Appendix

| C2 |

Estimation of Hansen Solubility Parameters

(HSP) from Binary DatadPES
This appendix builds on the techniques described
in Appendix C1. Published data of compatibility
between PES polymer and many solvents are the
feedstock for use of an optimization technique. The
outcome is not only numerical values of solubility
parameters for the polymer, but can be an understanding of the vulnerability of the integrity of those
values.

B. ESTIMATION OF HSP USING

BINARY DATAdPOLYETHERSULFONE
POLYMER (PES)
The optimization procedure1 is simple, and can be
described as follows:
A. In a spreadsheet, create columns A and B2 following

A. EXAMPLE DATA
FOR PARAMETER
EVALUATION
Note in the experimental data within Table C2-1A that 41
solvents were tested for compatibility with polyethersulfone (PES) polymer. The solvents are noted in
column A; results of the solubility tests are noted in column
B for each solvent in column A.
Of these 41 solvents, the polymer is only soluble in ve.
They are noted in boldface type.
Figures C2-1 (two-dimensional plots with the dispersion HSP not plotted) and C2-2 (three-dimensional plot)
are made using the solvent/polymer test data in Table
C2-1.
The symbols in dark green within both Figures C2-1 and
C2-2 are those ve data points in which the PES polymer
dissolved in the test solvent. Compare either gure to
Figure 2.42, and recognize the complexity of real-world
data and that the optimization can't often be done by
visual inspection of plots, and why instead tabulated
results are analyzed by a spreadsheet or a programmed
optimization routine.

Table C2-1. These are the identications of the

individual solvents tested, and the solubilities of PES
polymer in each solvent. The outcomes are taken to be
binary: 1 for solubility, 0 for insolubility.
B. Choose trial values of the ddisperse, dpolar, dhydrogen bonding,
and RO parameters for the soil/polymer/solventdPES
polymer in this case. Usually this choice is not critical to
the outcomedbut occasionally it can be very much so.
As noted in Appendix C1, useful starting values can be
15, 5, 5, and 5 MPa respectively3.
C. For each solvent (a row of the spreadsheet) look up in
Table A1-5, or in the reference noted in endnote A, the
associated ddisperse, dpolar, dhydrogen bonding values. Put
them in columns C, D, and E, respectively.
D. For each solvent, apply Equation 2.8 to calculate the
distance in three-dimensional solubility parameter
space from the trial value of the location of the center
1
Dr. Charles Hansen developed a computer program called SPHERE which
does this optimization using programmed routines rather than optimization
functions in a spreadsheet. It is the author's belief from evaluation of various
cases, and study of the methodology, that both SPHERE and the spreadsheet
optimization functions provide essentially identical results. Certainly, the
same desirability and error functions are used.
2
These are the columns in Table C2-1, which may be labeled otherwise
in the spreadsheet one creates.
3
These trial values were chosen because they are integer median values
of all 210 cleaning solvents listed in the database of Appendix A1, and
they were successfully used in Appendix C4.

569

Table C2-1 Initial Determination of Solubility Parameters for PES Polymer [Trial Values are: ddisperse [ 15.0; dpolar [ 5.0; dH
SOLUBILITY
GRADE 1
OK; 0 NOT

d
Disperse

d
H
Bonding

d
Polar
MPa

SOLVENT
A

Acetone

15.5

10.4

Acetophenone

19.6

Benzene

18.4

1-Butanol

16

5.7

Butyl Acetate

15.8

GammaButyrolactone

19

Carbon
Tetrachloride

Chlorobenzene

9
10
11

HSP RA

GOOD
& IN

BAD
& OUT

BAD
& IN

[ 5.0; RO [ 5.0]

GOOD
& OUT

bonding

Sphere Error

DF

R A < RO

RA > RO

0.385402

5.845

0.43

2.327

8.6

3.7

9.964

0.007

143.228

0.007

8.958

0.019

52.336

15.8

11.006

0.002

405.819

12.972

3.7

6.3

2.437

0.077

0.0077

16.6

7.4

14.294

10,873.18

17.8

0.6

8.702

0.025

40.517

19

4.3

8.573

0.028

35.61

Chloroform

17.8

3.1

5.7

5.955

0.385

2.598

Cyclohexanol

17.4

4.1

13.5

9.803

0.008

121.883

Diacetone Alcohol

15.8

8.2

10.8

6.815

0.163

6.139

12

o-Dichlorobenzene

19.2

13

Diethylene Glycol

16.6

14

Diethyl Ether

14.5

15

Dimethyl
Formamide

16

Dimethyl Sulfoxide

17

6.3

3.3

8.668

0.026

39.187

20.7

17.485

264,378.36

2.9

5.1

2.328

14.468

0.069

0.069

17.4

13.7

11.3

11.765

0.001

867.142

867.142

18.4

16.4

10.2

14.256

10,469.55

1,4-Dioxane

19

1.8

7.4

8.944

0.019

51.639

18

Ethanol

15.8

19

Ethanolamine

17

20

Ethyl Acetate

15.8

21

Ethylene Dichloride

19

22

Ethylene Glycol

17

12

8.8

19.4

14.979

21,561.22

15.5

21.2

19.715

2,458,894.82

5.3

7.2

2.737

9.614

0.104

0.104

8.401

0.033

29.982

22.204

29,609,
425.90

7.4
11

4.1
26

23

Ethylene Glycol
Monobutyl Ether

16

5.1

12.3

7.57

0.077

13.062

24

Ethylene Glycol
Monoethyl Ether

16.2

9.2

14.3

10.483

0.004

240.53

25

Ethylene Glycol
Monomethyl Ether

16.2

9.2

16.4

12.384

0.001

1609.79

26

Formamide

17.2

26.2

19

25.784

1,062,314,
431.89

27

Hexane

14.9

7.074

0.126

7.956

28

Isophorone

16.6

8.2

7.4

5.122

0.885

1.13

29

Methanol

15.1

12.3

22.3

18.778

963,356.51

30

Methylene
Dichloride

18.2

6.3

6.1

6.623

0.197

5.067

0.197

31

Methyl Ethyl
Ketone

16

5.1

4.473

1.693

0.591

0.0591

32

Methyl Isobutyl
Ketone

15.3

6.1

4.1

1.543

0.032

0.032

33

Methyl-2Pyrrolidone

18

12.3

7.2

9.702

0.009

110.174

0.009

34

Nitroethane

16

15.5

4.5

10.7

0.003

299.007

31.73

35

Nitromethane

15.8

18.8

5.1

13.893

7279.40

36

2-Nitropropane

16.2

12.1

4.1

7.549

0.078

12.788

37

Propylene
Carbonate

20

18

4.1

16.426

91,664.84

38

Propylene Glycol

16.8

9.4

23.3

19.163

1,415,
119.07

39

Tetrahydrofuran

16.8

5.7

4.738

1.299

0.77

0.77

40

Toluene

18

1.4

7.613

0.073

13.642

41

Trichloroethylene

18

3.1

5.3

6.301

0.272

3.672

1
1.05307175865665
E-13
41

Appendix

C2

Estimation of Hansen Solubility Parameters (HSP) from Binary DatadPES

Figure C2-1 Two-Dimensional Plot HSP of Solvents and their Solubility Response with PES Polymer

Figure C2-2 Three-Dimensional Plot HSP of Solvents and their Solubility Response with PES Polymer

of the sphere (step B) to each solvent. This value is

called RA and it is placed in Column F. For example,
use as condition 1 in Equation 2.8, the three trial HSP
values. Use as condition 2 in Equation 2.8 the three
HSP values for each solvent tested.
E. Examine the solubility data on each row of Table C21, and sort4 them into four baskets or groups. The
groups are: GOOD & IN (Column G), BAD &
OUT (Column H), BAD & IN (Column I), and
GOOD & OUT (Column J). Enter5 a 1 in the
column in which each data point belongs, and enter
a 0 (or a blank) in the other three columns6. For visual

convenience, one can apply background colors to

Columns G and H (where the solvent is properly
positioned relative to the sphere); apply another color
background to Columns I and J (where the solvent is
not properly positioned relative to the solubility
sphere).
F. Calculate the numerical error values using Equation
C1-3. Pay attention to the type of error for each data
point; either BAD & IN (Column I), or GOOD &
OUT (Column J)7.
For solvents producing a ducing a BAD & IN
response (with a 1 being in column I), enter the
formula in Column K8.

Spreadsheets make this sorting based on comparative values trivial.

The entry should be by calculation of the spreadsheet, and not
a manual entry which cannot change as the optimization proceeds.
6
In every row of Table C2-1 (for every solvent), there must be only one
of columns G through J in which a value of 1 has been entered. The
other columns must have entries of 0, or be blank.
5

572

For those solvents correctly positioned GOOD & IN or BAD &

OUT the errors are zero.
The reason for being particular about the values placed in columns K
and L is that the errors stated in Equation C1-3 are taken to be positive
values in every case.
7
8

Estimation of Hansen Solubility Parameters (HSP) from Binary DatadPES

For solvents producing a ducing a GOOD & OUT
response (with a 1 being in column J) enter the
formula in column L.
Prior to the start of the optimization, as in Table
C2-1, no solvent is correctly positioned (because the
values of the ddisperse, dpolar, dhydrogen bonding, and RO
parameters are only guessed at and not yet
determined by optimization).
G. In column M, evaluate the dfi function for each solvent.
 Insert the number 1 if there is a 1 in columns G or H
(as determined in step E above), meaning that the
solvent is properly positioned relative to the
position and size of the solubility sphere.
 Insert the value from column K if there is a 1 in
column I (a BAD & IN response), or insert the
value from column K if there is a 1 in column J (a
GOOD & OUT response). Both are determined
in step F above.
H. Evaluate the overall desirability function (DF) for the
polymer/soil/solvent being tested through use of
Equation C1-4. This is the product of the individual
dfi values taken to the power of the reciprocal of the
number of solvents tested (41 in this example). It is the
number in red above column M.
I. Activate the optimization algorithms in the spreadsheet
to seek a value of 1 for the overall DF. They will do that
by separately and independently modifying the values
of the trial ddisperse, dpolar, dhydrogen bonding, and RO
parameters until the value of 1 for DF is reached, or
until it reaches a value of DF close to 1 which it cannot
improve to be closer to 1. One then takes those nal
trial values as being the outcome of the optimization
based on the provided experimental data.
Optimization details and outcome are shown in Table
C2-2 which should be compared with the starting information in Table C2-1.


Note that in every case, the optimized value of RO

(5.390 MPa) is greater than the value of RA
(column F) for the ve solvents which were shown to
dissolve the PES polymer. This means that these
solvents are correctly positioned inside the solubility
sphere (all GOOD and IN).
And, in every case, the optimized value of RO (5.390
MPa) is less than the value of RA (column F) for the 36
solvents which were shown not to dissolve the PES
polymer. This means that these solvents are correctly
positioned outside the solubility sphere (all BAD and
OUT).

The optimized9 HSP values for PES polymer, which produce

a DF 1, are ddisperse 18.8135752121003, dpolar

9
With the Quattro-Pro spreadsheet optimization function, the author
has found the following settings to provide superior results: Quadratic
estimates of subsequent trial values, central derivatives, and a conjugate
search routine for fertile regions in which to locate an optimum.

C2

Appendix

11.2436835925914, dhydrogen bonding 7.86109904033407,

and RO 5.38982536111856 MPa10.
The progress of the optimization can be seen in Figure
C2-3. Only the rst 20 iterations are shown for simplicity.
Note that the right-most values in Figure C2-3, the nal
optimized values for each parameter, are not those fully
determined at the 21st iteration.
Note also the substantial level of convergence obtained in
this case with eight or so iterations. The entire optimization
takes less than 1 to 5 seconds on a personal computer.
Corresponding to the two-dimensional idealized Figure
C2-1 (or Figure 2.42) is the two-dimensional data-based
Figure C2-411.
Because DF 1, all 41 solvents have been positioned
correctly relative to the three- dimensional solubility sphere12.
The green and orange lines in Figure C2-4 are the calculated
two-dimensional projections of the three- dimensional
periphery of the solubility sphere (columns I and J). They are
calculated from Equation 2.8, which is a quadratic equation
with two roots (the green and orange colored lines).
Occasionally, the starting values recommended here will
not produce an optimum with a DF value at or near 1. Then,
one must select new trial values by trial and error which are
somewhat remote from the starting values recommended
here. In this case the optimization routine has located a local
optimum rather than the desired global optima.

B.1 Spherical Reasoning

The dfi function, as Equation C1-3, is based on an arithmetical
difference of two distances in three-dimensional HSP space.
The difference is point-to-pointda straight line separation.
10

Values reported in Chapter 5 of the reference of Endnote A for

ddisperse, dpolar, dhydrogen bonding, and RO are 19.6, 10.8, 9.2, and 6.2
respectively. The DF value is given as 0.999. The excess of digits reported
for each of the four parameters is not signicant. The values are printed
here to show the power of the optimization algorithms.
11
Figure C2-5 is plotted at the ddisperse value for the center of the
solubility sphere.
12
One might inquire why spherical geometry was chosen to represent
the coherence (or lack of it) between soils/polymers/solvents, and
solvents. Consider these three reasons: (1) There are three different HSP
values and a sphere incorporates three dimensions, (2) it's simple, easily
visualized, and communicated, and (3) the use of spherical geometry
generally works well enough.
Yet, Equation 2.8 is not that of a sphere; the number 4 would not
be present in the three-dimensional equation for the surface of a sphere.
So the actual geometry used in determination of HSP is not spherical, but
it is stated to be such.
Actually, there are numerous literature examples of ellipsoidal
geometry being used to describe the solubility envelope between
soluble/miscible, and not so. Dr. Hansen comments upon this point on
page 161 of Endnote A. Further, US Department of Energy Project
DE-FC26-98FT40323-036 was completed earlier in the last decade to
develop a visualization technique to examine various non-spherical
geometries as effective methods of characterizing solubility data
involving crude oil and derivatives.
Further, there is no a priori reason to assume that the shape of the
solubility envelope must be free of inection points, where some
sectors of composition are more or less effective in incorporating all
materials into a solution.

573

SOLUBILITY
GRADE 1
OK; 0 NOT

d
Disperse

MPa
B

Acetone

15.5

10.4

Acetophenone

19.6

8.6

Benzene

18.4

1-Butanol

16

5.7

Butyl Acetate

15.8

GammaButyrolactone

19

Carbon
Tetrachloride

Chlorobenzene

HSP
RA

GOOD
& IN

BAD &
OUT

BAD
& IN

GOOD
& OUT

Sphere
Error

DF

RA < RO

RO > RA

1.00000

6.736

0.26

3.842

3.7

6.175

1.24

0.806

12.707

0.001

1,505,305

15.8

11.199

0.003

333.43

3.7

6.3

9.781

0.012

80.747

16.6

7.4

5.389

1.001

0.999

17.8

0.6

13.537

3543.98

19

4.3

9.094

0.025

10.629

Chloroform

17.8

3.1

5.7

8.666

0.038

26.474

10

Cyclohexanol

17.4

4.1

13.5

9.53

0.016

62.818

11

Diacetone
Alcohol

15.8

8.2

10.8

7.634

0.139

7.2

12

oDichlorobenzene

19.2

6.3

3.3

6.771

0.251

3.978

13

Diethylene Glycol

16.6

20.7

13.602

3684.82

12

14

Diethyl Ether

14.5

2.9

5.1

12.315

0.001

1017.95

15

Dimethyl
Formamide

17.4

13.7

11.3

5.085

1.357

0.737

16

Dimethyl
Sulfoxide

18.4

16.4

10.2

5.722

0.717

1.394

17

1,4-Dioxane

19

1.8

7.4

9.462

0.017

58.701

18

Ethanol

15.8

19

Ethanolamine

17

20

Ethyl Acetate

15.8

8.8

19.4

13.246

2580.442

15.5

21.2

14.464

8724.396

5.3

7.2

8.491

0.045

22.216

Estimation of Hansen Solubility Parameters (HSP) from Binary DatadPES

A
1

d
H
Bonding

[ 7.8611;

C2

SOLVENT

d
Polar

Bonding

Appendix

574

Table C2-2 Final Determination of Solubility Parameters for PES Polymer [Optimized Values are: ddisperse [ 18.8136; dpolar [ 11.2437; dH
RO [ 5.390]

Ethylene
Dichloride

19

7.4

22

Ethylene Glycol

17

23

Ethylene Glycol
Monobutyl Ether

16

5.1

12.3

24

Ethylene Glycol
Monoethyl Ether

16.2

9.2

25

Ethylene Glycol
Monomethyl
Ether

16.2

9.2

11

4.1

0.999

1.001

493,746.82

9.44

0.017

57.407

14.3

8.542

0.43

23.378

16.4

10.218

0.008

125.014

26

5.391
18.5

26

Formamide

17.2

26.2

19

18.926

756,036.88

27

Hexane

14.9

15.795

33,029.64

28

Isophorone

16.6

8.2

7.4

5.392

0.998

1.002

29

Methanol

15.1

12.3

22.3

16.271

53,190.17

30

Methylene
Dichloride

18.2

6.3

6.1

5.39

31

Methyl Ethyl
Ketone

16

5.1

6.658

0.281

3.553

32

Methyl Isobutyl
Ketone

15.3

6.1

4.1

9.486

0.017

60.11

33

Methyl-2Pyrrolidone

18

12.3

7.2

2.05

22.228

0.035

34

Nitroethane

16

15.5

4.5

7.815

0.088

11.307

35

Nitromethane

15.8

18.8

5.1

10.052

0.009

105.895

36

2-Nitropropane

16.2

12.1

4.1

6.496

0.331

3.024

37

Propylene
Carbonate

20

18

4.1

8.089

0.067

14.86

16.8

9.4

23.3

16.062

43,123.79

16.8

5.7

6.853

0.231

4.322

11.571

0.002

483.772

8.691

0.037

27.133

40

Toluene

18

1.4

41

Trichloroethylene

18

3.1

5.3

41

1
41

575

Appendix

Propylene Glycol
Tetrahydrofuran

C2

38
39

Estimation of Hansen Solubility Parameters (HSP) from Binary DatadPES

21

Appendix

C2

Estimation of Hansen Solubility Parameters (HSP) from Binary DatadPES

Figure C2-3 Progress of Iteration to Produce an Optimum



One distance, RO, is the range over which one chemical

(solvent, soil, or polymer) has been shown to dissolve (or
be dissolved by) other solvents. It is found by optimization
based on solubility data with a set of solvents.
The other distance, RA, is the separation distance from
that chemical to any specic solvent. It is found by
calculation based on the locations of each material in
three-dimensional HSP space.

The volume of the sphere for the chemical is 4 RO/3.

The volume of the sphere for the solvent is 4 ROi/3. The
center of both spheres is presumed to be the samedthe
HSP values of the chemical. In this consideration, the dfi
value is given by Equation C2-1:

There is an RA for every solvent relative to that soil

chemical. If RA is less than RO, then there is a high likelihood that the solvent will dissolve that chemical. If RA is
larger than RO, then there is a low likelihood that the
solvent will dissolve that chemical.
Another way to consider this behavior is to treat
the chemical as a sphere of some volume in three-dimensional HSP space, and to enquire if the location of any
solvent is within that volume. In other words, nd out if
the two spherical volumes overlap one another.

When the two denitions of dfi as Equations C1-3 and

C2-1 are used with the desirability function (DF) of
Equation C1-4, the same optimized results are produced.
One gets the same optimized values of ddisperse, dpolar,
dhydrogen bonding, and RO without regard to whether the
solvent and chemical (solvent, soil, or polymer) are
sufciently proximate to one another on either a linear or
volumetric basis. This outcome is because of the properties
of DF. In other words, the numeric values in columns K
and L of Table C2-1 are different depending upon which



df i exp R A  R O 3 jfFACTORgi

Figure C2-4 Two-Dimensional Plot of Outcome of Optimization Routine

576

C2  1

Estimation of Hansen Solubility Parameters (HSP) from Binary DatadPES

dfi is used, but the optimized HSP and the RA values are
the same.
The geometric perspective about solubility is directly
derived from the dening relationship among three independent solubility parameters, Equation 2.6, and its
derived relationship, Equation 2.8.

C2

Appendix

Endnotes
A. Hansen CM. Hansen Solubility Parameters: A User's Handbook, 2nd
ed. Boca Raton FL: CRC Press; June 2007. ISBN 0849372488. They
are reproduced from Table 5.4A. HSP values of many solvents are
listed in Appendix A Table A-2 of this reference.

577