Role of the Drying Technique on the Low-Acyl Gellan Gum Gel Structure: Molecular and Macroscopic Investigations
- 482 Downloads
The effect of three drying processes (freeze, oven and supercritical CO2 drying) on CP Kelco low-acyl gellan gum gel was investigated, highlighting the role of the water removal mechanism (i.e. sublimation, evaporation and solvent replacement/extraction) and the process parameters on the gel structure, rather than focusing on the drying kinetics. It is the first time that a research paper not only compares the drying methods but also discusses and investigates how the molecular and macroscopic levels of gellan gum are affected during drying. Specifically, the dried gel structures were characterised by bulk density and shrinkage analyses as well as scanning electron microscope (SEM) and micro-computed tomography (μCT) microscopy. Micro-differential scanning calorimetry (μDSC) was used in a novel way to investigate the effect of the drying technique on the polymer disorder chains by partial melting of the gel. The resulting water uptake during rehydration was influenced by the obtained dried structure and, therefore, by the employed drying process. It was found that freeze-dried (FD) structures had a fast rehydration rate, while both oven-dried (OD) and supercritical CO2-dried (scCO2D) structures were slower. After 30 min, FD samples achieved a normalised moisture content (NMC) around 0.83, whereas OD and scCO2D samples around 0.33 and 0.19, respectively. In this context, depending on the role of the specific hydrocolloid in food (i.e. gelling agent, thickener, carrier), one particular dried-gel structure could be more appropriate than another.
KeywordsDrying LA gellan gum Gel microstructure Rehydration
Dried foods have been widely produced to extend the product shelf life and to be consumed on demand (Ratti 2001; Marabi et al. 2004). In relatively complex products, such as instant foods or dairy products, the formulation may contain additives and preservatives as ingredients, which need to be considered during the drying process (de Vries 2002; Norton and Foster 2002; Renard et al. 2006). In this context, hydrocolloids are often used in food formulations to modify the product properties, acting as thickener, stabiliser or gelling agents (Phillips and Williams 2004, 2009), and to improve the dried product quality and shelf life (Brown et al. 2010).
Dried foods often need to be rehydrated before consumption (Marabi et al. 2006; Joardder et al. 2015), and the speed of this process can vary according to the specific application. In this light, both the properties and the structure of the dried gel contained in the product formulation may affect the water uptake. Thus, the rehydration rate of both the gel and, therefore, the final product can be modulated, since it is strictly dependent on the drying process (Marabi et al. 2004). Dried gel systems are often used also as carriers for active ingredient release (e.g. drugs, bioactive molecules or sugars) (Hoffman 1987; Nishinari and Fang 2016; Lin and Metters 2006; Tønnesen and Karlsen 2002) and for packaging applications (Suderman et al. 2016).
Dried gel systems are of interest not only in the food industry, but also in biomedicine, and especially in tissue engineering. The design and modulation of biopolymeric dried scaffolds with different patterns and structures may functionalise the material, providing specific mechanical and chemical responsive properties (Sachlos and Czernuszka 2003).
A common gelling agent in the food industry is low-acyl or deacylated (LA) gellan gum. It is a microbial polysaccharide presenting as a primary structure a tetrasaccharide unit composed of glucuronic acid, rhamnose and glucose (Morris et al. 2012). The molecular structure leads to specific mechanical properties that can be further engineered if it is blended to high-acyl (HA) gellan gum, known to be softer and more elastic due to the presence of acyl substituents (Morris et al. 2012; Phillips and Williams 2009). LA gellan gum properties are direct consequence of the double helix formation during gelation and the ion-induced association of the same helices, forming the junction zones (Morris et al. 2012). The resulting network consists of disordered and flexible chains with few highly ordered domains (Chandrasekaran et al. 1988). Reducing the amount of water, the gellan gum chains aggregate and the network becomes more packed (Morris et al. 2012).
Among the industrial drying techniques, freeze drying is widely used in the food industry since, being a non-thermal method, it preserves the food structure (Evans 2009; Krokida et al. 1998) and nutrients (Avila and Silva 1999). The process is based on the freezing of the product, followed by sublimation of the ice crystals into vapour at reduced pressure (Scherer 1990). Throughout the process, the capillary stress is avoided, preventing the collapse of the structure and minimising the shrinkage of the material (Krokida et al. 1998). Another common technique is air drying, which generates a compacted structure (Joardder et al. 2015; Krokida and Maroulis 1997), since local stresses within the material are produced on evaporation (Scherer 1990). This mechanism is affected by the surface tension of the liquid (Deshpande et al. 1992), which is relatively high for water. Supercritical fluid (SCF)-assisted techniques have been proposed recently in the food industry (Brown et al. 2010; Brown et al. 2008), as they are generally used for other food applications, such as extraction of active ingredients from natural resources (Peker et al. 1992), isoelectric precipitation of proteins (Hofland et al. 2000), micronization (Prosapio et al. 2016), and substrate impregnation (De Marco and Reverchon 2017). The general employed fluid is carbon dioxide (CO2), being an inert gas, with a relatively low critical point (31.1 °C and 73.8 bar). By changing both pressure and temperature, it is possible to modulate the fluid properties, such as viscosity and density (Wang 2008; Benali and Boumghar 2014; Brunner 1994). For drying applications, the hydrogel-alcogel transition is carried out, followed by solvent removal using supercritical CO2 (Ulker and Erkey 2017).
In terms of produced dried-gel gellan gum structure, the aforementioned drying techniques have already been investigated. Tiwari et al. (2015) and Silva-Correia et al. (2011) reported information about freeze-drying in gellan gum systems. However, their analyses were based only on scanning electron microscopy (SEM) observations, lacking data about porosity throughout the whole dried-gel volume. In order to have a 3D sample reconstruction, micro-computed tomography (μCT) can be a useful method to complete the freeze-dried microstructure understanding. Ratti (2001) accurately compared freeze drying with air drying, suggesting that the former is more suitable to achieve a high-quality product although it is more expensive. The effect of air drying on the material structure has been deeply explained and rationalised by Joardder et al. (2015). The drying kinetics and the effect of the different drying processes on the dried-gel macrostructure were reported by Sundaram and Durance (2008). However, in that work, locust bean gum, pectin, and starch were the hydrocolloids investigated and they were mixed to form a single gel system, which is considerably more complex and different from LA gellan gum in terms of gelation mechanism and molecular configuration. Brown et al. (2010) studied the effect of the supercritical CO2 (scCO2) drying on agar gels in comparison with both oven and freeze drying processes. However, this system was characterised only at the macroscopic scale by μCT, without providing information on the effect on the molecular and network levels. Furthermore, the collected data on agar are significant only for gelling agents with comparable gelation mechanism and molecular structure, unlike gellan gum.
In this work, for the first time, the effect of freeze, oven, and scCO2 drying on gel microstructure at both the molecular and macroscopic levels was investigated. Precisely, the role of the physical mechanism of water removal for the specific drying method was highlighted (i.e., sublimation, evaporation, and solvent replacement/extraction), rather than the drying kinetics. LA gellan gum was used as a model gelling agent in a quiescent form, yet other hydrocolloids are expected to behave similarly, especially if they present a similar gelation mechanism, based on the physical interactions of the polymer chains (Gulrez et al. 2011), such as carrageenan (Aguilera and Stanley 1999). This study proposes micro-differential scanning calorimetry (μDSC) analysis as a method to investigate the role of the drying technique on the gel network, characterising the molecular aggregation extent and structure order. In fact, although μDSC has already been used for gellan gum investigations (Sudhamani et al. 2003), it has never proposed for dried gellan gum gels to show the presence of possible changes in the dried sample thermal behaviour. After drying, the samples were rehydrated to study how the structure affects the water uptake into the material.
Materials and Methods
Low-acyl gellan gum was provided by Kelcogel F, CP Kelco, UK. After heating distilled water to 85 °C, the LA gellan gum powder at 2% weight/weight (w/w) was slowly added to avoid the formation of clumps. At complete hydration, the solution was poured into sample moulds (13.5 mm in diameter and 65 mm in height) and left to cool at room temperature (20 ± 1 °C). After setting, a maturation period at room temperature (20 ± 1 °C) for 24 h was performed. The gels were cut with a knife in samples of 10 mm in height. All the materials were used with no further treatment or purification.
The gel samples were frozen in a − 18 °C freezer (LEC U50052W, UK) for 24 h, applying a freezing rate of around 0.2 °C/min, previously measured by use of thermocouples (Cole-Parmer® Instrument Company, USA) at both sample core and surface. Afterwards, they were placed onto shelf trays at room temperature in the freeze dryer (SCANVAC 110-4 PRO, LoboGene, UK) for 48 h. The experiments were carried out in a conservative freeze-drying cycle, without providing additional heat from the shelves in order to avoid potential structure collapse. The chamber pressure was set at 0.18 mbar and the temperature of the condenser at − 110 °C, condition that is defined by the equipment. After the drying process, the samples were stored in a desiccator with silica gel beads under low vacuum conditions until characterisation.
Oven drying was performed in a vacuum oven (Fistreem International Co. Ltd., Leicestershire, UK) at 20, 40, and 60 °C under static air, room pressure and a constant relative humidity (RH) of 20%. Since oven drying kinetics depends on the temperature, the process time was set accordingly to reach a normalised moisture content value lower than 0.1, as described in the “Moisture Content and Water Activity” section.
Carbon dioxide was supplied from BOC (Guildford, UK). Before drying using scCO2, an ethanol (purity 99.9%, AnalaR NORMAPUR, VWR, UK) pre-treatment was performed to replace water, while the supercritical CO2 drying was carried out to remove the liquid ethanol from the sample and, therefore, to obtain a solid dried matrix (normalised moisture content (NMC) below 0.1). The gel samples were left, stepwise, in the alcoholic solutions at 25, 50, and 80% wt. Each step was carried out for 6 h, before using absolute ethanol for 24 h. This gradual pre-treatment was needed to reduce the shrinkage extent due to the use of ethanol.
The supercritical drying process necessary to remove ethanol was carried out in two configurations: batch and semi-continuous. In the batch configuration, alcogels, obtained from a high-pressure pipe, were placed into the vessel and then it was pressurised with CO2 and heated until the desired operating conditions were achieved. The effects of temperature (40–50 °C) and pressure (85–100 bar) were investigated. In the semi-continuous configuration, alcogels were placed in the same high-pressure vessel, which was pressurised applying a continuous CO2 flow (1–2.5 L/min) throughout the experiment using an air-driven liquid pump (MS-71, Haskel, USA). The same conditions of pressure and temperature were tested.
In both configurations, temperature was controlled by a thermostatic water bath, in which the rig is submerged. Pressure was monitored by using a manometer, while the CO2 flow rate was read by a digital mass flow meter (RHE08, Rheonik, Germany) and adjusted by opening the metering valve downstream, since steady-state conditions were applied. All the experiments in the batch configuration were carried out for 8 h, followed by a 20-min depressurisation. If the forced flow was applied, a 3-h process was performed.
Moisture Content and Water Activity
Since the gel concentration is known, and equal to 2% w/w, it is possible to estimate Ms from M0. Brown (2010) suggested a value of NMC < 0.1 as the goal to achieve to have negligible moisture content. Equation 1 was also used to monitor the rehydration process, considering Md as a function of the rehydration time. At the NMC < 0.1, water activity was measured by using the Aqualab dew point water activity meter 4te (Labcell LTD, UK). Samples were placed into the test chamber at 25 °C, after being crumbled, to analyse aw throughout the sample.
The dried gel microstructure was analysed by X-ray μCT and scanning electron microscope (Philips XL-30 FEG ESEM, Netherlands), in conjunction with the analysis of physical/geometrical properties such as bulk density, shrinkage and shape.
High-resolution micro-computed tomography was performed by using the Skyscan 1172 (Bruker, Belgium). This system allows the visualisation and a complete 3D structure reconstruction of 2D cross-sections without any chemical fixation and sample preparation. The acquisition mode can be set at a maximum current of 96 μA and voltage of 100 kV. After binarisation into black and white images, both qualitative and quantitative analyses were performed by using a CT analyser (188.8.131.52), obtaining porosity information on the whole bulk structure.
ESEM FEG (Philips XL30) was used to collect high-quality images of the dried gel structure. The samples were cut after cooling in liquid nitrogen to highlight both the vertical and horizontal cross-section. The maximum voltage was set up to 10 kV and the magnification up to ×1500.
The absolute (true) density of LA gellan gum was measured by using the AccyPyc II 1340 pycnometer (Micromeritics, USA), using helium as a displacement medium.
Once the sample volume was calculated, the bulk density of the sample was measured as the mass was known.
The water uptake was calculated by measuring the sample weight every 6 min for 30 min. A distilled water bath (100 ml) at room temperature (20 ± 1 °C) was used as a medium (Vergeldt et al. 2014). Rehydrated samples were carefully blotted before weighing to remove surface water.
A micro DSC 3 evo (Setaram Instrumentation, France) was used to investigate the thermal transitions. The sample was placed in the “close batch cell” (0.6 ± 0.1 g). The reference cell was filled with an equal mass of distilled water. Two sets of analyses were carried out on LA gellan gum gel before drying, from 5 to 80 °C and from 5 to 55 °C, applying a scan rate of 1 °C/min. The latter temperature range was used to isolate the melting of the disordered chains and avoid the “second” thermal transitions that depict the disruption of the junction zones (Picone and Cunha 2011).
The dried samples were rehydrated for 6 h in 100-mL distilled water to enhance the mobility of the polymer disorder chains and observe the thermal event on cooling. In this case, thermal cycles were applied from 5 to 55 °C.
The μDSC curves were presented as an average of the first cycles in triplicate, while the values of transition temperature, enthalpy and entropy were expressed with plus/minus a single standard deviation.
All the experiments were performed in triplicate and, for each measurement, six dried gel samples were analysed. Data were analysed by one-way analysis of variance (ANOVA) and Tukey’s multiple comparison tests, using SigmaPlot 12.5 Statistical Software. The level of significance was defined as p ≤ 0.05.
Results and Discussion
In order to highlight the effect of the drying process on gel structure, the gel formulation was kept constant throughout the experiments. The gel composition may lead to a different drying kinetics and final microstructure, due to a different crystal distribution in the case of freeze drying (Tiwari et al. 2015), case hardening for air drying (Joardder et al. 2015) and different CO2 penetration for SCF-assisted technology. The drying efficacy of the three techniques was evaluated in terms of NMC and water activity (aw) (Barbosa-Cánovas et al. 2008). At the end of the drying processes (time specified in the “Material and Methods”), NMC was below 0.1 ± 0.01 for all samples. Water activity was 0.28 ± 0.01 for oven-dried samples, 0.25 ± 0.05 for freeze-dried samples and 0.20 ± 0.01 for scCO2-dried samples; all the values were considerably lower than the threshold of 0.6, below which bacteria and microorganisms cannot grow and proliferate (Barbosa-Cánovas et al. 2008; Rahman 2009). The difference between the drying processes was negligible, within a standard deviation.
Dried Microstructure: Effect of the Water Removal Mechanism
Bulk densities and shrinkage values
Gradual EtOH pre-treatment + scCO2 drying (batch)
Bulk density (kg/m3)
0 ± 10a
17 ± 10b
50 ± 10c
26.20 ± 1.69a
82.55 ± 0.12b
54.65 ± 1.00c
The shrinkage extent defines the bulk sample density after drying. The freeze-dried sample had the lowest bulk density, being highly porous. The bulk density for the oven-dried structure was the highest due to its collapse during the drying process. For supercritical CO2 drying, the final product was homogeneously shrunk, yet not collapsed. On the other hand, the absolute density was not dependent on the drying process and, in fact, it was found to be 1.7 g/cm3 by using the pycnometer, close to the value reported in Upstill et al. (1986).
Dried Microstructure: Effect of the Process Parameters
In addition to the effect of the water removal mechanism on the gel network, the influence of the process parameters was investigated in terms of the produced dried structure. Considering freeze drying, process parameters such as the pressure chamber and the product temperature could likely affect the drying rate (Chang and Patro 2004), rather than the material structure, which is, by contrast, strongly dependent on the gel formulation. Hence, in this work, all these variables were kept constant. The collapse temperature (Tc) in the freeze drying application is specific for each substance and it is the temperature above which the collapse of the frozen structure occurs. This irreversibly leads to the failure of the material and to defect formation (Kett et al. 2005; Bellows and King 1972; Abdelwahed et al. 2006). During gelation, hydrocolloids generate a network, specific for each gelling agent, in which water is embedded. Specifically, the sol-gel temperature of gellan gum is around 30 °C. As the material temperature in the freeze dryer is expected to be around − 18 °C, the structure stability is likely ensured and its collapse avoided (Figs. 1a and 2a).
The water uptake and its diffusion into the material were affected by both the surface and bulk properties (Aguilera and Stanley 1999). The former is more likely influenced by the chemical properties of the gel type and its formulation. On the other hand, the latter considers the mechanical, morphological and physical properties.
Figure 6 suggests that for freeze drying two different rehydration rates, expressed as NMC over time, describe the water uptake. After 6 min, the rate was 0.110 min−1, while for longer timescales, the rehydration slowed down to 0.006 min−1. This trend became less evident for the other drying techniques. For scCO2 drying, this rate passed from 0.15 to 0.04 min−1, while for oven drying, it became negligible.
These results suggest that freeze drying is suitable for applications where a fast rehydration is required, while oven-dried and scCO2-dried structures are more appropriate for applications where the rehydration rate should be slower.
Effect on Gellan Gel Molecular Structure
After drying, the structure needs to be rehydrated to enhance the polymer chain mobility and distinguish clear thermal transitions on cooling. Otherwise, flat thermal events would be recorded. For this reason, 6-h rehydrated samples were analysed with μDSC to assess the effect of the drying process on the gel network.
Peak temperatures, enthalpies and entropies for the gel before drying and after the drying process, followed by rehydration for 6 h
Gradual EtOH pre-treatment + scCO2-dried (batch)
32.6 ± 0.1a
27.0 ± 2.5b
38.9 ± 0.4a
24.1 ± 0.8b
ΔH (kJ kg−1)
− 0.167 ± 0.007a
− 0.084 ± 0.029b
− 0.070 ± 0.017b
− 0.047 ± 0.007b
ΔS (kJ kg−1°C−1) × 10−3
− 5.1 ± 0.2a
− 3.0 ± 0.7a
− 1.8 ± 0.4b
− 1.9 ± 0.2b
Freeze drying was expected to force the alignment of the polymer chains during the freezing step along the ice crystal edges, similarly to what occurs during cryogel formation (e.g., xanthan). The reduction in enthalpy to − 0.084 ± 0.029 J g−1 compared to the gel before drying suggests that fewer disordered chains are involved in the thermal transition. The peak temperature reduction (27.0 ± 2.5 °C) indicates that the chains should be more aggregated due to the previous formation of ice crystals. However, the overall order of the system slightly decreased, as the entropy was reduced to − 3.0 · 10−3 ± 0.7 · 10−3 J g−1 °C−1. This can suggest that the polymer chains in this structure were less ordered although more packed.
Supercritical CO2 drying in batch configuration is a gentle process since the capillary stress suppressed. However, the need to perform an ethanol pre-treatment tends to irreversibly influence the polymer conformation, hardening the material (Eltoum et al. 2001; Buesa 2008; Cassanelli et al. 2017). A significant drop to 24.1 ± 0.8 °C in transition temperature was observed, likely due to the presence of ethanol that altered the water network around the polymer, which obstructs polymer rearrangement (Cassanelli et al. 2017). More time was required for the exothermic transition to happen, in comparison with the gel before drying. Similar considerations were applied to the scCO2 process in the presence of a continuous flow.
In Table 2, the values of peak temperatures, enthalpies and entropies are summarised. On a second thermal cycle, these dried and rehydrated gels showed a similar thermal behaviour recorded on the first cycle.
Conclusions and Future Work
The present work shows for the first time the effect of freeze, oven and scCO2 drying on low-acyl gellan gum gel systems. All the techniques successfully reduced the water activity below the microbial growth threshold. The drying process influenced the dried gel structure. Specifically, freeze drying generated a highly porous material with more aggregated polymer chains. By contrast, the oven-dried gel was completely collapsed, resulting in a gel that slowly and partially reabsorbs water. The scCO2 drying did not induce the structure collapse and only partially shrank the material, leading to a slower water uptake than the freeze-dried gel, yet quicker than the oven-dried gel. However, the necessity to perform an alcoholic pre-treatment made the material harder, changing the polymer network order and increasing the aggregation extent.
The understanding of the relationship between the drying techniques and the produced dried structure can help to design both food products with gelling agent in their formulation and gel agents alone, in either quiescent form or gel particle suspension, throughout the whole production process.
According to with the final application, the most suitable drying technique in terms of produced dried microstructure and the following water uptake might be suggested.
Future works will involve the study of more complex food products, containing hydrocolloids in the formulation. For example, this research can be considered the starting point to design and optimise both the product quality and the production process of freeze-dried ice cream and dairy, dried meet substitutes, ready meals, etc.
This study was funded by the Engineering and Physical Sciences Research Council [grant number EP/K030957/1], the EPSRC Centre for Innovative Manufacturing in Food.
- Aguilera, J. M., & Stanley, D. W. (1999). Microstructural principles of food processing and engineering. New York: Springer Science & Business Media, Springer-VerlagGoogle Scholar
- Barbosa-Cánovas, G. V., Fontana Jr., A. J., Schmidt, S. J., & Labuza, T. P. (2008). Water activity in foods: fundamentals and applications (Vol. 13). John Wiley & Sons. https://doi.org/10.1002/9780470376454.
- Benali, M. & Boumghar, Y. (2014). Supercritical fluid-assisted drying. In A. S. Mujumdar (Ed.), Handbook of industrial drying. 4th ed., ebook edition (pp. 1261–1269). Boca Raton: CRC Press.Google Scholar
- Brown, Z. K. (2010). The drying of foods using supercritical carbon dioxide. Doctoral dissertation, University of Birmingham.Google Scholar
- Brown, Z. K., Fryer, P. J., Norton, I. T., & Bridson, R. H. (2010). Drying of agar gels using supercritical carbon dioxide. Journal of supercritical fluids. 54. Numb, 1, 89–95.Google Scholar
- Brunner, G. (1994). Gas extraction : an introduction to fundamentals of supercritical fluids and the application to separation processes. Darmstadt: Steinkopff Verlag; New York: Springer-Verlag.Google Scholar
- Chang, B. S., & Patro, S. Y. (2004). Freeze-drying process development for protein pharmaceuticals. In H. R. Costantino & M. J. Pikal (Eds.), Lyophilization of biopharmaceuticals (pp. 113–138). Arlington: American Association of Pharmaceutical Scientists.Google Scholar
- de Vries, J. (2002). Interaction of carrageenan with other ingredients in dairy dessert gels. Special Publication - Royal Society of Chemistry, 278, 201–210.Google Scholar
- Evans, J. A. (Ed.). (2009). Frozen food science and technology. Hoboken: John Wiley & Sons.Google Scholar
- Gulrez, S. K., Al-Assaf, S., & Phillips, G. O. (2011). Hydrogels: methods of preparation, characterisation and applications. In A. Carpi (Ed.), In progress in molecular and environmental bioengineering-from analysis and modeling to technology applications (pp. 117–150). Winchester: InTech.Google Scholar
- Joardder, M. U., Karim, A., Kumar, C., & Brown, R. J. (2015). Porosity: establishing the relationship between drying parameters and dried food quality. Springer. https://doi.org/10.1007/978-3-319-23045-0.
- Marabi, A., Dilak, C., Shah, J., & Saguy, I. (2004). Kinetics of solids leaching during rehydration of particulate dry vegetables. Journal of Food Science, 69(3).Google Scholar
- Norton, I., & Foster, T. (2002). Hydrocolloids in real food systems. Special Publication - Royal Society of Chemistry, 278, 187–200.Google Scholar
- Phillips, G. O., & Williams, P. A. (2004). Gums and stabilisers for the food industry 12. Cambridge: Royal Society of Chemistry.Google Scholar
- Phillips, G. O., & Williams, P. A. (2009). Handbook of hydrocolloids. Cambridge: Woodhead Publishing Limited.Google Scholar
- Sachlos, E., & Czernuszka, J. (2003). Making tissue engineering scaffolds work. Review: the application of solid freeform fabrication technology to the production of tissue engineering scaffolds. European Cells & Materials, 5(29), 39–40.Google Scholar
- Suderman, N., Isa, M. I. N., & Sarbon, N. M. (2016). Effect of drying temperature on the functional properties of biodegradable CMC-based film for potential food packaging. International Food Research Journal, 23(3), 1075–1084.Google Scholar
- Wang, L. (2008). Energy efficiency and management in food processing facilities. CRC press.Google Scholar
Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.