of IEEE Nuclear Science Symposium: 〈http://ieeexplore.
ieee.org/document/1596368/〉, 2005.
[9] Now marketed as SensComp model 600 transducer, 〈http://www.senscomp.com/
ultrasonic-sensors/〉
[10] T. Blake, et al., Nucl. Instr. Methods Phys. Res. A 791 (2015) 27.
[11] R. Bates, et al JINST 8 P02006, 〈http://iopscience.iop.org/article/10.1088/1748-
0221/8/02/P02006/pdf〉, 2013.
[12] E. Nappi, J. Seguinot, Riv. Del. Nuovo Cim. 28 (8–9) (2005) 1.
[13] R. Bates, et al., IEEE Trans. Nucl. Sci. 61 (2014) 2059.