9 views

Uploaded by Rauf Aliev

Does shaking increase the pressure inside a bottle of champagne?

- Plug Flow Reactor Module Using Matlab
- D 6558 â€“ 00 ;RDY1NTGTMDA_
- CH 353 Notes-1
- The Hydrophobic Effect
- template_ije.doc
- 1985 Flow Characteristics of Downflowbubble Columns
- Chemical Engineering Mass Transfer Notes
- Problem 11 - Optimum Slug Size
- 2010 Test 1
- CO2 vimp
- Fizzy Balloon Extractor
- 19245.pdf
- Marine Bilge water treatment system
- Properties of Pure Substances
- 1grade 8 june ME 2018-2019
- LT catalyst reduction
- Sensitivity of Pipeline Gas Flow Model
- Rigorous simulation and design Plate Column.pdf
- States of Matter
- 11th Chemistry - Unit 1,6,7 Question Paper - TamilNadu TN State Board English Medium - Brainkart.com

You are on page 1of 12

www.elsevier.com/locate/jcis

Regular Article

A. Vreme a,b, B. Pouligny a, F. Nadal c, G. Liger-Belair b,⇑

a

Centre de Recherche Paul Pascal, CNRS, 115 avenue Schweitzer, 33600 Pessac, France

b

Equipe Effervescence (GSMA), UMR CNRS 7331, Université de Reims Champagne-Ardenne, BP 1039, 51687 Reims, France

c

Commissariat à l’Énergie Atomique, 33114 Le Barp, France

a r t i c l e i n f o a b s t r a c t

Article history: Colas, beers and sparkling wines are all concentrated solutions of carbon dioxide in aqueous solvents. Any

Received 28 July 2014 such carbonated liquid is ordinarily conditioned inside a closed bottle or a metal can as a liquid–gas 2-

Accepted 9 October 2014 phase system. At thermodynamic equilibrium, the partial pressure of carbon-dioxide in the gas phase

Available online 22 October 2014

and its concentration in the liquid are proportional (Henry’s law). In practical conditions and use (trans-

port, opening of the container, exterior temperature change, etc.), Henry’s equilibrium can be perturbed.

Keywords: The goal of this paper is to describe and understand how the system responds to such perturbations and

Henry’s equilibrium

evolves towards a new equilibrium state. Formally, we investigate the dynamics around Henry’s equilib-

Carbonated beverages

Molecular diffusion

rium of a closed system, through dedicated experiments and modeling. We focus on the response to a

Bubble dynamics sudden pressure change and to mechanical shaking (the latter point inspired the article’s title). Observa-

tions are rationalized through basic considerations including molecular diffusion, bubble dynamics

(based on Epstein–Plesset theory) and chemi-convective hydrodynamic instabilities.

Ó 2014 Elsevier Inc. All rights reserved.

that occur on the rather small scale of laboratory experiments, in

The capacity of carbon dioxide to get dissolved in aqueous volumes on the order of a liter. Most of our experimental work is

media is of paramount importance, in nature and industrial prod- done with pure water and CO2 but the study is relevant to common

ucts and productions. At large scale, CO2 ﬂuxes between the atmo- situations involving carbonated beverages. Beyond the chemically

sphere and oceans are determinant in controlling the general most simple case of a water–carbon dioxide solution (model sys-

temperature of the planet, and solutions for storing huge amounts tem), our experiments and analysis equally apply to real sparkling

of the gas into aquifer reservoirs [1] are currently the matter of beverages such as champagne, beers and colas. The system under

intensive search. How much gas can be stored, in conditions of consideration is a closed bottle of water–carbon dioxide solution

constant temperature T, is ruled by the well-known Henry’s law, (either the model system or a real beverage). We address a few fre-

which states that the equilibrium concentration of dissolved CO2 quently asked and apparently simple questions such as: (i) How

is proportional to the partial pressure of gas P: fast the liquid inside can ingest or release carbon dioxide in a sit-

uation where the thermodynamic equilibrium between vapor

c ¼ kH P ð1Þ and liquid phases is perturbed? (ii) What is the consequence of

shaking the bottle on the pressure inside the bottle and the later

kH in Eq. (1) is Henry’s constant and c is CO2 mass concentration. By occurrence of effervescence and gushing, when the bottle is

deﬁnition, Henry’s law only applies to a static conﬁguration, which, opened? As we will see, though the above questions may appear

strictly speaking, requires an inﬁnite equilibration time to be real- childish, the involved phenomena may not be that simple, and

ized. In practical conditions, true equilibrium is not met, meaning some of the answers are not intuitive. It may come as a paradox

that the system is constantly at ﬁnite distance from a true equilib- to the reader that such questions have not been thoroughly exam-

rium conﬁguration. Real situations, from global climatic phenom- ined and solved decades ago. Nevertheless and rather surprisingly,

ena down to micrometer-sized living cells, involve CO2 exchanges they have not been, as far as we can tell based on our literature

between phases, a gas phase and a liquid phase in simplest search, particularly in the ﬁeld of champagne chemical-physics

situations. [2–6]. In this context, the reported research should also have a

didactical interest, beyond the ﬁeld of sparkling beverages

⇑ Corresponding author. engineering.

E-mail address: gerard.liger-belair@univ-reims.fr (G. Liger-Belair).

http://dx.doi.org/10.1016/j.jcis.2014.10.008

0021-9797/Ó 2014 Elsevier Inc. All rights reserved.

A. Vreme et al. / Journal of Colloid and Interface Science 439 (2015) 42–53 43

and methods. We explored the above questions with dedicated

experiments, with specially designed laboratory hardware, some

of them with real champagne bottles. In Section 3 we shortly

review basic concepts of gas solution equilibrium thermodynam-

ics, essentially Henry’s law for carbon dioxide in water. The section

provides quantitative bases on the amount of dissolved carbon

dioxide as a function of both pressure and temperature. Henry’s

law is illustrated in the case of a bottle of champagne that has been

left at rest and constant temperature. In Section 4 we address the

related question of how fast Henry’s equilibrium can be reached,

in practical conditions, and discuss the roles of molecular diffusion

and convection. We ﬁrst explore the case of super-saturation, i.e.

when the gas pressure on top of the bottle is smaller than Henry’s

pressure. Next, we tackle the case of under-saturation. In both sit-

uations the perturbation is induced by applying a small pressure

step to the system. The objective is to assess the time it takes for

the pressure inside the bottle to come back to the new equilibrium

value. As we will see, the pressure proﬁle measured for carbon

dioxide rejection observed after a negative pressure step (super-

saturated solution), is consistent with a purely diffusive model,

provided the whole system is kept at constant temperature. The

case of a positive pressure step (under-saturated solution) is more

difﬁcult to interpret as the transient situation of an aqueous solu-

tion ingesting carbon dioxide through a free ﬂat interface is

mechanically unstable [7].

Section 5 is dedicated to effects of shaking, starting from a solu-

tion that is initially close to equilibrium. Essentially we report

experiments where the pressure in the bottle neck is measured,

Fig. 1. Sketch of the experimental setups (a) aluminum container ﬁlled with water/

in parallel with observations of bubbles produced by the mechan- carbon dioxide solution. (b) Glass bottle ﬁlled with champagne. The cork has been

ical perturbation. Contrary to the belief that shaking increases the replaced by a home-made pneumatic connector and a small optical window has

pressure, we show that the main effect is a small transient pressure been ﬁtted to the bottle neck (optical axis in dashed line).

drop. As we explain, the drop may be understood through simple

considerations about bubble kinetics, based on Epstein–Plesset

theory [8]. type). P and T signals are numerized using an acquisition ampliﬁer

Our main results are summarized in the conclusion, together MX 840 A (HBM, Germany), and recorded on a PC computer with

with a few prospects. We end the paper with a short discussion Catman software.

on necessary conditions for the onset of effervescence when the The metal container has a capacity V T ¼ 1:27 L, with a circular

bottle is uncorked. Supplementary information is provided at the cross section S ¼ 63:6 cm2 . The total height (20 cm) is shared

end of the paper. Appendix A is a supplement to Section 4; essen- between a liquid phase of height hL , and a gas phase on top, of

tially the detailed mathematical resolution of the diffusion model height hG . Both heights were varied among different experimental

of pressure kinetics. Useful information on Epstein’s article runs.

about bubbles dynamics in water/gas solutions is reminded in Preparing a carbonated solution is straightforward, as we only

Appendix B. have to inject CO2 under a few bar pressure. While keeping the

solution under constant CO2 pressure (the intake valve remains

open), we just wait for the gas to get dissolved until equilibrium

2. Materials and methods is achieved. Purely diffusive dissolution is extremely slow, as we

discuss in Section 4. Not surprisingly, the process can be much

Experiments about pressure kinetics consist in recording the accelerated if the liquid is stirred, simply by vigorously shaking

pressure P inside a container as a function of time t, for instance the container. In our standard procedure, the container is located

when the bottle is subject to a sudden perturbation of Henry’s inside a water bath, whose temperature T 0 is controlled within

equilibrium. Such experiments can be made using a simple airtight 0.02 K. To stir the solution, the container is pulled out of the bath,

container. We used two different types of container/solution cou- shaken several times over a 12-h period, and the intake valve is

ple: most of the quantitative experiments have been performed closed. The system is then left at rest back in the bath for about

using a cylindrical metal tank (cf. Fig. 1a) ﬁlled with pure water/ 48 h. After this preparation, the solution is supposed to be

carbon dioxide binary solution. Some qualitative observations close to equilibrium, though it is never strictly so, as discussed in

has been done on a real glass champagne bottle (cf. Fig. 1b) ﬁlled Section 5.

with champagne (Trouillard Elexium). A ﬂat optical window was In the case of the real system, the aqueous solution (cham-

ﬁtted to the champagne bottle neck to view bubbles, see Fig. 1b. pagne) was already saturated. We had to replace the cork of the

In the following, the ﬁrst system is referred to as model system bottle with a home made pneumatic device (see Fig. 1b), connected

and the second one is referred to as real system. to the same kind of pressure sensor as in the model system. The

The lid of the metal tank is equipped with three ports, for gas original cork was gently removed, to avoid foaming, and was

admission and purge, and for pressure and temperature measure- immediately replaced by the pneumatic device. The loss in carbon

ments, see Fig. 1a. The pressure sensor (Druck Unik 5000, GE, US) dioxide caused by the brief opening of the bottle was negligible, as

is attached to the exterior of the container, and the temperature we could check from the value of the pressure at rest, about 6 bar

is measured inside the gas phase by means of a thermocouple (K at room temperature. The volumes of gas and liquid phases inside

44 A. Vreme et al. / Journal of Colloid and Interface Science 439 (2015) 42–53

He ¼ ; ð5Þ

respectively. M

In principle, the experimental system should be perfectly gas-

proof. In reality the pneumatic technology used in our experiments M being the molar mass of carbon dioxide. Eq. (4) gives the pressure

had very small but measurable leaks, which caused drifts in pres- corresponding to Henry’s equilibrium as a function of the total

sure on the order of a millibar per hour, in typical conditions of amount of carbon dioxide in the container.

6 bar-CO2 pressure. The leak comes as a limitation in long-term Note that P depends on the temperature, directly and indirectly

observations, but, fortunately enough, has no visible inﬂuence on through the dependence of kH on T. Henry’s constant monoto-

the scale of few-hour recordings; which turned out enough for nously decreases with T following a Van’t Hoff-like equation:

our purposes, as we will see.

A few complementary experiments were dedicated to viewing

0 DH D 1 1

ﬂuid convection driven by CO2 dissolution, see Section 4. These kH ðTÞ ¼ kH exp ; ð6Þ

experiments were carried out with a specially designed cubic con-

R T T0

tainer equipped with ﬂat transparent windows, see Ref. [9] for

0

technical details. In this case a minute amount of ﬂuorescein dye where kH is Henry’s constant of CO2 at T 0 ¼ 298 K, and DHD is the

(1.6 106 M L1) was added to the water solution as a marker dissolution enthalpy of CO2 in the liquid of interest (in J mol1).

of CO2 dissolution within the ﬂuid. The method exploits the fact Illustrative values of parameters entering the Vant’ Hoff for-

that the ﬂuorescence intensity of the dye strongly decreases when mula can be obtained from Agabaliantz data [13] about champagne

0 1

the solution pH becomes acid [10–12], as in the case of carbonated and other sparkling wines: kH ’ 1:21 g L1 bar ,

1

solutions [11,10,12]. The effect is very intense with ﬂuorescein; DHD ¼ 24800 J mol . The temperature dependences of kH and P

CO2 rich regions show up as dark zones in the ﬂuorescence image within a typical 75 cL champagne bottle is shown in Fig. 2 (data

of the solution, even under CO2 pressures as small as 0.1 bar. from ref [14]). The concentration of dissolved gas can be deduced

Examples of such images are given in Section 4.4. from Fig. 2 using Eq. (1). It is worth noting that the concentration

of dissolved carbon dioxide is only slightly temperature-depen-

3. Thermodynamic equilibrium dent, whereas the pressure is much more sensitive (about

200 mbar K1).

In this section, we address the ideal situation of thermodynamic

equilibrium of a carbonated solution. Though we deal with well-

known concepts, we want to set out basic relations prior to moving 4. Henry’s equilibrium recovery

to non equilibrium phenomena. We consider a closed container

partially ﬁlled with water, and derive the expression of the partial We now suppose that the equilibrium of the liquid–gas system,

pressure of carbon dioxide in the gaseous phase, given the total initially at pressure P 0 and at c0 ¼ kH P 0 concentration, is suddenly

number of CO2 molecules. We start from Eq. (1), where c is the perturbed by quickly changing the pressure to a new value,

mass concentration of dissolved CO2 molecules, P is the partial P1 ¼ P0 þ DP. The perturbation is achieved either by injecting or

pressure of carbon dioxide in the vapor phase. kH , the Henry’s con- removing a small amount of gas, through the purge-intake port

stant (usually in g L1 bar1) is a function of temperature, and of the container (see Fig. 1). According to whether DP > 0 or

depends on the particular composition of the liquid. DP < 0, the liquid becomes under- or super-saturated, respectively,

In the whole paper, we suppose that the gas phase is composed of and the system starts evolving towards a new Henry equilibrium.

carbon dioxide only, such that the total pressure P and the carbon We want to know the characteristics of the ﬁnal state and how

dioxide partial pressure P CO2 are considered as identical (this is not long it takes for the system to reach it.

rigorously true since the gas phase is a mix of all the components In paragraph Section 4.1, we give the expression of the ﬁnal

present in the liquid phase, each of them being at saturation pressure. Next, we present a simple diffusion-based model for

pressure). the pressure relaxation, within the simplifying assumption that

We consider now a closed cylindrical container with a volume there is no convection (see Section 4.2). In Section 4.3, results of

V G of gas on top of a volume V L of liquid, with a ﬂat horizontal experiments concerning the dynamics of dissolution (or release)

interface, of surface S, in between. For simplicity, we suppose that of carbon dioxide in water are reported.

both volumes remain constant (we neglect the minute changes of

the liquid volume due to gas dissolution or rejection). The total

number of moles of carbon dioxide nT is a conserved quantity that 3.0

decomposes into nG moles in the gaseous phase and nL moles in the

8

liquid:

2.5

kH (g L−1 bar−1 )

nT ¼ nG þ nL : ð2Þ

7

PCO2 (bar)

suppose that equilibriums in gaseous phases are ruled by the ideal 2.0

gas law. Then:

6

PV G ¼ nG RT: ð3Þ

1.5

where R is the ideal gas constant (8.31 J K1 mol1). Combining Eqs.

5

(1)–(3), we obtain:

1.0

nT RT 5 10 15 20

P¼ ; ð4Þ

V G þ HeV L T (◦ C)

In the above equation, He stands for ‘‘Henry number’’, deﬁned Fig. 2. Typical temperature dependences of kH and P within a 75 cL champagne

by: bottle. Pressure: solid line and left ordinate axis. Henry’s constant: dashed line and

right ordinate axis.

A. Vreme et al. / Journal of Colloid and Interface Science 439 (2015) 42–53 45

h

s¼ 2

¼ ; ð14Þ

ðHeÞ D D

The ﬁnal asymptotic value of the pressure after equilibrium

recovery is denoted by P 1 . Following the same reasoning as in Sec- We recall that, in Eq. (12), hG ¼ V G =S is the height of the gaseous

phase. The equivalent liquid height is deﬁned as h ~ ¼ pﬃﬃﬃﬃ

phG =He. Then

tion 3, we obtain (for t ! 1Þ:

s, given by (14), may be thought of as the time taken by CO2 to dif-

1 ~

VL fuse in the liquid over the distance h.

P1 ¼ P0 þ DP 1 þ He : ð7Þ Let us put numbers in Eq. (14) to get some useful orders of mag-

VG

nitude. The Henry number He ¼ 0:83. With D ¼ 1:8 109 m2 s1

Note that, when the volume of liquid is much larger than the (see Refs. [15–18]), kH ’ 1:5 105 kg m3 Pa1 (see Refs.

volume of gas, the pressure returns back to its initial value [19,20]), and hG ’ 5 cm, we obtain ~ ’ 0:11 m

h and

(P 1 ¼ P 0 ). s ’ 6:3 106 s. We thus come out with a very long time (about

73 days!), but this number may be misleading in suggesting that

4.2. Dynamics of recovery the diffusive process is extremely slow. In fact, the square root

term in Eq. (12) is singular at t ¼ 0, corresponding to an inﬁnite

We now elaborate a simple model based on the following time derivate. The pressure in fact is predicted to very rapidly vary

assumptions: at short time. The short time limit, according to (13), corresponds

to t 6 7 103 s.

(a) The gas phase is permanently at equilibrium, meaning that

the ideal gas law is veriﬁed at all times; 4.3. Response to decompression

(b) The temperature T is uniform and constant;

(c) Dissolution only proceeds by molecular diffusion (convec- In this paragraph, we examine the effects of a small negative

tive effects in the liquid phase will be discussed further on); pressure step (DP < 0), meaning that we decrease the pressure

(d) The growth time of the pressure step is much smaller than inside the container down to a prescribed value by opening the

any other characteristic time of the system (in other words, purge valve for a short while dt, around t ¼ 0, conventionally. We

the pressure step is achieved instantaneously on the time observe the response in P to this perturbation. In some of the

scale of diffusion); experiments we record both pressure and temperature responses,

(e) Saturation is immediately reached at the liquid–gas inter- see Fig. 3 for illustration. The ﬁgure indicates that the system

face. Consequently, Henry’s law is permanently satisﬁed at responds in two steps. The ﬁrst one is a short-lived response, in

z ¼ 0, and the shift in CO2 mass concentration, Dc ¼ c c0 , the form of negative peaks in both P and T, on the order of a minute

is related to the shift in pressure, DP ¼ P P0 , by the in duration. In the second step, the temperature is back to the con-

equality stant value imposed by the thermostat (T0 = 20.35 °C), and the

pressure linearly increases as a function of t1=2 . Qualitatively, the

Dc ¼ kH DP on Sðz ¼ 0Þ: ð8Þ latter response is in line with the prediction of the diffusion model,

Dc obeys the classical one-dimensional (1D) linear diffusion see Eq. (12).

equation The drops in P and T are very likely due to the direct thermal

response of the gas phase to the decompression. If dt is short

@ t Dc ¼ D@ zz Dc; ð9Þ enough, the decompression is close to adiabatic. The resulting tem-

perature drop dT may be estimated as:

where D is the diffusivity of carbon dioxide in the liquid of interest.

The diffusion equation must be completed by the boundary condi-

tion (8) – assumption (e) – and a condition of vanishing ﬂux at

the bottom of the tank:

P0 =5.4 bar

@ z Dc ¼ 0 at z ¼ hL ; ð10Þ

5.12

During the relaxation process, the molecules of carbon dioxide P2

diffuse through the interface from the gas phase into the liquid 5.10

(under-saturation, DP > 0) or in the opposite direction (over-satu-

P (bar)

ration, DP < 0). Because the system is closed, dissolution or rejec- 5.08

tion of CO2 directly modiﬁes the pressure in the gas phase through δPe

a feed back process. The system formed by Eqs. (9), (8) and (10) is 5.06

closed when adding the conservation equation:

5.04

dDnG dDnL P1

¼ for t > 0þ : ð11Þ

dt dt 20.5

In Appendix A, we offer a full resolution of the above system, in

T (◦ C)

δTe

are such that only the short time limit of the general solution is

19.5

of practical interest. We ﬁnd:

h i 0 20 40 60 80 100

P ¼ P0 þ DP 1 ðt=sÞ1=2 ; ð12Þ t1/2 (s1/2 )

which is valid in the limit Fig. 3. Pressure and temperature response to a small negative pressure step

(experiment #6 in Table 1). P and T are plotted versus t1=2 for direct comparison

4 102 with the diffusion model, Eq. (12). The ﬁgure shows pressure values between 5.03

t6 s; ð13Þ and 5.13, to focus on details of the pressure response at ‘‘short times’’ (Eqs. (12) and

p2

(13)). The arrow on top indicates the initial pressure, P 0 ¼ 5:4 bar, which is well

In (12) and (13) s is a diffusion time, given by: above the represented interval.

46 A. Vreme et al. / Journal of Colloid and Interface Science 439 (2015) 42–53

" ð1cÞ=c #

P1 0.05

dT ¼ T 0 1 ; ð15Þ

P0

0.04

(P − P2 )/ΔP

" ð1cÞ=c #

P1 0.02

dP ¼ P1 1 : ð16Þ

P0

0.01

Estimates from the above equations, with c ¼ 1:28, are in line

with the amplitudes of the short time responses in experiments 0.00

where decompression was operated ‘‘brutally’’, meaning that dt

was on the order of a few seconds. In the example shown in -0.01

Fig. 3, decompression was operated more gently, in about 2 min. 0.01 0.02 0.03 0.04

The process then was intermediate between adiabatic and isother- (t/τe )1/2

mal, and the drops in P and T are inferior to values (0.08 bar and

4 K, respectively) estimated with Eqs. (15), and (16). Fig. 4. Dimensionless pressure ðP P 2 Þ=DP as a function of ðt=se Þ1=2 (se being the

An experimental characteristic time, denoted as se , can be experimental characteristic time), for the experiments listed in Table 1. Note that

the time has been rescaled using a different se for each experiment (cf. Table 1).

deduced from the slope of the linear response in Fig. 3. Because

of the short-lived thermal response, the amplitude of the pressure

perturbation to be retained for comparison with the diffusion The source of convection is most likely attributable to the ther-

model is not the pressure drop imposed by the experimenter, mal perturbation, which causes transient gradients of the ﬂuid

DP ¼ P1 P0 , but the drop corrected for the thermal effect, density. Density gradients have a vertical component, which may

DP ¼ P2 P0 (see Fig. 3). Results from experiments carried out with be either stabilizing or destabilizing (as in the classical Rayleigh-

different values of the pressure drop are gathered in Table 1. The Bénard-Taylor problems [22,23]), and horizontal components.

table indicates the measured se next to the theoretical diffusion Because the bottle and the liquid inside have ﬁnite thermal con-

time calculated using Eq. (14), and the corresponding DP. Using ductivities, temperature variations (dT), whatever their origins,

the parameters listed in the table, we may transform the time cannot be uniform in horizontal sections of the system. This situa-

and the pressure into dimensionless variables, ðt=se Þ1=2 and tion, involving a horizontal temperature gradient, causes a hori-

ðP P 2 Þ=DP, respectively. Results from the whole set of experi- zontal gradient in density of the liquid, which is unstable to

ments can then be gathered into a single graph, as shown in convection, whatever the amplitude and the sign of dT.

Fig. 4. Note that all data merge onto a unique curve, meaning that The amplitude of the convection may be hardly perceptible if

the experimental records for pressure relaxation are indeed well decompression has been operated ‘‘mildly’’, meaning that cooling

represented by Eq. (12). However the agreement is only qualitative has been minimal, resulting in very small gradients. Of course

because se in general is less than what we expect based on Eq. (14), the thermal response might be about eliminated if decompression

see Table 1. Quantitative agreement is met only for the smallest is operated very gently (dt ! 1). But the pressure step has to be

amplitude of the pressure drop and mild operation, as in Fig. 3. fast on the scale of the time limit set by Eq. (13) for the t 1=2

When the perturbation is large and sudden, the pressure response response to be observed. Due to the latter requirement, dt cannot

still seemingly follows the diffusion law, Eq. (12), but with an be larger than about 2 min. Completely eliminating convection in

anomalously short characteristic time. The anomaly suggests the real experiments may then turn elusive.

existence of a convective ﬂow, that inevitably accelerates the gas

exchange through the interface. However, in the case of decom- 4.4. Response to compression – chemi-convection

pression, diffusion should not be destabilizing, essentially because

the density of carbonated water increases with the concentration We now turn to the response to a positive pressure step, mean-

of CO2 , see Ref. [21]. In response to the pressure drop, molecular ing that a ﬁnite amount of gas is injected through the intake valve

diffusion has the effect of decreasing the amount of CO2 in the of the container (Fig. 1). Experimental records of the pressure

layer below the interface. The density of this layer is then less than responses in this case show complex behaviors, far from the rather

that of the bulk solution underneath. Therefore the diffusive ﬂux of simple picture of the former paragraph on decompression. Com-

CO2 , from the liquid to the gas phase on top, results in a density pression ﬁrst heats the gas phase, through the same mechanism

proﬁle that monotonously decreases from the bottom to the top as that described before, but with an opposite sign. Heating of

of the liquid. Such a proﬁle is stable[22,23], meaning that it cannot the liquid through the interface now leads to lowering the density

promote convection. on top of the liquid phase. Adiabatic – or nearly so – compression

then has a stabilizing effect, and cannot cause convection.

In the compression experiment, carbon dioxide is added into

Table 1

Table of the characteristic time se measured for each experiment. The experimental

the system and progressively gets dissolved in the liquid phase.

and theoretical adiabatic pressure drops (denoted by dPe and dP resp.), together with The process starts by molecular diffusion, creating a diffuse layer

the expected theoretical time s and the experimental adiabatic temperature drop that is heavier than the ﬂuid underneath. The resulting unstable

dT e ¼ T 1 T 0 are given for information. density proﬁle is the source of convection [24,25], a process known

# DP (bar) se (s) s (s) dP e (bar) dP (bar) dT e (K) as ‘‘chemi-convection’’. Contrary to the case of decompression, the

source of chemi-convection is intrinsic to carbon dioxide dissolu-

1 0.092 2:3 10 6

4:6 106 0.0078 0.0217 –

tion, and then subsists as long as complete dissolution has not been

2 0.165 2:1 106 4:6 106 0.0349 0.0432 –

achieved [7,9]. In practical conditions, the heavy diffuse layer gets

3 0.250 1:9 106 4:6 106 0.0512 0.0643 –

destabilized within a few seconds; consequently the diffusive

4 0.106 9:6 106 9:64 106 0.0032 – 0.107

5 –

regime of pressure relaxation is not observable.

0.215 7:4 106 9:64 106 0.0026 0.122

6 – Convective motion of the ﬂuid cannot be directly viewed

0.414 5:1 106 9:64 106 0.0021 0.113

through the opaque metal container, and even not in the cham-

A. Vreme et al. / Journal of Colloid and Interface Science 439 (2015) 42–53 47

pagne bottle, which has not enough optical quality. In this case we 5.05

used the special cubic container [9] mentionned in Section 2. We

observed a sample of water inside a parallelepipedic glass cell 5.04

(width a ¼ 70 mm, thickness b ¼ 5 mm, height c ¼ 50 mm) that

was exposed to positive pressure steps of carbon dioxide, from

0.2 bar to 0.8 bar in amplitude. The cell was illuminated by a laser 5.03

P (bar)

sheet (wavelength 514 nm), parallel to (a, c)-sides to excite the

ﬂuorescent dye. 5.02

This type of experiment is limited to pure water as the initial

state of the liquid, and to volumes much inferior to a liter, but

5.01

has enough generality as it clearly shows the onset and evolution

of chemi-convection. Main features should be similar starting from

Henry equilibrium states with ﬁnite concentrations of CO2 and lar- 5.00 (a)

ger volumes. Fig. 5 shows a typical Rayleigh–Taylor instability that

70 80 90 100 110 120 130

has evolved into characteristic ﬁngers, called ‘‘plumes’’. The dark

5.030 P0

features are CO2 -rich zones, while clear zones correspond to about P∞

pure water. The dark horizontal band on top of the photos is the

P (bar)

diffuse layer, which is strongly unstable, as we explained. Plumes ΔPs

5.025

are the source of general convection inside the cell, and convection

acts back on them, producing complex patterns, as the ﬁgure

illustrate.

(b)

want to investigate the effect of vigorously shaking the bottle con- 20.2

taining the carbonated solution. A popular belief is that the pres- (c)

sure increases when the bottle is shaken, and that this is the 70 80 90 100 110 120 130

reason for gushing of the liquid when the bottle is opened. We t (s)

show in this section that shaking in fact produces a small but oppo-

site effect, in the form of a transient pressure decrease. We start Fig. 6. Typical responses to shaking recorded with water in the aluminum tank

(model system). (a) Raw pressure signal; (b) ﬁltered pressure signal; and (c)

with experimental observations and move to qualitative interpre-

temperature signal.

tations afterwards.

5.1. Experimental observations stands for ‘‘shaking’’), followed by a monotonous raise up of P over

about a minute, up to a plateau value denoted as P1 .

We tested the effect of shaking with champagne, sparkling Different experiments of the same type, with different contain-

white wine (‘‘Blanc Foussy’’), and simply water. The signals shown ers and liquids, revealed that P1 in general differs from P 0 , but the

in Fig. 6 were obtained with the metal container (Fig. 1) and car- difference may be of positive or negative sign. What seems to be a

bonated water, which was initially at equilibrium (as far as it could non reproducibility in the sign of the difference between the ﬁnal

be, as we explained in Section 2) at about 5 bar of CO2 pressure. and initial pressures, is very likely due to the fact that the solution

Shaking was operated by hand for about 20 s, and the container is never completely at thermodynamic equilibrium. As the diffu-

was put back in the temperature controlled bath (’20 °C). sion time s is of the order of months, it is practically impossible

The graph in Fig. 6a features an oscillation regime followed by a to start from a real equilibrium (Henry) state. Shaking just helps

smooth evolution of the pressure. Oscillations are located within accelerating CO2 exchange between the gas and liquid phases. As

the period of active shaking of the container, between 70 and a result the solution after shaking is closer to Henry’s equilibrium.

90 s along the time axis. These oscillations are not of primary inter- As a corollary of the latter statement, repeating shaking only a

est as they are most probably a signature of the ﬂuid dynamic pres- short while (less than one hour, say) after the ﬁrst perturbation

sure which raises and drops in the course of shaking. Interestingly, produces about no change in pressure – i.e P 1 ¼ P 0 .

the oscillations can be eliminated by low-frequency ﬁltering of the The graph in Fig. 6c shows a transient temperature decrease, in

signal, resulting in the graph shown in Fig. 6b. The graph reveals an parallel to the pressure transient. The experimental records then

overall decrease of the pressure, DP s ’ 10 mbar (the subscript s suggest that P and T transients have a common cause.

Fig. 5. Images of plumes and convective cells consecutive to positive pressure steps of (a) 0.2, (b) 0.4 and (c) 0.8 bar at t = 10, 7 and 4 s respectively. The horizontal dimension

of each image is 6 cm. Horizontal streaks are optical parasites due to imperfections in the windows of the cubic chamber. See text for details of experimental conditions and

method.

48 A. Vreme et al. / Journal of Colloid and Interface Science 439 (2015) 42–53

P0 , VG,0 P0 , nG,1 , VG,1 P1 , VG,1

it was possible to observe bubbles inside the bottle, through the

nG,0

optical window that was ﬁtted to the bottle neck (see Fig. 1b). VB

We thus could make videos of the bubbles a few centimeters below

the liquid–gas interface, within the liquid bulk. Since the device

only compensates for defects on the exterior interface of the bottle,

the optical quality is not optimal, but bubbles were well discern-

ible and their sizes could be measured from video images within

5 lm in uncertainty.

Within a few seconds after shaking, videos show huge amounts

of bubbles that very rapidly ﬂow up across the images towards the

bottle top. Initially the motion is too fast with too many bubbles to

provide well resolved images (at 25 frames/s), but very large bub-

bles (larger than 100 lm in diameter) can be discerned. The

motion, together with the population of bubbles, deﬁnitely

decreases within about a minute, and largest bubbles disappear. N bubbles

After about 2 min, remaining bubbles tend to roughly the same

size, which we estimate 60 ± 5 lm in diameter. It takes about (a) (b) (c)

220 s for bubbles to completely disappear from video records.

Fig. 7. Sketch of the bubble collapse scenario (no buoyancy). (a) Before shaking. (b)

After shaking: N bubbles have been engulfed in the bulk. (c) After collapse. In (c),

V G;1 is the volume of the gas phase (white region + gray region); the gray region is

5.2. Dynamics of the bubble assembly the volume gain due to bubbles collapse (V B ).

the pressure in the container after the bubble collapse (P 1 ). The

strong correlation between the pressure transient and bubble

amplitude of the pressure variation is:

dynamics. We now come to our proposition to explain the

observed transients. We start with a most simpliﬁed model, that VB

DPs ¼ P1 P0 ¼ 2P 0 ; ð17Þ

hopefully conveys the key ideas about the evolution of the bubble V G;0

population and its consequence on the pressure. This model is

Eq. (17) indicates that bubble collapse has a very sensitive

based on Epstein–Plesset theory for bubble dissolution [8]. In a sec-

impact on the pressure. Suppose e.g. V G;0 ¼ 100 mL, and

ond part we propose a less restrictive version of the model, that

P0 ¼ 6 bar. A very small amount of gas, e.g. V B ¼ 0:05 cm3 initially

takes into account the polydispersity of bubble sizes. This part ends

in the bubble phase is sufﬁcient to produce 50 mbar as the ampli-

with a numerical resolution of the model.

tude DP s of the pressure drop.

The above reasoning was built on the assumption that all bub-

5.2.1. One-bubble-size model bles would collapse and contribute to the pressure decrease. We

Let us consider a single bubble in the liquid bulk. For simplicity, ignored the effects of buoyancy, essentially the rising up of bubbles

we ﬁrst ignore the rising up motion of the bubble due to buoyancy due to the Archimedes force. Bubbles that reach the upper inter-

and suppose that it stays in bulk water. In the absence of capillary face before collapsing hardly contribute to the pressure decrease.

tension (and buoyancy), a bubble would be at Henry’s equilibrium Clearly very small bubbles collapse in very short times, losing the

and would stay stable. The effect of the interfacial tension (r) is to whole volume of their gas into water. These bubbles efﬁciently

increase the bubble internal pressure. As a consequence, the bub- contribute to the pressure drop as estimated by Eq. (17) and the

ble ‘‘sees’’ the surrounding liquid as under-saturated. Surface ten- effect lasts for a long time, because the dissolved gas can return

sion forces the gas content of the bubble to dissolve in the liquid. to the top gas phase only by advection/diffusion through the inter-

The initial state is unstable and the bubble has to decrease in size. face. Conversely, large bubbles survive much longer, leaving them

The process creates an excess of carbon dioxide concentration enough time to reach the top interface and return their gas content

around the bubble. This extra concentration progressively gets directly into the top phase.

diluted and, at long times, relaxes by diffusion across the macro- Let be sc the collapse time of a bubble of initial radius a0 ; sr the

scopic water/gas interface, bringing the system back to Henry’s time it takes for the same bubble to move up over the length hL

equilibrium. If we regard the top gas phase as a giant bubble (if (height of liquid), and aH 0 the limit size for which both times are

convection is neglected), the transfer of gas from the collapsing about equal, explicitly:

small bubbles to the macroscopic gas phase may be viewed as a

case of Ostwald ripening [26]. The important consequence of this sr ðaH0 Þ ¼ sc ðaH0 Þ ¼ sH : ð18Þ

process is that the gas contained in the bubbles is temporarily Bubbles of radius a0 < aH0 will collapse before reaching the surface

‘‘eaten’’ by the liquid. and will contribute the pressure drop for long times (t sH ). Large

In the following, we consider that the water can accommodate bubbles a0 > aH0 will brieﬂy and slightly contribute the pressure

the molecules from the bubbles with about no volume change (the drop since they will have hardly recessed when they reach the

approximation may be justiﬁed quantitatively). We adopt the fol- surface.

lowing notations, see Fig. 7 for illustration: P0 ; V G;0 ; nG;0 are the We may give estimates of the above characteristic time and size

initial pressure, volume of gas phase and number of CO2 moles in using Epstein–Plesset (EP) theory of bubble dissolution [8]. In the

gas phase; see Fig. 7a. We suppose that shaking instantaneously case of a bubble surrounded by a water/dissolved gas solution at

generates N bubbles, all of them of size (radius) a (see Fig. 7b). Henry’s equilibrium with a gas top phase (coefﬁcient f ¼ 1 in EP

The bubble phase (of volume V B ) and the top gas phase (of volume original paper), the theory gives the following expression for the

V G;1 ) contain nB and nG;1 moles of gas, respectively. recession time (Eq. (41) in the original paper):

We want to estimate the decrease in pressure after the bubbles

have collapsed. Combining equations for volume conservation, 1 a20 a0

sc ¼ 1þ

; ð19Þ

V G;1 ¼ V B þ V G;0 , and mass conservation, nG;1 ¼ nG;0 nB , we obtain 3d D a

A. Vreme et al. / Journal of Colloid and Interface Science 439 (2015) 42–53 49

where d ¼ c=qG is the ratio between the concentration of dissolved choose equal to 1 mm. Initial radii and altitude are denoted as an0

gas in bulk and the concentration in gas in the top phase (qG ). In and zn0 respectively (n being the index of the bubble).

, which has the dimension of a length, is given by:

Eq. (19), a To represent the time evolution of the bubble radius, we use

Eq. (40) of original EP article, which is written as Eq. (B3) in Appen-

2rM dix B. The latter equation only holds approximately in our case of a

¼

a : ð20Þ

qG RT near to Henry equilibrium system, but is sufﬁcient in the context of

our crude model, see the discussion in Appendix B and Fig. 11.

Eq. (19) indicates that the bubble lifetime considerably increases

Given the value an of the radius and the altitude zn at time ti (i

is very small (we

with a0 . In practice, the characteristic length a

being the time index), we deduce the altitude of the bubble at time

estimate a 100 nm). So, except for extremely small bubbles, we

tiþ1 according to:

may approximate Eq. (19) as:

2qL n 2

a30 zn ðtiþ1 Þ ¼ zn ðt i Þ þ a ðt i Þ Dt: ð25Þ

sc ’

; ð21Þ 9g

3dDa

The bubble rise time may be estimated directly from the Stokes The previous equation results from the balance between buoyancy

drag coefﬁcient of a sphere of initial size a0 : and viscous drag [27]. We also implicitly ignored hydrodynamic

couplings between bubbles. This assumption is probably an over-

9ghL

sr ’ ; ð22Þ simpliﬁcation at times shortly after shaking, but hopefully tends

2qga20 to correct in the later stage of the process when only few bubbles

are left.

where the mass density difference between the liquid and gas

The evolution in time of the size distribution Dt ðaÞ is presented

phases has been approximated to the liquid mass density q; g is

on Fig. 8. As might be anticipated, the distribution quickly narrows;

the gravity acceleration and g is the viscosity of the liquid.

after 30 s the size amax at which Dt ðaÞ is extremum (most repre-

Combining the latter expression with Eqs. (18) and (22), we R

sented size) and the mean value hai ¼ Dt ðaÞada are almost the

obtain:

same, as shown in the insert of Fig. 8. The right side of the distribu-

3=5 tion, corresponding to large radii, recedes under the action of buoy-

9ghL

sH ’ ð3dDaÞ2=5 ; ð23Þ ancy. The left side, corresponding to small bubbles, recedes due the

2qg

1=5 collapse process. The value of amax quickly drops down to about

ghL

27dDa 50 lm, which later slowly decays down to 20 lm ðt ’ 200 sÞ, and

aH

0 ¼ : ð24Þ

2qg ﬁnally falls to zero within a few seconds ðt ’ 215 sÞ. Note that

the maximum of the distribution eventually goes to zero together

Using q ¼ 103 kg m3 and the numerical values listed in Table 2

with the total number N of bubbles, as shown in Fig. 9. The [50 lm,

to feed Eqs. (23) and (24), we ﬁnd aH 0 ¼ 33 lm and s ¼ 42 s. As

H

20 lm] interval includes the radius aH 0 provided by the simple

reported in paragraph Section 5.1 A, microscope observations per-

model (33 lm) and is in line with the size observed in the experi-

formed near the interface (z 1 cm) through the glass of the

ment (2a = 60 ± 5 lm). The main improvement of the model, com-

champagne bottle showed that the characteristic diameter of the

pared to the ‘‘one-bubble-size’’ model, resides in our estimate of

bubbles tends towards a limit value of 60 ± 5 lm and we observed

how the bubble population evolves in time. We ﬁnd that it takes

that the last visible bubble would vanish at t 220 s after shaking.

about 30 s for the distribution to focus around the ultimate bubble

The simple model presented above predicts that bubbles smal-

size, and that no bubbles are left after 215 s (extinction time).

ler than 33 lm in radius should get dissolved, and then should not

The relaxation time of the bubble population can be estimated

be observed, while bubbles larger than 33 lm should survive long

from the linear part of the log-lin plot of Fig. 9. We ﬁnd

enough to reach the interface. The model then successfully repro-

duces one main feature of the observations, namely the average

size of the bubbles near the end of the bubble population lifetime. D t (a)

However the estimated time sH beyond which no bubble can 4

survive in the bulk (because they already collapsed or reached 3.0×10

-4

1×10

the free ﬂat surface) seems too small compared to what observa-

tions indicate, namely 200 s, approximately. 2.5×10

4

80 s -5

8×10

a, amax (m)

-5

4 6×10

5.2.2. Improved model 2.0×10

A weakness of the above model stems from the fact that only a -5

4×10

constant radius is envisaged in Eq. (22). We now propose an 1.5×10

4

-5

improved version of the model where the EP collapse dynamics 2×10

Instead of a single bubble, we reason on a size-polydisperse distri- 1.0×10 0.0 1 2 2 2

5.0×10 1.0×10 1.5×10 2.0×10

bution. We consider an ensemble D0 of N ¼ 106 bubbles uniformly t (s)

3

10 s

distributed in the bulk (z 20; hL ½) and whose radii are also uni- 5.0×10 5s

1s 0s

formly distributed between 0 and a maximum size, which we

0.0

-4 -4 -4 -4

0 1×10 2×10 3×10 4×10

Table 2

Numerical values of the parameters used to feed Eqs. (23) and (24). a (m)

3

D ¼ 1:8 109 m2 s1 (Ref. [16]) c ¼ 7:5 kg m Fig. 8. Evolution in time of the size distribution Dt ðaÞ. The initial ﬂat distribution

qG ¼ 8:31 kg m3 d ¼ 1:17 quickly evolves to a narrower distribution whose maximum lies between 50 and

¼ 0:13 106 m

a g ¼ 103 Pa s 20 lm (see text for further details). In the insert are plotted the maximum and the

hL ¼ 0:1 m mean values of the bubble size. We see that both values are almost equal once

g ¼ 9:81 m s2

buoyancy has ‘‘ﬁltered’’ the large bubbles.

50 A. Vreme et al. / Journal of Colloid and Interface Science 439 (2015) 42–53

10

6 to 60 s and 325 s repectively, consistently with the sH / g3=5 scal-

ing provided by the simple model.

5 The conclusion is then that differences in surface tension and

10

τN ∼ 40 s viscosity of champagne, compared to values for pure water, do

change predictions of the model, but not so strongly, in so far as

4

10 the model cannot do better than providing estimates of character-

istic times for bubbles dynamics. The differences may not be signif-

N (t)

10

3

icant in view of the limitations of the model. As an important

simpliﬁcation, we neglected hydrodynamic interactions between

2 the bubbles and the large scale convection due to ascending bub-

10

bles. This type of convection is evident in observations, but only

1

in the ﬁrst seconds after shaking; it is hopefully negligible in the

10 later evolution of the system, when large bubbles have disap-

peared. Another possibly important point, in the case of cham-

0

10 0 pagne, is the effect of the liquid–gas interface rigidiﬁcation,

50 100 150 200

which expectedly plays a role in slowing down the collapse of bub-

t (s)

bles reaching the surface [29]. This mechanism should be taken

Fig. 9. Evolution in time of the total number of bubbles. The characteristic time into account in a more elaborate version the model.

sN 40 s extracted from the linear part of the log-lin curve can be seen as the decay

time of the number of bubbles left in the bulk.

6. Conclusion

sN ’ 40 s, not far from the characteristic time s found with the H

A carbonated water solution inside a closed system, the main

‘‘one-bubble-size’’ model. In fact, neglecting the variation of the

object of our study, is the formal equivalent of a sparkling beverage

drag with the radius, as we ﬁrst did, amounts to make no distinc-

inside a closed bottle. We have investigated how such a system

tion between the extinction time and the decay time sN . Finally,

evolves between a given equilibrium conﬁguration, ruled by

the numerical model provides a satisfying value of the ‘‘extinction’’

Henry’s law, to another equilibrium. As causes of perturbation to

time (215 s), close to what is observed experimentally (220 s).

drive such changes, we investigated the effects of a small variation

of the gas content of the system and of mechanical shaking.

We showed that the system might respond to a pressure change

5.2.3. Discussion essentially through molecular diffusion, in conditions where con-

In the above analysis, we used values of r and g corresponding vection might be reduced to a minimum. We could elaborate an

to pure water. Values for champagne are deﬁnitely different, due to experimental procedure whereby these conditions were satisﬁed.

the presence of active molecules (amphiphilic molecules, proteins, We offered an exact resolution of the diffusion problem in this sit-

alcohol, etc.) [28]. The static surface tension of a ‘‘pure’’ hydroalco- uation, and found that the pressure response would follow a

holic solution at 12.5% of alcohol in volume is about 48 mN m1, square root law in time, within a few hours after the perturbation.

whereas it is about 46 mN m1 for champagne. Ethanol is mainly This prediction was shown to be in agreement with experimental

responsible for this drop (compared to water for which data recorded in the case of gentle decompression

r 70 mN m1 ), the other molecules being responsible for the (DP 0:1 bar). Conversely to conditions for molecular diffusion

small extra drop of 1 or 2 mN m1. In the ‘‘one-bubble-size’’ model, to be the dominant mechanism, we observed that compression of

the surface tension dependence is contained in the quantity a (cf. the system in general leads to chemi-convection. Severe decom-

Eqs. (23) and (24)), which is proportional to r. Consequently, sH pression also drives convection due to transient cooling of the

and aH are proportional to r2=5 and r1=5 , respectively. Thus the gas phase.

dependence on r is weak, and then the bubbles dynamics should The main outcome of the experiments on shaking is that the

not differ much from that with pure water. pressure inside the bottle does not change ‘‘much’’, in so far as

Viscosity also varies according to both the chemical composi- the system initially was close to Henry equilibrium. This observa-

tion of the beverage and the temperature. Viscosity of champagne tion then rules out the common belief that shaking a bottle of

at 20° is about 1.6 mPa s (but goes up to 2.5 mPa s at 4 °C). The vis- champagne increases the internal pressure. Signiﬁcant changes of

cosity has an inﬂuence on the rising time of the bubble up to the the pressure do occur if the system initially was far from equilib-

ﬂat surface and on the shrinking time due to the local undersatu- rium, meaning that stirring the liquid just helps in quickly estab-

ration. This twofold inﬂuence is reﬂected in the expression for lishing the equilibrium pressure. The latter statement can be

the time sH , which is proportional to g3=5 . Note that changing the illustrated in everyday-life with a cola conditioned in a plastic bot-

viscosity from 1 to 2.5 mPa s increases sH by less than a factor 2. tle. The envelope of the bottle initially is rigid due to the high pres-

Using the numerical model, we studied the sensitivity of the sure inside. If the bottle is opened, for instance to serve a glass of

bubbles population dynamics to changes in g and r. We tested the cola, the pressure in the gas phase is temporarily lowered to

two combinations: r ¼ 40 mN m1 with g ¼ 103 Pa s, and the outside room pressure, and the envelope becomes ﬂaccid. It

r ¼ 70 mN m1 with g ¼ 2 103 Pa s. remains so even after the bottle has been closed back. At that step,

We found that decreasing the surface tension has a small but shaking has the immediate effect of restituting the rigidity of the

visible inﬂuence. The characteristic time sN (which is the equiva- envelope.

lent of sH for the improved model) is shifted from 40 to 50 s, and Looking at the pressure response more closely, we discovered

the extinction time is shifted from 220 s to 275 s. Both changes that shaking systematically produces a small transient pressure

are consistent with the sH / r2=5 scaling provided by the simple drop, having a few tens of seconds lifetime. We observed that

model. the latter effect was concomitant to the generation of bubbles.

Increasing the viscosity from 1 to 2 mPa s has a more sensitive We could give an interpretation to the pressure based on

inﬂuence. However, the general shape of the size distribution is not Epstein–Plesset theory of bubble dissolution and collapse. The sim-

changed. The decay time sN and the extinction time are raised up ple analytical and numerical models presented in Section 5.2 were

A. Vreme et al. / Journal of Colloid and Interface Science 439 (2015) 42–53 51

satisfactory in reproducing the experimentally observed evolution where HðtÞ is the Heavyside function. The solution to Eq. (9) com-

of the bubble population, in size and lifetime. However, deriving a pleted by the conditions (8), (A1) and (A2) admit the classical

quantitative relationship between the bubble distribution and the solution

small pressure drop (amplitude and proﬁle) turned out difﬁcult

z

since the initial conditions just after shaking are unknown. DcH ðz; tÞ ¼ HðtÞDP 1 erf pﬃﬃﬃﬃﬃﬃ ; ðA3Þ

We end this Section with a few remarks about effervescence. 2 Dt

The latter phenomenon is typically observed when a bottle (of where erf stands for the error function. In our case, however, the

champagne, say) has been shaken and immediately opened, result- pressure is not constant in time. Thus, we have to generalize result

ing in foaming and gushing out of the bottle neck. Stated more for- (A3) to the case of a non perfect (still unknown) function DPðtÞ. We

mally, this situation amounts to gathering both types of the above begin by writing the pressure as a sum of impulsions, namely

mentioned perturbations, namely mechanically induced bubbling Z 1

and application of a large pressure drop DPo (the subscript ‘‘o’’ DPðtÞ ¼ DPðuÞdðt uÞdu; ðA4Þ

stands for opening), about 6 bar in amplitude in the case of cham- 0

pagne. In this situation, bubbles lie in a highly super-saturated

since for t < 0; DPðtÞ ¼ 0. Integrating by part once leads to

liquid, and almost all of them quickly grow in size and start rising

Z 1

up, generating a kind of explosive foam. As well-known, bubble dDP

DPðtÞ ¼ ½DPðuÞHðt uÞ1

0 þ Hðt uÞdu; ðA5Þ

growth only concerns those bubbles which are larger than a lower 0 du

limit size, called ‘‘critical radius’’ and given by ac ¼ 2r=DP o , see e.g.

which can ﬁnally be rewritten as

[30,31].

If opening occurs just after shaking, the immense majority of Z t

bubbles are super-critical (a > ac 0:2 lm), and then efferves- DPðtÞ ¼ DP0 ðuÞHðt uÞdu; ðA6Þ

0

cence starts ﬁercely. There is no pressure increase prior to opening,

and the pressure in the bottle neck keeps close to zero once the where P0 ðuÞ ¼ dDP=du. Since equations are linear, Dcðz; tÞ may sim-

bottle is opened. Gushing only lasts for a few seconds, but can be ply be derived by adding the responses to a sum of steps of inﬁni-

prolonged by continuous shaking of the bottle. As the concentra- tesimal amplitudes P 0 ðuÞdu, which yields

tion of carbon dioxide in the liquid quickly decreases, so does the Z !

t

corresponding Henry pressure PH [32]. Conversely, the critical z

Dcðz; tÞ ¼ kH DP0 ðuÞf pﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ du; ðA7Þ

radius, ac 2r=P H , divergently increases. Thus, gushing lasts as 0 2 Dðt uÞ

long as the generated bubbles are larger than the critical size.

As a ﬁnal remark, we stress again that characteristics (shapes where f ðxÞ ¼ 1 erfðxÞ. Integrating (A7) from z ¼ 0 to inﬁnity, one

and sizes) of bubbles generated by shaking are not known a priori. gets the number of moles of dissolved gas in excess (compared to

Because predictions about the evolution of the system are sensitive the initial quantity) in the liquid phase

Z 1

to what is supposed as the initial size distribution, this problem S

deserves a dedicated study. Exploiting the Faraday instability DnL ðtÞ ¼ Dcðz; tÞdz

M 0

[33] may be a route to producing approximately controlled bubble 1=2 Z

D SkH t 0

sizes by shaking. Other methods to generate bubbles, based on ¼ DP ðuÞqðt uÞdu; ðA8Þ

ultrasound induced cavitation [34] and shocks [35] are interesting p M 0

alternatives to study effervescence in a controlled framework. where qðxÞ ¼ x1=2 .

The conservation Eq. (11) is valid for t > 0. Its generalized form,

Acknowledgments which is valid at any time, can be written as:

¼ þ DndðtÞ: ðA9Þ

frame of VINEAL project. We thank E. Laurichesse, the instrumen- dt dt

tation and mechanics groups of CRPP for their continuous help The differential pressure proﬁle can be written as the sum of a

with the experiments, A. De Wit and F. Nallet for illuminating heavyside step DP HðtÞ and a continuous function DPgðtÞ whose

discussions. qualitative proﬁles are presented in Fig. 10. Doing this, Eq. (A8)

can be rewritten as:

Appendix A. Diffusion-based model of relaxation 1=2

D SkH

DnL ðtÞ ¼ ½qðtÞ þ g 0 qðtÞ: ðA10Þ

To make the calculation tractable, the volume of liquid is sup- p M

posed to be much larger than the volume of the vapor phase, so

Taking the Laplace transform of (A10) and using the initial con-

that we assume V L =V G ! 1. A direct consequence of the previous

dition gð0Þ ¼ 0 leads to

simpliﬁcation is that the pressure in the gas phase will relax to its

1=2

initial value – i.e. P1 ¼ P0 . On the other hand, the vanishing ﬂux D SkH

d

D nL ðsÞ ¼ ^ðsÞ½1 þ sg^ðsÞ

sq

condition in system (8)–(11) can be replaced by a condition of van- p M

ishing differential concentration at inﬁnity: 1=2

D SkH cPðsÞ;

Dc ! 0 for z ! 1; ðA1Þ ¼ ^ðsÞ D

sq ðA11Þ

p M

Now, consider ﬁrst the basic case of a perfect pressure step – i.e. ^ ðsÞ ¼ 1=s. Finally, considering that DnL ð0Þ ¼ 0, the Laplace

since H

a situation in which the differential pressure DP ¼ P P0 is kept at

transform of the time derivative of DnL takes the form

a constant value DP for t > 0þ (for example, by letting the intake

1=2

tap open, in contact with a regulation manometer) D SkH 2

d

D n0L ðsÞ ¼ s q cPðsÞ;

^ðsÞ D ðA12Þ

DPðtÞ ¼ DPHðtÞ; ðA2Þ p M

uous part, and supposing that DnG (resp. Dn) is linked to DP (resp.

52 A. Vreme et al. / Journal of Colloid and Interface Science 439 (2015) 42–53

ΔP, ΔnG

-5

5×10

-5

4×10

t

Θ 40

-5

a (m)

3×10 30

δa/a (%)

20

-5

2×10 10

t 0

g

-5 -10

t 1×10

-20 -6 -5 -4 -3

10 10 10 10

a0 (m)

0

0 20 40 60 80 100 120 140

ΔP , ΔnG t (s)

t relative error dsc =sc between the approximate and exact solutions to Eq. (B1)

evaluated at the time of collapse is shown in insert.

Henry’s equilibrium with the vapor phase. In this case and in

Fig. 10. Shapes of the functions involved in the resolution of Eqs. (A8) and (A9). See absence of surface tension, a bubble of any size would be stable

text for deﬁnitions. The Dirac distribution is represented by a vertical arrow. and would stay in the bulk endlessly (if buoyancy is neglected).

However the surface tension slightly increases the bubble inner

pressure (by an amount equal to Laplace pressure), making the sur-

DP) by the ideal gas law, one gets for the Laplace transform of Eq. rounding solution slightly under-saturated: the contained gas then

(A9): empties in the bulk.

RT d0 We consider a bubble of initial radius a0 , in a saturated carbon

cPðsÞ ¼

sD DnL ðsÞ þ DP: ðA13Þ

VG dioxide/water solution, that is to say a solution for which EP

parameter f is equal to 1. Using the same notations as in paragraph

^ðsÞ ¼ p1=2 =ð2s3=2 Þ, introducing (A12) in (A13) yields

Given that q Section 5.2.1 B1, the equation which rules the time evolution of the

radius a can be written (Eq. (34) of EP original paper) as:

cPðsÞ ¼ DPs1=2 ½a þ s1=2 ;

D ðA14Þ " #

da b=ðaqG Þ 1 1

where ¼ Dd þ ; ðB1Þ

dt 1 þ 2b=ð3aqG Þ a ðpDtÞ1=2

SkH D1=2 RT

a¼ : ðA15Þ where b ¼ 2Mr=ðRTÞ. This equation can be put in a dimensionless

2MV G 1=2

form by taking ¼ a=a0 and x ¼ ð2DdtÞ =a0 , and one gets

Once taken the inverse Laplace transform, one ﬁnally obtains

d n= hx i

2

DPðtÞ ¼ DPea t ½1 erfðat 1=2 Þ: ðA16Þ ¼ þ 2b ; ðB2Þ

dx 1 þ 2n=ð3Þ

Expanding the previous function in successive power of t 1=2 leads to and b ¼ ½d=ð2pÞ1=2 . An approximate solution corre-

where n ¼ a0 =a

DPðtÞ p sponding to Eq. (B2) where the constant b has been neglected, can

¼ 1 ðt=sÞ1=2 þ ðt=sÞ þ O½ðt=sÞ3=2 ðA17Þ be written in the following polynomial form:

DP 2

At early times, the ﬁrst two terms could be sufﬁcient to properly 1 2 þ nð1 2 Þ ¼ ð3n=2Þx2 : ðB3Þ

render the pressure relaxation process, provided the third one is Once solved the previous polynomial equation and going back

negligible, that is to say ðt=sÞ1=2 > ð10p=2Þðt=sÞ – if we consider that to dimensioned variables, we obtain the radius of the bubble as

the second term must be an order of magnitude greater than the an explicit function of time. The latter is quite cumbersome and

third one. This leads to the limit beyond which the expansion the will not be given here. We stress that Eq. (B3) is valid for small

O½ðt=sÞ3=2 term has to be taken into account: b, since this parameter has been neglected in Eq. (B2). Whereas b

is small in the case of over-saturated or under-saturated solution,

4 102

t6 s: ðA18Þ it is not the case for the saturated solution at Henry’s equilibrium.

p2 Thus, we can expect some discrepancy between the approximate

The previous results have been established in the case of an inﬁnite solution given by (B3) and the exact solution to the differential

volume of liquid. We guess that Eq. (A16) still holds for early times Eq. (B1). The relative difference in collapse time between the two

in the case of a ﬁnite volume of liquid since, considering the concen- solutions as a function of the initial radius lies between 0% and

tration proﬁles (at early time), the vanishing ﬂux boundary condi- 20% in absolute value, as can be seen in Fig. 11.

tion has nearly no effect on the diffusion process.

References

Appendix B. Simpliﬁed Epstein–Plesset solution

[1] S.M. Benson, D.R. Cole, Elements 4 (5) (2008) 325–331.

[2] B. Duteurtre, Le Champagne: de la tradition à la science, Ed. Lavoisier, Paris,

We recall in this appendix some theoretical results presented at 2010.

the end of EP paper [8], concerning the collapse of a bubble in a sat- [3] G. Liger-Belair, G. Polidori, V. Zéninari, Anal. Chim. Acta 732 (2012) 1–15.

A. Vreme et al. / Journal of Colloid and Interface Science 439 (2015) 42–53 53

[4] G. Liger-Belair, Uncorked: The Science of Champagne, Princeton University [22] F. Charru, Instabilites Hydrodynamiques, CNRS Editions, Paris, 2007.

Press, New Jersey, 2013. [23] E. Guyon, J.-P. Hulin, L. Petit, Physical Hydrodynamics, Ed. Oxford, Oxford,

[5] K.M. Valant, Hétéronucléation de bulles dans des liquides sursaturés en CO2. 2001.

Ph.D. Thesis, Thèse de l’Université de Marne-la-vallée, 2005. [24] T.J. Kneafsey, K. Pruess, Transp. Porous Med. 82 (2010) 123–139.

[6] M. Vignes-Adler, Angew. Chem. Int. Ed. 52 (2013) 187–190. [25] L. Lemaigre, M.A. Budroni, L.A. Riolfo, P. Grosﬁls, A. De Wit, Phys. Fluids 25

[7] A. Okhotsimskii, M. Hozawa, Chem. Eng. Sci. 53 (14) (1998) 2547–2573. (2013). 014103–18.

[8] P.S. Epstein, M.S. Plesset, J. Chem. Phys. 18 (1950) 1505–1509. [26] P. Taylor, Adv. Colloid Interface Sci. 75 (1998) 107–163.

[9] F. Nadal, P. Meunier, B. Pouligny, E. Laurichesse, J. Fluid Mech. 719 (2013) 203– [27] Note that we used a viscous drag coefﬁcient of the form 6pga instead of

229. 4pga, which can seem irrelevant in the case of a bubble. However, in real

[10] A.D. Britt, W.B. Moniz, IEEE J. Quant. Electron. 8 (12) (1972) 913–914. sparkling beverages, bubbles mostly behave as solid spheres due the presence

[11] C. Arcoumanis, J.J. McGuirk, J.M.L.M. Palma, Exp. Fluids 10 (1990) 177–180. of amphiphilic components at the liquid/gas interface (see C. Ybert, J.M. Di

[12] R. Sjoback, J. Nygren, M. Kubista, Spectrochim. Acta Part A 51 (1995) L7–L21. Meglio, Eur. Phys. J. B 4 (1998) 313–319).

[13] G.G. Agabaliantz, Bull. OIV 36 (1963) 703–714. [28] J. Senée, B. Robillard, M. Vignes-Adler, Food Hydrocolloids 13 (1999) 15–26.

[14] G. Liger-Belair, M. Bourget, C. Cilindre, H. Pron, G. Polidori, J. Food Eng. 116 (14) [29] K. Abou-Saleh, V. Aguié-Béghin, L. Foulon, M. Valade, R. Douillard, Colloids

(2013) 78–85. Surf. A 344 (2009) 86–96.

[15] A. Tamimi, Edward B. Rinkerand, O.C. Sandall, J. Chem. Eng. Data 39 (1994) [30] C.A. Ward, A. Balakrishna, F.C. Hooper, J. Fluids Eng. 92 (1970) 695–704.

330–332. [31] C.A. Ward, P. Tikuisis, R.D. Venter, J. Appl. Phys. 53 (1982) 6076–6084.

[16] M.J.W. Franck, J.A.M. Kuipers, W.P.M. van Swaaij, J. Chem. Eng. Data 41 (2) [32] We call Henry pressure the saturation pressure corresponding to the actual

(1996) 297–302. concentration in the bulk, which, after opening, does not match the external

[17] W. Lu, Huirong Guo, I.M. Chou, R.C. Burruss, L. Li, Geochim. Cosmochim. Acta partial pressure of carbon dioxide.

115 (2013) 183–204. [33] C.L. Goodridge, W. Tao Shi, D.P. Lathrop, Phys. Rev. Lett. 76 (11) (1996) 1824–

[18] S.P. Cadogan, G.C. Maitland, J.P. Martin Trusler, J. Chem. Eng. Data 59 (2014) 1827.

519–525. [34] K.K. Sahu, Y. Hazama, K.N. Ishihara, J. Colloid Interface Sci. 302 (2006) 356–

[19] J.J. Carroll, A.E. Mather, J. Sol. Chem. 21 (1992) 607–621. 362.

[20] L.W. Diamond, N.N. Akinﬁev, Fluid Phase Equilib. 208 (1-2) (2003) 265–290. [35] J. Rodriguez-Rodriguez, A. Casado, D. Fuster, Why Does a Beer Bottle foam Up

[21] A. Hebach, A. Oberhof, N. Dhamen, J. Chem Eng. Data 49 (2004) 950–953. After a Sudden Impact on its Mouth?, 2013 (arXiv (arxiv.org/abs/1310.3747)).

- Plug Flow Reactor Module Using MatlabUploaded byTaofik Nurdiansah
- D 6558 â€“ 00 ;RDY1NTGTMDA_Uploaded bybogginoc
- CH 353 Notes-1Uploaded byImilie Chang
- The Hydrophobic EffectUploaded byEsteban Araya
- template_ije.docUploaded byPadmasri Girirajan
- 1985 Flow Characteristics of Downflowbubble ColumnsUploaded byRonaldo Leite
- Chemical Engineering Mass Transfer NotesUploaded byLebohang Czar Nku
- Problem 11 - Optimum Slug SizeUploaded byRavindr Kumar
- 2010 Test 1Uploaded byJeremie Lacroix
- CO2 vimpUploaded byRimika Madan Kapoor
- Fizzy Balloon ExtractorUploaded byDanielle
- 19245.pdfUploaded bynach
- Marine Bilge water treatment systemUploaded byrobjsimon
- Properties of Pure SubstancesUploaded byaathithyayan
- 1grade 8 june ME 2018-2019Uploaded byLucille Gacutan Aramburo
- LT catalyst reductionUploaded byRajeshkumar Elango
- Sensitivity of Pipeline Gas Flow ModelUploaded byedgareler
- Rigorous simulation and design Plate Column.pdfUploaded byCatherine Ccasa
- States of MatterUploaded byKerri Satherley
- 11th Chemistry - Unit 1,6,7 Question Paper - TamilNadu TN State Board English Medium - Brainkart.comUploaded byBrainKart Com
- Form 2 Air pressureUploaded bylccjane8504
- Modelling of Radial Heat Flow- Article IlievUploaded bysnegorin
- Maths Investigation_This Troubled PlanetUploaded byCameron_Stewart
- 12_LectureOutlineThermalPropertiesOfMatter.pptUploaded byOrlando
- Homeowork-state of Matter Week 8Uploaded byUsha Peru
- Lecture No. 5 Phase DiagramUploaded byarslan
- Gas CloudUploaded byCebolinha Cebola
- carbon cycle ppt notes-1Uploaded byapi-293001217
- Monitoring pond water quality to improve shrimp and fish productionUploaded byInternational Aquafeed magazine
- h2 in AluminumUploaded byAnonymous m1OSMajSJ

- TEAM PORTFOLIO: Project 14 Feature List (SAP HYBRIS)Uploaded byRauf Aliev
- [slides] SAP Hybris: Thinking Outside the Box [English]. Part 6. Page Fragment Caching, Varnish. Nginx/Memcached. PDF generationUploaded byRauf Aliev
- SAP Hybris Thinking Outside the Box. PART 4. Automatic Facet DiscoveryUploaded byRauf Aliev
- Rauf's Collection - Sheet MusicUploaded byRauf Aliev
- TEAM PORTFOLIO: Project 43 Feature List (SAP HYBRIS)Uploaded byRauf Aliev
- TEAM PORTFOLIO: Project 60 Feature List (SAP HYBRIS)Uploaded byRauf Aliev
- Каган. Как молодежь расходует и как надо расходовать получкуUploaded byRauf Aliev
- SAP Hybris Thinking Outside the Box. PART 1Uploaded byRauf Aliev
- SLIDES: SAP HYBRIS: Thinking Outside the Box. PART 5 [VIDEO, RUSSIAN + ENGLISH]. Bulk search. Geofencing. Drools Fusion. Rule-based recommendations.Uploaded byRauf Aliev
- Rauf Aliev's SAP hybris Commerce certificateUploaded byRauf Aliev
- Steve Jobs: the lost interviewUploaded byRauf Aliev
- All articles from hybrismart.com (2016-08-23)Uploaded byRauf Aliev
- SAP Hybris Thinking Outside the Box. PART 2. Search AnalyticsUploaded byRauf Aliev
- SAP Hybris Thinking Outside the Box - ext2-35-43.pdfUploaded byRauf Aliev
- SAP Hybris Thinking Outside the Box - ext2-35-43.pdfUploaded byRauf Aliev
- Опросник по технологиямUploaded byRauf Aliev
- RaufAliev CV Rus FullUploaded byRauf Aliev
- Benchmark Full EnglishUploaded byRauf Aliev
- Events I took partUploaded byRauf Aliev
- JavaTechChecklistUploaded byRauf Aliev
- RaufAliev CV FullUploaded byRauf Aliev
- The Forrester Wave B2B Commerce Suites 2013Uploaded byRauf Aliev
- The Forrester Wave B2B Commerce Suites 2013Uploaded byRauf Aliev
- задачка по информатикеUploaded byRauf Aliev
- RaufAliev Rus AchievementsUploaded byRauf Aliev
- Jack London - To Build A FireUploaded byapi-3784429
- RaufAliev CV 2013 (docx version) - sign in required to download the fileUploaded byRauf Aliev
- Rauf Aliev CV 2013 Eng COMPACTUploaded byRauf Aliev
- Rauf Aliev CV 2013 Rus COMPACTUploaded byRauf Aliev

- Dynamic Collapse Simulation of 3-Bay RC Frame Under Extreme Earthquake LoadingsUploaded byNimish Prabhukhanolkar
- 6. Temp Effect on CompactionUploaded byvidyaranya_b
- Sequential Algorithms for Multiplication and DivisionUploaded byAhmad Rahan
- 510Uploaded byEdgar Chuquipiondo Vargas
- lecture3_11Uploaded byAralechan Carpy
- Lecture 1Uploaded byAmierson Tilendo
- analysis of pumping testUploaded byΒενιαμίν στέμμα
- psbyDevilsDukeUploaded byramarvs
- 198948917-Time Travel Theory Avoids Grandfather ParadoxUploaded bysmss_proe5945
- SSRN-id283308Uploaded bythejournalofeconomic
- 5 kpis for services companies.pdfUploaded byNam Duy Vu
- Cfdb TurbulenceUploaded bycristi01068468
- Syllabus MGU BTech 26-03-3ffUploaded byrathish14u
- forwards and backwards number patterns term 1Uploaded byapi-277245562
- 12 Dwdm NotesUploaded byOzioma Ihekwoaba
- Energy Levels in One DimensionUploaded bySaiia Gustti
- Hazard ModelUploaded byShejj Peter
- SPSSUploaded byMohamed H. Jiffry
- E204Uploaded byClar Cabundocan
- comp1hwUploaded byTienanh Nguyen
- Slope Calculation From Contour Lines in a Topographic MapUploaded byzartol
- Low-Power and Area-Efficient Carry Select AdderUploaded byrajidisahithi
- 8 queen perfectUploaded bynitishramudu
- MOCK DESCRIPTION.docxUploaded byMerupu Debba
- 2.1 Derivatives Limits&ContinuityUploaded byJuanAntonioPoblete
- 233369101 Root Sequence Planning 1Uploaded byTulipe Pivoine
- Crawford Kulak 1968Uploaded bygimusi
- MITSOB SRM SEM 1 QUE PAPERUploaded byKuldip
- 7.1 NotesUploaded byjhaselhorst
- Alexander v. Ivanov Auth. Asymptotic Theory of Nonlinear RegressionUploaded byjjg8116