Foam in a two-dimensional Couette shear: a local measurement of bubble deformation
Extracted source text · arXiv cond-mat/0306590
%Utiles francois / alterner
\documentstyle[epsfig]{jfm}
%\documentstyle[referee,epsfig]{jfm}
%\documentclass[referee]{jfm} %documentclass plus recent
%\usepackage{graphicx,subfigure} %rico
%\usepackage{amstext} %rico
%\usepackage{amsfonts} %rico
%\usepackage{amssymb} %rico
%\usepackage[applemac]{inputenc} %rico
%\usepackage{pdfsync} %rico
%Inutiles:
%\documentstyle{jfm} %rico
%\documentstyle[referee]{jfm}
%\documentclass{jfm}
%\documentclass[nocite]{epl} %francois
%\documentstyle[aps,epsfig]{revtex} %francois
%\documentstyle[twocolumn,aps,epsfig]{revtex} %francois
%\documentstyle[aps, preprint, epsfig]{revtex} %francois
%\documentclass[aps,pdftex,floats]{revtex4} %francois
%\usepackage[pdftex]{graphicx} %francois
%\usepackage{times} %francois
%\usepackage[pdftex]{hyperref} %francois
%francois (eric, tu peux virer ce qui suit):
\def\bigcirc{open \; circles}
\def\bullet{closed \; circles}
\def\triangle{open \; triangles}
\def\blacktriangle{closed \; triangles}
\def\square{open \; squares}
\def\blacksquare{closed \; squares}
\def\boxplus{crossed \; boxes}
\def\Diamond{diamonds}
\def\times{crosses}
\begin{document}
%\draft
\title[Local measurement of bubble deformation
]{Foam in a two-dimensional Couette shear: \\
a local measurement of bubble deformation. }
\author[E. Janiaud and F. Graner]{
E. Janiaud$^{1,2}$ \and F. Graner$^{3}$
\thanks{$\;$to whom correspondence
should be addressed at graner@ujf-grenoble.fr}
}
\affiliation{$^1$
Laboratoire des Milieux D\'esordonn\'es et
H\'et\'erog\`enes, case 78, Universit\'e Paris
6 and CNRS UMR 7603, 140 rue de Lourmel, 75015 Paris,
France.\\[\affilskip]
$^2$
Universit\'e Paris 7, Denis Diderot, F\'ed\'eration de Recherche
FR2438 Mati\`ere et Syst\`emes Complexes, 2 place Jussieu, 75251 Paris Cedex
05, France.
\\[\affilskip]
$^3$
Laboratoire de Spectrom\'etrie Physique,
CNRS UMR 5588 et Universit\'e Grenoble I,
BP 87, F-38402 St Martin d'H\`eres Cedex, France.}
%\begin{document}
\maketitle
%\date{today}
\begin{abstract}
We re-analyse experiments on a foam sheared in a two-dimensional
Couette geometry [Debr\'egeas {\it et al.}, Phys. Rev. Lett. {\bf 87},
178305 (2001)]. We characterise the bubble deformation by a texture
tensor. Our measurements are local in time: they show two regimes,
one transient and one stationary. They provide both the
average and fluctuations of the anisotropy.
Measurements are also local in space: they show that both the
deformation and the elastic contribution to the stress field do
not localise, varying smoothly across the shear gap. We can thus
describe the foam
as a continuous medium with elastic properties.
\end{abstract}
%\maketitle
\vskip 2\baselineskip
\begin{figure}
\centering
\includegraphics[width=0.9\textwidth]{fig2phototrans1.eps}
\includegraphics[width=0.9\textwidth]{fig2phototrans2.eps}
\caption{Run 1 (transient regime): foam under steady quasistatic shear in a
Couette geometry. Pictures of the foam correspond to different values of applied shear strain $\gamma$:
(a) 0, immediately after switching from counter-clockwise to clockwise rotation; (b) 0.15; (c)
0.5, when the inner boxes have reached the yield strain
(see Fig. \protect\ref{Gamma s}); and (d) 7.6, well into the stationary regime.
The cross-component ${r\theta}$ of the texture tensor, defined in eq. (\protect\ref{defM}), is positive (a), zero (b), negative (c), and stationary (d).
}
\label{images}
\vspace{1cm}
\end{figure}
\section{Introduction}
A liquid foam exhibits ``complex" behaviour under stress: it is
elastic for small deformation, plastic for large deformation, and flows
at large deformation rates \cite{hutzler,jiangpre}. It is a
model for the study of
complex
fluids: its typical length scale permits direct
observation. In particular, a foam with only one bubble layer
(so-called ``two-dimensional foam'' \cite{hutzler}) is easy to image,
and image analysis yields information on all the geometrical
properties of the foam.
Recently, Debr\'egeas {\it et al.} \cite{debregeas1}
quasistatically sheared a fairly dry foam in a two-dimensional Couette
geometry (Fig. \ref{images}).
They studied the velocity field and showed that it localises
in a shear band:
its changes of value happen as a steep gradient
on a layer much thinner than the gap, rather than over a long distance.
It decreases exponentially across the gap $(r_1-r_0)$, as $\exp\left[-(r-r_0)/d\right]$. Here $r$ is the
distance to the wheel center, $r_0$ the radius of the inner wheel, $r_1$ the radius of the outer wheel,
and $d \ll (r_1-r_0)$ is a characteristic length
comparable to one bubble diameter. Such
shear-banding
has been explained and numerically simulated by the combination of
frozen disorder, elastic
behaviour at low deformation, and local discrete bubble
rearrangements, leading to an instability more likely to occur at
the inner side where fluctuations are larger
\cite{debregeas2}.
What dominates the macroscopic behaviour of the material: averages or
fluctuations?
This question is important in the rheology of foams, as well as in
other disordered materials.
Briefly, if the signature of fluctuations averages out at large
scales, a ``thermodynamical limit" exists in which the average
dominates; a constitutive equation relating
the stress and strain fields can then describe the material as a continuous medium
(like a solid or a liquid).
If the fluctuations have the same order of magnitude as the average
at all scales, they can dominate the physical behaviour, which need
not correspond to any continuous medium.
Whenever the fluctuations dominate, we can ask
whether they correlate (leading to large stress drops) or have a white spectrum.
The answer need not be the same for different materials (say, foam
and granular materials)
nor for different quantities (say, velocity and deformation).
Similarly, the existence and origin of localisation raise different
questions: the effect of boundary conditions,
the respective roles of strain and strain rate, the precise role of fluctuations, the differences between
various (disordered) materials, the distinction between sharp \cite{coussot} or exponential
\cite{debregeas1} decreases,
the dimension of
space (2D or 3D), and even the field itself (velocity or
deformation).
We would like to extract additional physical information, beyond the velocity field.
Most suggestions to quantify statistically the anisotropy and the
deformation
of the internal structure of the foam, that is, the
bubble wall network, are based on scalar quantities
\cite{elias}.
Since the deformation encompasses information on both anisotropy and
orientation,
it is actually a tensor \cite{landau,alexander}.
Recent suggestions of possible tensorial descriptions
based on bubble details are promising
\cite{kraynik2000,kraynik2003,ball,blumenfeld}.
Here, we use
the ``texture tensor'' (eq. \ref{defM}), which has the following
advantages
\cite{aubouy}: (i) it is purely geometrical, independent of stresses
and forces; (ii) it applies
to both small and large numbers of bubbles; (iii) it is local in time
and space \cite{delft}, so it can
characterise a sub-region of the foam; (iv) it applies to disordered
or ordered, 2D or 3D materials, in elastic, plastic or fluid regimes.
The plan of this paper is as follows.
Section (\ref{MatMet}) reviews the original experimental set-up, and
our image analysis protocol. The theoretical section
(\ref{theory}) reviews the definitions and measurements
of stress, texture and strain tensors; it can be read separately.
The next section (\ref{results}) presents our analyses of
experiments.
The first (section \ref{transient}),
resolved in both time and space, studies the transient regime before a
stationary flow is
established. The second (section \ref{permanent}), resolved in space but averaged in time,
shows that in
stationary flow, both the bubble anisotropy and the elastic contribution to the
deviatoric stress field vary smoothly across the
Couette gap; they do not
localise near the inner rotating wheel. The third (section \ref{fluctuat}),
again resolved in time,
shows that in stationary flow the fluctuations are larger near the
inner wheel. In section (\ref{discu}), we discuss whether we can describe the foam
as a continuous medium with elastic properties.
%\section{Materials: Experiments}
\section{Materials and Methods} \label{MatMet}
\subsection{ Experiments}
Georges Debr\'egeas generously provided experimental data obtained in two
different runs. For experimental details see Ref.
\cite{debregeas1}.
Briefly, the set-up consists of an inner shearing wheel and an outer fixed one, with no-slip conditions due to
tooth-shaped boundaries (tooth depth: radius 1.2 mm).
The foam is confined
between
two transparent plates separated by $h=2$~mm spacers (Fig. \ref{FigPhoto}(d)); this foam
thickness is smaller than
the
distance between bubble centers (see below),
enforcing a single layer of bubbles
\cite{Cox}.
The fluid fraction, measured by the weight of water
introduced in the cell, is $\Phi=5.2\%$. The bubble size dispersity
is $\sim 30\%$,
enough to prevent crystallisation.
Coarsening and size sorting under shear are negligible during the
experiment.
A CCD
digital camera records the position and the shape of the bubbles.
The inner wheel rotates slowly enough ($V_{wheel}=0.25$~mm.s$^{-1}$)
that the flow is well in quasistatic regime: \label{qstat}
the foam passes through a succession of equilibrium states.
All averaged static physical quantities, like the elastic strain or
stress, remain the same if the applied
shear rate
$V_{wheel}/(r_1-r_0)$
varies.
Kinetic quantities, like the
velocity field, the rate of T1s, and the input power and
dissipation, scale as the shear rate.
This scaling is typical of solid-like friction (only at
higher velocity does the foam exhibit a fluid-like friction, when
the input power and dissipation scale like the square of the
shear rate).
As
required for comparison to other experiments,
we thus express all results as a
function of the total applied shear strain:
\begin{equation}
\gamma= \frac{V_{wheel}\; t}{r_1-r_0}.
\label{prop_time}
\end{equation}
Note that $\gamma$ is
proportional to time $t$,
not to the local shear in each box.
Run 1 (unpublished data) contains a transient
regime. To prepare the foam, the inner disc is rotated counter-clockwise,
until
a stationary regime is reached. Then, at an arbitrary time, chosen as the origin
($t=\gamma=0$), the shear direction is switched
to clockwise (Fig. \ref{images}),
the experiment
begins and 500 pictures are recorded.
The $r\theta$ components
of the deformation thus switch from positive to negative (Fig. \ref{FigTransi}).
Each picture shows 200 bubbles, in an angular
sector of the experimental cell (Fig. \ref{images}).
The average distance between the centres of mass of neighbouring bubbles
is $3.0$ mm, with a standard deviation of $\pm 0.5$ mm, corresponding to a
mean bubble wall length of $\langle \ell \rangle = 1.7$ mm.
The internal
radius is
$r_0=71$~mm (same for both runs) and the external radius is $r_1=112$~mm
(different for each run).
Thus $\gamma=1$ corresponds
to a wheel displacement of $ V_{wheel}\; t = r_1-r_0=41$~mm: {\it
i.e.} 15-20 bubble
diameters.
\begin{figure}
%vspace{14cm}
\includegraphics[width=0.92\textwidth]{fig1PhotoM2.eps}
\caption{ Run 2 (stationary flow): image analysis and notations. (a) {\it Bottom left:} top view of the original
experiment by Debr\'egeas {\it
et al.} \protect\cite{debregeas1}, rotating counter-clockwise.
{\it Top right:} image analysis. The reconstructed network of
vertices linked by vectors
(white) and the boundaries of the 8 equi-spaced concentric circular boxes used
as representative volume elements (RVE) for spatial
analysis (black circular arcs) are superimposed on the snapshot of the experiment. We exclude the regions near both
wheels, where the bubbles are not entirely visible.
(b) We denote by $\vec \ell$ the vector linking both vertices of a given wall.
(c) If one of these vertices is outside of the box (defined
in (a), top right), we denote by $f \in [0,1]$ the
fraction of the bubble wall length inside the box; $f=0$ if the wall is
entirely outside, $f=1$ if it is entirely inside. (d) Side-view sketch of a vertical bubble wall cross-section, between two horizontal glass plates.
}
\label{FigPhoto}
%\vspace{1cm}
\end{figure}
Run 2 (published in
Ref. \cite{debregeas1}) focusses on the stationary regime.
The preparatory rotation is clockwise.
Then, at an arbitrary time, chosen as the origin
($t=\gamma=0$), it is switched to counter-clockwise
(Fig. \ref{FigPhoto}).
Pictures are recorded only after
a full $2\pi$-turn, corresponding to
$\gamma= 8$; the
$r\theta$ components
of the deformation are thus positive (Fig. \ref{FigPerm}).
A quarter of the Couette cell is filmed, with 700 bubbles in
each picture.
To improve the statistics, the camera is then displaced to record successively
the four quarters; 2000 pictures are obtained as four 500-images movies.
Data are measured and averaged over the four
quarters of the foam.
The average distance
between the centres of mass of neighbouring bubbles is $2.4$ mm, with
a standard
deviation of $\pm 0.4$ mm, corresponding to a mean bubble wall length of
$\langle \ell \rangle = 1.4$ mm.
The internal
radius is
$r_0=71$~mm (same for both runs) and the external radius is $r_1=122$~mm
(different for each run).
Thus $\gamma=1$ corresponds
to a wheel displacement of $51$~mm: {\it i.e.} 20-25 bubbles
diameters.
% \section{Methods: measurements}
\subsection{Image analysis}
We analyse the images using a home-made extension module to the
public software
NIH-Image. We threshold the image, and skeletise the
bubble walls to a one-pixel
thickness. We identify the points where three walls meet (``vertices'').
We replace each wall by the vector $\vec \ell$ linking its vertices
(white lines in Fig. \ref{FigPhoto}a, top right).
To respect the symmetry of
the Couette cell,
this wall vector $\vec \ell$ (Fig. \ref{FigPhoto}b) is expressed in local
polar coordinates $(r,\theta)$
(Fig. \ref{FigPhoto}a). This neglects the variation of the polar
referential between both its end points. This assumption is correct as long as $\ell \ll r$.
For the same reason, we choose equi-spaced orthoradial, circular boxes (Fig.
\ref{FigPhoto}a, top right) to divide the foam and measure
the deformation using spatial averages to improve the statistics.
In the quasistatic regime, the bubble deformation remains very
small and barely visible by eye. After trying various box sizes, we choose 8 boxes as a compromise
between the quality of the statistics and the level of visible
details.
For Run 1, each box is $ 2.7 \langle \ell\rangle $ wide,
and box number $i$ is at position $r_1^{i}=r_0 + [0.4+2.7
(i-0.5)] \langle \ell \rangle $.
The number of bubbles in each box grows linearly,
typically from 17 bubbles in the 1st box, to 32 in the 8th box.
For Run 2, $r_2^{i}=r_0 + [2.8+3.1(i-0.5)] \langle\ell \rangle $.
Due to the shape of the pictures (Fig.
\ref{FigPhoto}a), the number of bubbles first increases with $i$ (from
typically 105 bubbles in the 1st box), then decreases (to 41 in the
8th box).
\section{Definitions and measurements of deviatoric elastic stress and strain} \label{theory}
\subsection{Capillary contribution to the stress} \label{presentation stress}
\subsubsection{Theoretical definition}
In fluid foams, the
three main contributions to stress $\overline{\overline{\sigma}}$
are the capillary term
$\overline{\overline{\sigma}}^{cap}$
due to the tension of the liquid walls, the
pressure of the gas within each bubble, and the viscous dissipation%
.
%:
%\begin{equation}
% \overline{\overline{\sigma}}= \overline{\overline{\sigma}}^{cap} +
% \overline{\overline{\sigma}}^p
% +\overline{\overline{\sigma}}^{diss}.
%\end{equation}
In what follows, the ``deviatoric" part of the capillary stress refers to the traceless tensor, {\it i.e.}
$\overline{\overline{\sigma}}^{cap}- {\rm Tr}\left(\overline{\overline{\sigma}}^{cap}\right)
\overline{\overline{I}}/2$, where $ \overline{\overline{I}}$ is
the identity tensor in 2 dimensions. It consists in the
shear stress $\sigma^{cap}_{r\theta}$ and the normal stress difference
$\sigma^{cap}_{rr}-\sigma^{cap}_{\theta\theta}$.
In the extreme ``wet foam" limit $\Phi \sim 1$, bubbles are well separated from each other. The capillary contribution
in a representative volume element (RVE) ${{V}}$ which contains
$N$ bubbles with
surface tension $\Gamma$ is
\cite{rosenkilde,batchelor}:
\begin{eqnarray}
\overline{\overline{\sigma}}^{cap} &=& \frac{\Gamma}{{{V}}} \;
\sum_{S \in {{V}}} \oint_S \left(\overline{\overline{I}} - \vec{n}
\otimes \vec{n} \right) \; ds,
\nonumber\\
\sigma_{ij}^{cap} &=& \Gamma \; \frac{N}{{{V}}} \; \left\langle \oint_S
\left(\delta_{ij} - n_in_j \right) \; ds
\right\rangle_{{V}}.
\end{eqnarray}
Here $ N/{{V}}$ is the density of bubbles; $S$ is the surface of each bubble;
$\vec{n}$ is the unit vector normal to the surface element $ds$; and
$\langle . \rangle_{{V}}$ is the average over bubbles in ${{V}}$.
In the other extreme, ``dry foam" limit $\Phi \ll 1$, pairs of bubble surfaces merge into a
film with surface tension $2 \Gamma$. The capillary contribution is
\cite{khan1986,khan1987,kraynik2000,kraynik2003}:
\begin{equation}
\overline{\overline{\sigma}}^{cap} = \frac{2 \Gamma}{N{{V}}}
\oint_S \left(\overline{\overline{I}} - \vec{n} \otimes \vec{n}\right) ds
+ {\cal O}\left( \Phi^2\right).
\label{2gamma}
\end{equation}
The correction of order $\Phi^2$ is a contribution of the integral over
the surface $S$ of its mean
curvature \cite{mecke}. Since the mean curvature is mostly concentrated in bubble edges
(``Plateau borders"), this contribution is physically equivalent to the recently identified line tension
which affects the bubble edges \cite{kern,geminard}.
\subsubsection{Operational definition in two dimensions} \label{operational}
A true 2D foam has an interesting property: at equilibrium, or in a
quasistatic regime, the coarse-grained Laplacian of the pressure is zero
\cite{graner2001,wuppertal}.
The small scale fluctuations of bubble pressures, according to bubble side numbers,
average out to zero at large scales \cite{hutzler}.
In the present Couette geometry, where no large-scale pressure
gradient exists, pressure does not vary significantly
\cite{morgan}.
Hence, in-plane curvatures remain much smaller than the inverse of bubble
sizes. They contribute to the stress through their algebric
average (not their second moment), which is small along
any path within the foam: curvatures are counted positively if the path crosses the wall from the concave side \cite{wuppertal}.
They thus induce only a negligible correction to the deviatoric
capillary stress, with respect to
straight walls
\cite{asipauskas}.
However, adapting the above definition (eq. \ref{2gamma}) to
measure deviatoric capillary stress fields in bubble monolayers requires care. In
the present case, the curvature radius in the third dimension, not
visible in images, is smaller or equal to the half distance between plates
(Fig. \ref{FigPhoto}(d)).
Its inverse, namely the curvature in the third dimension, is larger
than the curvature visible in the image. This case is
thus intermediate between two limiting bubble monolayers: bubble
rafts, and Langmuir foams.
In bubble rafts, stress has already been measured, in a
circular Couette geometry, by mechanical measurements of the force
at the boundaries \cite{pratt}, or indirectly by local measurements of bubble
ellipticity \cite{ybert}. Since bubble top surfaces are curved (they are
approximately spherical caps), direct local measurements of stress are
difficult.
At the other extreme, Langmuir foams are one molecule thick: their
aspect ratio is of the order of nanometers divided by centimeters, with
no well-defined curvature in the third dimension. They are
thus probably the closest approximation to a true 2D foam (as long as
3D viscosity effects remain negligible), and capillary stress fields
can be measured as described below (eq. \ref{sigma cap})
\cite{langmuir_foam}.
In the present case,
for a single layer of bubbles \cite{Cox} in the quasistatic regime,
the dry part of the film remains vertical (Fig. \ref{FigPhoto}(d)).
By symmetry,
we can assume that
the curvatures on both sides of a film are almost opposite, so that
their effects almost cancel each other.
We thus neglect the effect of curvature in
the third dimension.
We simplify further
eq. (\ref{2gamma}) using two approximations. First, although strictly speaking
it applies
to closed surfaces
(complete surfaces of bubbles), we extend it to averages on open surfaces (walls between bubbles),
assuming that any possible non-compensated term (due for instance to bubbles at the foam boundary) has a negligible contribution to the average \cite{kraynik2003}.
Second, on closed surfaces
in 2D, the integral of the mean curvature only contributes a
constant, so that the ${\cal O}\left( \Phi^2\right)$ term
need not be considered here; in what follows, on open surfaces, we again neglect
the contribution of possible non-compensated terms.
Finally, eq. (\ref{2gamma})
simplifies to an expression which has already been used (but not always justified) for
various types of bubble monolayers
\cite{hutzler,debregeas2,asipauskas,langmuir_foam}:
\begin{equation}
\overline{\overline{\sigma }}^{cap} =\frac{1}{{{V}}} \sum_{\vec{\ell}}
f\; \left( \lambda \hat{e} \otimes \vec{\ell}\right).
\label{sigma cap dim}
\end{equation}
Here (Fig. \ref{FigPhoto}) the 2D RVE ${{V}}$
is a surface area; the $\vec{\ell}$s are the chords linking two
connected vertices, approximating bubble walls;
$f$ is the fraction of the wall length contained in ${{V}}$;
$\hat{e}= \vec{\ell}/\ell$ is the unit vector tangent to $\vec{\ell}$.
The line tension $\lambda$ is equal to $\Gamma h$ multiplied by a geometric
constant which depends on $\Phi$ (Fig. \ref{FigPhoto}(d)): it varies from 2 for a dry foam (two flat interfaces of length $h$)
to $\pi$ for a wet foam (two half-circles of diameter $h$).
\subsubsection{Comments on this definition} \label{comments stress}
Eq. (\ref{sigma cap dim}) is analogous
to the expression
for the elastic stress in a (2D or 3D) network with a density $\rho$
of links $\vec{\ell}$ undergoing two-body interactions with
tensions $\vec{\tau}$ \cite{kruyt}:
\begin{equation}
\overline{\overline{\sigma }}^{cap}= \rho \left\langle
\vec{\tau} \otimes \vec{\ell} \right\rangle_{{V}}.
\label{sigma gene}
\end{equation}
In the continuum limit, the average $\langle . \rangle_{{V}}$ is the average of the wall probability distribution function, and
this expression for the stress is exactly the same (see for instance
ref. \cite{kruyt} or the appendix of ref. \cite{aubouy}) as the
classical definition based on the links crossing the boundaries of
${{V}}$ \cite{landau}. This equivalence validates {\it a posteriori} our
application of eq. (\ref{2gamma}) to open surfaces. In
practice, experimental data have a finite size:
the number of links used in the calculation, and so the
signal/noise ratio of the resulting measurement \cite{asipauskas}, is much
larger if we use all links in ${{V}}$, rather than only the
links crossing its boundaries.
Eq. (\ref{sigma gene}) shows that, in a network of
points which interact
through a pairwise potential, we can measure
the stress from an image if we know the expression for the two-points force law
$\vec{\tau}= \vec{\tau}(\vec{\ell})$ explicitly
(for instance Lennard-Jones, or
harmonic).
For quasistatic 2D foams (eq. \ref{sigma cap dim}), all tensions
have the same modulus $\lambda$; hence
$\vec{\tau}(\vec{\ell}) = \lambda \hat{e}$.
The image
provides
$\overline{\overline{\sigma }}^{cap}/\lambda$, which is enough for
our purposes:
\begin{eqnarray}
\frac{\overline{\overline{\sigma }}^{cap}}{\lambda}&=&
\rho \left\langle
\hat{e} \otimes \vec{\ell}
\right\rangle_{{V}}
,\nonumber\\
\frac{\sigma^{cap}_{ij}}{\lambda}&=&
\rho \left\langle \frac{\ell_{i}\ell_{j}}{\ell}
\right\rangle_{{V}} .
\label{sigma cap}
\end{eqnarray}
Here
$\rho$ is the area density of bubble walls: the number of bubble walls
in ${{V}}$ weighted by $f$, divided by ${{V}}$;
$\langle .
\rangle_{{V}}$ is the average over the walls in ${{V}}$, also
weighted by $f$.
\subsection{Status of the elastic strain tensor}\label{presentation deformation}
\subsubsection{Macroscopic definition of the elastic strain}
A macroscopic definition of elastic strain \cite{macosko,farahani}
is usually a function of the {\it deformation gradient} $\overline{\overline{F}}$:
\begin{eqnarray}
\overline{\overline{F}}
&=& \left( \frac{\partial \vec{x}}{\partial \vec{x}^0}\right)^t = \left( \vec{\nabla}\; \vec{x} \right), \nonumber \\
F_{ij} &=& \frac{\partial x_i}{\partial x^0_j},
\label{defo_gradient}
\end{eqnarray}
where $\vec{x} $ is the current position of a point initially at position $\vec{x}^0$.
This deformation gradient $\overline{\overline{F}}$ coincides with the identity $\overline{\overline{I}}$ in a solid translation, and has a determinant 1 in a volume-conserving transformation.
From $\overline{\overline{F}}$, we can define two symmetric tensors $\overline{\overline{\cal U}}$ and $\overline{\overline{\cal V}}$ through a rotation
$\overline{\overline{R}}$ \cite{phan}:
\begin{equation}
\overline{\overline{F}} = \overline{\overline{R}} \; \overline{\overline{\cal U}} = \overline{\overline{\cal V}} \; \overline{\overline{R}}.
\label{stretch}
\end{equation}
They are called the {\it right} and {\it left stretch tensors}. They coincide with the identity $\overline{\overline{I}}$ in a solid transformation (translation and rotation), and also have a determinant 1 in a volume-conserving transformation.
In turn, they determine a {\it right} and a {\it left deformation tensors}, called the Finger tensor and the
Cauchy-Green tensor, respectively \cite{phan}:
\begin{eqnarray}
\overline{\overline{B}}
&=& \overline{\overline{F}} \; \overline{\overline{F}}^t =
\overline{\overline{\cal V}}^2, \nonumber \\
\overline{\overline{C}}
&=& \overline{\overline{F}}^t \; \overline{\overline{F}} = \overline{\overline{\cal
U}}^2.
\label{defo_tensor}
\end{eqnarray}
They too are symmetric, coincide with the identity $\overline{\overline{I}}$ in a solid transformation (translation and rotation), and have a determinant 1 in a volume-conserving transformation.
Finally, a {\it strain measure tensor} $\overline{\overline{E}}$ is for instance equal to
$\left(\overline{\overline{B}}-\overline{\overline{I}}\right)/2$, $\left(\overline{\overline{C}}-\overline{\overline{I}}\right)/2$, or more generally to one of the Seth-Hill strain measure tensors based on the successive powers of
$ \overline{\overline{\cal U}}$ or $ \overline{\overline{\cal V}}$:
\begin{eqnarray}
\overline{\overline{E}}^{(m)} &=&
\frac{ \overline{\overline{\cal U}}^m - \overline{\overline{I}}}{m},
\nonumber \\
{\rm or} \quad
\overline{\overline{E}}^{(m)} &=&
\frac{ \overline{\overline{\cal V}}^m - \overline{\overline{I}}}{m},
\label{seth_hill}
\end{eqnarray}
where $m$ is a positive or negative integer
\cite{farahani}. They all tend
towards the linear elasticity definition (the gradient of the displacement field)
in the limit of infinitesimal
transformation.
The particular case of $m=0$ corresponds to the
logarithm-based definition \cite{hoger} and generalises the Hencky
strain tensor \cite{hencky} to arbitrary deformations:
\begin{eqnarray}
\overline{\overline{E}}^{(0)} &= &
\log \overline{\overline{\cal U}} = \frac{1}{2} \; \log \overline{\overline{C}}
,
\nonumber \\
{\rm or} \quad
\overline{\overline{E}}^{(0)} &=&
\log \overline{\overline{\cal V}} = \frac{1}{2} \; \log \overline{\overline{B}}.
\label{logstrain}
\end{eqnarray}
This {\it true strain tensor} $\overline{\overline{E}}^{(0)} $ correctly quantifies both volume-invariant deformations (which
appear as traceless strain tensors) and dilations (which appear as
the logarithm of the volume, as, for example, in gases).
Constitutive equations of a material, that is stress-strain relations, can use either of these tensors; some arguments favor right tensors \cite{farahani,larson} while other favor left ones \cite{macosko}.
\subsubsection{Micromechanical definitions of elastic strain}
A microscopic definition of elastic strain
\cite{kr96,liao,alexander,ball,blumenfeld,kruyt,gold3,zimmerman,aubouy}
adopts a different point of view: it constructs the macroscopic
elastic strain from
the positions of individual ``sites" ({\it i.e.} atoms, grains, or
vertices of a network), just as stress is constructed
from individual forces, see eq. (\ref{sigma gene}).
For instance, Zimmerman considers elastic deformations of 2D or 3D
ordered atomic crystals
\cite{zimmerman}. He constructs separately the stress (from atomistic
potentials) and the elastic strain (from displacements). He shows
that, in the ordered cases he considers, his strain yields a
continuum-mechanics limit identical to
that derived from the Cauchy-Green tensor.
Kruyt (2003) expands the analysis he used in a series of papers with
Rothenburg, which, for instance, predicted the elastic moduli of
disordered or ordered 2D granular materials \cite{kr02}.
He constructs separately the stress from individual forces (obeying
equilibrium conditions) and the elastic strain from individual
displacements (obeying geometric compatibility conditions) small
enough to avoid rearrangements \cite{kruyt}.
Goldhirsch and Goldenberg (2002) rewrite elasticity theory on a microscopic
basis, take the continuous limit, and show its compatibility with
classical elasticity, including the energetics and conjugation
relations \cite{gold3}. In particular, they show that their
definition of elastic strain is better than those based, for instance,
on fits of the displacements to a macroscopic strain field
\cite{liao} and that, even for granular materials, microscopic
fluctuations progressively average out when going to macroscopic
scales.
\subsection{Extension to plastic and fluid regimes}
Macroscopic and micromechanic elastic strain tensors become difficult to define after the first topological
rearrangement. They finally lose their meaning after a few rotations of the
Couette-cell inner wheel, and mathematical singularities appear.
This breakdown of formalism is counterintuitive; experimentally, the mechanical behaviors of elasto-plasto-viscous materials do not appear to change discountinuously.
We need to treat the (even large) elastic deformation, independently from the flow involving
rearrangements
of the connections between neighbouring sites (changes in network topology).
The total applied shear strain decomposes into plastic and elastic strains.
The plastic strain is much larger, since it increases linearly
with time (its derivative, corresponding to the velocity gradient, is
constant in steady flow). The elastic strain is much smaller:
here, in steady flow, it remains constant; but it is physically
important, because it characterises the current state of the foam.
We use below the approach of
Aubouy {\it et al.} (2003), which consists in three steps. It first characterises the current state of the material by a
``texture" (or ``fabric") tensor $\overline{\overline{M}}$, built
from statistical averages over microscopic positions, which exists in all regimes. It then defines a reference state characterised by a reference tensor $\overline{\overline{M}}_0$. It finally
constructs the elastic
strain by a comparison of both tensors.
We now review these definitions.
\subsubsection{Texture tensor} \label{def_text}
The pattern of the walls in a region of the foam can be statistically
characterised by the local {\it texture tensor}
$\overline{\overline{M}}$ \cite{aubouy}. In 2D, it writes
(again with averages weighted by $f$):
\begin{eqnarray}
\overline{\overline{M}}
&=& \left\langle \vec{\ell}\otimes \vec{\ell} \right\rangle_{{V}}
,\nonumber\\
M_{ij}&=&
\langle
\ell_{i}\ell_{j}
\rangle_{{V}}
= \left\langle \left(
\begin{array}{ll}
\ell_r^2 &\ell_\theta \ell_r\\
\ell_r\ell_\theta & \ell_\theta^2
\end{array}
\right)\right\rangle_{{V}} .
\label{defM}
\end{eqnarray}
It accounts for
the direction and length $\ell$ of the walls;
however, since the symmetry $\vec{\ell} \to -\vec{\ell}$ does not affect the
pattern,
the orientation of $\vec{\ell}$ plays no role.
The diagonal components $M_{rr}$ and $M_{\theta\theta}$ of this tensor are
both of
order $\left\langle \ell ^{2}\right\rangle $. Conversely, the
off-diagonal component
$M_{r\theta}=M_{\theta r}$ (the tensor is symmetric), and the
difference $M_{rr} - M_{\theta\theta}$, together characterise the
anisotropy: they
are both much smaller than $\left\langle \ell ^{2}\right\rangle $. They
vanish
when the foam is isotropic; in which case $\overline{\overline{M}}
= (\langle \ell^2 \rangle /2) \overline{\overline{I}}$ is
proportional to
$ \overline{\overline{I}}$, and only
encodes the r.m.s. wall length.
Hence the tensor $\overline{\overline{M}} $ has two strictly positive
eigenvalues, also of
order $\left\langle \ell ^{2}\right\rangle $, the
larger being in
the direction in which bubbles elongate.
It thus quantifies the average size, as well as the
anisotropy direction and amplitude, of the network consisting of the
bubble vertices (Fig. \ref{FigPhoto}a, top right).
In a stationary regime, it is constant, as are all averaged physical quantities.
\subsubsection{Reference state} \label{reference}
At any time, the current state of the foam corresponds to
a current reference state, namely
the state that the foam reaches after relaxation
\cite{porte}.
This reference state defines
the {\it reference texture tensor }
$\overline{\overline{M}}_0$.
It varies with time;
but in a stationary regime, it is constant, as are all averaged physical quantities
including $\overline{\overline{M}}$.
The reference state is not necessarily equal to the initial state \cite{porte}.
In the present experiment, letting the inner wheel rotate back freely is difficult,
so that in practice we do not
have direct access to a reference state. But we do not need microscopical details
to estimate $\overline{\overline{M}}_0$:
it suffices to make a physical approximation on the average properties of the foam at rest.
We chose below to assume that the reference state is statistically isotropic.
We thus approximate
$\overline{\overline{M}}_0$ by
$ M_0 \overline{\overline{I}}$, where
the constant
$M_0 = {\rm Tr} \left(\overline{\overline{M}}_0\right) / 2$ is half the average over all boxes of the squared wall length.
We implicitly assume that the volume is conserved, which should be
correct in Couette flow. More precisely, to avoid
overestimating $M_0$ due to bubble stretching, we take the average
between $\langle \ell \rangle^2/2$ and $ \langle \ell^2 \rangle/2$;
since here $\langle \ell \rangle=1.40$ mm and $\sqrt{\langle \ell^2
\rangle}=1.45$ mm, we take $M_0=1.0$ mm$^2$.
\subsubsection{Elastic strain tensor}
We can operationally measure the {\it statistical elastic strain tensor } $\overline{\overline{U}}$ in each box
from \cite{aubouy}:
\begin{equation}
\overline{\overline{U}}(\vec r) =
\frac{\log{\overline{\overline{M}}}(\vec r) -
\log{ \overline{\overline{M}}_0 }}{2}.
\label{U log}
\end{equation}
It obeys \cite{aubouy}
the mathematical requirements for a strain tensor: symmetry
properties under translation, rotation and index permutation.
It
is always defined: $\log{\overline{\overline{M}}}$ is the tensor
with the same eigenvectors as $\overline{\overline{M}}$, but with the
logarithm of its (strictly positive) eigenvalues.
It is the only dimensionless function of $\overline{\overline{M}}$
and $\overline{\overline{M}}_0$ which is always defined (there is no equivalent of the division for tensors, except when they commute).
It also obeys the physical requirements for an elastic strain tensor. It statistically quantifies
the deformation reversibly stored in the present state of the
material. Its differential $d\overline{\overline{U}}$ (and hence the
elastic moduli for infinitesimal deformation around the current state)
depends explicitly only on the current state.
Moreover, let us consider the case where the (small or large) deformation is elastic
and affine. Then:
\begin{eqnarray}
\vec{\ell} &=& \overline{\overline{F}} \; \vec{\ell}_0,\nonumber \\
\vec{\ell} \otimes \vec{\ell} &=&
\vec{\ell} \; \vec{\ell}^t = \overline{\overline{F}} \; \vec{\ell}_0 \;
\vec{\ell}_0^t \; \overline{\overline{F}}^t
=
\overline{\overline{F}} \;
\left( \vec{\ell}_0 \otimes \vec{\ell}_0 \right) \; \overline{\overline{F}}^t ,
\nonumber \\
\overline{\overline{M}}&= &\left\langle \vec{\ell} \otimes \vec{\ell} \right\rangle
= \overline{\overline{F}} \;
\left\langle \vec{\ell}_0 \otimes \vec{\ell}_0 \right\rangle \; \overline{\overline{F}}^t
=
\; \overline{\overline{F}} \; \overline{\overline{M}}_0 \;
\overline{\overline{F}}^t
.
\end{eqnarray}
If we now assume that $\overline{\overline{M}}_0$ commutes with
$\overline{\overline{F}}$, which is the case if for instance
$\overline{\overline{M}}_0$ is isotropic, then
$\overline{\overline{M}}
= \overline{\overline{M}}_0 \; \overline{\overline{F}} \;
\overline{\overline{F}}^t$.
We then prove that $\overline{\overline{U}}$
coincides with the true strain
\cite{hoger,zimmerman}:
\begin{equation}
\overline{\overline{U}} =
\frac{\log{\overline{\overline{M}}} -
\log{\overline{\overline{M}}_0} }{2} =
\frac{\log{
\left(
\overline{\overline{M}}_0 \; \overline{\overline{F}} \;
\overline{\overline{F}}^t
\right)
} -
\log{\overline{\overline{M}}_0} }{2} =
\frac{1}{2}\;
\log{ \left(\overline{\overline{F}} \; \overline{\overline{F}}^t \right) }
= \overline{\overline{E}}^{(0)} ,
\end{equation}
see eq. (\ref{logstrain}).
The Finger tensor appears
because we base our elastic description on a tensor,
$\vec{\ell} \otimes \vec{\ell} =
\vec{\ell} \; \; \vec{\ell}^t $. This constrast with
classical elasticity, based on a scalar product $d\vec{x} \cdot d\vec{x} =
d\vec{x}^t \; d\vec{x} = d\vec{x}_0^t \; \overline{\overline{F}}^t \; \overline{\overline{F}} \; d\vec{x}_0 $, where the Cauchy-Green tensor appears
\cite{phan}.
A specific feature of the quasistatic regime is that the deformation (and
hence the elastic strain) remains small, barely visible by eye (Fig.
\ref{images}), even for large applied shear strain (and hence large
plastic strain). We thus assume, and check {\it a posteriori}
(Fig. \ref{U(r)}), that the components of
$\overline{\overline{U}}$ are much smaller than one, so we approximate
eq. (\ref{U log}) by its linearisation:
\begin{equation}
\overline{\overline{U}}(\vec r) \approx \frac{1}{2}\left(
\frac{\overline{\overline{M}}(\vec r)}{M_0}-
\overline{\overline{I}}
\right).
\label{def U}
\end{equation}
\section{Results} \label{results}
\subsection{Transient regime}\label{transient}
All data in this section are taken from Run 1 (Fig. \ref{images}),
which has two distinct regimes: a transient
and a stationary flow (Fig. \ref{FigTransi}).
\begin{figure}
\centering
\includegraphics[width=\linewidth]{fig3transitoire2.eps}
\caption{Run 1 (transient regime):
texture tensor $\overline{\overline{M}} $
{\it versus} the applied shear strain $\gamma$.
The components
of $\overline{\overline{M}} $
defined in eq. (\protect\ref{defM}) and measured
from the first 200 images
($\blacktriangle$: $M_{rr}$; $\square$: $M_{\theta\theta}$;
$\bullet$: $M_{r\theta}$)
are expressed in units of the average length of the bubble
wall, $ \langle \ell^2 \rangle$.
(a) Innermost box (number 1):
each point represents one image;
% (17 bubbles);
(b) outermost box (number 8):
for clarity, we show only every second image.
% , 32 bubbles.
The vertical dashed line marks the cross-over $\gamma_Y$ between the transient and
stationary regimes, defined by the intersection between the
straight line and the zero slope line fitting $M_{r\theta}$ in the transient
and stationary regimes, respectively (solid lines in (a)).
}
\label{FigTransi}
\vspace{1cm}
\end{figure}
\subsubsection{Texture tensor}
In the {\it transient regime}, the average wall length does not
change
much, but the walls tend to
align with the direction of the rotation (Fig. \ref{images}). Hence,
it is mostly the cross-component $M_{r\theta}$ which correlates
to the shear (Fig. \ref{FigTransi}). Since $\gamma=0$ marks the
change from clockwise to
counter-clockwise shear,
$M_{r\theta}$ decreases linearly
and changes sign.
The correlation with shear of both other components, if any, is of order of their fluctuations.
\begin{figure}%[!thb]
\center
\includegraphics[width=9cm]{fig4mrt-lin.eps}
\caption{Run 1 (transient regime):
cross component $U_{r \theta}$ {\it versus} the applied shear strain
$\gamma$, for box 1: $\bullet$, box 3: $\blacksquare$, box 5: $\triangle$, and box 7: $\times$.
Since $\gamma=0$ marks the reversal of the rotation
(and not an initially unstrained state), we expect $U_{r \theta}$ to be zero when $\gamma=0.15$,
within experimental error ($\pm$ 0.03).
A linear fit of $U_{r \theta}$ {\it versus} $\gamma$ yields the derivative $d U_{r
\theta}/d\gamma$, plotted in Insert, where
the line is a linear fit with zero intercept to $d U_{r
\theta}/d\gamma$
{\it versus} $1/r^2$.
}
\label{Urt lin}
\vspace{1cm}
\end{figure}
\subsubsection{Elastic strain tensor}
During the transient, no bubble rearrangement (neighbour swapping, also called ``T1
processes'') is visible for $\gamma < 0.3 $. This probably
corresponds to the elastic regime \cite{hohler}.
The cross-component $U_{r \theta}$ varies linearly with $\gamma$, and
a linear fit yields its derivative $d U_{r \theta}/d\gamma$
(Fig. \ref{Urt lin}).
It varies smoothly with $r$, and a fit to $r^{-2}$
yields the prefactor $(- 1600 \pm 90)\: \left\langle\ell\right\rangle^2$.
It is interesting to compare $d U_{r \theta}/d\gamma$
with the expression which would be predicted by classical elasticity in a
Couette geometry:
\begin{equation}
%%u_{r\theta} = \frac{R_1^2 R_0}{R_1+R_0} \; \frac{1}{r^2},
%u_{r\theta} = \gamma \; \frac{r_1^2 r_0}{r_1+r_0} \; \frac{1}{r^2},
%\frac{du_{r\theta} }{d\gamma} = \frac{r_1^2 r_0}{r_1+r_0} \; \frac{1}{r^2}.
\frac{dU_{r\theta} }{d\gamma} = \frac{r_1^2 r_0}{r_1+r_0} \; \frac{1}{r^2}.
\label{u couette}
\end{equation}
The prefactor of eq. (\ref{u couette}) is
%$ (R_1^2 R_0)/[(R_1+R_0)] = (- 1680 \pm 80)\: \left\langle\ell\right\rangle^2 -$,
$ (r_1^2 r_0)/[(r_1+r_0)] = (- 1680 \pm 80)\: \left\langle\ell\right\rangle^2 $,
where we have taken 1.2 mm (the depth of the teeth) as the uncertainty of the wheel radius.
This agrees with the experimental prefactor of $r^{-2}$
(Fig. \ref{Urt lin}).
\subsubsection{Yield }\label{yield}
\begin{figure}%[!thb]
\center
\includegraphics[width=8cm]{fig5gammay2.eps}
\caption{Run 1 (transient regime): variation of $\gamma_{Y}$
{\it versus} box number. Data are from Fig. (\protect\ref{FigTransi}).
The line is a linear fit.
}
\label{Gamma s}
\vspace{1cm}
\end{figure}
The flow becomes {\it stationary } at a
value $\gamma_Y$ which we measure
by determining the intersection between the transient build-up of $M_{r
\theta}$ fitted by a straight line and the
stationary plateau fitted by a horizontal line (Fig. \ref{FigTransi}).
%For simplicity, we call $\gamma_Y$ the
%``yield shear strain".
It would be interesting to determine whether $\gamma_Y$ actually marks
the onset of
irreversibility in the stress-strain
relation at macroscopic scale (plasticity).
We can at least observe on the movie (data not shown) the
``T1"
bubble rearrangements, which concern local topology.
In the inner boxes,
$\gamma_Y$
correlates with the appearance of a large number of T1s.
The first isolated T1s
appear at values of $\gamma$ between 0.26 and 0.32, of the
same order of magnitude as, but significantly smaller than $\gamma_Y$.
The outer boxes
lack T1s, except once near $\gamma \sim 3$, well beyond $\gamma_Y$ and in the stationary regime.
Though
$\gamma_Y$
is well defined, it marks a saturation in the local shear strain applied to the bubbles,
which probably oscillates and remains below the actual yield strain.
The value of $\gamma_Y$
increases gradually
from
0.35 for the inner box, to 0.55 for the outer box
(Fig. \ref{FigTransi}). Within experimental error ($\pm$ 0.03, corresponding to $\pm$ 2 images),
this variation is linear (Fig. \ref{Gamma s}).
Explaining this variation of $\gamma_Y$ with $r$ would certainly
help us to understand the yielding of foams (see section \ref{disc yield}).
\begin{figure}
%vspace{8cm}
\begin{center}
\includegraphics[width=9cm]{fig6Mbox.eps}
\end{center}
\caption{
Run 2 (stationary flow): texture tensor
$\overline{\overline{M}}_{stat} $ {\it versus} box number.
The components of $\overline{\overline{M}} $
($\blacktriangle$: $M_{rr}$; $\square$: $M_{\theta\theta}$;
$\bullet$: $M_{r\theta}$),
again expressed in units of $ \langle \ell^2 \rangle$, are
averaged over the four quarters of the whole foam,
and also over time (2000 pictures).
}
\label{FigPerm}
\vspace{1cm}
\end{figure}
\subsection{Stationary regime - average values}\label{permanent}
All data of this section are taken from Run 2 (Fig. \ref{FigPhoto}),
which allows us to improve our statistics, using temporal averages over
the stationary flow.
\subsubsection{Texture tensor}
Fig. (\ref{FigPerm}) plots the average $M_{\mathrm{stat}}$ of
the instantaneous texture tensor $\overline{\overline{M}}$ in the
stationary regime.
We can now detect, not only the (small) anisotropy of the foam,
$M_{r\theta}$, but also
its small spatial variations.
$M_{r\theta}$ decreases across the gap: it is larger in the inner
boxes than in the outer ones,
reflecting the preferred
orientation imposed by the shear.
$M_{rr}$ decreases,
$M_{\theta\theta}$ increases, and
$M_{rr} - M_{\theta\theta}$ changes sign; this sign-change is more
difficult to understand (section
\ref{nonzero}).
\subsubsection{ Elastic strain}
\begin{figure}
\center
\includegraphics[width=9cm]{fig7Ur.eps}
\caption{
Run 2 (stationary flow): tensor $\overline{\overline{U}} $ {\it
versus} box number.
The components of $\overline{\overline{U}} $
($\blacktriangle$: $U_{rr}$; $\square$: $U_{\theta\theta}$;
$\bullet$: $U_{r\theta}$)
are derived from the data in Fig. (\protect\ref{FigPerm})
through eq.
(\protect\ref{def U}), using the value $M_0=1.0$ mm$^2$ (see section
\protect\ref{reference}).
The solid line is a best fit of $U_{r\theta}$ to a function
cst$/r^2$.
%$ here cst$= (410 \pm 7 ) \langle \ell \rangle ^2 $.
}
\label{U(r)}
\vspace{1cm}
\end{figure}
$U_{r\theta}$ does not localise
near the inner wheel: it varies smoothly with the distance to the
inner wheel. It decreases as $\sim 1/r^{2}$ (Fig. \ref{U(r)}),
as expected if
the foam behaves like a continuous elastic medium, see eq. (\ref{u couette}).
To compare to the transient regime in Run 1
(which has slightly different bubble and wheel diameters),
we express the prefactor determined in Fig. (\ref{U(r)}) in the
same units as in eq. (\ref{u couette}):
\begin{equation}
%\langle U_{r\theta} \rangle= \gamma_{el} \frac{R_1^2 R_0}{R_1+R_0}
\langle U_{r\theta} \rangle= \gamma_{el} \frac{r_1^2 r_0}{r_1+r_0}
\frac{1}{r^2}.
\label{gamma elast}
\end{equation}
The prefactor is $\gamma_{el} = 0.171 \pm 0.003 $,
see section (\ref{disc yield}).
\begin{figure}
\center
\includegraphics[width=9cm]{fig8div-stress.eps}
\caption{
Run 2 (stationary flow): elastic stress tensor
$\overline{\overline{\sigma}}^{cap} $ {\it versus} box number.
The deviatoric components of $\overline{\overline{\sigma}}^{cap} $
($\bullet$: shear stress $\sigma^{cap}_{r\theta}$;
$\square$: normal stress difference
$\sigma^{cap}_{rr}-\sigma^{cap}_{\theta\theta}$)
are derived from eq.
(\protect\ref{sigma cap}) and expressed in units of $\lambda/\langle
\ell \rangle$.
The solid line is a best fit of $\sigma^{cap}_{r\theta}$ to a function
cst$/r^2$.
}
\label{div stress}
\vspace{1cm}
\end{figure}
\subsubsection{Deviatoric elastic stress} \label{deviatoric}
In the quasistatic regime, mechanical equilibrium implies that the
elastic stress is divergence free.
Hence we expect that its off-diagonal component, which reduces to the capillary
contribution, varies as for a continuous elastic medium in a Couette geometry
\cite{dennin}:
$ \sigma^{cap}_{r\theta} \sim {\mathrm
cst}/r^2$ . Such $r^{-2}$ dependence occurs experimentally
(Fig. \ref{div stress}).
Again, the spatial variation of $\sigma^{cap}_{rr} -
\sigma^{cap}_{\theta\theta}$, which changes sign,
is more difficult to understand (section
\ref{nonzero}).
\subsubsection{Non-dimensional shear modulus} \label{hooke}
\begin{figure}%[!thb]
\center
\includegraphics[width=9cm]{fig9module3.eps}
% \includegraphics[width=9cm]{fig9module2.eps}
\caption{Run 2 (stationary flow):
Hooke relation.
The deviatoric part of the capillary stress
$\overline{\overline{\sigma^{cap}}} $, expressed in units of
$\lambda/\langle \ell \rangle$,
plotted {\it versus} the corresponding components of
$\overline{\overline{U}} $
($\bullet$: $\sigma^{cap}_{r\theta}$ {\it versus}
$U_{r\theta}$; $\square$:
$\sigma^{cap}_{rr}-\sigma^{cap}_{\theta\theta}$ {\it versus}
$U_{rr}-U_{\theta\theta}$).
Each data point comes from Figs.
(\protect\ref{U(r)}) and (\protect\ref{div stress}),
{\it i.e.} it derives from averages at one position in the foam.
The solid line is a linear fit through all points, slope
$2\mu \langle \ell \rangle/\lambda= 1.17 \pm 0.04$.
}
\label{Hook}
\vspace{1cm}
\end{figure}
In elasticity, the stress and the deformation are conceptually
independent quantities (see section \ref{difference}). In a
continuous medium, they
correlate physically; their relation is the elastic part of the material's
constitutive relation \cite{landau}.
Let us focus on the components we measure here, namely the
deviatoric terms $rr-\theta\theta$ and $r\theta$. If the material is
linear, the components of strain and stress are proportional to each
other, the slope being twice the shear modulus $\mu$. If the material
is isotropic, $rr-\theta\theta$ and $r\theta$ play a similar role.
In polar coordinates, the Hooke equations becomes:
\begin{eqnarray}
\sigma^{cap}_{r \theta} &=& 2 \mu U_{r \theta} ,\\
(\sigma^{cap}_{rr}-\sigma^{cap}_{\theta \theta}) &=& 2 \mu
(U_{rr}- U_{\theta
\theta}).
\label{loi Hook}
\end{eqnarray}
Fig. (\ref{Hook}) indicates that, in the quasistatic regime, the foam
indeed behaves like a continuous medium:
the deviatoric part of the capillary stress
$\overline{\overline{\sigma^{cap}}} $ and of $\overline{\overline{U}}
$ correlate strongly, and details of the
microstructure appear only through mesoscopic averages.
Moreover, the foam is both isotropic and linear, with a shear
modulus $\mu= (0.59 \pm 0.02) \lambda / \langle \ell \rangle$. With a
smaller or a larger box size, we obtain the same value for $\mu$ with a
larger error.
\subsection{Permanent regime - temporal fluctuations}\label{fluctuat}
\subsubsection{Time series of $\overline{\overline{M}}$ in Run 1}
The stationary flow results from a balance between the increase due to
the applied shear strain and relaxation due to bubble
rearrangements \cite{langer}. The texture tensor and the elastic strain
fluctuate
around their mean value.
In Run 1, the short-term fluctuations are much larger near the inner
wheel (Fig.
\ref{FigTransi}a), where
shape deformations and bubble rearrangements are both larger
\cite{debregeas1},
than near the
outer wheel (Fig. \ref{FigTransi}b).
Fig. (\ref{FigTransi}) does not show any significant physical
feature of the fluctuation statistics, {\it e.g.} abrupt, large-scale
relaxations of deformations {\it via}
correlated, large stress drops.
\subsubsection{Histograms of increments in Run 2} \label{sec_histo}
\begin{figure}
\center
\includegraphics[width=0.9\linewidth]{fig10histo.eps}
\caption{
Run 2 (stationary flow): statistics for the increments of the
texture tensor $\overline{\overline{M}} $.
(a) Histogram of increments of $M_{rr}$
in box 1, again expressed in units of $ \langle \ell^2 \rangle$. Solid line
is a Gaussian
with the same standard deviation and zero average.
(b, c) The same for
$M_{r \theta}$ and
$M_{\theta \theta}$.
(d) Standard deviation of increments of $\overline{\overline{M}}$
{\it versus} box number ($\blacktriangle$: $M_{rr}$; $\square$: $M_{\theta\theta}$; $\bullet$: $M_{r\theta}$).
Since the fluctuations in box 8 are extremely small (see
Fig. \protect\ref{FigTransi}), the data are noisy: image filtering
and thresholding would have been required to obtain a
statistically significant
value.
}
\label{fig_histo}
\vspace{1cm}
\end{figure}
We perform a more quantitative analysis on Run 2, at a small scale, namely
between two successive images, corresponding to an increase of $d\gamma$
in the applied shear strain.
The increment of the texture tensor is defined as
$\overline{\overline{M}}(\gamma +d\gamma)-\overline{\overline{M}}(\gamma)$.
The signature of ``avalanche-like" events at such a small scale would
be extremely large fluctuations and/or a large asymmetry between
positive and negative increments.
We find no such signature.
The increments are almost Gaussian (Fig. \ref{fig_histo}a-c).
The width of the histograms decreases by a factor of 10 from inner to
outer boxes (Fig. \ref{fig_histo}d): as we mentioned, most large
fluctuations occur near the inner wheel.
Only $M_{r\theta}$ displays a slight asymmetry between positive and
negative increments
(Fig. \ref{fig_histo}b), probably because this cross-component is
increased by the applied shear strain and decreased by the
anisotropy relaxation due to T1 events.
\subsubsection{Temporal autocorrelations in Run 2}
\begin{figure}
\center
\includegraphics[width=\linewidth]{fig11auto-impaire.eps}
\caption{
Run 2 (stationary flow): temporal
autocorrelation $ C_{MM}$ of texture tensor
$\overline{\overline{M}}$ (eq. \protect\ref{def auto}), {\it
versus} $\delta{\gamma}$, in boxes 1, 3, 5, 7
($\blacktriangle$: $M_{rr}$; $\square$: $M_{\theta\theta}$;
$\bullet$: $M_{r\theta}$).
We recall that the total applied shear $\gamma$ is proportional to time,
not to the local shear in each box (eq. \ref{prop_time}).
}
\label{all autocor}
\vspace{1cm}
\end{figure}
The rotation of the inner wheel does not perturb the foam near the outer wheel.
In the outer boxes, the texture tensor thus decorrelates very slowly. Let us quantify this observation
using the data of Run 2.
To analyse the fluctuations at all scales of the shear strain $\delta
\gamma$, we calculate the temporal autocorrelation $ C_{MM}$ of the fluctuations:
\begin{equation}
C_{MM}( \delta{\gamma})= \frac{
\left\langle
\left(M(\gamma)-M_{\mathrm{stat}}\right)
\left(M(\gamma+ \delta{\gamma})-M_{\mathrm{stat}}\right)
\right\rangle}{
\left\langle
\left(M(\gamma)-M_{\mathrm{stat}}\right)^{2}
\right\rangle},
\label{def auto}
\end{equation}
where
$M(\gamma)$ (resp: $M_{\mathrm{stat}}$)
is the instantaneous value (resp: the average over the stationary regime)
of the texture tensor $\overline{\overline{M}}$.
\begin{figure}
\center
\includegraphics[width=0.6\linewidth]{fig12autocorMrt.eps}
\caption{
Run 2 (stationary flow):
autocorrelation $C_{M_{r\theta}M_{r\theta}}$ of the cross-component
$M_{r\theta}$.
Box 1: $\boxplus$, box 2: $\bigcirc$, box 3: $\blacksquare$,
box 4: $ \triangle$, box 5: $\blacktriangle$, box 6: $\bullet$,
box 7: $\square$, box 8: $\Diamond$.
Insert:
initial decorrelation rate $\gamma_{\mathrm{decorr}}^{-1}$
for each box (eq. \protect\ref{def initial}).
}
\label{autocor-M}
\vspace{1cm}
\end{figure}
Fig. (\ref{all autocor}) displays a rapid initial decay in the correlation,
followed by a slower decay,
reflecting the destruction of the initial conditions \cite{durian,lauridsen,ybert,pratt}.
An exponential fit at small $\delta{\gamma}$ yields the
shear strain $\gamma_{\mathrm{decorr}}$ characterising the
initial
loss of correlation, especially quick for the cross
component $M_{r \theta}$ (Fig.
\ref{autocor-M}):
\begin{equation}
M_{r \theta} = \exp \left( -
\frac{\delta \gamma}{
\gamma_{\mathrm{decorr}}}\right).
\label{def initial}
\end{equation}
As we expect, the correlation decays much more quickly near the inner
wheel, $\gamma_{\mathrm{decorr}}\approx 1/30 $, than near the outer
one, $\gamma_{\mathrm{decorr}}\approx 1 $ (insert of Fig.
\ref{autocor-M}).
The decorrelation rate $\gamma_{\mathrm{decorr}}^{-1}$ decreases exponentially
with the distance from the inner wheel.
The fluctuations of the
texture tensor are thus localised
near the inner wheel, over a distance comparable to 4 bubble diameters.
This localisation resembles that of the
velocity field over 1 bubble diameter
\cite{debregeas1}, and is probably the signature of the
effect of T1s on the deformation \cite{dennin}.
\section{Discussion}\label{discu}
\subsection{Discussion of the strain and stress}\label{disc_stress_strain}
%We now turn to a discussion of the measurements.
\subsubsection{Measurement method of the stress }
Let us discuss
the contributions to the stress we do not measure.
The dissipation is probably dominated by the
friction on both glass plates, which we expect to be much larger than the
internal viscous stress \cite{cantat}.
Whatever its physical origin, it
should on average balance
the input power. It is thus never negligible, especially during the
relaxation after each T1, see section (\ref{qstat}).
However, the results we present here, which concern only the elastic
terms, do not require us to measure the dissipation.
We measure the deviatoric elastic stress using eq. (\ref{sigma
cap}), which is formally equivalent to the stress in a network where sites
interact through a two-body force. Foams do not include only
two-vertex interactions. Their stability is due to the
pressure inside each bubble, which relates to the bubble volume.
Hence pressure is a function of all wall curvatures and all the vertex
positions. It is an $n$-body interaction, which we can decompose
into two-point interactions in principle \cite{alexander}, but not
really in practice.
It is not important here to measure the pressure inside bubbles, for the following reasons \cite{wuppertal}.
Since the pressure stress is
isotropic, it does not contribute to
the deviatoric elastic stress. Hence the shear modulus is independent of the
pressure field.
Finally, the compression modulus of a foam is typically that of
an ideal gas, which scales like the pressure inside bubbles $\sim 10^5$ Pa
(multiplied by $h$, when expressed as a 2D compression modulus).
This modulus is not very interesting and is much larger
than the shear modulus of the foam, which is thus effectively incompressible
\cite{hutzler}.
%Pressure differences between bubbles curve the bubble walls.
%As we mentioned in section (\ref{operational}), in the quasistatic
%regime the in-plane curvature radii
%induce only a negligible correction to the deviatoric capillary
%stress for
%straight walls.
%The out-of-plane curvature (in the third dimension) also relates to
%pressure differences, and we neglect its effect on the deviatoric
%components of a dry foam (section \ref{nonzero}).
The line tension is not known, but unimportant for our present
purposes. The stress, and hence the shear
modulus, are expressed in non-dimensional form, as
required for comparison to other foams. A direct measurement of $\lambda$
can complement
the local image
measurement
when necessary, for instance to
predict the actual
force exerted by
the flowing foam on an obstacle, or to validate
eq. (\ref{sigma cap}) by checking
its agreement with an independent,
macroscopic measurement of the force \cite{langmuir_foam}.
\subsubsection{Difference between
$\overline{\overline{\sigma}}^{cap}$ and $\overline{\overline{M}}
$} \label{difference}
Both $\overline{\overline{\sigma}}^{cap}$ and $\overline{\overline{M}} $
are measured from the same image, and their definitions
(eqs. \ref{sigma cap}, \ref{defM})
seem similar. However, they are really independent
both physically and mathematically.
Physically, their significance and status are different.
The value of $\overline{\overline{\sigma}}^{cap}$ can be
determined from image analysis only if the sites interact through a
two-body force $\vec{\tau}$, and if we explicitly know the
form of this force $ \vec{\tau}(\vec{\ell})$ with respect to
the distance between sites.
The value of $\overline{\overline{\sigma}}^{cap}$ is sensitive to physical forces:
first, it doubles if the
density of links doubles; second, it does not change if we cut a link in two by artificially
introducing a new site at its middle; third, in foams,
it formally depends on wall curvatures.
On the other
hand,
$\overline{\overline{M}} $ is purely geometric.
It is a characteristic of the pattern, independent of physical
interactions; it can be defined and measured for various kinds of
cellular patterns for which no $\vec{\tau}$ can be defined, including
solid foams, and even grain boundaries in crystals
\cite{durand_weiss}.
The value of $\overline{\overline{M}} $ is sensitive to the detailed network
topology:
first, it does not change if the
density of walls doubles; second, it decreases if we cut a wall in two by artificially
introducing a new vertex at its middle; third, in foams, it is insensitive to the (in-plane or
out-of-plane) curvature of walls.
Mathematically, $\overline{\overline{\sigma}}^{cap}$ and
$\overline{\overline{M}} $
are independent in all cases, except when $\vec{\tau} \propto
\vec{\ell}$, {\it i.e.} a harmonic spring, with a length
$\ell$ much greater than its length at equilibrium. For instance, writing
the stress
of an
isolated polymer molecule in a shear solvant as $\vec{\ell} \otimes \vec{\ell}$
is legitimate \cite{bird}.
However, foams, for which $
\tau = $ const, fall in the general case where
$\overline{\overline{\sigma}}^{cap}$ and $\overline{\overline{M}}
$ are mathematically different averages.
The difference is even more visible in a 3D foam, where
$\overline{\overline{\sigma}}^{cap}$
is an integral over a surface \cite{batchelor}, while
$\overline{\overline{M}} $ remains based on vectors
\cite{aubouy}.
Moreover, $\overline{\overline{M}} $ does not even need be defined on the vectors $\vec{\ell}$ linking neighbour vertices. Definitions other than eq. (\ref{defM}) could be acceptable in principle \cite{dollet}. For instance, we could use the vector $\vec{g}$ which links the centers of mass of neighbouring bubbles, and construct $\overline{\overline{M}}$ as $
\left\langle f\; \vec{g}\otimes \vec{g} \right\rangle$. This would apply to bubble rafts \cite{dennin}, granular materials \cite{kruyt}, or 3D foams \cite{kraynik2003}.
This would
change both $\overline{\overline{M}} $
and $\overline{\overline{M}}_0 $ by a multiplicative constant:
the elastic strain $\overline{\overline{U}} $
(eqs. \ref{U log}, \ref{def U}),
and hence the shear modulus $\mu$ (eq. \ref{loi Hook})
would almost not change.
We emphasise that $\overline{\overline{\sigma}}^{cap}$ and
$\overline{\overline{U}} $ are also independent.
They represent different averages of
microscopic details, so we cannot express one as a
function of the other \cite{kruyt,gold3}, except empirically,
through numerical simulations
or physical approximations
approximations
\cite{kruyt}.
The difference is even more visible
at large deformations:
the logarithmic term of eq. (\ref{U log})
deviates from the linear approximation when the maximum value of the elastic shear
strain (0.15 in Fig. \ref{U(r)}) reaches large values, {\it e.g.}
0.4 \cite{langmuir_foam} or
0.6 \cite{asipauskas}.
There is no trivial relation between them, except in a regular pattern, where averages are replaced by exact identities; for instance, in the case of a honeycomb, their relation is exactly as expected \cite{dollet}.
\subsubsection{Sign of normal differences}\label{nonzero}
The normal differences $M_{rr}-M_{\theta\theta}$
and $\sigma^{cap}_{rr}-\sigma^{cap}_{\theta\theta}$
change sign in the middle of the gap,
between boxes 4 and 5 (Figs. \ref{FigPerm}, \ref {U(r)}, \ref{div stress}).
Normal stresses appear in
elastic solid \cite{poynting} or granular \cite{bagnold} materials sheared at constant volume
(this is equivalent to ``dilatancy": volume increase under shear at constant pressure \cite{dilatancy}).
In complex fluids such as foams, these normal stresses exist but are difficult to measure \cite{khan1988} and to predict \cite{dilatancy}.
Predictions and 2D simulations of their average are accurate: under steady shear, the pressure of a foam with fluid fraction $\Phi= 5\%$ increases by 15\% \cite{dilatancy}. This is the order of magnitude of the spatial variations
of
$M_{rr}$ and
$M_{\theta\theta}$ (Fig. \ref{FigPerm}).
We can not predict
their anisotropy (hence the normal stress differences), but normal stresses due to dilatancy are a second-order effect in the strain \cite{larson}: they are likely smaller that the large value of $M_{rr}-M_{\theta\theta}$ we observe, and
might be masked by dynamical effects due to bubble rearrangements.
But we can imagine other explanations.
First, this change could reflect an initial anisotropy resulting from the
procedure used to fill the cell
with foam \cite{debregeas1},
which the preparatory rotation does not suffice to relax in the outer boxes
(Debr\'egeas, personal communication). The initial state is not necessarily
a zero-stress state.
More generally, trapped stresses make notoriously difficult to measure
normal stress differences
\cite{khan1988}, and carefully relaxed simulations are more reliable \cite{kraynik2003}. The present measurement is probably one of the first direct experimental ones,
along with that of Ref. \cite{labiausse}.
Second, it could be an artefact due to our measurement method, which
does not accounts for the out-of-plane curvature of bubble walls
(in the third dimension). Since the foam is not perfectly dry, the sides
of a wall can have slightly different curvatures, so that their contributions would
not exactly cancel out.
Third, this change could arise from a deviation from axisymmetric Couette flow,
due to a secondary recirculation
with radial velocity components.
A signature of the discrete and disordered nature of the material is the
appearance of solid body rotation
of bubble clusters. It results in vortices,
generic for amorphous elastic materials,
generated by the ``backflow" of the non-affine displacements
\cite{tanguy}.
Their lifetime is as
short as the time the internal wheel takes to move one bubble diameter
\cite{debregeas1}.
\subsection{Discussion of the stationary flow}
The present measurements clarify a few open debates.
\subsubsection{To localise or not to localise?}
The experimental set-up, the fluid fraction, the friction on the plates, and the boundary conditions at
the wheel, are those of Ref. \cite{debregeas1}.
How can the same data alternatively display and not display localisation?
There is no fundamental incompatibility: different fields
have different spatial variations.
Ref. \cite{debregeas1}
measures the velocity field, and the T1 rate; while we measure the elastic
stress and strain fields.
In a quasistatic regime, where the
foam is almost always close to mechanical equilibrium, we expect
the stress to vary smoothly across the gap, as observed here and in
simulations \cite{debregeas2}.
The elastic strain, which correlates to the
elastic stress, does not localise, although the plastic strain,
which correlates to the velocity gradient, does.
\subsubsection{Yield}\label{disc yield}
The value of the yield strain and stress, characterised {\it e.g.} by the
equality of the real and imaginary parts of the complex shear modulus
\cite{asnacios},
is not completely understood yet.
It generally depends on the shear rate
\cite{coussot,rouyer}, but not in the present quasistatic regime.
We expect that, to store a deformation $
\gamma_{el}$, requires us to apply an equivalent shear strain
$\gamma_Y $.
Since Run 1 corresponds to an inversion from counter-clockwise to
clockwise rotation, we expect that $\gamma_Y = 2 \gamma_{el}$.
We observe an excellent agreement
in box 1, in which $\gamma_Y=0.35 \pm 0.02$
(Fig. \ref{FigTransi})
and $\gamma_{el}= 0.171 \pm 0.003$ (Fig. (\ref{U(r)}) and eq.
(\ref{gamma elast})).
But $\gamma_Y \neq 2 \gamma_{el}$ in other boxes, for larger $\gamma_Y$
(Fig. \ref{Gamma s}).
Thus the deformation stored in the foam changes with $\gamma$ even
above
$\gamma_Y$. Kabla and Debr\'egeas
reach the same conclusion: their simulations indicate that, after the
foam yields,
T1s localise in the shear band only after a certain delay time
\cite{debregeas2}.
\subsubsection{Averages and fluctuations}
Our data show that stress and strain have a well defined, physically meaningful average
at the scale of a representative volume element,
while their fluctuations remain small.
A coarse-grained shear modulus can be defined and measured,
unlike {\it e.g.} in granular
media, where the importance of fluctuations
at large scales
is still an open debate \cite{gold4}.
More precisely, the statistics of the texture tensor are nearly Gaussian.
Visual inspection of the images confirms that T1s remain isolated. They
correlate \cite{debregeas2}, but over a very
small space and time range, probably due to the experimental conditions: disordered foam,
quasistatic regime, large dissipation due to top and bottom plates.
We do not observe
extremely large fluctuations, nor sudden relaxations of deformations and stress.
Averages seem to
dominate the physics
Fig. (\ref{FigTransi}), but fluctuations too are important.
For instance, the width of the histogram of the velocity or stress field fluctuations
might constrain the velocity field, and explain why
shear-banding occurs preferentially near the inner wheel, due to a
feedback loop on which fluctuations grow \cite{debregeas2}.
Both the
average
and fluctuations of T1 rate manifest themselves in the texture tensor.
The signature of the
average T1
rate is the existence of a stationary regime, where the decrease in deformation due to T1s balances
the increase due to applied shear.
The signature of T1 rate fluctuations is the
exponential localisation of the texture tensor fluctuations.
\subsubsection{ Measurement of the shear modulus} \label{comments modulus}
Fig. (\ref{Hook}) suggests a few remarks.
First, the classical method to measure a shear modulus consists in increasingly
deforming a sample, and measure the stress and strain in successive states.
We use here another method \cite{asipauskas}.
Each point in Fig. (\ref{Hook}) comes from a different
region of the same foam. We thus observe different deformation states
simultaneously from a single
heterogeneous image.
Averaging over successive images reduces noise \cite{asipauskas}, but is not
necessary \cite{langmuir_foam,durand_weiss}.
Second, when determining the constitutive relation for plastic materials, the elastic stress
is classically plotted {\it versus } the
total applied shear strain: see Refs. \cite{jiangpre,kraynik2000,dilatancy} to quote a few.
Different applied strains correspond to the same stress.
Here we observe a one to one correspondance
between elastic stress and elastic strain (Fig. \ref{Hook}).
The
foam is elastic, although the plastic strain increases steadily, and the foam
is well beyond the yield strain.
Third, although we do not know the value of $\lambda$, we can compare
our non-dimensional measurement of shear modulus $\mu$ with published values.
An ideal two-dimensional dry monodisperse foam, where bubbles form a
regular honeycomb lattice
\cite{princen,khan1986,hutzler}, has a shear modulus
$\mu_{hc}=\frac{\lambda}{\sqrt{3} \langle \ell \rangle}
=0.577\frac{\lambda}{\langle \ell \rangle} $.
Our $\mu$ is slightly larger, probably due to disorder; here $\frac{\langle \ell^2
\rangle}{\langle \ell \rangle^2}=1.08$. The non-zero fluid
fraction tends to decrease $\mu$ \cite{princen,khan1986,mason,dilatancy}.
In a non-quasistatic, and more heterogeneous, stationary flow, either through a
constriction or around an obstacle, a dry foam (fluid fraction much
lower than $5\%$), even with a small bubble size dispersity $<5\%$,
displays a shear modulus 20\% higher than
$\mu_{hc}$, due to
the stretched bubble shape ($\frac{\langle \ell^2 \rangle}{\langle \ell
\rangle^2}=1.22$
\cite{asipauskas}.
\section{Conclusion}
To summarise, we re-analyse the data generously provided by Georges
Debr\'egeas on a foam sheared in a two-dimensional Couette geometry, with
the velocity field localised near the inner wheel
\cite{debregeas1}. Since the flow is quasistatic, the
deformation of the bubbles is very small. We take advantage of the foam's
invariance under rotation
(and, in the stationary regime, of its time invariance) to
average over many bubbles. We then measure the texture tensor of the foam, a general tool to
quantify the deformation of 2D or 3D networks and cellular patterns
\cite{aubouy}.
For small applied shear strain, a transient regime has the
cross-component of the texture tensor increase linearly with shear,
with no bubble rearrangements (``T1"), and the foam seems to
behave like a classical elastic, continuous medium.
The
only large scale manifestation of the discrete, cellular nature of
the foam is that, in
a mixture of
two fluids (soap solution and air),
the surface tension of bubble walls produces
elastic behaviour and a shear modulus.
The deformation saturates
first near the inner wheel, then gradually in more distant boxes,
at an applied shear strain $\gamma_{Y}$.
This onset of plastic behaviour correlates, but does not strictly coincide
with the appearance of T1s.
The foam then enters a stationary regime, where the deformation increase
due to
shear and the relaxation due to T1s balance each other. The
fluctuations of the texture tensor around its mean value
remain small.
Despite the T1s, the elastic behaviour of the foam again resembles
that of a classical elastic, continuous medium. The
maximum shear strain that the foam can sustain in the quasistatic regime
is $\gamma_{el}$, compatible with the value of $\gamma_{Y}$.
The
strain and stress are highly correlated, even proportional
to each other; the shear modulus is slightly higher than for a
honeycomb with similar average bubble wall length.
The deviatoric components of the elastic strain and stress tensors,
as well as the texture tensor, do not localise.
The only trace of localisation is the much quicker loss of correlation of the
texture tensor fluctuations near the inner wheel.
Future studies aim at understanding the origin of the value of the yield strain,
the change of sign in the normal stress and strain differences, the
precise role of boundary conditions, and the relation to granular
materials.
They include the measures of pressure and dissipation, and the
analytical calculation of the shear modulus.
% \acknowledgments
{\bf Acknowledgments:}
We are extremely grateful to G. Debr\'egeas for providing us with
many recordings of his published and unpublished
experiments, and for detailed discussions. We would also like to thank
M. Dennin, R. H\"ohler, A. Kabla, A. Kraynik, N. Kruyt, J. Scheibert for
discussions,
M. Aubouy, B. Dollet, F. Elias and J.A. Glazier for critical reading of the manuscript,
O. Cardoso for help with image analysis. This
work was partially
supported by CNRS ATIP 0693.
\begin{thebibliography}{0}
% \begin{references}
%\bibitem[\ast]{cauthor}
% Author for correspondence. E-mail:
% graner@ujf-grenoble.fr. Fax: (+33) 4 76 51 45 44.
\bibitem[(Alexander 1998)]{alexander}
{\sc Alexander S.}
{1998},
{\it Phys. Reports}, {\bf 296}, {65}.
\bibitem[(Asipauskas {\it et al.} 2003)]{asipauskas}
{\sc Asipauskas M., Aubouy M., Jiang Y., Graner F.
\& Glazier J.A.}
{2003},
{\it Granular Matt.}, {\bf 5}, {71}.
\bibitem[(Aubouy {\it et al.} 2003)]{aubouy}
{\sc Aubouy M., Jiang Y., Glazier J. A. \& Graner F.}
{2003},
{\it Granular Matt.}, {\bf 5}, {67}.
\bibitem[(Bagnold 1941)]{bagnold}
{\sc Bagnold R.A.} {1941},
{\it The Physics of Blown Sand and Desert Dunes}, Methuen, London.
\bibitem[(Ball \& Blumenfeld 2002)]{ball}
{\sc Ball R. \& Blumenfeld R.}
{2002},
{\it Phys. Rev. Lett.}, {\bf 88}, { 115505}.
\bibitem[(Batchelor 1970)]{batchelor}
{\sc Batchelor G.K.} {1970}, {\it J. Fluid Mech.}, {\bf 41}, {545}.
\bibitem[(Bird, Armstrong \& Hassager 1987)]{bird}
{\sc Bird R.B., Armstrong R.C. \& Hassager O.} {1987},
{\it Dynamics of polymeric liquids},
John Wiley \& Sons, NY (USA).
\bibitem[(Blumenfeld 2004)]{blumenfeld}
{\sc Blumenfeld R.}
{2004},
{\it Physica A}, {\bf 336}, { 361}.
\bibitem[(Cantat, Kern \& Delannay 2004)]{cantat}
{\sc Cantat I., Kern N. \& Delannay R.} {2004},
{\it Europhys. Lett.} {\bf 65}, {726}.
\bibitem[(Courty {\it et al.} 2003)]{langmuir_foam}
{\sc Courty S., Dollet B., Elias F., Heinig P. \& Graner F.}
{2003},
{\it Europhys. Lett.} {\bf 64}, {709}.
\bibitem[(Coussot {\it et al.} 2002)]{coussot}
{\sc
Coussot P., Raynaud J.-S., Bertrand F., Moucheront P., Guilbaud
J.-P., Huynh H.T., Jarny S. \& Lesueur D. } {2002},
{\it Phys. Rev. Lett. } {\bf 88}, { 218301}.
\bibitem[(Cox, Weaire \& Vaz 2002)]{Cox}
{\sc Cox S.J., Weaire D., \& Vaz M.F.}
{2002},
{\it Eur. Phys. J. E}, {\bf 7}, {311}.
\bibitem[(Debr\'egeas, Tabuteau \& Di Meglio 2001)]{debregeas1}
{\sc Debr\'egeas G., Tabuteau H. \& Di Meglio J.-M.}
{2001},
{\it Phys. Rev. Lett.}, {\bf 87}, {178305}.
\bibitem[(Dennin 2004)]{dennin}
{\sc Dennin M.} {2004},
%{\it Statistics of Bubble Rearrangements in a Slowly Sheared Two-dimensional Foam}, {\it Phys. Rev. E}, {\bf 70}, { 041406}.
\bibitem[(Dollet {\it et al.} 2004)]{dollet}
{\sc Dollet B., Mecke K., Aubouy A. and Graner F.} {2004},
in preparation.
\bibitem[(Durand, Weiss \& Graner 2004)]{durand_weiss}
{\sc Durand G., Weiss J. \& Graner F.}
{2004},
%{\it Deformation of grain boundaries in polar ice,}
{\it Europhys. Lett.}, {\bf 67}, {1038}.
\bibitem[(Durian 1997)]{durian}
{\sc Durian D. J.}
{1997},
{\it Phys. Rev. E}, {\bf 55}, {1739}.
\bibitem[(Elias {\it et al.} 1999)]{elias}
{\sc Elias F., Flament C., Glazier J. A., Graner F. \&
Jiang Y. }
{1999}, {\it Philos. Mag. B}, {\bf 79}, {729}.
\bibitem[(Farahani \& Naghdabadi 2000)]{farahani}
{\sc Farahani K. \& Naghdabadi R.} {2000},
{\it Int. J. Solids Struct.}, {\bf 37}, {5247}.
\bibitem[(G\'eminard {\it et al.} 2004)]{geminard}
{\sc G\'eminard J.-C., {\.{Z}}ywoci\'{n}ski A., Caillier F. \& Oswald P.} {2004},
{\it Phil. Mag. Lett.} {\bf 84}, {199}.
\bibitem[(Goldhirsch \& Goldenberg 2002)]{gold3}
{\sc Goldhirsch I. \& Goldenberg C.} {2002},
{\it Eur. Phys. J. E}, {\bf 9}, {245}.
\bibitem[(Goldhirsch \& Goldenberg 2004)]{gold4}
{\sc Goldhirsch I. \& Goldenberg C.} {2004},
%{\it Small and Large Scale Granular Statics},
{\it Granular Matter}, {\bf 6}, {87}.
\bibitem[(Graner {\it et~al.} 2001)]{graner2001}
{\sc Graner F., Jiang Y., Janiaud E. \& Flament C.}
{2001},
%{Equilibrium states and ground state of two-dimensional fluid
% foams}.
{\it {Phys. Rev. E}}, {\bf{63}}, {011402}.
\bibitem[(Graner 2002)]{wuppertal}
{\sc Graner F.} {2002},
{\it Two-dimensional fluid foams at equilibrium}, in Ref.
\protect\cite{mecke}, pp. 187-214.
\bibitem[(Hoger 1987)]{hoger}
{\sc Hoger A.} {1987},
{\it Int. J. Solids Struct.}, {\bf 23}, {1645}.
\bibitem[(H\"{o}hler, Cohen-Addad \& Asnacios 1999)]{asnacios}
{\sc H\"{o}hler R., Cohen-Addad S. \& Asnacios A.}
{1999},
{\it Europhys. Lett.}, {\bf 48 }, {93}.
%Europhys. Lett, 48 (1), pp. 93-98 (1999)
%Rheological memory effect in aqueous foam
%R. Hhler, S. Cohen-Addad and A. Asnacios
\bibitem[(H\"{o}hler, Cohen-Addad \& Hoballah 1997)]{hohler}
{\sc H\"{o}hler R., Cohen-Addad S. \& Hoballah H.}
{1997},
{\it Phys. Rev. Lett.}, {\bf 79}, {1154}.
\bibitem[(Jiang {\it et al.} 2000)]{delft}
{\sc Jiang Y., Asipauskas M., Glazier J. A., Aubouy M. \& Graner F.}
{2000}.
In {\it Eurofoam 2000} ed. Zitha P., Banhart J. \& Verbist G., p.
297, MIT Verlag,
Bremen.
\bibitem[(Jiang {\it et al.} 1999)]{jiangpre}
{\sc Jiang Y., Swart P.J., Saxena A., Asipauskas M.
\& Glazier J.A.}
{1999},
{\it Phys. Rev. E}, {\bf 59}, {5819}.
\bibitem[(Kabla \& Debr\'egeas 2003)]{debregeas2}
{\sc Kabla A. \& Debr\'egeas G.}
{2003},
{\it Phys. Rev. Lett.}, {\bf 90}, {258303}.
\bibitem[(Kern \& Weaire 2003)]{kern}
{\sc Kern N. \& Weaire D.} {2003},
{\it Phil. Mag. } {\bf 83},{ 2973}.
\bibitem[(Khan \& Armstrong 1986)]{khan1986}
{\sc Khan S.A. \& Armstrong R.C.} {1986},
{\it J. Non-Newt. Fluid
Mech.}, {\bf 22}, {1}.
\bibitem[(Khan \& Armstrong 1987)]{khan1987}
{\sc Khan S.A. \& Armstrong R.C.} {1987}, {\it J. Non-Newt. Fluid
Mech.}, {\bf 25}, {61}.
\bibitem[(Khan, Schnepper \& Armstrong 1988)]{khan1988}
{\sc Khan S.A., Schnepper C.A. \& Armstrong R.C.} {1988},
{\it J. Rheol.}, {\bf 32}, {69}.
\bibitem[(Kraynik, Reinelt \& van Swol 2003)]{kraynik2003}
{\sc Kraynik A., Reinelt D., \& van Swol F.}
{2003},
{\it Phys. Rev. E}, {\bf 67}, { 031403}.
\bibitem[(Kruyt 2003)]{kruyt}
{\sc Kruyt N.P.} {2003},
{\it Int. J. Solids Struct.}, {\bf 40}, {511}.
\bibitem[(Kruyt \& Rothenburg 1996)]{kr96}
{\sc Kruyt N.P. \& Rothenburg L.} {1996},
{\it J. Appl. Mech.}, {\bf 118}, {706}.
\bibitem[(Kruyt \& Rothenburg 2002)]{kr02}
{\sc Kruyt N.P. \& Rothenburg L.} {2002},
{\it Int. J. Solids Struct.}, {\bf 39}, {311}.
\bibitem[(Labiausse, H\"ohler \& Cohen-Addad 2004)]{labiausse}
{\sc Labiausse V., H\"ohler R. \& Cohen-Addad S.} {2004}
in {\it Proceedings of Eufoam 2004}, {\sc Adler M.} ed., Elsevier, to appear.
\bibitem[(Landau \& Lifschitz 1986)]{landau}
{\sc Landau L. D. \& Lifschitz E. M.} {1986},
{\it Theory of Elasticity},
{Reed, Oxford, 3$^{rd}$ ed.}.
\bibitem[(Langer \& Liu 1997)]{langer}
{\sc Langer S.A. \& Liu A.J.} {1997},
%Effect of Random Packing on Stress Relaxation in Foam
{\it J. Phys. Chem. B} {\bf 101}, { 8667}.
\bibitem[(Larson 1997)]{larson}
{\sc Larson R.} {1997},
{\it J. Rheol.} {\bf 41}, {365}.
\bibitem[(Lauridsen, Twardos \& Dennin 2002)]{lauridsen}
{\sc Lauridsen J., Twardos M. \& Dennin M.}
{2002},
{\it Phys. Rev. Lett.}, {\bf 89}, {098303}.
\bibitem[(Liao {\it et al.} 1997)]{liao}
{\sc Liao C.L., Chang T.P., Young D.H. \& Chang C.S.} {1997},
{\it Int. J. Solids Struct.}, {\bf 34}, {4087}.
\bibitem[(Macosko 1994)]{macosko}
{\sc Macosko C.} {1994},
{\it Rheology: principles, measurements and applications},
Wiley-VCH, New-York.
\bibitem[(Mason, Bibette, \& Weitz 1995)]{mason}
{\sc Mason T., Bibette J., \& Weitz D.} {1995}
{\it
Phys. Rev. Lett. }, {\bf 75}, {2051}.
%Elasticity of Compressed Emulsions
\bibitem[(Mecke \& Stoyan 2002)]{mecke}
{\sc Mecke K. \& Stoyan D. eds.} {2002},
{\it
Morphology of Condensed Matter - Physics and Geometry of Spatially
Complex Systems},
Lecture Notes in Physics 600, Springer, (Heidelberg), proceedings
of the 2nd conference ``Spatial Statistics and Statistical Physics",
Wuppertal (Germany), March 2001.
\bibitem[(Phan-Thien 2002)]{phan}
{\sc Phan-Thien N.} {2002},
{\it Understanding viscoelasticity}, Springer, Berlin.
\bibitem[(Porte, Berret \& Harden 1997)]{porte}
{\sc Porte G., Berret J.F. \& Harden J.L.} {1997},
{\it J. Phys. II France}, {\bf 7}, {459}.
\bibitem[(Poynting 1909)] {poynting}
{\sc Poynting J.H.} {1909},
{\it Proc. Roy. Soc. London}, {\bf A82}, 546.
\bibitem[(Pratt \& Dennin 2003)]{pratt}
{\sc Pratt E. \& Dennin M.}
{2003},
{\it Phys. Rev. E}, {\bf 67}, {051402}.
\bibitem[(Princen 1983)]{princen}
{\sc Princen H. M.} {1983},
{\em J. Coll. Interf. Sci.} {\bf 91} {160}.
\bibitem[(Reinelt \& Kraynik 2000)]{kraynik2000}
{\sc Reinelt D. A. \& Kraynik A.}
{2000},
{\it J. Rheology}, {\bf 44}, {453}.
\bibitem[(Rosenkilde 1967)]{rosenkilde}
{\sc Rosenkilde C.E.} {1967}, {\it J. Math. Phys.}, {\bf 8}, {84}.
\bibitem[(Rouyer {\it et al.} 2003)]{rouyer}
{\sc Rouyer F., Cohen-Addad S., Vignes-Adler M. \& H\"ohler R.}
{2003}, {\it Phys. Rev. E}, {\bf 67}, {021405}.
\bibitem[(Scheibert \& Debr\'egeas 2004)]{debregeas4}
{\sc Scheibert J., Debr\'egeas G.}
{2004},
in preparation.
\bibitem[(Tanguy {\it et al.} 2002)]{tanguy}
{\sc Tanguy A., Wittmer J.-P., Leonforte F. \& Barrat J.-L.}
{2002},
{\it Phys. Rev. B}, {\bf 66}, {174205}.
\bibitem[(Tanner \& Tanner 2003)]{hencky}
{\sc Tanner R.I. \& Tanner E.} {2003},
{\it Rheo. Acta}, {\bf 42}, {93}.
\bibitem[(Weaire \& Hutzler 1999)]{hutzler}
{\sc Weaire D. \& Hutzler S.}
{1999},
{\it Physics of Foams},
{Oxford Univ. Press, Oxford}.
\bibitem[(Weaire \& Hutzler 2003)]{dilatancy}
%Dilatancy in liquid foams
{\sc Weaire D. \& Hutzler S.}
{2003},
{\it Phil. Mag.}, {\bf 83}, { 2747}.
\bibitem[(Weaire {\it et al.} 2004)]{morgan}
{\sc Weaire D., Kern N., Cox S.J., Sullivan J.M. \& Morgan F.} {2004},
{\it Proc. Roy. Soc.: Math. Phys.}, {\bf 460}, {569}.
\bibitem[(Ybert \& di Meglio 2002)]{ybert}
{\sc Ybert C. \& di Meglio J.-M.}
{2002},
{\it C.R. Physique}, {\bf 3}, {555}.
\bibitem[(Zimmerman 1999)]{zimmerman}
{\sc Zimmerman J.A.} {1999},
PhD thesis, Univ. Stanford, chap. 4, pp. 67-106, unpublished.
\end{thebibliography}
%\end{references}
\end{document}