Will My Polymers Mix?

: Methods for
Studying Polymer Miscibility
Adapted from a Review published in Trends in Polymer Science, Volume 2,
number 8, August 1994 by F.H. Case and J.D. Honeycutt

Summary
A number of different modeling techniques are available for studying polymer
miscibility, ranging from quantitative structure property relationship
methods and simple lattice based models, through atomistic modeling, to
state-of-the-art statistical mechanics approaches. A survey of several
available methods is presented along with example applications. The
strengths and weaknesses of each approach are highlighted. Often the most
insight is obtained by using a combination of different modeling
techniques,incorporating experimental data where available. Although the
computer cannot yet be used as a black box giving a simple "yes" or "no" to
the question "Will my polymers mix?", appropriate use by an informed
polymer scientist of the techniques now available can give considerable
insight into the behavior of polymer mixtures.

Introduction
The design and synthesis of new polymers for specific applications is a time
consuming and costly process. A highly successful alternative is to blend
existing polymers to obtain a balance, or even an enhancement, of the
desired properties that are exhibited by the individual components. Questions
about what actually happens when materials interact on the molecular level
cannot be answered by experiments alone. Liquid/liquid blends have been
the subject of many years of theoretical research 1-5. However, solubility
relations in polymer systems are more complex than those among lowmolecular weight compounds 6. One of the most promising areas for polymer
modeling is in the application of these theories combined with simulations, to
the study of polymer/polymer or polymer/solvent interactions.
Here, we consider a range of different modeling techniques for studying
polymer miscibility:
Using Solubility Parameters
Prediction of solubility parameters using Quantitative Structure Property

Relationship (QSPR) methods and their application in the Flory-Huggins

theory and in the "repulsion" theory of copolymer blends. This approach is
used to predict miscibility behavior for simple systems not dominated by
specific interactions (such as hydrogen bonding), or by non- combinatorial
entropy effects.
Studying local interactions using Molecular Mechanics
Use of simple atomistic calculations on realistic samples of polymer fragment
pair configurations to predict interaction energies and estimate enthalpies of
mixing for polymer blends, including the effects of specific interactions.
Analysis of bulk atomistic simulations
Analysis of bulk atomistic simulations of blend and single component systems
to calculate energies of mixing directly, without use of an adjustable
coordination number.
Application of statistical mechanics
Application of the statistical mechanical Polymer Reference Interaction Site
Model (PRISM) to predict blend structure and phase diagrams including
enthalpy, and both combinatorial and non-combinatorial entropy effects (e.g.,
packing effects).
Putting it all together to predict the phase diagram
Application of a generalized Flory-Huggins theory theory to extrapolate
results from theoretical and/or experimental work on model systems to
predict the effects of temperature, concentration, molecular weight, and
molecular weight distribution on blend miscibility and modes of phase
separation.
The list is not exhaustive, but it does sample the major classes of
computational methods for treating polymer miscibility. The methods
considered deal primarily with the thermodynamics of mixing. Of course,
whether a material is blended or not also depends on its processing history -the polymers may be in a non-equilibrium state 7. But, before a polymer
scientist attempts to blend two polymers to create a new material it is useful
to know, based on thermodynamics, if there is a good chance of success, i.e.
- "will my polymers mix?"

Theoretical Background
The miscibility of two polymers is determined by the free energy of mixing
which includes both entropic and enthalpic terms.

Flory-Huggins theory2 is the classical theory for calculating the free energy of
mixing. Originally derived for small molecule systems, it assumed that each
molecule occupied one site on a lattice. The theory was expanded to model
polymer systems by assuming that the polymer consisted of a series of
connected segments each of which occupied one lattice site. Assuming that
the segments are randomly distributed, the free energy of mixing per mole of
lattice sites for a mixture of polyA and polyB is:

Where R is the gas constant, phi(i)the volume fraction of polymer "i" and chi
the Flory Huggins interaction parameter. The first two terms on the right hand
side of equation 2 represent the combinatorial contribution to the entropy of
mixing, which is derived by calculating all the differnt ways that chains can
randomly pack on a lattice. In the original theory, the entropy of mixing was
assumed to be purely combinatorial, but it has been known for many years
now that non-combinatorial contributions are also important in polymer
systems. These are variously known as free volume effects, equation of state
effects, and packing effects. Once the Flory Huggins chi parameter (which, in
classical FH theory, includes only the enthalpic contibution to the free energy
of mixing) is known, the entire phase behavior of the system can be
calculated. This includes the binodal curve, which is the equilibrium boundary
between the one phase (miscible) and two phase (immiscible) regions, and
the spinodal curve, which gives the limit of stability of the one phase system.
The location of a blend system on the phase diagram determines both
whether it will phase separate, and the mode of phase separation if it does
(either nucleation and growth or spinodal decomposition). The following
phase diagram is for mixtures of deuterated and non-deuterated
polybutadiene. Systems in region1 are miscible; systems in region2, between
the binodal and spinodal curves, are metastable and are predicted to phase
separate by a process of nucleation and growth; systems in region 3 will
spontaneously phase separate.

In the original formulation of Flory-Huggins theory, chi is strictly proportional

to 1/T, and can only give rise to phase diagrams with upper critical solution
temperatures (UCSTs). It has long been known that polymer systems exhibit a
much richer range of phase behaviors than this. The original form is
inadequate to account quantitatively, or even qualitatively, for many
experimentally observed polymer blend phase diagrams. The simple
modification equating chi with a + b/T improves agreement with experiment
and also allows prediction of lower critical solution temperatures (LCSTs). But
even this form is not adequate to describe all the types of experimentally
observed phase diagrams, and chi must be further modified if the classic
form of the FH equation is to be retained. Below, we describe a wide range of
techniques which vary greatly in their degree of sophistication and
computational requirements. In deciding among the available techniques, a
polymer scientist should be aware of the strengths and weaknesses of each,
and the type of system to which it applies

Solubility Parameter -- Flory-Huggins theory Theory

Approach
The solubility parameter of a liquid, delta, is defined as (Ecoh/V)1/2 where
Ecoh is the cohesive energy. According to Hildebrand1, the Flory-Huggins
theory interaction parameter 8, can be estimated from the solubility
parameters of the components of a polymer blend as follows:

where Vseg is the volume of one polymer segment, and deltaA and deltaB
are solubility parameters for components A and B. (The choice of Vseg is
arbitrary as long as one consistently uses the same value to calculate the
number of segments per chain in Equation 2.) Solubility parameters for small

molecule solvents can be obtained reliably from experiment. However, there

is a relatively small amount of polymer data, and reported values vary
greatly 9.
A popular alternative is the use of Quantitative Structure Properties
Relationship (QSPR) methods. The traditional group additivity approaches,
for example those developed by van Krevelen 10 and Fedors 11, require a
database of group contributions for each functional group in the polymer.
Recently, a new method based on connectivity indices has been
developed 12. This allows prediction of solubility parameters for polymers of
arbitrary structure, removing one major restriction on the use of solubility
parameters.
The major limitation of the above approach to calculating chi is that is is
based on the traditional Flory-Huggins theory theory. This theory does not
include any structural correlations between polymer groups. All segments are
assumed to be the same size and shape and it is assumed that there is no
volume change on mixing. Non-combinatorial entropy is ignored.
Concentration dependence of chi cannot be predicted from solubility
parameters, and only UCST phase behavior can be predicted. Using solubility
parameters of the individual components to estimate miscibility also assumes
that specific interactions, such as hydrogen bonds, are neither created nor
destroyed as the blend is formed. This limitation can be lifted, to a certain
degree, by the addition of a semi-empirical specific interaction term to the
Flory- Huggins equation 13. The main advantage of the solubility parameter
approach is that it gives a rapid (within minutes) estimate of miscibility for
simple systems for which the above assumptions are valid.

Example: Studying the Miscibility of PAN-SAN Blends

We were recently confronted with the following question: 'what is the
maximum styrene content in poly(styrene-co-acrylonitrile) (SAN) for which a
fully miscibile blend with pure polyacrylonitrile (PAN) is possible?' Using the
"repulsion" model of copolymer blends, and assuming random
mixing 14,15,16,17, for a blend of poly(A-co-B) and poly(B-co-C), the effective
chi for a coplymer mixture is:

where x is the volume fraction of A in poly(A-co-B) and y is the fraction of C in

poly(C-co-D). Chi(AB) etc. are the chi parameters for the corresponding
homopolymer blends (polyA + polyB etc.) which are estimated using
solubility parameters (equation 3) Miscibility is then calculated as a function
of SAN copolymer composition. The following graphs show the predicted
SAN/PAN phase diagrams at two different copolymer compositions (SAN
copolymer with 90% acrylonitrile, and SAN with 95% acrylonitrile). Note that
as the amount of styrene in the copolymer is reduced the miscibile region (1)
increases. SAN copolymers with higher acrylonitrile will be more miscible with
pure PAN. The asterisk indicates the critical temperature above which the
system is fully miscible.

The third graph shows these critical temperatures as a function of SAN

copolymer composition for a SAN molecular weight of 53,000. This can be
used to predict the amount of AN required in the SAN copolymer for full
miscibility with pure PAN at a given temperature. For example, at room
temperature (300K) the SAN copolymer must have 94.1% by volume
acrylonitrile.
Results can be compared with data reported by Molau 18 on related systems:
SAN/SAN blends.

For a molecular weight of 106,000 (which is close to that used in Molau's

experiments) the model predicts that the SAN copolymer must contain at
least 95.8% acrylonitrile. Thus if more than 4.2% of the SAN copolymer is PS
it will not mix with PAN. This is in good agreement with the experimental
results with suggest that a SAN/SAN mixture would phase separate if the
difference in styrene content of the two componants was greater than 3.54.5%

Atomistic Calculations On Polymer Fragments

If it is assumed that miscibility is determined by the enthalpy of mixing, that
the enthalpy of mixing is dominated by the local interactions between
segments of the polymer chains, and furthermore that the volume change on
mixing an be ignored, then polymer miscibility can be estimated from
interaction energies calculated from molecular mechanics studies on pairs of
small polymer fragments.
This simple molecular mechanics approach has been proposed by a number
of groups 20-23. In our implementation, which we call Flexiblend, relatively
large polymer segments (typically 10 backbone bonds) are used 24. To obtain
an accurate prediction of the local interaction energies between chain
segments, it is important to use a representative sample of the available
conformations for each chain. These are generated using simulated annealing
or using a Monte Carlo approach based on rotational isomeric state (RIS)
statistical weights 19,25.
To predict the miscibility of polyA with polyB, AA, BB and AB segment
interaction energies must be calculated. To calculate the AA interaction
energy, two conformations of the polyA fragment are selected at random.
These are initially placed so that their centers of mass coincide. One oligomer
is then oriented at a random angle and the centers of mass moved apart until
either the intermolecular energy drops below a predefined threshold
(resulting in an acceptable starting configuration), or they move beyond a
prespecified distance (unacceptable -- the pair is discarded). This process is
repeated many times for AA, BB and AB oligomer pairs. It is then important to
allow the local structure to relax. This is done using either molecular
mechanics minimization or molecular dynamics. The output is analyzed to
obtain EAA, EBB and EAB, and to estimate errors. This is used to estimate D
mix for the polymer blend.

Here, the energy values are energies per unit volume; Z* is the effective
coordination number. This parameter arose originally from the lattice-based
formulation of Flory-Huggins theory theory. Its meaning is rather ill-defined in
continuous space, although efforts have been made to compute it. 23 The
major advantage of this method is that it is relatively quick and easy to use.
(A large number of different configurations for the two polymers can be
explored within hours on a workstation.) It gives a fast estimate of whether
deltaEmix is positive or negative, and of the relative magnitude of deltaEmix
in different systems. Unlike the solubility parameter approach, it includes the
effects of specific interactions on polymer miscibility. However, to get a
quantitative prediction of chi, some allowance must be made for the packing
of the polymer chains, as this will determine the number of AA, AB, and BB
contacts. Z* remains an adjustable parameter in this approach. Noncombinatorial entropy contributions are not calculated. Also, the effects of
energy fluctuations and forcefield quality on the results must be considered,
as in any atomistic method.

Full Atomistic Simulation

In principle, full molecular dynamics simulation of the blending process could
include effects due to: concentration fluctuations, correlations between
segments, packing effects, tacticity and chain configuration, and volume
changes on mixing. But there is a problem. Consider bulk polystyrene (PS) of
molecular weight 100,000 (degree of polymerization of about 1000). The
longest relaxation time in the melt is about 0.1 sec. 26To be able to
distinguish miscible from immiscible systems, a reasonable system size might
be 50 chains (of order 800,000 atoms). Even if it takes only one Cray-second
of CPU time per one femtosecond timestep to simulate such a system, it will
take at least 3 x 106 years of Cray time for the system to equilibrate. Clearly,
then, direct atomistic simulation of polymer mixing for realistic systems is not
feasible. We cannot simply simulate bulk models of polyA and polyB, place
them in contact, and wait to see if they mix.
We cannot directly model the process of polymer mixing or demixing.
However, if realistic models of the mixed and de-mixed systems can be
generated, the differences in their calculated cohesive energy densities can
be used to predict the energy of mixing per unit volume:

Here, phiA and phiB are the volume fractions of the two components in the
blended system.
Note that in contrast to the segment pair approach, a full model of the bulk
polymer blend is used. deltaE mix here is the total energy of mixing per unit
volume, not just a segment- segment energy; thus, an adjustable
coordination number is not needed. The value of chi can be calculated using
an equation similar to Equation 6, but without Z*.

This equation has been added to clarify the original paper.

When performing atomistic calculations on bulk polymer structures and
blends, it is important that realistic models be used. As stated above, it is not
possible on a molecular dynamics time scale for a dense, high molecular
weight, polymer system to equilibrate starting from an arbitrary initial
configuration. Thus, the initial configuration should be as realistic as possible.
If we model an isolated cluster of molecules an appreciable fraction of the
atoms would lie close to the periphery of the system, and significant surface
effects would be evident. Sophisticated building algorithms have been
developed to generate realistic bulk models 19,27,28. The standard approach
is introduce periodic boundary conditions in which the system is considered
to be surrounded on all sides by replicas of itself forming an infinite
macrolattice.

The strength of this bulk simulation approach is that once realistic models
have been generated, they can be analyzed to predict a whole range of
structural, thermodynamic and mechanical properties for the proposed
system19. The technique is not limited to binary polymer blends. More
complex systems can be generated, incorporating several polymers and
small molecules such as solvents, gases or plasticisers. Careful analysis can
give insight into the polymers' behavior and aid intuition as new candidate
molecules are proposed. Insight can also be gained into the behavior of
phase separated systems by modeling the phase boundary using two
dimensional periodic boundary conditions. Atomistic modeling of the bulk
systems may also provide input parameters for equation of state
theories29,30. These can be used to estimate the noncombinatorial entropic
contribution to the free energy of mixing.
The major limitation is that entropic effects cannot be predicted directly. The
effects of energy fluctuations, which can be highly significant on the atomistic

scale, must be considered. To obtain statistically meaningful output, results

must be averaged over a large number of bulk polyA, polyB and blend
configurations. This makes the calculations somewhat CPU intensive; a full
study may require several days, or more, of CPU time on a workstation.

Forcefield Quality
The quality of any atomistic study is, of course, strongly dependent on the
quality of the force field used to describe the system. Validation studies have
been carried out on the Accelrys PCFF forcefield31 comparing predicted
solubility parameters ((Ecoh/V)1/2 ) with experimental values for a wide
range of small molecule solvents and polymers. These have shown that
typical values of the standard error in the mean when calculating solubility
parameters for polymers are 0.2-0.3 (0.02 for low molar mass compounds).
The relative error is about 5% with greater than 90% confidence (better than
0.5% for low molar mass compounds). Low molar mass compounds give
solubility parameters in excellent agreement with experiment. The
agreement with experiment for polymers is fair -- although this situation is
harder to judge because of the wide range in reported experimental results9.
Cohesive energy [Ecoh] = Increase in internal energy per mole of material
when all intermolecular forces are eliminated.

solubility parameter =
A sample of predicted and experimental solubility parameters from forcefield
validation work:
Compound

delta calc (Mpa^1/2)

delta expt(Mpa^1/2)

ethanol

25.16 (0.13)

26.5

acetone

19.43 (0.09)

20.0

methyl ethyl ketone

19.12 (0.14)

19.0

hexane

15.03 (0.08)

14.9

trans decane

17.73 (0.09)

18.0

cis decane

18.21 (0.09)

18.8

tetralin

19.69 (0.06)

20.0

benzene

19.20 (0.22)

18.6 - 18.8

toluene

18.5 (0.09)

18.2 - 18.3

ethyl benzene

18.3 (0.1)

18.0 - 18.1

i-propyl benzene

17.6 (0.06)

17.6 - 18.1

di-N butyl phthalate

20.03 (0.01)

20.2

polystyrene (atactic)

15.2 (0.8)

16.6 - 20.2

polyethylene

16.4 (0.4)

15.7 - 17.1

isotactic pmma

17.3 (1.1)

18.6-26.3 (atactic)

Syndiotactic PMMA

16.3 (1.3)

18.6-26.3 (atactic)

To obtain these results, accurate atomistic models of the bulk amorphous

materials were constructed and fully refined using the pcff forcefield. Each
PBC model contained 1000-1500 unique atoms. Each result was averaged
over 25 independent models. The full method is discussed in both the
InsightII based Polymer User Guide (volume 1) and the Cerius2 based
Materials Science Tutorial Guide.
Our results suggest that cohesive energy densities can be accurately
predicted using molecular mechanics calculations, and that for systems in
which entropy effects are not important, differences in predicted values for
single component and blend systems may provide a good prediction of
miscibility as long as the magnitude of deltaHmix is sufficiently large relative
to the statistical uncertainty. Although we have not yet applied this method to
a wide range of polymer systems, results for PEO/PP and PEO/PAA systems
show good agreement with predictions from other methods and with
experimental observations24.

Statistical Mechanics: PRISM Theory Approach

PRISM is a theory of the structure and thermodynamics of polymer systems in

the liquid state 32-34. It allows one to calculate the properties of polymer
melts and blends, including systems containing copolymers. It is an offlattice, integral equation theory based on the successful reference interaction
site model theory of molecular liquids 33. It is able to calculate phase
diagrams (spinodal curves) while incorporating chemical detail, including
bond lengths and angles, chain conformation statistics (at the rotational
isomeric state level) and nearly atomistic (united atom) potentials.
PRISM theory makes its predictions using single chain structure factors
(obtainable from RIS calculations25) and potential energy functions
(obtainable, in principle, from an atomistic forcefield). An alternative integral
equation theory of polymers has been proposed by Lipson,35 although most
work to date using that theory has been on lattices.
PRISM theory is best suited for studying trends in systems whose phase
behavior depends on polymer microstructure -- including, e.g., tacticity,
monomer distribution in copolymers, and chain branching -- as that
microstructure is changed. For example, PRISM theory calculations will
predict the experimentally observed 36-39 difference in the miscibility of
isotactic and syndiotactic polymethylmethacrylate (PMMA) with PVC.

Predicted Phase Diagram:

The modeling suggests that the difference in miscibility is the result of the
greater "openness" of the syndiotactic chain, arrising from the greater
abundance of trans conformations in the isotactic case 40 In another
study 41 workers at Allied Signal Inc. were able to reproduce trends in
miscibility behavior of two rubber blends (natural rubber with a high vinyl
content polybutadiene rubber, and styrene-butadiene rubber with a low vinyl

content polybutadiene rubber). Predictions of concentration dependence of

the average interchain spacings, as determined by wide-angle X-ray
scattering, were also in good agreement with experimental results.
The ability to compute non- combinatorial entropy effects, not accounted for
by the other methods considered here, is the strength of PRISM theory.
Limitations include the need to obtain site-site (united atom) potentials for
each system studied, and the requirement that an RIS model exist for the
chains in the system in order to calculate their conformational statistics.

Generalized Flory-Huggins Theory Approach

The methods described above can provide varying amounts of data about the
phase behavior of polymeric materials. It may be the case that limited
experimental data, perhaps on low molecular weight model systems, are
available as well. In order to fit these data to predict the full miscibility
behavior of the system the Flory-Huggins theorytheory can be generalized
giving the chi parameter a functional form that better represents its
temperature and concentration dependence, thus capturing both enthalpic
and entropic effects42-44:

Here, chieff is the "effective" chi parameter. In this approach, the form of
the Flory-Huggins theory free energy expression (Equation 2) is retained, but
chi is treated as an adjustable parameter. This allows incorporation of noncombinatorial entropy effects, while leaving the original formalism unaltered.
This approach differs from equation of state theories, 30 in which the original
form of chi is retained, but other terms are added to the free energy
expression to account for these entropy effects. The coefficients in Equation 8
can be obtained from the literature, from the other modeling methods
described in this paper, or by fitting to experimental data. If the parameters
are derived by fitting to experimental data, or data from a PRISM calculation,
then free volume and other non-combinatorial entropy effects are implicitly
included in chieff. Given chieff, the full phase diagram of the
polymer/polymer or polymer/solvent system can be predicted44. This
approach has been shown to model all experimentally observed types of
polymer phase diagram. The effects of temperature, concentration, molecular
weight and molecular weight distribution 45 on miscibility can be predicted,
as well the mode of phase separation for immiscible systems.

Additional examples are given in references 46 and 47

Conclusions
A range of methods is now available for studying and predicting
polymer/polymer and polymer/solvent phase behavior. The scope of the
traditional Flory-Huggins theory/solubility parameter approach has been
extended by development of new methods which allow solubility parameters
to be calculated for polymers of arbitrary structure without a database of
group contributions. However, this approach, although fast and easy to use,
still fails to account for specific interactions, such as hydrogen bonding, on
polymer miscibility.
A method has been developed for estimating the interaction energies
between polymer segments using molecular mechanics (details). This allows
for conformational freedom and averaging over representative polymer
configurations and can provide a fast estimate of the sign and relative
magnitudes of enthalpies of mixing. It can predict the effects of specific
interactions, but still requires an adjustable packing parameter if a value for
chi is to be obtained.
Methods have been developed which allow realistic atomistic models of the
bulk polymer and polymer blend to be readily constructed and analyzed.
Validation studies suggest that forcefields are now of sufficient accuracy for

enthalpies of mixing to be calculated from differences in calculated cohesive

energy densities, as long as the magnitude of deltaHmix is sufficiently large
relative to the statistical uncertainty.
The PRISM approach includes both enthalpic and entropic contributions, and
may be the best method for studying miscibility differences arising from
changes in microstructure. A modified Flory-Huggins theory theory has been
developed and tested. This can take data on the concentration and
temperature dependence of mixing from other modeling methods, or from
limited experimental data on model systems, and fit these data to obtain an
interaction parameter. This can be used to predict the effects of temperature,
concentration, molecular weight and polydispersity on polymer miscibility
and the mode of phase separation in immiscible systems.(details)
Although the computer cannot yet be used as a black box giving a simple
"yes " or "no" to the question "Will my polymers mix?", we believe that by
appropriate use of the range of available techniques an informed polymer
scientist can gain considerable insight into the behavior of polymer mixtures.
The scientist wishing to answer this critical question should make a choice
from the available methods, balancing the effort and computation time
required with the accuracy and approximations made in each. Often the most
insight will be gained by using a combination of different approaches.

References
1) Hildebrand, J.H. and Scott R.L. (1949) "The Solubility of Non-Electrolytes",
Reinhold, New York
2) Flory, P.J. (1953) "Principles of Polymer Chemistry" Cornell University Press
3) Guggenheim, E.A. (1952) "Mixtures" Clarendon Press, Oxford
4) Prigogine, I. and Defay, R. (1954) "Chemical Thermodynamics", Longmars
Green & Co
5) Marcus, Y. (1977) "Introduction to Liquid State Chemistry" John Wiley &
Sons
6) Billmeyer, F.W. Jr, (1984) "Textbook of Polymer Science",Wiley-Interscience
7) Kozlowski, M., (1993) Polymer Networks & Blends, 3(4), 283
8) Flory, P.J. (1970) Discussions of the Faraday Society, 49, 7 (1970).
9)"Polymer Handbook" (1989)J. Brandrup and E.H. Immergut, Editors, Wiley
Interscience.

10) van Krevelen, D.W.(1976) "Properties of Polymers, Their Estimation and

Correlation With Chemical Structure" Elsevier
11) Fedors, R.F. (1974) Polym. Eng. Sci. 14, 147, 472
12) Bicerano, J. (1993), "Prediction of Polymer Properties" Marcel Dekker Inc.
13) Coleman, M.M., Graf J.F., and Painter, P.C. (1991) "Specific Interactions
and the Miscibility of Polymer Blends", Technomic
14) ten Brinke, G., Karasz, F.E., MacKnight, W.J. (1983), Macromolecules. 16,
1827
15) Paul, D.R. and Barlow, J.W. (1984) Polymer, 25, 487
16) Kanbour, R.P., Bendler, J.T., Bopp, R.C. (1983) Macromolecules, 16, 753.
17) Roe, R.J, Rigby, D. (1987) "Phase Relations and Miscibility in Polymer
Blends Containing Copolymers" in "Advances in Polymer Science 82",
Springer Verlag Berlin Heidelberg
18) Molau,G.E. (1965) J. Poly. Sci., Poly. Lett, 3, 1007
19) Polymer User Guide, version 6.0, (1994) Biosym Technologies, San Diego.
20) Hopfinger, A.J., Koehler, M.G. (1993) ACS PMSE Preprints , 69, 43.
21) Nelson, G.V., Jacobson, S.H., Gordon, D.J. (1992), Chem Design Autom.
News, 39.
22) Jacobson, S.H., Gordon, D.J, Nelson, G.V., Balazs, A. (1992) Adv. Materials,
4, 198
23) Fan C.F., Olafson B.D., Blanco M., & Hsu S. L. (1992) Macromolecules, 25,
14, 3667
24) Tiller A.R., Gorella B., "Prediction of Polymer Miscibility From Molecular
Mechanics Calculations" Polymer (in press).
25) Flory, P.J. (1989) "Statistical Mechanics of Chain Molecules" Hanser
26) Ferry, J.D. (1980) "Viscoelastic Properties of Polymers," John WIley & Sons,
New York, p. 388.
27) Theodorou, D.N. and Suter, U.W. (1985) Macromolecules, 18, 1467
28) Brown, D. and Clarke, J.H.R. (1991) Macromolecules, 24, 2775
29) Eichinger, B.E., Flory, P.J. (1968) Transactions of the Faraday Society. 64,
2035;

30) Sanchez, I.C. (1992)."Polymer Compatibility and Incompatibility," MMI

Press Symp. Ser., Vol. 2, K. Solc, ed., Harwood Acad. Publ., New York, p. 59.
31) Maple, J.A., et al, (1994), J. Comp. Chem. 15(2) 162
32) Chandler, D., and Andersen, H.C. (1972) J. Chem. Phys. 57, 1932
33) Schweizer K. S. and Curro, J.G. (1989) J. Chem. Phys. 91, 5059
34) Honeycutt, J.D., (1992) ACS Polymer Preprints 41(1), 657
35) Lipson, J.E.G. (1991) Macromolecules, 24, 1334.
36) Schurer, J.W., de Boer, A., Challa, G. (1975) Polymer, 16, 201.
37) Vanderschueren, J., Janssens, A., Ladang, M., Niezette, J. (1982) Polymer,
23, 395.
38) Vorenkamp, E.J., ten Brinke, G., Meijer, J.G., Jager, H., Challa, G. (1985)
Polymer, 26, 1725.
39) Lemieux, E., Prud'homme, R.E., Forte, R., Jerome, R., Teyssie, P. (1988)
Macromolecules, 21, 2148.
40) Honeycutt, J.D., (1994) Macromolecules (submitted).
41) Ding, K., Williams, B. and Murthy, N.S. (1994) Society of Plastics
Engineers ANTEC
42) Koningsveld, R., Staverman, A.J. (1968) J. Polym. Sci. A2, 6, 305.
43) Koningsveld, R., Kleintjens, L.A., Shultz, A.R. (1970) J. Polym. Sci. A2, 8,
1262.
44) Qian, C., Mumby, S.J. and Eichinger, B.E. (1991), Macromolecules. 24,
2255
45) Mumby, S.J., Qian, C., Eichinger, B.E. (1992) Polymer, 33(23) 5105
46) Mumby, S.J. and Sher, P. (1994) Macromolecules 27(3) 689
47) Case, F.H. and Honeycutt, J.D. (1994) Trends in Polymer Science, 2(8),
259-266

Solubility Parameters: Theory and Application

John Burke

Solvents are ubiquitous: we depend on them when we apply pastes

and coatings, remove stains or old adhesives, and consolidate flaking
media. The solubility behavior of an unknown substance often gives
us a clue to its identification, and the change in solubility of a known
material can provide essential information about its ageing
characteristics.
Our choice of solvent in a particular situation involves many factors,
including evaporation rate, solution viscosity, or environmental and
health concerns, and often the effectiveness of a solvent depends on
its ability to adequately dissolve one material while leaving other
materials unaffected. The selection of solvents or solvent blends to
satisfy such criterion is a fine art, based on experience, trial and error,
and intuition guided by such rules of thumb as "like dissolves like" and
various definitions of solvent "strength". While seat-of-the-pants
methods are suitable in many situations, any dependence on
experiential reasoning at the expense of scientific method has
practical limitations. Although it may not be necessary to understand
quantum mechanics to remove masking tape, an organized system is
often needed that can facilitate the accurate prediction of complex
solubility behavior.
Solubility Scales

Product literature and technical reports present a bewildering

assortment of such systems: Kaouri-Butanol number, solubility grade,
aromatic character, analine cloud point, wax number, heptane
number, and Hildebrand solubility parameter, among others. In
addition, the Hildebrand solubility parameter, perhaps the most widely
applicable of all the systems, includes such variations as the
Hildebrand number, hydrogen bonding value, Hansen parameter, and
fractional parameter, to name a few. Sometimes only numerical values
for these terms are encountered, while at other times values are

presented in the form of two or three dimensional graphs, and a

triangular graph called a Teas graph has found increasing use
because of its accuracy and clarity.
Understandably, all this can be slightly confusing to the uninitiated.
Graphic plots of solvent-polymer interactions allow the fairly precise
prediction of solubility behavior, enabling the control of numerous
properties in practical applications that would be very difficult without
such an organizing system. Yet the underlying theories are often
extremely complex, and an understanding of the "why" of a particular
system can be very difficult, enough to discourage the use of such
systems. Many of the systems mentioned, however, are actually quite
simple (this is especially true of the Teas graph) and can be used to
advantage with little understanding of the chemical principles at work.
This paper will attempt to bridge these two realities by briefly
introducing solubility theory as well as its application so that the
conservator will be both better able to understand and profitably apply
the concepts involved. The discussion will center on Hildebrand
solubility parameters and, after laying a theoretical foundation, will
concentrate on graphic plots of solubility behavior. It should be
remembered that these systems relate to non-ionic liquid interactions
that are extended to polymer interactions; water based systems and
those systems involving acid-base reactions cannot be evaluated by
simple solubility parameter systems alone.
Solutions and Molecules

A solvent, usually thought of as a liquid, is a substance that is

capable of dissolving other substances and forming a uniform mixture
called a solution. The substance dissolved is called the solute and is
usually considered to be the component present in the smallest
amount. According to this definition, an almost-dry or slightly swollen
resin film comprises a solution of a liquid (the solute) in a resin (the

solvent), even though conventionally the liquid is usually referred to as

the solvent, and the resin as the solute.
Molecular Attractions

Liquids (and solids) differ from gases in that the molecules of the
liquid (or solid) are held together by a certain amount of intermolecular
stickiness. For a solution to occur, the solvent molecules must
overcome this intermolecular stickiness in the solute and find their
way between and around the solute molecules. At the same time, the
solvent molecules themselves must be separated from each other by
the molecules of the solute. This is accomplished best when the
attractions between the molecules of both components are similar. If
the attractions are sufficiently different, the strongly attracted
molecules will cling together, excluding the weakly attracted
molecules, and immiscibility (not able to be mixed) will result. Oil
and water do not mix because the water molecules, strongly attracted
to each other, will not allow the weakly, attracted oil molecules
between them.
Van der Waals Forces

These sticky forces between molecules are called van der Waals
forces (after Johannes van der Waals who first described them in
1873). Originally thought to be small gravitational attractions, Van der
Waals forces are actually due to electromagnetic interactions between
molecules.
The outer shell of a neutral atom or molecule is composed entirely
of negatively charged electrons, completely enclosing the positively
charged nucleus within. Deviations in the electron shell density,
however, will result in a minute magnetic imbalance, so that the
molecule as a whole becomes a small magnet, or dipole. These
electron density deviations depend on the physical architecture of the
molecule: certain molecular geometries will be strongly polar, while
other configurations will result in only a weak polarity. These

differences in polarity are directly responsible for the different degrees

of intermolecular stickiness from one substance to another.
Substances that have similar polarities will be soluble in each other
but increasing deviations in polarity will make solubility increasingly
difficult.
Van der Weals forces, then, are the result of intermolecular
polarities. As we shall see, accurate predictions of solubility behavior
will depend not only on determining the result of intermolecular
attractions between molecules, but in discriminating between
different typesof polarities as well. A single molecule, because of its
structure, may exhibit van der Waals forces that are the additive result
of two or three different kinds of polar contributions. Substances will
dissolve in each other not only if their intermolecular forces are
similar, but particularly if their composite forces are made up in the
same way. (Such types of component interactions include hydrogen
bonds, induction and orientation effects, and dispersion forces, which
will be discussed later.)
The Hildebrand Solubility Parameter

It is the total van der Waals force, however, which is reflected in the
simplest solubility value: the Hildebrand solubility parameter. The
solubility parameter is a numerical value that indicates the relative
solvency behavior of a specific solvent. It is derived from thecohesive
energy density of the solvent, which in turn is derived from the heat
of vaporization. What this means will be clarified when we
understand the relationship between vaporization, van der Waals
forces, and solubility.
Vaporization

When a liquid is heated to its boiling point, energy (in the form of
heat) is added to the liquid, resulting in an increase in the temperature
of the liquid. Once the liquid reaches its boiling point, however, the
further addition of heat does not cause a further increase in

temperature. The energy that is added is entirely used to separate the

molecules of the liquid and boil them away into a gas. Only when the
liquid has been completely vaporized will the temperature of the
system again begin to rise. If we measure the amount of energy (in
calories) that was added from the onset of boiling to the point when all
the liquid has boiled away, we will have a direct indication of the
amount of energy required to separate the liquid into a gas, and thus
the amount of van der Waals forces that held the molecules of the
liquid together.
It is important to note that we are not interested here with
the temperature at which the liquid begins to boil, but the amount of
heatthat has to be added to separate the molecules. A liquid with a
low boiling point may require considerable energy to vaporize, while a
liquid with a higher boiling point may vaporize quite readily, or vise
versa. What is important is the energy required to vaporize the liquid,
called the heat of vaporization. (Regardless of the temperature at
which boiling begins, the liquid that vaporizes readily has less
intermolecular stickiness than the liquid that requires considerable
addition of heat to vaporize.)
Cohesive Energy Density

From the heat of vaporization, in calories per cubic centimeter of

liquid, we can derive the cohesive energy density (c) by the
following expression

where:
c=Cohesive energy density
h=Heat of vaporization
r=Gas constant

t=Temperature
Vm = Molar volume
In other words, the cohesive energy density of a liquid is a numerical
value that indicates the energy of vaporization in calories per cubic
centimeter, and is a direct reflection of the degree of van der Waals
forces holding the molecules of the liquid together.
Interestingly, this correlation between vaporization and van der
Waals forces also translates into a correlation between vaporization
and solubility behavior. This is because the same intermolecular
attractive forces have to be overcome to vaporize a liquid as to
dissolve it. This can be understood by considering what happens
when two liquids are mixed: the molecules of each liquid are
physically separated by the molecules of the other liquid, similar to the
separations that happen during vaporization. The same intermolecular
van der Waals forces must be overcome in both cases.
Since the solubility of two materials is only possible when their
intermolecular attractive forces are similar, one might also expect that
materials with similar cohesive energy density values would be
miscible. This is in fact what happens.
Solubility Parameter

In 1936 Joel H. Hildebrand (who laid the foundation for solubility

theory in his classic work on the solubility of nonelectrolytes in 1916)
proposed the square root of the cohesive energy density as a
numerical value indicating the solvency behavior of a specific solvent.

It was not until the third edition of his book in 1950 that the
term "solubility parameter" was proposed for this value and the
quantity represented by the symbol . Subsequent authors have

proposed that the term hildebrands be adopted for solubility

parameter units, in order to recognize the tremendous contribution
that Dr. Hildebrand has made to solubility theory.
Units of Measurement

Table 1 lists several solvents in order of increasing Hildebrand

parameter. Values are shown in both the common form which is
derived from cohesive energy densities in calories/cc, and a newer
form which, conforming to standard international units (SI units), is
derived from cohesive pressures. The SI unit for expressing pressure
is the pascal, and SI Hildebrand solubility parameters are expressed
in mega-pascals (1 mega-pascal or mpa=I million pascals).
Conveniently, SI parameters are about twice the value of standard
parameters:
Table I. Hildebrand Solubility Parameters
Standard Hildebrand values from Hansen. Journal of Paint
TechnologyVol. 39, No. 505, Feb 1967
Sl Hildebrand values from Barton.Handbook of Solubility
Parameters, CRC Press, 1983
Values in parenthesis from Crowley. et al., Journal of Paint
Technology Vol. 38, No. 496. May 1966
solvent

(SI)

n-Pentane

(7.0)

14.4

n-Hexane

7.24

14.9

Freon TF

7.25

n-Heptane

(7.4)

15.3

Diethyl ether

7.62

15.4

1,1,1 Trichloroethane

8.57

15.8

n-Dodecane

16.0

White spirit

16.1

Turpentine

16.6

Cyclohexane

8.18

16.8

Amyl acetate

(8.5)

17.1

Carbon tetrachloride

8.65

18.0

Xylene

8.85

18.2

Ethyl acetate

9.10

18.2

Toluene

8.91

18.3

Tetrahydrofuran

9.52

18.5

Benzene

9.15

18.7

Chloroform

9.21

18.7

Trichloroethylene

9.28

18.7

Cellosolve acetate

9.60

19.1

Methyl ethyl ketone

9.27

19.3

Acetone

9.77

19.7

Diacetone alcohol

10.18

20.0

Ethylene dichloride

9.76

20.2

Methylene chloride

9.93

20.2

Butyl Cellosolve

10.24

20.2

Pyridine

10.61

21.7

Cellosolve

11.88

21.9

Morpholine

10.52

22.1

Dimethylformamide

12.14

24.7

n-Propyl alcohol

11.97

24.9

Ethyl alcohol

12.92

26.2

Dimethyl sulphoxide

12.93

26.4

n-Butyl alcohol

11.30

28.7

Methyl alcohol

14.28

29.7

Propylene glycol

14.80

30.7

Ethylene glycol

16.30

34.9

Glycerol

21.10

36.2

Water

23.5

48.0

/calcm-3/2 =

0.48888 x /MPa1/2

(3)

/MPa =

2.0455 x /calcm-3/2

(4)

Literature published prior to 1984 should contain only the common

form, designated , and it is hoped that where the newer SI units are
used, they are designated as such, namely /MPa or (SI).
Obviously, one must be careful to determine which system of
measurement is being used, since both forms are called Hildebrand
parameters. This paper will primarily use the SI values, and the use of
standard values will be noted.
Solvent Spectrum

In looking over Table 1, it is readily apparent that by ranking

solvents according to solubility parameter a solvent "spectrum" is
obtained, with solvents occupying positions in proximity to other
solvents of comparable "strength". If, for example, acetone dissolves a
particular material, then one might expect the material to be soluble in
neighboring solvents, like diacetone alcohol or methyl ethyl ketone,
since these solvents have similar internal energies. It may not be
possible to achieve solutions in solvents further from acetone on the
chart, such as ethyl alcohol or cyclohexaneliquids with internal
energies very different from acetone. Theoretically, there will be a
contiguous group of solvents that will dissolve a particular material,
while the rest of the solvents in the spectrum will not. Some materials
will dissolve in a large range of solvents, while other might be soluble
in only a few. A material that cannot be dissolved at all, such as a

crosslinked three-dimensional polymer, would exhibit swelling

behavior in precisely the same way.
Solvent Mixtures

It is an interesting aspect of the Hildebrand solvent spectrum that

the Hildebrand value of a solvent mixture can be determined by
averaging the Hildebrand values of the individual solvents by volume.
For example, a mixture of two parts toluene and one part acetone will
have a Hildebrand value of 18.7 (18.3 x 2/3 + 19.7 x 1/3), about the
same as chloroform. Theoretically, such a 2:1 toluene/acetone mixture
should have solubility behavior similar to chloroform. If, for example, a
resin was soluble in one, it would probably be soluble in the other.
What is attractive about this system is that it attempts to predict the
properties of a mixture a priori using only the properties of its
components (given the solubility parameters of the polymer and the
liquids); no information on the mixture is required.

Fig. I Swelling of Linseed Oil Film in Solvents Arranged According to

Solubility Parameter (adapted from Feller, Stolow, and Jones,On
Picture Varnishes and Their Solvents)
Polymer Cohesion Parameters

Figure 1 plots the swelling behavior of a dried linseed oil film in

various solvents arranged according to Hildebrand number. Of the
solvents listed, chloroform swells the film to the greatest degree,
about six times as much as ethylene dichloride, and over ten times as
much as toluene. Solvents with greater differences in Hildebrand
value have less swelling effect, and the range of peak swelling
occupies less than two hildebrand units. By extension, we would
expect any solvent or solvent mixture with a Hildebrand value
between 19 and 20 to severely swell a linseed oil film. (The careful
observer will notice certain inconsistencies in Fig. 1 which will be
discussed later.)
Since a polymer would. decompose before its heat of vaporization
could be measured, swelling behavior is one of the ways that
Hildebrand values are assigned to polymers (the general
term cohesion parameter is often preferred to the term solubility
parameter when referring to non-liquid materials). Another method
involves cloud-point determinations in which a resin is dissolved in a
true solvent and titrated with another solvent until the mixture
becomes cloudy, thus identifying the range of solubility. Testing cloudpoints with a variety of solvents and diluents enable a precise
determination of cohesion parameter values for polymers. Other
methods include a combination of empirical tests, such as cloud-point
and solubility/swelling tests, with the addition of theoretical
calculations based on comparing chemical structure to other materials
of known Hildebrand value.
Other Practical Solubility Scales

Similar empirical methods have been used to develop other

solubility scales, unrelated to the Hildebrand parameter, that quantify
solvent behavior. Many of these other systems have been developed
for particular applications and are appropriate for use in those
applications but, although agreement between unrelated systems is

somewhat loose, it is possible to correlate most of these other

systems to the Hildebrand parameter. While such correlations are not
always practicable, it does support the Hildebrand theory as a unifying
approach, and allows the translation of solubility information into
whatever system is best for the application at hand.
Kauri-Butanol Value

A particularly common cloud-point test for ranking hydrocarbon

solvent strength is the Kauri-Butanol test. The kauri-butanol value
(KB) of a solvent represents the maximum amount of that solvent that
can be added to a stock solution of kauri resin (a fossil copal) in butyl
alcohol without causing cloudiness. Since kauri resin is readily soluble
in butyl alcohol but not in hydrocarbon solvents, the resin solution will
tolerate only a certain amount of dilution. "Stronger" solvents such as
toluene can be added in a greater amount (and thus have a higher KB
value) than "weaker" solvents like hexane.

Fig. 2 Relationship Between Kauri-Butanol Number and Hildebrand

Parameter
Figure 2 illustrates an almost direct relationship between KB values
and Hildebrand values. This relationship is linear for solvents with KB
values greater than 35 and can be expressed:

/MPa= 0.04 KB + 14.2 (5)

For aliphatic hydrocarbons with KB values less than 35, the
relationship, while also linear, involves calculations that include
corrections for molecular size.
Solubility Grade

While the Kauri-Butanol test measures the relative strength of a

solvent, another cloud-point test, developed by the National Gallery of
Art Research Project, is used to determine the Solubility Grade of a
polymer. In this test, 10% mixtures of the polymer in n-dodecane ( an
aliphatic hydrocarbon, boiling point 213C) are diluted with varying
percentages of toluene. The Solubility Grade of the polymer is the
minimum percent of toluene needed to give a clear solution, thus
indicating the strength of the solvent needed to dissolve the polymer.
The higher the percentage of toluene in the blend, the "stronger" is the
solvent strength of the blend; the Solubility Grade is therefore the
mildest blend that can be used to dissolve the polymer. Table 2 gives
the Solubility Grades of several polymers, along with the
corresponding Hildebrand number (SI) of the toluene-dodecane
solution.
Table 2. Solubility Grades (% Toluene) of Polymers at 10%
solids with Hildebrand Values of the Corresponding TolueneDodecane blends
Polymer

Solubility Grade

/MPa

Poly vinyl acetate

89

18.05

Poly methyl methacrylate

87

18.00

Acryloid B-72 (Rohm and Haas) 80

17.84

Poly n-butyl methacrylate

25

16.58

Poly isobutyl methacrylate

23

16.53

Acryloid B-67 (Rohm and Haas) 18

16.41

Resin AW-2

16.05

44

Although the Solubility Grade gives us a conveniently broad scale

for judging the solubility of polymers in mild solvents, the Hildebrand
value provides a slight additional advantage: the ability to assess the
solubility of the polymer in solvent blends other than toluenedodecane. To do this, the ratio is calculated of the relative
contributions of the two new solvents in terms of their distance (in
Hildebrand units) from the Hildebrand value of the polymer Solubility
Grade. In this way we might determine that poly isobutyl methacrylate
should form clear solutions above I0% solids in a solvent of heptane
containing at least 42% xylene ( 16.53 -15.3)+(18.2 - 15.3). While the
principle here is sound, it should be noted that the fine divisions
between Hildebrand values in this instance can only give approximate
results.
Other Solubility Scales

Other empirical solubility scales include the aniline cloudpoint (aniline is very soluble in aromatic hydrocarbons, but only
slightly soluble in aliphatics), the heptane number (how much heptane
can be added to a solvent/resin solution), the wax number (how much
of a solvent can be added to a benzene/beeswax solution), and many
others. The aromatic character of a solvent is the percent of the
molecule, determined by adding up the atomic weights, that is
benzene-structured (benzene is the simplest hexagonal aromatic
hydrocarbon). Benzene therefore has 100% aromatic character,

toluene 85%, and diethyl benzene 56% aromatic character. By loose

extension, the aromatic character of a mixed solvent, such as V.M.
and P. naphtha or mineral spirits, is the percent of aromatic solvent in
the otherwise aliphatic mixture.
These diverse solubility scales are useful because they give concise
information about the relative strengths of solvents and allow us to
more easily determine what solvents or solvent blends can be used to
dissolve a particular material. Because most of these other systems
can be more or less directly related to the Hildebrand solubility
parameter, and because the Hildebrand solvent spectrum
encompasses the complete range of solvents, it is the Hildebrand
solubility parameter that is most frequently encountered in
contemporary technical literature.
Component Polarities

As was mentioned above, there are inconsistencies in Fig. 1 that

are difficult to explain in terms of single component Hildebrand
parameters. The graph shows chloroform and ethylene dichloride
(with Hildebrand values of 18.7 and 20.2 respectively) swelling a
linseed oil film considerably more than methyl ethyl ketone (MEK) and
acetone. And yet the Hildebrand values for MEK and acetone are 19.3
and 19.7, both between the values for the two high swelling solvents.
Theoretically, liquids with similar cohesive energy densities should
have similar solubility characteristics, and yet the observed behavior
in this instance does not bear this out. The reason for this is the
differences in kinds of polar contributions that give rise to the total
cohesive energy densities in each case.
It was mentioned that van der Waals forces result from the additive
effects of several different types of component polarities. The
inconsistencies in Fig. 1 are due to the fact that, while the sum total
cohesive energy densities are similar in the four solvents in question,

the addends that make up those individual totals are different. These
slight disparities in polar contributions result in considerable
differences in solubility behavior. If these component differences are
taken into account, quantified, and included in solubility theory, the
prediction of solubility behavior can become more accurate. To do
this, different types of polar contributions must be examined, and
differentiated.
The following section is an introduction to the three types of polar
interactions that are most commonly used in solubility
theories:dispersion forces, polar forces, and hydrogen bonding
forces. In some systems, the Hildebrand parameter is used in
conjunction with only one or two of these forces (i.e. Hildebrand value
and hydrogen bonding value), while more recent developments
subdivide the Hildebrand parameter into all three forces, or derivatives
of them. The concepts discussed provide an excellent foundation for
understanding the inner workings of the practical systems introduced
later. It should be stressed, however, that it is possible to use these
practical systems without a thorough understanding of the molecular
dynamics on which they are based.
Dipoles and Dipole Moments

Strong electromagnetic forces are present in every atom and

molecule. At the center of a molecule is a positively charged atomic
nucleus, while the outer surface is covered by a dispersed cloud of
negatively charged electrons. These positive and negative charges
balance out, and the molecule as a whole is neutral. If, for reasons we
will investigate, the distribution of the electron cloud is uneven (maybe
thicker in one place and thinner in another), small local charge
imbalances are created: the parts of the molecule with a greater
electron density will be negatively charged, and the electron deficient
parts will be positively charged. The molecule as a whole, while still
neutral, will have the properties of a small magnet, with equal but
opposite poles, called dipoles.

A single molecule, because of its structure, can have several dipoles

at once, some strong and some weak, some which cancel out, and
some which reinforce each other. The resulting sum of all the dipoles
is what is known as the dipole moment of the molecule. Molecules
that have permanent dipole moments are said to be polar, while
molecules in which all the dipoles cancel out (zero dipole moment) are
said to be nonpolar.
This molecular polarity is at the heart of intermolecular attractions
(imagine a pile of small magnets sticking together). The strength with
which the molecules cling together, and therefore the cohesive energy
density and the solubility parameter, is directly related to the strength
of the molecular dipoles. But since the overall polarity of a molecule is
often the combined result of several contributing polar structures, it is
not enough to know the dipole moment of a molecule. The component
polarities must be considered as well. Molecules like to be with other
molecules of their own electromagnetic kind, both in terms of polar
strength and in terms of polar composition.
Dispersion Forces

Nonpolar liquids, such as the aliphatic hydrocarbons, have weak

intermolecular attractions but no dipole moment. Magnets without
poles; how can this be? The source of their electromagnetic
interactions can be described by quantum mechanics, and is a
function of the random movement of the electron cloud surrounding
every molecule. From instant to instant, random changes in electron
cloud distribution cause polar fluctuations that shift about the
molecular surface. Although no permanent polar configuration is
formed, numerous temporary dipoles are created constantly, move
about, and disappear.
When two molecules are in proximity, the random polarities in each
molecule tend to induce corresponding polarities in one another,
causing the molecules to fluctuate together. This allows the electrons

of one molecule to be temporarily attracted to the nucleus of the other,

and vise versa, resulting in a play of attractions between the
molecules. These induced attractions are called London dispersion
forces, or induced dipole-induced dipole forces.
The degree of "polarity" that these temporary dipoles confer on a
molecule is related to surface area: the larger the molecule, the
greater the number of temporary dipoles, and the greater the
intermolecular attractions. Molecules with straight chains have more
surface area, and thus greater dispersion forces, than branched-chain
molecules of the same molecular weight. This dependence on surface
area explains why conversions between Kauri-Butanol numbers and
Hildebrand values for paraffin must include calculations for molecular
size. The intermolecular forces between paraffin molecules are
entirely due to dispersion forces, and are therefore size dependent.
Polar Forces

Dispersion forces are present to some degree in all molecules, but

in polar molecules there are also stronger forces at work. Some
atomic elements attract electrons more vigorously than others, and
permanent dipoles are created when electrons are unequally shared
between the individual atoms in a molecule. If the molecule is
symmetrical, these dipoles may cancel out. If, on the other hand, the
electron density is permanently imbalanced, with some atoms in the
molecule harboring a greater share of the negative charge
distribution, the molecule itself will be polar. The polarity of a molecule
is related to its atomic composition, its geometry, and its size. Water
and alcohol are strongly polar molecules, toluene is only slightly polar,
and the paraffin hydrocarbons such as hexane and Stoddard solvent
are considered to be nonpolar (again, the attractions between
nonpolar molecules are due entirely to dispersion forces).
Polar molecules tend to arrange themselves head to tail, positive to
negative, and these orientations lead to further increases in

intermolecular attraction. These dipole-dipole forces, called

Keesom interactions, are symmetrical attractions that depend on the
same properties in each molecule. Because Keesom interactions are
related to molecular arrangements, they are temperature dependent.
Higher temperatures cause increased molecular motion and thus a
decrease in Keesom interactions.
On the other hand, any molecule, even if nonpolar, will be
temporarily polarized in the vicinity of a polar molecule, and the
induced and permanent dipoles will be mutually attracted. These
dipole-induced dipole forces, called Debye interactions, are not as
temperature dependant as Keesom interactions because the induced
dipole is free to shift and rotate around the nonpolar molecule as the
molecules move. Both Debye induction effects and Keesom
orientation effects are considered similar in terms of solubility behavior
and are collectively referred to as polar interactions or
simply polarities.
Hydrogen Bonding

A particularly strong type of polar interaction occurs in molecules

where a hydrogen atom is attached to an extremely electron-hungry
atom such as oxygen, nitrogen, or fluorine. In such cases, the
hydrogen's sole electron is drawn toward the electronegative atom,
leaving the strongly charged hydrogen nucleus exposed. In this state
the exposed positive nucleus can exert a considerable attraction on
electrons in other molecules, forming a protonic bridge that is
substantially stronger than most other types of dipole interactions.
This type of polarity is so strong compared to other van der Waals
interactions, that it is given its own name: hydrogen bonding.
Understandably, hydrogen bonding plays a significant role in solubility
behavior.
The inconsistencies in Fig. 1 stem from a difference in hydrogen
bonding between the chlorinated solvents and the ketones. The

intermolecular forces in linseed oil are primarily due to dispersion

forces, with practically no hydrogen bonding involved. These polar
configurations are perfectly matched by the intermolecular forces
between chloroform molecules, thus encouraging interpenetration and
swelling of the linseed oil polymer. Acetone and methyl ethyl ketone,
however, are more polar molecules, with moderate hydrogen bonding
capabilities. Even though the total cohesive energy density is similar
in all four solvents, the differences in component forces, primarily
hydrogen bonding, lead to the observed differences. Acetone and
MEK would much rather be attracted to each other than to linseed oil.
Two Component Parameters

A scheme to overcome the inconsistencies caused by hydrogen

bonding was proposed by Harry Burrell in 1955. This simple solution
divides the solvent spectrum into three separate lists: one for solvents
with poor hydrogen bonding capability, one for solvents with moderate
hydrogen bonding capability, and a third for solvents with strong
hydrogen bonding capability, on the assumption that solubility is
greatest between materials with similar polarities. This system of
classification is quite successful in predicting solvent behavior, and is
still widely used in practical applications. The classification according
to Burrell may be briefly summarized as follows:
Weak hydrogen bonding liquids: hydrocarbons, chlorinated
hydrocarbons, and nitrohydrocarbons.
Moderate hydrogen bonding liquids: ketones, esters, ethers, and
glycol monoethers
Strong hydrogen bonding liquids: alcohols, amines, acids,
amides, and aldehydes
Later systems assign specific values to hydrogen bonding capacity,
and plot those values against Hildebrand values on a two dimensional

graph. Although hydrogen bonding values are generally determined

using IR spectroscopy (by measuring the frequency shift a particular
solvent causes in deuterated methanol), another interesting method
uses the speed of sound through paper that has been wet with the
solvent being tested. Since paper fibers are held together largely by
hydrogen bonds, the presence of a liquid capable of hydrogen
bonding will disrupt the fiber-fiber bonds in preference to fiber-liquid
bonds. This disruption of paper fiber bonding will decrease the velocity
of sound travelling through the sheet. Water, capable of a high degree
of hydrogen bonding, is used as a reference standard, and the
hydrogen bonding value of a liquid is the ratio of its sonic disruption
relative to water. In this test, alkanes have no effect on fiber hydrogen
bonding, giving the same sonic velocities as air dried paper.
Hydrogen bonding is a type of electron donor-acceptor interaction
and can be described in terms of Lewis acid-base reactions. For this
reason other systems have attempted the classification of solvents
according to their electron donating or accepting capability. Such
extensions of the Hildebrand parameter to include acidity-basicity
scales, and ultimately ionic systems, are relatively recent and outside
the scope of this paper.
Three Component Parameters

Solubility behavior can be adequately described using Hildebrand

values, although in some cases differences in polar composition give
unexpected results (Fig. 1 , for example). Predictions become more
consistent if the Hildebrand value is combined with a polar value (i.e.
hydrogen bonding number), giving two parameters for each liquid.
Even greater accuracy is possible if all three polar forces (hydrogen
bonding, polar forces, and dispersion forces) are considered at the
same time. This approach assigns three values to each liquid and
predicts miscibility if all three values are similar.

As long as data is presented in the form of a single list, or even a

two dimensional graph, it can be easily understood and applied. With
the addition of a third term, however, problems arise in representing
and using the information; the manipulation of three separate values
presents certain inconveniences in practical application. It is for this
reason that the development and the use of three component
parameter systems has centered on solubility maps and models.
3-D Models

While polymer solubilities may be easily presented as a connected

group of solvents on a list, or as a specific area on a graph, the
description of solubilities in three dimensions is understandably more
difficult. Most researchers have therefore relied on three-dimensional
constructions within which all three component parameters could be
represented at once.
In 1966, Crowley, Teague, and Lowe of Eastman Chemical
developed the first three component system using the Hildebrand
parameter, a hydrogen bonding number, and the dipole moment as
the three components. A scale representing each of these three
values is assigned to a separate edge of a large empty cube. In this
way, any point within the cube represents the intersection of three
specific values. A small ball, supported on a rod, is placed at the
intersection of values for each individual solvent ( Figure 3).

Fig. 3. A three dimensional box used to plot solubility information

( after Crowley, Teague and Lowe) a=Hildebrand value, p=dipole
moment, h=hydrogen bonding value
Once all the solvent positions have been located within the cube in
this way, solubility tests are performed on individual polymers. The
position of solvents that dissolve a polymer are indicated by a black
ball, nonsolvents by a white one, and partial solubilities are indicated
by a grey ball. In this way a solid volume (or three dimensional area)
of solubility is formed, with liquids within the volume being active
solvents (black balls), and liquids outside the volume being nonsolvents (white balls). Around the surface of the volume, at the
interface between the area of solubility and the surrounding nonsolvent area, the balls are grey.
Once the volume of solubility for a polymer is established, it
becomes necessary to translate that information into a form that is
practical. This means transforming the 3-D model (difficult to carry
around) into a 2-D graph (easier to publish). This is usually done in
one of two similar ways. In both cases, the data is plotted on a
rectangular graph that represents only two of the three component
parameter scales (one side of the cube).

Fig. 4. Approximate Representations of Solid Model and Solubility

Map for Cellulose Acetate (from Crowley, at al, Journal of Paint
Technology Vol 39, 504, Jan 1967) graph that represents either a

single slice through the volume at a specified value on the third

component parameter scale, or a topographic map that indicats
several values of the third parameter at the same time (see Figure 4).
Because the volume of solubility for a polymer usually has an unusual
shape, several graphs are often needed for an individual polymer if its
total solubility behavior is to be shown.
Maps such as these can' be used in conjunction with a table of three
component parameters for individual solvents, and in this way provide
useable information about solvent-polymer interactions and allow the
formulation of polymer or solvent blends to suit specific applications.
Data presented in this way is not only concise, but saves considerable
time by allowing the prediction of solubility behavior without recourse
to extensive empirical testing. !t is for these reasons that solubility
maps are often included in technical reports and manufacturer's
product data sheets. How graphs are actually used to accomplish
these purposes will be described later in terms of the triangular Teas
graph, in which these procedures are similar but greatly simplified.
Hansen Parameters

The most widely accepted three component system to date is the

three parameter system developed by Charles M. Hansen in
1966.Hansen parameters divide the total Hildebrand value into three
parts: a dispersion force component, a hydrogen bonding component,
end a polar component. This approach differs from Crowley's in two
major ways: first, by using a dispersion force component instead of
the Hildebrand value as the third parameter, and second, by relating
the values of all three components to the total Hildebrand value. This
means that Hansen parameters are additive:
t2=d2 + p2 + h2 (6)
where

t2=

Total Hildebrand parameter

d2=

dispersion component

p2=

polar component

h2 =

hydrogen bonding component

The numerical values for the component parameters are determined

in the following way: First, the dispersion force for a particular liquid is
calculated using what is called the homomorph method. The
homomorph of a polar molecule is the nonpolar molecule most closely
resembling it in size and structure (n-butane is the homomorph of nbutyl alcohol). The Hildebrand value for the nonpolar homomorph
(being due entirely to dispersion forces) is assigned to the polar
molecule as its dispersion component value. This dispersion value
(squared) is then subtracted from the Hildebrand value (squared) of
the liquid, the remainder designated as a value representing the total
polar interaction of the molecule a (not to be confused with the polar
component p). Through trial and error experimentation on numerous
solvents and polymers, Hansen separated the polar value into polar
and hydrogen bonding component parameters best reflecting
empirical evidence. Table 3 lists Hansen parameters for several
solvents.
Table 3, Hansen Parameters for Solvents at 25C
(values selected from Hansen's 1971 parameters listed inHandbook
of Solubility Parameters, Allan F. M. Barton. Ph.D., CRC Press,
1983, page 153-157)
/MPa

Solvent
t

Alkanes
n-Butane

14.1

14.1

0.0

0.0

n-Pentane

14.5

14.5

0.0

0.0

n-Hexane

14.9

14.9

0.0

0.0

n-Heptane

15.3

15.3

0.0

0.0

n-Octane

15.5

15.5

0.0

0.0

Isooctane

14.3

14.3

0.0

0.0

n-Dodecane

16.0

16.0

0.0

0.0

Cyclohexane

16.8

16.8

0.0

0.2

Methylcyclohexane

16.0

16.0

0.0

0.0

Benzene

18.6

18.4

0.0

2.0

Toluene

18.2

18.0

1.4

2.0

Napthalene

20.3

19.2

2.0

5.9

Styrene

19.0

18.6

1.0

4.1

Aromatic Hydrocarbons

o-Xylene

18.0

17.8

1.0

3.1

Ethyl benzene

17.8

17.8

0.6

1.4

p- Diethyl benzene

18.0

18.0

0.0

0.6

Chloro methane

17.0

15.3

6.1

3.9

Methylene chloride

20.3

18.2

6.3

6.1

1,1 Dichloroethylene

18.8

17.0

6.8

4.5

Ethylene dichloride

20.9

19.0

7.4

4.1

Chloroform

19.0

17.8

3.1

5.7

1,1 Dichloroethane

18.5

16.6

8.2

0.4

Trichloroethylene

19.0

18.0

3.1

5.3

Carbon tetrachloride

17.8

17.8

0.0

0.6

Chlorobenzene

19.6

19.0

4.3

2.0

o-Dichlorobenzene

20.5

19.2

6.3

3.3

1,1,2 Trichlorotrifluoroethane

14.7

14.7

1.6

0.0

Halohydrocarbons

Ethers

Tetrahydrofuran

19.4

16.8

5.7

8.0

1,4 Dioxane

20.5

19.0

1.8

7.4

Diethyl ether

15.8

14.5

2.9

5.1

Dibenzyl ether

19.3

17.4

3.7

7.4

Acetone

20.0

15.5

10.4

7.0

Methyl ethyl ketone

19.0

16.0

9.0

5.1

Cyclohexanone

19.6

17.8

6.3

5.1

Diethyl ketone

18.1

15.8

7.6

4.7

Acetophenone

21.8

19.6

8.6

3.7

Methyl isobutyl ketone

17.0

15.3

6.1

4.1

Methyl isoamyl ketone

17.4

16.0

5.7

4.1

Isophorone

19.9

16.6

8.2

7.4

Di-(isobutyl) ketone

16.9

16.0

3.7

4.1

29.6

19.4

21.7

5.1

Ketones

Esters
Ethylene carbonate

Methyl acetate

18.7

15.5

7.2

7.6

Ethyl formate

18.7

15.5

7.2

7.6

Propylene 1,2 carbonate

27.3

20.0

18.0

4.1

Ethyl acetate

18.1

15.8

5.3

7.2

Diethyl carbonate

17.9

16.6

3.1

6.1

Diethyl sulfate

22.8

15.8

14.7

7.2

n-Butyl acetate

17.4

15.8

3.7

6.3

Isobutyl acetate

16.8

15.1

3.7

6.3

2-Ethoxyethyl acetate

20.0

16.0

4.7

10.6

Isoamyl acetate

17.1

15.3

3.1

7.0

Isobutyl isobutyrate

16.5

15.1

2.9

5.9

Nitromethane

25.1

15.8

18.8

5.1

Nitroethane

22.7

16.0

15.5

4.5

2-Nitropropane

20.6

16.2

12.1

4.1

Nitrobenzene

22.2

20.0

8.6

4.1

Nitrogen Compounds

Ethanolamine

31.5

17.2

15.6

21.3

Ethylene diem me

25.3

16.6

8.8

17.0

Pyridine

21.8

19.0

8.8

5.9

Morpholine

21.5

18.8

4.9

9.2

Analine

22.6

19.4

5.1

10

N-Methyl-2-pyrrolidone

22.9

18.0

12.3

7.2

Cyclohexylamine

18.9

17.4

3.1

6.6

Quinoline

22.0

19.4

7.0

7.6

Formamide

36.6

17.2

26.2

19.0

N,N-Dimethylformamide

24.8

17.4

13.7

11.3

Carbon disulfide

20.5

20.5

0.0

0.6

Dimethylsulphoxide

26.7

18.4

16.4

10.2

Ethanethiol

18.6

15.8

6.6

7.2

29.6

15.1

12.3

22.3

Sulfur Compounds

Alcohols
Methanol

Ethanol

26.5

15.8

8.8

19.4

Allyl alcohol

25.7

16.2

10.8

16.8

1-Propanol

24.5

16.0

6.8

17.4

2-Propanol

23.5

15.8

6.1

16.4

1-B utanol

23.1

16.0

5.7

15.8

2-Butanol

22.2

15.8

5.7

14.5

Isobutanol

22.7

15.1

5.7

16.0

Benzyl alcohol

23.8

18.4

6.3

13.7

Cyclohexanol

22.4

17.4

4.1

13.5

Diacetone alcohol

20.8

15.8

8.2

10.8

Ethylene glycol monoethyl ether

23.5

16.2

9.2

14.3

Diethylene glycol monomethyl ether

22.0

16.2

7.8

12.7

Diethylene glycol monoethyl ether

22.3

16.2

9.2

12.3

Ethylene glycol monobutyl ether

20.8

16.0

5.1

12.3

Diethylene glycol monobutyl ether

20.4

16.0

7.0

10.6

1 -Decanol

20.4

17.6

2.7

10.0

Acids
Formic acid

24.9

14.3

11.9

16.6

Acetic acid

21.4

14.5

8.0

13.5

Benzoic acid

21.8

18.2

7.0

9.8

Oleic acid

15.6

14.3

3.1

14.3

Stearic acid

17.6

16.4

3.3

5.5

Phenol

24.1

18.0

5.9

14.9

Resorcinol

29.0

18.0

8.4

21.1

m-Cresol

22.7

18.0

5.1

12.9

Methyl salicylate

21.7

16.0

8.0

12.3

Ethylene glycol

32.9

17.0

11.0

26.0

Glycerol

36.1

17.4

12.1

29.3

Propylene glycol

30.2

16.8

9.4

23.3

Diethylene glycol

29.9

16.2

14.7

20.5

Phenols

Polyhydric Alcohols

Triethylene glycol

27.5

16.0

12.5

18.6

Dipropylene glycol

31.7

16.0

20.3

18.4

Water

47.8

15.6

16.0

42.3

Hansen Model

Charles Hansen also used a three-dimensional model (similar to

that used by Crowley et al.) to plot polymer solubilities. He found that,
by doubling the dispersion parameter axis, an approximately spherical
volume of solubility would be formed for each. polymer. This volume,
being spherical, can be described in a simple way (Figure 5): the
coordinates at the center of the solubility sphere are located by means
of three component parameters (d, p, h), and the radius of the
sphere is indicated, called the interaction radius (R).Table 4 gives
the Hansen parameters and interaction radius of several polymers.

Fig. 5, The Hansen volume of solubility for a polymer is located

within a 3-D model by giving the coordinates of the center of a
solubility sphere (d, p, h) and its radius of interaction (R). Liquids
whose parameters lie within the volume are active solvents for that
polymer.
Table 4. Hansen Parameters and Interaction Radius of Polymers
(from: Solubility in the coatings industry, C. M. Hansen, Sksnd.

Tidskr. Faerg. Lack, 17, 69. 1971)

/MPa1/2
Polymer (trade name, supplier)

Cellulose acetate (Celliclore A, Bayer)

18.6 12.7 11.0

7.6

Chlorinated polypropylene (Parlon P10, Hercules)

20.3 6.3

10.6

Epoxy (Epikote 1001, Shell)

20.4 12.0 11.5

12.7

Isopreneelastomer (Ceriflex IR305,

Shell)

16.6 1.4

9.6

Cellulose nitrate (1 /2 sec, H-23, H forn)

15.4 14.7 8.8

Polyamide, thermoplastic (Versamid

930, General Mills)

17.4 -1.9

14.9 9.6

Poly( isobutylene) Lutonal IC-123,

BASF)

14.5 2.5

4.7

12.7

Poly(ethyl methacrylate) (Lucite 2042,

DuPont)

17.6 9.7

4.0

10.6

Poly(methyl methacrylate) (Rohm and

Haas)

18.6 10.5 7.5

8.6

Polystyrene (Polystyrene LO, BASF)

21.3 5.8

12.7

5.4

-0.8

4.3

11.5

Poly(vinyl acetate) (Mowilith 50,

Hoechst)

20.9 11.3

9.6

13.7

Poly(vinyl butyral) (Butvar B-76,

Shawnigan)

18.6 4.4

13.0 10.6

Poly(vinyl chloride) (Vilpa KR, k=50,

Montecatini)

18.2 7.5

8.3

Saturated polyester (Desmophen 850,

Bayer)

21.5 14.9 12.3 16.8

3.5

A polymer is probably soluble in a solvent (or solvent blend) if the

Hansen parameters for the solvent lie within the solubility sphere for
the polymer. In order to determine this (without building a model) it
must be calculated whether the distance of the solvent from the center
of the polymer solubility sphere is less than the radius of interaction
for the polymer:
D(S-P) = [4(ds - dp)2 + (ps - pp)2 + (hs - hp)2] (7)
where
D(S-P) = Distance between solvent and center of polymer solubility
sphere
xs=Hansen component parameter for solvent
xp=Hansen component parameter for polymer
(The figure "4" in the first term of equation (7), which doubles the
dispersion component scale, is intended to create a spherical volume
of solubility.) If the distance ( D(s-p) ) is less than the radius of
interaction for the polymer, the-solvent would be expected to dissolve
the polymer. This method avoids the reliance on graphic, plots of
solubility behavior and can be used effectively in solely numerical

form. The mathematics involved are inconvenient however (especially

when solvent blends are concerned), and it is perhaps for this reason
that the use of this excellent system is not more widespread.
Hansen Graph

Hansen parameters are both reasonably accurate in predicting

solubility behavior and concise in their representation of that
information.. Accurate because precise values for all three component
parameters are utilized, and concise because the entire solubility
volume for a polymer can be numerically indicated by four terms: one
set of parameters and a radius.
On the other hand, a two-dimensional graph sacrifices some of that
accuracy and conciseness in return for a system that clearly illustrates
the relative positions of numerous materials, and can be easily used
in practical applications. Predicting whether a polymer is soluble in a
mixture of two solvents, for example, while possible mathematically, is
accomplished on a graph by drawing a line between the two solvents
and seeing whether that line passes through the area of solubility for
the polymer.

Figure 6. Hansen graph of solubility areas for poly(methyl

methacrylate) (MMA) and poly(ethyl methacrylate) (EMA). Liquid
parameters are indicated by symbols; small circles indicate center of

solubility spheres. Liquids outside the solubility area of a polymer are

non-solvents. The dotted line illustrates all the possible mixtures of
MEK and ethanolnotice that MMA will tolerate a greater proportion
of ethanol than will EMA Accordingly, MMA should be soluble in
toluene/acetone 3:1, but not in 100% toluene.

Figure 7. Hansen graph of solubility areas for polyvinyl acetate)

(PVA), poly(vinyl butyral) (PVB), and poly(vinyl chloride). This type of
graph uses only two of the three Hansen parameters.
As was the case with Crowley's solubility maps, Hansen's three
dimensional volumes can be similarly illustrated in two dimensions by
plotting a cross-section through the center of the solubility sphere on a
graph that uses only two of the three parameters, most commonly
p and h. Figures 6 and 7 illustrate this approach by plotting the
volumes of solubility for five polymers: polyvinyl acetate, polyvinyl
butyral, polyvinyl chloride, polymethyl methacrylate, and polyethyl
methacrylate. The graphs use the hydrogen bonding component
parameter and the polar component parameter as the X and Y axis,
respectively, and plot the circle generated by the radius of interaction
for each polymer; the symbols indicate the respective locations of
solvents.

Hansen graphs are easy to use because solvent positions are

constant and polymer solubility areas may be drawn with a compass;
furthermore, solvent blending calculations can be done with a pencil
and ruler. The accuracy of predicting solubility behavior is about 90%,
with solvent locations nearest the edge of a solubility area being the
least predictable. This is due to the three-dimensional nature of the
actual solubility sphere. When reduced to two dimensions, solvents
that appear near the edge inside the solubility area may in fact
be outside it, in front or behind, in three dimensions.
Fractional Parameters

The division of the Hildebrand parameter into three component

Hansen parameters (dispersion force, polar force, and hydrogen
bonding force) considerably increases the accuracy with which nonionic molecular interactions can be predicted and described. Hansen
parameters can be used to interpret not only solubility behavior, but
also the mechanical properties of polymers, pigment binder
relationships, and the activity of surfactants and emulsifiers.
Being a three component system, however, places limitations on the
ease with which this information can be practically applied. Translating
this three component data onto a two-dimensional graph (by ignoring
one of the components) solves this problem but sacrifices a certain
amount of accuracy at the same time. What is n is a simple, planar
graph on which polymer solubility areas can be drawn in their entirety
in two dimensions. A triangular graph meeting these qualifications was
introduced by Jean P. Teas in 1968, using a set of fractional
parameters mathematically derived from the three Hansen
parameters. Because of its clarity and ease of use, the Teas graph
has found increasing application among conservators for problem
solving, documentation, and analysis, and is an excellent vehicle for
teaching practical solubility theory.

The Teas Graph

In order to plot all three parameters on a single planar graph, a

certain departure must be made from established solubility theory.
The construction of the Teas graph is based on the hypothetical
assumption that all materials have the same Hildebrand value.
According to this assumption, solubility behavior is determined, not by
differences in total Hildebrand value, but by the relative amounts of
the three component forces (dispersion force, polar force, and
hydrogen bonding force)that contribute to the total Hildebrand value.
This allows us to speak in terms of percentages rather than unrelated
sums.
Hansen parameters are additive components of the total Hildebrand
value (Equation 6). In other words, if all three Hansen values
(squared) are added together, the sum will be equal to
the Hild ebrand value for that liquid (squared). Teas parameters,
calledfractional parameters, are mathematically derived from
Hansen values and indicate the percent contribution that each
Hansen parameter contributes to the whole Hildebrand value:

In other words, if all three fractional parameters are added together,

the sum will always be the same ( 100).

For example, the alkanes, with intermolecular attractions due

entirely to dispersion forces, are represented by a dispersion
parameter of 100, indicating totality, with both polar and hydrogen
bonding parameters of zero. Molecules that are more polar have
dispersion parameters of less than 100, the remainder proportionately

divided between polar and hydrogen bonding contributions as the

particular Hansen parameters dictate.
Because Hildebrand values are not the same for all liquids, it should
be remembered that the Teas graph is an empirical system with little
theoretical justification. Solvent positions were originally located on
the graph according to Hansen values (using Equation 8), and
subsequently adjusted to correspond to exhaustive empirical testing.
This lack of theoretical foundation, however, does not prevent the
Teas graph from being an accurate and useful tool, perhaps the most
convenient method by which solubility information can be illustrated.
Fractional parameters for solvents are listed in Table 6, at the end of
this paper.

Figure 8. The Teas graph is an overlay of three solubility scales.

The Triangular Graph

The layout of a triangular graph is confusing at first to people who

are accustomed to the common Cartesian rectangular coordinate
graph. Instead of two axes perpendicular to each other, there are
three axes oriented at 60, and instead of these three axes requiring
three dimensions in space, the triangular graph is flat. Furthermore,
on a artesian graph all the scales graduate out from the same origin,
but on a triangular graph the zero point of any one scale is the upper
limit of another one.
This unusual construction derives from the overlay of three identical
scales, each proceeding in a different direction (Figure 8). In this way,

any point within the triangular graph uses three coordinates, the sum
of which will always be the same: 100 (Figure 9).

Figure 9. At any point on a triangular graph, all three coordinates

add up to 100.
Solvent Locations

By means of a triangular graph, solvents may be positioned relative

to each other in three directions (Figure 10). Alkanes, whose only
intermolecular bonding is due to dispersion forces, are located in the
far lower right corner of the Teas graph, the corner that corresponds to
100% dispersion force contribution, and 0% contribution from polar or
hydrogen bonding forces. Moving toward the lower left corner,
corresponding to 100% hydrogen bonding contribution, the solvents
exhibit increasing hydrogen bonding capability, culminating at the
alcohols and water, molecules with relatively little dispersion force
compared to their very great hydrogen bonding contribution. Moving
from the bottom of the graph upwards we encounter solvents of
increasing polarity, due less to hydrogen bonding functional groups
than to an increasingly greater dipole moment of the molecule as a
whole, such as the ketones and nitro compounds.

Figure 10. The Teas Graph

Numbers indicate solvent locations and refer to Table 5, pg. 43-45.
Overall, the solvents are grouped closer to the lower right apex than
the others. This is because the dispersion force is present in all
molecules, polar or not, and determining the dispersion component is
the first calculation in assigning Hansen parameters, from which the
Teas fractional parameters are derived. Unfortunately, this greatly
overemphasizes the dispersion force relative to polar forces,
especially hydrogen bonding interactions.
Solvent Classes

Figure 1 1 illustrates solvents on a Teas graph grouped according to

classes. Increasing molecular weight within each class shifts the

relative position of a solvent on the graph closer to the bottom right

apex. This is because, as molecular weight increases, the polar part
of the molecule that causes the specific character identifying it with its
class, called the functional group, is increasingly "diluted" by
progressively larger, nonpolar "aliphatic" molecular segments. This
gives the molecule as a whole relatively more dispersion force and
less of the polar character specific to its class.

Figure 11. Solvents grouped according to classes. Within each

class, increasing molecular weight shifts the solvent position toward
the right axis, corresponding to an increase in dispersion contribution
relative to polar contributions.
This trend toward less polarity with increasing molecular weight
within a class also accounts for the observation that lower molecular
weight solvents are often "stronger" than higher molecular weight
solvents of the same class, although determinations of solvent
strength must really be made in terms of the solvents position relative
to the solubility area of the solute. (Another reason for low molecular
weight solvents seeming more active is that smaller molecules can
disperse throughout solid materials more rapidly than their bulkier
relatives.)
The only class in which increasing molecular weight places the
solvent further away from the lower right corner is the alkanes. As
previously stated, the intermolecular attractions between alkanes are

due entirely to dispersion forces, and accordingly, Hansen parameter

values for alkanes show zero polar contribution and zero hydrogen
bonding contribution. Since fractional parameters are derived from
Hansen parameters, one would expect all the alkanes to be placed
together at the extreme right apex.
Observed behavior indicates, however, that different alkanes do
have different solubility characteristics, perhaps because of the
tendency of larger dispersion forces to mimic slightly polar
interactions. For this reason, Teas adjusted the locations of the
alkanes to correspond to empirical evidence, using Kauri-Butanol
values to assign alkane locations on the graph. Several other solvent
locations were also shifted slightly to properly reflect observed
solubility characteristics.
The position of water on the chart is very uncertain, due to the ionic
character of the water molecule, and the placement in this paper is
according to recent published values (Teas, 1976). The presence of
water in a solvent blend, however, can alter dramatically the accuracy
of solubility predictions.
[It should be noted that the position of methylene chloride is also
correct according to recent values. Many earlier publications have
given methylene chloride incorrect parameters properly corresponding
to Hansen's values for methyl chloride, a different chemical, possibly
due to calculation error.]
Polymer Solubility Windows

Given the solvent positions, it is possible to indicate polymer

solubilities using methods similar to those used by Crowley and
Hansen: a polymer is tested in various solvents, and the results
indicated on the graph (a 3-D model is no longer necessary). At first,
individual liquids from diverse locations on the graph are mixed with
the polymer under investigation, and the degree of swelling or

dissolution noted. Liquids that are active solvents, for example, might
have their positions on the graph marked with a red dot. Marginal
solvents might be marked with a yellow dot, and nonsolvents marked
with black. Once this is done, a solid area on the Teas graph will
contain all the red dots, surrounded with yellow dots (see Figure 12).
The edges of this area, or polymer solubility window, can be more
closely determined in the following way. Two liquids near the edge of
the solubility window are chosen, one within the window (red dot), and
one outside the window (black dot). Dissolution (or swelling) of the
polymer is then tested in various mixtures of these two liquids, using
cloud-point determinations if accuracy is essential, and the mixture
just producing solubility is noted on the graph, thus determining the
edge of the solubility window. (The mixture would be located on a line
between the two liquids, at a point corresponding in distance to the
ratio of the liquids in the mixture.) If this procedure is repeated in
several locations around the edge of the solubility window, the
boundaries can be accurately determined. Interestingly, some
composite materials (such as rubber/resin pressure sensitive
adhesives, or wax resin mixtures) can exhibit two or more separate
solubility windows, more or less overlapping, that reflect the degree of
compatibility and the concentration of the original components.

Figure 12. The solubility window of a hypothetical polymer (circles

indicate solvents).

This method of solubility window determination can be performed on

samples under a microscope, and the results plotted on a Teas graph.
In cases where the solubilities of artifactual materials are to be a prior
to treatment, it is often unnecessary to delineate the entire solubility
window of the materials in question. It can suffice to record the
reaction of the materials to the progressive mixtures of a few selected
solvents under working conditions in order to determine appropriate
working solutions.
Temperature, concentration, viscosity

The solubility window of a polymer has a specific size, shape, and

placement on the Teas graph depending on the polarity and molecular
weight of the polymer, and the temperature and concentration at
which the measurements are made. Most published solubility data are
derived from 10% concentrations at room temperature.
Heat has the effect of increasing the size of the solubility window,
due to an increase in the disorder (entropy) of the system. The more
disordered a system is (increased entropy), the less it matters how
dissimilar the solubility parameters of the components are. Since
entropy also relates to the number of elements in a system (more
elements=more disorder), polymer grades of lower molecular weight
(many small molecules) will have larger solubility windows than
polymer grades of higher molecular weight (fewer large molecules).
Concentration also has an effect on solubility. As stated, most
polymer solubility windows are determined at 10% concentration of
polymer in solvent. Because an increase in polymer concentration
causes an increase in the entropy of the system (more
elements=more disorder), solubility information can be considered
accurate for solutions of higher concentration as well. Solvent
evaporation as a polymer film dries serves to increase the polymer
concentration in the solvent, thus insuring that the two materials stay
mixed. It is possible, however, for polymer solutions of less than 10%

to phase separate (become immiscible), due to a decrease in entropy.

This is particularly susceptible to polymer-solvent combinations at the
edge of the polymer solubility window. In other words, with lower
polymer concentration there is an increase in the order of the system
(less entropy); therefore, the size of the solubility window becomes
smaller (there is less difference tolerated between solvent and
polymer solubility values).
Solution viscosity also varies depending on where in the polymer
solubility window the solvent is located. One might expect viscosity to
be at a minimum when a solvent near the center of a polymer
solubility window is used. However, this is not the case. Solvents at
the center of a polymer solubility window dissolve the polymer so
effectively that the individual polymer molecules are free to uncoil and
stretch out. In this condition molecular surface area is increased, with
a corresponding increase in intermolecular attractions. The molecules
thus tend to attract and tangle on each other, resulting in solutions of
slightly higher than normal viscosity.
When dissolved in solvents slightly off-center in the solubility
window, polymer molecules stay coiled and grouped together into
microscopic clumps which tend to slide over one another, resulting in
solutions of lower viscosity. As solvents nearer and nearer the edge of
the solubility window are used to dissolve the polymer, however, these
clumps become progressively larger and more connected and
viscosity again increases until ultimately polymer-liquid phase
separation occurrs as the region of the solubility window boundary is
crossed.
Dried film properties

The position of a solvent in the solubility window of a polymer has a

marked effect on the properties of not only the polymer-solvent
solution, but on the dried film characteristics of the polymer as well.
Because of the uncoiling of the polymer molecule, films (whether

adhesives or coatings) cast from solvent solutions near the center of

the solubility window exhibit greater adhesion to compatible
substrates. This is due to the increase in polymer surface area that
comes in contact with the substrate. (Hildebrand parameters can be
related to surface tension, and adhesion is greatest when the
polarities of adhesive and adherend are similar.)
Many other properties of dried films, such as plastic crazing or gas
permeability are related to the relative position that the original solvent
occupied in the solubility window of the polymer. The degree of both
crazing and permeability is predictably less when solvents more
central to the solubility window have been used.
Evaporation rates and solubility

Solvent evaporation rates can also have a marked affect on dried

film properties. The solubility parameters of solvent blends can
change during film drying because of the difference in evaporation
rates of the component liquids. If a volatile true solvent is mixed with a
slow evaporating non-solvent, the compatibility between solvents and
polymer can shift as the true solvent evaporates. The predominance
of the non-solvent during the last stages of drying can result in a
discontinuous, porous film with higher opacity and decreased
resistance to water and oxygen deterioration. (There may be
instances where these properties are desirable.)
This can be avoided, however, by either blending a small amount of
a high boiling true solvent into the solvent mixture (this solvent
remains to the last and insures miscibility), or by making sure that, if
an azeotropic mixture is formed on evaporation, the parameters of the
azeotrope lie within the polymer solubility window.
An azeotrope is a mixture of two or more liquids that has a constant
boiling point at a specific concentration. When two liquids are mixed
that are capable of forming an azeotrope, the more volatile liquid will

evaporate more quickly until the concentration reaches azeotropic

proportions. At that point, the concentration will remain constant as
evaporation continues. If the position of the azeotropic mixture lies
within the solubility window, compatibility with the polymer will
continue throughout the drying process. This can be determined by
consulting a table of azeotropes and checking the location of the
mixture on the Teas graph in relation to the polymer solubility window.
(Methods of plotting solvent mixtures are described in the next
section.)
Blending solvents

Teas graph is particularly useful as an aid to creating solvent

mixtures for specific applications. Solvents can easily be blended to
exhibit selective solubility behavior (dissolving one material but not
another), or to control such properties as evaporation rate, solution
viscosity, degree of toxicity or environmental effects. The use of the
Teas graph can reduce trial and error experimentation to a minimum,
by allowing the solubility behavior of a solvent mixture to be predicted
in advance.
Because solubility properties are the net result of intermolecular
attractions, a mixture with the same solubility parameters as a single
liquid will, in many cases, exhibit the same solubility behavior.
Determining the solubility behavior of a solvent mixture, therefore, is
simply a matter of locating the solubility parameters of the mixture on
the Teas graph. There are two ways by which this may be
accomplished: mathematically, by calculating the fractional
parameters of the mixture from the fractional parameters of the
individual solvents, and geometrically, by simply drawing a line
between the solvents and measuring the ratio of the mixture on the
graph. The mathematical method is the most accurate, and is
appropriate for mixtures of three or more solvents. The geometrical
method is the most convenient and is suitable for mixtures of two
solvents, or for very rough guesses when three solvents are involved.

The mathematical method

The solubility parameter of a mixture of liquids is determined by

calculating the volume-wise contributions of the solubility parameters
of the individual components of the mixture. In other words, the
fractional parameters for each liquid are multiplied by the fraction that
the liquid occupies in the blend, and the results for each parameter
added together. For example, given a mixture of 20% acetone and
80% toluene:
In this way, the position of the solvent mixture can be located on the
Teas graph according to its fractional parameters. Calculations for
mixtures of three or more solvents are made in the same way.
The geometric method

The geometric method of locating a solvent mixture on the Teas

graph involves simply drawing a line between the two solvents in the
mix, and finding the point on the line that corresponds to the volume
fractions of the mixture.
fd

fd

fh

Acetone:

47 (x.20) = 9.4

32 (x.20) = 6.4

21 (x.20) = 4.2

Toluene:

80 (x.80) = 64.0

7 (x.80) = 5.6

13 (x.80) = 10.4

fd =73.4

fh =12.0

fp = 14.6

20/80 Mix:

Figure 13. A mixture of 20% acetone and 80% toluene can be

located on the Teas graph by using a pencil and ruler. The mixture lies
on a line connecting the two liquids, at a distance equal to the ratio of
the mixture, and closer to the liquid present in the greatest amount.
This is illustrated in Figure 13 for the same 20% acetone, 80% toluene
mixture. A line connecting acetone and toluene is drawn on the Teas
graph. A point is then located on the line, 20% of the length of the line
away from toluene. It is important to remember that the location of a
mixture will be closer to the liquid present in the greatest amount.
Solvent blends and solubility windows

What is interesting about visualizing solvent blends on the Teas

graph is the control with which effective solvent mixtures can be
formulated. For example, two liquids that are non-solvents for a
specific polymer can sometimes be blended in such a way that the
mixture will act as a true solvent. This is possible if the graph position
of the mixture lies inside the solubility window of the polymer, and is
most effective if the distance of the non-solvents from the edge of the
solubility window is not too great. This is illustrated in Figure 14.

Figure 14. A mixture (M) of non-solvents (A,B) may act as a true

solvent for a polymer if the mixture is located inside the solubility
window for the polymer on the Teas graph.
This phenomenon is particularly valuable when selective solvent
action is required. Often it is necessary to selectively dissolve one
material while leaving other materials unaltered, as in the case of
removing the varnish from a painting, some adhesive tape from the
image area of a print, or when a consolidant must not dissolve the
material being consolidated. Sometimes the solubilities of all the
materials involved are so similar that selecting an appropriate and
safe solvent can be difficult. In such cases it is helpful to indicate the
solubility windows of both the material that needs to be dissolved
(varnish, adhesive, consolidant), and the materials that must be
protected (media), on a Teas graph. This can be accomplished by
simple solubility testing, noting the results of the tests on the graph.

Figure 15. In situations where one material must be dissolved (dark

circle) while another must remain unaffected (shaded area), it is
helpful to plot the solubility of both materials on the graph. A solvent
blend can then be formulated (triangle) that selectively dissolves only
the proper material (see text).
Once this has been done, it is easy to see the overlap of solubilities,
and the areas where solubilities are mutually exclusive, if they exist. A
solvent blend can then be formulated that actively dissolves the
proper material, while positioned as far away from the solubility
window of the other material as possible (Figure 15). It is important to
remember that differences in evaporation rates can shift the solubility
parameter of the blend as the solvents evaporate, and this must be
taken into account. Additionally, while a material may not shown signs
of solution in a solvent or solvent blend, the solvent may still adversely
affect the material, for example by softening the material or leaching
out low molecular weight components. Such changes can be
irreversible and must be considered prior to embarking on a
treatment.
Solvent mixture scales

Although the Teas graph is useful and informative when dealing with
complex solubility questions, in most day to day situations choosing a
solvent is a straightforward procedure that would be unnecessarily
complicated by having to plot entire solubility windows. In most cases,
the degree of solubility of a material is simply tested in various
concentrations of two or three solvents, in order to determine the
mildest solvent capable of forming a solution.
Perhaps the most often used solvent mixtures are blends of aliphatic
and aromatic hydrocarbons, sometimes with the addition of acetone.
This is because the search for the mildest solvent is often
synonymous with the search for the least polar solvent (and the
aliphatic hydrocarbons are the least polar possible).

Figure 16. Some common solvent blends (a=acetone, t=toluene,

h=heptane, e=ethanol, m=methylene chloride, f=Freon TF). Inmost
cases, dispersion force values give a relative indication of solvent
strength (100=weakest, 30=strongest).
Testing whether a polymer is suitable for use in conservation, for
example, usually involves determining the mildest solvent mixture that
will dissolve both aged and un-aged samples. For this purpose,
various concentrations of toluene in cyclohexane are used; should the
polymer prove insoluble in straight toluene, however, increasing
amounts of acetone are a until solubility is achieved. This type of
solubility test anticipates the choices that will be made in working
situations.
Looked at in terms of fractional parameters, what is being
determined in such tests is essentially the location of the edge of the
solubility window for the polymer in relation to, the lower right corner
of the Teas graph. Figure 15 illustrates various mixtures of heptane,
toluene, and acetone. It can been clearly seen that solvent strength
increases with greater distance from the 100% dispersion axis. Blends
of trichlorotrifluoroethane (Freon TF) and methylene chloride as well

as blends of toluene ethanol are also illustrated. In all cases,

increasing solvent strength follows decreasing dispersion force
contribution.
For this reason, the use of fractional dispersion values (Table 5) is
an excellent method for concisely designating relative solvent
strength, in place of other more limited scales (Kauri-Butanol number,
aromatic content, etc.). The benefits of this approach include the use
of a standard designation that encompasses the entire range of
solvent strengths, and the ability to easily enlarge the designation to
include more precise solubility parameter data if necessary.
Table 5. Mixtures of heptane. toluene, and acetone, with
corresponding dispersion force values. fd locations are illustrated
in Figure 15
% Heptane

% Toluene

% Acetone

Approx. fd

100

100

75

25

95

50

50

90

25

75

85

100

80

85

15

75

70

30

70

55

45

65

40

60

60

26

76

55

10

90

50

100

47

Solvents and Health

As we have shown, the Teas graph can be a useful guide in tailoring

solvent blends to suit specific applications. By adjusting the position of
the blend relative to the solubility window of a polymer such properties
as solution viscosity and adhesion can be optimized. Evaporation
rates can be controlled independently of solvent strength, and the
effects of temperature and concentration can be anticipated.
A further advantage that can be derived from this latitude in creating
solvent mixtures is the possibility of choosing solutions based on
degree of toxicity. A solvent mixture having a graph position close to
another solvent will have such similar solubility characteristics to that
solvent that it can be used interchangeably in many applications. For
example, a petroleum solvent of 30% aromatic character is more or
less the same whether the aromatic content is due to benzene (very
toxic) or to toluene (moderately toxic). By extension, a mixture of
ethanol/toluene 50:50 might be used in place of tetrahydrofuran in
some applications, and toluene might be replaced with a 3:1 mixture
of Stoddard solvent and acetone. In such cases, it should be pointed
out that the similarity between solvents and blends having the same
numerical parameters decreases as the distance between the
components of the blend increases. Where alternate blends are

effective, however, the use of a less toxic replacement can be a

sensible choice, and the Teas graph a useful tool.
Conclusion

The theory of solubility parameters is a constantly developing body

of scientific concepts that can be of immense practical assistance to
the conservator. Through the media of solubility maps, complex
molecular interactions can be visualized and understood, and in this
way, solubility theory can simply function as a ladder to be left behind
once the basic concepts are assimilated. On the other hand, the
solution to an unusual problem can often be put within reach by
graphically plotting solubility behavior on a Teas graph.
In the near future, the extension of solubility theory to encompass
ionic and water based systems is conceivable, and the development
of simple computer programs to manipulate multi-component solubility
parameter data, along with accessible data bases of material
solubilities, is probable. Until that time, both conservators and the
objects in their charge can continue to profit by the use, either
conceptual or real, of solubility parameter theory.
John Burke
The Oakland Museum August 1984

Table 6. Fractional Solubility Parameters

(Values from Gardon and Teas Treatise on Coatings, Vol.2,
Characterization of Coatings: Physical Techniques, Part II, Meyers
and Long, Eds., Marcel Dekker, NY, 1976. Values in brackets
derived from Hansen's 1971 parameters inHandbook of Solubility
Parameters, A. Barton. CRC Press. 1983, using Equation 4)
Numbers in left column refer to solvent positions in Teas graph,
Figure 10, paw 30.
solvent

100 fd

100 fp 100 fh

Alkanes
1

n-Pentane

100

n- Hexane

100

n-Heptane

100

n-Dodecane

100

Cyclohexane

94

V M& P Naphtha

94

Mineral Spirits

90

Aromatic Hydrocarbons
5

Benzene

78

14

Toluene

80

13

o-Xylene

83

12

Naphthalene

70

22

Styrene

78

18

10

Ethylbenzene

87

10

11

p-Diethyl benzene

97

Halogen Compounds
12

Methylene chloride

59

21

20

13

Ethylene dichloride

67

19

14

14

Chloroform

67

12

21

15

Trichloroethylene

68

12

20

16

Carbon tetrachloride

85

13

17

1,1,1 Trichloroethane

70

19

11

18

Chlorobenzene

65

17

19

Trichlorotrifluoroethane

90

10

Ethers
20

Diethyl ether

64

13

23

21

Tetrahydrofuran

55

19

26

22

Dioxane

67

26

23

Methyl Cellosolve

39

22

39

24

Cellosolve 8

42

20

38

25

Butyl Cellosolve

46

18

36

26

Methyl Carbitol

44

21

35

27

Carbitol

48

23

29

25

Butyl Carbitol

46

18

36

Ketones
28

Acetone

47

32

21

29

Methyl ethyl ketone

[53]

[30]

[17]

30

Cyclohexanone

55

28

17

Diethyl ketone

56

27

17

Mesityl oxide

55

24

21

31

Methyl isobutyl ketone

58

22

20

32

Methyl isoamyl ketone

62

20

18

Isophorone

51

25

24

Di-isobutyl ketone

[67]

[16]

[17]

33

Esters
34

Methyl acetate

45

36

19

35

Propylene carbonate

48

38

14

36

Ethyl acetate

51

18

31

Trimethyl phosphate

[39]

[37]

[24]

Diethyl carbonate

64

12

24

Diethyl sulfate

42

39

19

n-Butyl acetate

60

13

27

Isobutyl acetate

60

i5

25

38

Isobutyl isobutyrate

63

12

25

39

Isoamyl acetate

60

12

28

40

Cellosolve acetate

51

i5

34

Ethyl lactate

44

21

35

Butyl lactate

40

20

32

37

Nitrogen Compounds
41

Acetonitrile

39

45

16

42

Butyronitrile

44

41

15

43

Nitromethane

40

47

13

44

Nitroethane

44

43

13

45

2-Nitropropane

50

37

13

46

Nitrobenzene

52

36

12

47

Pyridine

56

26

18

48

Morpnoline

57

15

28

49

Aniline

50

19

31

50

N-Methyl-2-pyrrolidone

48

32

20

Diethylenetriamine

38

30

32

Cyclohexylamine

[64]

[12]

[24]

Formamide

28

42

30

N N-Dimethylformamide

41

32

27

41

36

23

30

22

48

51

52

Sulfur Compounds
88
53

Carbon disulfide

54

Dimethylsulfoxide
Alcohols

55

Methanol

56

Ethanol

36

18

46

57

1-Propanol

40

16

44

58

2-Propanol

[41]

[16]

[43]

59

1-Butanol

43

15

42

2-Butanol

[44]

[16]

[40]

Benzyl alcohol

48

16

36

60

Cyclohexanol

50

12

38

61

n-amyl alcohol

46

13

41

62

Diacetone alcohol

45

24

31

2-Ethyl-1-hexanol

50

41

Polyhydric Alcohols
63

Ethylene glycol

30

18

52

64

Glycerol

25

23

52

65

Propylene glycol

34

16

50

66

Diethylene glycol

31

29

40

67 Water

18

28

54

Miscellaneous Liquids
68

Phenol

46

15

39

69

Benzaldehyde

61

23

16

70

Turpentine

77

18

71

Dipentene

75

20

Formic acid

[33]

[28]

[39]

Acetic acid

[401

[22]

[38]

Oleic acid

[62]

[14]

[24]

Stearic acid

[65]

[131

[22]

Linseed oil

66

17

17

Cottonseed oil

67

15

18

Neets foot oil

69

14

17

Pine oil

70

14

16

Sperm oil

75

11

14

Mineral oil

100

1
References

Barton, Allan F. M., Handbook of Solubility Parameters end Other

Cohesion Parameters Boca Raton, Florida: CRC Press, Inc., 1983.
Burrell, Harry, "The Challenge of the Solubility Parameter
Concept," Journal of Paint Technology, Vol. 40, No. 520, 1968.
Crowley, James D., 0. S. Teague, Jr., and Jack W. Lowe, Jr., "A
Three Dimensional Approach to Solubility," Journal of Paint
Technology, Vol. 38, No. 496, 1966.
Durrans, Thomas H., Solvents. London: Chapman and Hall Ltd.,
1971.
Feller, Robert L., Nathan Stolow, and Elizabeth H. Jones, On
Picture Varnishes and Their Solvents. Cleveland: The Press of Case
Western Reserve University, 1971.
Feller, Robert L., "The Relative Solvent Power Needed to Remove
Various Aged Solvent-Type Coatings," in Conservation and
Restoration of Pictorial Art, Bromelle and Smith, Eds., London:
Butterworths 1976
Hildebrand, J. H., The Solubility of Non-Electrolytes New York:
Reinhold, 1936
Hansen, Charles M., "The Three Dimensional Solubility Parameter
Key to Paint Component Affinities: 1. Solvents Plasticizers, Polymers,
and Resins," Journal of Paint Technology, Vol. 39, No. 505, 1967.
Hansen, Charles M., "The Three Dimensional Solubility Parameter Key to Paint Component Affinities: 11. Dyes, Emulsifiers, Mutual
Solubility and Compatibility, and Pigments," Journal of Paint
Technology, Vol. 39, No. 51 1, 1967.
Hansen, Charles M., "The Three Dimensional Solubility Parameter Key to Paint Component Affinities: 111. Independent Calculations of

the Parameter Components," Journal of Paint Technology, Vol. 39,

No. 511, 1967.
Hansen, Charles M., "The Universality of the Solubility Parameter
Concept," I & E C Product Research and Development, Vol. 8, No. 1,
1969.
Hedley, terry, "Solubility Parameters and Varnish Removal: A
Survey," The Conservator, No. 4, 1980.
Teas, Jean P., "Oraphic Analysis of Resin Solubilities," Journal of
Paint Technology, Vol. 40, No. 516, 1968.
Teas, Jean P., Predicting Resin Solubilities. Columbus, Ohio:
Ashland Chemical Technical Bulletin, Number 1206.
Torraca, Giorgio, Solubility and Solvents for Conservation
Problems Rome: ICCROM, 1978.

Publication History
Received: Fall 1984
This paper was submitted independently by the author, and was not
delivered at the Book and Paper specialty group session of the AIC
Annual Meeting. It has not received peer-review