Journal of Colloid and Interface Science 396 (2013) 187–196

Contents lists available at SciVerse ScienceDirect

Journal of Colloid and Interface Science

www.elsevier .com/locate / jc is

Gelation of microsphere dispersions using a thermally-responsive graft polymer

Nur Nabilah Shahidan a, Cameron Alexander b, Kevin M. Shakesheff b, Brian R. Saunders a,⇑a Biomaterials Research Group, School of Materials, The University of Manchester, Grosvenor Street, Manchester M13 9PL, UKb School of Pharmacy, The University of Nottingham, University Park, Nottingham NG7 2 RD, UK

a r t i c l e i n f o

Article history:Received 4 December 2012Accepted 11 January 2013Available online 31 January 2013

Keywords:Particle gelNetworkPolycaprolactoneThermoresponsive polymerPMe2OMAIsostrainIsostress

0021-9797/$ - see front matter � 2013 Elsevier Inc. Ahttp://dx.doi.org/10.1016/j.jcis.2013.01.025

⇑ Corresponding author. Fax: +44 161 306 3586.E-mail address: brian.saunders@manchester.ac.uk

a b s t r a c t

Dispersions of microspheres (MSs) that form self-supporting particle gels are fundamentally interestingfrom the viewpoints of gel formation and mechanical properties. Here, we investigate model mixed MS/thermally responsive polymer dispersions that exist as particle gels at 37 �C. The MS comprised poly(cap-rolactone) (PCL) and was prepared by solvent evaporation. The thermally responsive polymer contained acationic backbone and poly(2-(2-methoxyethoxy)ethyl methacrylate) side chains and is abbreviated asPMA. Mixed PCL/PMA dispersions formed weak gels due to depletion at 20 �C. At higher temperaturesthey formed stronger gels due to a combination of bridging of PCL MS by PMA and reinforcement by aPMA network. A key parameter controlling the mechanical properties of the reinforced MS particle gelswas the volume fraction of PMA with respect to total polymer present (UPMA). Self-healing behaviour wasobserved for the gels using dynamic rheology and this depended on UPMA. The MS particle gel mechanicalproperties were conceptually described in terms of isostress and isostrain blending laws. At UPMA lessthan or greater than 0.057 the gels were dominated by the PCL or PMA networks, respectively. The lattervalue is suggested to be analogous to a phase inversion point.

� 2013 Elsevier Inc. All rights reserved.

1. Introduction

Particle gels are space-filling networks comprising inter-con-nected polymer particles [1–4]. They have attracted considerableacademic interest because they are believed to form by the sametype of thermodynamic phase instability that applies to solid–gasequilibrium. The gel formation is termed arrested phase separation[3]. The elasticity of a particle gel is determined by the numberdensity of elastically effective particle chains and the inter-particlebond strength. We have previously investigated particle gels forpotential application as injectable scaffolds [4–6]. Biocompatibleinjectable gel forming dispersions can be readily prepared [4,5].They have usually comprised a mixed dispersion of microspheres(MSs) and thermoresponsive polymer. The use of MS particles fol-lowed earlier work where emulsion droplets were used [7]. Thefundamental aspects of gel formation and elasticity of thermallyresponsive particle gels have received little attention. Here, weinvestigate poly(caprolactone) (PCL) MS dispersions containingadded thermally responsive polymer. The latter consisted of a pos-itively charged backbone and poly(2-(2-methoxyethoxy) ethylmethacrylate) side chains (PMe2OMA) and is abbreviated as PMA.PMA triggered formation of MS particle gels. Here, we investigatethe roles of each network (PCL–PCL and PMA–PMA) in determining

ll rights reserved.

(B.R. Saunders).

the mechanical properties of the mixed gel. We examine whetherthe elasticities of the mixed gels could be conceptually describedin terms of the isostrain and/or isostress blending law models pro-posed by Morris [8]. The models provided in that seminal studyenvisaged hard filler particles dispersed within a soft matrix (iso-stress) and soft filler particles dispersed within a hard matrix (iso-strain). Here, we vary the proportions of the MS andthermoresponsive polymer to move between composition regionswhere each model should apply.

Thermoresponsive polymers, particles and gels have attractedconsiderable interest [9–15]. PMA is shown in Scheme 1a and con-tains PMe2OMA side-chains (shown in black). The latter belong tothe thermally responsive polymers developed by Lutz and Hoth[11] and provides PMA with thermoresponsive properties. PMAdiffers in composition, charge and architecture to previous ther-mally responsive polymers that have been used to prepare biode-gradable dispersions that form gels when heated [4–6]. Ourprevious work showed that PMA existed as unimolecular micellesat room temperature [16] in dilute solutions. As part of the presentstudy, we show here for the first time that concentrated PMA solu-tions form physical gels when heated (Scheme 1a).

Polymers that form gels when heated to 37 �C have receivedconsiderable attention for potential biomaterials applications andhave been the subject of a number of reviews [17–20]. By forminga space-filling gel they can occupy the whole space of a defect orinjury in the body. Furthermore, they can be delivered using aminimally-invasive injection. However, gels that form through

(a)

(b)

Scheme 1. Thermally-triggered bridging gel formation of PCL/PMA particle gels. (a) PMA exists as unimolecular micelles at low temperature and forms micellar aggregates attemperatures greater than the cloud point temperature (Tcp). At higher concentrations physical gels form. (b) Mixed PCL/PMA dispersions formed weak depletion gels attemperatures less than Tcp. At temperatures greater than Tcp reinforced MS particle gels form.

188 N.N. Shahidan et al. / Journal of Colloid and Interface Science 396 (2013) 187–196

thermally-triggered physical gel formation tend to have low tomoderate elasticities. Their potential regenerative medicine appli-cations are principally for non-load-bearing soft tissue repair.Injectable gels have improved prospects for delivery if they canbe loaded with high concentrations of actives. We have investi-gated mixtures of thermally responsive polymers and biodegrad-able particles [4,21] for use as injectable scaffolds. Inclusion ofbiodegradable particles offers potential advantages in the form ofhigher active loadings than conventional polymer gels as well asbuilt-in porosity, due to space-filling aggregate structures whichtend to form when thermally-triggered gelation occurs. The princi-ples governing gel formation and mechanical properties need to befully understood in order to maximise the potential of injectablegel forming MS dispersions for biomaterial use. Here, we focuson the fundamental aspects of a model system.

The approach used in this study is depicted in Scheme 1. PCLMS particles were prepared using solvent evaporation in the pres-ence of cetyltrimethylammonium bromide (CTAB) and this gavecationic microspheres (Scheme 1b). CTAB would not be suitablefor application as a biomaterial because it is well known to becytotoxic. The present system was a model gel forming mixture.PCL MS particles have also been prepared using PVA in our labo-ratories and the dispersions can also form gels when mixed withPMA. Those systems are not considered here. PMA existed as uni-molecular micelles at [16] 20 �C and underwent micellar aggrega-tion and gelation when heated [16] (Depicted in Scheme 1a). Itwill be shown that PCL/PMA dispersions formed depletion gelsat temperatures less than the cloud point temperature (Tcp) dueto non-adsorbed PMA (Scheme 1b). At temperatures greater thanTcp we envisage mixed PMA/PCL particle gels containing two net-works (PCL–PCL and PMA–PMA). We aimed to identify the factorsgoverning the elasticity of each of the networks and, hence, theircontributions to the elasticity of PCL/PMA gels. Here, we invokethe isostress and isostrain blending law models [8] to provideuseful frameworks for describing the mechanical properties ofthese reinforced MS particle gels. The results of this study shouldbe generally applicable to gel-forming mixtures of colloid parti-cles and polymers, especially those that involve temperature-responsive gelation.

2. Experimental section

2.1. Reagents

Polycaprolactone (PCL, Aldrich, Mn of about 10,000 g/mol) andCTAB (P99%) were purchased from Sigma Aldrich. CH2Cl2 (99%,Fisher Scientific) was analytical reagent grade. Water was Milli-Qgrade quality.

2.2. Thermally responsive cationic polymer synthesis

The method used to synthesise PMA is depicted in Scheme S1and was described fully earlier [16]. This involved use of a polycat-ionic macroinitiator established by Chen et al. [22]. PMA was pre-pared using ATRP [16]. The polycationic backbone for PMAcontained 46 units and the side chains contained 101 MeO2MAunits (Scheme 1a). The number-average molecular weight forPMA determined in previous work was 450,600 g/mol. PMA wascationic and adsorbed strongly to negatively charged surfacesand the polar phase used for GPC analysis. Because this preventeduse of GPC the polydispersity for PMA was not able to be deter-mined [16].

2.3. PCL dispersion preparation

PCL dispersions were prepared using solvent evaporation. PCL(1.5 g) was dissolved in CH2Cl2 (100 mL). The CH2Cl2 solutionwas then fed into a beaker of aqueous 1 wt.% CTAB solution(200 mL) with a feeding rate of 2.5 mL/min whilst high shear mix-ing using a Silverson LR4 high speed mixer. Throughout this pro-cess the beaker was cooled to 0 �C. The final product was stirredovernight at room temperature to remove CH2Cl2. The PCL disper-sion was then centrifuged and the supernatant removed and re-placed with water and the particles redispersed. These stepswere repeated once more. The PCL dispersion was filtered usingpre-CTAB washed filter paper (11 lm pore size). Finally, freeze-drying was used to obtain the PCL MS particles as a redispersable

N.N. Shahidan et al. / Journal of Colloid and Interface Science 396 (2013) 187–196 189

powder. The zeta potential of the as-made particles was 67 mV.This decreased to 15 mV after the washing procedure used here.

2.4. PCL(x)/PMA(y) dispersion preparation

The abbreviation PCL(x)/PMA(y) is used to describe the mixedgels investigated here. The parameters x and y are the volume frac-tions of PCL or PMA (/PCL and /PMA), respectively, with respect tothe total dispersion volume. The densities of the PCL and PMA usedwere 1.146 and 1.0 g/ml. Gel-forming dispersions were preparedby mixing PCL dispersions and PMA solutions with the requiredconcentrations. The following describes the preparation ofPCL(0.27)/PMA(0.012). The dispersion was prepared by adding aconcentrated dispersion of PCL (0.15 g, 41 wt.%) to an aqueoussolution of PMA (0.05 g, 4.8 wt.%). The vial was placed in an icebath and subjected to frequent vortex mixing. The dispersionswere then placed on a rotating wheel and gently mixed (end-over-end) at room temperature overnight. The mixed gels are alsoidentified in this work in terms of the volume fraction of solid PMA(UPMA) with respect to the total polymer content of mixed disper-sions according to the following equation:

UPMA ¼/PMA

/PMA þ /PCLð1Þ

2.5. Physical measurements

Photon correlation spectroscopy (PCS) and zeta potentialmeasurements were obtained using a Malvern Zetasizer instru-ment with variable temperature capability. The instruments’software automatically accounted for the effect of temperatureon viscosity. The zeta potential measurements were conductedin the presence of 0.001 M NaCl. SEM measurements were per-formed using a Philips FEGSEM instrument. The freeze-dried par-ticle gel samples were prepared using samples that had formedgels at 37 �C (bridging gels) or 20 �C (depletion gels). The bridg-ing gels were prepared in an incubator oven at 37 �C for 30 mins.The samples were dropped into liquid nitrogen and immediatelyfreeze-dried. An Olympus BX41 microscope was used to obtainoptical images and to determine the particle size. Dynamicrheology measurements were performed using a TA InstrumentsAR G2 temperature-controlled rheometer and measured theelastic (G0) and loss (G00) modulus values as a function offrequency or strain (c). A 20 mm diameter plate geometry wasused. The gap was 100 lm. A small quantity of dodecane wasapplied to the edges of the samples to prevent evaporation. Afrequency of 1 Hz was used for the strain amplitude measure-ments. A strain of 0.1% was used for the frequency sweepmeasurements.

3. Results and discussion

3.1. Thermally-responsive PMA polymer characterisation

In the following we briefly review key properties of the ther-mally-responsive PMA polymer and also present new data. Thecomposition of PMA was [16] PTMAþ23-g-(PMeO2MA101)23. The cal-culated unimolecular micelle length was [16] 60 nm based on thenumber-average molecular weights of the side-chains(Scheme 1a). PMA is related to bottlebrush polyelectrolytes [23].However, PMA had a much lower aspect ratio and the unimolecu-lar micelles and is more star-like (Scheme 1a). The cloud pointtemperature (Tcp) for PMA was 27 �C in water as judged byvariable temperature transmittance measurements (Fig. 1a). PMAformed unimolecular micelles at temperatures less than Tcp. The

hydrodynamic diameter (dh) of the micelles was measured as110 nm at 20 �C (Fig. 1b). Aggregation occurred when the temper-ature exceeded Tcp. However, this was limited in dilute dispersionsand colloidally stable dispersions resulted at temperatures greaterthan 25 �C. The zeta potential (f) increased for the micelles withtemperature (Fig. 1b) which was due to the cationic backbonesmoving closer to the aggregate surface. The micellar aggregateswere charge stabilised in water. A representative SEM image(Fig. 1c) shows nanoparticles were deposited and these originatefrom the micelles (Fig. 1c). Although a range of sizes is present(80–160 nm) the diameters are consistent with the dh value of110 nm measured at 20 �C (Fig. 1b). It is possible that polydisper-sity of PMA contributed to the distribution of nanoparticle sizesapparent from Fig. 1c.

We report here for the first time that concentrated PMA solu-tions formed gels when the temperature was increased beyond21 �C (see Fig. 1d). Fig. 1e shows a temperature ramp study of G0

(elastic modulus) and tand for a concentrated PMA solution(/PMA = 0.12). As the temperature increased beyond 21 �C, the G0

value increased and tand decreased below 1.0. The G0 increasedfrom 10 Pa at 21 �C to 325 Pa at 37 �C. Fig. 1f shows strain sweepdata for PMA gels at 37 �C that contained /PMA = 0.012, 0.017 and0.050. These weak gels were very ductile and tand remained below1.0 at strain values (c) of 100% or more. The yield strain (cc) is thevalue for c where tand = G00/G0 = 1.0 (G00 is the loss modulus). The cc

value was greater than or equal to 100%. The elastic energy perPMA molecule (UPMA) can be estimated from the G0 and /PMA valuesgiven above for /PMA = 0.017, 0.05 and 0.12 using G = m kT (m isnumber density of PMA). The values are 0.1, 0.4 and 0.5 kT perPMA molecule, respectively. Because gels were present at 37 �Cthe UPMA value per elastically effective chain must have had an en-ergy of at least kT. Average UPMA values of less than kT imply that asignificant proportion of the PMA chains were not elastically effec-tive. The PMA micelles were aggregated within the gels. The vari-ation of G0 with /PMA is shown in Fig. 1g and a linear relationshipwas apparent over the range studied. Related, poly(N-isopropylac-rylamide) graft polymers showed supra-linear behaviour [24]. Itmay be that the greater flexibility for PMeO2MA side-chains de-creased the gradient for PMA.

3.2. PCL MS particle characterisation

The PCL microspheres had a number-average size of 1.1 lm(coefficient of variation, CV = 23%) as determined from opticalmicroscopy (Fig. 2a). The MS particles were polydisperse due tothe emulsification/solvent-evaporation method used for theirpreparation. The measured hydrodynamic diameter (dh) was also1.1 lm. The zeta potential measured for the PCL MS particles was15.1 mV, showing that they were cationic, which was due to resid-ual CTAB.

SEM images of deposited PCL microspheres (Fig. 2b–d) gave anumber-average diameter (dSEM), which was 0.95 lm (CV = 21%).High magnification SEM images of the PCL MS particles (Fig. 2cand d) revealed that the PCL microsphere surfaces had a wrinkledmorphology. A related morphology has been reported for PCL par-ticles prepared using electrospraying by Ding et al. [25] and Bocket al. [26] The wrinkled particle morphology is most likely due topartial MS shrinkage. This could also account for the dSEM valuebeing smaller than the value measured by optical microscopy. Fur-thermore, frequent holes were apparent in a high proportion of theMS particles (arrows in Fig. 2c and d). The higher magnification im-age (Fig. 2d) shows the holes were indentations caused by partialcollapse of the MS particles. It would seem likely that the unusualPCL surface morphologies are due to the evaporation treatmentused for SEM.

(c)

μμ (d)

0

20

40

60

80

100

15 20 25 30 35 40 45

Tran

smitt

ance

/ %

Temperature / oC

(a)

0

50

100

150

200

d h(n

m) o

r ζ(m

V)

(b)

ζ

dh

0.1

1

10

100

G' (

Pa) o

r tan

δ

Temperature (°C)

G'tan δ

(e)

G' = 2990φPMA -38.8R2 = 1.00

0.0

0.5

1.0

1.5

2.0

0

100

200

300

400

0.00 0.05 0.10 0.15

tan δ

G' /

Pa

φPMA

G'tan δ

(g)

0.1

1

10

100

0.1 1 10 100

G' (

Pa) a

nd ta

nδ

strain / %

G' (0.05) tan (0.017)G'(0.017) tan (0.05)G'(0.012) tan (0.012)

(f)

δ

δδδ

15 20 25 30 35 40 45Temperature / oC

15 20 25 30 35 40 45

Fig. 1. Thermally-responsive PMA behaviours. (a) The variation of transmittance of PMA solutions in water with temperature. (b) The variation of hydrodynamic diameterand zeta potential with temperature. (c) An SEM image of deposited micelles. (d) An image of a PMA gel (/PMA = 0.05) at 37 �C. (e) The variation of G0 and tand withtemperature for /PMA = 0.12. (f) Strain sweep data for PMA at 37 �C using /PMA = 0.012, 0.017 and 0.050. (g) The variation of G0 with /PMA at 37 �C (see text). Data from (a) and(b) were taken from Ref. [16]. All other data presented are new.

190 N.N. Shahidan et al. / Journal of Colloid and Interface Science 396 (2013) 187–196

3.3. Mixed PCL/PMA dispersions at 20 �C

Mixtures of PCL MS particles and PMA formed weak particlegels at 20 �C. That latter was well below Tcp (27.0 �C). SEM imagesof a PCL/PMA gel freeze-dried from 20 �C (Fig. 3a and b) show thatthe wrinkles of the original PCL MS particles (Fig. 2c and d) were nolonger evident. However, particle deformation remained (arrow inFig. 3a). The higher magnification image (Fig. 3b) clearly shows anodular surface and this morphology is reminiscent of recently re-ported images for polymer vesicles coated with latex particles [27].

The small particles that reside on the PCL MS particles have a min-imum size of about 60 nm, with some larger nodules present. Thenodules are consistent with the range of sizes apparent for depos-ited micelles (Fig. 1c). These data show that PMA adsorbed to thesurface of the PCL MS particles.

Did adsorption of PMA decrease the /PMA value in solution? ForPCL MS particles with a diameter of 1.1 lm the calculated specificsurface area is 5.4 m2/g. The highest /PCL value used in this studywas 0.31. It is easy to show that monolayer coverage of PCL MSby PMA unimolecular micelles would only require /PMA of

(a)

0.5 µm

20 μμm

(c)

2 μm

2 µm(b)

(d)

Fig. 2. PCL particle size and morphology. (a) An optical micrograph of PCL MS particles dispersed in water. (b–d) SEM images of PCL MS particles deposited from water atdifferent magnifications. The arrows in (c) and (d) identify PCL MS particles containing indentations.

N.N. Shahidan et al. / Journal of Colloid and Interface Science 396 (2013) 187–196 191

1.5 � 10�4. This value can be compared to the minimum /PMA usedin this study for the mixed PCL/PMA dispersions, which was 0.012.It follows that the values of /PMA in solution would not been signif-icantly different to the nominal values. From this and Fig. 3b wepropose that the mixed PCL/PMA dispersions at room temperatureconsisted of PCL MS particles coated with a monolayer of PMA dis-persed in a solution of PMA micelles (see Scheme 1b).

It can be seen from the dynamic strain sweep data (Fig. 3c andd) that PCL/PMA dispersions formed soft gels at 20 �C. Given thatthere was non-adsorbed PMA and gelation was indicated at lowstrain (tand < 1.0 in Fig. 3d), we conclude that depletion floccula-tion [28] occurred at 20 �C. If we assume a simple model wherethe thickness of the adsorbed PMA layer is equal to the diameterof the unimolecular micelles in solution at 20 �C (110 nm) thenthe range at which depletion would begin would be when the sep-aration between the PCL surfaces was less than 330 nm (ca. 30% ofthe PCL MS diameter). The particle gel had a low elasticity at 20 �Cand was brittle with a cc value of 0.65%. Particle gels are normallybrittle [1] because the attractive inter-particle interactions respon-sible for maintaining the elastic network are short-range relative tothe particle size [3].

3.4. Effect of temperature on mixed PCL/PMA gel mechanical properties

The temperature response of a mixed PCL/PMA system wasprobed using dynamic rheology (see Fig. 4a and b). For comparison

a PCL dispersion (/PCL = 0.18) was also measured. The PCL disper-sion was a fluid at all temperatures with G0 values less than 1 Paand tand values which varied widely with values greater than1.0 (not shown for clarity). For PCL(0.23)/PMA(0.022) gel the val-ues for G0 increased with temperature (Fig. 4a). The tand value de-creased slightly with increasing temperature (Fig. 4b). The absenceof a sharp change in G0 near Tcp (27 �C) (cf. Fig. 1a) implies that thechange in the hydrophobicity that occurred for PMA are gradual interms of their influence elasticity within the particle gels. This isgenerally consistent with the temperature-dependent G0 behaviourfor concentrated PMA solutions (Fig. 1e). The PCL(0.23)/PMA(0.022) gel was very soft at 20 �C (G0 = 13 Pa). When heatedto 37 �C, the G0 value increased to 446 Pa.

There are three contributions to the G0 increase with tempera-ture for PCL/PMA mixed dispersions. Firstly, stronger PCL–PCL link-ages formed because of the temperature-triggered hydrophobicityof the surface PMA. Secondly, an increase in the number ofPCL–PCL contacts occurred as a result of temperature-triggeredMS particle aggregation. Thirdly, formation of the PMA networkfrom previously non-adsorbed PMA will have occurred which rein-forced the PCL MS network. The line of best fit for the G0 vs. /PMA

data (Fig. 1g) shows that G0 for a PMA gel at 37 �C with/PMA = 0.022 would be 27 Pa. Using this G0 value it can be estimatedthat a fraction of 0.061 of the modulus of PCL(0.23)/PMA(0.022) gelwas provided by the PMA network. Interestingly, this value isreasonably close to the value of UPMA for this system (0.087).

(f, 20 oC) 0.5 µm

1 µm

(a) (b) 0.5 µm

1

10

100

G' /

Pa

Strain / %

0

1

2

3

4

5

6

0.1 1 10 100 0.1 1 10 100

tanδ

Strain / %

(c) (d)Fig. 3. SEM images and dynamic strain sweeps for PCL/PMA mixed dispersions at 20 �C. (a) and (b) SEM images for freeze-dried mixtures of PCL(0.18)/PMA(0.044). The arrowin (a) highlights an indented MS particle surface. (c) and (d) The variation of G0 and tand, respectively, with strain for PMA(0.27)/PMA(0.012).

0.1

1

10

100

1000

G' /

Pa

Temperature / oC

PCL(0.23)/PMA(0.022)PCL(0.18)

(a)

0.0

0.2

0.4

0.6

0.8

1.0

20 25 30 35 40 20 25 30 35 40

tanδ

Temperature / °C

PCL(0.23)/PMA(0.022) (b)

0.0

0.5

1.0

1.5

2.0

tanδ

Strain / %

0.27, 0.012 at 20 0.27, 0.012 at 37

(d)

oCoC

1

10

100

1000

0.1 1 10 1000.1 1 10 100

G' /

Pa

Strain / %

0.27, 0.012 at 37 0.27, 0.012 at 20 (c)

oCoC

Fig. 4. Effect of temperature on PCL/PMA gel mechanical properties. (a and b) The variation of G0 and tand with temperature for a mixed PCL(0.23)/PMA(0.022) and also aPCL(0.18) dispersion. The arrow shows the Tcp value. (c and d) Strain sweep measurements for PCL(0.27)/PMA(0.012). The data measured at 20 �C are replotted from Fig. 3 tofacilitate comparison. The arrows in (d) show the estimated cc values.

192 N.N. Shahidan et al. / Journal of Colloid and Interface Science 396 (2013) 187–196

N.N. Shahidan et al. / Journal of Colloid and Interface Science 396 (2013) 187–196 193

Fig. 4c and d shows strain sweep data for PCL(0.27)/PMA(0.012)at 20 and 37 �C. The G0 value measured at ca. c = 0.1% increasedfrom 32 Pa at 20 �C to 910 Pa at 37 �C. The value of G0 for PMA at/PMA = 0.012 was measured as 0.8 Pa at 37 �C from the data inFig. 1f. The G0PMA/G0 value is 0.001; whereas, UPMA = 0.043. In thiscase the PMA network contributed less to G0 for the network thanfor PCL(0.23)/PMA(0.022) (above). It follows that PCL–PCL interac-tions were more dominant for PCL(0.27)/PMA(0.012). The cc valuewas 1.65% at 37 �C (Fig. 4d). This low cc value is additional supportthat PCL–PCL network interactions were dominant.

3.5. Reinforced PCL(x)/PMA(y) gels at 37 �C: effect of PCL volumefraction

In order to probe the effects of PCL–PCL interactions on G0, dy-namic rheology measurements were performed using mixed PCL/PMA gels containing low proportions of PMA. Here, the UPMA val-ues were less than 0.06. For these studies /PMA = 0.012 and /PCL

was varied between 0.20 and 0.31. The PCL/PMA gels had low fre-quency dependencies of G0 and tand (Fig. 5a and b). The low tandfrequency dependence is an indication of an ideal gel [29,30].The strain sweep data (Fig. 5c and d) showed that the PCL(x)/PMA(0.012) gels were brittle as evidenced by the rapid decreasein G0 (and increase in tand values) at strain values greater thanabout 0.2%.

Fig. 5e shows the G0 and tand data plotted in semilog form. Thedata show a strong increase in the G0 values with /PCL. The G0 valuesfor the PCL(x)/PMA(0.012) gels (Fig. 5e) were in the range of 285–2445 Pa. Fig. 5f shows a log–log plot for G0 vs. /PCL. The power law

0.1

1

10

100

1000

0 0.1 0.2 0.3 0.4

G' (

Pa) a

nd ta

nδ

φ

G'

tan

(e)

δ

1

10

100

1000

10000

0.1 1 10 100

G' /

Pa

Strain / %

0.31

0.27

0.23

0.2

(c)

10

100

1000

10000

0.1 1 10

G' /

Pa

Frequency / Hz

0.310.270.230.2

(a)

PCL

Fig. 5. Rheology data measured at 37 �C for PCL(x)/PMA(0.012) gels. (a and b) Frequencyvalues of /PCL are shown in the legends. (e) G0 and tand data. (f) Log–log plots for the G

exponent (a) for G0 was 4.8. This is the same exponent value thatwas reported for polystyrene MS gels [31] and suggests thatshort-ranged attractions were present [32]. This is additional sup-port for the view that the elasticity of PCL(x)/PMA(0.012) gels withlow UPMA values was controlled by PCL–PCL linkages. The role ofPMA was mostly to cover the PCL surface, and trigger PCL–PCL net-work formation. It may also have provided a secondary PMA net-work; however, this contribution should not have been majorbecause UPMA was less than 0.06.

We proposed above that excess PMA should have been presentfor all of the mixed dispersions because of the relatively low /PMA

required for surface coverage. If secondary PMA networks werepresent then evidence for them should be evident from SEM.Fig. 6 shows SEM images of representative samples of gel freeze-dried from 37 �C. The gel morphologies are highly porous, as ex-pected, because space-filling gels were obtained. The MS particlesurfaces appeared smooth and contrasted to those of the barePCL MS particles (Fig. 2b). Importantly, there was evidence of fea-tureless, non-spherical, structures for both gels (arrows). Theywere not present in the SEM images for PCL alone (Fig. 2b). Thesespecies are almost certainly non-adsorbed PMA. The features areconsistent with Scheme 1b. Presumably, strands of PMA micellarchains had collapsed during SEM sample preparation.

3.6. Reversibility of strain-induced PCL/PMA gel breakdown at 37 �C

The reversibility of the mixed gel networks to fracture wasprobed to learn more about the role of PMA within the MS particlegels. We selected two gels for this study. One where UPMA was low

G'= 624900 φPCL4.8

R² = 0.96

0.11

10

100

1000

10000

10.1

G'

(Pa)

and

γc

(%)

φPCL

G'

(f)

γc

0

1

2

3

4

5

tanδ

Strain / %

0.2

0.23

0.27

0.31

(d)

0.1 1 10 100

0.0

0.2

0.4

0.6

0.8

1.0

0.1 1 10

tan

Frequency / Hz

0.2

0.23

0.27

0.31

(b)

0.27

sweep data for PCL(x)/PMA(0.012) gels. (c and d) Strain sweep data for the gels. The0 and cc data.

Fig. 6. Morphology of PCL(x)/PMA(y) gels. SEM images of freeze-dried bridging gelsare shown. The gels were prepared at 37 �C. The arrows highlight examples ofaggregated PMA.

194 N.N. Shahidan et al. / Journal of Colloid and Interface Science 396 (2013) 187–196

(PCL(0.27)/PMA(0.012), UPMA = 0.043) and the other where it wasrelatively high (PCL(0.20)/PMA(0.044), UPMA = 0.18)). The strainprofile employed is shown in Fig. S2a. Each sample was measuredat low strain (test cycle) and then, after short delay (1 min at zerostrain), subjected to a strain-sweep experiment (strain cycle).Fig. S2b and c shows the self-healing measurement for PCL(0.27)/PMA(0.012). Data for PCL(0.20)/PMA(0.044) are shown in Fig. S3aand b. The strain used during the strain sweeps exceeded cc whichensured that the particle gel networks were broken during eachstrain sweep (see arrows in Fig. S2c). For PCL(0.27)/PMA(0.012)the G0 values recovered rapidly during each subsequent test period(Fig. S2b). Most of the recovery occurred before measurementsduring the subsequent test period began. The full recovery of theinitial mechanical properties after application of destructive strainrepresents complete self-healing. The data shown in Fig. S2 are thefirst report of self-healing for a temperature-triggered particle gelto the best of our knowledge. Self-healing is a useful characteristicfor gels and elastomers [33–36] because it offers potentially in-creased durability.

Fig. 7 shows the values for G0, tand and cc as a function of testperiod or cycle number. The PCL(0.27)/PMA(0.012) gels showedrelatively stable G0, cc and tand values. However, this was not thecase for the PCL(0.20)/PMA(0.044) (UPMA = 0.18). These data indi-cate that rapid repair of the fragmented particle gels is favouredat low UPMA values. For PCL(0.20)/PMA(0.044) (UPMA = 0.18) frag-mentation must have caused major reorganisation of both thePCL and PMA gel networks. Because the PMA gel network wouldhave been extensive for that system, it must have interfered withre-establishing PCL–PCL contacts, by providing more PMA networkbetween the PCL aggregates. In contrast, the PMA network musthave been smaller for PCL(0.27)/PMA(0.012) (UPMA = 0.043) andPCL–PCL MS particle contacts should have rapidly re-established.

Support for this proposal comes from the increase in cc and tandwith strain cycle which occurred for PCL(0.20)/PMA(0.044)(Fig. 7b and c). This indicates shear-induced transformation to amore viscous, ductile gel as a consequence of network. Further-more, from Fig. 1f G0 = 91 Pa for /PMA = 0.044. The initial G0 was640 Pa, and the modulus due to the PMA network would then bea fraction of 0.14 of the total which is close to UPMA = 0.18. How-ever, as the strain cycles proceed, the G0 decreased to 213 Pa (Straincycle 2) and 138 Pa (Strain cycle 3), which corresponds to the PMAnetwork particles being responsible for increasing fractions (of0.43 and 0.65, respectively) of the modulus assuming that thePMA network was not broken. (The cc values for PMA networkswere much higher than 100% (Fig. 1e) and so it is plausible thatthe PMA network did not break at 100% strain.) This analysis sup-ports the view that strain-induced disruption of the reinforced par-ticle network irreversibly decreased the proportion of PCL–PCLlinkages over the timescale involved in the rest cycles (ca. 4 min)for PCL(0.20)/PMA(0.044).

3.7. Proposed gel structures for reinforced microsphere particle gels

We propose from the data presented above that PCL/PMA gelscan be considered as mixed gels. There were two networks (PCL–PCL and PMA–PMA) and these were interconnected via PCL–PMAcontacts at the PCL MS particle surface. The PMA network, whichformed upon temperature increase, reinforced the PCL particlegel. We have found that for the systems studied the value of UPMA

controlled the network mechanical properties. This is explored fur-ther in Fig. 8a and b which show the variation of G0 and cc as a func-tion of UPMA for all of the systems studied here. (Unfortunately,strain sweeps were not conducted for PCL(0.23)/PMA(0.022),which had UPMA = 0.087.) There was a pronounced change inbehaviour for G0 with a negative gradient changing to a positivegradient when a critical value, UPMA(c) = 0.057, was reached. Thischangeover occurred in a region where /PCL did not change greatly(Fig. 8a) and marked a transition. When UPMA < UPMA(c) the particlegel was dominated by PCL–PCL linkages and G0 � /PCL4:8 (Fig. 5f).When UPMA > UPMA(c) the gradient is positive and a more gradualincrease in G0 occurred.

A mechanical model was developed for biphasic hydrogels byMorris [8] and considered discrete high modulus particles dis-persed within a lower modulus continuous phase as well as the in-verse structure. Inspired by that work we have depicted relatedstructural models for our system in Fig. 8c and d. We propose thatat low UPMA values (less than UPMA(c)) PMA network units are dis-persed within a continuous PCL MS particle network (Fig. 8c. Thiswould correspond to the isostrain model [8]. For this model theG0 values should have been dominated by the PCL network. It canbe seen from Fig. 8a that the G0 behaviour was dominated by thehigh modulus phase, which followed G0PCL � /PCL4:8 in that region(Fig. 5f) as /PCL increased with decreasing UPMA (Fig. 8a). Further-more, for the isostrain model the strain should have been domi-nated by that of the stiffer, PCL, network, i.e., c = cPCL. Fig. 8bshows that the cc values in this region were low and dominatedby PCL–PCL interactions.

We further propose that when UPMA > 0.057 the stiffer PCLphase was dispersed in the softer PMA network (Fig. 8d). This sit-uation would correspond to the isostress model. Accordingly, G0

should increase with UPMA and this was observed (Fig. 8a). Further-more, the cc values should be higher than for the isostrain case be-cause it should be dominated by the softer PMA network. There isstrong evidence from the self-healing experiment (PCL(0.20)/PMA(0.044), UPMA = 0.18) that cc was much higher (Fig. 8b). Ourresults shown in Fig. 8a and b can be generally explained by thefeatures depicted in Fig. 8c and d. There was a changeover froma brittle isostrain gel to a more ductile isotress gel at

10

100

1000

G' /

Pa

Test period

0.27, 0.0120.20, 0.044

(a)

0.1

1

10

100

0 1 2 3 4 5 0 1 2 3 4 5

tanδ

orγ

c/ %

Strain cycle

(0.20, 0.044)(0.27, 0.012)

tan (0.20, 0.044)tan (0.27, 0.012)

(b)

δδ

γcγc

Fig. 7. PCL(x)/PMA(y) gel response to excessive strain (a) and (b) show the variation of G0 , tand and cc, for PCL(0.27)/PMA(0.012) at 37 �C after a network failure due toexcessive strain using the strain ramps shown in Fig. S2a. The G0 and tand values were measured at 4 min. The values for c were measured during the strain cycles. The fulldata set are shown in Figs. S2 and S3.

0

500

1000

1500

2000

2500

0

2

4

6

8

10

0.0

0.5

1.0

1.5

2.0

2.5

0.00 0.05 0.10 0.15 0.20 0.00 0.05 0.10 0.15 0.20

(a) (b)

(c) (d)

Fig. 8. Reinforced network mechanical properties and model. (a) The variation of G0 and with UPMA. Data for /PCL are also shown for comparison. (b) The variation of cc andtand with UPMA. (c and d) Depictions of the gel structures that are proposed to exist either side of UPMA(c) – see text. The dashed lines illustrate continuous networks. The boldlines illustrate dispersed (isolated) networks. The equivalent mechanical models are also depicted.

N.N. Shahidan et al. / Journal of Colloid and Interface Science 396 (2013) 187–196 195

UPMA(c) (=0.057). UPMA(c) can be thought of as a stress distributioninversion point. This is a new observation for particle gels. Thesethermally triggered gels have interesting tuneability potentialand it appears that good mechanical property control can beachieved through adjusting the responsive polymer and MSvolume fractions as well as their relative proportions, i.e., /PMA,/PCL and UPMA. This should be useful for potential injectablescaffold applications in future work because mechanical propertytuneability is important in that context.

4. Conclusions

The new hypothesis proposed from this study is that themechanical properties of thermally-triggered particle gels [4–6]

obey the isostress and isostrain blending models originally pro-posed by Morris [8] for covalent gels. This fundamental study hasinvestigated the gelation properties of mixed MS/thermorespon-sive polymer dispersions and the mechanical properties of thereinforced particle gels. The mixed dispersions formed weak deple-tion gels 20 �C. When heated the PMA at the surface triggeredstronger, adhesive, interactions and a secondary PMA networkformed. These interactions increased the elasticity. A low value of/PMA enabled rapid repair of the structure when cc was exceeded.An important result from this study is the first example of self-healing particle gel. The results suggest that if UPMA was less than0.057 the isostrain model can be used to describe the particle gel,which was brittle (and self healing). If UPMA was greater than 0.057then the isostress model appeared to apply and these gels were

196 N.N. Shahidan et al. / Journal of Colloid and Interface Science 396 (2013) 187–196

more ductile. Fig. 8a and b shows that there are two ways to in-crease G0 (by increasing or decreasing UPMA relative to UPMA(c))and they have different effects on the brittleness of the reinforcedMS particle gels. There are a number of thermally-responsive par-ticle gels that have been reported and they have potential applica-tion as injectable tissue scaffold. We propose that the blendinglaws described here should generally apply to those systems[4–6] and others containing dispersed particles and gel-formingpolymers. For the broader field of injectable particle gels, which in-cludes [4–6], the results of this study should enable selection offormulation parameters to construct injectable MS particle gelsthat have desired modulus and ductility for use in low shear,non-load bearing, environments. Moreover, this study illustratesthe versatility of the original blending models of Morris [8] fordescribing mechanical properties of complex soft matter.

Acknowledgments

We would like to thank the Malaysian government for funding.This work was in part funded by the EPSRC Centre for InnovativeManufacturing in Regenerative Medicine. Research leading to theseresults received funding from the European Research Council un-der the European Community’s Seventh Framework Programme(FP7/2007-2013)/ERC Grant agreement 227845 and also the EPSRC.

Appendix A. Supplementary data

Supplementary data associated with this article can be found, inthe online version, at http://dx.doi.org/10.1016/j.jcis.2013.01.025.

References

[1] B.H. Bijsterbosch, M.T.A. Bos, E. Dickinson, J.H.J. van Opheusden, P. Walstra,Faraday Discuss 101 (1995) 51.

[2] A.D. Dinsmore, V. Prasad, I.Y. Wong, D.A. Weitz, Phys. Rev. Lett. 96 (2006)185502.

[3] P.J. Lu, E. Zaccarelli, F. Ciulla, A.B. Schofield, F. Sciortino, D.A. Weitz, Nature 453(2008) 499.

[4] W. Wang, H. Liang, R.C. Al Ghanami, L. Hamilton, M. Fraylich, K.M. Shakesheff,B. Saunders, C. Alexander, Adv. Mater. 21 (2009) 1809.

[5] R.C. Al Ghanami, B.R. Saunders, C. Bosquillon, K.M. Shakesheff, C. Alexander,Soft Matter 6 (2010) 5037.

[6] M. Fraylich, R. Liu, S.M. Richardson, P. Baird, J. Hoyland, A.J. Freemont, C.Alexander, K. Shakesheff, F. Cellesi, B.R. Saunders, J. Colloid. Interface Sci. 344(2010) 61.

[7] A.Y.C. Koh, C. Prestidge, I. Ametov, B.R. Saunders, Phys. Chem. Chem. Phys. 4(2002) 96.

[8] E.R. Morris, Carbohydr. Polym. 17 (1992) 65.[9] J. Madsen, S.P. Armes, Soft Matt. 8 (2012) 592.

[10] B. Liu, S. Perrier, J. Polym. Sci. (A): Polym. Chem. 43 (2005) 3643.[11] J.F. Lutz, A. Hoth, Macromolecules 39 (2006) 893.[12] O.J. Cayre, N. Chagneux, S. Biggs, Soft Matter 7 (2011) 2211.[13] R. Liu, M. Fraylich, B.R. Saunders, Colloid Polym. Sci. 287 (2009) 627.[14] A. Miasknikova, A. Laschewsky, G. De Paoli, C.M. Papadakis, P. Muller-

Buschbaum, S.S. Funari, Langmuir 28 (2012) 4479.[15] A.M. Budgin, Y.A. Kabachii, Z.B. Shifrini, P.M. Valetsky, S.S. Kochev, Langmuir

28 (2012) 4142.[16] N. Shahidan, R. Liu, F. Cellesi, C. Alexander, K.M. Shakesheff, B.R. Saunders,

Langmuir 27 (2011) 13868.[17] N.C. Hunt, L.M. Grover, Biotechnol. Lett. 32 (2010) 733.[18] H.J. Chung, T.G. Park, Nano Today 4 (2009) 429.[19] L. Yu, J. Ding, Chem. Soc. Rev. 37 (2008) 1473.[20] M.K. Nguyen, D.S. Lee, Macromol. Biosci. 10 (2010) 563.[21] M. Fraylich, W. Wang, K. Shakesheff, C. Alexander, B. Saunders, Langmuir 24

(2008) 7761–7768.[22] X.Y. Chen, S.P. Armes, S.J. Greaves, J.F. Watts, Langmuir 20 (2004) 587.[23] T. Pettersson, A. Naderi, R. Makuska, P.M. Claesson, Langmuir 24 (2008) 3336.[24] R. Liu, F. Cellesi, N. Tirelli, B.R. Saunders, Polymer 50 (2009) 1456.[25] L. Ding, T. Lee, C.H. Wang, J. Controlled Release 102 (2005) 395.[26] N. Bock, M.A. Woodruff, D.W. Hutmacher, T.R. Dargaville, Polymers 3 (2011)

131.[27] R. Chen, D.J.G. Pearce, S. Fortuna, D.L. Cheung, S.A.F. Bon, J. Am. Chem. Soc. 133

(2011) 2151.[28] P. Jenkins, M. Snowden, Adv. Colloid. Interface Sci. 68 (1996) 57.[29] H.H. Winter, Polym. Eng. Sci. 27 (1987) 1698.[30] H.H. Winter, F. Chambon, J. Rheol. 30 (1986) 367.[31] C. Zhao, G. Yuan, D. Jia, C.C. Han, Soft Matter 8 (2012) 3036.[32] C. Zho, G. Yuan, D. Jia, C.C. Han, Soft Matter 8 (2012) 3036.[33] F. Liu, F. Li, G. Deng, Y. Chen, B. Zhang, J. Zhang, C.-Y. Liu, Macromolecules 45

(2012) 1636.[34] J. Fox, J.J. Wie, B.W. Greenland, S. Burattini, W. Hayes, H.M. Colquhoun, M.E.

Mackay, S.J. Rowan, J. Am. Chem. Soc. 134 (2012) 5362.[35] Y. Amamoto, H. Otsuka, A. Takahara, K. Matyjaszweski, Adv. Mater. 24 (2012)

3975.[36] R. Martin, A. Rekondo, J. Echeberria, G. Cabanero, H.J. Grande, I. Odrozola,

Chem. Commun. 48 (2012) 8255.