Modeling of anhydrite swelling with FLAC

J.M. Rodríguez-Ortiz
Gamma Geotécnica SL, Madrid, Spain
P. Varona & P. Velasco
Itasca Consultores SL, Asturias, Spain

ABSTRACT: Anhydrite and rocks containing argillaceous minerals experience swelling phenomena when
they come into contact with water. In tunneling, this can lead to a strong heave of the floor and to a high level
of stresses in the lining. Although characterization of swelling potential, monitoring of swelling process, and
a lot of relevant case histories of tunnel construction in swelling rocks are currently available, the design of
the support in swelling rocks usually do not consider an accurate stress-strain relationship for the swelling.
Current trend in tunneling design considers numerical modeling of the rock-support interaction, but the
available geotechnical codes do not include the swelling formulation. This paper presents the implementation
in FLAC (via FISH routines) of the analytical stress-strain formulation for the swelling presented by Wittke
(1999) and the validation of this algorithm against the swelling tests carried out by different authors and
presented by Wittke (1999).

Previous data refer to pure anhydrite, but in case

of interbedded mudstone-anhydrite the maximum
1 DESCRIPTION OF THE ANHYDRITE swelling volume is lower but the swelling stress is
SWELLING PHENOMENA larger (in the order of 2 to 5 MPa).
In the case of pure anhydrite, as the hydration
1.1 Chemical description
process begins, the thin layer of impervious gypsum
Calcium sulphate appears naturally as two different created at the surface of the grains inhibits the water
minerals: gypsum (CaSO4⋅H2O), in which water penetration stopping the process. In the case of
appears within the crystalline structure, and interbedded anhydrite-mudstone the swelling
anhydrite (CaSO4). The hydration of anhydrite is a process leads into the disintegration of the rock,
complex process that depends on the pressure and reducing its strength. Steiner (1993) quantifies this
the temperature. For ambient conditions the reduction of strength with an angle of friction of 20º.
chemical reaction is illustrated in Table 1.

Table 1. Transformation of anhydrite into gypsum.

_____________________________________________ 1.2 Characterization of the swelling behavior
Anhydrite + water Gypsum
_____________________________________________
Equation CaSO4 + H2O CaSO4⋅2H2O The International Society of Rock Mechanics has
Mass (gr) 136.14 + 36 172.14 proposed a set of tests to quantify the swelling of
Density (gr/cm3) 2.96 1 2.32 argillaceous rocks: the Maximum Axial Swelling
Volume (cm3) 46.2 + 36 74.3
_____________________________________________ Stress test, the Axial and Radial Free Swelling
Strain test, and the Axial Swelling Law test (axial
With an external inflow of water, the volumetric swelling stress as a function of axial swelling strain,
increment associated to this process is presented in or Huder-Amberg swelling test).
Equation 1: An illustration of the results from the Huder-
Amberg swelling test (total vertical strain of the
∆V 74.3 − 46.2 sample, εz in %, versus vertical load, σz in kPa) is
= 100% = 61% (1)
V 46.2 presented in Figure 1, taken from Wittke (1999).
Transformation of anhydrite into gypsum can be Stages 1, 2 and 3 correspond to the initial loading
inhibited at 20ºC with a pressure of 1.6 MPa; this phase with 2 load cycles; stage 4 corresponds to the
process is reversible, being necessary a pressure of watering of the sample (no stress increment but
80 MPa at 58ºC to transform gypsum into anhydrite.
 2 ANHYDRITE SWELLING LAW
The following description of swelling law is taken
from Wittke (1999), and starts from the axial stress-
strain relationship previously presented (Equation
2). The swelling law only applies for compressive
stresses, where σc (Figure 2) is a minimum stress
representing the lower limit of validity of the
swelling law. As the swelling strain equals to zero
for compressive stresses larger than the swelling
stress, the swelling law can be finally formulated as
Equation 3:
⎧0 ⎫ ⎧σ i ≥ σ 0 ⎫
⎪ ⎪ ⎪ ⎪
ε q
i∞ = ⎨K q log(σ i / σ 0 ) ⎬ for ⎨σ c ≤ σ i < σ 0 ⎬ (3)
⎪ ⎪ ⎪σ < σ ⎪
⎩K q log(σ c / σ 0 )⎭ ⎩ i c ⎭
where εqi∞= final axial strain due to swelling in the
direction i; Kq= swelling deformation parameter; σi=
axial stress in the direction i; σ0= axial swelling
Figure 1. Swelling test of an interbedded anhydrite-mudstone
sample (Wittke, 1984, in Wittke 1999).
stress; and σc= minimum limit for the axial stress.
Equation 3 represents the 3D (i=1,2,3, means the
strain increment), and finally stage 5 corresponds to 3 directions in the space) isotropic (the same Kq
the different points of the unload-swelling process. parameter is considered for the 3 directions)
The swelling strain equals the total strain (stage 5 of swelling law, where the final axial strain is reached
the test) minus the elastic strain (stages 2 and 3 of at the end of the swelling process.
the test). This swelling law considers that the principal
If the strain due to swelling is plotted against the directions of swelling εqi∞ (i=1,2,3) are coaxial with
stress in a semi-logarithmic scale (Figure 2) a the principal stresses σi (i=1,2,3), and therefore the
straight line is obtained which may be described by value of the swelling principal strain depends only
the relationship presented in Equation 2. on the value of its coaxial principal stress.
Previous relations refer to the strain reached at
the end of the swelling process. Furthermore, for the
kinetics of the process Witkke (1999) presents the
following Equation 4 for the swelling strain rates at
time t:
∂ε iq (t ) 1
∂t ηq
[
= ⋅ ε iq∞ − ε iq (t ) ] (4)

where ηq= swelling time parameter; εqi∞= principal

swelling strains for t=∞; and εqi(t)= principal
swelling strains which already occurred until time t.
According to Wittke (1999), the time dependence
of swelling is adequately described by Equation 4 as
long as the strength of rock is not exceeded. The
plastic deformations occurring if the rock strength is
Figure 2. Axial swelling law (Grob, 1972, in Wittke, 1999).
exceeded lead to a volume increase and to an
increase of permeability, that accelerates the
⎛σ ⎞ penetration of water, increasing the swelling strain
ε qz∞ = K q ⋅ log⎜⎜ z ⎟⎟ (2) rate.
⎝ σ0 ⎠ To consider this effect in the formulation, Wittke
where εqz∞= final axial strain due to swelling (final (1999) includes the following relationship for the
deformation means deformation at the end of the swelling time parameter ηq, Equation 5:

{ }
swelling process); Kq= swelling deformation 1
parameter; σz= axial stress; and σ0= axial swelling =a 0 + a el ⋅ ε elv + a vp ⋅ min ε plv , maxEVP (5)
stress (as showed in Figure 2, it is intersection of the ηq
straight line with εqz∞= 0). where a0, ael, avp= constant values. The parameter a0
represents the dependence of the swelling velocity
on the anhydrite content, regardless of whether a Then, a loop is performed until the swelling time
strain occurred before or not; εelv is the elastic reaches the expected simulation age. Within this
volumetric strain occurred prior to the beginning of loop, the strain tensor due to complete swelling is
swelling that also influence the permeability; εplv is calculated according to Equation 3. The stress
the volumetric plastic strain; and maxEVP increments associated to these strains are calculated
represents an upper limit of the plastic volume strain with the following lineal elastic relationship,
with regard to an eventual acceleration of swelling. Equation 6:
According to Wittke (1999), plastic volumetric
strains larger than maxEVP do not lead to a further ⎧∆σ 1 = (λ + 2G ) ⋅ ε 1q∞ + λε 2q∞ + λε 3q∞ ⎫
⎪ ⎪
increase of the swelling velocity because the ⎨∆σ 2 = λε 1∞ + (λ + 2G ) ⋅ ε 2 ∞ + λε 3∞ ⎬
q q q
(6)
penetration of water into the rockmass cannot be ⎪ q ⎪
⎩∆σ 3 = λε 1∞ + λε 2∞ + (λ + 2G ) ⋅ ε 3∞ ⎭
q q
further accelerated by these.
Following Equation 5 the swelling time
parameter is no longer constant but dependent on where λ and G are constants known as Lamé’s
time as elastic-plastic volumetric strains varies parameters (λ= K - 2/3G; K is the bulk modulus and
during the swelling process. G is the shear modulus).

t=0
3 FINITE DIFFERENCES CALCULATION
ALGORITHM
σ i=1, 2 = f (σ xx ,σ yy ,τ xy ,θ )
The swelling law presented in previous paragraphs Principal
has been implemented in FLAC, coupling the σ 3 = σ zz stresses
swelling phenomena with the built-in elastic-plastic
constitutive relationships via FISH routines. The t< T_fin
principal concept of this algorithm is that the
volumetric strain is reached in the zones of the
model introducing of small increments of isotropic ε iq∞ = K q log⎛⎜σ i σ ⎞⎟ Swelling strains
stress within them, Noorany et al (1999). ⎝ 0 ⎠ Associated stress
The sketch of the algorithm is to calculate the (
∆σ i = f εiq∞ ; λ, G ) increment
final swelling strain tensor for all the zones of the
model, transform the strains into an increment of
stresses, and then “inject” the stresses in small
increments into the zones.
1
{ }
= a0 + ael ⋅ εelv + a vp ⋅ min εvpl , max EVP
Swelling time
ηq parameter
The flowchart of the calculation algorithm is
presented in Figure 3, and can be resumed in the
following points: ⎧10% ⋅σ i ⎫
∆t = ηq ⋅ min ⎨ ⎬
− Determination of the principal stresses for all the i =1, 2, 3
⎩ ∆σ i ⎭ Minimum
elements of the model. timestep
− Determination of the swelling principal strains ∆t min = min {∆t}
and of the stress increments associated with these
strains. ∆t min
σ i = σ i + ∆σ i
− Determination of the swelling time parameter. ηq “Injection” of a
− Determination of the minimum timestep,
σ xx ,σ yy, τ xy = f (σ i =1, 2 ,θ )
fraction of the
necessary for numerical convergence of the stress increment
algorithm, and to synchronize the swelling rate of σ zz = σ 3
all the element in the model.
− “Injection” of a fraction of the stress increment Solve to
associated to the swelling strain. Mechanical equilibrium mechanical
equilibrium
− Solve to mechanical equilibrium of the current
timestep t = t + ∆t min Accumulation of
− Accumulation of swelling strains and time. ∆t min
εiq = εiq + (εiq∞ − εiq ) swelling time and
− Repetition of the algorithm until the expected age ηq swelling strains
of the simulation is reached.
A detailed description of the algorithm is as
END
follows. After initiation of swelling time to zero, the
principal stress tensor for all the elements is
Figure 3. Flowchart of the calculation algorithm
calculated from the current stress state. As the
routine has been implemented in a 2D model, the
out-of-plane stress is a principal stress.
 Previously to the calculation of the swelling time ⎧ 1% ⋅ σ i ⎫
parameter the plastic component of the volumetric ⎪ri ⋅ ∆σ i = 1% ⋅ σ i ; ri =
∆σ i ⎪⎪
strain has to be determined. ⎪
⎨ ⎬ (11)
The total volumetric strain, addressed in FLAC ⎪r = min {r } = min ⎧ 1% ⋅ σ i ⎫ ⎪
with a FISH variable, is the sum of the following ⎪⎩ i=1, 2 , 3 i i=1, 2 ,3 ⎨⎩ ∆σ i ⎬⎭ ⎪⎭
components, Equation 7:
where ri= fraction of the stress increment ∆σi;
ε tot
v = εv + εv + εv + εv
el 0 el pl q
(7)
σi= stress state; and r= minimum fraction of the 3
where εtotv= total volumetric strain; εel0v= elastic principal directions.
volumetric strain produced in the model previous to From equations 6 and 11 the fraction r of the
any calculation; εelv= elastic volumetric strain stress increment that are going to be “injected” in
produced during the calculation; it can be calculated the elements of the model can be expressed with the
as εelv=(σ1+σ2+σ3)/(3K), being K the bulk modulus; following Equation 12:
εqv= swelling volumetric strain accumulated during ∆t
calculation, εqv=εq1+εq2+εq3; therefore, the plastic r= (12)
volumetric strain, εplv, can be calculated with the ηq
following Equation 8: and therefore, the timestep for each element can be
ε =ε
pl
v
tot
v −ε el 0
v −ε −ε
el
v
q
v (8) obtained as, Equation 13,
⎧ 1% ⋅ σ i ⎫
The swelling strain does not occur ∆t = η q ⋅ r = η q min ⎨ ⎬ (13)
instantaneously but following the kinetics i =1, 2 , 3
⎩ ∆σ i ⎭
formulated with Equation 4. Expressing this
differential equation in finite differences we obtain It is necessary to synchronize the rate of swelling
the following Equation 9: for all the elements of the model adopting the same
timestep for all; the minimum timestep of all the
ε iq (t + ∆t ) − ε iq (t )
∆t
=
1
ηq
[ε q
i∞ (t ) − ε iq (t )] (9) elements is the searched,
∆t min = min{∆t} (14)
and therefore, The “injection” of stresses associated to the
ε iq (t + ∆t ) = ε iq (t ) +
∆t
ηq
[ε q
i∞ (t ) − ε iq (t )] (10)
swelling behavior can be expressed with the
following Equation 15,
∆t min
As in all finite difference algorithm schemes, this σ i (t + ∆t ) = σ i (t ) + ∆σ i (15)
equation applies only for values of ∆t that are ηq
significantly low. This means that the swelling strain
that give the relationship between the current stress
at time t that still remains to produce, [εqi∞(t) -
state, σi(t) the total increment of stresses due to
εqi∞(t)] cannot be induced in the model
swelling, ∆σi, the minimum timestep, ∆tmin, and the
instantaneously because the model would
swelling time parameter ηq.
degenerate. Thus, the next phase is to determine a
These stress increments have to be transformed
critical value of ∆t to use in the finite difference
from the principal axes reference to the coordinated
scheme.
axes reference, assuming that the principal stresses
To determine a value of ∆t small enough, only a
have not rotated during the swelling processes, and
fraction of the stress increment associated to the
the angle between the principal stresses and the
remaining swelling strain should be “injected” in the
horizontal remains the same.
elements of the model.
Finally, once the mechanical equilibrium has
A criterion of a maximum of 1% of the current
been reached for this fraction of stresses injected
stress state has been adopted to determine de
into the elements of the model, it is necessary to
fraction of ∆σi to “inject”. The minimum fraction
actualize the accumulated swelling strain of each
obtained from the 3 principal directions in each
element of the model, Equation 16:
element is adopted. These relationships are
ε (t + ∆t ) = ε (t ) + ∆t min
[ε (t ) − ε iq (t )]
illustrated in Equation 11 for every element in the q min q q
model. i∞ (16)
ηq
i i

and for the swelling time, Equation 17.

t = t + ∆t min (17)
 This procedure is repeated until the accumulated
swelling time reaches the expected age of the To simulate this test, a FLAC model has been set
swelling simulation. up; the constitutive model is elastic with the same
properties presented in Figure 4. The results from
FLAC simulation are presented in Figure 5 (stress in
4 VALIDATION OF THE ALGORITHM kPa versus time in hrs).
The algorithm introduced in previous paragraphs As the model is isotropic, the 3 components of
has been implemented in FLAC via FISH routines. the stress are identical in the simulation.
Now, the validation of this algorithm against the
swelling tests presented in Wittke (1999) is
presented. 4.2 Huder-Amberg swelling test in elasticity

Figure 6 shows the swelling strain-time curves

4.1 Swelling pressure test obtained for the different stages of loading in a
Huder-Amberg swelling test on an anhydritic
The first validation test is a swelling pressure test of mudstone from de Gypsum Keuper.
a cubic sample of swelling mudstone carried out in a
triaxial test apparatus. After a load-unload cycle the
sample was flooded. Preventing the strains in the 3
directions the swelling pressures were measured in
function of time. Figure 4 presents the evolution of
the swelling pressure, in the 3 directions, versus
time, and the parameters for the swelling law.

Figure 6. Huder-Amberg swelling test in elasticity (in Wittke,

1999)

Figure 4. Swelling pressure test (in Wittke, 1999)

 500

400

300
σ (kN/m )
2

200

100

0
0 5 10 15
t (horas)
sx sy sz FLAC

Figure 5. Swelling pressure test simulation with FLAC

 -5

-4

-3
εz (%)
q

-2

-1

0
0 5 10 15
t (días)
sz = 520 kN/m2 sz = 260 kN/m2 sz = 130 kN/m2 sz = 65 kN/m2 sz = 32.2 kN/m2
FLAC sz=520 kPa FLAC sz=260 kPa FLAC sz=130 kPa FLAC sz=65 kPa FLAC sz=32.2 kPa

Figure 7. Huder-Amberg swelling test in elasticity simulated with FLAC

A FLAC model with 1 element has been set up; the The elastic constants of the material are E= 2800
constitutive model is elastic with E= 1000 MPa and MPa and ν= 0.33, and swelling parameters are Kq=
ν= 0.33. The results from FLAC simulation are 6.4% and σ0= 89.2 MPa.
presented in Figure 7 (strain in % versus time in Wittke (1999) suggests that it is necessary to
days) against the results of the test. consider the plastic strain of the sample to reproduce
this test, as it is shown in Figure 8 where the back-
analysis with elastic-plastic stress-strain relationship
4.3 Huder-Amberg swelling test in plasticity fits better with the measured values than the back-
analysis with elastic stress-strain relationship. The
The objective now is to validate the algorithm plastic constants are c’= 0, φ= 11º and ψ= 5.5º.
against a test in which the strength of the sample is Figure 9 presents the results from the FLAC
exceeded and therefore, plastic strains develop in model, also for elastic and elastic-plastic behaviors,
addition to the elastic and swelling deformations. together with the results from the tests and the
The test was carried out following the Huder- values fitted by Wittke (1999).
Amberg procedure. The initial vertical load (applied
in two cycles) is 15 MPa; the sample is then flooded
and unloaded to a vertical pressure of 6.5 MPa. The 4.4 Combined swelling pressure and swelling strain
results for this test are presented in Figure 8. test

This is a swelling test on a sample taken from

Gypsum Keuper. The test was carried out in a
confined compression test apparatus (horizontal
strains of the sample were prevented during the test)
with boundary conditions which were variable with
time for a period of more than 14 years. The test
sequence and the test results are presented in Figure
10.
The description of the test, taken from Wittke
(1999) is as follows. Phase 1 may be divided into 4
partial stages, form 1a to 1d. In phase 1a strains in
the vertical direction were also prevented and the
vertical stress was monitored. After 2.7 years, a
vertical stress of 4.2 MPa was reached. Starting
Figure 8. Huder-Amberg swelling test in plasticity (in Wittke, phase 1b it was allowed for a small vertical strain
1999)
(that is not recognizable in Figure 10 because of the
chosen scale) that results in a reduction of the
vertical stress to approximately 3.8 MPa.
 -24

-20

-16

-12
ez (%)

-8

-4

4
0.1 1 sz (MPa) 10 100

Ensayo: Carga-descarga inicial Ensayo: hinchamientos

Wittke elástico Wittke plástico
FLAC elástico FLAC plástico

Figure 9. Huder-Amberg swelling test in plasticity simulated with FLAC

function of time. Phase 2 was stopped after a vertical

strain of approximately 28% had occurred without
stabilization of the deformations.
During the phase 3 of the test, a further increase
of the vertical strain was prevented and the increase
of the vertical stress was registered as a function of
time. The slope of the stress-strain curve decrease
continuously with time and after a period of 5.7
years a vertical stress of 4 MPa was measured.
Wittke (1999) reproduced this test with the
elastic-plastic properties presented in Table 2:
Table 2. Elastic-plastic parameters used in Wittke (1999) to
reproduce the test.
____________________________________________
Parameter Elastic Plastic
___________ ______________
Figure 10. Combined swelling pressure and swelling strain test
on a sample from Gypsum keuper (in Wittke, 1999) E (MPa) ν c(MPa) φ(º) ψ(º)
____________________________________________
Gypsum keuper 4000 0.2 0.65
____________________________________________30 30

Following to this, the vertical deformation of the For the swelling parameters, Wittke (1999) uses
sample was again prevented, and consequently, the the following values, σ0= 16 MPa and kq= 15%.
vertical stress increased again to the same value of Nevertheless, regarding on the kinetics of the
4.2 MPa. swelling, Wittke found necessary to change the
The course of phase 1c was equivalent to the one swelling time parameter during the course of the
of phase 1b. During phase 1d it was allowed for a test. The parameters proposed are presented in Table
vertical strain slightly larger than during the 3.
preceding phases. At the beginning this led to
decrease the vertical stress to less than 0.5 MPa. Table 3. Swelling kinetics parameters used in Wittke (1999) to
Subsequently, the vertical stress was increased to 2.5 reproduce the test.
________________________________________________
MPa over a period of 0.3 year without stabilization Kinetic of the swelling
_____________________________________
of the vertical stress. a0(year-1) ael(year-1) avp(year-1) maxEVP(%)
________________________________________________
During phase 2 of the test, the vertical stress was 1a - 2 0.0018 0.0 40.0 0.1
lowered to 0.5 MPa keeping it constant for more 3 0.0018 0.0 2.0 0.1
________________________________________________
than 5 years. The vertical strain was measured as a
 5 CONLUSIONS
As Wittke (1999) refers, to reproduce accurately
the phase 3 of the test it is necessary to reduce the The formulation for the swelling behavior presented
value of the coefficient avp from 40 year-1 to 2 year-1, by Wittke (1999) has been reviewed and a
that is equivalent to a reduction of the permeability calculation algorithm, based in this formulation, has
of the sample during phase 3 due to the increment of been implemented in FLAC, via FISH routines. This
the vertical stress in this phase of the test. algorithm allows the simulation of the swelling
Changes in permeability of the sample during the behavior with FLAC code.
load process are not taken into account in the The algorithm has been checked against different
formulation of the kinetic, thus Wittke (1999) swelling tests presented by Wittke (1999), and the
suggests that the coefficient avp should vary during results from the model fit quite well to the results of
the calculation for an accurate simulation of the the different tests. Therefore, these routines can be
swelling process. used to simulate the swelling behavior of expansive
The test described in this paragraph has been grounds in real engineering problems.
simulated with FLAC, considering the same Nevertheless, when using these routines to
parameters (in Table 2 and Table 3). Figure 11 simulate a swelling behavior, the following
presents the evolution of the vertical stress (in MPa) limitations of the formulation have to be
versus time (in years), comparing the results from remembered:
FLAC simulation against the test. − The direction of the principal stresses does not
Figure 12 presents the evolution of the vertical change during the swelling process.
strain (in %) versus time (in years), comparing the − All the elements of the model are susceptible to
results from FLAC model against the test. swell; this means that the whole rockmass is
saturated and the penetration of water is enough
5
to permit the complete swelling of the anhydrite.
− The proposed kinetics describe adequately the
4 swelling process when the strength of the rock is
not exceeded, but the parameters of the
formulation need to be changed in case of large
3
plastic deformations.
σz(MPa)

6 REFERENCES
1
ISRM 1989. Suggested Methods for Laboratory Testing of
Argillaceous Swelling Rocks. In Int. J.Rock Mech. Min.
Sci. & Geomech. Abstr, Vol. 26, No. 5:414-426.
0
0 5
t (años)
10 15 Huder J. & G. Amberg 1970. Quellung in Mergel, Opalinus-
Ensayo FLAC ton und Anhydrit. Schweizer, Bauzeit, 83: 975-980.
Figure 11. Combined swelling pressure-strain test simulated Noorany, I., Frydman S. & Detournay C. 1999. Prediction of
with FLAC; comparison of stresses. soil slope deformation due to wetting, In Detournay & Hart
(eds), FLAC and Numerical Modeling in Geomechanics:
40 101-107. Rotterdam: Balkema.
Saïta A., Robert A. & Le Bissonnais H. 1999. A Simplified
Finite Element Approach to Modeling Swelling Effects in
Tunnels. In Alten et al. (eds), Challenges for the 21st
30 Century: 171-178. Rotterdam: Balkema.
Steiner W. 1993. Swelling Rock in Tunnels: Rock
Characterization, Effect of Horizontal Stresses and
Construction Procedures. In Int. J.Rock Mech. Min. Sci. &
εz(%)

20
Geomech. Abstr. Vol. 30. No. 4: 361-380.
Witke W. 1999. Stability Analysis for Tunnels. Fundamentals.
Geotechnical Engineering in Research and Practice. WBI-
10
Print 4. Ed. WBI Prf.Dr.Ing. W. Wittke. Consulting
engineers for Foundation and Construction in Rock Ltd.
Verlag Glückauf GmbH. Essen.
Witke W. 1990. Rock Mechanics. Theory and Applications
0 with Case Histories. Springer-Verlag, New-York.
0 5 10 15
t (años)
Ensayo Cálculo FLAC

Figure 12. Combined swelling pressure-strain test simulated

with FLAC; comparison of strains.