Estimation of HSP from Solvent Mixtures

a cold cleaning process, but only the effect of the

solvent7.
Test results and analysis for the PR1 photoresist are
collected into a spreadsheet as Table C4-1 with columns A
through I and rows from 1 to 80.
Columns A and B contain identication of the 80 solvent
blends and the time to dissolve the photoresist (PR) in
a test protocol8 applied throughout the experiment. The
blend composition noted in column A is the volumetric
percent of the rst-noted solvent.
The preferred value of Solubility Grade of 1 was assigned
to a test result if solution occurred within 30 seconds. The
desirable value of 2 was for solution within 60 seconds, the
acceptable value of 3 for within 120 seconds, the unacceptable value of 4 for within 300 seconds, and the useless
value of 5 for anything beyond that. The assigned Solubility Grades are shown in column C.
Blend values of the three HSP are given in columns D9, E,
and F. Each is separately calculated using the general
volumetric blend rule of Equation A2-7 below. The symbol
v is the volumetric fraction of the ith blend component, and
symbol di is any10 solubility parameter for the ith blend
component. See Appendix A3 for a derivation of Equation
A2-7 from an energy balance of cohesive energy.
qX



dany
A2  7
vi  di 2
Equation A2-7, when restated for only two components, is
Equation A2-7A. For a binary system, of course, v2 1  v1.
The values of d for each solvent are obtained from Appendices
A1 or C12 in this book, or the reference noted in Chapter 2,
Endnotes U and Z (Hansen) and endnote N (Barton).
q
X
dany
v1  d1 2 1  v1  d2 2 A2  7A
The values of RA separating each solvent and the trial values of
HSP for the PR1 photoresist are noted in column G of Table
C4-1. The unoptimized trial values are 15.0, 5.0, 5.0, and 5.0
MPa for ddisperse, dpolar, dhydrogen bonding, and RA respectively.
The optimization described in Appendices C2 and C3,
using Equation C1-4, was completed using all 80 data
points in Table C4-1.

7
Absence of agitation motivates the above concern about whether or
not the solubility results were limited by mass transfer of the solvent into
the PR, or were equilibrium results.
8
Blanket PR layers were obtained by coating the PR onto chemically
cleaned p-type (100) silicon substrates, followed by a soft bake at
130 C. Dissolution tests were performed on wafer pieces in a beaker at
room temperature under gentle manual agitation. Between 30 s and
10 min of immersion, samples were withdrawn from the solvent at
regular time intervals. Samples were then either directly dried with a N2
gun or rst rinsed in DI water before drying. When samples looked
visually clean, Fourier Transform Infrared (FTIR) spectroscopy was used to
examine for typical peaks and accurately determine the time needed for
complete dissolution of the PR layer.
9
No solubility parameters were found for NEP and its position in the plot
is based on the position of NMP and is only approximate.
10
Disperse, hydrogen bonding, polar, or Hildebrand.

C4

Appendix

Actually, the optimization was completed ve times, each

with different standards of cleaning performance (Solubility
Grades) applied, ranging from 1 to 5. Recall that the value 1
meant that the solution of the PR1 was accomplished within
30 seconds, and 5 meant that the solution of the PR1 was
accomplished outside 300 seconds, or not at all. In other
words, the successive optimizations accepted increasingly
poorer cleaning speed as acceptable.
The results, for all ve optimizations, are shown in
Table C4-2 (above).
In Table C4-2, since the value of DF (1.1806) is greater
than 1, a complete and successful optimization was
not obtained. This means that some solubility results
were BAD (Solubility Grade >1) and located within the
solubility sphere of radius RO (Column J) or GOOD
(Solubility Grade 1) but located outside the solubility
sphere of radius RO (Column K). There were eight of the
former and seven of the latter; fteen in total.
The remaining 65 solvent blends dissolved the PR1 in 30
seconds and were located within the solubility sphere of
radius RO (Column I), or did not dissolve the PR1 in 30
seconds and were located outside the solubility sphere of
radius RO (Column H). These are the outcomes expected
of successful use of HSP technology: differentiation and
prediction.
Results with other standards of performance and longer
solution times are listed in Table C4-2.
One should notice in Table C4-2 that the full optimization (DF 1) is not obtained until the performance of
all 80 solvent blends is forced to be located within a threedimensional solubility sphere with an outlandish and
useless radius (RO 29.33 MPa). This outcome provides
no discrimination about solution performance, and no
useful information. It is not a valuable optimization.

E. ANALYSIS OF UNEXPECTED
SOLVENCY PERFORMANCE
The reason for the incomplete or awed optimization of data
in Table C4-1 is that 15 solvents or blends did not perform as
expected. Some solvent combinations were outside expectations by a small and nearly negligible amount. A few were
outside expectations by a signicant amount.
This is not an abnormal outcome.
Most of the 15 data points of Figure C4-1 have values
of excess spherical radius11 within  1 unit of radius

11

Only the 15 data points whose result was not properly predicted
are plotted in Figure C4-1. The vertical axis is excess sphere radius
compared to the optimum value (RO), i.e, (RO  RA). Positive values
represent data points outside the solubility sphere (which did dissolve
PR1 better than expected). Negative values represent data points inside
the solubility sphere (which did not dissolve PR1 as well as expected).

589