HAL Id: jpa-00246502
https://hal.archives-ouvertes.fr/jpa-00246502
Submitted on 1 Jan 1992
HAL is a multi-disciplinary open access archive for the deposit and dissemination of sci- entific research documents, whether they are pub- lished or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.
Random media with temperature controlled connectivity
F. Carmona, E. Valot, L. Servant, M. Ricci
To cite this version:
F. Carmona, E. Valot, L. Servant, M. Ricci. Random media with temperature controlled connectivity.
Journal de Physique I, EDP Sciences, 1992, 2 (5), pp.503-510. �10.1051/jp1:1992111�. �jpa-00246502�
Classification Physics Abstracts
64.60A 73.40C 81.20T
Short Communication
Random media with temperature controlled connectivity
F-
Carmona,
E.Valot(*),
L. Servant and M.Ricci(**)
Centre de Recherche Paul Pascal-CNRS, Av. A. Schweitzer, 33600 Pessac, France
(Received
3 March 1992, accepted 12 March1992)
Abstract We show that the electrical resistance of a percolating medium made of a polymer
filled with conductive particles may undergo a steep increase over a narrow temperature range above room temperature. We demonstrate that a likely mechanism for the effect is the loss of interparticle electrical contact inducing a weakening of the connectivity of the medium as
temperature is increased.
There has been much interest in recent years in the non-linear electrical
properties
of randompercolative
systemscomprising
non-linear elements like fuses [1, 2], diodes [3], weak supracon- ductive links [4, 5],Analogies
with the rupture of materials havelargely
been stressed[6-8].
Most of the
published
studies are concerned with the theoreticalpredictions
of behavior or report computerexperiments.
On the otherhand,
verylarge
and reversible increases of the resistance ofcomposites
made of apolymer
filled with conductiveparticles
when heated aboveroom temperature have
long
beenreported
and known as thepositive
temperature coefficientor CTP efTect [9, 10]. But no
convincing interpretation
of such efTect had beenprovided
sofar
especially
when thepolymer
matrix is notthermoplastic [11,
12]. Besides it haslong
beenreported that,
at room temperature, such disordered mediaundergo
an insulator-to-conductive transition aspredicted by percolation theory
when the conductive filler volume concentration is increased [13]. In this letter wegive
evidence that alikely origin
of such a behavior is a tem-perature induced
change
of theconnectivity
of the random network ofinterparticle
contacts, in otherwords,
it is anexperimental
illustration of the behavior of a random fuse network withtemperature controlled fuse
breaking.
The materials
reported
in this letter are made of an epoxy whoseglass
transition temper-ature is 50°C and conductive short carbon fibers with I mm
length
and 8 pm diameter. Theprocessing
and electricalproperties
of suchcomposites
have beenreported previously
[14]. It has been shown[IS] that,
contrary to similar materials in which thefilling particles
are carbonblacks,
conductionproceeds through
direct contact among theparticles
and notby tunneling (*)
Now at ATOCHEM, Centre d'applications, 95 rue Danton, 92300 Levallois, France.(**) Now at DIAS Industries, BPS, 33380 Marcheprime, France.
504 JOURNAL DE PHYSIQUE I N°5
across a thin
polymer
barrier. As the volume concentration 4l of the fibers is increased fromzero, an insulator-tc-conductive transition is
observed,
the behavior of the electrical conduc-tivity
in theneighborhood
of the critical concentration 16*being largely compatible
with thepower law behavior
predicted by percolation theory.
The fibers critical volume fraction and theconductivity
exponent, determined from the concentrationdependence
of the room tempera-ture
composite conductivity
as shown on thelog-log plot
offigure
I, are 16* = 0.9~ and t = 3respectively.
Bothfigures
are ingood
agreement withprevious findings
on similar systems [13]. Let usjust
recall that the very low value for 16* is due to the geometry of theparticles (aspect
ratiolarger
than ahundred)
and it is sensitive to thephysicc-chemical properties
of thesuspending
resin(viscosity
and surface tensionessentially).
On the other hand the value for t which issignificantly larger
than the universal value t = 2 is still notfully
understood [16].io
P
io ~
i~,cm)
i o4
1o i
1.5 1-o o.5. o-o o.5 1-o
tog (tb q~#)
Fig, i.
Log-log
plot of the concentration dependence of the resistivity of the short carbon filled composites.Straight
line has slope t = 3,o.In
figure
2 wegive
atypical example
of the temperaturedependence
above room temperature of acomposite
resistance on the conductive side of the transition. One observes alarge
increaseby
several orders ofmagnitude
of the resistance over a narrowtemperature
range. The location of this efTect and theapparent slope depend
on the fiber volume concentration(more precisely
on the relative "distance" from the
transition)
and on the epoxy. Outside theabrupt change
region,
the resistance isonly weakly temperature dependent.
Belowit,
there exists aslight
anomily
in theglass
transition temperatureneighbourhood.
Above thisregion,
the resistance is almost temperatureindependent
and we have also found that its value is of order I to 10 x 10~ kQpractically independent
of the fiber concentration. Note that suchlarge
resistancechanges
areessentially
due toresistivity changes,
the efTects of thermalexpansion
of thesample being negligibly
small. However, theresistivity change
itself isobviously
to be attributed to thethermal
expansion
of thecomposite.
But it cannot besimply
due to fiber volume concentrationchanges
inducedby
the difTerence between therespective
coefficients of thermalexpansion (CTE)
of thepolymer
and of the carbon fibers asactually
observed when the temperature is decreased below room temperature[15]. Although
such an efTect wouldlead,
in agreementwith the
data,
to an increase of the resistance since thepolymer expands
more(CTE
of order10~~ °C~~)
than the fibers(CTE
of order10~~ °C~~),
itpredicts
too smalla relative
change of,
at most, a few tens of percent.~6
R/Ra
~~5
~WW+
+*
io4
+ RJRa exp
lo ~
R/Ro calc
i o2
o i
io°
-.*'*****+~
T(°c)
i o
50 loo lso 200
Fig.
2. Relative variations with temperature of the resistance of a composite sample containing 4i = 2.5% carbon fibers. Full line is calculated with relations (3) and(4)
and Ts =162°C,
S=
19°C and Tc
= 187°C.
A
possible explanation
of the observed behavior is that there is achange
in the connectivity of the system as the temperature isincreased,
that is to say,- electrical connection among the conductiveparticles
ischanged
when thepolymer
matrixexpands.
Such aconjecture
isconvincingly supported by
the thermal behavior of the electrical resistance of a test systemconsisting
in twoinitially contacting
carbon filaments or wires embedded in the epoxy. Such asystem
can be considered as anexperimental
model of a "contact" similar to those present in506 JOURNAL DE PHYSIQUE I N°5
the
composites.
Oneactually
observes that it mayreversibly
switch from about I kQ(which
represents
essentially
the resistancealong
thefilaments)
to about 10~ kQ when the temperature iscyclically changed
from room temperature to 200 °C and back. This proves that the "contact resistance" among the wires isnegligible
at room temperature and increasesdramatically
underheating:
theinitially
"closed"junction
may"open"
but one must notice that in the open state the resistance remains finite. For agiven junction,
the"opening"
and"closing"
temperatures in acycle
are difTerent andthey
both vary from onecycle
to the other.Furthermore,
when aset of n identical
junctions
isstudied,
allopening
andclosing
temperatures are different. Thisexperiment
then evidences actual reversible randomopenings
of the electrical contact between two wires imbedded in an epoxy.Figure
3 reports atypical histogram
of contactopenings
asa function of temperature,
Q(T) being
the cumulatedproportion
of the n studiedjunctions (in
thisexample
n=
20)
which are open at temperature T when the temperature has beencyclically changed
N times(N
= 20 in thisexample)
between room temperature and 150°C.Opening
occursstatistically
over a limited temperature range above room temperature in fairsimilarity
with theshape
of thecomposite
resistance versus temperature curve. Notespecially
the few contact
openings
in theregion
of thepolymer glass
transition temperature. We found that theshape
of theQ(T)
curvedepends
on the mechanical tension which isapplied
to thejunctions
whileprocessed
in order toget "good"
electrical contact at room temperature. The mechanism of contactopening
above room temperature,although
nottotally clear,
is then seenas the result of the two
competing
actionsof,
on the onehand,
the epoxy thermal expansion which tends to separate the wiresand,
on the otherhand,
of the mechanical stressacting
on the
junction
which opposes to it. Besides theapplied
mechanicaltension,
this stress alsocomprises
thoseappearing
oncooling
frompolymerization
temperature(which
is 120° C in thesematerials)
and in the course of resinpolymerization.
The latterare known to
reorganize locally
when the
polymer glass
transition iscrossed,
whileremaining grossly statistically
identical below and above the transitionrespectively
over successive thermalcycles.
This insures that contactopening
is a random process which can be describedby
a definite distribution function.Admitting
from now on that such contactopenings
among conductiveparticles
is alikely
mechanism forconnectivity change
incomposites
we mayanalyze
which consequences on the temperaturedependence
of thecomposites resistivity
are to beexpected.
Note that the fol-lowing analysis
will not consider the weakanomaly
around Tgsince, although
it has the sameorigin,
it is more difficult to model.The
assembly
of short carbon fibers inside thecomposite
is viewed as a random lattice of fiber-te-fiber contacts with numberdensity
p. We make the first reasonableassumption
that there is
negligible
Jouleheating
whilemeasuring
thecomposite
resistance so that the temperature is uniform over thesample.
It results that contactopening
may involve any twoinitially
connectedparticles
whetherthey belong
to the infinite cluster or to a finite size cluster.We make the second
assumption
that contactopening
statistics is describableby
aprobability density
functionq(T)
which has non-zero values in agiven
temperature range(Tm, TM)
aboveroom temperature: the
probability
for a contact to open when the temperature T in the convenient interval ischanged by
dT will beq(T)dT
and theproportion
of open contacts~~~~~"~~ '~~~~ ~ "'
T
Q(T)
=/ q(r)dr (I)
Tm
Let po be the number
density
ofinterparticle
contacts at room temperature in agiven composite
with fiber volume concentration 16. It must be anincreasing
function of16. Therelationship
between po and 16 is not known. The excluded volume estimate of critical number of contacts per fiber is of no
help
here since in the present system it has been demonstrated that contact formationduring
materialprocessing
is nottotally
accounted forby geometrical
argumentsa(T) (%)
80
40
20
O
oUnoUnco~u~~u~o-UnoUnoUnc3~c3~Comc3mCo~u~~u~~u
~u~ummm~~UnUn~o~o~~CoCoCoc3c3oo--~u~umm+~Un
Fig. 3. Typical contact opening histogram
Q(T)
obtained during 20 successive heatings from room temperature to 160°C of 20 model fiber-fiber contacts,[13, 17].
Let pc be the value of po for the critical concentration 16*(as
observed at roomtemperature).
As a result ofheating,
the actualdensity
p will decrease as:P(T)
=Po(I Q(T)) (2)
Depending
on both theshape
ofq(T)
and the initial povalue, p(T)
willeventually equal
p~ at a
given
temperature T~ and thecomposite
will becomeinsulating Q(T~)
= ~° ~~Pc is the
proportion
of contacts which have to be broken in order tobring
the random lattice ofinterparticle
contacts to thepercolation
threshold. It isexpected
to increase with'the fiber volume concentration since theprobability
for a contact tobelong
to the infinite clusterincreases with16
(or po).
Provided that the po value is such that the usual power lawdependence (p
pc)~ for theconductivity
holds at room temperature(this
is the case for allsamples
we aregoing
to consider below as seen inFig, I),
then the temperaturedependence
of thecomposite
resistivity
is:p cc
(p(T) pc)~~ (when
T <Tc)
whichfinally
leads toj
~
~
PO~~~j
~~~
~ @j ~~
~~~
Po Po Pc
Q (Tc)
where po is some reference material
resistivity:
it should be the room temperatureresistivity
but,
due to the weakanomaly
atTg,
it will be theresistivity just
above it. Note that all these power laws would be exactif,
in the open state, the contacts had infinite resistance which we508 JOURNAL DE PHYSIQUE I N°5
know not to be true: when a contact is
"broken",
it isreplaced by
alarge
but finite resistance in the random lattice and for p closeenough
to pc(high enough temperature)
the power lawdependence
is nolonger
valid.However,
the actual"open
state" resistance islarge enough
forthese
relationships
to remain vali@ up to temperatures which are very close to thebeginning
of theplateau. Besides,
as there is no reason for therelationship
between po and F to be critical at the roomtemperature percolation transition,
theconductivity
exponent value to be used is the one which was deduced from the concentrationdependence
of the room temperatureconductivity,
I,e, t = 3.The relative
change
with temperature of thecomposite resistivity
can then be calculated ifq(T)
is known. This is not the case sincefirst,
it is notsimply
related to theexperimental Q(T)
of the modeljunctions (there
is noequivalent
in thecomposit~
to theapplied
mechanicaltension)
andsecond,
itdepends
on the internal stress distribution inside thecomposite
which is not known.Assuming
forq(T)
a function which satisfies therequirement
of non-zeroopening probability
within a limited temperature range, we may on the contraryby
curvefitting
to theexperimental
data obtainq(T). Taking
forq(T)
a Gaussian distributionq(T)
=(I /s/&)
exp
[(T Ts)~ /2s~j (4)
we
performed
a curvefitting
on the data above Tg withTc,
7~ and S as threeadjustable
parameters. Note that Tc isonly slightly adjustable
since it must be below and very close to thebeginning
of theplateau. Figure
2 shows thetypical
result obtained with the data for thesample containing
2.5~ fibers with Tc = 187 +2°C,
Ts = 162 + 4°C and S= 19 + 1°C. These values
satisfy
therespective requirements
ofplateau
closeness andnegligible
contactopening
outside a finite temperature range as shown
by
thecorresponding Q(T) offigure
4 which has been calculated with the above set of parameters. Asexpected,
the temperature range over which there isagreement
between calculated andexperimental points
is limited on the lower sideby
the Tganomaly,
and on the upper sideby
the finite open state resistance effect whichactually precludes
any resistancedivergence,
Table I reports the values of the parameters as obtained on a limited series of
samples
withvarying
fiber volume concentration between I and 3.5~. Given theexperimental
andfitting uncertainties,
one observes theinteresting
feature of concentrationindependent q(T) (constant
7~ andS).
This demonstratesthat,
at least for materials with fiber concentrations in the range of thisstudy,
contactopening
is a local effect with no correlation amongneighbouring
contacts. In other
words,
all material andprocessing
parametersbeing
held constant, theopening probability
functiondepends only
on the matrix material constants and not on thecomposite
material constants(which depend on16)
thisinteresting
conclusion contradictssome earlier works. The distribution function
q(T) being independent
ofconcentration,
the Tc increase is ingood
agreement with the necessary increase with 16 oftheproportion
of contacts to beopened
forreaching
the critical contactdensity
pc.Noting
that the values of n and Sobtained for the
sample containing
2.5~ fibers are very close to the means of therespective
values for all
compositions,
we have located therespective
temperatures Tc(Fig. 4).
For eachcomposition,
the valueQ (Tc), gives
theproportioi
of contacts to beopened
forbringing
thecorresponding
medium topercolation
threshold. It increases with the fiber concentration.In summary, we have shown that
percolating
media made ofshort carbon fibersdispersed
ina
polymer
have apeculiar
temperaturedependence
of the resistance which may be accounted forby
aphenomenological
model ofvarying connectivity.
We have shown that the latter may be describedby
some functionq(T)
which is concentrationindependent:
it can be used forcharacterizing
different series of materials. Therelationship between,
on the onehand,
the contactopening density
functionand,
on the otherhand,
thepolymer
thermomechanical con- stants andprocessing
parameters has beenextensively
studied [18]: the results will bereported
ioo
Q(T) (ib)
80
60
40
20
I'
O
loo 200
T
(oc)
Fig. 4. Cumulated contact opening function
Q(T)
in composites calculated with relations(1)
and(4)
and values of the parameters: Ts = 162°C and S= 19°C. Vertical lines show the location of the
respective values of Tc for the five investigated composites.
Table I. Concentration
dependence
of the parameters7~, S,
andTc.
Note that the values of the first two parameters arefairly
constant whereas Tc increases with 16.~(i~)
158.5 19 161
2 161 19 167
2.5 162 19 187
3 164 19 199
3.5 158 20 205
510 JOURNAL DE PHYSIQUE I N°5
elsewhere. In this
article,
we have considered theonly
case ofconnectivity
as controlledby
temperature when the material is heated.Temperature
increase may also be achievedby
Jouleheating:
there isexperimental
evidence that contactopening
also occursgiving
rise tostrongly
non-linear resistance versus
voltage
characteristics [18]. Work is now in progress forstudying
this different case where contact
opening
involves theonly
fibers whichbelong
to the backbone of the infinite cluster.Finally,
the sameapproach
may beapplied
to the case of random normalmetal/superconducting grain composites
in whichgrain-tc-grain superconducting coupling
is a function of temperature andmagnetic
field: the results of thisstudy
will bepublished shortly.
Acknowledgements.
We
acknowledge
financial support of thisstudy by
Le Carbone LorraineCy.
References
Ill
DUXBURY P-L-, LEATH P-L- and BEALE P-D-, Pbys. Rev. B 36(1987)
367.[2] SODERBERG M., Pbys. Rev. B 35
(1987)
352.[3] Roux S., HANSEN A. and GUYON E., J. Pbys. France 48
(1987)
2125.[4] LEATH P-L- and TANG W., Pbys. Rev. B 39
(1989)
6485.[5] HINRICHSEN El., Roux S. and HANSEN A., Pbysica C 167
(1990)
433.[6] GILABERT A., VANNESTE C., SORNETTE D. and GUYON E., J. Pbys. France 48
(1987)
763.[7] De ARCANGELIS L. and REDNER S., J. Pbys. Lett. France 46
(1985)
L585.[8] Roux S. and HANSEN A., Europbys. Lett. ii
(1990)
37.[9] HEYWANG W., Solid. State Electron. 3
(1969)
51.[10] MEYER J., Polym. Eng. Sci. 13
(1973)
462.[llj
OHE K. and NAITO Y., Jpn J. Appl. Pliys. 10(1971)
99.[12] MEYER J., Polym. Eng. Sci. 14
(1974)
706.[13] CARMONA F., Ann. Cbim. Fr. 13
(1988)
395.[14] CARMONA F., CANET R. and DELHAES P., J. Appl. Pbys. 61
(1988)
2550.[15] CARMONA F. and MOUNEY C., J. Mat. Sci., in press.
[16] VINCHES C., SALOME L., COULON C. and CARMONA F., J. Pbys. France 51
(1990)
2505.[17] MOUNEY C., PhD Thesis, Bordeaux University
(1989).
[18] VALOT E., PhD Thesis, Bordeaux University