Journal of Colloid and Interface Science 406 (2013) 51–59

Contents lists available at SciVerse ScienceDirect

Journal of Colloid and Interface Science

www.elsevier.com/locate/jcis

Mechanism and kinetics of nanostructure evolution during early stages

of resorcinol–formaldehyde polymerisation
Katarzyna Z. Gaca ⇑, Jan Sefcik ⇑
Department of Chemical and Process Engineering, University of Strathclyde, 75 Montrose Street, G1 1XJ Glasgow, UK

a r t i c l e i n f o a b s t r a c t

Article history: Resorcinol and formaldehyde react in aqueous solutions to form nanoporous organic gels well suited for a
Received 17 February 2013 wide range of applications from supercapacitors and batteries to adsorbents and catalyst supports. In this
Accepted 2 May 2013 work, we investigated the mechanism and kinetics of formation of primary clusters in the early stages of
Available online 3 June 2013
formation of resorcinol–formaldehyde gels in the presence of dissolved sodium carbonate. Dynamic Light
Scattering measurements showed that size of freely diffusing primary clusters was independent of both
Keywords: reactant and carbonate concentrations at a given temperature, reaching the mean hydrodynamic radius
Sol–gel
of several nanometres before further changes were observed. However, more primary clusters formed at
Gelation
Resorcinol–formaldehyde
higher carbonate concentrations, and cluster numbers were steadily increasing over time. Our results
Organic gels indicate that the size of primary clusters appears to be thermodynamically controlled, where a solubil-
Dynamic Light Scattering ity/miscibility limit is reached due to formation of certain reaction intermediates resulting in approxi-
Primary clusters mately monodisperse primary clusters, most likely liquid-like, similar to formation of micelles or
Spontaneous emulsification spontaneous nanoemulsions. Primary clusters eventually form a particulate network through subsequent
Nanoemulsions aggregation and/or coalescence and further polymerisation, leading to nanoscale morphologies of result-
Thermodynamic control ing wet gels. Analogous formation mechanisms have been previously proposed for several polymerisation
and sol–gel systems, including monodisperse silica, organosilicates and zeolites.
Ó 2013 Elsevier Inc. All rights reserved.

1. Introduction
Organic gels based on formaldehyde and resorcinol were first
synthesised in the late 1980s [1] through an organic sol–gel pro-
cess, and since then, they have become increasingly important in
a wide range of applications [1–3]. Due to their electrical conduc-
tivity, very large surface areas and internal porosity, resorcinol–
formaldehyde aerogels have high potential as superior materials
for electrodes in electrochemical double-layer supercapacitors
and batteries [4,5], for capacitive deionisation units, as well as
materials for hydrogen storage [6]. Therefore, further development
and tailoring of these materials for specific applications is of great
importance for energy storage and low carbon technologies. Fur-
ther important applications of carbon aerogels include catalyst
supports and adsorbents [7].
Previous research has shown that resorcinol–formaldehyde gel
properties can be tailored to specific requirements, such as pore
surface area and pore size distribution, by changing the conditions
of their synthesis in the sol–gel process [2]. The sol–gel process has
⇑ Corresponding authors. Present address: Faculty of Chemical Engineering and
Technology, Cracow University of Technology, ul. Warszawska 24, 31-155 Cracow,
Poland (K.Z. Gaca).
E-mail addresses: katarzyna@gaca.cat (K.Z. Gaca), jan.sefcik@strath.ac.uk (J.
Sefcik).
been widely used in materials synthesis [8], and although it has
been traditionally associated with the synthesis of inorganic mate-
rials, a wide range of organically modified or purely organic mate-
rials have been produced in this way [2]. The organic sol–gel
process is a convenient and environmentally friendly synthesis
method, where dissolved monomers undergo series of reactions,
forming primary particles or clusters at nanometre to micrometre
scale, which then subsequently coalesce and/or aggregate forming
a system-spanning network immersed in the solvent: a wet gel.
The resulting organic gel is then dried, carbonised and chemically
activated. Additional treatments like solvent exchange and ageing
may also be introduced prior to drying, in order to achieve a gel
with properties appropriate for a given application. Although there
exists considerable empirical experience regarding the synthesis of
organic gels, fundamental understanding of their nanostructure
evolution is lacking, and therefore, our ability to rationally tailor
their structure for specific applications has been limited.
The widely accepted reaction mechanisms for the synthesis of
resorcinol–formaldehyde polymers are shown in Fig. 1. It can be
divided into two steps: substitution of resorcinol with formalde-
hyde to form (simply, doubly or triply) substituted hydroxymethyl
resorcinol and step-growth polymerisation of substituted resor-
cinol [1,3].
Although resorcinol–formaldehyde polymerisation can take
place in aqueous solutions at room temperature without any added

0021-9797/$ - see front matter Ó 2013 Elsevier Inc. All rights reserved.
http://dx.doi.org/10.1016/j.jcis.2013.05.062
52 K.Z. Gaca, J. Sefcik / Journal of Colloid and Interface Science 406 (2013) 51–59
Fig. 1. Overall reaction mechanism of resorcinol–formaldehyde polymerisation.
catalyst, the overall rate is relatively low [9], and it can be
expected, by analogy with a well-known phenol–formaldehyde
system, that the substitution reaction is both acid or base catalysed
[1,3,10]. The polymerisation step, where the hydroxymethyl group
reacts with another hydroxymethyl group, forming an ether bridge
–CH2–O–CH2–, or with an unsubstituted resorcinol site, forming a
methylene bridge –CH2–, can also be expected to be both acid and
base catalysed [10]. Polymerisations are typically performed at ele-
vated temperatures (up to 90 °C) in order to speed up the process.
Moreover, previous research showed that in order to obtain very
high pore surface areas of resulting gels, neutral or slightly basic
conditions within a relatively narrow range of pH values are re-
quired [10], and this is often achieved by adding weak bases such
as sodium carbonate as reaction catalysts to resorcinol–formalde-
hyde solutions. More recent research indicated that not all bases
are equally effective at same pH values in promoting high surface
areas of resulting gels and that both concentration and type (anion
and cation) of basic catalysts added to modify solution pH affect
the properties of the final gel [11], suggesting that the role of these
catalysts is beyond just providing basic conditions. Electrolytes,
such as sodium carbonate, are well known to destabilise colloidal
suspensions due to screening repulsive electrostatic interactions
between colloidal particles. Also, electrolytes are known to induce
liquid–liquid phase separation in water–alcohol mixtures leading
to colloidal emulsions.
We note that in aqueous solutions, formaldehyde is mainly
present in its hydrated form as methylene glycol and its oligo-
mers/polymers, and commercial formaldehyde solutions are nor-
mally stabilised with methanol in order to prevent formaldehyde
polymerisation, so that methylene glycols are accompanied by
methoxymethylene glycol and their small oligomers, while the
concentration of formaldehyde in its aldehyde form is in fact very
low (usually not exceeding 100 ppm) [12]. Therefore, the initial
reaction mixture in resorcinol–formaldehyde polymerisation is
an aqueous solution of resorcinol and methylene glycols in the
presence of strong electrolytes (such as sodium carbonate) added
as catalysts. As reactions proceed, a range of products starting with
variably substituted resorcinol and their condensation products
appear over time, with unknown solubility/miscibility properties
in respect to the background solution at relevant temperatures.
While it is expected that sufficiently long polymer chains will
eventually become insoluble and undergo microphase separation,
it may well be possible that solubility/miscibility limits are
reached earlier in the polymerisation process as relatively small
molecular weight intermediates and oligomers are produced.
There are two principal theories which can be found in the lit-
erature aiming to explain the mechanism of the gel formation in
polymerising resorcinol–formaldehyde systems: microphase sepa-
ration [13,14] or aggregation of primary colloidal particles [15].
The microphase separation scenario is based on the idea of poly-
meric chains growing too large and becoming insoluble and there-
fore triggering demixing, resulting in two interpenetrating phases,
one polymer rich and the other solvent rich, a process well known
from other polymer systems. The colloidal aggregation scenario is
based on the idea of formation of primary colloidal particles which
then subsequently aggregate and form a space-filling network of
interconnected clusters. Recent analysis of Small Angle X-ray Scat-
tering (SAXS) measurements from resorcinol–formaldehyde gels
suggests that these two scenarios are unlikely to be distinguishable
from the resulting gel structures and in any case may well be two
extremes of a more general process [16]. From published transmis-
sion electron micrographs [13], it is apparent that the dry network
of resorcinol–formaldehyde gels produced in the presence of car-
bonates consists of interconnected dense primary particles of a
few to a few tens on nanometres, with smaller primary particles
for higher carbonate concentrations. However, it is not clear what
is a mechanism of formation of these primary particles and how is
it controlled by solution pH and the nature of cations and anions
used. Formation of primary clusters was previously investigated
by SAXS [17], where clusters with radius of gyration of 2–4 nm
were observed before the onset of gelation at 25 °C. In a recent
study, SAXS and X-ray photon correlation spectroscopy was used
to monitor gel formation at 70 °C [18], where it was observed that
primary clusters were gradually growing with their radius of gyra-
tion from about 2 to about 10 nm, after which development of
increasingly stiff network dynamics was observed. Sol–gel transi-
tion process in resorcinol–formaldehyde solutions at 25 °C was
investigated by Dynamic Light Scattering, where monomodal de-
cay time distributions were observed initially with apparent clus-
ter hydrodynamic diameters growing gradually up to about
5 nm. The initial formation of these primary clusters was followed
by the development of much slower autocorrelation function de-
cay modes, which were presented in term of steeply increasing
apparent hydrodynamic diameters of growing colloidal particles
[15].
The nanoscale structure of resorcinol–formaldehyde gels as ob-
served by electron microscopy appears to be morphologically sim-
ilar to the one of silica-based gels from silicon alkoxides through
hydrolysis and subsequent step-growth polymerisation in near
neutral or slightly basic aqueous solutions, where the sol–gel pro-
cess is similar in some respects to synthesis of resorcinol–formal-
dehyde gels. It has been previously proposed that formation of
primary particles in silica sol–gel systems may include formation
of transient nanoemulsions through molecular demixing driven
 K.Z. Gaca, J. Sefcik / Journal of Colloid and Interface Science 406 (2013) 51–59
by production of small molecular weight reaction intermediates
before any substantial polymerisation commences [19]. Similar
mechanisms based on formation of nanoscale liquid-like interme-
diates have been proposed for several polymerisation and sol–gel
systems, including monodisperse Stoeber silica particles [20],
zeolites [21] and organosilicates [22,23]. We note that while small
silica/organo-silica oligomers appear to be poorly miscible/soluble
in water–alcohol mixtures at basic and neutral pH conditions,
resulting in formation of colloidal primary particles, there appear
to be no solubility issues in acidic solutions where transparent gels
are formed without any apparent nanoscale phase separation
phenomena.
Interestingly, it appears that the mechanism of resorcinol–
formaldehyde gel formation under basic conditions is also different
from that in acidic conditions, where polymerisation-induced
critical opalescence, as inferred from SAXS measurements [24],
has been proposed to be linked to polymerisation-induced phase
separation, leading to formation of micron-sized particles
subsequently aggregating and forming a gel. Nevertheless, the
mechanism of formation of primary clusters in resorcinol–formal-
dehyde polymerisation under basic and neutral conditions and the
role(s) played by basic catalysts, such as sodium carbonate, are still
poorly understood.
The objective of this work was to investigate mechanism and
kinetics of formation of primary clusters during early stages of res-
orcinol–formaldehyde polymerisation in the presence of sodium
carbonate by analysing effects of carbonate and reactant concen-
trations and temperature, using Dynamic Light Scattering to mon-
itor formation of primary clusters and subsequent structure
development towards gel formation.
2. Experimental
Resorcinol (99 wt.%), formaldehyde (37 wt.%, aqueous solution
stabilised by 13 wt.% methanol; the concentration of methanol
was determined experimentally using the NMR spectroscopy
[25]) and sodium carbonate (P99.5 wt.%, anhydrous ACS reagent)
were all purchased from Sigma–Aldrich. Syringe filters (0.45 lm
pore size, PTFE membrane and PP housing; 0.02 lm pore size, Ano-
top 10; manufactured by Whatman) were purchased from Fisher
Scientific. Filtration was carried out with rubber-free PVC/PP syrin-
ges. Vials for DLS experiments were made from borosilicate glass,
with diameter of 10 mm and height 75 mm, and were also pur-
chased from Fisher Scientific. Filtration was performed in order
to eliminate impurities, including dust, which could strongly affect
DLS results. The reacting mixtures were filtered after the initial
mixing, but before thermal treatment, where only unsubstituted
and substituted resorcinol is present, but no larger oligomers are
formed under these conditions [25].
The compositions of resorcinol–formaldehyde solutions used
for sol–gel processes are customarily defined by ratios of amounts
of reagents, carbonate and water used. The most frequently used
ratios are R/C (resorcinol to catalyst; mol/mol), R/F (resorcinol to
formaldehyde; mol/mol) and R/W (resorcinol to water; g/ml; W re-
fers to water added as a solvent, not the water present in formal-
dehyde solution). In all experiments, resorcinol to formaldehyde
ratio was kept constant at 0.5 mol/mol, while resorcinol to water
and overall carbonate concentrations were varied. The composi-
tions of mixtures used in this work are shown in Table 1. Please
note that F in the R/F ratio refers to mols of formaldehyde, while
F solution refers to the volume of the added formaldehyde solution.
All resorcinol–formaldehyde solutions were prepared as fol-
lows. The required amounts of deionised water (W) and resorcinol
(R) were transferred to a beaker and stirred using a magnetic stir-
rer for 5 min until resorcinol fully dissolved. Sodium carbonate (C)
53
was then added to the beaker with dissolved resorcinol, and after
10 min of further stirring to dissolve sodium carbonate, the
required amount of formaldehyde solution (F solution) was added
to the mixture and stirring continued for another 30 min. The solu-
tions were then filtered in two steps, first using 0.45 lm and then
0.02 lm pore size syringe filter, directly into clean vials pre-
washed with deionised water and filtered resorcinol–formalde-
hyde solution. Vials were then sealed with a cap and parafilm,
placed in the DLS measurement cell where they were thermostated
at 55 °C and measurements were taken at certain time intervals.
Since DLS measurements were not possible to do in situ at 80 °C
with the instrument available, they were conducted in an alterna-
tive manner. Solution was divided upon filtration into several par-
allel samples which were placed in an electric oven preheated to
80 °C. The vials were then taken out from the oven at certain time
intervals, rapidly cooled down to ambient temperature in an ice
bath to quench the reaction and taken for DLS measurements per-
formed at 25 °C.
In Dynamic Light Scattering experiments, autocorrelation func-
tions g1(s) were measured using a digital correlator (ALV/LSE-
5004). The wavelength of incident laser light was k = 632.8 nm,
and the scattering angle h was 90°. When the autocorrelation decay
was showing an exponential decay, indicating unhindered Brown-
ian diffusion, we used the cumulant method [26] to estimate in the
initial decay rate C [s 1]. From this, one can determine the mean
diffusion coefficient D, using C = Dq2, where q = (4p/k)sin(h/2),
and the mean hydrodynamic radius Rh, using the Stokes–Einstein
equation, D = kBT/6plRh, where kB is the Boltzmann constant, T is
the absolute temperature, and l is dynamic viscosity (taken here
to be equal to that of pure water at a given temperature).
For estimation of gelation times, solutions were transferred into
25 ml PP flasks (diameter 25 mm, height 100 mm), sealed and
placed in an electric oven preheated to a desired temperature.
The process of gelation was monitored by visual observations of
changes in the solution flowability, and gelation time was taken
here as time measured from the moment a sample of reacting
mixture is placed in an electric oven set at the desired temperature
and periodically checked until a lack of flow at tilting by 45° is ob-
served. Although this kind of gel time estimation is subject to sev-
eral factors, including the diameter of the vessel and the total
volume of the solution, it is a suitable tool for the assessment of
differences in gel formation kinetics among similar samples sub-
ject to different reaction conditions. It has been reported that there
has been a reasonable correlation between visual and rheological
determination of gel times in similar resorcinol–formaldehyde sys-
tems, where visually estimated gel times were 5–45% longer than
those determined rheologically [27]. With this caveat, we can use
visual estimation of gel times for approximate determination of
time windows for DLS measurement before networking of primary
clusters and onset of gelation.
3. Results and discussion
Initial gelation time observations were performed to estimate
time scales appropriate for DLS measurements in order to monitor
formation of primary clusters well before the gel formation itself,
since previous scattering studies typically focused on later stages
of the process where extensive aggregation and/or cross-linking
of primary clusters resulted in complicated interplay of cluster
size, structure and dynamics. Fig. 2 shows gelation times deter-
mined for a range of carbonate concentrations at two different
temperatures. Here, ratios of resorcinol to formaldehyde (R/F)
and resorcinol to water (R/W) were kept constant at 0.5 mol/mol
and 0.10 g/ml, respectively, while concentration of carbonate,
expressed as the ratio of resorcinol to catalyst (R/C), was varied
54 K.Z. Gaca, J. Sefcik / Journal of Colloid and Interface Science 406 (2013) 51–59

Table 1
Compositions of investigated reaction mixtures.

R/W R/C (mol/mol) R/F (mol/mol) W (ml) R (g) C (g) F solution (ml) Initial pH
(g/ml) (mol/mol)
0.10 0.0164 50 0.5 5.00 0.50 0.0096 0.74 7.51
0.10 0.0164 100 0.5 5.00 0.50 0.0048 0.74 7.08
0.10 0.0164 200 0.5 5.00 0.50 0.0024 0.74 6.63
0.10 0.0164 300 0.5 10.00 1.00 0.0032 0.74 6.13
0.10 0.0164 600 0.5 10.00 1.00 0.0016 0.74 6.11
0.20 0.0328 178 0.5 5.00 1.00 0.0054 1.47 –
0.30 0.0492 241 0.5 5.00 1.50 0.0060 2.21 –
0.40 0.0656 292 0.5 5.00 2.00 0.0066 2.95 –
0.50 0.0820 334 0.5 5.00 2.50 0.0072 3.69 –

between 50 and 600 mol/mol (see Table 1). As expected, increasing
carbonate concentration (i.e. decreasing R/C ratio) leads to decreas-
ing gel time, with an approximately linear dependence of the ob-
served gel time on the R/C value (i.e. inversely proportional to
the carbonate concentration). We also note that the concentration
of carbonate (base) causes a systematic variation of solution pH,
varying from 7.5 for R/C = 50 mol/mol to 6.1 for R/C = 300 mol/
mol (these are initial solution pH values measured at ambient tem-
perature before solutions were heated), related to neutralisation of
slightly acidic reactants (pH of the formaldehyde stock solution
was 4.3 due to small amount of formic acid present, and resor-
cinol–water solution at R/W = 0.1 g/ml had pH of 4.8), see Table 1.
Increasing concentration of resorcinol, corresponding to pro-
portionally increasing concentration of formaldehyde concentra-
tion (as the R/F ratio was kept constant at 0.5 mol/mol), also
caused the gelation time to become shorter, as expected, with an
approximately linear dependence of the observed gelation time
on the inverse of resorcinol concentration. In terms of temperature
effect, increasing the temperature by 25 °C (from 55 °C to 80 °C)
caused the gelation time to decrease by a factor of about 5, what
is consistent with expectations and the previous literature [27].
A series of in situ DLS experiments were performed at 55 °C
where the overall gelation process is slow enough to investigate
formation of primary clusters in more detail. The raw data
resulting from these experiments were autocorrelation functions,
which were then fitted to obtain the initial decay rates and from
that the apparent mean hydrodynamic radii of primary clusters
formed in the solutions. A typical example of how measured auto-
correlation functions change over time is shown in Fig. 3. Although
autocorrelation functions were measured every 3–5 min, only
three representative ones are shown here for clarity.
It can be seen in Fig. 3 that the autocorrelation function
collected after 10 min of heating has a shape very close to that
for an ideal single exponential decay, indicating that the primary
clusters observed are initially approximately monodisperse and
subject to unhindered Brownian motion. As the time proceeds,
the shape of the autocorrelation function changes in two respects,
as seen in the shape of the autocorrelation function collected after
75 min of heating: the initial decay rate is lower and there is a sec-
ondary decay significantly deviating from an ideal monodisperse
exponential decay. This is most likely due to aggregation of pri-
mary particles forming larger clusters; however, most of them still
remain free to move around and appear to be subject to Brownian
motion, since the measured autocorrelation function fully decays
within about 1 ms. Finally, the shape of the autocorrelation func-
tion measured after 103 min of heating is consistent with one that
is expected for a gel, consistently with the macroscopic gel time
estimated by tilting a vial containing the gelling solution (see
Fig. 2). At this point, the shape of the autocorrelation function de-
cay shows clear power law features, typical for cases where mobil-
ity of clusters is severely restricted as would be expected for a gel
network consisting of a three-dimensional structure of intercon-
nected clusters, while unconnected clusters can diffuse within
the pores of the gel structure, as indicated by the initial decay still
visible.
A standard analysis of autocorrelation functions can be used to
estimate the mean hydrodynamic radius of clusters in the solution,
assuming that the clusters are freely diffusing, and they are not yet

Fig. 2. Gelation time as a function of composition for solutions kept at 55 °C (squares) and 80 °C (circles). Lines are guides to eye only. (a) R/W and R/F ratios equal to 0.10 g/
ml and 0.5 mol/mol, respectively, (b) R/C and R/F ratios equal to 100 mol/mol and 0.5 mol/mol, respectively, and sodium carbonate concentration equal to 7.9 mmol/dm3.
 K.Z. Gaca, J. Sefcik / Journal of Colloid and Interface Science 406 (2013) 51–59 55

Fig. 3. DLS autocorrelation functions (log-linear and log–log plots shown) for solutions with R/C, R/W and R/F ratios equal to 50 mol/mol, 0.10 g/ml and 0.5 mol/mol,
respectively, kept at 55 °C for times as indicated. Dashed lines show theoretical decays calculated assuming a monodisperse cluster population with the hydrodynamic radius
equal to the mean value determined from the initial decay of the autocorrelation function.

interconnected, as evidenced by the autocorrelation function
following an exponential decay rather than stretched exponential
or power law decays characteristic of gel networks. We have there-
fore restricted our attention to the early times where monomodal
primary cluster populations are present, so that true mean hydro-
dynamic radii of primary clusters can be determined. The initial
decays of measured autocorrelation functions were fitted as de-
scribed in the experimental section, and the resulting initial decay
rates (C) and the corresponding mean hydrodynamic radii are
shown in Fig. 4.
Fig. 4b shows clearly that the (true) mean hydrodynamic radius
of primary clusters during the initial stage of their formation (up to
about 50 min of heating at 55 °C) is essentially independent of the
R/C ratio: it grows from about 1.5 nm at 10 min to about 3.5 nm at
50 min, following a power law dependence of size on time. We
note that the hydrodynamic radius of a hydrated resorcinol mole-
cule is about 0.4 nm, which we were able to directly measure in an
aqueous solution of resorcinol using our DLS instrument, while
variously substituted hydroxymethyl derivatives of resorcinol are
expected to be somewhat larger, so it is clear that primary clusters
found after the first 10 min are several times larger than single res-
orcinol–formaldehyde molecules.
Further change of the initial decay is very slow, but the second
decay starts developing at longer times, indicating subsequent
aggregation of primary clusters. During this stage, the apparent
mean hydrodynamic radius grows only slightly. Eventually, the
secondary decay increases in relative magnitude and assumes a
power law scaling indicative of a gel network (see Figs. 3 and 4).
Although one could still technically fit the initial decay rate and
determine the corresponding apparent mean hydrodynamic radius
during the next stage of the gel formation process, it would not be
a physically meaningful quantity describing a real size of growing
clusters.
These observations are consistent with previously published re-
sults [15] for experiments performed at 25 °C, although those
authors presented the apparent hydrodynamic diameter deter-
mined by DLS for times extending well into the stage of cluster
aggregation and network formation, where their reported decay

Fig. 4. Time evolution of the initial decay rate (a) and the mean hydrodynamic radius (b) measured at 55 °C. Ratios R/F and R/W were 0.5 mol/mol and 0.10 g/ml, respectively
and R/C as shown in legends.
56 K.Z. Gaca, J. Sefcik / Journal of Colloid and Interface Science 406 (2013) 51–59
Fig. 5. DLS autocorrelation functions (log-linear and log–log plots shown) for solutions with R/W and R/F ratios equal to 0.50 g/ml and 0.5 mol/mol, respectively, and sodium
carbonate concentration 7.9 mmol/dm, kept at 55 °C for times as indicated.
time spectrum became bimodal and very long decay times were
observed. However, in the early stage where monomodal decay
spectra were observed (see their Fig. 1), initial sizes of primary
clusters appear to be quite similar for various values of the R/C ra-
tio, although there are only few data points reported for hydrody-
namic diameters below 5 nm (see their Fig. 2). Similarly, the
growth of primary clusters was investigated at 25 °C by SAXS
[17], and from the data reported in that work, it can be seen that
the radius of gyration of primary clusters appears to be around
2 nm (see their Fig. 4) for all conditions investigated at the same
time of 500 min, which is well before any increase in viscosity indi-
cating approaching gelation. The same authors also showed scaling
exponents for their SAXS intensity data increasing over time from
their initial values around 2 to plateau at values around 4 at longer
times. The power law scaling exponent of 4 is indicative of smooth
surfaces (Porod regime), while value below 3 is considered to be
indicative of mass fractal structures. However, in the reported data
(see their Fig. 3), there is an extremely narrow range of q values
over which a power law scaling is fitted, and it can be argued that
such data cannot be used reliably to extract structural information
from scattered intensities. In fact, it has been shown previously
that for small fractal-like aggregates, the apparent power law scal-
ing determined from scattering structure factors can be signifi-
cantly lower than the actual mass fractal scaling exponent and
only approached the actual value when aggregates are sufficiently
large [28].
A series of experiments with varied R/W ratios were also per-
formed at 55 °C in order to determine the influence of overall reac-
tants concentration on the evolution of autocorrelation function
and the apparent hydrodynamic radius. Ratio of resorcinol to form-
aldehyde was kept at 0.5 mol/mol, and instead of keeping R/C ratio
constant, the concentration of sodium carbonate was kept constant
at and equal to 7.9 mmol/dm, while the ratios of resorcinol to
water were changing from 0.10 g/ml to 0.50 g/ml. Fig. 5 shows typ-
ical autocorrelation functions for a sample with R/W 0.50 g/ml.
Interestingly, the measured autocorrelation functions are very sim-
ilar to those shown in Fig. 3, and the decay times are comparable,
despite the fivefold increase in the resorcinol concentration (cf.
curves for 10 min in Figs. 3 and 5).
Mean hydrodynamic radii were calculated from the autocorre-
lation functions for samples with varied R/W ratios for a range of
times where decays were still following a single exponential pat-
tern, and so, there was no hindrance on free diffusion of primary
clusters (in the same manner as it was done for samples with var-
ied R/C ratios). The results of this analysis are shown in Fig. 6. One
can clearly see that the mean hydrodynamic radius is again very
similar for all values of R/W ratios investigated here, indicating
that the overall concentration of reactants does not control the
sizes of primary clusters. This is also consistent with previous
observations of Yamamoto et al. [15] at 25 °C, where reported
hydrodynamic diameters of primary clusters were similar for a
range of R/W ratios at early times (see their Fig. 2), while they be-
came very different at later times (where these can only be seen as
apparent hydrodynamic diameters which do not reflect real cluster
sizes anymore, as discussed above).
From the data reported in Figs. 4 and 6 above, we can see that
the size of primary clusters does not appear to vary significantly
with the concentration of carbonate or reactants. This might seem
surprising, because carbonate is a catalyst for the substitution of
resorcinol with formaldehyde, and therefore, it is expected that
there is more resorcinol–formaldehyde present and hence faster
polymerisation at higher carbonate concentration. However, there
is a clear effect of the carbonate concentration as well as the reac-
tants concentration on the measured values of scattered intensities
as shown in Fig. 7.
In Fig. 7a, we can see that scattered intensities (expressed here
in terms of mean count rates measured at the scattering angle of
90°) stay approximately constant during the first 5–10 min and
then start increasing steadily over time. However, unlike mean
hydrodynamic radii, there is a clear dependence of scattered inten-
sities on the R/C ratio, with higher intensities for smaller R/C ratios
(i.e. higher carbonate concentrations). Since hydrodynamic radii
were very similar at all three R/C ratios (see Fig. 4), while the
scattered intensity is higher for higher carbonate concentration
at a given time, it follows that the scatterers (i.e. primary clusters)
are either present at higher concentrations or they possess higher
optical contrast (difference in refractive index respective to sur-
rounding solvent) at higher carbonate concentrations. Since the
surrounding solvent is very similar in all cases (same reactant con-
centrations in water, with carbonate being present at comparably
low concentrations), it is most likely that in the case of various
R/C ratios, the differences in scattered intensity are due to more
 K.Z. Gaca, J. Sefcik / Journal of Colloid and Interface Science 406 (2013) 51–59 57

Fig. 6. Time evolution of the initial decay rate (a) and the mean hydrodynamic radius (b) at 55 °C. Ratio R/F was 0.5 mol/mol, sodium carbonate concentration was 7.9 mmol/
dm3, and R/W was as shown in the legend.

primary clusters formed at higher carbonate concentrations. This is
reasonable, since higher carbonate concentration should imply fas-
ter substitution reaction and so faster production of substituted
resorcinol and presumably also faster subsequent condensation
reaction.
In Fig. 7b, we can see that scattered intensities are lower for
higher concentrations of resorcinol (and formaldehyde, since the
R/F ratio was kept constant equal to 0.5). By the same argument
as above, the primary clusters are either present at lower concen-
trations or they have lower optical contrast. Since it would be ex-
pected that the primary clusters would become more (not less)
numerous at higher concentration of reactants (while the carbon-
ate concentration was held constant at all R/W ratios), it follows
that they ought to have decreasing optical contrast with increasing
reactant concentration. This is sensible, since composition of
solvent surrounding primary clusters becomes more similar to
clusters as concentration of reactants increases and water concen-
tration decreases in the reaction mixture with increasing R/W ra-
tios. Refractive indices of reaction mixtures used here were
estimated to vary between 1.35 for R/W = 0.1 g/ml and 1.40 for
R/W = 0.5 g/ml [29], while the typical literature values for dense li-
quid phase systems chemically similar to the clusters of reaction
intermediates have been reported between 1.44 (saturated aque-
ous solution of phenol; [30]) and 1.48 (phenol–formaldehyde resin
at 50% w/w solids; [31]). Therefore, it can be expected that primary
clusters composed of resorcinol–formaldehyde reaction intermedi-
ates have higher optical density than surrounding bulk solution,
and thus indeed, their optical contrast will be lower in solutions
with higher reactant concentrations, i.e. at higher R/W ratios,
which have higher refractive index.
A series of DLS experiments were also performed on solutions
reacting at 80 °C, while quenching them at specified times to
20 °C where reactions are expected to be much slower and DLS mea-
surements could be performed. An example of results from these
experiments is shown in Fig. 8. Comparison of Figs. 3 and 8 shows
that the evolution of the autocorrelation function indicates the
same overall pattern at both temperatures – from an initially mono-
disperse population of primary clusters, through a polydisperse

Fig. 7. Time evolution of the scattered intensity measured at 55 °C. (a) Ratios R/F and R/W were 0.5 mol/mol and 0.10 g/ml, respectively, and R/C was as shown in the legend.
(b) Ratio R/F was 0.5 mol/mol, sodium carbonate concentration was 7.9 mmol/dm3, and R/W was as shown in the legend.
58 K.Z. Gaca, J. Sefcik / Journal of Colloid and Interface Science 406 (2013) 51–59

Fig. 8. DLS autocorrelation functions (log-linear and log–log plots shown) for solutions with R/C, R/W and R/F ratios equal to 200 mol/mol, 0.10 g/ml and 0.5 mol/mol,
respectively, kept at 80 °C for times as indicated and then rapidly quenched and measured at 20 °C.

collection of aggregated clusters to a gel network of interconnected
clusters.
However, there is a significant difference in time when the
autocorrelation function falls into one of these three stages. For
example, at the composition corresponding to results shown in
Fig. 3, the aggregation stage was observed only after more than
50 min at 55 °C, while it appeared in just 10 min at 80 °C. Similarly,
the solution gelled about five times later at 55 °C than at 80 °C (see
Fig. 3). These results indicate that the temperature influences the
overall rate of the gel formation, but it does not appear to influence
an overall mechanism of its formation, at least within the range of
temperatures investigated.
In order to assess the effect of temperature on the mean hydro-
dynamic radius of primary clusters, in Fig. 9, we show a compari-
son of mean hydrodynamic radii of primary clusters obtained by
heating at 55 °C and 80 °C. It can be that the primary clusters ob-
tained at 80 °C are larger than those at 55 °C, while the time evolu-
tion is following the same overall trend irrespective of the R/C
ratio.
We have seen that under the range of conditions investigated
here, the size of primary clusters varies only little when changing
concentrations of reactants or carbonate at a given temperature.
Furthermore, primary clusters are more numerous, and overall
gelation process is faster at higher carbonate concentrations. If pri-
mary clusters are more numerous, then it is sensible that their
aggregation is faster, and space-filling by growing aggregates leads
to gel formation at shorter times.
It may seem intriguing that the primary cluster size is only little
sensitive to the reaction mixture composition. This appears to be
inconsistent with the scenario where growth of primary clusters
is driven by homogeneous polymerisation of substituted resor-
cinol, as proposed previously [17], since faster reactions should
lead to faster growth of cluster size, but this is not the case at early
stage of the process before the onset of primary cluster aggrega-
tion. However, our results are consistent with a process of nano-
scale demixing, controlled by solubility/miscibility equilibria,
similar to that involved in micellisation or spontaneous emulsifica-
tion, such as in Ouzo effect. Unlike the classical Ouzo effect in
non-reacting mixtures, though, resorcinol–formaldehyde species
in primary clusters undergo further reactions, and hence, chemical
and physical nature of clusters evolves over time.
We propose that certain intermediates produced in resorcinol–
formaldehyde polymerisation (either multiply substituted
resorcinol or its oligomers) reach a limit of their solubility/misci-
bility in the solution mixture at a given temperature, and the
solution phase undergoes a process similar to spontaneous nano-
emulsification or micelle formation. This leads to primary clusters
observed in this work as well as by others using in situ DLS or SAXS
measurements (see above). Such a process is controlled by solubil-
ity/miscibility equilibria with demixing resulting in formation of
discrete non-interconnected primary clusters freely diffusing in
surrounding solution matrix, in contrast to two interpenetrating
phases resulting from a classical polymerisation driven microphase
separation.
As indicated in the overall reaction mechanism shown in Fig. 1,
there are a series of reactions starting with the first resorcinol sub-
stitution with formaldehyde, followed by the second and possibly
further ones. Then, the substituted resorcinol species undergo sub-
sequent condensation reactions to form dimers and further oligo-
mers. It is likely that a certain substituted and/or condensed
species has limited solubility/miscibility in the background reac-
tion mixture, and as it accumulates over time, it reaches a critical
concentration at which primary clusters are formed. As can be seen
in Fig. 7a, there is an earlier onset of an increase in scattered inten-
sity at higher carbonate concentrations indicating that such a crit-
ical concentration of reaction intermediates is reached earlier, as
Fig. 9. Early evolution of mean hydrodynamic radius for solutions kept at 55 °C
(squares) and 80 °C (circles). (R/C ratios 50, 100 and 200 mol/mol 1; R/F is 0.5 mol/
mol 1; R/W is 0.10 g/ml 1).
 K.Z. Gaca, J. Sefcik / Journal of Colloid and Interface Science 406 (2013) 51–59
expected for a series of reactions catalysed by bases, such as car-
bonates. At higher carbonate or reactant concentrations, reaction
rates are higher, more intermediates are produced, and therefore,
more primary clusters (of approximately the same size) are formed
within the same solution matrix, eventually resulting in faster
aggregation and gelation. Since intermediate species in primary
clusters undergo further reactions, chemical and physical nature
of clusters continually evolves, which may help to explain why
their size is changing over time even though their formation is
thermodynamically controlled. Primary clusters may be initially li-
quid-like and subject to subsequent coalescence as well as aggre-
gation. Further polymerisation may proceed faster within the
clusters than in the bulk solution due to locally higher concentra-
tions of intermediates, eventually leading to solidification and
colloidal destabilisation or polymerisation driven microphase
separation.
The final size of particles observed to form the gel network then
results from a delicate interplay of polymerisation, phase equilib-
ria, coalescence and aggregation phenomena. Furthermore, car-
bonates might play role in colloidal stability of primary clusters
or their aggregates, resulting in faster aggregation and subsequent
gelation at higher carbonate concentrations, which could also con-
tribute to the observed effect of carbonates on gel times.
4. Conclusions
Formation of primary clusters and their subsequent aggregation
and gelation during resorcinol formaldehyde polymerisation was
monitored by DLS at two different temperatures, 55 °C and 80 °C.
We found that the kinetics of growth of primary clusters, before
they were subject to further aggregation, changed only very little
when the carbonate concentration was varied. Although more pri-
mary clusters were formed at higher carbonate concentrations, as
deduced from scattered light intensities, and corresponding gela-
tion times were shorter, the size of primary clusters was very sim-
ilar at all carbonate concentrations. We also found that the growth
kinetics of primary clusters was only little dependent as the con-
centration of reactants (resorcinol and formaldehyde) was varied.
Resulting gelation times were again shorter at higher reactant con-
centrations, indicating that more primary clusters were produced,
although this could not be unequivocally established from scatter-
ing intensities.
Our results indicate that size of primary clusters appears to be
thermodynamically controlled, where a miscibility limit is reached
due to formation of certain reaction intermediates, most likely
with relatively small molecular weight, resulting in nanoscale
molecular demixing leading to formation of approximately mono-
disperse primary clusters, similar to formation of micelles or spon-
taneous nanoemulsions. At higher carbonate concentration or
reactant concentrations, more primary clusters were produced
due to faster reactions. Primary clusters may be initially liquid-like,
and further polymerisation may proceed faster within them than
in the bulk solution due to locally higher concentrations of
59
intermediates. This leads to a competition between polymerisa-
tion, phase separation, coalescence and aggregation resulting in
nanoscale morphologies of wet gels. Analogous formation mecha-
nisms have been previously proposed for several polymerisation
and sol–gel systems, including monodisperse silica, organosilicates
and zeolites.
There are likely to be multiple roles played by carbonates in the
process of resorcinol–formaldehyde gelation. In addition to car-
bonates being catalysts for reactions of resorcinol and formalde-
hyde, resulting in faster reactions and higher numbers of primary
clusters formed, they may also influence solubility/miscibility
equilibria in complex multicomponent reaction mixtures responsi-
ble for formation of primary clusters. Furthermore, carbonates may
also influence colloidal stability of primary clusters resulting in
faster aggregation and subsequent gelation at higher carbonate
concentrations.
References
[1] R.W. Pekala, J. Mater. Sci. 24 (1989) 3221.
[2] A.C. Pierre, G.M. Pajonk, Chem. Rev. 102 (2002) 4243.
[3] S.A. Al-Muhtaseb, J.A. Ritter, Adv. Mater. 15 (2003) 101.
[4] M. Mirzaeian, P.J. Hall, Electrochim. Acta 54 (2009) 7444.
[5] M. Mirzaeian, P.J. Hall, J. Mater. Sci. 44 (2009) 2705.
[6] H. Kabbour, T.F. Baumann, J.H. Satcher, A. Saulnier, C.C. Ahn, Chem. Mater. 18
(2006) 6085.
[7] N. Tonanon, Y. Wareenin, A. Siyasukh, W. Tanthapanichakoon, H. Nishihara,
S.R. Mukai, H. Tamon, J. Non-Cryst. Solids 352 (2006) 5683.
[8] A.C. Pierre, in: Kirk-Othmer Encyclopedia of Chemical Technology, John Wiley
& Sons Inc., New York, 2006.
[9] R.B. Durairaj, Resorcinol: Chemistry, Technology and Applications, Springer,
Berlin, 2005.
[10] C. Lin, J.A. Ritter, Carbon 35 (1997) 1271.
[11] N. Job, C.J. Gommes, R. Pirard, J.-P. Pirard, J. Non-Cryst. Solids 354 (2008) 4698.
[12] H.R. Gerberich, G.C. Seaman, Formaldehyde, in: Kirk-Othmer Encyclopedia of
Chemical Technology, John Wiley & Sons, Inc., 2008.
[13] R.W. Pekala, D.W. Schaefer, Macromolecules 26 (1993) 5487.
[14] D.W. Schaefer, G. Beaucage, R.W. Pekala, J. Non-Cryst. Solids 186 (1995) 159.
[15] T. Yamamoto, T. Yoshida, T. Suzuki, S.R. Mukai, H. Tamon, J. Colloid Interface
Sci. 245 (2002) 391.
[16] C.J. Gommes, A.P. Roberts, Phys. Rev. E 77 (2008).
[17] H. Tamon, H. Ishizaka, J. Colloid Interface Sci. 206 (1998) 577.
[18] O. Czakkel, A. Madsen, Europhys. Lett. 95 (2011).
[19] J. Sefcik, A.V. McCormick, Catal. Today 35 (1997) 205.
[20] D.L. Green, S. Jayasundara, Y.-F. Lam, M.T. Harris, J. Non-Cryst. Solids 315
(2003) 166.
[21] D.D. Kragten, J.M. Fedeyko, K.R. Sawant, J.D. Rimer, D.G. Vlachos, R.F. Lobo, M.
Tsapatsis, J. Phys. Chem. B 107 (2003) 10006.
[22] S.E. Rankin, J. Sefcik, A.V. McCormick, J. Phys. Chem. A 103 (1999) 4233.
[23] J. Ambati, S.E. Rankin, J. Colloid Interface Sci. 362 (2011) 345.
[24] C.J. Gommes, N. Job, J.-P. Pirard, S. Blacher, B. Goderis, J. Appl. Crystallogr. 41
(2008) 663.
[25] K.Z. Gaca, in: Kinetics and Mechanisms of Early Stages of Resorcinol–
Formaldehyde Polymerization; PhD thesis, University of Strathclyde,
Glasgow, United Kingdom, 2012.
[26] R. Finsy, Adv. Colloid Interface Sci. 52 (1994) 79.
[27] N. Job, F. Panariello, M. Crine, J.-P. Pirard, A. Léonard, Colloids Surf. A 293
(2007) 224.
[28] P. Sandkuhler, J. Sefcik, M. Morbidelli, Langmuir 21 (2005) 2062.
[29] A.K. Jain, Eur. J. Pharm. Biopharm. 68 (2008) 701.
[30] H. Ghanadzadeh Gilani, A. Ghanadzadeh Gilani, M. Sangashekana, J. Chem.
Thermodyn. 58 (2013) 142.
[31] J. Monni, L. Alvila, T.T. Pakkanen, Ind. Eng. Chem. Res. 46 (2007) 6916.