1 Introduction

Lard with crispy pieces of skin was the premier bread spread before being all but replaced by butter and margarine in the mid-1900s. The product is known as Griebenschmalz in German, smalec in Polish, and fedt in Danish. Global ambitions of energy conservation and resource utilization have led to a revival of lard as a raw material for food products [1]. Consumer satisfaction depends not only on the taste, but also critically relies on the mouthfeel, which is largely determined by the gradual transitions between various solid and liquid fat phases during biting, chewing, and swallowing. While an individual batch of lard may comprise triglycerides with acyl chain lengths or degrees of unsaturation yielding suboptimal melting behavior, a product with improved customer appeal could be achieved by blending multiple batches as well as including minor amounts of additives that stabilize the melt or certain crystal forms.

For a pure saturated triglyceride system, crystallization from the melt takes place by formation of the metastable pseudohexagonal α-form and subsequently transition to the stable triclinic β-form, sometimes via the orthorhombic β′-form [2, 3]. Simple mixed triglycerides often feature the same general sequence of events with the additional mechanism of fractionation of molecules with different chain lengths in the liquid → α and α → β transitions as illustrated in Fig. 1 [4]. More complex mixtures, including triglycerides with different acyl chain lengths and degrees of unsaturation connected to the same glycerol moiety, can be expected to include even more complex crystallization behavior with fractionation of molecular components between the melt and multiple (meta)stable crystal forms.

Fig. 1
figure 1

Reproduced from Ref [4] with permission from the American Chemical Society

Schematics of fat crystal forms and phase transitions induced by temperature changes in a two-component system comprising the triglycerides (triacyl glycerols, TAGs) trilaurin (LLL) and trimyristin (MMM) with short and long hydrocarbon chains, respectively. a High-temperature liquid mixture. b Temperature decrease: liquid–solid co-existence with partial fractionation of the long-chain triglycerides into α-form crystals. c Additional temperature decrease: fully crystallized system with multiple crystal forms having varying proportions of short- and long-chain triglycerides. d Storage or annealing: partial recrystallization to thermodynamically stable β-form crystals with single types of triglycerides. e Heating: co-existence of short-chain liquid and long-chain β-form crystals.

While X-ray diffraction gives a wealth of information about crystal symmetries, repeat distances, and, sometimes, atom positions within the unit cell in triglyceride system [3, 5,6,7], it faces great challenges in characterizing the liquid and the less-ordered solid phases in the multicomponent fat mixtures of relevance for the food industry. Likewise, low-field 1H NMR is a rapid and accurate method for quantifying the ratio between solid and liquid fat [8] but is confounded by the presence of other organic solids, such as pork cracklings, in addition to the fat crystals.

With the aim of enabling rational blending of lard-based spreads, we here apply 13C solid-state NMR with dynamics-based spectral editing [9,10,11] for detailed characterization of thermotropic phase transitions in store-bought and home-made lard products. The method, illustrated in Fig. 2, relies on magic-angle spinning (MAS) [12] and high-power 1H decoupling to give 13C spectra with sufficient resolution to identify multiple aliphatic, olefinic, and allylic carbons in the hydrocarbon chains [13, 14] even in the presence of skin proteins [15]. Spectral editing by cross-polarization (CP) [16] and insensitive nuclei enhanced by polarization transfer (INEPT) [17], with interpretation of signal intensities within a two-step model of C–H bond reorientation [18, 19] under moderate MAS, allows identification of molecular segments having combinations of correlation times τc and order parameters SCH that are classified as solid, intermediate, liquid, and liquid crystalline dynamics [20, 21].

Fig. 2
figure 2

Adapted from Ref [11] with permission from the authors

Principles of 13C MAS NMR spectroscopy with dynamics-based spectral filtering. a Diagrams of the pulse sequences DP, ramped CP, and refocused INEPT, wherein the 13C signal is detected during gated 1H decoupling and continuous sample spinning at the frequency ωMAS. Narrow and broad vertical lines symbolize 90º and 180º RF pulses applied at the 1H and 13C Larmor frequencies ω0H and ω0C. CP is performed by simultaneous application of 1H and 13C RF pulses of duration τCP and with amplitudes giving nutation frequencies ω1H and ω1C covering the matching conditions |ω1H – ω1C|/ωMAS = 1 and/or 2. In the INEPT sequence, conversions from in-phase to anti-phase 1H and anti-phase to in-phase 13C coherences occur during the delays τ1 and τ2. Longitudinal relaxation takes place during the recycle delay τR (not shown) between the end of signal detection and the following 90º pulse. b Correlation function g(τ) and spectral density j(ω) for the two-step C-H bond reorientation model with MAS wherein g(τ) decays in two exponential steps with the time constant τc and τs separated by an intermediate plateau given by the C–H bond order parameter SCH. Sample spinning yields modulations of g(τ) at the frequencies ωMAS and 2ωMAS. c Theoretical CP (blue) and INEPT (red) signal enhancement as a function of τc and |SCH| at constant τs = 1 ms calculated for a CH2 segment using the experimental parameters ω0C/2π = 125 MHz, ωMAS/2π = 5 kHz, τCP = 1 ms, ω1H/2π = 100 kHz, (ω1Hω1C)/ωMAS ramped from 2.5 to 0.5, τ1 = 1.8 ms, and τ2 = 1.2 ms.

2 Materials and Methods

Schweineschmalz and griebenschmalz were bought in a local Swabian butcher shop in southwestern Germany. The smalec produced by the company Skwierzyna was bought from a Polish food store in Gothenburg, Sweden. The Danish fedt was homemade from two different types of salami (pepper and pepperoni) manufactured by the Danish producer GØL and bought in a supermarket in Lund, Sweden. Salami pieces were fried in a pan until being very crispy and the fat was collected and allowed to solidify at room temperature. All products were used without further purification and stored for weeks at + 4 ºC before being prepared for NMR. The samples were placed in disposable HRMAS rotor inserts and transferred to 4 mm zirconia rotors (both from Bruker). The inserts are made from polychlorotrifluoroethylene thermoplastic (trade name: Kel-F) and have a sealed sample volume of approximately 40 μL.

NMR experiments were performed on a Bruker Avance Neo spectrometer operating at the Larmor frequencies ω0C/2π = 125 MHz and ω0H/2π = 500 MHz for 13C and 1H, respectively. The 13C signal was recorded on a 4 mm HX CPMAS probe at 5 kHz MAS using 400 ppm spectral width, 62 kHz SPINAL-64 1H decoupling [22], 50 ms acquisition time, accumulation of 256 transients, and τR = 5 s with the DP, ramped CP [16, 23], and refocused INEPT [17, 24] sequences displayed in Fig. 2. Hard radiofrequency (RF) pulses were applied at 80 kHz nutation frequencies for both 13C and 1H. CP was performed with the nutation frequencies ω1C/2π = 60 kHz and ω1H/2π ramped from 80 to 40 kHz during the contact time τCP = 1 ms. The delays τ1 = 1.8 ms, and τ2 = 1.2 ms were employed for INEPT.

The sample temperature was calibrated by 1H spectra for an external reference of methanol [25] at 5 kHz MAS. After an initial equilibration of the fat samples at – 8 ºC for 1 h, data was acquired as a function of temperature in increments of 5 ºC with 5 min equilibration after each temperature change before starting the sequence of DP, CP, and INEPT measurements lasting ~ 1 h.

Data was processed in Topspin 4.0 using zero-filling from 5000 to 16,384 points, exponential apodization yielding 20 Hz line broadening, Fourier transformation, and manual phase correction. The 13C chemical shift scale was internally referenced to 14.0 ppm for the terminal CH3 in the 32 ºC INEPT spectrum [26]. INEPT peak areas were estimated by deconvolution using the ‘Lorentzian–Gaussian’ shape option in MestReNova (Version 14.2.0) after baseline correction using the ‘Whittaker Smoother’ option.

3 Results and Discussion

Figure 3 shows 13C NMR spectra of griebenschmalz at 2 ºC where the comparable amplitudes of the CP and INEPT resonance lines indicate that the sample contains similar amounts of solids and liquids. According to the theoretical signal intensities in Fig. 2c, INEPT is observed for C–H bonds with reorientational correlation time τc smaller than ~ 10 ns and order parameter |SCH| lower than ~ 0.2, while CP is efficient in roughly the complementary area of the 2D τc-|SCH| space. Exceptions to this simple rule for distinguishing between solids and liquids are the intermediate (τc ≈ 1 μs and |SCH|< 0.95) and liquid crystalline (τc < 10 ns and |SCH| ≈ 0.1) regimes where either none or both of CP and INEPT are efficient [21].

Fig. 3
figure 3

13C MAS NMR data of griebenschmalz at + 2 ºC using dynamics-based spectral editing to highlight the co-existence of solids and liquids. The data was acquired at ω0C/2π = 125 MHz, ωMAS/2π = 5 kHz, 62 kHz SPINAL-64 1H decoupling, and τR = 5 s with the DP (gray), CP (blue), and INEPT (red) sequences in Fig. 2a using a ~ 40 μL sample of a fat-embedded pork crackling taken from the commercial product in the photo to the bottom left. CP was performed with τCP = 1 ms, ω1C/2π = 60 kHz, and ω1H/2π ramped from 80 to 40 kHz, while τ1 = 1.8 ms, and τ2 = 1.2 ms were used for INEPT. Peak assignments refer to the labeled carbon atoms in the chemical structure drawings of the generic amino acid residue in the protein and a representative fat molecule with linoleic, palmitic, and oleic acyl chains attached via ester bonds to the glycerol moiety. The labels at ~ 30 and ~ 32 ppm point out the distinct chemical shifts of interior methylene segments undergoing trans/gauche conformational isomerization or adopting a straight all-trans configuration as illustrated in Fig. 1e. The top spectra are magnified × 10 compared to the bottom ones. Stars indicate spinning sidebands and the label at 105 ppm shows the maximum of a broad DP resonance from the Kel-F material of the rotor insert

Based on literature data [14, 27], the resonance lines in Fig. 3 were associated to carbon atoms representative of the ends (ω, ω – 1, and ω – 2) and beginnings (1, 2, and 3) of the acyl chains, the glycerol moiety (g1, g2, g3), and acyl chains with double bonds (o1, o2, l1, l2, l3, and l4), including the special case of the two non-conjugated double bonds (l3 and l4) typical for linoleic acid [14]. Since all spectra are acquired with the same spectral width, acquisition time, MAS frequency, and high-power 1H decoupling, as well as processed with the same moderate apodization, the larger linewidths in CP mainly reflect the spread of isotropic chemical shifts originating from a distribution of non-interchanging conformations and packings of otherwise chemically identical molecular segments in the poorly crystalline solids [28, 29]. Conversely, conformational isomerization at rates far exceeding the intrinsic chemical shift-induced frequency differences, for instance ~ 5 ppm × 125 MHz = 625 Hz for the trans and gauche conformations of the saturated hydrocarbon chains [30, 31], results in narrow INEPT peaks with resolution limited by the 50 ms acquisition time and the 20 Hz line broadening required to avoid truncation artifacts. The chemical shift difference between the interior methylene peaks for the solid (all-trans) and liquid (trans/gauche) fat mainly reflects the fractional populations of the various conformational isomers in the two aggregation states [31].

Differences in melting behavior of the fat blends are apparent in the variable temperature data in Fig. 4. The overall balance between solid and liquid fat is estimated by comparison of the all-trans CP and trans/gauche INEPT peaks at ~ 32 and ~ 30 ppm, respectively. At the lowest investigated temperature – 8 ºC, all samples are dominated by solid fat, but also contain a substantial liquid fraction increasing in the order schweineschmalz < griebenschmalz ≈ smalec < fedt. With increasing temperature, solid fat is gradually replaced with liquid while preserving the relative order between the samples. All samples contain both solid and liquid fat over a temperature interval spanning more than 40 ºC consistent with a mixture of multiple components with individual transition temperatures determined by their acyl chain lengths and degrees of unsaturation. Closer inspection of the CP and INEPT resonance lines shows differences in chemical shift and linewidth for all types of carbons, including o2/l2, l3, and ω, implying that all types of acyl chains occur in both solid and liquid states. Conversely, none of the samples feature the characteristic peaks of anisotropic liquid crystals—CP and INEPT with similar amplitudes at identical chemical shifts and lineshapes [20]—as often observed for phospholipid bilayers with molten acyl chains [32,33,34,35,36,37,38]. The absence of these distinctive pairs of CP and INEPT peaks at all investigated temperatures contradicts the sometimes suggested lamellar order in liquid fat just above the melting point [2, 3].

Fig. 4
figure 4

Variable temperature 13C MAS NMR data for lard-based spreads with zoom-in on chemical shift ranges dominated by a aliphatic, b olefinic, c carbonyl, and d glycerol carbons. For each shift range, the spectra are organized in columns for the samples schweineschmalz, griebenschmalz, smalec, and fedt, as well as rows for the selected temperatures –8, 7, 17, and 32 ºC. The DP, CP, and INEPT spectra are jointly scaled to the same maximum intensity for each combination of sample, temperature, and shift range. Peak assignments refer to the carbon labels in Fig. 3 except for X1 and X2, indicating aliphatic and carbonyl peaks originating from different crystal forms, and i from non-triglyceride impurities

More detailed inspection of the temperature dependence of the CP resonances from all-trans methylene at ~ 32 ppm and carbonyl at ~ 173 ppm in Fig. 4 reveals at least two sets of peaks tentatively originating from crystal forms analogous to the low-melting α- and high-melting β-forms found in pure triglyceride systems [7]. While the ~ 1 ppm higher shift of the all-trans methylene for the high-melting form agrees with literature data for the α/β′- and β-forms of tripalmitin [13], and corresponds to the differences observed for the interior methylene carbons in triclinic and other crystal forms of all-trans alkanes [39], the carbonyl shifts are not consistent with the conventional α- (171.9 ppm), β′- (172.5 and 173.3 ppm), or β-forms (172.9, 173.4, and 174.6 ppm) [13, 40], indicating that the molecular conformations and crystal structures in the present fat blends differ from those found in pure triglycerides. Although X-ray diffraction of complex fat blends is often interpreted using crystal form terminology developed for simpler triglyceride systems [41, 42], the data in Fig. 4 suggest that further solid-state NMR studies may be useful to elucidate the details of the structures and possibly identify sub-groups of crystal forms. Contrary to the temperature-sensitivity of the fat, the CP resonances from the protein remain nearly unaffected by the temperature changes as previously observed for skin proteins at dry conditions [15]. Although the general trends of the melting of the solid phases can be estimated from the CP spectra in Fig. 4, actual quantification of the relative proportions of the phases would require recycle delays on the order of a minute to compensate for the unusually slow 1H longitudinal relaxation for the β-form of long-chain saturated triglycerides [40], as well as specialized pulse sequences being less sensitive to differences in relaxation and buildup rates during the CP block [43].

In Fig. 5, the temperature dependence of the composition of the liquid phase can be gleaned from the INEPT peak areas normalized by the value for the terminal methyl groups (ω), which can be expected to melt at lower temperature than any of the other parts of the fat molecules—possibly even being present as liquid-like end-group regions in otherwise solid fat crystals [3]. While the normalized area for the penultimate methylene (ω – 1) remains nearly constant at unity, the values for the interior methylene carbons increase and the olefinic (o2, l2, l3) and allylic (o1, l1, l4) carbons decrease with temperature, showing that the low-temperature fluid is enriched in fat components with short acyl chains and high degree of unsaturation. At the highest temperatures, where all fat is liquid, the peak areas report on the overall chemical composition. The lowest degree of unsaturation is observed for schweineschmalz, which also has the lowest fraction of liquid at low temperature. Conversely, the higher fraction of liquid for fedt may not be attributed to any differences in acyl chain length or unsaturation, but rather to the thermodynamic mechanism of freezing-point depression induced by the presence of low molecular weight impurities labeled in Fig. 4d. More accurate quantification of the liquid composition could be achieved by systematic variation of the τ1 and τ2 delays in the INEPT sequence to account for the dynamics, scalar couplings, and multiplicities of the different carbons at the expense of an order of magnitude longer measurement times [44].

Fig. 5
figure 5

Temperature dependence of the area of selected INEPT peaks relative to the value for the terminal methyl group ω. The carbon atom labels are explained in Fig. 3 and approximate chemical shifts are given in parenthesis. All panels are shown with the same temperature axis

4 Conclusions

Solid-state NMR with dynamics-based spectral filtering enables detailed characterization of co-existing solid and liquid phases in lard-based spreads over wide temperature ranges, showing enrichment of fat components with shorter acyl chains and higher degrees of unsaturation in the low-temperature liquid, the higher melting points of crystal forms with more efficient acyl chain packing, the freezing-point depression induced by low molecular weight impurities, and the minimal effects of temperature on pork cracklings embedded in the fat. These rich details may prove useful for the development of lard products with optimal balance between solid and liquid fractions to achieve favorable texture properties and consumer appeal over an extended range of serving temperatures.