Anda di halaman 1dari 88

Eur. Phys. J.

Special Topics 201, 188 (2012)


c EDP Sciences, Springer-Verlag 2012

DOI: 10.1140/epjst/e2012-01528-0

THE EUROPEAN
PHYSICAL JOURNAL
SPECIAL TOPICS

Review

The physics behind the zz in champagne and


sparkling wines
G. Liger-Belair1,2,a
1

Equipe Eervescence, Groupe de Spectrometrie Moleculaire et Atmospherique (GSMA),


UMR CNRS 6089, UFR Sciences Exactes et Naturelles, BP. 1039, 51687 Reims Cedex 2,
France
2
Laboratoire dnologie et Chimie Appliquee, UPRES EA 2069, URVVC, UFR Sciences
Exactes et Naturelles, Universite de Reims, BP. 1039, 51687 Reims Cedex 2, France
Received 30 September 2011 / Received in nal form 16 November 2011
Published online 12 March 2012

G
erard Liger-Belair:
He received his PhD in physical sciences in 2001 from the University
of Reims, in France. He received an associate professor position at the
University of Reims in 2002, and a full professor position, in 2007, in
the same University. He has been researching the physics and chemistry
behind the bubbling properties of champagne and sparkling wines for
several years. His current interests include the science of bubbles, foams
and thin lms, and their broad interdisciplinary applications. He is the
author of several academic and popular science books. His rst book,
Uncorked: the science of champagne, published by Princeton University
Press, won the 2004 award for the Best Professional/Scholarly Book in
Physics from the Association of American Publishers.
Abstract. Bubbles in a glass of champagne may seem like the acme
of frivolity to most of people, but in fact they may rather be considered as a fantastic playground for any physicist. Actually, the so-called
eervescence process, which enlivens champagne and sparkling wines
tasting, is the result of the ne interplay between CO2 dissolved gas
a

e-mail: gerard.liger-belair@univ-reims.fr

The European Physical Journal Special Topics


molecules, tiny air pockets trapped within microscopic particles during
the pouring process, and some both glass and liquid properties. Results obtained concerning the various steps where the CO2 molecule
plays a role (from its ingestion in the liquid phase during the fermentation process to its progressive release in the headspace above the
tasting glass as bubbles collapse) are gathered and synthesized to propose a self-consistent and global overview of how gaseous and dissolved
CO2 impact champagne and sparkling wine science. Physicochemical
processes behind the nucleation, rise, and burst of gaseous CO2 bubbles found in glasses poured with champagne and sparkling wines are
depicted. Those phenomena observed in close-up through high-speed
photography are often visually appealing. I hope that your enjoyment
of champagne will be enhanced after reading this fully illustrated review dedicated to the science hidden right under your nose each time
you enjoy a glass of champagne.

1 Introduction
Since the end of the 17th century, champagne has been a world-wide renowned French
sparkling wine. Nevertheless, only quite recently much research has been devoted to
depict each and every parameter involved in its bubbling process [1]. From a strictly
chemical point of view, Champagne wines are multicomponent hydroalcoholic systems
with a density close to unity, a surface tension 50 mN m1 , and a viscosity about
50% larger than that of pure water [1]. Champagne wines are also supersaturated
with CO2 dissolved gas molecules formed together with ethanol during the second
fermentation process (called prise de mousse, and promoted by adding yeasts and
a certain amount of sugar inside bottles lled with a base wine and sealed with a
cap). Actually, during this second fermentation process which occurs in cool cellars,
the bottles are sealed, so that the CO2 molecules cannot escape and progressively
dissolve into the wine [14]. Therefore, CO2 molecules dissolved into the wine and
gaseous CO2 molecules trapped under the cork progressively establish equilibrium
(an application of Henrys law which states that the concentration of a given gas
molecule dissolved into a solution is proportional to its partial pressure in the vapor phase, above the solution). Champagne wines therefore hold a concentration of
dissolved CO2 proportional to the level of sugar added to promote this second fermentation (see reference [3] for example, and references therein). Actually, a standard 75
centiliters champagne bottle typically holds about 9 grams of dissolved CO2 , which
correspond to a volume close to 5 liters of gaseous CO2 under standard conditions
for temperature and pressure [14]. This very signicant volume of dissolved CO2 is
responsible for bubble formation once the bottle is uncorked and the wine poured
into a glass. To get an idea of how many bubbles are potentially involved all along
the degassing process from this single bottle, we can divide this volume of CO2 to
be released by the average volume of a typical bubble of 0.5 mm in diameter. A huge
number close to 108 is found!
In champagne and sparkling wine tasting, the concentration of dissolved CO2 is
indeed a parameter of high importance since it directly impacts the four following
sensory properties: (i) the frequency of bubble formation in the glass [58], (ii) the
growth rate of rising bubbles [3,6], (iii) the mouth feel, i.e., the mechanical action of
collapsing bubbles as well as the chemosensory excitation of nociceptors in the oral
cavity (via the conversion of dissolved CO2 to carbonic acid) [912], and (iv) the
aromatic perception of champagne (i.e., its so-called bouquet [13,14]), as collapsing
bubbles release their content in gaseous CO2 and volatile organic compounds above

Champagne Fizzics

the champagne surface. Actually, a link has been recently evidenced between carbonation and the release of some aroma compounds in carbonated waters [15,16]. Sensory
results revealed that the presence of CO2 increased aroma perception regardless of the
sugar content in mint-avored carbonated beverages [16]. CO2 seemed thus to induce
large modications of the physicochemical mechanisms responsible for the aroma release and avor perception in soft drinks. Therefore, following these recent highlights,
a link is also strongly suspected between the level of dissolved CO2 and the release
of aromatic compounds during Champagne and sparkling wine tasting.
In this tutorial review, the evolving nature of the dissolved and gaseous CO2
found in Champagne wines is done, from the bottle to the tasting glass, in order
to propose a global overview of how CO2 impact champagne and sparkling wine
science. Moreover, physicochemical processes behind the nucleation, rise, and burst
of gaseous CO2 bubbles found in glasses poured with champagne and sparkling wines
are depicted. Impacts of various physicochemical and geometrical parameters of the
wine, bottle, and glass itself are presented, in tasting conditions, through various
analytical techniques.

2 The CO2 within the bottle


The modern production of champagne is not so far removed from that developed
empirically by the Benedictine monk Dom Pierre Perignon, in the late 17th century.
This method is also used outside the Champagne region. Sparkling wines produced
as such are labelled methode traditionelle. Indeed, most American and Australian
sparkling winemakers use this method to elaborate their own sparkling wines. This
method involves several distinct steps [14].

2.1 A rst alcoholic fermentation


Three grape cultivars are grown in the 75 000 acres of the Champagne vineyards:
Chardonnay (a white grape), Pinot Meunier and Pinot Noir (both dark grapes).
Usually around mid-September (but more often around early September since the
past few years, probably due to global warming), the grapes harvested from these
vineyards are pressed to make a juice, called the grape must. After pressing, the
must is transferred into open vats where yeast (Saccharomyces cerevisiae) is added. In
Champagne, the must typically holds about 180 grams per liter of sugars. Generally
speaking, the key metabolic process during winemaking is alcoholic fermentation:
the conversion of sugars into ethanol and carbon dioxide by yeast. The process of
fermentation was rst scientically described by the French chemist Joseph-Louis
Gay Lussac in 1810, when he demonstrated that glucose is the basic starting block
for producing ethanol:
C6 H12 O6 CH3 CH2 OH + 2 CO2 .

(1)

The manner in which yeast contributes to the fermentation process was not clearly
understood until 1857, when the French microbiologist Louis Pasteur discovered that
not only does the fermentation process not require any oxygen, but alcohol yield
is actually reduced by its presence. The amount of ethanol generated by this rst
alcoholic fermentation is about 11%. At this step, champagne is still actually a noneervescent white wine, because the huge amount of gaseous carbon dioxide produced
during the rst alcoholic fermentation (close to 50 liters of gaseous CO2 per liter of
must) is allowed to escape into the atmosphere.

The European Physical Journal Special Topics

Fig. 1. A cellar master (here Thierry Gasco, from Champagne Pommery) will blend several
dierent still wines from various grape varieties, vineyards and vintages to produce one
champagne. (Alain Cornu/Collection CIVC.)

2.2 The art of blending


Because it is rare that a single wine of a single vintage from a single vineyard and grape
variety will provide the perfect balance of avor, sugar level, and acidity necessary
for making a ne champagne, wine makers will often mix several dierent still wines.
This is called the assemblage (or blending) step, and it is carried out directly after
the rst alcoholic fermentation is complete. Blending is considered a key step in the
art of champagne-making (see Fig. 1). A cellar master will sometimes blend up to 80
dierent wines from various grape varieties, vineyards and vintages to produce one
champagne. The blending of still wines originally made from the three grape cultivars
forms a base wine, which will then undergo a second fermentation the key step in
producing the sparkle in champagne and other sparkling wines.
2.3 The prise de mousse: A second alcoholic fermentation in a sealed bottle
Once the base wine is created, sugar (about 24 grams per liter) and yeast are added.
The entire concoction is put into thick-walled glass bottles, and sealed with caps.
The bottles are then placed in a cool cellar (10 to 12 C), and the wine is allowed
to slowly ferment for a second time, producing alcohol and carbon dioxide again.
Following Eq. (1), 24 grams per liter of sugar added in closed bottles to promote
the second alcoholic fermentation produce approximately 11.8 grams of CO2 per liter
of wine. The concentration of dissolved CO2 in champagne (in grams per liter) is
therefore roughly equivalent to half of the concentration of sugar (in grams per liter)
added into the base wine in order to promote the prise de mousse. Finally, a standard
75 centiliters champagne bottle holds close to 9 grams of dissolved CO2 molecules.
By use of the molar mass of CO2 (44 g/mol), and the molar volume of an ideal gas
(close to 24 L/mol at 12 C), it can be deduced that about 5 litres of gaseous CO2
are trapped into a single bottle of champagne (i.e., 6 times its own volume!).
Actually, during this second fermentation process which occurs in cool cellars,
the bottles are sealed, so that the CO2 molecules cannot escape and progressively

Champagne Fizzics

Table 1. Henrys law constant of CO2 in champagne as a function of temperature, for a


typical champagne with 12.5% (v/v) of ethanol and 10 g/L of sugars (compiled from the
data by Agabaliantz [17]).
Temperature
( C)
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20

Henrys law constant kH


(g/L/bar)
2.98
2.88
2.78
2.68
2.59
2.49
2.41
2.32
2.23
2.16
2.07
2.00
1.93
1.86
1.79
1.73
1.67
1.60
1.54
1.48
1.44

dissolve into the wine. Therefore, CO2 molecules dissolved into the wine and gaseous
CO2 molecules under the cork progressively establish equilibrium an application of
Henrys law which states that the partial pressure of a given gas above a solution is
proportional to the concentration of the gas dissolved into the solution, as expressed
by the following relation:
c = kH PCO2

(2)

where c is the concentration of dissolved CO2 molecules, PCO2 is the partial pressure
of CO2 molecules in the vapor phase, and kH is its Henrys law constant (i.e., its
solubility into the wine). Since the present work specically deals with the dissolved
CO2 concentration found in champagne samples, and to facilitate the reading of the
article, the subscript CO2 found in the formulas will be omitted in the following.
Generally speaking, the solubility of a given gas into a solution is strongly
temperature-dependent (the lower the temperature of the solution, the higher the
gas solubility). In champagne and other sparkling wines, Agabaliantz thoroughly examined the solubility of dissolved CO2 molecules as a function of both temperature
and wine parameters [17]. His empirical relationships are still in use nowadays by
champagne and other sparkling winemakers. For a typical sparkling wine elaborated
according to the methode traditionnelle, Agabaliantz established the temperature dependence of the Henrys law constant, which is displayed in Table 1.
Thermodynamically speaking, the behavior of Henrys law constant as a function
of temperature can be conveniently expressed with a Vant Ho like equation as
follows:



1
Hdiss 1

kH (T ) = k298K exp
(3)
R
T
298

The European Physical Journal Special Topics

where k298K is the Henrys law constant of dissolved CO2 at 298 K ( 1.21 g/L/bar),
Hdiss is the dissolution enthalpy of CO2 molecules in the liquid medium (in J/mol),
R is the ideal gas constant (8.31 J/K/mol), and T is the absolute temperature (in
K). By tting Agabaliantz data with the latter equation, it is worth noting that the
dissolution enthalpy of CO2 molecules in champagne may be evaluated [18]. The best
t to Agabaliantz data was found with Hdiss 24 800 J/mol. In comparison, the
dissolution enthalpy of CO2 molecules in pure water is about 19 900 J/mol [19].
2.4 Disgorging and corking
This second fermentation is followed by a minimum period of 15 months of aging,
in contact with dead yeast cells (called aging on lees). The autolysis of yeast occurs during the aging of champagne on lees (and of other sparkling wines elaborated
through the methode traditionnelle). During this process, the yeast releases dierent
compounds that modify the organoleptic properties of the wine. This aging period is
required to give these wines their roundness and characteristic aroma and avour, as
detailed in the review by Alexandre and Guilloux-Benatier [20]. More details about
chemical and biochemical features involved in sparkling wine production (from a traditional to an improved winemaking technology) can be found in two recent reviews
by Pozo-Bayon et al. [21], and by Torresi et al. [22].
After aging on lees, bottles then undergo disgorging. The necks of the bottles are
frozen, creating a small ice plug that traps the sediment of dead yeast cells. The bottle
caps are then removed in order to remove the sediment of dead yeast cells, as shown
in the photograph displayed in Fig. 2.
Specic liquor, which consists of a mixture of sugar and old wines, is then added in
order to replace the bit of wine lost during disgorging. A bit of dissolved CO2 is also
inevitably lost at this step, but bottles are quickly corked with traditional or technical
cork stoppers to prevent an excessive loss of CO2 [14]. After corking the bottle,
dissolved and gaseous CO2 quickly recover equilibrium. Dissolved CO2 molecules
progressively desorb from the liquid medium to promote the raise of gaseous pressure
under the cork, which nally and quickly recovers a stable value. Nevertheless,
because the solubility of CO2 strongly depends on the champagne temperature,
the pressure of gaseous CO2 under the cork also strongly depends, in turn, on the
champagne temperature. The physicochemical equilibrium of CO2 molecules within a
champagne bottle is ruled by both Henrys law (for CO2 -dissolved gas molecules) and
the ideal gas law (for the gaseous CO2 in the headspace under the cork). Moreover,
the conservation of the total mass of CO2 molecules (dissolved into the wine, and
in the vapor phase, under the cork) applies, since bottles are hermetically closed.
Therefore, by combining the two aforementioned laws with mass conservation, the
following relationship was determined which links the pressure P of gaseous CO2
under the cork (in bars) with both temperature and bottles parameters as [23]:
P

mRT
4, 4 103 v + (kH RT ) V

(4)

where m is the total mass of CO2 within the bottle (in grams), kH is the Henrys law
constant given in Table 1 (in g/L/bar), V is volume of champagne within the bottle
(in liter), and v is the volume of the gaseous headspace under the cork (in liter).
Most commonly, champagne is elaborated in three various types of bottle, namely
the standard bottle (with a volume of 75 centiliters), the magnum bottle (with a
volume of 150 centiliters), and the half bottle (with volume of 37.5 centiliters). It is

Champagne Fizzics

Fig. 2. During disgorging, the necks of the bottles are frozen, creating a small ice plug
that traps the sediment of dead yeast cells. Bottle caps are removed, and the CO2 pressure
that is released shoots the ice out of the bottleneck, along with all the yeast sediment.
(Kumesawaga/Collection CIVC).

worth noting that the three aforementioned bottles present indeed dierent volume
content, but the same headspace volume v 25 mL, as well as similar cork stoppers
(in terms of length, and cross section).
The temperature dependence of the pressure P under the cork, within the three
aforementioned bottle types, is displayed in Fig. 3. At the temperature of champagne
tasting (usually between 8 and 10 C), the pressure within bottles is close to 5 bars
(i.e., 5 105 N/m2 ), whatever the bottle type. It is worth noting that the pressure
inside the bottle is very slowly bottle type-dependent (provided that 24 g/L of sugar
were added in the base wine to promote the prise de mousse).
Moreover, since dissolved and gaseous CO2 experience equilibrium through
Henrys law, as stated in Eq. (2), the concentration c of CO2 dissolved molecules

The European Physical Journal Special Topics

12
Standard Bottle
Magnum Bottle
Half Bottle

P (bars)

10

2
0

10

15

20

25

30

T (C)
Fig. 3. Temperature dependence of the pressure P under the cork, within the three most
common champagne bottle types, namely, the standard bottle, the magnum bottle, and the
half bottle, respectively.

found in champagne expresses as follows:


c = kH P

kH mRT
.
4.4 103 v + (kH RT ) V

(5)

The temperature dependence of the dissolved CO2 concentration within champagne,


for the three aforementioned bottle types, is displayed in Fig. 4. It can be seen from
Fig. 4 that the smaller the bottle volume, the lower the concentration of dissolved
CO2 . It is also worth noting from Fig. 4 that the concentration of dissolved CO2 is
rather slightly temperature-dependent (despite a pressure within the bottle which is
highly temperature-dependent, as seen in Fig. 3).
Concentrations of CO2 molecules dissolved into champagne are indeed easily accessible by use of carbonic anhydrase, which is the ocial method recommended by
the OIV (namely the International Oce of Vine and Wine) for measuring the CO2
content in Champagne and sparkling wines. This method is thoroughly detailed in
a recent paper [24]. Numerous experiments with early disgorged champagne samples
were done recently [2426]. The characteristic concentration of dissolved CO2 measured inside freshly opened standard bottles was found to be of order of 11.5 g/L, and
therefore in very good accordance with Eq. (5).
2.5 The impact of champagne aging
The cork of a wine bottle is far from being completely hermetic with regard to gas
exchanges. Cork is a porous material [2730]. Gas species are therefore able to slowly
diuse through the cork along their respective inverse partial pressure gradients (see
Fig. 5) [31].

Champagne Fizzics

-1

C (g L )

12

11

Standard Bottle
Magnum Bottle
Half Bottle
10
0

10

15

20

25

30

T (C)
Fig. 4. Temperature dependence of the dissolved CO2 concentration within champagne,
within the three most common champagne bottle types.

In case of a corked champagne bottle, the huge dierence in CO2 partial pressure
between the headspace (close to 6 bars at 12 C) and the ambient air (of order of only
0.0004 bar) forces CO2 molecules to slowly diuse through the cork. Therefore, during
its aging period of time, a corked champagne bottle progressively loses its gaseous
CO2 content, as well as its dissolved CO2 content (which both experience equilibrium
through the so-called Henrys law). As seen in the preceding paragraph, early disgorged Champagne wines (elaborated with 24 g/L of sugar added in the base wine to
promote the prise de mousse) hold a concentration of dissolved CO2 of c 11.5 g L1
[2426]. Recently, the concentration of dissolved CO2 was measured in various samples of an older Champagne wine elaborated in the late 1990s (also with 24 g/L of
sugar added in the base wine to promote the prise de mousse), disgorged, and aged
in a cool cellar ever since. Approximately ten years after having been disgorged and
corked, this old champagne was found to hold a much lower level of dissolved CO2
of only c 8 g L1 [30,31]. As expected, the old champagne contains substantially
less dissolved CO2 than the young one, undoubtedly due to the slow but continuous
diusion of gaseous CO2 through the cork, all along the aging period. Actually, deeper
investigations about the way cork stoppers alter with time in a bottle neck are needed
to better understand how their porosity and permeability are aected with regards to
gas transfers during champagne aging. Such experiments are to be conducted along
that line in a near future.
2.6 The chemical composition of champagne
From the point of view of the chemist, champagne can indeed be viewed as a multicomponent aqueous solution. The ne chemical composition of a typical Champagne
wine is reported in Table 2 [32]. Actually, gases like CO2 undergo specic reactions
with water. Equilibrium is established between the dissolved (CO2 )aq and H2 CO3 ,

10

The European Physical Journal Special Topics

>>

Fig. 5. Scheme of the bottle neck, and compilation of the various cork, bottle, and wine
parameters.

the carbonic acid:

(CO2 )aq + H2 O H2 CO3 .

(6)

Moreover, carbonic acid is a weak acid that dissociates in two steps:


H2 CO3 + H2 O H3 O+ + HCO
3
2
+
HCO
3 + H2 O H3 O + CO3

pKa1 (at25 C) = 6.37

(7)

pKa2 (at25 C) = 10.25.

(8)

However, as the pH of champagne and sparkling wines is relatively low (of order of

3.2), no carbonated species (CO2


3 , HCO3 ) should coexist with dissolved CO2 . The
13
C Magnetic Resonance Spectroscopy (MRS) technique was used as an unintrusive
and non-destructive method to determine the amount of CO2 dissolved in closed
bottles of Champagne and sparkling wines [33]. Dierent well-separated peaks were
recorded in a 13 C spectrum, as can be seen in the Fig. 6: (i) the quadruplet of CH3
group of ethanol appears at 17.9 ppm, (ii) the triplet of CH2 (-OH) group of ethanol
at 57.3 ppm, and (iii) the singlet of CO2 appears at 124,4 ppm, thus conrming the

absence of other carbonated species (CO2


3 , HCO3 ) in the liquid matrix (contrary to

Champagne Fizzics

11

Table 2. Average composition of a typical Champagne wine, from Dussaud [32]; typically,
pH 3.2 and the ionic strength of champagne is 0.02 M .
Compound
Ethanol
CO2
Glycerol
Tartaric acid
Lactic acid
Sugars
Proteins
Polysaccharides
Polyphenols
Amino acids
Volatile Organic Compounds (VOC)
Lipids
K+
Ca2+
Mg2+
SO2
4
Cl

Concentration
12.5% v/v
1012 g/L
5 g/L
2.5 to 4 g/L
4 g/L
1050 g/L
510 mg/L
200 mg/L
100 mg/L
0.82 mg/L
700 mg/L
10 mg/L
200450 mg/L
60120 mg/L
5090 mg/L
200 mg/L
10 mg/L

CH2(-OH)
CH3

Ethanol

CO2

130

120

110

100

90

80

70

60

50

40

30

20

10

(ppm)
13

Fig. 6. C spectrum recorded to measure the CO2 concentration in champagne; it is clear

that no carbonated species (CO2


3 , HCO3 ) coexist with dissolved CO2 .

zzy waters for example, where pH values near neutrality enable the above mentioned
carbonated species to cohabit with CO2 -dissolved gas molecules).

3 Champagne cork popping revisited through high-speed


infrared imaging
Even if it is far more gentle to uncork a bottle of champagne with a subdued sigh,
anyone of us has already experienced popping the cork with a bang. To the best
of our knowledge, no scientic study dealing with the temperature dependence of
the cork popping process has been reported up to now. Champagne cork popping
out of standard 75 cL bottles was examined through high-speed infrared imaging.
The cloud of gaseous CO2 escaping during the cork popping process (invisible by

12

The European Physical Journal Special Topics

the naked eye) was visualized and followed with time while diusing in ambient
air [34]. By considering that gases under pressure in the bottleneck experience an
adiabatic expansion during the uncorking process, a thermodynamic model was built
which links the volume of gaseous CO2 gushing out of the bottleneck, its drop of
temperature, and the potential energy stored in the bottleneck, with the champagne
temperature, the solubility of CO2 into the wine, and the volume of the bottleneck.
3.1 The batch of champagne corked bottles
A batch of standard commercial Champagne wine, recently elaborated in 75 cL bottles, with a blend of 100% chardonnay base wines (vintage 2008 Cooperative Nogent
lAbbesse, Marne, France), was used for this set of experiments. Bottles were elaborated with 24 g/L of sugar added in the base wine to promote the prise de mousse,
so that the same amount of about 9 g of CO2 was produced into every bottle of this
batch. After a classical period of aging of 15 months, bottles were then corked with
traditional natural cork stopper (from a same batch of corks), as the one shown in
Fig. 7.
As seen in Fig. 7, each cork stopper is composed of two well-distinct dierent
parts: (i) a upper part composed of agglomerated cork granules, and (ii) a lower part
made of two massive cork slices stuck together. The mass M of each cork is about
10 grams. It is also worth noting that, into every bottle of this batch, the volume of
the headspace under the cork is equal to 25 mL. After having been corked with cork
stoppers, bottles were stored in a cool cellar, at 12 C.

3.2 Infrared imaging technique


A visualization technique based on the Infrared (IR) thermography principle has
been used to lm the gaseous CO2 uxes outgassing from champagne (invisible in the
visible light spectrum) [35]. The CO2 absorptions observable by the IR camera are
quite weak because this gas molecule has only a strong absorption peak in the detector
bandwidth at 4.245 m. Consequently, the best way to visualize the ow of gaseous
CO2 desorbing from champagne is to t the camera with a band-pass lter (centered
on the CO2 emission peak), as shown in a recent article [36]. The experimental device
consists of a CEDIP middlewaves Titanium HD560M IR video camera, coupled with
a CO2 lter ( 50.8 mm 1 mm thick Laser Components SAS ). In complement, the
technique involves an extended high-emissivity (0.97) blackbody (CI systems provided
by POLYTEC PI ), used at a controlled uniform temperature of 80 C, and placed
behind the bottleneck. The IR video camera was used at a 100 frames per second
(fps) lming rate. A photograph of the whole set-up is displayed in Fig. 8.

3.3 Visualizing the cloud of gaseous CO2 escaping during the cork popping
process
The time-sequence displayed in Fig. 9 illustrate the champagne cork popping step
as seen through the objective of the IR video camera, for a bottle stored at 12 C.
Because CO2 molecules absorb the 4.245 m wavelength found in the IR spectrum
emitted by the blackbody, the presence of gaseous CO2 in front of the blackbody is
betrayed as a dark zone in the eld of view of the camera. Therefore, the gaseous
CO2 trapped in the headspace under the cork appears as a gray cloud gushing out of
the bottleneck during the cork popping process.

Champagne Fizzics

13

Agglomerated
cork granules

Two massive
cork slices
stuck together

Fig. 7. Side view of a natural cork stopper freshly uncorked from a standard champagne
bottle used in this set of experiments; it clearly appears that it is composed of two dierent
parts: (i) a upper part composed of agglomerated cork granules, and (ii) a lower part made
of two massive cork slices stuck together.

3.4 An adiabatic expansion while cork popping


While cork popping, the gaseous volume in the headspace under the cork instantaneously experiences a drop of pressure (from the pressure inside the bottle before
opening strongly temperature dependent, as seen in Fig. 3 to the ambient pressure close to 1 bar). This sudden drop of pressure is therefore linked with an adiabatic
expansion of this gaseous volume trapped under the cork (considered as an ideal gas).
Assuming an adiabatic expansion experienced by the gas volume v of the headspace
while cork popping (from the pressure inside the bottle, denoted P , to the ambient
pressure of about 1 atm), the corresponding volume expansion of gaseous CO2 may
easily be accessed by the following and well known relationship:
P v = constant

(9)

14

The European Physical Journal Special Topics

High-speed infrared
video camera

black body
placed behind
the bottleneck

control monitor

Fig. 8. Setup used to visualize, through high-speed infrared imaging, the volutes of gaseous
CO2 gushing out of the bottleneck while cork popping.

Fig. 9. Time-sequences showing the ow of gaseous CO2 gushing out of the bottleneck while
cork popping, as seen through high-speed infrared imaging, for a bottle stored at 12 C; the
time interval between each frame of time-sequences is 10 ms (reprinted from ref. [34]).

where is the ratio of specic heats of the gas volume experiencing adiabatic expansion (mainly composed of gaseous CO2 and being equal to 1.3).
By application of Eq. (9), the following relationship can be deduced, which links
the nal volume vf of gaseous CO2 gushing out of the bottleneck with its initial
volume v, and with both the pressure inside the bottle before cork popping (denoted
P ), and the ambient pressure (denoted P0 and equivalent to 1 bar):

vf = v

P
P0

1/


v

P
P0

0.77
.

(10)

Champagne Fizzics

15

130

120

vf (cm3)

110

100

90

80

70
2

10

12

14

16

18

20

T (C)
Fig. 10. Theoretical nal volume vf of gaseous CO2 vapors (in cm3 ), which expand out of
the champagne bottleneck assuming adiabatic expansion while cork popping, as a function
of initial champagne temperature T ; this graph was built by replacing in Eq. (11), each
parameter by its numerical value corresponding to our batch of bottles.

By replacing P in the previous equation, by its theoretical expression found in Eq. (4),
the volume of gaseous CO2 gushing out of the bottleneck while cork popping may be
expressed as:

0.77
mT
vf v
.
(11)
4, 4 103 v + kH T V
The dependence of the nal volume vf of gaseous CO2 gushing out of the bottleneck
with champagne temperature is displayed in Fig. 10. It can clearly be seen from
Fig. 10 that the volume of gaseous CO2 gushing out of the bottleneck increases with
the champagne temperature.
This adiabatic expansion of gases trapped in the headspace under the cork during
cork popping is also inevitably linked with a drop of temperature of the gaseous
volume owing out of the bottle. By combining Eq. (9) with the ideal gas law which
states that P v = nRT (with n being the number of moles of gas experiencing adiabatic
expansion), the corresponding theoretical drop of temperature experienced by the gas
volume may be accessed by the following and well known relationship:
P (1) T = constant.

(12)

By application of Eq. (12), the following relationship can therefore be deduced, which
links the nal temperature Tf of gaseous CO2 gushing out of the bottleneck with its
initial temperature T before cork popping, the pressure inside the bottle before cork
popping, denoted P , and the ambient pressure (denoted P0 and equivalent to 1 bar):

Tf = T

P
P0

 1


T

P0
P

0.23
.

(13)

16

The European Physical Journal Special Topics


-75

Tf (C)

-80

-85

-90

-95
2

10

12

14

16

18

20

T (C)
Fig. 11. Theoretical nal temperature Tf of gaseous CO2 vapors (in cm3 ), which expand
out of the champagne bottleneck assuming adiabatic expansion while cork popping, as a
function of initial champagne temperature T ; this graph was built by replacing in Eq. (14),
each parameter by its numerical value corresponding to our batch of bottles.

By replacing P in the previous equation, by its theoretical expression found in Eq. (4),
the nal temperature of the gaseous CO2 volume gushing out of the bottleneck while
cork popping may be expressed as:

0.23
4.4 103 v + kH T V
Tf T
.
(14)
mT
The dependence of the nal temperature Tf of gaseous CO2 gushing out of the bottleneck with the initial champagne temperature is displayed in Fig. 11.
Paradoxically, and quite counter intuitively, the higher the initial champagne temperature, the lower the temperature reached by the volume of gaseous CO2 gushing
out of the bottleneck while cork popping and adiabatic expansion. Now, it is worth
noting that the lower the temperature of the volume of gaseous CO2 , the higher its
density. Actually, the overall dynamics of gaseous CO2 gushing out of the bottleneck
should indeed therefore vary with the initial champagne temperature.
Further experimental investigations should be conducted now with bottles stored
at dierent temperatures in order to corroborate (or not) the strong inuence of
temperature on both the volume and the overall dynamics of gaseous CO2 gushing
out of the bottleneck while cork popping [37].
3.5 Potential energy released while cork popping
Does all the energy released while cork popping is converted into the form of
kinetic energy with the ying cork stopper? A temptative answer based on classical
thermodynamics follows.

Champagne Fizzics

17

The total energy released while cork popping, during adiabatic expansion,
expresses as follows:
 vf
ET =
P dv.
(15)
v

By using Eq. (9), and by replacing the term P by P = constant/v , Eq. (15) transforms as:
 vf
dv
ET = constant
.
(16)
v
v
By replacing in the previous relationship the constant by, for example, constant =
P0 vf , which corresponds to the nal stage after adiabatic expansion, Eq. (16) transforms as:
 vf
dv
ET = P0 vf
.
(17)
v
v
Integrating Eq. (17) between the initial and the nal volume of gaseous CO2 during
complete adiabatic expansion, and developing leads to the following relationship:



 v 1 
1
1
f

1
1
ET = P0 vf
P 0 vf
v
v
.
(18)
1
(1 ) f
(1 )
v
By replacing, in the preceding equation vf by its theoretical expression given in
Eq. (11), leads to the following relationship, which links the energy released while
cork popping with the pertinent parameters of our system:

 
0.77

mT
mT
1
. (19)
ET P0 v

( 1)
4.4 103 v + kH T V
4.4 103 v + kH T V
Finally, replacing each and every parameter by its numerical value allows us to retrieve
the total energy released while cork popping, as a function of champagne temperature,
as displayed in Fig. 12. In the graph displayed in Fig. 12, the energy released into
the form of kinetic energy (EK = 1/2M U 2 ) with the cork popping also appears
for the champagne bottles stored at 12 C. As clearly seen in Fig. 12, only a small
fraction of the total energy released while cork popping is converted into the form of
corks kinetic energy (only about 5%). Most of the total energy seems therefore to be
dissipated into the form of a sound shock wave (the very characteristic bang).

3.6 A structural analogy


I would like to end this paragraph by evidencing a nice structural analogy. The two
photographs displayed in Fig. 13 illustrate the similarity between the vortex ring
which follows an aboveground nuclear test, and the one which follows the champagne
cork popping process.
It is worth noting that, despite tremendous dierences in their respective Reynolds
numbers (Re = DU /, with being the density of gaseous ow, D and U being its
characteristic diameter and velocity, and being its dynamic viscosity), the same
large-scale structure may be observed between both ascending gaseous ows, with a
primary vortex ring followed by tails. More details about the vorticity generation
mechanisms can be found in the work by Sigurdson [38].
After the uncorking of the bottle, the pressure of gaseous CO2 experienced by
the champagne inside the bottle suddenly falls. Consequently, the thermodynamic
equilibrium of CO2 is broken, and the dissolved CO2 content found in champagne

18

The European Physical Journal Special Topics

100

ET , EK (J)

10

0,1
2

10

12

14

16

18

20

T (C)
Fig. 12. Total energy released while cork popping during the adiabatic expansion of the
gaseous volume of CO2 trapped under the cork, as a function of champagne temperature
(blue line), compared with the kinetic energy (EK = 1/2M U 2 ) of the cork popping out of
the bottleneck at 12 C (white circle); in the term EK = 1/2M U 2 , the mass of the cork
stopper M is expressed in kg, and its velocity U in m/s.

must inevitably escape to recover equilibrium, and fulll Henrys law, as stipulated
by Eq. (2).
After having been uncorked by the taster, the bottle is then generally inclined
to pour champagne into a ute or a coupe. The aim of the following part is to
closely examine how dissolved CO2 progressively escapes from the liquid medium to
progressively invade the headspace above glasses, thus progressively modifying the
gaseous chemical space perceived by the consumer.

4 Losses of dissolved CO2 during champagne serving


The impact of the pouring step on the loss of dissolved CO2 from champagne has
remained completely unexplored until very recently. During the pouring step, the free
air/champagne interface indeed considerably increases. Actually, champagne ows
from the bottle to the tasting glass into the form of a liquid tongue, which presents
a very large interface with regard to its own volume [39]. Moreover, the pouring step is
a hugely turbulent phase, with possible entrapment of air bubbles inside champagne,
which also contribute to the diusion of dissolved CO2 from the liquid phase to the
ambient air. Besides, champagne is strongly mixed during the pouring step, with
formation of eddies of various sizes in the liquid matrix, which also undoubtedly
contribute to signicant losses of dissolved CO2 .
The impact of the pouring step on the loss of dissolved CO2 from champagne was
examined, depending on the way champagne is served into the ute [39]. Moreover,
impacts on losses of dissolved CO2 of three champagne service temperatures were also
examined (i.e., 4 C, 12 C, and 18 C, respectively).

Champagne Fizzics

(a)

19

(b)

Fig. 13. Large-scale structural analogy experienced by the ow of gaseous CO2 gushing
out of the bottleneck while cork popping (with a Reynolds number of order of several 104 )
(a) and the gaseous ow driven by the reball following an atomic blast in Nevada (with a
likely Reynolds number of several billions) (b). (US department of Energy.)

4.1 Impact of the way champagne is served, and the role of temperature
After uncorking a bottle, two ways of serving 100 mL of champagne into the ute
were examined with regard to their respective impacts on losses of dissolved CO2 :
(i) a standard champagne-like, and (ii) a standard beer-like way of serving. Photographs displayed in Fig. 14 illustrate both possible ways of serving champagne into
the ute.
(i) During the standard champagne-like way of serving, champagne vertically falls
and hits the bottom of the ute (thus usually providing a thick head of foam, which
quickly vertically extends, and then progressively collapses during serving). This way
is the traditional way of serving champagne and sparkling wines in bars, clubs, and
restaurants.
(ii) During the beer-like way of serving, champagne ows along the inclined ute
wall and progressively lls the ute. The axis of symmetry of the ute inevitably
and progressively recovers its vertical position during serving. This beer-like way of
serving champagne is less turbulent, with usually much less foam generation.
Impacts on losses of dissolved CO2 in champagne were investigated for three sets of
champagne temperatures, i.e., 4 C, 12 C, and 18 C, respectively, and for both ways
of serving champagne. A volume of 100 4 mL of champagne was carefully poured
into the ute. It is worth noting that the heat capacity of the glass ( 0.8 kJ kg1
K1 ) being much lower than that of champagne ( 4.2 kJ kg1 K1 ), the temperature of champagne remains almost constant during the few seconds of the pouring
process [40]. To enable a statistical treatment, four successive pouring and dissolved

20

(1)

The European Physical Journal Special Topics

(2)

(1)

(2)

(3)

(3)

(4)

(4)

(a)

(b)

Fig. 14. Time-sequences illustrating both ways of serving 100 milliliters of champagne into
the ute; the champagne-like way of serving (this way is the traditional way of serving
champagne and sparkling wines in bars, clubs, and restaurants) (a); the beer-like way of
serving, traditionally reserved for beer serving (to prevent an excessive formation of foam)
(b). (Gerard Liger-Belair.)

CO2 measurements were done for both ways of serving champagne, and for each
champagne temperature. An arithmetic average of the four data provided by the four
successive pouring were done, to nally produce one single average dissolved CO2
concentration corresponding to a given way of serving champagne at a given temperature (with standard deviations corresponding to the root-mean-square deviations of
the values provided by the four successive pouring). Concentrations of dissolved CO2

Champagne Fizzics

21

10

-1

CL (g L )

Beer-like
Champagne-like

7
2

10

12

14

16

18

20

T (C)
Fig. 15. Concentrations of dissolved-CO2 , as chemically measured once champagne was
poured into the ute, for both ways of serving champagne, and for each champagne temperature; each dot is the arithmetic average of four successive values provided from four
successive pouring (for both ways of serving champagne); standard deviations correspond
to the root-mean-square deviations of the values provided by the four successive data
recordings.

data, denoted cL , as chemically measured immediately after pouring champagne into


the ute, are displayed in Fig. 15 (for both ways of serving, and for each champagne
temperature). It is clear from Fig. 15 that the higher the temperature of champagne,
the lower its concentration of dissolved CO2 after pouring. Moreover, it also clearly
appears that the beer-like way of serving champagne is signicantly less impacting
its concentration of dissolved CO2 after the pouring step.
4.2 The pouring of champagne revisited through high-speed infrared imaging
The visualization technique based on infrared imaging has also been used to lm the
gaseous CO2 uxes outgassing from champagne during the pouring step [39]. Snapshots displayed in Fig. 16 illustrate the pouring steps, as seen through the objective
of the IR video camera, for both ways of serving champagne (at 4 C, and at 18 C,
respectively).
It therefore clearly appears, through IR video camera recordings, that the higher
the champagne temperature is, the more easily dissolved CO2 escapes during the
pouring process, thus conrming the tendency underscored by chemical measurements (for both ways of serving champagne). It is also worth noting that the ow of
gaseous CO2 seems to ow down from the opening of the ute, by licking the glass
walls, rather than by diusing isotropically all around it. This observation betrays
the fact that gaseous CO2 is approximately 1.5 times denser (CO2 = 1.87 gL1 at
20 C) than dry air is (air = 1.20 gL1 at 20 C), and therefore tends to naturally
ow down.
Ambient air being considered as a huge thermal tank, it quickly thermalizes the
volume of gaseous CO2 outgassing from champagne during the pouring step. Therefore, by considering the ux of gaseous CO2 discharging from champagne as an ideal

22

The European Physical Journal Special Topics

(a)

(c)

(b)

(d)

Fig. 16. Snapshots illustrating both ways of serving champagne, as seen through the objective of the IR video camera for a bottle stored at 4 C (a) and (b), and for a bottle stored
at 18 C (c) and (d).

gas, the volume of gaseous CO2 discharging during the pouring step, denoted V20 C
(expressed in cm3 ), expresses as follows,
V20 C = 106

cvf RT
MCO2 P0

(20)

with c being the loss of dissolved-CO2 concentration during the pouring step (i.e.,
c = ci cL , expressed in g L1 ), vf being the volume of champagne poured into
the ute (in L, namely 0.1 L in the present case), T being the ambient temperature
(namely 293 K in the present case), MCO2 being the molar mass of CO2 (44 g mol1 ),
and P0 being the atmospheric pressure (close to 105 Pa).
Besides, Table 3 compiles losses of dissolved CO2 concentrations (with regard to
the initial concentration ci before pouring of ci 11.5 g L1 ) as well as the corresponding volume of gaseous CO2 escaping during the pouring step. It is worth noting
that during the few seconds of the champagne-like way of serving, champagne loses
approximately as much dissolved CO2 as during the rst 10 minutes following pouring,
as shown in two papers [24,25]. To prolong the drinks chill and to help champagne
retain its CO2 content (and therefore its eervescence), the champagne-like way of
serving is certainly not the most appropriate.

Champagne Fizzics

23

Table 3. Losses of dissolved-CO2 concentrations in champagne during the pouring step (in
g L1 ), and corresponding volumes of gaseous CO2 discharged from champagne (converted
in cm3 ), for both ways of serving, and at three champagne temperatures.
Temperature

Champagne-like


 1 
VCO2 cm3
c gL

Beer-like
 1 


c gL
VCO2 cm3

4 C
12 C
18 C

3.0 0.2
3.3 0.2
4.0 0.5

1.6 0.2
2.0 0.3
3.7 0.3

166 11
182 11
221 28

88 11
110 17
204 17

4.3 The molecular mechanism behind the loss of dissolved-CO2 during serving
Molecular diusion is actually the mechanism behind the progressive desorption of
dissolved gas species from the free surface area of a supersaturated liquid medium
(as dissolved-CO2 molecules continuously do from the free air/champagne interface,
once the bottle is uncorked). The number of CO2 moles that cross the air/champagne
interface per unit of time is ruled by:

dN

J dS
(21)
=
dt
air/champagne
interface


where J is the ux of CO2 molecules dened by the rst Ficks law, J = D c.
In the latter equation, D is the diusion coecient of CO2 molecules in the liquid
 is the gradient of CO2 -dissolved molecules between the champagne
matrix, and c
bulk and the air/champagne interface in equilibrium with the gaseous CO2 in the
vapor phase outside the liquid phase.
By assuming a linear gradient of dissolved-CO2 between the champagne bulk and
 may be rewritten as c/, with being the thickness
the air/champagne interface, c
of the boundary layer where a gradient of dissolved-CO2 exists, and c = cL cI is
the dierence in dissolved-CO2 concentrations between the liquid bulk (denoted cL ),
and the air/champagne interface (denoted cI ) in equilibrium with the gaseous CO2
in the vapor phase (see Fig. 17).
Generally speaking, desorption of dissolved gas species is ruled by pure diusion
or by diusion-convection, whether the supersaturated liquid medium is perfectly
stagnant or in motion [41]. In case of a liquid medium agitated with ow patterns,
convection forbids the growing of the diusion boundary layer by supplying the liquid near the free surface area with dissolved gas molecules freshly renewed from the
liquid bulk. Pouring champagne into a ute being a hugely turbulent phase with formation of eddies and convection currents, no doubt that the mechanism behind the
loss of dissolved CO2 molecules from champagne during the pouring step is ruled by
diusion-convection.
It clearly appears from Table 3 that, the higher the temperature of champagne, the
higher its loss of dissolved-CO2 during the pouring step. Why such a dependence?
Actually, the diusion coecient of the CO2 molecule denoted D, which rules
the diusion rate through the air/champagne interface, is strongly temperaturedependent. D may indeed be approached through the well-known Stokes-Einstein
equation as follows:
kB T
D
(22)
6a

24

The European Physical Journal Special Topics

Fig. 17. Detail of the air/champagne boundary layer where a gradient of dissolved-CO2
exists; turbulences of the pouring process impact the thickness of the boundary layer, and
therefore the ux of dissolved-CO2 diusing through the air/champagne interface.

with kB being the Boltzman constant (1.38 1023 J K1 ), being the champagne
viscosity, and a being the characteristic size of the CO2 molecules hydrodynamic
radius (a 1010 m).
In the range of usual champagne tasting temperature (varying from approximately 5 C to 15 C), the temperature dependence of champagne was measured with
a thermostated Ubelhode capillary viscosimeter (with a sample of champagne rst
degassed), and found to classically obey the following Arrhenius-like equation [3]:
(T ) 1.08 104 exp(2806/T )

(23)

with the dynamic viscosity being expressed in mPa s, and the temperature T being
expressed in K.
It is worth noting that, since the viscosity of champagne is strongly temperaturedependent, as seen in Eq. (23), the diusion coecient of dissolved-CO2 molecules
is also in turn strongly temperature-dependent. By combining Eqs (22) and (23), D
may be rewritten as follows:


2806
D T exp
.
(24)
T
Following Eq. (24), the lower the champagne temperature is, the lower the diusion
coecient of dissolved-CO2 molecules. Therefore, it is no wonder that, the lower the

Champagne Fizzics

25

champagne temperature is, the lower the loss of dissolved-CO2 during the pouring
step, as reported in Table 3.
Besides, since the lower the champagne temperature the higher its viscosity, ow
patterns induced by turbulences of the pouring step certainly calm down more rapidly
(by viscous dissipation) at low champagne temperatures, thus limiting also the loss
of dissolved CO2 molecules by diusion through the air/champagne interface. As
clearly seen in Fig. 15, the beer-like way of serving champagne much less impacts
its dissolved-CO2 concentration than the champagne-like way of serving it, and especially at low champagne temperatures (4 C and 12 C). The beer-like way of serving champagne is much softer than the champagne-like one. Turbulences in the
ute are therefore expected to be much less important during the pouring step in
case of the beer-like way of serving, thus reducing in turn the loss of dissolved-CO2
molecules by diusion-convection in comparison with the champagne-like way of serving champagne. Moreover, turbulences being less important during the beer-like way
of serving, the champagne is supposed to calm down even more rapidly at low champagne temperatures (by viscous dissipation), thus explaining the rather low loss of
dissolved-CO2 during the beer-like way of serving (at low temperatures) compared
with the champagne-like one. Therefore, the gap in the loss of dissolved-CO2 between
12 C and 18 C for both ways of serving champagne (as seen in Table 3 and in
Fig. 15) is interpreted as an eect of the lower viscosity of champagne at 18 C.

5 The bubble nucleation process


Champagne, sparkling wines, and to a larger extent carbonated beverages in general,
are characterized by the presence of bubbles rising gracefully into the form of very
characteristic bubble columns, or bubble trains. By the naked eye, bubbles seem to
continuously nucleate from several specic sites of the glass wall, called nucleation
sites. But why, how, and where do all these bubbles really form?
5.1 Champagne: A liquid phase supersaturated with CO2 dissolved molecules
In addition to this sudden temperature drop experienced by gases from the headspace,
the fall of CO2 partial pressure above the champagne surface linked with bottle uncorking leads to a huge consequence concerning the thermodynamic equilibrium of
CO2 -dissolved molecules. Since the partial pressure of CO2 falls above the champagne surface, the CO2 dissolved in champagne is not in equilibrium any longer with
its partial pressure in the vapor phase. Champagne enters a metastable state, i.e.,
it contains CO2 molecules in excess in comparison with what Henrys law states. To
recover a new stable thermodynamic state corresponding to the partial pressure of
CO2 molecules in the atmosphere (about only 3.5 104 atm), almost all the carbon
dioxide molecules dissolved into the champagne must escape. The champagne becomes
supersaturated with CO2 . Before proceeding further, it is important to dene the
supersaturating ratio, used for quantifying CO2 molecules in excess in a carbonated
liquid. The supersaturating ratio S is dened as follows [42]:
S=

cL
1
c0

(25)

where cL is the concentration of CO2 in the liquid bulk, and c0 is the equilibrium
concentration of CO2 corresponding to a partial pressure of gaseous CO2 of 1 atm.
As soon as S > 0, a supersaturated liquid enters a metastable state and must degas
to recover a supersaturating ratio equal to zero. In the case of Champagne wines, just

26

The European Physical Journal Special Topics

after uncorking the bottle, cL is the equilibrium concentration of CO2 in the liquid
bulk corresponding to a partial pressure of CO2 of about 5 atm. Because there is
a strict proportionality between the concentration of dissolved CO2 and its partial
pressure in the vapor phase (as expressed by Henrys law), cL /c0 5. Therefore, just
after uncorking the bottle, the supersaturating ratio of champagne is approximately
S 4, and champagne must degas. Actually, there are two mechanisms for gas loss:
(i) losses due to diusion through the surface of the liquid (invisible by the naked
eye), and (ii ) losses due to bubbling (the so-called eervescence process).
5.2 A critical radius required for bubble nucleation
Generally speaking, carbonated beverages are weakly supersaturated with
CO2 -dissolved gas molecules. In weakly supersaturated liquids such as champagne
and sparkling wines, bubbles do not just pop into existence ex nihilo. Actually, to
cluster into the form of bubbles, CO2 -dissolved gas molecules must cluster together
and push their way through the liquid molecules that are held together by Van der
Waals attractive forces. Bubble formation is therefore limited by an energy barrier
(for a complete review see the paper by Lugli and Zerbetto [43]). This is the reason
why in weakly supersaturated liquids, bubble formation and growing require preexisting gas cavities with radii of curvature large enough to overcome the nucleation
energy barrier and grow freely [4446]. This critical radius, denoted r , can easily be
accessed by using standard thermodynamic arguments, or by using simple arguments
based on classical diusion principles. The critical radius r of gas pockets required to
enable bubble production in a carbonated beverage expresses as follows (see reference
[3] and references therein),
2
r
(26)
P0 S
where is the surface tension of the liquid medium (of order of 50 mN/m in champagne and sparkling wines), and P0 is the atmospheric pressure (P0 105 N/m2 ).
Immediately after having uncorked a champagne bottle, because S 4, and following Eq. (26), the critical radius required to enable bubble nucleation is therefore
of order of 0.25 m. Jones et al. made a classication of the broad range of nucleation likely to be encountered in liquids supersaturated with dissolved gas molecules
[45]. Bubble formation from preexisting gas cavities larger than the critical size is
referred to as non-classical heterogeneous bubble nucleation (type IV bubble nucleation, following their nomenclature). Generally speaking, eervescence in a glass of
champagne or sparkling wine may have two distinct origins. It can be natural or
articial.
5.2.1 Natural bubble nucleation
Natural eervescence is related to the bubbling process from a glass which has not
experienced any specic surface treatment. Closer inspection of such glasses poured
with champagne and other sparkling wines was conducted through a high-speed video
camera tted with a microscope objective (see the photograph displayed in Fig. 18).
It revealed that most of the bubble nucleation sites were found to be located on preexisting gas cavities trapped inside hollow and roughly cylindrical cellulose-ber-made
structures on the order of 100 m long with a cavity mouth of several micrometers
(see Fig. 19) [3,6,7,47].
The hollow cavity (a kind of tiny channel within the bers) where a gas pocket
is trapped during the pouring process is called the lumen. It can be clearly noticed

Champagne Fizzics

27

Fig. 18. High-speed video camera used to visualize bubble nucleation sites in a glass poured
with champagne. (Photograph by Hubert Raguet.)

Fig. 19. Three typical cellulose bres adsorbed on the wall of a glass poured with champagne;
the gas pockets trapped inside the bres lumen and responsible for bubble formation clearly
appear (bar = 100 m). (Photographs by Gerard Liger-Belair.)

28

The European Physical Journal Special Topics

from Fig. 19 that the radii of curvature of gas pockets trapped inside the bres
lumen are much higher than the above-mentioned critical radius r. Fibers probably
adhere on the ute wall due to electrostatic forces (especially if the glass or the ute
is vigorously wiped by a towel). Natural eervescence may also arise from tartrate
crystals precipitated on the glass wall and resulting from the evaporation process after
rinsing the glass with tap water. Therefore, there is a substantial variation concerning
the natural eervescence between utes depending on how the ute was cleaned
and how and where it was left before serving.
5.2.2 Artificial bubble nucleation
Articial eervescence is related to bubbles nucleated from glasses imperfections done
intentionally by the glassmaker to promote or to eventually replace a decit of natural nucleation sites. Actually, it has been known for decades that bubbles may arise
from microscratches on the glass wall [48,49]. Those microscratches are geometrically
able to trap tiny air pockets when champagne is poured into the glass (as cellulose
bers do). Those microscratches on a glass can be done by essentially two techniques:
sandblast or laser engraving. Nevertheless, eervescence produced from scratches intentionally done by the glassmaker does not resemble that arising from tiny individual
cellulose bers. A rendering of such micro scratches releasing bubbles at the bottom
of a champagne ute is displayed in Fig. 20. It is worth noting that the repetitive
bubbling process arising from articial bubble nucleation is much more vigorous and
chaotic than the bubbling process from tiny cellulose bers. Glasses engraved at their
bottom are thus indeed easily recognizable, with a characteristic bubble column rising on their axis of symmetry [5052]. Eervescence promoted by engraved glasses
is indeed visually quite dierent than that naturally promoted by cellulose bers,
but the dierence is also suspected to go far beyond the sole esthetical (and rather
subjective) point of view. Dierences are strongly suspected concerning the kinetics
of CO2 and avour release all along champagne tasting [2325].
5.3 Entrapping an air pocket within a ber
Cellulose bers are in the form of hollow tubes of several hundreds of micrometers
long and with a cavity mouth of several micrometers wide. The ber wall section consists of densely packed cellulose micro brils, with a preferential orientation along the
ber axis. Cellulose micro brils consist of glucose units bounded in a -conformation
favouring straight polymer chains. The dierent structural levels of a cellulose ber
are presented in Fig. 21. For a current review on the molecular and supra molecular
structures of cellulose, see the article by OSullivan and references therein [53].
From the physics point of view, cellulose bers can indeed be considered as tiny
roughly cylindrical capillary tubes of radius r and length h. Consequently, a wetting
liquid placed into contact with this highly hydrophilic material penetrates it by capillary action. Actually, in capillaries with radii much smaller than the capillary length,
gravity may be neglected. Therefore, being the viscosity of the liquid phase, being
the surface tension of liquid, z being the distance of penetration at time t, and the
eective contact angle between the liquid and the capillary wall, the overall balance
of forces on the liquid in the capillary may be expressed as,

d2 z
z 2 +
dt

dz
dt

2

2 cos 8z dz
2
.
r
r dt

(27)

Champagne Fizzics

(b)

29

(c)

(a)

Fig. 20. At the bottom of this ute, on its axis of symmetry, a small ring (done with adjoining laser beam impacts) has been etched by the glassmaker (a); single laser beam impact
as viewed through a scanning electron microscope (bar = 100 m) (b); eervescence in this
ute is promoted from these articial micro scratches into the form of a characteristic
and easy recognizable vertical bubbles column rising on its axis of symmetry (bar = 1 mm)
(Photographs by F. Beaumont) (c).

The left-hand side of the latter equation is related to the liquid inertia, whereas
both terms in the right-hand side are related to capillarity (the driving force), and
viscous resistance, respectively. Under steady conditions, capillarity is balanced by the
viscous drag of the liquid, and the famous Lucas-Washburns equation can be derived
[54,55]:
z2 =

r cos
t.
2

(28)

Lets imagine a liquid edge spreading with a velocity v along a solid surface where
cellulose bers are adsorbed. This is basically what happens when you ll a glass with
a liquid. Actually, a liquid edge progressively advances along the vertical glass wall at
a velocity v of order of several cm/s. As soon as the wetting liquid gets in touch with
the ber, some liquid progressively penetrates and lls the bers lumen by capillary
rise. Finally, a gas pocket may be trapped within the ber if the time taken by the
liquid to completely ll the lumen by capillary action overcomes the characteristic
time T taken by the liquid edge to completely submerge the ber inside the liquid
(see the scheme displayed in Fig. 22).

30

The European Physical Journal Special Topics

h 100 m 1 mm
fiber wall
e 1-5 m

d 10 20 m

lumen

cellulose micro fibrils network


Fig. 21. The dierent structural levels of a typical cellulose bre; the ber wall consists of
closely packed cellulose micro brils oriented mainly in the direction of the bre.

By retrieving Eq. (28) with the characteristic bers parameters dened in Fig. 21,
the characteristic time required to completely ll the bers lumen by capillary action
may be expressed as,
=

2h2
.
r cos

(29)

Considering a ber with a length h, inclined by an angle with regard to the liquid
edge advancing over it at a velocity v, leads to the following time required for the
ber to be completely submerged:
T =

h sin
.
v

(30)

Champagne Fizzics

31

air

v
v

u
d = 2r

champagne
Fig. 22. From the physics point of view, a ber may be seen as a tiny capillary tube
which gets invaded by a wetting liquid placed into contact with one of the bers tip; v is
the velocity of the liquid edge advancing over the ber, and u is the velocity at which the
meniscus advances inside the bers lumen by capillary action.

The condition of gas entrapment inside the ber therefore expresses as > T , i.e.,
h sin
2h2
>
.
r cos
v

(31)

Because cellulose is a highly hydrophilic material, the contact angle of an aqueous


liquid on it is relatively small (about 30 with pure water). Consequently, cos 1.
Finally, the condition of entrapment may be rewritten as follows,

h
>
r sin
2v

(32)

with the geometric parameters of the cellulose ber lying on the left-hand side of
Eq. (32), and the liquid parameters lying on the right-hand side of Eq. (32).
The entrapment of an air pocket inside the lumen of a ber during the lling of a
glass is therefore favoured by the following conditions, depending on both ber and
liquid parameters: (i) as elongated bers as possible (h long) (ii) of small lumens radii
r, (iii) bers as horizontal as possible with regard to the liquid edge (i.e., sin small),
(iv) liquids with a small surface tension , (v) and a high viscosity , and nally (vi)
a high velocity for the liquid edge advancing along the glass wall. It is worth noting
that both conditions (iv) and (v) imply that hydroalcoholic carbonated beverages are
more favourable than zzy waters to entrap air pockets inside cellulose bers during
the pouring process. Actually, the surface tension of champagne and beer is of order
of 50 mN/m (i.e., about 20 mN/m less than the surface tension of pure water), and
their dynamic viscosity is about 50% higher than that of pure water [3].

32

The European Physical Journal Special Topics

Fig. 23. A curved-shaped ber naturally increases the time taken by the liquid to fully
invade the bers lumen (bar = 50 m).

Remember that this model considers bers as straight capillary tubes. It is worth
noting that real bers are nevertheless often curved-shaped, which certainly increases
the likelyhood of air entrapment when pouring, by increasing the time required for
the liquid to completely invade the bers lumen by capillary action during the full
immersion of the ber (see for example the ber displayed in Fig. 23).
5.4 Modeling the bubble dynamics inside the bers lumen
A typical time-sequence illustrating one period of the cycle of bubble production
from a typical cellulose ber is displayed in Fig. 24. The gas pocket trapped inside
the bers lumen strikingly resembles, in miniature, a Taylor-like bubble trapped inside a micro-channel.
Because it is immersed into a liquid supersaturated with dissolved gas molecules,
the gas pocket surface acts as a nucleation area. Gas molecules dissolved into the
liquid matrix therefore diuse inside the trapped gas pocket through the gas/liquid
interface. In turn, the trapped gas pocket grows inside the bers lumen until it
reaches a bers tip (from Fig. 24(a) to Fig. 24(e)). As the gas pocket reaches the
tip of the ber, a tiny bubble is ejected (Fig. 24(f)), but a portion of the gas pocket
remains trapped inside the ber, shrinks back to its initial position, and the cycle
starts again until bubble production stops through lack of dissolved gas molecules.
The whole process leading to the production of a bubble can be coarsely divided in
two main steps: (i) the growth of the gas pocket trapped inside the bers lumen,
and (ii) the bubble detachment as the gas pocket reaches the bers tip. Actually, the

Champagne Fizzics

33

(a)

(b)

(c)

(d)

(e)

(f)

Fig. 24. Time-sequence illustrating one period of the cycle of bubble production from a
typical cellulose ber adsorbed on the wall of a glass poured with champagne; from frame
a to frame e, the time interval between each frame is about 100 ms, whereas it is only 1 ms
between frame e and f (bar = 50 m).

time scale of the bubble detachment is small ( 1 ms) compared with the relatively
slow growth of the gas pocket (several tens to several hundreds of ms). Therefore,
the cycle of bubble production seems to be largely governed by the growth of the gas
pocket trapped inside the bers lumen.
To go further in with the dynamics of a gas pocket growing trapped inside the
lumen of a ber, a theoretical model has been developed [7]. Actually, those elongated gas pockets trapped inside the lumen of cellulose bers are bounded by two
hemispherical gas/liquid interfaces of radius r, in touch with the liquid bulk inside
the ber (both ends of the gas pocket), and by a nearly cylindrical border of radius r and length z, in touch with the ber inner wall. A scheme is displayed in
Fig. 25, where the geometrical parameters of this growing Taylor like bubble are
dened.

34

The European Physical Journal Special Topics

Molecular diusion is the mechanism behind the bubble growth, and the number
of CO2 moles that cross the gas pocket interface per unit of time is ruled by:
 
dN

J dS
(33)
=
dt
gaz pocket
surface


where J is the ux of CO2 molecules dened by the rst Ficks law, J = D c.

In the latter equation, D is the diusion coecient of CO2 molecules, and c is
the gradient of CO2 -dissolved molecules between the liquid bulk and the boundary
layer in equilibrium with the CO2 gas molecules in the vapor phase inside the gas
pocket.
Assuming the gas into the trapped gas pocket as an ideal gas, and assuming a nearly
constant section ( r2 ) for the cylindrical body of the gas pocket, the rate at which
the Taylor-like bubble grows inside the bers lumen, dz/dt, may be linked to the
number of CO2 moles, dN/dt, that cross the gas pocket interface through,
 dN
dz
=
dt
PB r2 dt

(34)

where  is the ideal gas constant, is the absolute temperature, and PB is the pressure
inside the gas pocket.
Actually, there are two ways in which CO2 molecules can diuse toward the gas
pocket: (i) through the two nearly spherical caps directly in touch with the liquid
bulk, and (ii) through the cylindrical border in touch with the ber wall. The diusion coecient of CO2 molecules diusing through the spherical caps is denoted D0 ,
whereas that of CO2 molecules diusing through the ber wall (and therefore perpendicular to the cellulose microbrils) is denoted D . D0 is the diusion coecient
of CO2 molecules in the liquid bulk. It was measured by NMR (Nuclear Magnetic
Resonance) techniques in various carbonated beverages [56]. In champagne, it was
found to be D0 1.4 109 m2 s1 , at 20 C. The transversal diusion coecient
D of CO2 molecules through the ber wall was approached and properly bounded
by D /D0 0.1 and D /D0 0.3 [57]. For modeling purposes, an intermediate
value of about D 0.2 D0 was proposed and will be used hereafter [57]. By taking
into account the two dierent ways for the diusing CO2 molecules, and because there
are two spherical caps, Eq. (30) may be rewritten as follows,


dNSC
dNF W
dN
+

JSC dS
JF W dS
=
+
=2
(35)
dt
dt
dt
spherical cap

fiber wall

where the subscripts SC and FW correspond to the spherical cap gas/liquid interface
and ber wall, respectively.
Geometrically, the whole surface of the two spherical caps is SSC 4r2 , whereas
that of the cylindrical border in touch with the ber wall is SF W 2rz. Finally, by
assuming the ux of CO2 molecules as being constant along each given surface, and
by assuming a linear gradient of CO2 concentration in the boundary layer around the
gas pocket, Eq. (32) may transform as,
c
c
dNSC
dNF W
dN
=
+
4r2 D0
+ 2rzD
dt
dt
dt
SC
F W

(36)

where is the boundary layer thickness, and c = cL cB (the driving force responsible for the diusion of CO2 into the rising bubble) is the dierence in CO2 -dissolved

Champagne Fizzics

35

(a)

gas pocket

(b)

z
gas
pocket

cB

boundary layer

liquid bulk

cL

Fig. 25. Real gas pocket trapped inside the lumen of a cellulose ber acting as a bubble
nucleation site in a glass poured with champagne (bar = 50m) (a), modelled as a slugbubble trapped inside an ideal cylindrical microchannel and being fed with CO2 -dissolved
molecules diusing, (i) directly from the liquid bulk through both ends of the gas pocket,
and (ii) through the wall of the microchannel (b).

concentrations between the liquid bulk, denoted cL , and the close vicinity of the
bubble surface in equilibrium with the gaseous CO2 into the gas pocket, denoted cB .
Finally, by combining Eq. (36) with Eq. (34), the rate at which the Taylor-like
bubble grows inside the bers lumen may be rewritten as follows,


2c D
dz
2D0

z+
.
dt
PB
rF W
SC

(37)

36

The European Physical Journal Special Topics

Actually, in the close vicinity of the bubble surface, the CO2 -dissolved concentration cB is forced by the partial pressure of gaseous CO2 molecules inside the bubble,
and Henrys law locally applies, i.e., cB = kH PB , where kH is the Henrys law constant (strongly temperature-dependent). Strictly speaking, the pressure PB inside the
Taylor-like bubble is the sum of three terms: (i) the atmospheric pressure P0 , (ii) the
hydrostatic pressure gH, H being the depth at which the cellulose ber lies, and (iii)
the Laplace pressure 2/r, originated in the bubbles curvature. However, H varying
from several millimeters to several centimeters, the contribution of hydrostatic pressure is clearly negligible in front of both atmospheric and Laplace pressures. Thus,
cB = kH PB kH (P0 + 2/r). Finally, the only parameters unknown in Eq. (37) are
SC and F W , the boundary layers thicknesses around the spherical caps and the
ber wall, respectively. The aim of the two following paragraphs is to discuss around
how to reasonably approach SC and F W .
5.5 Pure diusion without convection
Generally speaking, in a stagnant medium, heat and mass transfer are ruled by pure
diusion. In the present situation, in a purely diusive case, each CO2 molecule
which diuses through the gas/liquid interface into the gas pocket is removed from
the boundary layer around. In turn, the thickness of the boundary layer depleted
with CO2 -dissolved molecules progressively expands around the gas pocket during
the diusing process. Due to the cylindrical shape of the gas pocket trapped inside
the bers lumen, the boundary layer depleted with CO2 along the cylindrical ber
wall is a cylindrical sleeve of liquid with a thickness F W . Contrary to the CO2
molecules diusing through the ber wall, the CO2 molecules diusing through the
two spherical bubble caps are removed from a nearly spherical zone of radius SC
expanding around the bers tips. Fig. 26 illustrates the shapes taken by the boundary
layers around an ideal cellulose ber, in the purely diusive case.
By use of mass conservation between the boundary layer and the gas pocket,
SC and F W may be approached. Actually, by considering the two above-dened
geometries for the boundary layers, the mass conservation of CO2 molecules between
the boundary layer and the gas pocket during the diusive process may be written
as,

82SC dSC c dNSC
.
(38)
2 (r + F W ) zdF W c dNF W
Then, by inserting the two above-mentioned expressions for dNSC and dNF W into
Eq. (36), and by integrating, the growth of SC (t) and F W (t) may be deduced as
time proceeds as follows,


1/4

SC 2D0 r2 t
1/2
(39)
F W (2D t)
for short times,
i.e., F W  r .

1/3
F W (3D rt)
for long times, i.e., F W r
Thus, as time goes on, a zone depleted with CO2 progressively expands around the
Taylor-like bubble which continuously consumes CO2 -dissolved molecules while releasing newly created bubbles. Consequently, as time goes on, the period required for
the bers tip to release a bubble should inexorably increase compared with the period required to release the preceding bubble, which was clearly not observed during
the clockwork release of more than 200 successive bubbles from the ber displayed in
Fig. 24.

Champagne Fizzics

37

SC

cL
zf

gas

pocket

cB

z0

FW

SC
Fig. 26. Sketch of the Taylor-like bubble growing trapped inside the cellulose bers lumen;
the dotted-lines limited zones are the boundary layers depleted with CO2 -dissolved molecules
and where a gradient of dissolved molecules exists because of pure diusion of CO2 molecules
from the liquid bulk to the Taylor-like bubble.

5.6 When convection plays its part


Actually, the liquid around the ber is far from stagnant mainly due to bubbles
nucleating elsewhere in the ute and rising in the bers neighborhood. In a previous
work, convection induced by rising bubbles was indeed easily underscored in a glass of
champagne, by following the fall of a dye drop after it striked the liquid surface [58].
Natural convection in the bers neighborhood is likely to forbid the growing of this
boundary layer around the ber, thus keeping it roughly constant by continuously
supplying the liquid around the ber with CO2 -dissolved molecules freshly renewed
from the liquid bulk. Therefore, the mass transfer of CO2 -dissolved molecules from
the liquid bulk to the gas pocket trapped inside the bers lumen is rather believed
to be governed by diusion-convection than by pure diusion. Consequently, during
the bubbling process, the boundary layer thickness around the gas pocket growing
trapped inside the bers lumen is believed to remain reasonably constant during
the growth of numerous successive gas pockets from a given bers lumen. Thus,
SC F W = .

38

The European Physical Journal Special Topics

z0

zf
Fig. 27. Between the release of two bubbles from the bers tip, the gas pocket trapped
inside the bers lumen oscillates between two extreme positions where its length varies from
its initial length denoted z0 and its nal length denoted zf .

Equation (37) may therefore be easily integrated. Striclty speaking, c in Eq.


(37) is time dependent. Actually, c progressively decreases with time because CO2
molecules progressively escape from the liquid medium during the bubbling process.
Nevertheless, during the growth of a single gas pocket growing trapped in the bers
lumen, c may also be considered as a constant. Finally, Eq. (37) admits a simple
analytical solution, and by integrating it, the growth of a Taylor-like bubble trapped
inside a bers lumen may be linked with both liquid and ber parameters as,

z(t) (z0 + aT) exp(t/T) aT

(40)
4D0 c
(P + 2/r) r

withT = 0
, and a =
2D c
(P0 + 2/r)
where z0 is the initial length of the Taylor-like bubble before it starts its growth
through the lumen, before each cycle of bubble production (see for example Fig. 24(a)
and 24(f)).
The whole cycle of bubble production being largely governed by the growth of the
gas pocket trapped inside the bers lumen, the period of bubble formation from a
single cellulose bre is therefore equal to the total time T required by the tiny gas
pocket to grow from its initial length, denoted z0 , to its nal length, denoted zf , as
it reaches the bers tip (see Fig. 27).
By retrieving Eq. (40), it is therefore possible to access the frequency of bubble
formation f from a single ber as follows:
f

2D c
1

.
T
r (P0 + 2/r) ln [(zf + 10r)/(z0 + 10r)]

(41)

To go further on with the dependence of the bubbling frequency with both liquid
and ber parameters, we can replace in Eq. (41) the diusion coecient D0 by its

Champagne Fizzics

39

theoretical expression approached through the well-known Stokes-Einstein equation


(D0 kB /6a). By replacing in Eq. (41) each and every parameter by its theoretical expression and each constant by its numerical value, the variation of the bubbling
frequency as a function of the various pertinent parameters involved may be rewritten
as follows (in the MKSA system):
f 2.4 1014

2 [cL kH (P + 2/r)]
.
r (P + 2/r) ln [(zf + 10r)/(z0 + 10r)]

(42)

The boundary layer thickness was indeed indirectly approached in a recent paper
and found to be of order of 1020 m [7]. Finally, let us apply the latter equation to
the standard textbook case ber displayed in Fig. 24 and modelled in Fig. 25 (i.e.,
r 5 m, z0 20 m and zf 100 m). Equation (42) may therefore be rewritten as
follows, by replacing the bers parameters r, z0 , zf and by their numerical value:
f 5.2 108

2 [cL kH (P + 0.2)]
.
(P + 0.2)

(43)

In the latter expression, cL is expressed in g/L, kH in g/L/atm, P in atm, and in


kg/m/s, to t the standards used in enology.
We will discuss the relative inuence of the following parameters on the average
bubbling frequency: (i) the concentration cL of CO2 -dissolved molecules, (ii) the liquid
temperature , and (iii) the ambient pressure P .
(i) Following Eq. (43), every other parameter being constant, the dependence of
the theoretical average bubbling frequency f with the CO2 -dissolved concentration
cL is in the form f = acL b. By use of a high-speed video camera tted with a
microscope objective, a few cellulose bers acting as bubble nucleation sites on the
wall of a glass poured with champagne were followed with time during the whole gas
discharging process (which may last up to several hours). This method is developed in
minute details in Liger-Belair et al. [6]. The dependence of the experimental bubbling
frequency fexp with cL was found to follow a linear-like cL dependence, as expected
from the model developed above. Therefore, the frequency of bubble formation from
a given nucleation site is found to progressively decrease with time, because the concentration cL of CO2 -dissolved molecules progressively decreases as CO2 continuously
desorbs from the champagne matrix. Furthermore, it is worth noting that the bubbling
frequency of a given nucleation site vanishes (i.e., the bubble release ceases, f 0
bubble/s), although the CO2 -dissolved concentration cL remains higher than a critical
value, as shown in reference [6]. Actually, following both Laplaces and Henrys law,
the curvature r of the CO2 pocket trapped inside the bers lumen induces in the close
vicinity of the trapped CO2 pocket a concentration cB of CO2 -dissolved molecules of
order of kH (P0 + 2/r). Consequently, as soon as the concentration of CO2 -dissolved
in the liquid bulk reaches a critical value cL = cB kH (P0 + 2/r), the diusion
toward the gas pocket ceases and the given nucleation site stops releasing bubbles
(simply because c, the driving force of diusion, vanishes as cL cL ). Let us apply
the latter condition to the characteristic radius of a cellulose ber (r 5 m). At
10 C, the critical concentration cL below which bubble release becomes impossible is
therefore:



cL kH (P0 + 2/r) 2.07 105 105 + 2 5 102 5 106 2.5g/L. (44)
(ii) The dependence of the bubbling frequency with the liquid temperature is much
more dicult to test experimentally in real consuming conditions. Actually, we needed
time to decrease or increase the liquid temperature, and we found no satisfying possibility of modifying the liquid temperature without signicantly loosing CO2 -dissolved

40

The European Physical Journal Special Topics


20

f (bubbles/s)

18

16

14

12

10
278

280

282

284

286

288

( K)
Fig. 28. Theoretical dependence of the bubbling frequency f with the temperature , as
expected from the model displayed in Eq. (43), in the range of usual champagne-tasting
temperatures (from 5 to 15 C), and for the textbook case ber displayed in Fig. 25.

molecules which continuously desorb from the supersaturated liquid matrix due to
diusion through the liquid surface and due to bubbling from the numerous nucleation sites found in the ute. We will nevertheless discuss the theoretical inuence of
the liquid temperature by retrieving Eq. (43). In Eq. (43), the temperature directly
appears as 2 , but the Henrys law constant kH , as well as the champagne dynamic viscosity are strongly temperature-dependent [3]. For example, by increasing the liquid
temperature by 10 K (let us say from 278 to 288 K, which is approximately the range
of champagne tasting temperatures), the theoretical bubbling frequency increases by
about 50%. For the ber displayed in Fig. 24 (r 5 m, and z0 20 mzf 100 m)
and with cL 12 g/L, the theoretical temperature dependence of the bubbling
frequency is displayed in Fig. 28.
(iii) Increasing or decreasing the ambient pressure P also signicantly modies
the corresponding average bubbling frequency f . For the ber displayed in Fig. 24
(r 5 m, z0 20 m and zf 100 m) and with cL 12 g/L, the theoretical
pressure dependence of the bubbling frequency is displayed in Fig. 29. Reducing the
ambient pressure to only 0,3 atm (on the top of Mount Everest for example) would
increase the average bubbling frequency by a factor of almost 3. This is basically the
same phenomenon which is responsible for gas embolism in divers who have breathed
high-pressure air under water if they resurface too quickly. Inversely, increasing the
ambient pressure to 2 atm decreases the average bubbling frequency by a factor of
about 2 compared to that at sea level.

5.7 Evidence for bubbling instabilities


The regular and clockwork release of bubbles from a cellulose ber is indeed the
most common and usual way of blowing bubbles, but cellulose bers were recently

Champagne Fizzics

41

f (bubbles/s)

1000

100

10

1
0

P (atm)
Fig. 29. Theoretical dependence of the bubbling frequency f with the ambient pressure
P (at 20 C), as expected from the model displayed in Eq. (43) for the textbook case ber
displayed in Fig. 25.

found to experience other various and sometimes very complex rhythmical bubbling
regimes [59,60]. After pouring champagne into a ute, thorough examination (even
by the naked eye) of the bubble trains rising toward the liquid surface recently
revealed a curious and quite unexpected phenomenon. As time proceeds, during the
gas discharging process from the liquid matrix, some of the bubble trains showed
abrupt transitions during the repetitive and rhythmical production of bubbles.
Visually speaking, the macroscopic pertinent parameter which is characteristic from
the successive bubbling regimes is the interbubble distance between the successive
bubbles of a given bubble train. In Fig. 30, micrographs of a bubble train in its
successive rhythmical bubbling regimes while degassing are displayed. The duration
of a given bubbling regime may vary from a few seconds to several minutes. In
frame (a), bubbles are seen to be generated from a period-2 bubbling regime which
is characterized by the fact that two successive bubbles rise by pairs. Then, the
bubbling regime suddenly changes, and a multiperiodic bubbling regime arises which
is displayed in frame (b). Later, in frame (c), clockwork bubbling in period-1 occurs
where the distance between two successive bubbles increases monotonically as they
rise, and so on. This nucleation site experienced other various bubbling regimes during its life, until it nally ended in a clockwork period-1 bubbling regime presented in
frame (g).
Such a curious and unexpected observation raises the following question: what
is/are the mechanism(s) responsible for the transitions between the dierent bubbling
regimes? To better identify the ne mechanisms behind this rhythmical production of
bubbles from a few nucleation sites, some of them experiencing bubbling transitions
were lmed in situ by use of a high-speed digital video camera. Two time sequences
are displayed in Fig. 31, and Fig. 32, where bubbles are blown in a period-2 and
in a very erratic way, respectively. The lumen of the cellulose bre displayed in
Fig. 31 presents only one gas pocket, whereas the bres lumen displayed in Fig. 32

42

The European Physical Journal Special Topics

Arrow of time

(a)
Period-2

(b)

(c)
Period-1

(d)
Period-2

(e)
Period-1

(f)

(g)
Period-1

Fig. 30. Time sequence (from left to right) showing a bubble nucleation site at the bottom
of a ute poured with champagne blowing bubbles through dierent and well-established
bubbling regimes (bar = 1 mm). (Photographs by Gerard Liger-Belair.)

clearly shows two gas pockets periodically touching and connecting themselves
through a tiny gas bridge (see frames 3 and 4 of Fig. 32). The micrometric gas
bridge connecting the two gas pockets and disturbing the overall production of
bubbles is enlarged in Fig. 33. This tiny gas bridge is a likely source of bubbling
instabilities.
Recently, a model was built which takes into account the coupling between the
bubbling frequency and the frequency of the single gas pocket which oscillates while
trapped inside the bres lumen (as in Fig. 31, for example). The previously published data showed a general rule concerning bubbling instabilities arising from some

Champagne Fizzics

(1)

(2)

43

(3)

Gas pocket trapped


within the fibers
lumen
Fig. 31. Close-up time sequence illustrating a tiny cellulose bre acting as a bubble nucleation site in its period-2 bubbling regime (i.e., bubbles are blown by pairs); the time
interval between two successive frames is 40 ms (bar = 50 m). (Photographs by Gerard
Liger-Belair).

Fig. 32. Two gas pockets are interacting in the lumen of this cellulose bre, thus disturbing the periodicity of the bubbling regime; the black arrows points the various gas pockets
interacting; the time interval between two successive frames is 10 ms (bar = 100 m). (Photographs by Gerard Liger-Belair.)

bres presenting just one trapped gas pocket. In this previous paper, the successive rhythmical bubbling regimes followed the so-called period-adding scenario [59].
Nevertheless, this previously published scenario does not t the various ways of blowing bubbles from more complex cellulose bres able to entrap numerous gas pockets
as shown in Fig. 32. Numerous bres, such as those shown in the present paper,
presented a sequence of various bubbling instabilities which is not reproduced by
our previous model. A huge collection of successive rhythmical bubbling regimes has
already been observed, and the highest recorded periodicity was observed for a ber
presenting a period-12 bubbling regime [60]. At the moment, we could not nd any
general rule with bres presenting numerous gas pockets interacting together, but the
close up observation and the discovery of the multiple gas pockets interacting together

44

The European Physical Journal Special Topics

Fig. 33. Detail of the cellulose bre displayed in Fig. 32, which clearly shows the establishment of a micrometric gas bridge between the two gas pockets trapped inside the bres
lumen (bar = 10 m). (Photograph by Gerard Liger-Belair).

is considered as a step toward a deeper understanding of the successive rhythmical


bubbling regimes arising from complex bres. The huge diversity of our observations,
in terms of the various successive bubbling regimes seems to be directly linked with
the natural variability of cellulose bres (in terms of size, lumen diameter, inner
wall properties. . . ).

5.8 A word about iers


Some of the particles acting as bubble nucleation sites (most of them including cellulose bers) may detach from the glass wall to nally get completely immersed into the
champagne bulk. Particles detached from the glass wall are nevertheless still active
(in terms of bubbling capacity) provided that a gas pocket with a radius of curvature larger than the critical radius has been trapped inside them. Those particles
immerged in the champagne bulk produce easily-recognizable bubble trains, which
seem to dance erratically inside the glass during champagne tasting. Those particles
in suspension in champagne glasses were called fliers (due to their often complex and
circling trajectories in the champagne bulk). Fliers are indeed a signicant source of
bubbles in glasses poured with champagne. The photograph of a typical ute poured
with champagne is displayed in Fig. 34, where some iers are recognizable. Fliers
undoubtedly catch the eyes of champagne-tasters, who also often are ne observers.
The dynamics of iers was recently investigated by use of long exposure time
photography and laser tomography techniques [61]. By use of long exposure time
photography, the trajectories of bubbles released by iers were found to leave very
elegant and characteristic prints as they crossed a section of champagne illuminated

Champagne Fizzics

(a)

45

(b)

Fig. 34. Photograph of a typical ute poured with champagne (a), and close-up on particles
acting as bubble nucleation sites freely oating in the bulk of the ute (called iers), thus
creating charming bubble trains in motion in the champagne bulk (b). (Photograph by Alain
Cornu Collection CIVC.)

with a 1 mm-thick laser sheet (see the for example the various examples of iers as
revealed through laser tomography techniques in Fig. 35). Because the ier immersed
in the champagne bulk is constantly moving, trajectories of bubbles released during
the 1s-exposure-time photography do not superimpose on each other. Therefore, the
print left by a ier which crosses the laser sheet during the exposure time of the
digital camera is a typical multiple laments structure, each lament materializing
the trajectory of a single rising bubble.

6 The bubble rise


6.1 Bubble velocity and bubble growth rate
After being born on micrometric gas pockets trapped inside impurities of the glass
wall, bubbles rise toward the liquid surface due to their own buoyancy. While rising,
bubbles continue to grow in size by continuously absorbing carbon dioxide molecules dissolved into the liquid matrix, as it is clearly illustrated on the photograph
displayed in Fig. 36. Growing bubbles thus continuously accelerate along their way
through the champagne. This continuous acceleration is also betrayed in Fig. 36, by

46

The European Physical Journal Special Topics

(a)

(d)

(b)

(e)

(c)

(f)

Fig. 35. Collection of various prints left by the bubbles released from various iers along
their 1s-way through the laser sheet which crosses the ute in its plane of symmetry; the
white arrow in frame (a) points toward the direction of move of the ier (bar = 5 mm).

Champagne Fizzics

47

Fig. 36. A characteristic bubble train promoted by the repetitive bubble formation process
from a single cellulose bre; bubbles are clearly seen growing during their way up (bar =
1 mm). (Photograph by Gerard Liger-Belair.)

the continuously increasing spacing e between the successive bubbles of a given bubble train.
High-speed photography and strobe lighting was used to monitor the dynamics of
ascending bubbles [13,5,6,62]. In Fig. 37, for rising bubbles of various bubble trains,
the bubble velocity U was plotted as a function of its radius R. In Fig. 37, U (R) is
2
also superimposed with the well-known Stokes velocity, UST = 2g
9 R , where and
are respectively the liquid density and viscosity, and g is the acceleration due to
gravity.

48

The European Physical Journal Special Topics

10,0

Re 100

UR
U (cm/s)

Re 10
1,0

Re 1
Re 0.1

500

400

300

200

70
80
90
100

50

60

40

30

0,1
R (m)

Fig. 37. Velocity of ascending champagne bubbles plotted as a function of their radius, U (R);
experimental data () are compared with the well-known Stokes velocity (black-dashed line)
which constitutes a guide for the eyes; the approximate Re experienced by ascending bubbles
is also reported close to the corresponding data.

The Reynolds numbers (Re = 2RU/) experienced by ascending champagne


bubbles are also reported in Fig. 37. To a quite good approximation, it was found
that the bubble velocity close to the glass wall can be modeled by a modication of
the numerical pre-factor in Stokes law:
U (R) =

2g 2
R (with 0.6 0.8).
9

(45)

More details about the determination of Eq. (45) can be found in Liger-Belair et al.
[62]. In terms of power-law, Fig. 37 therefore suggests a quadratic dependence with
the bubble radius for the velocity of rise, as it was also reported in the article by
Maxworthy et al. who closely examined the velocity of air bubbles rising in clean
viscous liquids in a wide range of Reynolds numbers [63].
By combining high-speed photography and strobe lighting, it was also found that
the bubble radius R of ascending champagne bubbles increases at a constant rate
k = dR/dt, as they rise toward the liquid surface (see Fig. 38) [13,5,6,62]. Thus:
R (t) = R0 + kt

(46)

where R0 is the bubble radius as it detaches from the nucleation site. R0 is of the
same order of magnitude than the radius of the mouth of the cellulose ber which
acts as a bubble nucleation site, i.e., around 10 to 20 m [3,7,8,47].
The rst experimental observations about the growth and rise of bubbles in carbonated beverages were conducted in the early 1990s, by Shafer and Zare, with bubbles rising and growing in a glass of beer [64]. Shafer and Zare also proved that
bubbles diameter linearly increases with time as bubbles rise toward the liquid surface.

Champagne Fizzics

49

300

R(t ) = R0 + kt

250

R (m)

200

150

100

50

0
0,0

0,2

0,4

0,6

0,8

1,0

t (s)
Fig. 38. Bubble radius increase versus time for a bubble rising toward the liquid surface;
two typical bubble trains at dierent steps of gas discharging are considered.

Three minutes after pouring champagne into a ute, bubble growth rates k of
order of 350 m/s (at 20 C) [3,62]. Experiments were also performed with bubbles
rising in beer glasses. In beer, three minutes after pouring, bubble growth rates were
found to lay around 100150 m/s, i.e., about three times less than those in found in
champagne and other sparkling wines [3,62].
By use of mass transfer equations, the growth rate k of bubbles rising in champagne and beer (at low to moderate Reynolds numbers) was indeed modelled and
linked with some physicochemical properties of the liquid medium [6]. The general
equation concerning the mass transfer of molecules from the bulk of a supersaturated
liquid to a bubble surface with time is:
dN
= KAc
dt

(47)

where N is the number of transferred CO2 moles, K the mass transfer coecient, A
the bubble area, and c (the driving force responsible for the diusion of dissolved
CO2 molecules into the rising bubble) is the dierence in dissolved CO2 concentrations
between the liquid bulk and the close vicinity of the bubble surface in equilibrium
with the gaseous CO2 into the rising bubble (see Fig. 39).
Assuming the gas into the rising bubble as ideal (PB V = N ), the number of
CO2 moles transferred into the bubble is connected with the variation of its radius
R with time as follows:
dN
PB dV
PB dR
=
=
A
(48)
dt
 dt
 dt
where V is the bubble volume, and PB is the pressure inside the rising bubble assumed
to be equal to the atmospheric pressure P0 , because both the hydrostatic pressure
(gH < 102 P0 ) and the Laplace pressure (2/R < 101 P0 ) inside the rising bubble
are negligible.

50

The European Physical Journal Special Topics

champagne
bubble

gas/liquid interface

liquid bulk

cL
CO2 bubble

c = c L c B

Boundary layer where a gradient


of dissolved CO2 exists

c B = k H PB k H P0

Adsorption layer in equilibrium with


the gaseous CO2 into the bubble and
where Henrys law locally applies

Fig. 39. Carbon dioxide concentrations in the close vicinity of the CO2 bubble surface.

By combining Eqs (47) and (48), the growth rate of ascending bubbles transforms
as follows:

dR
=
Kc.
(49)
dt
P0
Generally speaking, heat and mass transfers are functions of two dimensionless
numbers, the Sherwood and Peclet numbers, respectively Sh = 2KR/D0 and
P e = 2RU/D0 , U being the velocity of ascending bubbles, and D0 the diusion
coecient of CO2 molecules in champagne. In case of small and large P e, asymptotic
solutions have been derived in the literature. Most of them are listed in the book
by Sherwood et al. [65]. During ascent, champagne bubbles cover a range of high P e
between approximately 102 and 105 . At large P e, Sh becomes proportional to P e1/3 ,
with a numerical pre-factor very close to unity. Therefore,
1/3

1/3
KR
21/3 RU
2/3 U

K 0.63D0
.
(50)
D0
2
D0
R2/3
Combining Eq. (49) and (50) leads to:
 2/3 U 1/3
dR
= 0.63 D0
c.
dt
P0
R2/3

(51)

Champagne Fizzics

51

By replacing U in Eq. (51) by the empirical rising velocity of ascending champagne


bubbles given in Eq. (45), the growth rate of ascending bubble may be rewritten as:
dR
 2/3
k=
0.63 D0
dt
P0

2g
9

1/3
c.

(52)

Let us test the applicability of Eq. (52) in case of rising and expanding champagne
bubbles, at 20 C. By using known values of and in champagne, = 0, 7 [3,62],
D0 1.4 109 m2 /s, as measured by Nuclear Magnetic Resonance [56], and the
dierence in CO2 concentrations between the liquid bulk and the close vicinity of the
bubble surface, c 8 g/L 180 mol/m3 , one nds,

1/3

2 0.7 103 9.8
8.31 293 
9 2/3
k 0.63
1.4 10

180 400 m/s


105
9 1.5 103
(53)
which is in very good accordance with the order of magnitude of experimentally observed growth rates [3,62]. This value nevertheless slightly overestimates experimentally observed growth rates probably because, three minutes after pouring, champagne
already lost a signicant part of its dissolved CO2 content, thus decreasing c below
its initial value of 180 mol/m3 (as measured immediately after pouring champagne
into a ute). Quite recently also, Zhang and Xu proposed a model for the growth rate
of rising bubbles in both champagne and beer [66].

6.2 Average bubble size


Because champagne and other sparkling wine tasters are often concerned with the
size of bubbles formed in the wine (a proverb says that the smaller the bubbles,
the better the wine), much attention was paid recently to model the average size of
ascending bubbles. Actually, the nal average size of ascending bubbles is the result
of a combination between their growth rate and their ascending velocity.
The consumer pays indeed much attention to the size of bubbles as they reached
the liquid surface. This nal average bubble size depends on several parameters, such
as: (i) the growth rate k during ascent, (ii) the bubble velocity U , and (iii) the distance
traveled h by the bubble from its nucleation site. Actually, because the time of ascent
of a champagne bubble is very short ( 1 s), the bubble growth rate k during ascent
is constant during the bubble rise. Therefore, the rate of increase of the bubble radius
with h becomes,
dR
dR dh
dR
dR
k
k=
=
=
U, i.e.,
= .
(54)
dt
dh dt
dh
dh
U
Combining Eqs. (54) and (46), and performing the integration leads to,

1/3
27k
3
R = R0 +
h
.
2g

(55)

It is worth noting that, by neglecting R0 which is only of order of 10 to 20 m, the


latter expression transforms as,

R3

1/3

kh
.
2g

(56)

52

The European Physical Journal Special Topics


300

250

R (m)

200

150

100

50

0,06

0,05

0,04

0,03

0,02

0,01

0,00

k h (cm2/s)
Fig. 40. For various bubbles of various bubble trains, the radius of ascending bubbles is
plotted versus the product of the growth rate k (expressed in cm/s) by the traveled distance h
(expressed in cm); experimental data () are compared with the theoretical relation expressed
by Eq. (56) (black-dashed line).

In Fig. 40, for bubbles of various bubble trains, and all along the bubble rise, experimental bubbles radii were plotted versus the product kh. Experimental data
were also compared with the latter theoretical expression. The very good agreement
between Eq. (56) and our experimental data corroborates the validity of the preceding scaling law which links the bubble size with several parameters of the liquid
phase.
By replacing in Eq. (56), the theoretical relation (52) for the bubble growth rate
k, the following dependence of the bubble radius with some of the liquid parameters
was derived:

R 1.24

9
2g

2/9 


P0

1/3
2/9

D0

(cL cB )

1/3

h1/3 .

(57)

To go further on with the dependence of bubbles radii with some few parameters, we can also replace in (57) the diusion coecient D0 by its theoretical expression approached through the well-known theoretical Stokes-Einstein equation
(D0 kB /6a), the following relationship expressed in the MKSA system was
thus obtained:

R (h, , . . .) 2.5

3kB
4a

2/9 

1
g

2/9 

1
P0

1/3
1/3

5/9 (cL cB )

h1/3 .

(58)

It is worth noting that the dependence of the bubble size with the liquid viscosity
vanishes. Finally, by replacing in Eq. (58), kB , , and a by their known numerical

Champagne Fizzics

53

values, and by developing cB as kH P0 , one obtains:


R 2.7 103 5/9

1
g

2/9 

cL kH P0
P0

1/3
h1/3 .

(59)

Otherwise, because the liquid density does not signicantly vary from one champagne to another (and even from on carbonated beverage to another), we will discuss
and emphasize the inuence of the following parameters on the bubble size: (i) the
traveled distance h, (ii) the liquid temperature , (iii) the gravity acceleration g, (iv)
the ambient pressure P0 , and (v) the carbon dioxide content cL .
(i) The longer the traveled distance h, the larger the bubble size. This dependence
of the bubble size with its traveled distance through the liquid means that, during
champagne tasting, the average bubble size at the champagne surface varies from on
glass to another. In a narrow ute, for example, the level of champagne poured is
about three times higher than that in a typical coupe (with a shallower bowl and a
much wider aperture). Therefore, the average bubbles diameters in the ute will be
larger than those in the coupe by a factor of about Rflute /Rcoupe 31/3 1.45 (i.e.,
bubbles about three times larger in volume!), as seen in the photograph displayed in
Fig. 41, which compares the average size of bubbles after pouring, whether champagne
is served into a ute or into a coupe [25].
(ii) In Eq. (59), the temperature appears directly as 5/9 , but we should not forget
that the Henrys law constant kH is also strongly temperature-dependent (see Table 1)
and conveniently expressed by the Vant Ho-like Eq. (3). The temperature being
expressed in K, the temperature dependence of the bubble size is nevertheless quite
weak. Increasing the liquid temperature by 10 K (lets say from 278 to 288 K, which
is approximately the range of champagne tasting temperature) makes bubbles grow
from about only 56% in diameter.
(iii) The gravity acceleration which is the driving force behind the bubble rise
(through buoyancy) plays also a quite important role in the nal bubbles size. This
could indeed be easily evidenced during a parabolic ight where the acceleration
changes from micro-gravity (close to zero-g) to macro-gravity (up to 1.8 g). On the
Moon for example, where the gravity is about 1/6 the gravity on Earth, the average
bubbles size would increase by a factor of about gMoon /gEarth 62/9 1.49 (i.e.,
bubbles almost 50% larger in diameter and therefore more than 3 times larger in
volume).
(iv) The pressure inside the rising bubble is equivalent to the ambient pressure
P0 (for the reasons detailed in the latter paragraph). Usually, at the sea level, this
pressure is equivalent to 1 atm (or 105 N/m2 ). Reducing the atmospheric pressure to
only 0,3 atm (on the top of Mount Everest, for example) would increase the average
bubble diameter by about 55% (and therefore by a factor of almost 4 in volume).
(v) The carbon dioxide content of the liquid medium cL also inuences the nal
average bubbles size. This is the main reason why bubbles in beer are signicantly
smaller than bubbles in champagne and other sparkling wines. Actually, the carbon
dioxide content in beers may classically vary from about 4 g/L to 7 g/L, whereas the
carbon dioxide content in champagne and other sparkling wines may rather vary from
10 g/L up to 12 g/L (i.e., cL is approximately 2 times higher in champagne than in
beer). Reducing cL by a factor 2 in Eq. (59) would decrease the theoretical average
bubble size by about 40% (thus leading to bubbles almost 5 times smaller in volume).
The two photographs displayed in Fig. 42 illustrate the signicant dierence in bubble
size between a standard commercial champagne and a standard commercial beer,
both showing very typical bubbling behaviour. Moreover, after pouring champagne
into a ute, due to bubbling and diusion through the surface of champagne, CO2
molecules progressively escape from the liquid medium. Subsequently, the dissolved

54

The European Physical Journal Special Topics

(a)

(b)

Fig. 41. Bubbles size distribution at the free surface of champagne, 30 seconds after pouring,
whether champagne is served in the coupe (a) or in the ute (b) etched on its bottom.
Because of a drastic glass washing protocol, bubbles were generated only from nucleation
sites of the ring-shaped etchings (lying respectively 2.9 cm and 7.4 cm below the free surface
area, whether champagne is served in the coupe or in the ute). (Photographs by Gerard
Liger-Belair.)

carbon dioxide content cL in the liquid medium progressively decreases. Therefore,


as time proceeds during champagne tasting, the average bubbles size at the liquid
surface progressively decreases, as can be clearly seen in the sequence displayed in
Fig. 43.
6.3 Inuence of the surface-active macromolecules of champagne on the drag
coecient of ascending bubbles
As probably rst noticed by Bond [67], hydrodynamics of rising bubbles strongly
depends on the presence of surface-active substances in the liquid medium. During
ascent, surface-active materials progressively accumulate at the rear part of a rising

Champagne Fizzics

(a)

55

(b)

Fig. 42. Three minute after pouring, bubbles rising in a glass of beer (a) show diameters
much lower than those of bubbles rising in a ute poured with champagne (b) (bar = 1 mm);
the very signicant dierence between the bubble size in champagne and beer is mainly due to
amounts of dissolved-CO2 about two times higher in champagne than in beer. (Photographs
by Gerard Liger-Belair.)

bubble, thus increasing the immobile area of the bubble surface. It ensues a gradient
of surface tension between the front and the rear part of the bubble which induces a
modication of the hydrodynamic boundary conditions on the bubble via the onset
of the so-called Marangoni eect (see Fig. 44).
The gradient of surface tension around the rising bubble induces a viscous shear
stress which reduces its interfacial mobility. Viscous dissipation is therefore enhanced
which leads to a lower rising velocity. This model based on the rigidication of the
rear part of the bubble is known as the stagnant cap model. Numerous experimental,
numerical and theoretical researches on bubble motion have already conrmed this
phenomenon [6877]. The graph displayed in Fig. 45 illustrates this Marangoni eect
which progressively lowers the velocity of a millimetric bubble rising in water diluted
with proteins.
Hydrodynamically speaking, a rising bubble is rigidied by surfactants and
runs into more resistance than a rising bubble presenting a more exible interface
free from surface-active materials. Therefore, the drag coecient CD experienced
by a bubble of a xed radius rising in a surfactant solution progressively enhances
since the surface of the stagnant cap progressively increases. Bubbles of xed radii

56

The European Physical Journal Special Topics

(a)

(b)

(c)

(d)

Fig. 43. Time sequence showing successive top views of a ute poured with champagne
and followed as time proceeds; (a) immediately after pouring, (b) 3 mins after pouring, (c)
10 mins after pouring, and (d) 25 mins after pouring; it clearly appears that the average
bubbles size decreases as time proceeds, as well as the average number of oating bubbles.
(Photographs by Gerard Liger-Belair.)

ascending in surfactant solutions therefore experience a transient regime, where the


bubble behavior progressively changes from that of a uid to that of a rigid sphere
(when the bubble surface is completely rigidied by surfactants). Actually, there are
surface active substances in champagne and sparkling wines likely to be adsorbed at
the bubble surface, such as proteins and glycoproteins [7882]. Such materials will
certainly modify the bubble surface state during ascent, and therefore the drag exerted on bubbles in comparison with the drag exerted on bubbles rising in a liquid
free from surface-active compounds. Finally, the drag coecient experienced by a
single ascending champagne bubble may be used to reveal whether or not the bubble
interface is contaminated with surface-active substances.

Champagne Fizzics

57

Streamlines

Bubble

Fig. 44. The liquid streamlines sweep surfactant molecules around towards the bottom of
a rising bubble, thus creating a gradient of surface tension around the bubble interface;
In turn, counter-currents raise along the bubble interface and lower its velocity of rise; this
model based on the rigidication of the rear part of the bubble is known as the stagnantcap model.

6.4 The bubble drag coecient during ascent: An indirect determination of its
surface state
It is quite simple to measure the experimental drag coecient CD experienced by a
rising bubble. The simultaneous measurements of the bubble velocity and of its radius
allow a simple experimental determination of the drag coecient experienced by a
rising bubble through the expression,
CD =

8gR
.
3U 2

(60)

Details about the determination of Eq. (60) can be found in Liger-Belair [3]. During
the last decades, many empirical or semi-empirical equations have been proposed
to approximate CD for bubbles in free rise. Some of the most popular are listed
in the book by Clift et al. [83]. Our experimental measurements of CD (Re) are to
be compared with the drag coecients experienced by a bubble in the two limiting
boundary regimes in terms of interfacial mobility (on the one hand, a rising bubble
free from surface-active substances presenting a full interfacial mobility, and on the
other hand, a contaminated bubble with a zero velocity boundary condition for the
external uid and that hydrodynamically behaves as a rigid sphere of same density
and same diameter). We therefore retrieved three empirical drag coecients CD (Re)
that are listed in Table 4.

58

The European Physical Journal Special Topics

U
30
Fluid sphere
limit

Bubble
velocity
(cm/s)

15
the bubble velocity
decreases as soon as the
bubble interface traps
proteins

Rigid sphere
limit

10 cm

1 cm

Distance traveled by the bubble (cm)


Fig. 45. Velocity versus traveled distance of a millimetric bubble released in ultra-pure
water (black dashed-line) and in water added with only 10 mg/L of BSA (grey dotted-line);
redrawn from the article by Ybert and di Meglio [76].
Table 4. Relationships, derived for both rigid and uid sphere conditions, between the drag
coecient and the Reynolds number experienced by ascending bubbles. The subscripts RS
and FS refer to the rigid and uid sphere boundary conditions, respectively.
Authors

Range of
Validity

Relationship
for CD (Re)

Bubble Surface
State

Schiller and Naumann


24
(see the book by Clift (Re < 800)
CRS = Re
(1 + 0.15Re0.687 ) rigid
et al. [83])

16
Magnaudet et al. [84]
(Re < 50)
CF S = Re
(1 + 0.15 Re)
uid
Maxworthy et al. [63]
(1 < Re < 800) CF S = 11.1Re0.74
uid

In order to indirectly access the bubble surface state during the rise, the normalized

drag coecient CD
dened as follows was used,

CD
=

(CD CF S )
(CRS CF S )

(61)

where CRS and CF S are well-known empirical drag coecients derived for rigid and
uid sphere conditions, in their respective range of Reynolds numbers (see Table 4).
Therefore, bubbles with a fully mobile interface, behaving hydrodynamically as

uid spheres, will exhibit values of CD


close to zero, whereas polluted bubbles, be
having hydrodynamically as rigid spheres, will have values of CD
close to one. In Fig.
46, for bubbles of various bubble trains, the normalized drag coecient experienced
by bubbles was plotted as a function of their Reynolds number.
At low Re, just after the bubble detachment from a nucleation site stuck on the

glass wall, CD
is signicantly higher than the rigid sphere limit. Actually, during

Champagne Fizzics

59

( ) rigid sphere limit

Figure 47

(- - - -) fluid sphere limit


1

CD*

0
1

10

Re

100
Increasing h

CD
(Re)

Fig. 46. Normalised drag coecient


experienced by expanding champagne bubbles
during ascent (), compared with the normalised drag coecient that a bubble of xed
radius rising in a surfactant solution would experience (grey dashed-line); both the uid
(- - -) and rigid ( ) sphere limits are also presented.

the very rst moments after detachment from the nucleation site, wall eects are
certainly not negligible and probably modify the liquid ow around the rising bubble
resulting in a signicant increase of the drag coecient experienced by the bubble.

Then, from Re 34 to Re 10, CD


quickly decreases from the rigid sphere limit
to an intermediate value close to the uid sphere limit, thus suggesting that the
champagne bubble interface progressively increases its mobility. Then, from Re

10 to Re 100, CD
remains quite low and close to the uid sphere limit, which
suggests that champagne bubbles have reached a quasi-stationary stage in terms
of interfacial mobility. Strictly speaking, the level of the plateau seems to slightly
decrease during this second quasi-stationary stage, which means that the bubble
interface could continue to very progressively increase its mobility. In Fig. 46, the
line which progressively raises from the uid to the rigid sphere limit, qualitatively
symbolizes the transient regime that a bubble of a xed radius rising in a surfactant
solution would experience.

In the present situation, the main features of experimental CD


(Re) in Fig. 46
strongly suggest that, contrary to a bubble with a xed radius, the expanding champagne bubble interface experiences a transient regime during its way up, where the
bubble interface progressively changes its mobility from that of a rigid to that of
a uid sphere. We are therefore logically tempted to attribute this gain of interfacial mobility during ascent to the bubble growth which continuously oers newly
created surface to the adsorbed surface-active materials, thus diluting surface-active
compounds on ascending bubbles. The aim of the following paragraph is to discuss
this competition between the adsorption of surface-active compounds on ascending
bubbles and the growth of bubbles.

60

The European Physical Journal Special Topics

6.5 The champagne bubble surface state ruled by a competition between


surfactant adsorption and bubble growth
During ascent, surface-active substances accumulate at the bubble interface and contribute to its rigidifying, by increasing the amount of adsorbed materials. However, at
the same time, the bubble continuously grows as a result of supersaturating. Therefore, the area of the bubble interface increases, thus diluting the amount of the adsorbed materials. Bubbles experience a competition between two opposing eects.
The parameter which controls the bubble surface state is its surface concentration of
contaminants, denoted = M/A, which is the ratio of the mass M of contaminants
adsorbed to the bubble area A. The variation of the bubble surface concentration
with time, on expanding bubbles, is governed by the following equation [85]:




d
1 dM
M dA
=

.
(62)
dt
A
dt
A2 dt M

 A



Rate at which surfactants
adsorb on the rising bubble

Rate of dilatation
of the bubble area

Therefore, the variation with time of the surface concentration of contaminants on


ascending bubbles, depends on the ratio of the two terms in the right hand of Eq. (62).
If the rate of dilatation of the bubble area exceeds the rate at which surface-active
materials is transported to the bubble interface, the surface concentration of surfactants decreases and the bubble progressively cleans its surface. On the contrary, if the
surfactant transport exceeds the dilution eect due to the bubble growth, the bubble
interface progressively gets invaded by surfactants and becomes rigid. Finally, the
variation with time of the surface concentration of contaminants on ascending bubbles is ruled by the ratio of the characteristic time scale of bubble expansion to the
characteristic time scale of surfactant adsorption, respectively:
1
exp
=

2R
1 dA
=
,
A dt
R

1
and ads
=

1 dM
.
M dt

(63)

After the rst non-stationary regime where CD


> 1, it appears clearly from Fig. 46
that the expanding champagne bubbles experience a transient regime during ascent,
where the bubble behavior progressively changes from that of a rigid sphere to that
of a uid sphere. As a result, it seems that the bubble growth quickly overcomes the
adsorption rate of surface-active compounds on the ascending bubble. In a way, by
continuously growing during ascent, champagne bubbles would clean up themselves
[85].
To complete this set of experiments on expanding champagne bubble dynamics,
time scales of bubble expansion and surfactant adsorption are to be compared. The
time scale of bubble expansion, exp , is easily experimentally accessible by measuring
R and R all along the bubble rise, for each given bubble train. As concerns the
adsorption kinetics of surfactants, we retrieved theoretical calculations of diusive
and convective transport of solutes toward a bubble in free rise. At low Re, since
ascending champagne bubbles behave as rigid spheres, we retrieved a calculation
conducted by Levich for rigid sphere boundary conditions [69],



2/3
DS
dM 
2/3
AC (P e)
AC
(Re < 1)
dh RS
2RU

(64)

where M , h, A, C, DS , R and U are respectively, the mass of adsorbed surfactants


along the bubble rise, the traveled distance, the bubble area, the bulk concentration

Champagne Fizzics

61

of surfactants, the bulk diusion coecient of surfactants, the bubble radius and its
velocity of rise.
At higher Re, since ascending champagne bubbles progressively exhibit higher
interfacial mobility, we retrieved a theoretical result derived by Yang et al for uid
sphere boundary conditions [86]:


1/2
dM 
DS
1/2
AC (P e)
AC
(Re 1).
dh F S
2RU

(65)

The time scale for surfactant adsorption can be evaluated as follows,


1
ads
=

1 dM dh
1 dM
1 dM
=
=
U.
M dt
M dh dt
M dh

(66)

As a result, by replacing in Eq. (66) the expression for dM/dh given in eqs. (64) and
(65), the ratio of the two characteristic time scales expresses as follows (at low Re for
rigid sphere conditions, and at high Re for uid sphere conditions, respectively):


exp 
C

1/3
2/3

(DS ) (U R) (Re < 1)

ads RS
R
.
(67) and (68)


C
exp

1/2


(DS U R) (Re 1)

ads F S
R
Finally, let us test the compatibility of relations (67) with our experimental data on
the bubble surface state as reported in Fig. 46.
As far as we know, the pool of surface-active macromolecules in a typical Champagne wine is far from perfectly determined. It was nevertheless recently demonstrated
that most of surface-active macromolecules of champagne wines showed molecular
masses in the range of 104 to 105 (mostly composed of proteins and glycoproteins).
We will retrieve in Eqs. (67) and (68) a diusion coecient characteristic of proteins in water, DS 1011 m2 s1 . At low Re, since Fig. 46 suggests rigid sphere
boundary conditions, we will retrieve in Eq. (67) the average surface concentration,
RS 106 kg m2 , corresponding to the order of magnitude of the equilibrium surface concentration of proteins at the air/water interface [76]. At higher Re, since
Fig. 46 suggests that a rising bubble progressively recovers its interfacial mobility, we
will retrieve in Eq. (68) an average surface concentration, 107 kg m2 , corresponding to a low surface coverage of proteins [76]. Finally, by inserting respectively
at low and high Re in Eqs. (67) and (68), experimental data for U , R and R for each
given bubble train, the bulk concentration of proteins, C 5 103 kg m3 , as measured in our champagne samples, and DS 1011 m2 s1 , the expected ratios of the
two characteristic time scales, (exp /ads )RS and (exp /ads )F S , can be approached.
These ratios are displayed in Fig. 47, at low and high Re, and for both rigid and uid
sphere boundary conditions.
At low Re, (exp /ads )RS is lower than unity, which tends to indicate that the
rate of dilatation of the champagne bubble area exceeds the rate at which surfactants adsorb on the rising bubble. Consequently, the bubble interface progressively
and inexorably recovers its interfacial mobility as suggested in Fig. 46 at low Re. At
higher Re, (exp /ads )F S is nevertheless signicantly higher than unity, which tends
to show that the rate of dilatation of the champagne bubble area is becoming lower
than the rate at which surfactants adsorb on the rising bubble. Consequently, after
a rst detected transient regime where the rising bubble interface changes from that
of a rigid to that of a uid sphere, the interfacial mobility of ascending champagne

62

The European Physical Journal Special Topics

exp /ads

101

100

10-1
10-1

100

101

Re

102
Increasing h

Fig. 47. Ratio of the characteristic time scale of bubble expansion to the characteristic
time scale of surfactant adsorption, plotted as a function of Re all along the bubble rise, for
various bubble trains ascending close to the glass wall.

bubbles should be seen decreasing, which is clearly not observed from the observation

of CD
(Re) in Fig. 46.
In the work by Liger-Belair [85], two hypothesis were proposed in order to interpret this apparent contradiction between our experimental data and theory. First, due
to ow interactions between successive bubbles, the surfactant adsorption dynamics
on bubbles rising in-line could be rather dierent from that of a single bubble, thus
making obsolete the use of Eqs. (64) and (65). Second, above Re 10, the character
istic time scale needed for CD
(Re) to reach within 5 % of its steady state value could
forbid the indirect detection of a possible loss of interfacial mobility experienced by
ascending bubbles.

6.6 Ascending bubbles and ow patterns


During champagne or sparkling wine tasting, consumers certainly pay attention to
the continuous ow of ascending bubbles (often even before smelling and tasting the
wine). Ascending bubbles are indeed visually appealing, but in case of champagne
tasting, their role is suspected to go far beyond the sole aesthetical point of view.
Actually, at the bubble scale, the lower part of a rising bubble is a low pressure area
which literally attracts the uid molecules around. A rising bubble is thus able to drain
some uid along its path toward the free surface. Recently, it was demonstrated that
ascending bubbles act as so many swirling motion generators in champagne glasses.
Together with Guillaume Polidori and Fabien Beaumont, from Reims University, laser
tomography techniques were used in order to visualize the ow patterns induced by
the continuous ow of ascending bubbles in glasses poured with champagne [87]. The

Champagne Fizzics

63

principle of the experiment is to generate a 2 mm-thick laser sheet, built from a multiline argon laser source. The 2 mm-thick laser sheet crosses the plane of symmetry of a
glass poured with champagne. Before pouring champagne into the glass, champagne
R
was seeded with tiny and roughly spherical particles (called Rilsan
particles). Rilsan
particles are polymeric materials which exhibit a high degree of reectivity with regard
to the laser wavelength, and are therefore able to diuse the laser light as they cross
the 2 mm-thick laser sheet. Moreover, Rilsan particles were found to be completely
neutral with regard to bubble formation (this was denitely a crucial condition for
the feasibility of the work). Classical long exposure time photography of the laser
sheet was used in order to follow the motion of Rilsan particles, thus freezing the ow
patterns within the liquid bulk crossed by the laser sheet. Trajectories of convection
currents in champagne glasses were therefore made visible by numerous streaks of
light left by the Rilsan particles along their path through the laser sheet.
The photographs displayed in Fig. 48 show the ow patterns found in two glasses
freshly poured with champagne. In frame (a), the glass was perfectly washed with a
drastic technique using formic acid, so that bubble nucleation was simply forbidden.
In such a perfectly clean glass, free from any nucleation site, champagne looks like a
non eervescent at wine and, 30 s after pouring, the champagne is at rest. In frame
(b), the glass shows natural eervescence, with several nucleation sites giving rise
to several bubble trains in the champagne bulk. A particular bubble train crosses
the laser sheet and promotes ascending convection currents, as betrayed by vertically
oriented streaks of light left by the Rilsan particles along their path through the laser
sheet. By vigorously and continuously mixing the liquid medium all along tasting,
eervescence was indeed found to enhance the release of both gaseous CO2 and vapors
of ethanol above the champagne surface [88].
As shown in Fig. 49, the simultaneous quantication of CO2 and ethanol in the
headspace of a champagne glass was monitored, in real tasting conditions, all along the
rst 15 minutes following pouring, depending on whether the glass shows eervescence
or not [88]. Both CO2 and ethanol were found to be enhanced by the presence of
ascending bubbles, thus conrming experimentally, and for the rst time, a close link
between ascending bubbles and the rate at which dissolved CO2 and ethanol escape
from the liquid phase. Moreover, since a close link between eervescence and ethanol
release was evidenced, we are also tempted to extend our conclusions to the release of
the others numerous volatile organic compounds (VOCs) found in Champagne and
sparkling wines.

6.7 Marangoni and the tears of champagne


As seen earlier in Sect. 6.3, a Marangoni eect lowers the velocity of bubbles ascending in surfactant solutions. Another Marangoni-driven ow is responsible for
a macroscopic observation commonly known as tears of wine, which is related to
the formation and ow of wine drops on the internal walls of a glass [89,90]. This
observation may also easily be done in glasses poured with champagne or sparkling
wine (see Fig. 50).
Far from any myth often linked to this observation, a clear explanation can
be given with the knowledge of some chemical properties of the water-ethanol mixture. Since pure water has a much higher surface tension than ethanol ( 72, and
23 mN/m at 20 C, respectively), a mixture of water and ethanol, such as wine,
has a lower surface tension than that of pure water. When ethanol, which is more
volatile than water is, evaporates from the thinner region of the meniscus on the glass
wall, a surface tension gradient is generated. This surface tension gradient (upwardly
directed) forces the liquid phase to spontaneously climb along the glass wall, thus

64

(a)

The European Physical Journal Special Topics

(b)

(c)

Fig. 48. Laser tomography reveals the ow patterns found inside a ute poured with champagne; within a ute thoroughly washed and therefore showing no eervescence at all, a
few tens of seconds after pouring, champagne is completely at rest (a), whereas in a ute
showing eervescence, ascending bubbles drive characteristic ow patterns and convection
currents (b); detail in (c).

forming a thin lm of wine [89,90]. The lm stops at a certain height (determined


by the balance between surface tension gradient and gravity), where it accumulates
into the form of drops which, when large enough, roll downward under the action of
gravity, as seen in Fig. 50. Because the surface tension gradient is kept by continuous
evaporation of ethanol from the lm, the upwardly directed wine ow self propels
until it nally stops after complete evaporation of the liquid. Such a phenomenon can
be observed in any mixture where the more volatile compound has the lower surface
tension. Tears of wine close to the glass wall are therefore expected to enhance the
evaporation of ethanol compared to evaporation of ethanol from the champagne/air
interface, far from any boundary. Recently, the quantication of gaseous ethanol,
through micro-gas chromatography techniques (GC), was monitored above glasses

Champagne Fizzics

65

(a)

(b)
0,7

20

With effervescence
Without effervescence
15

0,5

% Ethanol

% CO2

With effervescence
Without effervescence

0,6

10

0,4
0,3
0,2

0,1
0

0,0
0

10

12

Time after pouring (min)

14

16

10

12

14

16

Time after pouring (min)

Fig. 49. Gaseous CO2 (a) and ethanol (b) concentrations found, in the headspace above the
champagne surface, through micro-gas chromatography techniques, depending on whether
the ute shows eervescence or not; each datum of each time series is the arithmetic average
of six successive values recorded from six successive pouring; standard deviations correspond
to the root-mean-square deviations of the values provided by the six successive data recordings.

Fig. 50. Evidence for the formation of tears (also called legs) on the wall of a glass poured
with champagne; tears of champagne are expected to enhance evaporation of ethanol close
to the utes wall; this photography was done by using the ombroscopy technique.

66

The European Physical Journal Special Topics


0,7

Edge
Center

0,6

% Ethanol

0,5
0,4
0,3
0,2
0,1
0,0
0

10

12

14

16

Time after pouring (min)


Fig. 51. Ethanol concentrations in the headspace above the champagne surface, all along the
rst 15 min after pouring champagne, depending on the position of the sampling valve; each
datum of each time series is the arithmetic average of six successive values recorded from
six successive pouring; standard deviations correspond to the root-mean-square deviations
of the values provided by the six successive data recordings.

poured with champagne [88]. Concentrations of gaseous ethanol appear signicantly


dierent according to the GC sampling valve position above the ute, as seen in
Fig. 51. It is worth noting that the headspace close to the edge of the ute is more
concentrated with vapors of ethanol than far from any boundary above the center
of the ute. It seems self-consistent with the presence of tears of champagne which
enhance the evaporation of ethanol close to the glass wall.

7 Close-up on bubbles bursting at the liquid surface


7.1 Shape of oating bubbles before collapse
Champagne bubbles, that traveled approximately 10 cm between their nucleation sites
and the liquid surface, reach the free surface with a radius R of order of 0.5 mm. At the
free surface, the shape of a bubble results from a balance between two opposing eects:
the buoyancy FB , of order of gR3 , which tends to make it emerge from the free
surface and the capillary force FC inside the hemispherical thin liquid lm, of order of
(/R)R2 = R, which tends to maintain the bubble below the surface. (/R) is the
excess of pressure, due to Laplace law, inside the curved thin liquid lm of the bubblecap, and r2 is the order of magnitude of the emerged bubble-caps area. Comparing
these two opposing forces comes down to compare the
 bubble radius, R 0.5 mm,
with the capillary length of the liquid medium, 1 = /g 2 mm, (where , and
g are respectively the liquid surface tension, the liquid density and the acceleration due
to gravity). Champagne bubbles radii being signicantly smaller than the capillary
length, gravity will be neglected in front of capillary eects. Consequently, like a tiny
iceberg, a oating bubble only slightly emerges from the liquid surface, with most of

Champagne Fizzics

67

(a)

1 mm
P0
P1 2 R1

P =

4 2
=
R1 R2

R1 = 2 R2 2 R

P 4 R1

bubble-cap

P0

air

liquid

bubble

R2

R1

FC

FB

1 mm

2R

(b)

Fig. 52. Oblique view of the bubble monolayer composed of quite millimetric bubbles
organized in a hexagonal pattern, each bubble being surrounded by an arrangement of six
neighbors (a); schematic transversal representation of ve aligned bubbles in touch in the
monolayer (b).

its volume remaining below the free surface. The scheme displayed in Fig. 52 presents
the various geometrical parameters linked with bubbles trapped near the free surface
in a bubble raft composed of quite millimetric bubbles (as are champagne bubbles).
The following reasoning does not take into account interactions between adjoining
bubbles (no dimple eects, for example).
Let us denote, R1 and R2 , the radii of curvature of the emerged bubble-cap and
the immerged bubble volume, respectively. Therefore, since the thin emerging bubblecap possesses two interfaces whereas the immerged part of the bubble possesses only
one, the excess of pressure inside a bubble, P , may be written as follows (according
to the Laplace law):
4
2
P =
=
R1 = 2R2 .
(69)
R1
R2
Consequently, the radius of curvature of the emerged bubble-cap is approximately
twice that of the submerged bubble volume. Furthermore, since most of the bubble
volume remains below the liquid surface, R2 R and consequently, R1 2R. Obviously, the transition between these two zones of dierent curvature does not appear

68

The European Physical Journal Special Topics

so abruptly. A complete numerical study of the shape taken by a bubble trapped


at a gas/liquid interface was conducted by Laurent Duchemin, in the limit of small
and large bubble radii [91,92]. The radius of the emerged bubble-cap, r, as shown in
Fig. 52, can also be estimated by equaling the bubble buoyancy with the capillary
pressure inside the thin lm as:



4 3
4 R2
2
R r r
0.14 mm.
(70)
g
3
R
3 1
Finally, the order of magnitude of the angle between the z-axis and the end of the
bubble-cap may also be easily determined as, sin1 (r/2R) 8 .

7.2 Dynamics of a bubble-caps aperture


At the free surface, since bubbles radii are signicantly smaller than the capillary
length, the liquid lms of bubble-caps progressively get thinner due to capillary
drainage. When the liquid lm of a bubble-cap reaches a critical thickness, it becomes fragile and nally ruptures. Since the pioneering work of Lord Rayleigh, in the
late 19th century [93], the rupture of thin lms has been widely experimentally, theoretically and numerically investigated. It was found that a hole appears in the lm,
surrounded by a rim which collects the liquid and that propagates very quickly driven
by surface tension forces. Balancing inertia with surface tension, Culick proposed the
following velocity for the growing hole in a thin rupturing liquid sheet [94],

uCulick

2
e

(71)

where e is the thickness of the liquid lm. The latter expression has already been
experimentally conrmed numerous times for the bursting of low viscous thin liquid
lms (see for example the pioneering work by McEntee and Mysels [95] and also that
by Pandit and Davidson [96], where beautiful snapshots of the disintegration of soap
bubbles are presented).
In case of a spherical cap with a radius of curvature 2R, rupturing from its axis
of symmetry (the z-axis) as schematized in Fig. 53, the velocity u of the propagating
liquid rim (of mass M ) is ruled by the following equation:
d
(M u) = 2 (2R) sin d.
dt

(72)

By replacing in the latter equation, the mass of the liquid rim by its expression,
M = e(2R)2 (1 cos )d, the velocity u by 2R(d/dt) and by developing, one
obtains the following dierential equation:

 2

2
d
2 d
e (2R)
(1 cos ) + sin
= 2 sin .
(73)
2
dt
dt
Finally, by considering only the constant velocity solution (d2 /dt2 = 0), one obtains
the following expression u for the growing hole,

d
2
=
= uCulick
u = 2R
(74)
dt
e

Champagne Fizzics

69

liquid rim of mass

M = e(2 R ) (1 cos )d
2

surface tension force

F = 2 (2 R )sin d

2R
0

Fig. 53. Schematic representation, in a rupturing spherical bubble-cap with a radius of curvature 2R, of the growing hole surrounded by a rim which collects the liquid and propagates
driven by surface tension forces.

which is the same as that derived by Culick. The thickness e of a millimetric champagne bubble-cap was already experimentally determined by the classical microinterferometric technique and found to be of order of 106 to 107 m [97]. As a
result, by replacing in Eq. (74) the known values of , ( 103 kg m3 ) and e, u is
expected to be of order of 10 m s1 for a rupturing champagne liquid lm. Finally,
the characteristic time scale of a millimetric bubble-caps disintegration should be
around:


(75)
r u 1.4 104 10 10 s.
7.3 An Estimation of the shear stresses during a bubble-caps aperture
To estimate the shear stress during the bubble-caps disintegration, we will use an
alternative approach, inspired from Clarkson et al., based on the energy dissipation
rate [98]. The energy dissipation rate is the rate at which energy is dissipated per
unit mass of liquid,
1 dED
=
(76)
m dt
where m is the mass of the liquid lm (m er2 ). Half of the total surface free energy
ET of the bubble-cap (where ET = 2r2 , due to the both gas/liquid interfaces) is
dissipated by viscous eects into the liquid rim around the growing hole, the other half
being converted into kinetic energy [9496]. Therefore, ED ET /2. As a result, during
the bubble-caps disintegration phase, the energy dissipation rate can be approached
as:


1
r2

.
(77)
2
er

70

The European Physical Journal Special Topics

Fig. 54. Scheme of the two production ways of droplets from a bursting bubble; redrawn
from the article by Resch et al. [100].

As presented in the work by Clarkson et al. [98], we retrieved the turbulence theory
to estimate the shear stress in the liquid rim, from the energy dissipation rate by
use of the following relationship [99],

 1/2

(78)

where is the kinematic viscosity dened as, = /. Finally, combining Eqs. (77)
and (78) yields to the following expression for the shear stress encountered as the
emerging bubble-cap disintegrates:

 1/2
e

(79)

By replacing in Eq. (79) the known values of , , e and in the present situation,
shear stresses in the propagating liquid rim are expected to be comprised in between
3 103 Nm2 and 9 103 N m2 , depending on the lm thickness.
7.4 The bursting process as frozen through high-speed photography
It is now generally recognized that bubbles bursting at a liquid surface eject two kinds
of droplets [100]: (i) small droplets called lm drops, formed as the lm of the emerged
bubble-cap disintegrates, and (ii) droplets formed by the collapse of the bottom of the
bubble, called jet drops (see Fig. 54). Nevertheless, it was shown that bubbles with
a diameter less than about 2 mm produce no lm drops as they burst [100]. Because
the champagne bubbles diameter rarely exceeds about 1 mm as they reach the liquid
surface, it can be concluded that only jet drops constitute the cloud of droplets above
the liquid surface.
Following Eq. (75), the characteristic time scale needed for the disintegration
of the bubble-cap of millimetric bubbles is of order of 10 s. During this extremely
brief initial phase, the bulk shape of the bubble is literally frozen, and a nearly
millimetric open cavity remains as a tiny indentation in the liquid surface (see for
example the two high speed photographs displayed in Fig. 55 [101]).
Then, a complex hydrodynamic process ensues, causing the collapse of the submerged part of the bubble and projecting into the air a liquid jet (the often so-called
Worthington jet [102]) which quickly breaks up into tiny droplets of liquid (called jet
drops). This process is indeed characteristic of every carbonated beverage. Generally

Champagne Fizzics

71

Fig. 55. High-speed photographs showing two cavities left by oating bubbles after the
disintegration of the bubble-cap (bar = 1 mm). (Photographs by Gerard Liger-Belair.)

speaking, the number, size, and velocity of jet drops produced during bubble collapse
depend on both the size of the initial bursting bubble and some liquid properties
(such as its surface tension, density and viscosity) [103106]. In Fig. 56, a reconstructed time-sequence frozen through high-speed photography is displayed. It shows
the formation of a tiny liquid jet, caused by the collapse of a champagne bubble, and
which breaks up into several droplets [101].

7.5 Close-up on the Worthington jet


The so-called Rayleigh-Plateau instability is a capillary wave, which quickly develops along the Worthington jet, and which is responsible for its breakup into several
droplets. The photograph displayed in Fig. 57(a) shows the split-second traveling
of the Rayleigh-Plateau instability which develops along the jet, as frozen through
high-speed photography. It is compared with the Rayleigh-Plateau instability which
develops along a cylindrical water stream falling from a leaking faucet (see Fig. 57(b)).
The length scale l of the upward jet radius being of order of 100 m, its corresponding Bond number is therefore expected to be Bo = (gl2 )/ 2103  1. Moreover,
the upward jet velocity U being of the order of 5 m/s [91,92], the Reynolds number
of the jet is therefore of order of Re = (U l)/ 102 1. Therefore, the RayleighPlateau instability, which develops along the champagne jet, results from a balance
between inertia (dU/dt U/R ) and surface tension gradients (P /l2 ), with
R being the characteristic time scale for the instability to develop. Finally, by equaling the two latter expressions, the characteristic time scale R for the Rayleigh-Plateau

72

The European Physical Journal Special Topics

Fig. 56. Reconstructed time-sequence showing the situation following the collapse of a single
millimetric bubble; the collapse of the empty cavity (frame 1) leads to the projection of a tiny
liquid jet (frames 2 and 3) which breaks up into several jet drops (frame 4) (bar = 1 mm).
(Photographs by Gerard Liger-Belair.)

instability to develop in the jet is of order of:


2

103 5 104
U l2
R

103 s 1ms.

5 102

(80)

In comparison, the characteristic time scale required for the Rayleigh-Plateau instability to develop in a millimeter-size water stream falling from a leaking faucet is
rather of order of 3 ms [107].
7.6 A paternoster for surface active molecules
As champagne or sparkling wine is poured into a glass, the myriad of ascending bubbles collapse and therefore radiate a multitude of tiny droplets above the free surface,

Champagne Fizzics

(a)

73

(b)

Fig. 57. Comparison between the Rayleigh-Plateau instability which develops along the
tiny champagne jet (a), and the one which develops along a cylindrical water jet falling from
c Gerard Liger-Belair/Daniel Schwen).
a leaking faucet (b) (

into the form of very characteristic and refreshing aerosols, as shown in the photograph displayed in Fig. 58. Those tiny droplets, ejected up to several centimeters
above the liquid surface, partly evaporate themselves, thus accelerating the transfer
of the numerous aromatic volatile organic compounds above the liquid surface. This
very characteristic zz considerably enhances the avor release in comparison with
that from a at wine for example. Laser tomography techniques were applied to freeze
the huge number of bursting events and the myriad of droplets ejected above champagne glasses in real consuming conditions (see the tomography of the droplets cloud
above the surface of a coupe displayed in Fig. 59) [23].
On a larger scale indeed, sea spray, which transports dissolved gases, salts, and
biological materials to the atmosphere (and therefore inuences climate), is largely
attributed to aerosols produced by an estimated 1018 to 1020 bubbles that rupture
every second across the oceans [108110]. Marine aerosols account for the majority of
the global natural aerosol ux and consequently may have a signicant impact on the
Earths radiative balance and biogeochemical cycling [111,112]. On a smaller scale,
it is worth noting that bubble-bursting-driven aerosols have been implicated in the
transmission of diseases in swimming pools and hot tubs [113].
It is indeed well known that preferential adsorption of surfactants at the airsolution interface occurs as a result of the amphiphilic properties of surfactants, with
the water-soluble moiety plunging into the solution and the hydrophobic component
in contact with the air. Enrichment of surfactant materials in the sea-surface microlayer, and in atmospheric aerosols has long been well characterized [114118].
Adsorption of surfactants is even considerably increased at the sea surface during
rough sea conditions, when wave breaking action causes air bubbles to be trapped
under the water surface [112]. Actually, bubbles trapped in the liquid bulk considerably increase exchange surfaces between the sea bulk and the atmosphere. Bubbles
drag surfactants along their way through the liquid bulk, reach the sea surface, to
nally burst and eject aerosol droplets into the atmosphere. Air bubbles trapped

74

The European Physical Journal Special Topics

Fig. 58. The collapse of hundreds of bubbles at the free surface radiate a cloud of tiny
droplets which is characteristic of champagne and other sparkling wines and which complec Alain Cornu / Collection CIVC).
ments the sensual experience of the taster (

during rough sea conditions were found to increase surfactant concentrations in


aerosols by several orders of magnitude compared with those found in the liquid
bulk [112]. Recently, Bird et al. proved that, by rupturing, a single large bubble (with
several centimeters in diameter) can fold and entrap air as it retracts, thus leading to
the creation of a ring of small daughter millimetric bubbles which burst in turn [119].
This phenomenon is believed to considerably increase the number and eciency of
aerosol dispersal.
From a conceptual point of view, the situation found in glasses poured with champagne or sparkling wine is nally very similar to that described above. Champagne
holds indeed hundreds of surface active compounds. Once champagne is poured into
a glass, bubbles nucleated on the glass wall drag champagne surfactants along their
way through the liquid bulk [3]. Surfactants dragged along with ascending bubbles
nally reach the free surface and concentrate themselves at the air/champagne interface. At the free surface of a glass poured with champagne, the ever-increasing
concentration of surfactants was indeed indirectly evidenced by observing the everincreasing lifetime of bubbles with time [3]. Actually, the ever-increasing surface concentration of surfactants progressively changes the boundary condition on the bubble

Champagne Fizzics

75

Fig. 59. The aerosol constituted by myriads of tiny droplets ejected from bubbles bursting
above the surface of a coupe, as seen through laser tomography technique; the droplets
trajectories are materialised by blue streaks of light during the 1s-exposure time of a digital
photo camera. (Photograph by G. Liger-Belair, F. Beaumont and G. Polidori.)

surface from slip to non-slip, thus reducing the drainage velocity and extending the
bubbles lifetime. Quite recently, the formation of adsorption layers of amphiphile
macromolecules at the air/champagne interface were directly evidenced through ellipsometry and Brewster angle microscopy (BAM) experiments [120,121].
Based on a phenomenological analogy between the zz of the ocean and the zz
in Champagne wines, it was hypothesized a few years ago, that aerosols found in the
headspace above a glass poured with champagne could considerably enhance the fragrance release of champagne by bringing chemical compounds to the tasters nostrils,
showing both surface activity and organoleptic interest [101]. Recently, ultrahigh resolution mass spectrometry (ICR-FT/MS) was used in order to analyze the aerosols
released by champagne bubbles [122]. In comparison with the champagne bulk, champagne droplets following bubble collapse were denitely found to be over-concentrated
with various surface active compounds, some of them showing indeed organoleptic interest (see Fig. 60). Each dot displayed in Fig. 60 represents the concentration factor
of a given compound found in the aerosols (i.e., the ratio of its concentration in the
aerosols to its concentration in the bulk, below the champagne surface). These compounds, mostly including saturated and unsaturated fatty acids, act as surfactants
(i.e., as double-ended compounds with one end attracted to the liquid phase and another that shuns it). It was suggested that champagne bubbles drain these compounds
out of the liquid bulk toward the liquid surface, with the hydrophobic end attracted
by the bubbles airy inside and the hydrophilic end attracted by the liquid outside.

76

The European Physical Journal Special Topics


35
30
25
20

Cf
15
10
5

1000

900

800

700

600

500

400

300

200

m/z
Fig. 60. Concentration factors analysis of all masses present in the mass spectra of champagne aerosols and bulk, respectively (in the whole mass range m/z 150-1000).

The bubbles then rise to the surface of the glass where they pop, releasing the compounds as aerosols. This recent discover supports the idea that rising and collapsing
bubbles act as a continuous paternoster lift for aromas in every glass of champagne.
Aerosols were thus found to hold the organoleptic essence of champagne.

7.7 When champagne bubbles dress up like owers


The close observation of bubbles collapsing at the free surface of a glass poured with
champagne also revealed another unexpected and lovely phenomenon. A few seconds
after pouring, and after the collapse of the foamy head, the surface of a champagne
ute is covered with a layer of monodisperse bubbles a kind of bubble raft, or 2D
foam, where each bubble is generally surrounded by six neighbouring bubbles (see
Fig. 61). Bubbles arrange themselves in an approximate hexagonal pattern.
While snapping pictures of the bubble raft after pouring, I also accidentally
took some pictures of bubbles collapsing close to one another in the raft. When
the bubble-cap of a bubble ruptures and leaves an open cavity at the free surface,
adjacent bubble-capse sucked towards this empty cavity and create unexpected and
short-lived ower-shaped structures, unfortunately invisible to the naked-eye (see for
example the high speed photographs displayed in Fig. 62) [123,124]. Shear stresses
induced by bubbles trapped in the close vicinity of a collapsing one are even better
visualized on the high-speed photograph displayed in Fig. 63, where the bubble raft
is not complete. Such behaviour rst appeared counter-intuitive to me. Paradoxically,
adjacent bubble-caps are sucked and not blown-up by bursting bubbles, contrary to
what could have been expected at rst glance.
Actually, after the disintegration of a bubble-cap, the hexagonal symmetry around
adjoining bubbles is suddenly locally broken. Therefore, the symmetry in the eld of
capillary pressure around adjoining bubbles is also locally broken. Capillary pressure
gradients all around the now empty cavity are detailed in Fig. 64.

Champagne Fizzics

77

Fig. 61. A few seconds after pouring, and after the collapse of the foamy head, the surface
of a ute is covered with a layer of quite mono disperse millimetric bubbles, where bubbles
arrange themselves in an approximate hexagonal pattern, strikingly resembling those in
beeswax (bar = 1 cm). (Photograph by Gerard Liger-Belair.)

Signs +/ indicate a pressure above/below the atmospheric pressure P0 . Finally,


inertia and gravity being neglected, the full Navier-Stokes equation applied to the
uid within the thin liquid lm of adjoining bubble-caps drawn by capillary pressure
gradients, reduces itself to a simple balance between the capillary pressure gradients
and the viscous dissipation as follows [124],



(u)S = P
(81)
S

where u is the velocity in the thin liquid lm of adjacent bubble-caps, is the


 are the capillary pressure gradients, and s being the axial
champagne viscosity, P
coordinate which follows the bubble-caps curvature and along which the uid within
the thin lm is displaced.

78

The European Physical Journal Special Topics

Fig. 62. Flower-shaped structure, as frozen through high-speed photography, found during
the collapse of bubbles in the bubble raft at the free surface of a ute poured with champagne
(bar = 1 cm). (Photograph by Gerard Liger-Belair.)

The asymmetry in the capillary pressure gradients distribution around a bubblecap adjacent to an empty cavity is supposed to be the main driving force of the
violent sucking process experienced by a bubble-cap in touch with a bursting bubble.
Actually, due to higher capillary pressure gradients, the liquid ows that develop in
the half part of the bubble-cap close to the open cavity are thus expected to be higher
than those which develop in the rest of the bubble-cap. It ensues a violent stretching
of adjoining bubble-caps toward the now empty cavity, which is clearly visible on the
photographs displayed in Figs. 62 and 63.
More recently, those ower-shaped structures have been observed during the coarsening of bi-dimensional aqueous foams, obtained by mixing a surfactant, sodium
dodecyl sulphate (SDS), with pure water [125]. But it is worth noting that this lovely
and short-lived process was rst observed at the top of a champagne ute.
During this sudden stretching process, adjacent bubble-caps areas signicantly
increase. A systematic image analysis of numerous time sequences similar to that
displayed in Fig. 5 demonstrated an average increase A around 15%, of bubble-caps
areas adjacent to a central collapsing bubble. The free energy of such a system will
be supposed to be mainly stored as surface free energy. Consequently, the density of
free energy per unit of volume in the liquid lm of a distorted bubble-cap (one petal
of the ower-shaped structure) can be evaluated as follows [123],


E
0.15 2
2A

104 105 J m3 = 104 105 N m2

(82)
V bubble-cap
Ae
e
where E is the corresponding surface free energy in excess during the stretching
process,V is the volume of the thin liquid lm of the emerging bubble-cap, A is the

Champagne Fizzics

79

Fig. 63. Shear stresses experienced by bubbles adjacent to a collapsing one at the free surface
of a ute poured with champagne (bar = 1 cm). (Photograph by Gerard Liger-Belair.)

emerging bubble-cap area, and e is the thickness of the thin liquid lm of the bubblecap (of order of 106 107 m). Therefore, the liquid ows induced by the capillary
pressure gradients are responsible for a density of energy per unit of volume dimensionally equivalent to shear stresses of order of 104 105 N m2 (depending on the
lm thickness). By comparison, in previous studies, numerical models conducted to
stresses of order of (only) 103 N m2 in the boundary layer around single millimetric
collapsing bubbles [91,126]. Therefore, stresses in the bubble-caps of bubbles adjacent
to collapsing cavities appear to be, at least, one order of magnitude higher than those
observed around single collapsing cavities, and also even higher than those developed
by viscous dissipation as a bubble-cap disintegrates. Intuitively, this is nally not so
surprising. Actually, after a bubble-caps aperture, the now empty cavity has to collapse to recover the horizontality of the liquid surface. The potential energy of such
an unstable situation is the sum of two contributions: rst, the gravitational energy
associated with the buoyancy of the open cavity (of order of gR4 ) and second, the
dierence of surface free energy between an hemispherical open cavity of radius R
and a at circular surface of radius R (of order of R2 ). The ratio of the gravitational energy to the surface free energy is called the Bond number, Bo = (gR2 )/.
In the present situation, Bo being of order of 5 102 , the main source of energy is
undoubtedly the surface free energy. Now, it should be noted that the same driving
potential energy is responsible for the single cavity collapse and for the collapse of the
cavity surrounded by neighboring bubble-caps. But, the volume of liquid displaced
in the thin lm of adjoining bubble-caps being much less than that of the boundary
layer drawn, it logically induces higher energy dissipation rates per unit of volume
and therefore higher strains in the petals of the ower-shaped structure. Finally,

80

The European Physical Journal Special Topics

P0
P0
P1 + R

(P )S

P + 2 R

s
bubble-cap

air
Free surface

2R
champagne

bubble

(P )S (PC P1 )

R 3 R 2

P0
PC 2 R
Fig. 64. Schematic transversal representation of the situation, as frozen after the disintegration of the central bubble-cap.

while absorbing the energy released during a bubble collapse, as so many tiny airbags would do, adjoining bubble-caps store this energy into the thin liquid lm of
emerging bubble-caps, leading nally to stresses much higher than those observed in
the boundary layer around single millimetric collapsing bubbles.
In addition to purely physicochemical reasons, biological reasons are also readily
found for the investigation of such ower-shaped structures around bubbles collapsing in a bubble raft. Actually, in the biological industry, animal cells cultivated in
bio-reactors were shown to be seriously damaged or even killed by the bursting of
gas bubbles used to aerate the culture medium [127,128]. It has even been suggested
that structural deformations of adjacent tissues are induced by bubble collapse during laser-induced angioplasty [129,130]. Kunas and Papoutsakis evaluated the critical
shear stresses needed to cause irreversible damages to classical animal cells [128]. They
found critical lethal stresses in the range between 103 N m2 and 104 N m2 . Therefore, by developing stresses of order of 104 105 N m2 , bubbles bursting in a bubble
raft should be potentially even more dangerous for micro-organisms or biological tissues trapped in the thin lm of these fast-stretched bubble-caps.
7.8 Avalanches of bursting events in the bubble raft?
Actually, avalanches of popping bubbles were put in evidence during the coarsening of bi-dimensional and three dimensional aqueous foams [125,131]. How does the

Champagne Fizzics

81

bubble raft behave at the surface of a ute poured with champagne? Does a bursting
bubble produce a perturbation which extends to the neighboring bubbles and induce
avalanches of bursting events which nally destroy the whole bubble raft? In case of
champagne wines, a few time sequences of bubbles bursting in the bubble raft have
been captured with a high-speed video camera. One of them is displayed in Fig. 65.
Between frame 1 and frame 2, the bubble pointed with the black arrow has
disappeared. In frame 2, neighboring bubbles are literally sucked toward this now
bubble-free area. Then, neighboring bubbles oscillate during a few milliseconds and
progressively recover their initial hemispherical shape. In conclusion, in case of bubbles adjacent to collapsing ones, despite high shear stresses produced by a violent
sucking process, bubbles adjacent to collapsing ones were never found to rupture
and collapse in turn, thus causing a chain reaction. At the free surface of a ute
poured with champagne, bursting events appear to be spatially and temporally non
correlated. The absence of avalanches of bursting events seems to be linked to the
champagne viscosity (which is about 50% higher than that of pure water) [125].
It can also be noted that a tiny daughter bubble, approximately ten times smaller
than the initial central bubble, has been entrapped during the collapsing process of the
central cavity (as clearly seen in frames 3 and 4 of Fig. 65). Bubble entrapment during
the collapsing process was already experimentally and numerically observed with single millimetric collapsing bubbles [91,132], including champagne bubbles [101]. This
bubble entrapment process can also be observed as drops impact on liquid surfaces
[133137].

7.9 And what about the Worthington jet?


Since we are dealing with bubbles collapsing at a free liquid/gas interface, we are logically tempted to wonder about the dynamics of the famous Worthington jet. Does
it exist, as in the single collapsing bubble case, or does the roughly hexagonal neighboring bubble pattern around a collapsing cavity strongly modify and even prevent
its formation?
Following Eq. (80), changes in the upward jet velocity U and/or changes in its
radius l, in comparison with the single bubble collapse case, could nally strongly
aect the overall dynamics of the Rayleigh-Plateau instability which breaks the jet
into droplets.
Once more, intuitively, dierences between the dynamics of the Worthington jet
could be expected. Actually, the bulk shape of bubbles adjacent to the empty cavity
left by the central collapsing bubble strongly changes the geometry of the system
beneath the free surface. Such a situation could therefore modify the converging liquid ows all around the empty cavity, thus probably modifying in turn the overall
dynamics of the upward liquid jet. Actually, a few snapshots froze the formation of a
Worthington jet for collapsing bubbles being surrounded by the hexagonal pattern of
six neighboring bubbles. Fig. 66 compares a liquid jet issued from a bubble collapsing in a bubble raft with a liquid jet issued from the collapse of a single champagne
bubble [124].
The jet following the single collapsing bubble seems to be better developed
than that following the bubble collapsing in touch with neighboring bubbles. The
same trend has been noticed on the few snapshots that froze the Worthington jet
in a bubble raft. Nevertheless, too few snapshots froze the Worthington jet in the
bubble raft. Further experimental investigations are to be conducted to enable us to
conclude about similitude and dierences with the dynamics of the jet following a
single bubble collapse.

82

The European Physical Journal Special Topics

This bubble-cap is
about to rupture

Fig. 65. Time sequence showing the dynamics of adjoining bubbles in touch with a collapsing
one at the free surface of ute poured with champagne; the whole process was lmed at 1500
frames/s; from frame 4, in the center of the empty cavity left by the collapsing bubble, a
tiny air-bubble entrapment is observed (bar = 1 mm).

Champagne Fizzics

83

(a)

(b)
Fig. 66. Comparison, as it divides into jet drops due to the Rayleigh-Plateau instability,
between the jet which follows the collapse of a bubble in the bubble raft (a), and that which
follows a single bubble collapse (b). (Photographs by Gerard Liger-Belair.)

84

The European Physical Journal Special Topics

Fig. 67. The liquid-jet surprisingly deviates from vertical as the neighbouring bubbles
symmetry is broken around a collapsing bubble (bar = 1 mm). (Photograph by Gerard
Liger-Belair.)

7.10 A particular situation: As the jet deviates from vertical


The high-speed photograph displayed in Fig. 67 froze the champagne liquid jet in
a quite particular situation [138]. On the right side of this picture, the collapsing
bubble is bordered by three neighboring bubbles, whereas on the left side, there
are no adjoining bubbles. The hexagonal symmetry is broken. In this case, the tiny
liquid jet (previously perfectly vertically oriented, as in the case of the single bubble
collapse displayed in Fig. 56) seems to deviate from vertical. The jet is tilted toward
the bubble-free area. There are certainly no enological consequences of such a
situation, but experts in the science of bubbles and foams ask themselves why such
a deviation from vertical is observed.
Further experimental observations, combined with numerical simulations, are soon
to be conducted in order to better understand the role of neighboring bubbles on the
dynamics of the jet formation and its breakup into droplets.

Champagne Fizzics

85

References
1. G. Liger-Belair, Uncorked: The Science of Champagne (Princeton University Press,
Princeton, 2004)
2. G. Liger-Belair, J. Rochard, Les Vins Eervescents : Du Terroir a
` La Bulle (Dunod,
Paris, 2008)
3. G. Liger-Belair, Ann. Phys. (Paris) 31, 1 (2006)
4. B. Duteurtre, Le Champagne : De La Tradition a
` La Science (Lavoisier, Paris, 2010)
5. G. Liger-Belair, R. Marchal, M. Vignes-Adler, B. Robillard, A. Maujean, P. Jeandet,
Am. J. Enol. Vitic. 50, 317 (1999)
6. G. Liger-Belair, C. Voisin, M. Vignes-Adler, B. Robillard, P. Jeandet, Langmuir 19,
1294 (2002)
7. G. Liger-Belair, C. Voisin, P. Jeandet, J. Phys. Chem. B 109, 14573 (2005)
8. G. Liger-Belair, M. Parmentier, P. Jeandet, J. Phys. Chem. B 110, 21145 (2006)
9. J.-M. Dessirier, C.T. Simons, M.L. Carstens, M. OMahony, E. Carstens, Chem. Senses
25, 277 (2000)
10. E. Carstens, M.L. Carstens, J.-M. Dessirier, M. OMahony, C.T. Simons, M. Sudo, S.
Sudo, Food Qual. Prefer. 13, 431 (2002)
11. J. Chandrashekar, D. Yarmolinsky, L. von Buchholtz, Y. Oka, W. Sly, N. Ryba, C.
Zuker, Science 326, 443 (2009)
12. A. Dunkel, T. Hofmann, Angew. Chem. Int. Ed. 49, 2975 (2010)
13. C. Priser, P.X. Etievant, S. Nicklaus, O. Brun, J. Agric. Food Chem. 45, 3511 (1997)
14. T. Tominaga, G. Guimbertau, D. Dubourdieu, J. Agric. Food Chem. 51, 1016 (2003)
15. M.A. Pozo-Bayon, M. Santos, P.J. Martin-Alvarez, G. Reineccius, Flavour Fragr. J.
24, 226 (2009)
16. A. Saint-Eve, I. Deleris, E. Aubin, E. Semon, G. Feron, J.-M. Rabillier, D. Ibarra, E.
Guichard, I. Souchon, J. Agric. Food Chem. 57, 5891 (2009)
17. G.G. Agabaliantz, Bull. OIV 36, 703 (1963)
18. G. Liger-Belair, J. Agric. Food Chem. 53, 2788 (2005)
19. D.R. Lide, H.P. Frederikse, CRC Handbook of Chemistry and Physics, 76th edn. (CRC
Press, Boston, 1995)
20. H. Alexandre, M. Guilloux-Benatier, Aust. J. Grape Wine Res. 12, 119 (2006)
21. M.A. Pozo-Bayon, A. Martinez-Rodriguez, E. Pueyo, M.V. Moreno-Arribas, Trends
Food Sci.Tech. 20, 289 (2009)
22. S. Torresi, M.T. Frangipane, G. Anelli, Food Chem. 129, 1232 (2011)
23. G. Liger-Belair, G. Polidori G, P. Jeandet, Chem. Soc. Rev. 37, 2490 (2008)
24. G. Liger-Belair, S. Villaume, C. Cilindre, P. Jeandet, J. Agric. Food Chem. 57, 1997
(2009)
25. G. Liger-Belair, S. Villaume, C. Cilindre, G. Polidori, P. Jeandet, J. Agric. Food Chem.
57, 4939 (2009)
26. C. Cilindre, G. Liger-Belair, S. Villaume, P. Jeandet, R. Marchal, Anal. Chim. Acta
660, 164 (2010)
27. S. Silva, M. Sabino, V. Fernandes, V. Correlo, L. Boesel, R. Reis, Int. Mater. Rev. 50,
345 (2005)
28. H. Pereira, Cork: Biology, Production and Uses (Elsevier, Amsterdam, 2007)
29. D.P. Faria, A.L. Fonseca, H. Pereira, O.M. Teodoro, J. Agric. Food Chem. 59, 3590
(2011)
30. G. Liger-Belair, S. Villaume, C. Cilindre, P. Jeandet, Anal. Chim. Acta 660, 29 (2010)
31. G. Liger-Belair, S. Villaume, J. Agric. Food Chem. 59, 4051 (2011)
32. A. Dussaud, Ph.D. thesis, ENSIAA, Massy, 1993
33. G. Autret, G. Liger-Belair, J.-M. Nuzillard, M. Parmentier, A. Dubois de Montreynaud,
P. Jeandet, B.T. Doan, J.-C. Beloeil, Anal. Chim. Acta 535, 73 (2005)
34. G. Liger-Belair, G. Polidori, Voyage au cur dune bulle de champagne (Odile Jacob,
Paris, 2011)
35. D. Gordge, R. Page, Exp. Fluids 14, 409 (1993)

86

The European Physical Journal Special Topics

36. H. Pron, D. Caron, F. Beaumont, G. Liger-Belair, G. Polidori, J. Visualization 13, 181


(2010)
37. G. Liger-Belair, M. Bourget, H. Pron, G. Polidori (in preparation)
38. L. Sigurdson, Phys. Fluids A 3, 2034 (1991)
39. G. Liger-Belair, M. Bourget, S. Villaume, P. Jeandet, H. Pron, G. Polidori, J. Agric.
Food Chem. 58, 8768 (2010)
40. P. Tipler, G. Mosca, Physics for Scientists and Engineers (W.H. Freeman ed.,
New York, 2003)
41. F. Incropera, D. Dewitt, T. Bergman, A. Lavine, Fundamentals of Heat and Mass
Transfer, 6th edn. (Wiley, 2007)
42. S.D. Lubetkin, M. Blackwell, J. Colloid Interface Sci. 126, 610 (1988)
43. F. Lugli, F. Zerbetto, Phys. Chem. Chem. Phys. 9, 2447 (2007)
44. M. Blander, J. Katz, AIChE J. 21, 836 (1975)
45. S.F. Jones, G.M. Evans, K.P. Galvin, Adv. Colloid Interface Sci. 80, 27 (1999)
46. S.D. Lubetkin, Langmuir 19, 2575 (2003)
47. G. Liger-Belair, R. Marchal, P. Jeandet, Am. J. Enol. Vitic. 53, 151 (2002)
48. A.D. Ronteltap, M. Hollemans, C.G. Bisperink, A. Prins, Master Brew. Ass. Am. Tech.
Quart. 28, 25 (1991)
49. D.M. Lynch, C.W. Bamforth, J. Food. Sci. 67, 2696 (2002)
50. G. Liger-Belair, J.-B. Religieux, S. Fohanno, M.-A. Vialatte, P. Jeandet, G. Polidori,
J. Agric. Food Chem. 55, 882 (2007)
51. G. Liger-Belair, F. Beaumont, M.-A. Vialatte, S. Jegou, P. Jeandet, G. Polidori, Anal.
Chim. Acta 621, 30 (2008)
52. G. Polidori, P. Jeandet, G. Liger-Belair, Am. Sci. 97, 294 (2009)
53. A. OSullivan, Cellulose 4, 173 (1997)
54. R. Lucas, Kolloid Z. 23, 15 (1918)
55. E. Washburn, Phys. Rev. 17, 273 (1921)
56. G. Liger-Belair, E. Prost, M. Parmentier, P. Jeandet, J.-M. Nuzillard, J. Agric. Food
Chem. 51, 7560 (2003)
57. G. Liger-Belair, D. Topgaard, C. Voisin, P. Jeandet, Langmuir 20, 4132 (2004)
58. G. Liger-Belair, Ann. Phys. (Paris) 27, 1 (2002)
59. G. Liger-Belair, A. Tufaile, B. Robillard, P. Jeandet, J.-C. Sartorelli, Phys. Rev. E 72,
037204 (2005)
60. G. Liger-Belair, A. Tufaile, P. Jeandet, J.-C. Sartorelli, J. Agric. Food. Chem. 54, 6989
(2006)
61. G. Liger-Belair, F. Beaumont, P. Jeandet, G. Polidori, Langmuir 23, 10976 (2007)
62. G. Liger-Belair, R. Marchal, B. Robillard, T. Dambrouck, A. Maujean, M. VignesAdler, P. Jeandet, Langmuir 16, 1889 (2000)
63. T. Maxworthy, C. Gnann, M. K
urten, F. Durst, J. Fluid Mech. 321, 421 (1996)
64. N. Shafer, R. Zare, Phys. Today 44, 48 (1991)
65. T. Sherwood, R. Pigford, C. Wilke, Mass Transfer (Mac Graw-Hill, Chemical
Engineering Series, 1975)
66. Y. Zhang, Z. Xu, Elements 4, 47 (2008)
67. V.N. Bond, Philos. Mag. 4, 889 (1927)
68. L.E. Scriven, C.V. Sternling, Nature 187, 186 (1960)
69. V.G. Levich, Physicochemical Hydrodynamics (Prentice Hall, Englewood Clis,
New Jersey, 1962)
70. R.E. Davis, A. Acrivos, J. Fluid Mech. 21, 681 (1966)
71. N.M. Aybers, A. Tapuccu, W
arme Sto
ubertragung 2, 171 (1969)
72. J.F. Harper, J. Fluid Mech. 58, 539 (1973)
73. N.N. Rulev, Colloid Journal 42, 210 (1980)
74. P.C. Duineveld, Ph.D. thesis, University of Twente, 1994
75. B. Cuenot, J. Magnaudet, B. Spennato, J. Fluid Mech. 339, 25 (1997)
76. C. Ybert, J.-M. di Meglio, Eur. Phys. J. B 4, 313 (1998)
77. Y. Liao, J.B. McLaughlin, J. Colloid Interface Sci. 224, 297 (2000)

Champagne Fizzics
78.
79.
80.
81.
82.
83.
84.
85.
86.
87.
88.
89.
90.
91.
92.
93.
94.
95.
96.
97.
98.
99.
100.
101.
102.
103.
104.
105.
106.
107.
108.
109.
110.
111.
112.
113.
114.
115.
116.
117.
118.
119.
120.
121.

87

F. Brissonnet, A. Maujean, Am. J. Enol. Vitic. 42, 97 (1991)


F. Brissonnet, A. Maujean, Am. J. Enol. Vitic. 44, 297 (1993)
J.B. Malvy, B. Robillard, B. Duteurtre, Sci. Aliments 14, 88 (1994)
R. Marchal, S. Bouquelet, A. Maujean, J. Agric. Food Chem. 44, 1716 (1996)
C. Cilindre, S. Jegou, A. Hovasse, C. Schaeer, A.J. Castro, C. Clement,
A. Van Doresselaer, P. Jeandet, R. Marchal, J. Proteome Res. 7, 1199 (2008)
R. Clift, J.R. Grace, M.E. Weber, Bubbles, Drops and Particles (Academic Press,
New York, 1978)
J. Magnaudet, M. Rivero, J. Fabre, J. Fluid Mech. 284, 97 (1995)
G. Liger-Belair, P. Jeandet, Langmuir 19, 801 (2003)
S.M. Yang, S.P. Han, J.J. Hong, J. Colloid Interface Sci. 169, 125 (1995)
G. Liger-Belair, J.-B. Religieux, S. Fohanno, M.-A. Vialatte, P. Jeandet, G. Polidori,
J. Agric. Food Chem. 55, 882 (2007)
C. Cilindre, A. Conreux, G. Liger-Belair, J. Agric. Food Chem. 59, 7317 (2011)
R. Vuilleumier, V. Ego, L. Neltner, A.-M. Cazabat, Langmuir 11, 4117 (1995)
M. Gugliotti, T. Silverstein, J. Chem. Educ. 81, 67 (2004)
L. Duchemin, Ph.D. thesis, Universite dAix-Marseille II, 2001
L. Duchemin, S. Popinet, C. Josserand, S. Zaleski, Phys. Fluids 14, 3000 (2002)
Lord Rayleigh, Proc. R. Inst. 13, 261 (1891): Reproduced in Scientic Papers, Vol. 3
(Dover, New-York, 1964)
F.E. Culick, J. Appl. Phys. 31, 1128 (1960)
W.R. McEntee, K.J. Mysels, J. Phys. Chem. 73, 3018 (1969)
A.B. Pandit, J.F. Davidson, J. Fluid Mech. 212, 11 (1990)
J. Senee, B. Robillard, M. Vignes-Adler, Food Hydrocolloids 13, 15 (1999)
J.R. Clarkson, Z.F. Cui, R.C. Darton, J. Colloid Interface Sci. 215, 323 (1999)
H. Schlichting, Boundary Layer Theory (McGraw-Hill, New York, 1979)
F.J. Resch, J.S. Darrozes, G.M. Afeti J. Geophys. Res. 91, 1019 (1986)
G. Liger-Belair, H. Lemaresquier, B. Robillard, B. Duteurtre, P. Jeandet, Am. J. Enol.
Vitic. 52, 88 (2001)
A.M. Worthington, A Study of Splashes (Macmillan, New York, 1963)
D.E. Spiel, J. Geophys. Res. 99, 10289 (1994)
D.E. Spiel, J. Geophys. Res. 100, 4995 (1995)
D.E. Spiel, J. Geophys. Res. 102, 5815 (1997)
D.E. Spiel, J. Geophys. Res. 103, 24907 (1998)
P.G. De Gennes, F. Brochard-Wyart, D. Quere, Capillarity and Wetting Phenomena
(Springer-Verlag, New York, 2004)
A.H. Woodcock, C.F. Kientzler, A.B. Arons, D.C. Blanchard, Nature 172, 1144 (1953)
F. MacIntyre, J. Geophys. Res. 77, 5211 (1972)
J. Wu, Science 212, 324 (1981)
C.D. ODowd, M.C. Facchini, F. Cavalli, D. Ceburnis, M. Mircea, S. Decesari, S. Fuzzi,
Y.J. Yoon, J.-P. Putaud, Nature 431, 676 (2004)
C.D. ODowd, G. de Leeuw, Philos. Trans. R. Soc. A 365, 1753 (2007)
L.T. Angenent, S.T. Kelly, A. St Amand, N.R. Pace, M.T. Hernandez, Proc. Natl.
Acad. Sci. USA 102, 4860 (2005)
W.R. Barger, W.D. Garret, J. Geophys. Res. 75, 4561 (1970)
D.C. Blanchard, Tellus B 42, 200 (1990)
R.S. Tseng, J.T. Viechnicki, R.A. Skop, J.W. Brown, J. Geophys. Res. 97, 5201 (1992)
C. Oppo, S. Bellandi, N. Innocenti, A. Stortini, G. Loglio, E. Schiavuta, R. Cini, Mar.
Chem. 63, 235 (1999)
S. Ekstr
om, B. Nozi`ere, M. Hultberg, T. Alsberg, J. Magner, E.D. Nilsson, P. Artaxo,
Biogeosciences 7, 387 (2010)
J.C. Bird, R. de Ruiter, L. Courbin, H.A. Stone, Nature 465, 759 (2010)
N. Peron, A. Cagna, M. Valade, C. Bliard, V. Aguie-Beghin, R. Douillard, Langmuir
17, 791 (2001)
N. Peron, J. Meunier, A. Cagna, M. Valade, R. Douillard, J. Microscopy 214, 89 (2004)

88

The European Physical Journal Special Topics

122. G. Liger-Belair, C. Cilindre, R. Gougeon, M. Lucio, I. Gebef


ugi, P. Jeandet, P. SchmittKopplin, Proc. Natl. Acad. Sci. USA 106, 16545 (2009)
123. G. Liger-Belair, B. Robillard, M. Vignes-Adler, P. Jeandet, C. R. Phys. 2, 775 (2001)
124. G. Liger-Belair, P. Jeandet, Langmuir 19, 5771 (2003)
125. H. Ritacco, F. Kiefer, D. Langevin, Phys. Rev. Lett. 98, 244501 (2007)
126. J.M. Boulton-Stone, J.R. Blake, J. Fluid Mech. 254, 437 (1993)
127. A. Handa, A.N. Emery, R.E. Spier, Dev. Biol. Stand. 66, 241 (1987)
128. K.T. Kunas, E.T. Papoutsakis, Biotech. Bioengng. 36, 476 (1990)
129. T. G. Van Leeuwen, J.H. Meertens, E. Velema, M.J. Post, C. Borst, Circulation 87,
1258 (1993)
130. E.A. Brujan, Europhys. Lett. 50, 175 (2000)
131. N. Vandewalle, J.F. Lentz, S. Dorbolo, F. Brisbois, Phys. Rev. Lett. 86, 179 (2001)
132. J. Herman, R. Mesler, J. Colloid Interface Sci. 117, 565 (1987)
133. H.N. Oguz, A. Prosperetti, J. Fluid Mech. 203, 143 (1990)
134. M.S. Longuet-Higgins, J. Fluid Mech. 214, 395 (1990)
135. A. Prosperetti, H.N. Oguz, Ann. Rev. Fluid Mech. 25, 577 (1993)
136. M. Rein, J. Fluid Mech. 306, 145 (1996)
137. D. Morton, M. Rudman, L. Jong-Leng, Phys. Fluids 12, 747 (2000)
138. G. Liger-Belair, Eervescence! La Science Du Champagne (Odile Jacob, Paris, 2009)