Progress in Organic Coatings 45 (2002) 101–117

New challenges for R&D in coating resins

Emiel Staring a , Aylvin A. Dias b,∗ , Rolf A.T.M. van Benthem b,∗
a DSM Coating Resins, P.O. Box 615, 8000 AP, Zwolle, The Netherlands
b DSM Research, P.O. Box 18, 6160 MD, Geleen, The Netherlands

Received 1 September 2001; accepted 13 March 2002

Abstract
The coatings industry has experienced a strong drive towards new products during the last couple of decades. New products must
either allow lower total “system” cost and/or have better performance, and/or lower impact on the environment. A number of trends like
globalisation, consolidation, specialisation and increased competitiveness can be discerned. These developments encourage a different
approach towards coatings development. New elements that come into play are speed of development and decomplexation. These elements
will influence the way the coating (resins) R&D is performed.
The central theme for addressing these questions, at least in the area of thermosetting coatings, in DSM’s view is a better, more thorough
understanding of the relationship between structure and properties: from craftsmanship to science. This relationship must be clarified on
at least three levels: the molecular, the mesoscopic and the temporal level.
One of the most important challenges for R&D in coating resins is to use and/or develop new characterisation techniques that relate
structure (of molecules and networks) to properties. Using recent developments from DSM laboratories various approaches to meet these
challenges are presented. It is our considered opinion that by having precise characterisation of raw materials with techniques that allow us
to monitor temporal and spatial changes we will be able to exercise control over chemistry, network and mechanical properties to achieve
performance coatings.
In order to characterise coating resins with non-linear architectures we have used the combination of size-exclusion chromatography
with mass spectrometry to exactly determine molecular weight distributions and molecular monomer compositions. This is illustrated with
the full characterisation of a set of a priori calculated and synthesised hyperbranched polyesteramides. Compliances as well as deviations
from theory have been established and explained.
The improved understanding of photocured networks and their heterogeneity has been derived using solid state proton NMR T2 relaxation,
where relaxation times can be related to the modulus of photocured coatings. This technique is of high utility to the coatings industry due
to the fact that it allows one to probe effects on a mesoscopic level with minimal sample preparation. Time-resolved chemical changes
have been monitored using RT-FTIR (real time also for photocuring systems). These techniques allow us to relate chemical changes to a
prescribed development of network topology and the eventual macroscopic properties like modulus and Tg .
© 2002 Elsevier Science B.V. All rights reserved.
Keywords: Structure–property relations; Hyperbranched polyesteramides; Photocured networks; NMR proton relaxation; Time-resolved FTIR

1. Introduction Coatings and paints are performance materials of which

the relation between molecular composition and macro-
The coatings industry has experienced a strong drive to- scopic properties is poorly understood. In many cases, the
wards new products during the last 20 years with ever in- most effective way to develop better coatings was and still
creasing diversity of applications. New products must either is, based on systematic testing of new resins, x-linkers
allow lower total “system” cost and/or have better perfor- and/or additives in a sheer endless variety of combinations
mance, and/or lower impact on the environment. The golden (Edisonian type of research). Most of this empirical knowl-
rule turned out to be that successful new products scored edge is present within industrial laboratories, and thus not
better on at least one of the above drivers with two others at universally accessible. This approach has resulted in a situa-
least equal. tion where coatings consist of a large number of ingredients.
Furthermore, numerous grades designed for specific appli-
cations have further broadened the range of raw materials
∗ Corresponding authors. used. The result is a product portfolio with ever-increasing
E-mail address: rolf.benthem-van@dsm.com (R.A.T.M. van Benthem). complexity and the accompanying logistical problems.

0300-9440/02/$ – see front matter © 2002 Elsevier Science B.V. All rights reserved.
PII: S 0 3 0 0 - 9 4 4 0 ( 0 2 ) 0 0 0 6 6 - 8
102 E. Staring et al. / Progress in Organic Coatings 45 (2002) 101–117

Global trends like consolidation and specialisation able to exercise control over chemistry, network topology
demand a different approach to coating development. New and mechanical properties to achieve performance coatings.
elements that come into play are speed of development and
decomplexation. These elements will influence the way the
coating (resins) R&D is performed. Important challenges 2. Characterisation of hyperbranched resins
are: how can we develop better coating (resins) in a faster
way than done until now? And how can we develop better 2.1. Calculated synthesis of polyesteramides
coating (resins) without increasing the complexity?
The central theme for addressing these questions, at least The basic concepts of the synthesis of hyperbranched
in the area of thermosetting coatings, in DSM’s view is polyesteramides as economically feasible counterparts of the
a better, more thorough understanding of the relationship well-known dendrimers, as well as their promising applica-
between structure and properties: from craftsmanship to tions in a variety of different coating formulations have been
science. This relationship may be addressed on at least three described previously [1,2] and presented at this conference
levels: in 1999 [3].
In summary, our synthetic approach bears a strong re-
• The molecular level. The molecular structure of the poly-
semblance with the classic A2 /B3 approach [4] which is
mers involved in the coatings network formation is often
unsuitable for higher molecular weights. In our approach,
not fully known. Monomer sequences, type and amount
however, one (A) group denoted (a) of the A2 -compound
of endgroups are known in general terms but not always
and one (B) group denoted (b) of the B3 -component, which
in sufficient detail to precisely understand their role in
is present in molar excess, are preferentially reactive towards
film formation. The introduction of new polymer architec-
each other. In this way, in the pre-reaction A-a-b-B2 -units
tures, like stars, combs, dendrimers, hyperbranched, and
(i.e. functional AB2 -type units) are formed while the molar
supramolecular assemblies promises much to further de-
excess of B3 -units is retained in the system [5]. This excess
velop the general field of polymer chemistry and in par-
limits molecular weight build-up and eliminates the risk of
ticular for coating resins. The determination of molecular
gel formation at higher molecular weights, contrary to the
weight distributions, however is no longer reliable with
classical A2 /B3 approach. Thus, the eventual hyperbranched
conventional calibration standards and new characterisa-
polymer is composed of n “Aa” units and (n + 1) “B2 b”
tion techniques must be applied. In addition improved un-
units and bears (n + 3) (B) end groups.
derstanding of the polymer chain topology is needed.
The exothermal reaction of 1,2-cyclohexane-dicarboxylic
• The mesoscopic level. It is generally acknowledged that
acid anhydride and diisopropanolamine normally leads
the network as formed by coating binders is a principal
to a molecule with one carboxylic acid group and two
feature that contributes to mechanical and some chemi-
2-hydroxy-propylamide groups. This molecule exemplifies
cal properties. Network topology of the resultant curing
the A-a-b-B2 unit. The 2-hydroxy-alkyl groups formed (B)
systems with attention to mean molar mass between
can react readily with the carboxylic acid groups (A). The
cross-links, trapped and temporary entanglements, dan-
esterification reaction between a 2-hydroxy-alkyl group and
gling chain ends and chain mobility as well unbound
a carboxylic acid is known to proceed much faster com-
migrateable materials are of prime importance for the
pared to normal aliphatic primary or secondary alcohols
coatings performance. There are, however, not many stan-
[6]. Instead of an addition–elimination reaction mechanism,
dard characterisation techniques available for studying
the formation of an intermediate oxazolinium–carboxylate
cross-linked materials in such physical/chemical detail.
ion pair is thought to prevail, which strongly facilitates
• The temporal level. In order to understand the final per-
the esterification reaction. Typically, this reaction can take
formance of industrial products, it is important to know
place at temperatures from about 140 ◦ C without a catalyst.
their relation to the production processes in which they
It also possible to introduce a monocarboxylic acid (A1)
find their origin. Especially, the network formation during
as a third component in the reaction mixture in a one-pot
cure of coatings is a strictly time-dependent phenomenon,
procedure. Depending on the molar ratio of A1 (m) to the
and accordingly is the development of properties as a
other two components in the system, the monocarboxylic
function of processing.
acid can be used merely as endgroup modifier (m < n + 3)
Using recent developments from DSM laboratories var- or as chain stopper (m > n + 3). In this study lauric acid
ious approaches to meet these challenges are presented. (1-dodecanoic acid) was used. When the reaction between
Section 2 deals with characterisation on the molecular level groups (a) and (b) is considered to occur instantaneously and
of polymers with new architectures, Section 3 deals with cycle formation is excluded according to Flory [4], the aver-
characterisation of photocured coatings in terms of network age molecular weights Mn and Mw of the resulting polymers
topology and using new time-resolved techniques. It is our can be directly calculated from the monomer composition as
considered opinion that by having precise characterisation a function of the chemical conversion according to Durand
of raw materials accompanied by powerful techniques that and Bruneau [7], optionally treating the mono-functional
allow us to monitor temporal and spatial changes we will be carboxylic acid A1 as a third component.
 E. Staring et al. / Progress in Organic Coatings 45 (2002) 101–117 103

Table 1 branched polyesteramides:

Description of the three series of target molecules
• What is the exact monomer composition of each polymer
Resin D/C Lauric ester
numbers modification (DS)
chain? (Are there side-reactions or other anomalies?)
• What is the exact molecular weight distribution? (Does it
Series I 1–5 1.5–1.14 –
Series II 6–9 1.14 or 1.20 Partially (0.20–0.80) fit with the calculated values?)
Series III 10–14 1.55–1.05 Fully • What is the degree of branching or the fractal dimension
of these polymers? (Are they really hyperbranched?)
The first question has been answered by MALDI-TOF
Experimental. Diisopropanolamine was introduced in a mass spectrometry [8]. The latter two questions could be
1 l glass reactor under a nitrogen atmosphere. The calcu- answered by size exclusion chromatography with on-line
lated amount of 1,2-cyclohexanedicarboxylic acid anhydride differential viscosimetry detection (SEC-DV) in principle.
was then added portionwise. As a result of the immediate The determination of exact the molecular weight distribu-
exothermic reaction, the temperature of the mixture raised tion with conventional SEC techniques (RI or UV detec-
to ca. 120 ◦ C. Lauric acid was optionally introduced, with tion) which are based on polystyrene calibration standards,
the total of the calculated amounts of raw materials amount- is not feasible since the hydrodynamic volume of a highly
ing to about 600 g. The reaction mixture was then heated to branched polymer is intrinsically different from a linear one.
180 ◦ C. The evolving reaction water was distilled off in 6 h, There are simply no suitable calibration standards for highly
ultimately under reduced pressure. At the desired acid con- branched polymers. SEC-DV relies on the universal cali-
centration but ultimately after 8 h, the hot resin was poured bration principle [9] which correlates the intrinsic viscos-
out and cooled to room temperature. ity of a polymer collection in one SEC-slice to its absolute
According to a set of calculated targets, three series of hy- molecular weight. Although this principle was thought to be
perbranched polyesteramides were synthesised, see Table 1, valid for hyperbranched polymers as well, and gives indirect
in which the ratio diisopropanolamine/anhydride (D/C) and information on the hyperbranched character of these poly-
the degree of substitution (DS) with lauric acid were varied. mers through the Mark–Houwink relation, its validity had
Since carboxylic acid groups are theoretically completely to checked. We have attempted this by combined SEC-DV
consumed in all these polycondensation reactions, the resid- and (off-line) SEC-MS (ESI and MALDITOF-MS) mea-
ual amount of carboxylic acid as determined by potentio- surements.
metric titration was taken as a relative quantitative measure Table 2 summarises the experimental SEC-DV results, in
of the total chemical conversion. The targeted and recalcu- correlation with the calculated molecular mass moments,
lated moments are given in Table 2. of the three series of hyperbranched polyesteramide resins
displayed in Table 1. Fig. 1 illustrates the SEC elugrams of
2.2. SEC-DV and SEC-MS for characterising three unsubstituted polyesteramides.
polyesteramides Experimental. The SEC measurements were performed
on a Hewlett-Packard chromatograph (HP 1090) equipped
Three imminent questions have to be answered in order with a differential refractometer (RI) and a differential
to address the molecular characterisation of these hyper- viscosimeter (DV) placed in parallel (Viscotek200). The

Table 2
Collected data of hyperbranched polyesteramide resins
No. DS D/C Target Mn a p(A) Calculated Found (SEC-DV)

Mn a Mw a Mn a Mw a

1 0 1.47 0.7 98.8 0.68 1.6 1.5 3.6

2 0 1.35 0.9 98.8 0.87 2.1 1.8 5.9
3 0 1.26 1.2 98.8 1.10 2.9 2.4 11
4 0 1.20 1.5 98.4 1.36 3.9 2.4 59
5 0 1.14 2.0 98.9 1.85 6.1 2.8 250
6 0.2 1.14 2.4 98.6 2.13 6.7 2.8 54
7 0.4 1.14 2.8 97.1 2.11 6.7 2.1 19
8 0.6 1.14 3.1 94.9 1.88 6.0 2.1 12
9 0.8 1.14 3.4 94.4 1.88 6.1 1.8 8.5
10 1.0 1.55 1.5 91.1 0.95 2.0 1.1 2.5
11 1.0 1.25 2.5 91.5 1.29 3.3 1.4 3.4
12 1.0 1.20 3.0 96.7 2.07 5.6 2.1 6.8
13 1.0 1.13 4.0 89.5 1.41 4.9 1.4 4.9
14 1.0 1.08 6.0 91.1 1.85 8.7 1.6 5.4
a All values in 103 g mol−1 .
104 E. Staring et al. / Progress in Organic Coatings 45 (2002) 101–117

Fig. 1. SEC elugrams of 1,3 and 5; CD2 (1), C2 D3 (2), C3 D4 (3), C4 D5 (4) oligomers indicated.

polyesteramides were dissolved in dichloromethane in a ni- was observed with the fully substituted resins, especially,
trogen atmosphere for at least 4 h; all polymer solutions and some of the partially modified resins (8 and 9). In the
were filtered (pore size 0.5 ␮m). Four Polymer Laboratories series of unsubstituted resins, the Mn and especially the Mw
mixed C columns were used. From both the concentration values found with SEC-DV were substantially larger than the
and viscosity chromatograms of the polyesteramide sam- a priori calculated values in all cases, as was also observed
ples, the intrinsic viscosity (η) as a function of elution vol- for partly modified resins 6 and 7.
ume was determined. From these data the average molecular These combined observations can be explained by the
weights (Mn , Mw and Mz ) were calculated using universal occurrence of some side reactions between unmodified
calibration. 2-hydroxy-propylamide endgroups leading to chain exten-
MALDI-TOF mass spectrometry was carried out us- sion, whereas lauric acid substitution prevents those. A
ing a Perkin Elmer/Perseptive Biosystems Voyager-DE-RP proposed mechanism for this side reaction involves nucle-
MALDI-TOF mass spectrometer equipped with delayed ophilic ring opening of the oxazolinium ion intermediate
extraction. A 337 nm UV Nitrogen laser producing 3 ns by a secondary amine, instead of the associated carboxylate
pulses was used, and the mass spectra were obtained in the ion [2].
linear and reflectron mode. Positive ion electrospray (ESI) As a consequence of this chain extension aberrant
mass spectra of the polyesteramides were recorded using a polyesteramides comprising n anhydride units and (n + 2)
PE Sciex API 150 single quadrupole mass spectrometer. diisopropanolamine units are formed. This is confirmed in-
For the validation of the universal calibration (SEC- dependently by the ESI mass spectrometry of the resins 1
MALDI-TOF-MS), fractions of one sample (1 in Table 2) and 2, in which substantial amounts of the aberrant struc-
were collected during an SEC run for off-line MALDI-TOF- tures with (n + 2) diisopropanolamine units have been
MS. The volume offset between the starting point of each identified. Fig. 2 shows these structures in a slice of the
fraction and the corresponding RI signal was determined SEC-DV of resin 1 together with the parent structures
by placing a UV detector behind the RI detector at the (n + 1) and cyclic structures (n).
fraction collector starting point. The offset volume was de- An increase in degree of lauric ester substitution can
termined using a PS calibration standard (molecular mass reduce the amount of chain extension side reactions in two
19,000 g mol−1 ). ways. On one hand, the higher concentration of carboxylic
The SEC fractions were diluted for MALDI-TOF-MS by acid groups with respect to the amine groups in the system
adding tetrahydrofuran. The MALDI-TOF-MS spectra were disfavours nucleophilic attack by the amine groups through
processed to yield values for Mn and Mw . The polydispersity protonation. On the other hand, the concentration of un-
(Mw /Mn ) of a fraction typically was 1.02. The calculated reacted 2-hydroxy-propylamide in the polycondensation
molecular masses via MALDI-TOF-MS were correlated to mixture decreases faster with progressing chemical con-
the SEC elution volume but additionally corrected for the version. It was confirmed by ESI-MS data that the partly
total fraction volume. modified polyesteramides contain smaller amounts of aber-
When the experimental SEC-DV results were compared rant structures with increasing degree of lauric ester subs-
to the calculated molecular mass moments, a close match titution.
 E. Staring et al. / Progress in Organic Coatings 45 (2002) 101–117 105

Fig. 2. Slice of SEC elugram of 1 as monitored by ESI-MS.

Both MALDI-TOF and ESI MS confirmed the generally in our calculative approach. Therefore, our calculated val-
expected structures including the increase of lauroyl sub- ues in Table 2 may need some correction; further studies to
stitution of the polyesteramide backbones as a function of establish this are underway.
the designed degree of substitution. However, next to the The hyperbranched character of the polyesteramides un-
aberrant structures in some samples, both MS techniques re- der investigation was confirmed in two independent ways
vealed the presence of cyclic structures in all the samples [2]. First, Mark–Houwink plots revealed slopes (α) of be-
investigated, see also Fig. 2. Although cyclisation reactions tween 0.295 and 0.373 for unsubstituted polyesteramides
were expected to occur, they could not be taken into account (see Fig. 3) and between 0.288 and 0.335 for the substituted

Fig. 3. Mark–Houwink plots of 1–5.

106 E. Staring et al. / Progress in Organic Coatings 45 (2002) 101–117

Fig. 4. Comparison of SEC-DV results with SEC-MALDI-TOF-MS.

ones. These values indicate a more or less spherical shape. photo-induced cationic polymerisation based on epoxide and
For comparison: a normal random coil gives α = 0.7 and vinylether functional groups. In the effort to relate mechan-
not less than 0.5 under Θ (poor solvent or melt) conditions, ical properties to chemistry it is essential to understand the
whereas a true dendrimer has α = 0.1. Secondly, the fractal network topology of the resultant cured coating. Mixtures
dimension as obtained with small-angle neutron scattering of a di- and mono-functional acrylate were used as a model
(SANS) was between 2.19 and 2.26 for the unsubstituted system for this study. The mean cross-link density and the
polyesteramides 1–5. Again in comparison, a linear polymer fraction of network defects were varied in these networks by
is expected to yield values around 1.6 and no more than 2 changing the content of mono-functional acrylate. Network
at conditions (“two-dimensional”), and a dendrimer would structure of UV-cured acrylates were analysed in the present
give value around 3 (truly “three-dimensional”). study by mechanical experiments, NMR T2 relaxation and
Fig. 4 finally shows the comparison the molar masses 13 C NMR spectroscopy [10]. The aim of the present study

calculated using universal calibration [9] (SEC-DV) for the was to establish correlation between the mechanical prop-
fractions of resins 1 and 2 with those obtained from the erties, as measured by DMA, and the network structure as
absolute MS method on the fractions of resin 1. Since the probed by NMR T2 relaxation experiments.
molecular mass in SEC-DV for each slice corresponds to
the number average molecular mass, comparison with the 3.1.2. Preparation of UV-cured coatings
number average molecular masses from MS is made. The Mixtures of polyethylene glycol di-acrylate (PEG700DA)
MS and SEC-DV values obviously agreed very well. Also with mono-functional 2-ethylhexyl acrylate (EHA, see
on a linear y-scale the differences are small, indicating that Scheme 1) were cured on glass plates at 27 ◦ C on a con-
for these highly branched samples the universal calibration veyor belt fitted with a Fusion F600 (6000 W) electrodeless
can very well be applied, at least in the molecular mass range H bulb, and with nitrogen inerting. The weight fraction of
studied here. EHA was varied from 0 to 90%. Of course, these numbers
are of little practical value, in view of the SHE properties of
EHA, but in this study its model function is highly appre-
3. Characterisation of coating networks ciated. The mixtures contained 1 mass% of photoinitiator
1-hydroxycyclohexyl phenyl ketone (HCPK). A UV dose
3.1. NMR T2 relaxation studies in photocured of 1 J/cm2 was determined using a UV Power PuckTM light
coatings [10] meter.

3.1.1. Introduction 3.1.3. Network density determination by 1 H T2 NMR

Photocuring has emerged as one of the most enabling relaxation experiments
technologies for the coatings industry with significant ad- Experimental: The proton transverse magnetisation de-
vantages like cure speed, reduced emission (solvent-free), cays, T2 relaxation decays, were measured on a Bruker
ease of applicability and lower overall systems costs. To Minispec NMS-120 spectrometer at a proton resonance fre-
date, free radical polymerisation of acrylates dominates over quency of 20 MHz. This spectrometer was equipped with a
 E. Staring et al. / Progress in Organic Coatings 45 (2002) 101–117 107

Scheme 1. Components of the photocuring formulations.

BVT-3000 variable temperature unit. The temperature gra- 1,1,2,2-C2 D2 Cl4 was added to the sample, then a Teflon plug
dient and stability was about 1 and 0.1 K, respectively. is inserted so that the bottom is just above the sample creat-
The decay of the transverse magnetisation was mea- ing a slight free volume above the sample. The samples were
sured with the Hahn-echo pulse sequence (HEPS), stored for 1 day before the measurements were performed.
90x ◦ –tHe –180x ◦ –tHe –acquisition, where tHe = 35 ␮s. An In this case, 1,1,2,2-C2 D2 Cl4 was solvent of preference due
echo signal is formed after the second pulse in the HEPS to its high boiling point, which permitted NMR experiments
with a maximum at time t = 2tHe after the first pulse. By at elevated temperatures.
varying the pulse spacing in the HEPS, the amplitude of the A distinguishing feature of T2 relaxation for thermoset-
transverse magnetisation, A(t), is measured as a function of ting systems is the high-temperature plateau that is observed
time t. at temperatures well above Tg (Fig. 5). The temperature in-
The T2 relaxation experiments were performed on sam- dependence of T2 is attributed to constraints which limit the
ples as whole and swollen samples. A certain amount of number of possible conformations of network chain with

Fig. 5. Schematic drawing of T2 relaxation time against temperature for amorphous polymers [12].
108 E. Staring et al. / Progress in Organic Coatings 45 (2002) 101–117

respect to those of a free chain. The T2 at the plateau is

(T2 p ) quantitatively related to the mean molar mass of net-
work chains with the relationships given below [10]. For a
Gaussian chain, in which the average, squared distance be-
tween network junctions is much shorter than the contour
chain length, the T2 p value is related to Z statistical segments
between the network junctions:
p
T2
Z= (1a)
aTrl2
where a is the theory coefficient, which depends on the angle
between the segment axis and the internuclear vector for
the nearest nuclear spins at the main chain. The T2rl value,
which is measured for swollen samples below Tg , is related
to the strength of intrachain proton–proton interactions in
the rigid lattice. Using the number of backbone bonds in the
statistical segment, Cω, the molar mass of network chains
Fig. 6. Tg of cured acrylates against the content of EHA. The line repre-
between chemical and physical network junctions, Mc+e is
sents the result of a linear regression analysis: intercept = −39.0±0.4 ◦ C;
calculated: slope = −0.381 ± 0.009 ◦ C (mass%)−1 . The correlation coefficient equals
ZCωMu 0.998.
Mc+e = (1b)
n
where Mu is the molar mass per elementary chain unit for
the polymer chain and n the number of backbone bonds in The NMR relaxation rate, 1/T2 , and the storage modulus
an elementary chain unit. increase proportionally with the content of difunctional acry-
In addition to the NMR analysis, the mean molar mass late (PEG700DA). The NMR relaxation rate, 1/T2 , revealed
of network chains was calculated from the slope of the lin- a linear dependence against the storage modulus, see Fig. 7.
ear part of the dependence of the modulus at temperatures Thus, the cross-link density in cured acrylates is inversely
above Tg . The following equation relates the slope of the proportional to the content of mono-functional monomer.
dependence (E /T) to the molar mass of network chains be- The molar mass of network chains, Mc+e , as calculated
tween chemical cross-links and chain entanglements, Mc+e , from mechanical and NMR T2 relaxation data, is compared
in kg/kmol [4,11]: in Fig. 8. The results of these two methods are in good
agreement, if one accounts for assumptions made for the
3ρRT(1 − x)
Mc+e = (2) calculation of the Mc+e from the NMR and mechanical data.
E The molar mass of network chains is significantly smaller
where ρ is the specific density in kg/m3 ; R the gas constant compared to the molar mass of di-functional monomer for
that equals 8.3 J/(mol K); E the modulus in Pa and x the cured acrylates that is apparently caused by zip-like origin
volume fraction of 2-ethyl hexyls fragment of EHA. Sug- of network junctions, see Fig. 9.
gesting density additivity, a value of x was calculated from
the specific density of network chains (PEGDA) and alkylic 3.2. Real time FTIR for both liquid and powder radiation
fragment of EHA 1.11 and 0.86 g/cm3 , respectively. Eq. (2) curable systems
is applicable to affine networks of the same chain length and
without network defects. It is recognised that the effective 3.2.1. Introduction
number of elastically active network chains reduces in net- Current photocuring processes vary from a multi-second
works with spatial clustering of cross-links [12–14]. This timescale to sub-second timescale, e.g. UV powder coatings
means that a value of Mc+e is overestimated with Eqs. (1a) and optical fibre coatings, respectively. The techniques used
and (1b) in the case of heterogeneous networks. to monitor the degree of cure and the kinetics of photoiniti-
As expected, the glass transition temperature of the com- ated polymerisation reactions are comprehensively reviewed
pounds in this study increased linearly upon an increase in by Rabek [16].
the content of di-functional monomer, as shown in Fig. 6. Infra Red spectroscopy has emerged as one of the most
Fox and Loshaek [15] or similar equations can describe the versatile techniques in radiation curing since it yields the
dependence: following information about a photocuring reaction:
kn
Tg = Tg∞ + (3) • the kinetics of UV-based radiation curing/polymerisation,
Mc • the degree of reactant conversion [17],
where Tg∞ is the Tg of a polymer of infinite molecular weight • the induction time related to, for example, air or moisture
and kn the material constant. inhibition.
 E. Staring et al. / Progress in Organic Coatings 45 (2002) 101–117 109

Fig. 7. The relaxation rate (1/T2s )max for samples partially swollen in 1,1,2,2-C2 D2 Cl4 against the storage modulus at 0 ◦ C. The rubbery plateau
is observed for all samples at this temperature. The line represents the result of a linear regression analysis: intercept = 1.1 ± 0.3 m s−1 ;
slope = 0.34 ± 0.02 m s−1 (MPa)−1 . The correlation coefficient equals 0.992 [10].

Employing thin film coatings and use of the Beer–Lambert tices have become common (depending on the type of
law allows the determination of the concentration of the application):
reactive groups, e.g. acrylates, vinylethers, maleates and
epoxides. The use of IR spectroscopy to monitor radia- • the use of nitrogen inerting to overcome oxygen inhibition
tion curing reactions was developed by Decker [18]—real in free radical polymerisations [19], e.g. in coatings for
time infra red, using a continuous wave spectrometer that optical fibres;
allowed only a single vibrational stretch. As the radi- • the use of elevated temperatures like in powder coatings
ation curing industry has developed a number of prac- [20];

Fig. 8. Mean molar mass between chemical cross-links and trapped chain entanglements in cured PEG700DA/EHA against EHA content. Mc+e is
determined by NMR T2 relaxation method using Eqs. (1a) and (1b) and from the slope of the dependence of plateau modulus on temperature. The molar
mass of PEG700DA is shown by arrow [10].
110 E. Staring et al. / Progress in Organic Coatings 45 (2002) 101–117

Fig. 9. Suggested network structure for compounds: (A) PEG700DA/EHA(80:20) and (B) PEG700DA(100) [10].

• the increase in the number of radiation curing chemistries 3.2.2. Equipment

such as free radical systems (acrylate, maleate–vinylether Acknowledging these developments, a new real time infra
[21], thiolene [22]), cationic-photogenerated acids (with red apparatus that had to meet the following requirements
epoxide [23], vinylethers and propenylethers [24]) was built [29]:
anionic-photogenerated bases [25];
• capable of measuring both liquid and solid radiation cur-
• the use of lamps with different emission spectra in the UV
able formulations;
and visible lamps for surface and through cure and for
• monitoring photocuring reactions under various atmo-
various pigmented systems (often done with photoinitiator
spheres, e.g. with nitrogen inerting;
optimisation based on their absorption spectra) [26];
• monitoring photocuring reactions at various tempera-
• the variety of monomers used in a given technology, e.g.
tures;
N-vinyl pyrollidone has been used in acrylate formula-
• allow the rates of consumption of different monomers
tions, etc. [27].
in a formulation to be monitored simultaneously during
photocuring, e.g. maleates and vinylethers, acrylates
The development of UV powder curing technology [28]
and N-vinylamides, epoxides and vinylethers in cationic
is an exciting development in the radiation curing industry.
systems.
In UV powder curing, solid powder formulations are applied
to a substrate, melted with IR heat allowing the resin to flow The RT-FTIR instrument shown in Fig. 10 may be divided
and coat the substrate and then exposing the coating to UV into three distinct parts: (i) the spectrophotometer, (ii) the
light for the curing/cross-linking reaction. illumination system and (iii) the sample.
 E. Staring et al. / Progress in Organic Coatings 45 (2002) 101–117 111

Fig. 10. RT-FTIR experimental setup for liquid and powder samples.

i) The spectrophotometer (Bruker IFS55) has been fit- The features of this cell are:
ted with a broadband MCT (mercury cadmium tel-
luride) detector. The spectrophotometer can measure • 70◦ LARI construction, to increase sensitivity.
between 20 and 40 spectra/s depending on the res- • Heatable sample holder with thermocouple (up to 500 ◦ C)
olution required. Further hardware updates included with water cooled housing.
were: acquisition processors (AQP) to store up to 600 • Digital temperature controller: this device allows to pro-
spectra in the memory of the instrument, and digital gram an initial temperature, a heating rate, and a final
input/output ports controlled via the software (TTL) temperature, similar to a gas chromatograph.
to control the UV illumination with high precision • Gas tight housing: this allows for partial evacuation of the
(1 ms). cell and/or purging with a suitable gas, e.g. a nitrogen.
ii) These systems are used together with liquid light guide
Thin coating of 10–50 ␮m is applied on a highly reflective
(10 mm) to focus the UV onto the sample. The UV light
Au-coated aluminium disk (thickness 0.5 mm) with a size
energy delivered in the test cell was 150 mW/cm2 for
of 7 mm × 7 mm and 8 mm × 8 mm. The surface of the Al
the Hg halide lamp and 75 mW/cm2 for the Hg lamp. A
disk is made of polished Al oxide, which introduced broad
shutter that is triggered by the spectrophotometer via a
bands in the single beam spectrum of the background in
delay timer. The electronic delay timer was constructed
the 3000–3700 and 1000–1800 cm−1 spectral range. These
at DSM Research.
bands not only caused an intensity reduction of the reflected
iii) A transflection (transmission–reflection) setup was cho-
light, but led to baseline problems because of differences in
sen for our system. In a transflection system, the IR
the reflection when the disk is coated with resin. To circum-
beam is reflected on a mirror plate, this plate is cov-
vent this effect, the disks were gold plated with a sputtering
ered with the sample substance. The cell which met our
technique normally used for electron microscopy.
demands as a very versatile system was a large angle
Liquid coating sample preparation: The gold plated alu-
reflectance infra red (LARI) cell. The construction is
minium disks were placed a recess (slot) of predetermined
given in Fig. 11.
depth (510 and 520 ␮m) in a steel block. A few drops of
The cell has an extra inlet that positions the optical light the liquid formulation were placed beside the slot and the
guide directly above the sample, which is in a horizontal po- formulation drawn over the Al disks. In this way a coating
sition. The end of the light guide is 1 cm away from sample. layer of 10–20 ␮m was obtained.
The horizontal setup was used to eliminate problems with Powder coating sample preparation. A 45 ␮m sieve car-
low viscosity samples. rying the powder sample was shaken above the gold-coated
112 E. Staring et al. / Progress in Organic Coatings 45 (2002) 101–117

Fig. 11. A large angle transflection (LARI) cell with horizontal, heated stage and facility for gas purging.

Al disks. The sample was then heated to 110 ◦ C to allow curve, t1 and t2 the time in seconds that correspond to the
the powder to melt and form a coating on the Al disk. This absorptions A1 and A2 . A0 is the initial absorbance at the
gave 20 ␮m thick coatings. start of the experiment. A review of such data analyses is
given by Rabek [16].
3.2.3. Photocuring experiments Some typical photocuring experiments examining the
The RT-FTIR experiment allows the entire mid-IR re- various options available with the new RT-FTIR are sum-
gion to be recorded. Twenty scans per second with a spec- marised in Figs. 14 and 15. The ability of the instrument to
tral resolution of 4 cm−1 was employed. The time to record record spectra in situ under different atmospheres is shown
one spectrum takes about 45 ms. Up to 600 spectra can be in Figs. 14 and 15, where photocuring results in wet and
recorded continuously, leading to 27 s overall duration of the dry air and under nitrogen are compared.
measurement. The effect of moisture in the air on the cationic photoini-
There are two possible ways of determining the region of tiated polymerisation of vinylether, cycloaliphatic epoxide
the IR spectrum that is undergoing the most change during and glycidylether-based formulations is shown in Fig. 15.
the curing reaction: (i) contour mapping and (ii) difference Formulations are based on 70% monomer, 28% polypropy-
spectra. leneglycol triol (Mwt-1500) and 2% cationic photoinitiator
Contour mapping allows one to identify the region of the UVI 6990.
IR spectrum undergoing maximum change during photoini- The cycloaliphatic epoxide, vinylether and glycidylether
tiated polymerisation. This is shown for the 810 cm−1 acry- peaks were monitored at 790, 815 and 840 cm−1 , respec-
late peak for the polymerisation of laurylacylate in Fig. 12. tively. The results indicate that both the vinylether and the
The difference spectra are obtained by subtracting the last glycidylether are adversely affected by moisture in the air
spectra from all the preceding spectra that were obtained (humidity) however the presence of moisture actually en-
during the photocuring reaction. In this way it is possible hances the levels of conversion of the cycloaliphatic epoxide
to monitor the peaks (functional groups) that are being con- but not the initial rate of polymerisation.
sumed as well as the bonds that are being formed. This is The ability of the instrument to monitor two different
shown in Fig. 13. monomers simultaneously and also at various tempera-
The kinetics of photoinitiated polymerisation was anal- tures is illustrated in Fig. 16. Photoinitiated acrylate ho-
ysed using the method described by Decker using the mopolymerisation and maleate–vinylether copolymerisa-
steady-state equation tion were monitored at various temperatures. The results
A1 − A 2 indicate as expected that reaction rate (Rp ) increases with
Rp (%/s) = × 100 (4) temperature for acrylate homopolymerisation. Based on
A0 (t2 − t1 )
the Arrhenius equation an activation energy of 4 kJ/mol
where Rp is the rate of polymerisation in %/s, A1 and A2 was calculated. Similar work based on PhotoDSC ex-
the absorbances in the straight line region of the sigmoid periments has already demonstrated that the activation
 E. Staring et al. / Progress in Organic Coatings 45 (2002) 101–117 113

Fig. 12. Difference spectra of laurylacrylate with 1% Irgacure 184 photocured with a 200 W, Hg lamp.

Fig. 13. Photopolymerisation of laurylacrylate with 1% Irgacure 184 with contour mapping.
114 E. Staring et al. / Progress in Organic Coatings 45 (2002) 101–117

Fig. 14. Effect of nitrogen inerting with benzophenone and dimethylethanolamine as photoinitiators of ethoxylated TMPTA (Mw-607) [25].

energy for this homopolymerisation is low (2 kJ/mol) trates that it is possible to analyse a solid formulation in the
[32]. However, with the maleate–vinylether copolymerisa- melt phase.
tion starts levelling off at 70 ◦ C. This may be attributed Similar experiments have also been performed to look at
to the disruption of the ground state donor acceptor maleimides as photoinitiators in the polymerisation of acry-
complex. lates where the acrylate group at 810 cm−1 was monitored
An experimental maleate–vinylether powder coating for- simultaneously with the maleimide group at 698 cm−1 The
mulation was photocured at 100 ◦ C and both maleate and improved consumption of both maleimide and acylate when
vinylether peaks monitored as shown in Fig. 17. This illus- photocured under nitrogen was demonstrated [30].

Fig. 15. Effect of moisture on the cationic photoinitiated polymerisation of cycloaliphatic, epoxide (CAE), vinylether (VE) and glycidylether (GE) based
formulations.
 E. Staring et al. / Progress in Organic Coatings 45 (2002) 101–117 115

Fig. 16. A comparison of the effect of temperature on the photoinitiated homopolymerisation of laurylacrylate and the copolymerisation of triethyleneg-
lycoldivinylether and dioctyl fumarate (with 1% (w/w) Irgacure I84).

3.3. Real time dynamic mechanical analysis to the modification of a dynamic mechanical analyser fitted
with UV transparent quartz parallel plates to allow simul-
To complement the developments in time-resolved spec- taneous UV illumination. Initial results suggest that this is
troscopy, time-resolved mechanical property measurements readily feasible for low modulus systems in the MPa range.
for curing systems are necessary. This is mainly due to the This is attributed to torque limitations that make measure-
fact that chemical conversion does not translate readily to ments prior to vitrification possible. Special modifications
similar developments in mechanical properties. The compli- are necessary to prevent the parallel plate from grinding to-
cation arises due to a number of factors, one of the most gether due to shrinkage associated with cross-linking poly-
important of which is the development of the network. merisation. Work is ongoing to examine UV powder coat-
In recent years real time dynamic mechanical analysis ings the results of which will be available at the time of
has been developed [31]. Our efforts have also been devoted presentation.

Fig. 17. UV curing of powder coating based on experimental maleate–vinylether reactive groups at 110 ◦ C with 1% (w/w) Irgacure 184.
116 E. Staring et al. / Progress in Organic Coatings 45 (2002) 101–117

4. Conclusions to the characterization of the hyperbranched polyesteramides

and Victor Litvinov, Paul Steeman and Jos Linsen for their
We have shown in this paper how new developments in contributions to the NMR relaxation and real time studies,
characterisation techniques on the molecular, mesoscopic respectively. We thank the management of DSM Research
and temporal level can improve our scientific understanding and DSM Coating Resins for the permission to publish this
of such complex industrial products as coatings. work.
In Section 2, we have described an absolute method to
determine the molecular weight distribution of a new coat-
ing resins with hyperbranched architecture in a combination References
of SEC-DV and SEC-MS. It was established that fully
lauric ester functionalised hyperbranched polyesteramides, [1] T.A. Misev, R.A.T.M. van Benthem, T.J.G. Zwartkruis, Bull. Chem.
Tech. Maced. 17 (1998) 77–88;
synthesised in a straight-forward one-pot procedure, have a
R.A.T.M. van Benthem, D. Muscat, D.A.W. Stanssens, Proc. Am.
molecular weight distribution in accordance with calculated Chem. Soc. Div. PMSE 80 (1999) 72;
values. MALDI-TOF and ESI-MS confirmed the expected H.J.J. Rutten, W. Grisnich, R.A.T.M. van Benthem, in: Proceedings
molecular structures, but also indicated the formation of of the 16th Congress Federation Scandinavian Paint and Varnish
cyclic structures that were excluded in the calculations. De- Technologists SLE, May 2000;
R.A.T.M. van Benthem, Prog. Org. Coat. 40 (2000) 203;
viations from the predicted molecular mass moments were
D. Muscat, R.A.T.M. van Benthem, in: Vogtle (Ed.), Topics in Current
found with unfunctionalised and partly lauric acid ester Chemistry, Dendrimers III, Vol. 212, 2001, p. 41.
functionalised (DS < 0.60) polyesteramides, due to side [2] R.A.T.M. van Benthem, E. Geladé, N. Meijerink, P. Froehling,
reactions of 2-hydroxy-propylamide groups. The compar- P.H.M. Hendriks, D. Muscat, T.J.G. Zwartkruis, C.G. de Koster,
ison between SEC-DV and SEC-MALDI-TOF-MS results Macromolecules 34 (2001) 3559–3566;
E. Gelade, N. Meijerink, C.G. de Koster, R.A.T.M. van Benthem, K.
appeared to confirm the principle of universal calibration
Mortensen, R. Fokkens, N.N. Nibbering, Macromolecules 34 (2001)
for these polymers. 3552–3558.
In Section 3, we have presented two techniques for the [3] R.A.T.M. van Benthem, in: Proceedings of the 25th Athens
characterisation of (photocured) coatings. With the first, Conference on Organic Coatings, 1999, p. 345.
NMR T2 relaxation, we were able to gather relevant quanti- [4] P.J. Flory, Principles of Polymer Chemistry, Cornell University Press,
Ithaca, NY, 1953.
tative information on the network structure in a facile way
[5] E. Malmström, A. Hult, Compare with the use of AB2 monomers
with minimal sample preparation. This method showed in combination with B3 starter molecules in the synthesis of
good correlation with mechanical tests to determine molar hyperbranched polyesters, Macromolecules 29 (1996) 1222.
mass between cross-links. An increase in the content of [6] D.A.W. Stanssens, R. Hermanns, H. Wories, Prog. Org. Coat. 22
mono-functional monomer in a UV-curable formulation re- (1993) 379–391.
sulted in the expected significant decrease in the cross-link [7] D. Durand, C.-M. Bruneau, Br. Polym. J. (1981) 33–40;
D. Durand, C.-M. Bruneau, Polymer 23 (1982) 69–72.
density; the mono-functional monomer acting as a chain ex- [8] D. Muscat, H. Henderickx, G. Kwakkenbos, R.A.T.M. van Benthem,
tender. Large fraction of mono-functional monomer (above C.G. de Koster, R. Fokkens, N.M.M. Nibbering, J. Am. Soc. Mass
80 mass%) caused significant fraction of network defects. Spectrom. 11 (2000) 218–227.
Obtained values of Mc were significantly smaller compared [9] Z. Grubisic, R. Rempp, H. Benoit, J. Chim. Phys. 63 (1966) 1507.
to molar mass of di-acrylate. As anticipated, it was apparent [10] V. Litvinov, A.A. Dias, Macromolecules 34 (2001) 4051.
[11] Y.Y. Gotlib, M.I. Lifshits, V.A. Shehelev, I.A. Lishanskii, I.V.
that classical rubber elasticity theories are not applicable for
Balanina, Polym. Sci. USSR 18 (1976) 2630.
characterisation of this type of networks containing zip-like [12] L.R.G. Treloar, The Physics of Rubber Elasticity, 3rd Edition,
network junctions. Clarendon Press, Oxford, 1975.
For the second, RT-FTIR, we have outlined a versatile [13] T.A. Vilgis, G. Heinrich, Kautsch, Gummi Kunstst. 45 (1992)
experimental setup. This instrument is capable of measuring 1006.
[14] K.-H. Schimmel, G. Heinrich, Coll. Polym. Sci. 269 (1991) 1003.
20 spectra/s under various atmospheres and temperatures.
[15] T.A. Vilgis, G. Heinrich, Makromol. Theory Simul. 3 (1994) 271.
The analyses of both liquid and solid (powder) samples [16] T.G. Fox, S. Loshaek, J. Polym. Sci. 15 (1955) 371.
can be performed. The ability to control the atmosphere in [17] J.F. Rabek, in: J.P. Fouassier, J.F. Rabek (Eds.), Radiation Curing in
the transflection cell has been put to good use in demon- Polymer Science and Technology Fundamentals and Methods, Vol.
strating the varied behaviour of cycloaliphatic epoxides, 1, Elsevier, London, 1993, p. 329.
glycidylethers and vinylethers under dry and humid con- [18] R.S. Davidson, K.S. Tranter, S. Wilkinson, in: D.R. Randell (Ed.),
Radiation Curing of Polymers, Vol. II, RSC, Special Publication No.
ditions. Our future work will be directed to explain these 64, 1991, p. 400.
initial findings. [19] C. Decker, K. Moussa, Makromol. Chem. 189 (1988) 2381;
C. Decker, K. Moussa, Macromolecules 22 (1989) 4455.
[20] R. Holman, N. Arsu, E. Cockburn, R. Whiting, in: Proceedings of
Acknowledgements the Conference on Aspects of Photoinitiation, PRA, Egham, UK,
1993, p. 245.
[21] J. Finter, I. Frischinger, T. Haug, R. Morton, in: Proceedings of the
The authors wish to thank their colleagues Eric Geladé, Fourth Nurnberg Congress, PRA, Germany, 1997, p. 38.
Chris de Koster and Theo Zwartkruis for their contributions [22] C. Decker, D. Decker, Polymer 38 (9) (1997) 2229.
 E. Staring et al. / Progress in Organic Coatings 45 (2002) 101–117 117

[23] C.R. Morgan, F. Magnotta, A.D. Ketley, J. Polym. Sci. Polym. Chem. [28] J. Doughherty, F.J. Vara, P.K. Wolf, in: Proceedings of the Radtech,
156 (1977) 627. Vol. 1, Nashville, 1996, p. 80.
[24] S.P. Pappas, in: N.S. Allen (Ed.), Photopolymerisation and [29] S. Udding, F. Witte, S. de Jong, Product Finishing, April 1997,
Photoimaging Science and Technology, 1989, p. 55 (Chapter 2). pp. 26 and 30.
[25] J.V. Crivello, S.A. Bratlavsky, J. Polym. Sci. A 32 (1994) 2755; [30] A.A. Dias, H. Hartwig, J.F.G.A. Jansen, Surf. Coat. Int. 83 (2000)
J.V. Crivello, S.A. Bratlavsky, J. Polym. Sci. A 32 (1994) 2919. 382.
[26] M. Tsunuoka, Y. Shigeru, K. Ito, in: Proceedings of the Radtech, [31] A.A. Dias, J.F.G.A. Jansen, M. van Dijck, Surf. Coat. Int. JOCCA
1996, p. 393. 83 (2000) 502.
[27] C. Armstrong, S. Herlihy, in: Proceedings of the Conference on [32] B.S. Chiou, R.J. English, S.A. Khan, Macromolecules 29 (1996)
Aspects of Photoinitiation, PRA, Egham, UK, 1993, p. 1. 5368.