Oil Mobility in Hazelnut Oil-Based Oleogels Investigated by NMR

The migration of triacyl glycerides such as hazelnut oil leads to quality losses in various foods (e.g., fat bloom formation on chocolate, also named “fat ripening”). Oleogelation, i.e., dispersion of oils in a solid matrix of gelators, is thought to immobilize oils and consequently to hinder oil migration, leading to questions about the translational, but also intramolecular mobility of triacyl glycerides in the oleogels. In addition to translational mobility measured by diffusion-NMR, the molecule-intrinsic dynamics is reflected in NMR-relaxation. In this study, transverse relaxation and diffusion were explored to obtain insight into the condition of the oils in the disperse materials. Oleogels based on sunflower seed wax are compared to oleogels based on mono- and diglycerides. In both types of oleogels NMR-measures depend on composition as well as on temperature. Studying both dimensions, concentration and temperature, reveals the restricted mobility of oil molecules in the oleogels.

Oleogelation could be a way to reduce oil migration by structuring the edible oils [15].A three-dimensional network by organogelators (e.g., waxes, mono-and diglycerides or phospholipids) forms, in which the liquid oil is enclosed.Thus, mobile oils are "solidified" without hydrogenation or other chemical modifications [16][17][18][19].The mechanism and extent of gelation depends on the gelator.Organogelators form stable gels even at low concentrations ranging between 1 and 4% ww −1 [16].Wax from sunflower seeds (SFSW) [20,21] or rice-bran wax [22] form needle-like structures, while mono-and diglyceride (MDG) crystals show rosette-like structures [23].These structures are characterized on diverse length scales, which becomes also evident in the diffusion experiments of this paper.Typical orders of the gelators clusters include the nanostructure of crystalline nanoplates which assemble into colloidal networks in the range of 1-10 μm up to aggregates in the order of 100 μm [12,23,24].
Migration occurs for mobile fats and oils, while immobilized, bounded and partially crystalline fats do not contribute in a first approach, they are thought to hinder other mobile components in migrating [3].Temperature has a major influence: below the glass transition temperature of the oils and fats, the molecular mobility is small [25,26].Above the melting temperature, the thermal energy of the molecules is larger compared to van der Waals forces [26].Accordingly, the low melting point of most vegetable oils such as sunflower oil or HNO corresponds to high mobility [27].
However, the term 'oil mobility' needs a definition [26].'Molecular mobility' is one aspect of the described migration of molecules, often caused by a concentration gradient in a first attempt given by Fick's second law.Brownian motion is due to thermal energy.An important contribution to oil migration is, therefore, translational diffusion.Conventional methods for measuring oil mobility, and synonymously oil binding capacity, like centrifugation or draining on filter paper are not standardized and are performed on soft or molten masses.Extraction and subsequent chromatographic or gravimetric determination of free fat are time consuming.Moreover, these methods do not provide information on the type and extent of oil binding [6,28].In the present context, oil binding is understood as change of the oils molecular mobility-either in translational or rotational dynamics-as reflected in diffusion and transverse NMR-relaxation.

3
Oil Mobility in Hazelnut Oil-Based Oleogels Investigated… NMR-transverse relaxation and diffusion are known for their potential to measure intramolecular mobility as well as translational mobility.Both methods were, therefore, applied to elucidate oil mobility on two different length and time scales and to gain deeper understanding about the immobilization of fats and oils in oleogels.The typical length scale in transverse relaxation concerns the Å to nm region, while the time scale is on the ms range according to the echo time τ e .In the case of PFG-STE (pulsed field gradient-stimulated echo), the time scale is crudely given by the diffusion time Δ, the length scale corresponds to the root-mean square displacement z of the molecules which in the present case of oils is in the order of several µm on the given time scale.Studies on mobility of free and bound water exist, oils and fats were less investigated in that respect [29,30].Temperature-dependent NMR ( 1 H Larmor frequency 400 MHz) was thus used to study HNO in SFSW and MDG oleogels.FA composition, melting-and crystallization behavior, viscosity and the oil binding capacity (OBC) were determined to complete the study and are summarized in the Supplementary Information.

Sample Preparation
Oleogels were prepared by direct dispersion of the gelator in HNO with concentrations of 5, 7.5 and 10% ww −1 SFSW or MDG (shortened to e.g., HNO + 5% SFSW).The mixture was heated to 358 K with a temperature rate of 1.33 K•min −1 in a temperature-controlled water bath (Julabo GmbH, Seelbach, Germany) while constantly shearing with a speed of 200 min −1 by a stirring bar (PHOENIX Instrument, Garbsen, Germany).The dispersion was cooled down to 338 K with a rate of 0.75 Kmin −1 using the same equipment.After the mixture reached room temperature without further stirring, it was stored at 277 K for 24 h to allow the oleogel network to form.A detailed characterization is summarized in the Supplementary Information.

Instrumentation and NMR Experiments
NMR experiments were performed on a 400 MHz spectrometer (Avance Neo WB ultrashield, Bruker BioSpin, Germany) with the software Topspin 4.1.1.The instrument was equipped with a 5-mm DiffBB gradient probe.Samples were tempered by a Variable Temperature Unit (BVT, Bruker BioSpin, Germany).Temperature calibration [31] was performed to correlate the sample temperature T and the BVT temperature.T was measured via the difference in the chemical shift between the -OH and the -CH 2 -1 H lines of ethylene glycol (ROTIPURAN ® , Carl Roth, Karlsruhe, Germany, purity ≥ 99.5%) [31].
Oleogels and their constituents were measured in 5 mm NMR sample tubes with a filling height of 1 cm to reduce convection especially at elevated temperatures.T was varied in 5 K steps between 298 and 358 K.A holding time of 15 min preceded each NMR measurement.The following sequence of experiments was recorded: 1 H-FID (free induction decay) for spectroscopy; PFG-STE NMR for diffusion [32] and CPMG (multi echo sequence according to Carr, Purcell, Meiboom, and Gill [33]) for transverse relaxation.The NMR parameters are listed in Table 1.
In a simple liquid, the PFG-STE magnetization decay of the liquid part follows the Stejskal-Tanner equation [34]: The signal amplitude S at the gradient amplitude g = 0 Tm −1 is given by S 0 , D is the diffusion coefficient, Δ the diffusion time, and δ the duration of the gradient pulse.q 2 is defined here as an abbreviation of the product γ 2 g 2 δ 2 , with the gyromagnetic ratio γ.This description allows the identification of diffusion timedependent effects.Please note that two delays in the stimulated echo sequence require sufficiently long transverse as well as longitudinal relaxation times leading to the fact that only the liquid part of the samples was observed.
Following Bloch equations in the case of "simple" spin systems, the signal decays are described by an exponential decay function characterized by the transverse relaxation rate R 2 .Please note that only the liquid part of the dispersed phase was measured in the CPMG experiments due to large R 2 in solids.In the case of 1 H NMR on macromolecules, R 2 reflects mainly fluctuating homonuclear Oil Mobility in Hazelnut Oil-Based Oleogels Investigated… dipolar interactions, which are caused by intramolecular mobility, for example, given by the rotational degrees of freedom of -CH 2 -groups.

Data Analysis
Oleogels cannot be regarded as "simple" spin systems, which is already evident from their chemical composition and additionally motivated by the state of aggregation of the different components.Thus, signal decays in both, diffusion experiments and relaxation measurements differ from the "simple" modelling sketched above.A variety of modelling approaches of magnetization decays is known like sums of exponential decays, the inverse Laplace transform as well as the discrete distributions of both, R 2 and D. The gamma distribution approach uses discrete, mathematically well described distributions and was applied for multi-component mixtures with molecular weight distributions like oils, gelators and oleogels [35,36].A bimodal distribution is needed to describe the PFG-NMR magnetization decays numerically well at temperatures below the melting temperatures T < T M .Transverse magnetization decays were also analyzed via the bimodal gamma distribution model which was established for HNO (Supplementary Information).Reasons for the bimodal distributions are the large variety of molecules with their individual molecular weight distributions and the hinderance and restriction of mobility at the high concentrations in these dispersed systems.All these facts lead to distributions of correlation times and root-mean square displacements of 1 H spin bearing functional groups.The signal decays of PFG-STE and CPMG experiments were modeled within Origin 2021 (OriginLab Corporation, Northampton, United States) using the Levenberg-Marquardt iteration algorithm.The coefficient of determination R 2 was larger than 99%.At melting and crystallization temperatures T M and T C , the microstructure in the oleogels changes.Below T M and T C , diffusion described by D eff (Δ) depends on the diffusion time Δ due to geometric hindrance and interactions of the oil molecules with the gelators.The tortuosity model [37][38][39] was applied similar to the findings in hydrogels [40] and in lubricating greases [41,42].The current conception is that the effective guest molecule diffusion depends on Δ due to the gelator networks resulting in an effective tortuosity τ tort > 1.A second parameter in the model is the surface-to-volume ratio SV −1 [37].SV −1 refers to the structure of the gelator.The model's parameter could be explored to improve the stability, texture and shelf life of gelator-containing products.A quantitative impression is given by additionally calculate the effective root-mean square displacement z of the oil molecules, which depends on ∆ and D eff (Δ):

Results and Discussion
Attempts to diminish fat bloom and quality losses in chocolate explore geometric restriction and hinderance of especially HNO in dispersed systems such as oleogels [43].By combining NMR relaxation and diffusion, complementary information is obtained about binding and mobility of HNO in the oleogels: While R 2 is determined by low frequency fluctuations of mainly the dipolar interactions within a molecule, diffusion provides insight into the translational mobility.)- Oil Mobility in Hazelnut Oil-Based Oleogels Investigated…

1 H-NMR Spectra of the Constituents of the Oleogels: HNO and Gelators
The temperature-dependent structuring and influence of the gelator was detected via 1 H-NMR spectroscopy (Fig. 1).Melting, solidification, and structural stability of compositions can be quantitatively observed. 1H spectra of HNO (Fig. 1a) at room temperature were assigned in accordance with [44][45][46] (Table 2).Oleic acid was identified as main component by gas chromatography (GC) (Table SI 1, Supplementary Information) as well as by 1 H-NMR: 1 H bound to allylic carbons (peak 4) and the olefinic 1 H (peak 9).The two gelators SFSW and MDG are organic solids where mainly dipolar interactions lead to line broadening.T > 353 K, above T M , was used for the spectra in Fig. 1.Unique, characteristic 1 H lines are in the chemical shift range of 3.8-4.0ppm for SFSW (Fig. 1b) and in the range of 3.0-5.5 ppm for MDG (Fig. 1c).Lines at 3.3-5.3ppm were attributed to glyceryl [47] and other alcohols.
All 1 H spectra are dominated by the signature of -CH 2 -groups with the highest signal intensity of about 60% (Fig. 1, peak 2).Transverse relaxation and diffusion were measured for this dominant line to study oil mobility in oleogels.

Oleogels
Oil mobility is expected to be sensitive to temperature, but also to the type of gelator.The gelators in the oleogels will not significantly contribute to the measured liquids signal decays due to the small S 0 or fast transverse relaxation at low temperatures (Supplementary Information).The concentration compared to HNO is also small.MDG oleogels are thereby expected to melt at lower T than SFSW oleogels (Differential scanning calorimeter (DSC), Fig. SI 2, Supplementary Information).

Transverse Relaxation in Oleogels
HNO is expected to interact with the gelator in oleogels with consequences for R 2 .
For example, R 2 of the liquid oil phase in a fat matrix was shown to be influenced by surface relaxation caused by crystal surfaces [48].The gelators phase transition should thus be observable in R 2 .
The transverse signal decays of -CH 2 -groups in the oleogels (Fig. 2) significantly differ from those of HNO (Fig. SI 4 and SI 5, Supplementary Information).They were modelled with the bimodal gamma distribution model (Figs. 2, 3).The mean relaxation rates of both types of oleogels are larger than those of HNO at As expected, the intramolecular mobility is reduced in all oleogels compared to HNO when the gelator is in its solid form.The solid-liquid phase transition shifts to lower T with increasing gelator concentration.In addition, the phase transition occurs at lower T for MDG compared to SFSW oleogels in agreement with the expectations (DSC, Fig. SI 2, Supplementary Information).
At T < T M , differences between the two types of oleogels are obvious: While < R 2,i > of MDG oleogels monotonically decrease with increasing T (T Table 2 1 H chemical shift assignments of the main components in the oleogels: HNO, SFSW and MDG "x" marks the presence of the functional group in the spectrum of the respective species.A distinction is made between bold (X), large (X) and small (x) representing the proportion of the functional group in the spectra Peak Oil Mobility in Hazelnut Oil-Based Oleogels Investigated… ∈ [298, 333] K, T < T M ), < R 2,i > of HNO + 10% SFSW are approximately constant up to T = 328 K (T < T M ).< R 2,1 > ∈ [11-14] s −1 for the MDG oleogel are larger than that for the SFSW oleogel with < R 2,1 > ∈ [10, 12] s −1 at T ∈ [298, 308] K, which suggests a stronger interaction in the nearest neighbourhood of gelator and oil molecules.MDG oleogels are thus more temperature sensitive than SFSW oleogels.At T > T M , relaxation rates are comparable to those of HNO (oleogels: < R 2,1 > ∈ [7, 9] s −1 and < R 2,2 > ∈ [2, 4] s −1 ) due to the melting of the gelators and consequently the disassembling of the gel network.

Diffusion in Oleogels
The solid phase of the gelators is considered to geometrically hinder HNO diffusion as it was early described in [49].Oil diffusion in fat crystal matrices was found to be influenced by the pore space in a fat network [48].Other parameters are the tortuosity τ tort of the network, the capillary flow, the diffusion of the liquid oil, the oil viscosity and the contact area between liquid and solid [48].Different length scales must be considered: Clusters and particles on the length scale of several 10 μm form aggregates in the order of 100 μm [24] apart from the mentioned nanostructures.
The aggregates concentration and their distribution as well as the pore size distribution determine the effective oil diffusion.PFG-NMR was thus measured on oleogels  with 5%, 7.5% and 10% ww −1 SFSW (Fig. 4a) and MDG (Fig. 4b) at T = 298 K and Δ = 120 ms.The PFG-signal decay of HNO in the oleogels is less pronounced than that of pure HNO (Fig. 4).Molecular diffusion seems to be hindered and is sensitive to the gelator's type and concentration: HNO diffusion in MDG oleogels is faster than that in SFSW oleogels.When choosing a suitable model for quantification, it should be considered that the shape of the oleogels' signal decay significantly changes.This is obvious in its T-dependence due to the gelators' solid-liquid phase transition (Fig. 5).
Diffusion data are described by bimodal distributions at T < T M .The form of the signal decay is caused by hindered diffusion and molecular diversity.Interactions on the molecular length scale cannot be excluded which lead to diminished diffusion.The simple sum of diffusion distributions of HNO and gelators as well as a monomodal distribution are numerically not sufficient.Two independent contributions Fig. 6 Exemplarily for PFG-signal decays in oleogels together with the modelled distributions: Normalised -CH 2 -magnetization decays of HNO + 10% SFSW ( ) at a T = 298 K (T < T M ): The monomodal gamma distribution describes the data not sufficiently well, decays were calculated with different distribution widths (dotted lines), the data are better described by a bimodal gamma distribution (solid line).b T = 338 K (T > T M ): The monomodal gamma distribution (dotted line) describes the data well with an error in the order of 10 -3 need to be assumed (Fig. 6a).Only at T > T M , modelling by a monomodal gamma distribution (Fig. 6b) is sufficient.< D eff,1 > is smaller than < D eff > of pure HNO in all oleogels.HNO diffusion is more reduced in SFSW (< D eff,1 > ∈ [8.96•10 -12 , 9.51•10 -12 ] m 2 s −1 and A 1 ≈ 0.90) than in MDG oleogels at the same gelator concentration (< D eff,1 > ∈ [9.79•10 -12 , 1.06•10 -11 ] m 2 s −1 and A 1 ≈ 0.96).Moreover, < D eff,1 > decreases with gelator concentration as expected.Concerning the second component in the oleogels: < D eff,2 > ∈ [3.17•10 -12 , 4.95•10 -12 ] m 2 s −1 is larger than < D eff > of the gelators.< D eff,2 > may be attributed to HNO in direct interaction to gelator structures.This fraction of the total magnetisation decay is small with A 2 ≈ 0.1 for SFSW oleogels and A 2 ≈ 0.04 for MDG oleogels.
The extend of geometric hindrance of diffusion depends on the physical dimensions in the gelators' network.Scanning electron microscope (SEM) images and polarisation microscope images [23,43,50] indicate the diverse typical length scales of the solid oleogels which are distributed over a wide range, but significant structures also are in the range of 50 μm for MDG and SFSW, a significantly larger dimension than the calculated z.However, the multiscalar structure of oleogelators needs to be considered making a direct interpretation of z in terms of interaction probability and geometric hindrance length difficult.
Temperature T is an important factor for HNO diffusion in oleogels which is exemplified on the oleogel HNO + 10% ww −1 SFSW (Fig. 7): < D eff,1 > in the oleogel is smaller than that for pure HNO at T < T M .< D eff,2 > (A 2 ≈ 0.1) in the oleogel is comparable to < D eff > of the gelator SFSW for T ∈ [298, 333] K.At T > T M , < D eff > of all substances show similar values when modelled with a Table 3 Mean effective diffusion coefficients, distribution widths, amplitudes and root-mean square displacement of HNO, SFSW, MDG and oleogels at Oil Mobility in Hazelnut Oil-Based Oleogels Investigated… monomodal gamma distribution.Summarizing, diffusion of HNO is reduced by the gelator, but not completely restricted as in droplets of emulsions [51,52].< D eff,1 > in oleogels depends not only on T, but also on Δ (Fig. 8).The gelator represents a geometric hindrance for HNO molecules, so that diffusion is not free, but hindered.Motivated by the structural similarity to structures such as to hydrogels [40] and lubricant greases [41], the diffusion of oil molecules is interpreted as a function of Δ via the tortuosity model [37] (Fig. 8).Often, the length scales of the root-mean square displacement of the molecules are in the order of the restricting geometry (in [37] dz −1 < 5, while d is the sphere's diameter).The restricting geometry in the oleogels is, however, multi-facetted: Taking a closer look to the images, for example in [23], structures on the small length scales of 1-5 µm are visible apart from the aggregates dimensions in the order of 50 µm and more.While knowing about the fact that the diffusing moieties should emphasis the restricting geometry, the experimental findings in the images and of the diffusion experiments  show the applicability of the tortuosity model at least as a first attempt to describe the data and being able to compare the measured data on the different oleogels reliably.The tortuosity model [37,38] results in τ tort = 3.9 and in a surface-to-volume ratio SV −1 = 1.9 µm −1 , exemplarily for the oleogel with 10% SFSW, and τ tort = 2.6, SV −1 = 1.5 µm −1 for the MDG oleogel (10%) at T = 298 K.For further interpretation, we assume a geometric diffusion hinderance of the oil molecules by spherical particles.A value SV −1 = 2 µm −1 would then correspond to a diameter of a sphere of 1.5 µm.Please note the different length scales in the solid state of oleogels [43] which comprise also the scale of a few µm, while the large structures of mainly agglomerates are in the order of 50 µm.The geometric hindrance reflected in τ tort is larger in SFSW oleogels than in MDG oleogels while SV −1 is in the same order of magnitude.Compared to greases and polysaccharide hydrogels, τ tort in oleogels is significantly larger [40][41][42] and in a similar order of magnitude than observed in porous rocks (τ tort ∈ [3.2, 4.4]) [39].
The parameters τ tort and SV −1 change significantly with T and depend on the oleogels composition.SV −1 usually depends on the microscopic structures and sizes.In general, the higher the density the lower SV −1 .SV −1 is an important parameter and enters the thermodynamic and kinetic descriptions, including that of melting and boiling points, solubility and reactivity and in consequence the applicability range of a dispersed system.

Fig. 1 a
Fig. 1 a 1 H-NMR spectra of HNO (black), SFSW (red) and MDG (blue) at T > 353 K.All spectra are dominated by peak 2 attributed to -CH 2 -groups.The zooms show the characteristic 1 H lines of the gelators b SFSW and c MDG. 1 H signals at 3.3-5.3ppm are attributed to glyceryl and other alcohols

5 Fig. 8
Fig.8<D eff,1 > and σ 1 (vertical lines) of the -CH 2 -peak of a HNO + 10% SFSW and b HNO + 10% MDG as a function of Δ. T < T M : The data were modelled by the tortuosity model (dotted lines), revealing τ tort (T) and SV −1 (T) in a first attempt.Above T M (red points, line: guide to the eye), < D eff,1 > increases with Δ due to convection

Table 1
Acquisition parameters of the 1 H-NMR experiments