LETTERS

Self-assembly in sugar–oil complex glasses

HITESHKUMAR DAVE1, FENG GAO1, JING-HUEI LEE2, MATTHEW LIBERATORE3, CHIA-CHI HO1 ANDCARLOS C. CO1*1Department of Chemical and Materials Engineering, University of Cincinnati, Cincinnati, Ohio 45221-0012, USA2Department of Biomedical Engineering, University of Cincinnati, Cincinnati, Ohio 45267-0586, USA3Department of Chemical Engineering, Colorado School of Mines, Golden, Colorado 80401-1887, USA*e-mail: carlos.co@uc.edu

Published online: 25 March 2007; doi:10.1038/nmat1864

In aqueous systems, the hydrophobic effect drives theself-assembly of amphiphiles into a broad range of micellar,rod-like, bicontinuous and liquid-crystalline complex fluids.Many of these are relevant to biological matter or technologicalapplications. However, amphiphilic self-assembly is not limitedto aqueous systems. Replacement of water with supercriticalcarbon dioxide, for example, results in complex fluids thatcombine the properties of gases and liquids1. Along this vein,we explore the self-assembly of surfactants in anhydroussugars. Our study reveals that anhydrous powders of sugarsand surfactants suspended in oil spontaneously form moltenglasses with nanometre-size domains of sugar and liquid oilwithout mixing. The low cost, water solubility, low toxicity andstabilizing properties of glassy sugars make them ideal waterreplacements for many pharmaceutical, food and materialssynthesis applications. The optical clarity and solid appearanceof these glasses at room temperature belie their inclusion ofmore than 50% (vol.) oil, which confers liquid-like diffusivity.The unique combination of solid- and liquid-like properties maylead to applications in sensors and optical devices.

The spontaneous solubilization of sugars (equimolarsucrose/trehalose) and oil (limonene) by surfactants (sucroseoleate/sucrose caprylate), as shown in Fig. 1, bears much similarityto the surfactant-driven spontaneous microemulsification of waterand oil mixtures2. However, on cooling below their glass-transitiontemperature (∼330 K), molten glasses solidify and crack intopieces. The hardness (0.7 mohs) and optical clarity of these glassesare remarkable considering that they contain ∼52 vol% liquidoil. Although a lower temperature bound for these one-phasemicroemulsion glasses is certain to be present, phase separationhas not been observed for samples cycled between molten and solidstates or slowly cooled through the glass transition in thermostatedwater baths. Dynamically arrested microemulsion glasses kept forover six months at room temperature also show no signs of phaseseparation or crystallization.

In the molten state at 365 K, the phase behaviour of sugar–oil microemulsions closely parallels that observed for water–oilmixtures (Fig. 2). For equal mass loadings of oil and sugar,the one-phase region (Fig. 2a) is bounded by upper and lowerlimits in the mass ratio of long-chain to short-chain surfactants(ε). At a fixed overall surfactant mass loading (γ) of 20 wt%(Fig. 2b), a continuous one-phase channel extends from the sugar–surfactant binary to ∼60 wt% oil loading on a surfactant-freebasis (α), corresponding to a sugar volume fraction of ∼22%,beyond which solid microemulsion glasses do not form. The

extreme viscosity of the sugar–oil mixtures precludes assignmentof the multiphase regions bounding the one-phase regions intotwo-phase and three-phase regions typically observed in aqueoussystems3. Nonetheless, the interfacial curvature, as set by the ratioof long-chain to short-chain surfactants4, remains the deterministicvariable controlling microemulsification of sugar–oil mixtures inthe molten-glass state.

Although solids at room temperature, the structure ofsugar–oil microemulsions is similar to that of liquid water–oilmicroemulsions. Their small-angle neutron scattering (SANS)spectra (Fig. 3) are consistent with those of bicontinuousmicroemulsions, and can be fitted well with the classical Teubner–Strey model5 to yield domain sizes (d) of 25–28 nm andcorrelation lengths (ζ) of ∼10 nm that are typical of aqueousmicroemulsions6. Matching the Ginzburg–Landau free-energyexpression that underlies the Teubner–Strey model to a gaussianrandom-field model7,8 yields a renormalized bending modulus(κR) of ∼0.95 kT, also typical of aqueous microemulsions. Thehigh viscosity of the sugar domains in the molten state does notsignificantly alter the bending moduli of the surfactant-coveredinterface. These SANS measurements also confirm that the opticalclarity of these oil/sugar glasses is not the result of fortuitousrefractive index matching.

Even in the solid-glass state at room temperature, the limoneneoil present in the microemulsion glasses exhibits liquid-likeself-diffusion coefficients that point to a sponge-like structurecontinuous in both solid sugar glass and liquid oil9. Usingmagnetic resonance imaging (MRI), to exclude cracks in thesamples (Fig. 4a), we observed limonene self-diffusion coefficientsthat increase from 3.0 × 10−6 to 6.3 × 10−6 cm2 s−1 (Fig. 2b)with increasing oil loading from α = 0.3 to 0.5 respectively. Incomparison, pure liquid limonene has a self-diffusion coefficient of1.3×10−5 cm2 s−1 at the same temperature. The spatial selectivityof self-diffusion coefficient measurements by MRI does come atthe loss of chemical-shift selectivity. We verified that the measuredself-diffusion coefficients are solely that of the limonene oil andexclusive of contributions from the surfactant chains by preparinga comparable microemulsion glass using deuterated toluene. Thissample yields no measurable proton signal from the surfactantalkyl chains (tube A in Fig. 4a) because of extensive linewidthbroadening arising from their slow dynamics. Thus, the measureddiffusion coefficient is solely that of the liquid oil present within thebicontinuous glass structure.

Structural relaxation processes in bicontinuous aqueousmicroemulsions are generally too rapid to probe rheologically.

nature materials VOL 6 APRIL 2007 www.nature.com/naturematerials 287

Untitled-1 1 10/3/07, 11:54:54 am

LETTERS

0.1 min 0.5 min 2 min 5 min After cooling

Figure 1 Spontaneous formation of a microemulsion glass. Sugar and surfactant powder dried to 99.5 wt% dryness dispersed in oil at room temperature ‘dissolves’ onheating to 365 K to form a one-phase molten microemulsion glass. The composition here corresponds to sample C in the phase diagram shown in Fig. 2a. Gradual cooling ofthe molten glass to room temperature yields a solid microemulsion glass (left) containing ∼52 vol% liquid oil with a Mohs hardness of 0.7.

0.16 0.17 0.18 0.19

EF

B

C

D

0.20 0.21

0.56

0.58

0.60

0.62

0.64

1

0

0

0.2

0.4

0.6

0.1 0.2 0.3 0.4 0.5 0.6 0.7

3.0 × 10–6 cm2 s–1

5.7 × 10–6 cm2 s–1

6.3 × 10–6 cm2 s–1

6.1 × 10–6 cm2 s–1

a b

ε εφ

1φ

γ α

Figure 2 Phase diagrams for molten sugar–limonene microemulsion glasses at 365K. Multiple phases are present above and below the boundaries delineating theone-phase (1φ ) microemulsion region. a, Phase diagram for a 50:50 limonene oil to sugar mass ratio corresponding to α = 50%. The upper and lower error bars correspondto single-phase and multiphasic samples bracketing the phase boundary. The letters correspond to compositions probed by SANS (Fig. 3) and MRI (Fig. 4). b, Phase diagramat fixed overall surfactant loading of γ = 20%. The values for phase boundaries at low oil loadings (α ≤ 0.3) were estimated (squares) from samples heated briefly to 400 Kthen observed at 365 K. Room-temperature self-diffusion coefficients of limonene oil present in the interstices of the microemulsion glasses measured by MRI (Fig. 4) areshown for compositions marked with stars. For comparison, the self-diffusion coefficient of pure limonene measured under identical conditions was 1.3×10−5 cm2 s−1.

Retardation of structural relaxation in molten microemulsionglasses by the high viscosity of the sugar domains however, doespermit their study using a sealed rheological cell in which themicroemulsions are prepared in situ. We constructed a mastercurve of the elastic and viscous moduli by time–temperaturesuperposition using the Williams–Landel–Ferry (WLF) equation10

(see Supplementary Information, Fig. S1) to probe the dynamicsof microemulsion glasses as they transition from molten- to solid-glass states. The WLF equation has been shown to be usefulfor systems near their glass-transition temperature11, which iscorroborated here. For sample C (Fig. 2a), the terminal relaxationtime was directly measured at 365 K from the inverse of thecrossover frequency to be 1.3 s and increasing to 6,300 s at 335 K

(determined by time–temperature superposition). In summary, therheological response confirms the transition of a viscoelastic fluidto a glass as the glass-transition temperature is approached.

Within the one-phase microemulsion channel shown in Fig. 2b,the midpoint heating glass transition (Tg,mph) of the sugars, asmeasured by modulated differential scanning calorimetry, increasesasymptotically from 325 to 358 K as the oil loading decreases fromα = 60 to 0 wt% (see Supplementary Information, Fig. S2). Thisdecrease in the glass-transition temperature with increasing oilloading is consistent with the depression of melting and glass-transition temperatures of liquids confined in pores12,13. Becausethe samples here were prepared using sugar/surfactant solutionsdehydrated to a fixed concentration of 99.5 wt%, the overall

288 nature materials VOL 6 APRIL 2007 www.nature.com/naturematerials

Untitled-1 2 10/3/07, 11:54:57 am

LETTERS

0.01 0.02 0.03 0.05 0.07 0.100.1

1.0

10

100

q (Å–1)

Inte

nsity

F: d = 28.3 nm; = 10.9 nm

E: d = 27.6 nm; = 10.9 nm

C: d = 25.1 nm; = 10.3 nm

ζ

ζ

ζ

Figure 3 SANS spectra of solid microemulsion glasses at room temperaturecorresponding to compositions F, E and C in the phase diagram shown inFig. 2a. The solid lines are fits to the Teubner–Strey model, which yields domainsizes (d ) ranging from 25 to 28 nm and correlation lengths (ζ ) of ∼10 nm.

water content does increase slightly with reduced oil loading (α).Although residual water molecules partition principally in thesugar and interfacial layer of surfactant sugar head groups, theslight variation in the ratio of sugar/surfactant/water moleculesmay also contribute to the observed variation of the glasstransition. Regardless, extracting a length scale of cooperativityfor the glass transition of sugars in these microemulsion glassesis obscured however, by the simultaneous changes in surfactantconcentration necessary to stay within the one-phase channel andthe accompanying changes in the topology of the sugar domainswith variations in oil loading.

New science and exciting applications await discovery at theintersection of glasses with complex fluids. The spontaneousmixing of molten sugars and oil to the nanometre scale,as demonstrated here using food-grade surfactants, sugarsand limonene oil, has broad consumer-product applications,potentially replacing many spray- and freeze-drying processeswith direct melt extrusion. Complex glasses may also beused as templates for nanomaterial synthesis by replacingthe limonene oil with polymerizable monomers, for example,tetrafluoroethyelene, or by non-aqueous sol–gel chemistries14–16

to permit complementary templating of the sugar-glass structure,which can be removed subsequently by simple immersion in excesswater. For applications where water solubility is undesirable, thesugar skeleton can be stabilized by reaction with multifunctionalisocynates to form urethane crosslinks. Broader applications ofthese complex glasses include sensor and optical applications wherethe clarity of the solid glasses may be coupled with the rapidresponse of molecularly mobile organic liquids.

METHODS

Sucrose laurate and sucrose oleate surfactants were supplied by MitsubishiKagaku-Foods Corporation, Japan and used as received. D-limonene (97%)with 0.1 wt% butylated hydroxytoluene, added as an antioxidant, waspurchased from Aldrich and used as received. Sucrose and trehalose dihydrate

F

E

D

C

B

G

A

π/2 π

δ δ

Δ

Radio-frequency

ADC

GSS

–G

TS

GPE

GR

0–1.6

–1.4

–1.2

–1.0

–0.8

–0.6

–0.4

–0.2

0

50 100 150 200 250

Ln (A

/A0)

2g2 2( – /3) × 1,000 s cm–2

a

b

c

γ δ Δ δ

Figure 4 MRI imaging and self-diffusion coefficient measurement ofmicroemulsion glasses. a, MRI (H1 intensity) of the cross-section of microemulsionglasses contained in a hexagonal array of seven tubes each similar to that shown inFig. 1. All of the samples contain a 50:50 mass ratio of limonene oil to sugar. SampleA was prepared using deuterated toluene instead of limonene. Samples B–G aremicroemulsion glasses prepared within the one-phase region of the limonene phasediagram (Fig. 2). b, The self-diffusion measurements were carried out using aconventional spin-echo pulse sequence, where GSS, GPE and GR are the sliceselection, phase encoding and read-out gradients and TS is the data acquisitionperiod of the analog-to-digital converter (ADC). c, The slope of the normalizedspin-echo intensity relative to the field gradient (g), gradient duration (δ), time delay(Δ) and proton gyromagnetic ratio (γ ) as shown here for sample C (Fig. 2a) yields aroom-temperature self-diffusion coefficient of 6.3×10−6 cm2 s−1.

(99%) were purchased from Acros Organics and used as received. Deionizedwater with a specific resistance of 18.2 M cm was used in the preparation ofall of the samples.

nature materials VOL 6 APRIL 2007 www.nature.com/naturematerials 289

Untitled-1 3 10/3/07, 11:55:06 am

LETTERS

PHASE-DIAGRAM DETERMINATIONSamples were prepared by gently heating mixtures of sugar, surfactant solutionsand water in flat-bottom screw-cap tubes, evaporating excess water by passingargon over the sample followed by vacuum drying. Complete dissolution andaccurate dehydration of the sugar/surfactant solutions is critical. Beforedehydration, samples were pre-weighed to an accuracy of ±0.1 mg. Initialdehydration with dry argon gas at room temperature was carried out withcontinuous rotation of the tubes to form thin viscous films of 90–92 wt%sugar/surfactant. The sample tubes were subsequently transferred to a vacuumoven and ramped carefully to 60 ◦C over ∼4 h to maximize the drying areathrough controlled foaming of the sugar/surfactant films. The sample tubeswere weighed periodically and dehydration was terminated when thesugar/surfactant mixtures reached 99.5 wt%. Drying times of 2 and 6 days aretypical for 1 and 5 g samples respectively. Complete dehydration ofsugar/surfactant solutions to 99.9+ wt%, for the sample sizes necessary here,would have required prolonged drying over 70 ◦C and riskedsample decomposition.

After the addition of d-limonene, the sugar/surfactant powders weredispersed using NdFeB stir bars at room temperature, then thermostated at365.00±0.02 with no further stirring. For samples containing α ≤ 20% oil,there was insufficient oil to disperse the sugar/surfactant powders at roomtemperature. Samples were thus heated and stirred at 400 K for 2 min to mixthe molten sugar/surfactant powder with limonene, then thermostated at 365 Kfor observation. Thermostated samples were observed for 1 h, althoughone-phase samples generally became optically transparent within 5 min. Thepresence of multiple phases was determined by visual inspection in bothtransmitted and scattered light, using crossed polarizers to check for thepresence of liquid-crystalline phases. The water content of one-phasemicroemulsion glasses, containing limonene, was validated by Karl Fischertitration, which yielded a water content of 0.6±0.1 wt% on an oil-free basisand corresponding closely to the expected water content following themass-controlled dehydration procedure.

NEUTRON SCATTERINGNeutron scattering spectra were acquired using the small-angle neutrondiffractometer at Argonne National Laboratory, using the instrument softwarefor background correction and circular averaging. Even without deuteration,we found that the coherent scattering, arising from the difference in massdensity between the sugar and oil, was sufficiently large compared with theincoherent scattering background.

SANS spectra were analysed using the Teubner–Strey model5 for thescattering I(q) from bicontinuous microemulsions:

I(q) = 1

a2 + c1q2 + c2q4+b.

In this model, a2,c1 and c2 are variables related to the Landau free-energyexpansion and b is the incoherent background. The physical parameters dTS

and ξTS, representing the domain size (periodicity) and correlation lengthpresent in the microemulsions, respectively are calculated from the fittedparameters through:

dTS = 2π

[1

2

(a2

c2

) 12 − 1

4

c1

c2

]− 12

ξTS =[

1

2

(a2

c2

) 12 + 1

4

c1

c2

]− 12

.

MODULATED DIFFERENTIAL SCANNING CALORIMETRYModulated differential scanning calorimetry measurements were carried outusing a TA Instruments Q100 calorimeter (New Castle) with modulationamplitude, time and heating rate set to ±1 ◦C, 60 s and 2 ◦C min−1,

respectively. N2 purge flow was set to 50 ml min−1. The samples were sealed inaluminium pans during the measurements. The reversible heat capacity ofmicroemulsion glasses was plotted against temperature to determine themidpoint heating glass-transition temperatures.

RHEOMETRYA TA Instruments AR-2000 (New Castle) controlled-stress rheometer equippedwith a high-pressure cell was used for the rheological measurements. Althoughthe cell is capable of being pressurized to 2,000 psi, these experiments werecompleted at ambient pressure inside the sealed chamber in the temperaturerange of 335–365 K. Frequency sweeps were completed at a strain of 0.005,which was determined to be in the linear viscoelastic regime byamplitude sweeps.

MAGNETIC RESONANCE IMAGINGMRI images and self-diffusion coefficients were obtained using a BrukerBiospec 70/30 MRI scanner (Billerica) equipped with a 7 T magnet and400 mT m−1 field gradients. Samples comprising seven glass tubes containingdifferent microemulsion glasses were positioned in the centre of theradiofrequency coil (72 mm diameter). Scout images using multi-slicemulti-echo were acquired to select regions for self-diffusion coefficientmeasurements that avoided cracks present in the samples. Self-diffusioncoefficient measurements were carried out using Bruker’s standard spin-echodiffusion sequence (field of view 6×6 cm2, data matrix 128×64,repetition/echo time (TR/TE) = 2,000/60 ms, slice thickness = 5 mm, gradientduration (δ) = 10 ms, time delay (Δ) = 40 ms, NEX = 1) with diffusiongradient strengths set to 0, 6, 12, 24, 48, 60, 72 and 84 mT m−1.

Received 1 November 2006; accepted 30 January 2007; published 25 March 2007.

References1. McClain, J. B. et al. Design of nonionic surfactants for supercritical carbon dioxide. Science 274,

2049–2052 (1996).2. Kahlweit, M. & Strey, R. Phase-behavior of ternary-systems of the type H2O-oil-nonionic amphiphile

(microemulsions). Angew. Chem. Int. Edn 24, 654–668 (1985).3. Schubert, K. V. & Kaler, E. W. Nonionic microemulsions. Ber. Bunsenges. Phys. Chem. Chem. Phys.

100, 190–205 (1996).4. Strey, R. Phase behavior and interfacial curvature in water-oil-surfactant systems. Curr. Opin. Colloid

Interface Sci. 1, 402–410 (1996).5. Teubner, M. & Strey, R. Origin of the scattering peak in microemulsions. J. Chem. Phys. 87,

3195–3200 (1987).6. Sottmann, T. et al. A small-angle neutron scattering study of nonionic surfactant molecules at the

water-oil interface: Area per molecule, microemulsion domain size, and rigidity. J. Chem. Phys. 106,6483–6491 (1997).

7. Gompper, G. et al. Measuring bending rigidity and spatial renormalization in bicontinuousmicroemulsions. Europhys. Lett. 56, 683–689 (2001).

8. Endo, H. et al. Effect of amphiphilic block copolymers on the structure and phase behavior ofoil-water-surfactant mixtures. J. Chem. Phys. 115, 580–600 (2001).

9. Olsson, U. et al. Change of the structure of microemulsions with the hydrophile lipophile balance ofnonionic surfactant as revealed by NMR self-diffusion studies. J. Phys. Chem. 90, 4083–4088 (1986).

10. Macosko, C. W. Rheology: Principles, Measurement, and Applications (Wiley, New York, 1994).11. Van Krevelen, D. W. Properties of Polymers (Elsevier, Amsterdam, 1994).12. Arndt, M. et al. Length scale of cooperativity in the dynamic glass transition. Phys. Rev. Lett. 79,

2077–2080 (1997).13. Jackson, C. L. & McKenna, G. B. The melting behavior of organic materials confined in porous solids.

J. Chem. Phys. 93, 9002–9011 (1990).14. Cheng, F. et al. Preparation of a mesoporous silicon boron nitride via a non-aqueous sol-gel route.

Chem. Commun. 242–243 (2003).15. Beckett, M. A. et al. Formation of borosilicate glasses from silicon alkoxides and metaborate esters in

dry non-aqueous solvents. J. Sol-Gel Sci. Technol. 39, 95–101 (2006).16. Niederberger, M. et al. Non-aqueous routes to crystalline metal oxide nanoparticles: Formation

mechanisms and applications. Prog. Solid State Chem. 33, 59–70 (2005).

AcknowledgementsThis work was made possible through grants from Givaudan Flavors Corporation, the NationalScience Foundation (CTS#0548364) and small-angle neutron-diffractometer beam time at ArgonneNational Laboratory supported by the Department of Energy (W-31-109-ENG-38).Correspondence and requests for materials should be addressed to C.C.C.Supplementary Information accompanies this paper on www.nature.com/naturematerials.

Competing financial interestsThe authors declare no competing financial interests.

Reprints and permission information is available online at http://npg.nature.com/reprintsandpermissions/

290 nature materials VOL 6 APRIL 2007 www.nature.com/naturematerials

Untitled-1 4 10/3/07, 11:55:17 am