Is there a diffusion of alkali in the activation of dissolving cellulose pulp at low NAOH stoichiometric excess?
- 42 Downloads
We conducted a quantitative study, following the degree of activation (i.e. the transformation to alkali cellulose, denoted as DoA) over time for dissolving cellulose pulp treated with different [NaOH] at low NaOH/anhydroglucose unit stoichiometric ratio (denoted as (r) ≤ 2.6). Our quantitative approach was based on Raman spectroscopy data, evaluated by partial least squares regression modelling. The results show strong influence of the (r) on DoA (increasing from DoA = 45% at (r) = 0.8, to DoA = 85% at (r) = 2.6), and its complex dependence on [NaOH]. At (r) = 0.8 the highest DoA (DoA ≳ 60%) was found at 30% [NaOH], while at (r) = 1.3 it was found at 20% [NaOH] (DoA ≳ 80%). Although activation of cellulose happens in minutes at the studied temperature (30 °C), it was found that the reaction may be slow when a low (r) is used. A gradual increase of the DoA from ≈ 30 to ≈ 70% in time was seen when samples were activated with 30% [NaOH] at (r) = 0.8. At the same (r), a similar increase of DoA from ≈ 30 to ≈ 60% was also observed when 40% [NaOH] was used. Slow diffusion of NaOH through poorly swollen cellulose fibres is proposed as an explanation for this phenomenon. Lastly, solid-state CP/MAS NMR measurements suggest that at a fixed temperature, the Na-Cell allomorph mostly depends on [NaOH]. However, in the transition area between Na-Cell I and Na-Cell II, its influence might be affected by (r).
KeywordsAlkali cellulose Activation Polymorphs Cellulose ether Diffusion
In the first step of manufacturing cellulose-based value-added products, e.g. viscose fibres and cellulose ethers, the cellulose, i.e. parallel crystalline structure cellulose I (Cell I), is treated with aqueous NaOH to alter its physical and chemical properties in a process called mercerisation. During mercerisation, Cell I transforms into a swollen crystalline structure, alkali cellulose (Na-Cell). This increases the reactivity of the cellulose and makes it more accessible for reagents to penetrate and react with the hydroxyl groups. Due to the importance of this process step, the cellulose-NaOH reaction has been extensively studied, where research has tended to focus at a relatively high stoichiometric ratio of NaOH/AGU (anhydroglucose unit), typically (r) ≥ 20 (e.g. Albán Reyes et al. 2016; Philipp et al. 1983; Porro et al. 2007; Ranby 1952; Schenzel et al. 2009; Sobue and Kiessig 1939; Yue et al. 2013) rather than at low (r). Mercerising cellulose in excess of NaOH may be appropriate when making viscose fibres (Mozdyniewicz and Sixta 2012; Wilkes 2001), but less so in processes related to cellulose ethers. The use of such high (r) then leads to the formation of unwanted by-products, which are generated by the reactions of the added chemicals with water and excess OH− ions (e.g. Saxell et al. 2015). To avoid this, modern cellulose ethers production is done at low-water-content (LWC) mercerisation i.e. low (r) ((r) = 1.1 − 5.0) in combination with high [NaOH] ([NaOH] = 40–50% w/w) (Berglund et al. 2009; Boström and Karlson 2013). Hereafter, we refer to low (r) when mercerisation is done at (r) ≤ 5.
The Na-Cell allomorph formed during the interaction of cellulose with an aqueous solution of NaOH as a function of temperature and [NaOH] using high (r) was earlier presented by Sobue and Kiessig (1939) where five different allomorphs Na-Cell were reported when X-ray diffraction was used. Later, Porro et al. (2007) suggested a reconsideration of such Na-Cell allomorphs by using 13C CP/MAS NMR. Only two types of Na-Cell were found, Na-Cell I and Na-Cell II, where Na-Cell I was more predominant at [NaOH] ≤ 25%, while Na-Cell II dominated at [NaOH] ≥ 30%. The transformation of these two allomorphs into an antiparallel chain structure, cellulose II (Cell II) is obtained by neutralising and drying (Langan et al. 2001; Okano and Sarko 1985). This transformation has been used as a standard procedure to investigate mercerisation (Gupta et al. 2013; Jin et al. 2016; Sisson and Saner 1941), and we used the same approach to define the degree of activation (DoA).
In mercerisation at higher (r), e.g. steeping, more contact between Cell I and NaOHaq take place, making mercerisation a rather quick process, however, highly affected by temperature and [NaOH] (Albán Reyes et al. 2016; Borysiak and Garbarczyk 2003; Ranby 1952; Sisson and Saner 1941). Despite these early observations, it remains unclear if this is the case for LWC mercerisation, where less contact between cellulose and NaOHaq takes place, and the system is drier compared to mercerisation at higher (r). In fact, in this type of mercerisation, it is essential to use a mixing system of high intensity, that helps to bring all the components together. Hence, it is of interest to know whether there is time dependence or diffusion of NaOH in this type of mercerisation or not. The study of the DoA over time is important for the cellulose ether producers, where a highly activate cellulose is desired, but the contact area between the cellulose and NaOH in their process conditions is relatively small. In a previous study, we have shown that during mercerisation at LWC, the variation in (r) and [NaOH] becomes important for the DoA, while the temperature shows to be of less importance in such a heterogeneous system (Reyes et al. 2017). We also suggested that an additional diffusion-controlled mercerisation mechanism directs the mercerisation when (r) = 0.8. This was based on the fact that DoA in ground cellulose pulp seemed to increase with time at the mentioned (r). This study, however, only covered a limited range of [NaOH] ([NaOH] = 45–55%) and reaction time (5 and 25 min). Therefore, the suggested mechanism is inconclusive. Hence, a comprehensive evaluation of the suggested diffusion-controlled mercerisation at LWC, covering a large [NaOH] span (in terms of its ratio to water) and at longer reaction time was, to our knowledge, still needed. Moreover, it remains uncertain whether the use of low (r) affects the type of Na-Cell possible to obtain. In the present study, we have attempted to fill these gaps. First, by exploring diffusion of NaOH from pre-mercerised cellulose into added unreacted cellulose pulp. Second, by mercerising cellulose pulp at two fixed low (r), with a broader variation of [NaOH]. Third, by mercerising cellulose pulp at different low (r) with a fixed [NaOH]. In mentioned studies, we have followed the DoA over time. Furthermore, we examine the type of Na-Cell formed at a sub-stoichiometric ratio with different [NaOH], as well as at different low (r) with a fixed [NaOH]. We have measured the DoA quantitatively, using a partial least square (PLS) regression modelling on Raman spectroscopic data of produced samples. The type of Na-Cell formed at the studied conditions was qualitatively evaluated, using CP/MAS NMR measurements. The presented approach thus offers substantial information that can be helpful for the understanding of mercerisation at low (r) over time.
Materials and methods
The sample material was sheet sulphite dissolving cellulose pulp from a blend of spruce and pine, with 544 ml g−1 of viscosity (ISO 2470:1999), 95.3% of R 18 (ISO 699:1982) and 89.6% of R 10 (ISO 699:1983) provided by Domsjö Fabriker AB, Örnsköldsvik, Sweden. The moisture content (MC) of the pulp was ≤ 5%, measured using a Mettler Toledo HG63 moisture analyser. For the preparation of alkaline aqueous solutions, NaOH from Sigma Aldrich (ACS reagent, ≥ 97.0%) and degassed (by boiling) ultrapure Milli-Q H2O was used. All the solutions were prepared in a temperature-controlled room at 25 ± 0.1 °C and kept in vessels protected from atmospheric CO2. All alkaline solutions were used within 24 h. Deionised H2O was used for washing the NaOH out of the mercerised samples.
The sheet dissolving pulp was ground before performed experiments in a fine cutting mill (Netzsch, Germany) located in the pilot grinder at AkzoNobel, Örnsköldsvik Sweden. The obtained cellulose power had an average arithmetic length of 0.17 mm.
Fibre length (size) measurement
The fibre length analysis was performed using PulpEye® fibre module. The instrument uses high-resolution optics to analyse fibre particles in a diluted suspension. The error of measurements is ≤ 1% on rayon fibres that are 2 mm long. The length limit for the analysed ground cellulose pulp was set to 0.01 mm.
Mercerisation at low NaOH stoichiometric excess
For mercerisation, 5 g of dry dissolving ground cellulose pulp (MC ≤ 5%), and NaOHaq were used. The NaOHaq was sprayed using a modified airbrush (Biltema Sweden AB, item no. 17372), to ensure even distribution onto the cellulose pulp. The minimum amount of NaOHaq sprayed onto ground cellulose pulp in this study was 1.28 ml of 50% [NaOH] at (r) = 0.8, while the maximum amount was 14.35 ml of 10% [NaOH] at (r) = 1.3. The mercerisation was carried out in a water jacketed z-mixer (kneader) at the viscose pilot plant at MoRe research in Örnsköldsvik, Sweden. The temperature during the mercerisation was 30 °C, and it was controlled by the z-mixer, as well as an external water bath. The temperature uncertainty was estimated to ± 2 °C. Photos of the mentioned setup are shown in the supplementary materials of one of our previous publications (Reyes et al. 2017). The samples were continuously mixed during the spraying of the NaOHaq. Samples were collected at the set reaction times by stopping the mixing and removing samples. The mercerisation reaction was interrupted by washing the cellulose in excess of deionised H2O, three to four times. The samples were then suspended in deionised H2O, and the pH was checked for neutrality using indicator paper. Finally, the samples were dried at 40 °C in vacuum until constant weight.
To analyse diffusion of NaOHaq from pre-mercerised ground dissolving cellulose pulp into added dry dissolving ground cellulose pulp (i.e. non-mercerised cellulose), two initial mercerisations were carried out, at (r) = 1.6 and at (r) = 2.6. In both mercerisations, 19.1 N NaOH (50% w/w) and 60 min reaction time was used. Thus, representing the pre-mercerised cellulose (Pre). The same amount of dry ground cellulose pulp was then added, and the diffusion of the NaOH from pre-mercerised cellulose into the added cellulose pulp was measured 5 (t5), 10 (t10) and 25 (t25) min after this addition. After the addition of cellulose, the average (r) values were 0.8 and 1.3 for the initial (r) values of 1.6 and 2.6, respectively.
Dry dissolving cellulose pulp was mercerised with 19.1 N (50% w/w), and at (r) = 0.8, 1.3, 2.6 (c.f. Fig. 1). The reaction time was 5, 25, 90, and 1440 min. As was mentioned previously, after 90 min of mercerisation in the kneader, the samples were taken out and left at room temperature until 1440 min was reached.
Calibration set samples
We created a calibration set samples that represent different DoA levels in order to build a calibration model based on their average Raman spectra data evaluated by PLS. The obtained model was then used to predict the DoA of the mercerised samples at low (r). The method used to create the calibration set samples was previously reported by Albán Reyes et al. (2016). The calibration set consists of a mix of different weight proportions of dissolving cellulose pulp, depicted Cell I and fully mercerised material, depicted Cell II to represent different DoA levels ranging from 0 (dissolving cellulose pulp) to 100 (fully mercerised material) in intervals of 10% (w/w), as was described by Schenzel et al. (2009). The fully mercerised material was produced by mercerising dissolving cellulose pulp with an excess of 30% (w/w) [NaOH], at 3 °C for 1 h. The sample was then left without stirring at room temperature for 24 h, followed by washing until neutrality and drying. After mercerisation, the sample was considered to represent a fully mercerised material (Porro et al. 2007). After washing and drying it was considered to contain very low levels of Cell I, high levels of Cell II (≈ 95%), as well as amorphous cellulose (Schenzel et al. 2009).
The Raman mapping data on a flat surface of the samples were acquired using a Renishaw InVia Raman spectrometer equipped with a CCD detector. The laser source was a 785 nm diode laser with maximum power of 300 mW and 1200 lines grating. The measurements were recorded in static mode, with a centre at 950 cm−1 (resulting in a spectral region of ca. 326–1494 cm−1), 1 s exposure time and 100% laser power. Raman shifts were calibrated using the built-in Si standard of the instrument. We used a 20 × lens, and 10 microns step size for mapping. For every mapping 150–350 spectra were recorded. Cosmic ray effects were removed, and data noise-filtered using WiRE (version 3.0, Renishaw Plc, UK). The Raman spectra of Set 2 samples, at (r) = 1.3, were recorded using a Bruker BRAVO portable instrument and OPUS version 7.8 (Bruker Optics GmbH, Ettlingen, Germany). In this case, the samples were placed in front of the laser beam exit manually, and spectra were recorded in the range of 326–1494 cm−1 at 4 cm−1 spectral resolution, using automatic integration times and 3 scans/sample. For all recorded Raman data, a baseline correction, Savitzky–Golay smoothing (Savitzky and Golay 1964), and a total area normalization was performed using a Matlab script provided by the Vibrational Spectroscopy Core Facility (http://www.kbc.umu.se/english/visp/download-visp/), the parameters used for baseline correction were λ = 1000 and p = 0.001, while for smoothing a polynomial order = 1 and frame rate = 3 was used. Finally, after pre-processing an average spectrum was calculated for each map and used for multivariate data analysis.
Partial least squares regression method
A PLS regression model (Eriksson et al. 2013; Geladi and Kowalski 1986) was used to correlate the averaged Raman spectra data of the calibration set samples to the known levels of dissolving cellulose pulp and fully mercerised material by using SIMCA (v.126.96.36.199, Umetrics AB, Umeå, Sweden). We used the Raman spectra data as X-data and the known levels of fully mercerised material as Y-data. The X-data were mean-centred, and the total number of components used in the model was two. The obtained PLS model was then used to predict the DoA of the mercerised samples at low (r) by using their average Raman spectra data as X-data and obtained their predict DoA value as a Y-data.
Nuclear magnetic resonance spectroscopy measurements
After 24 h of mercerisation, the appearance of the alkaline samples, Set 2 and Set 3 was still quite dry and heterogeneous. The produced alkaline samples were, in this semi-wet state, packed into a 4 mm zirconium oxide rotor for 13C CP/MAS (cross-polarisation magic angle spinning) Nuclear Magnetic Resonance (NMR) measurements. The NMR spectra of the samples were recorded with a Bruker Avance III 500 spectrometer (Bruker Biospin GmbH, Reinstetten, Germany) at 298 K with a spin rate of 10,000 Hz. Because of the high but differing alkalinity, the acquisition of NMR spectra varied for each sample within the given range, i.e. parentheses. A 3 µs (3–6 µs) 1H excitation pulse was used followed by cross polarisation at 62.5 kHz (31–62.5 kHz) with a 50–100% ramped proton field up to 82 kHz (37–82 kHz). The contact time for all samples were 1.5 ms, and SPINAL-64 decoupling at 83 kHz (37–83 kHz) was used. The recycle delay for all samples was 2.5 s, giving an acquisition time of approximately 1.15 h per sample. The Adamantane CH2 peak was used as an external reference, set at 38.5 ppm (Morcombe and Zilm 2003).
Results and discussion
Calibration samples and multivariate analysis of Raman spectroscopic data
Mercerisation at low NaOH stoichiometric excess
In the present study, the mercerisation at low (r) was performed using the same source of raw material, i.e. same batch, which consisted of softwood sulphite dissolving cellulose pulp from pine and spruce. Somewhat different results can be expected in DoA if other types of cellulose sources are used. We mercerised the dry dissolving cellulose pulp at (r) = 1.6 and 2.6 with 50% [NaOH] for 60 min (i.e. Pre-mercerised samples) and diluted the Pre-mercerised samples by adding dry dissolving cellulose pulp to an average value of (r) = 0.8 and (r) = 1.3. Thus, the diffusion of NaOH from the pre-mercerised cellulose (Pre) into added cellulose was analysed. DoA was re-measured for the total amount of cellulose (Pre-mercerised plus added cellulose pulp) 5 min after the addition of unreacted dry dissolving cellulose pulp, denoted here as t5. As expected, the subsequent addition of unreacted dry cellulose pulp (Fig. 3) results in a decrease of the measured DoA, and it also increases the standard deviation. The first effect is ascribed to the unreacted cellulose (increasing the total cellulose by a factor of 2), while the second effect can be ascribed to the high heterogeneity in the samples after addition of unreacted cellulose pulp. No increase of DoA with time can be observed for samples at (r) = 0.8. However, a slight increase of the measured DoA is observed after 10 min when activation was done at (r) = 1.3, indicating a possible diffusion of the NaOH. Although the sample at (r) = 0.8 does not show any increase in DoA after the cellulose is added, diffusion cannot be dismissed. It is important to note that any difference in the results obtained in these samples are not statistically significant due to the large standard deviation of the measured DoA values in the samples after subsequent addition of cellulose, which suggests a high heterogeneity of the mixed sample, and thus any weak increase of the DoA would be invisible.
The NMR spectra of semi-wet mercerised samples at (r) = 0.8 shows that pure Na-Cell is not formed (Fig. 6). This agrees with the measured DoA from the PLS model of the Raman data (Fig. 4). All the NMR spectra of samples mercerised at this (r), thus, shows that the samples still contain unreacted material. Keeping in mind that it is likely that the (r) must be ≥ 1 in order to have a stoichiometric amount of NaOH per AGU (Albán Reyes et al. 2017; Nishimura et al. 1991), it is interesting to observe that at this sub-stoichiometric ratio combined with such a low [NaOH] as 10% we can start to observe changes in C1, C4 and C6 of the NMR spectra of the cellulose (Fig. 6d). When the [NaOH] is increased to 20, and 30% (Fig. 6e, f), fewer features of the dissolving cellulose pulp are observed and there are more similarities with Na-Cell I, most noticeably for cellulose mercerised with 30% [NaOH]. When the [NaOH] increases to 40% and 50% (Fig. 6g, h), the NMR spectra of the samples have similarities to the cellulose pulp and Na-Cell II. The type of Na-Cell formed at 20%, 40% and 50% [NaOH] agree with those already published when (r) > 40 is used (Porro et al. 2007). However, the initial change observed in the starting material at 10% [NaOH] differs from the observation given in the mentioned study, where the NMR spectra of the cellulose remain intact at [NaOH] below 13%. In agreement with our observations, milled cotton linter treated with 10% [NaOH] at (r) ≈ 44 resulted in a transition from Cell I to Na-Cell I (Yokota et al. 1990). Furthermore, Schenzel et al. (2009) observed a decrease in Cell I content when sulphite pulp was mercerised using [NaOH] between 6 and 12% at (r) ≈ 15. The observed discrepancy thus seems to be due to, e.g. raw material and pre-treatment of the cellulose. A more interesting difference is seen when the cellulose is alkalised with 30% [NaOH] at (r) = 0.8 (Fig. 6f). Already in 1939, Sobue et al. found that Na-Cell II is formed when the [NaOH] was 30%. This was later confirmed by Porro et al. (2007). At our experimental conditions, however, only the signals assigned to Na-Cell I [e.g. 108.1 ppm (C1), 86.6 ppm (C4), and 62.7 ppm (C6)] are observed. This indication of Na-Cell I, which was unexpected from literature, is probably due to a change in the AGU/NaOH/H2O equilibrium due to the much lower (r) than what is reported previously. As noted by Porro et al. (2007) applied constraints conditions (e.g. here low (r) at 30% [NaOH]) can probably lead to different cellulose allomorphs. In contrast, Fig. 7 shows that the Na-Cell allomorph formed at higher [NaOH] (50% w/w) and at 0.8 < (r) < 2.6 seems not to be influenced by the (r). In this figure, all the spectra show similarities to Na-Cell II, where the feature of this type of Na-Cell become more pronounced with increased (r). In summary, NMR measurements suggest that at the studied temperature, the concentration of NaOH mainly determines the Na-Cell allomorph. However, close to the transition area (c.f. Porro et al. 2007), (r) also seems to be of importance.
Diffusion of NaOH at low NaOH stoichiometric excess
In our view, mercerisation at low (r) could, in part, proceed through diffusion of NaOH in already swollen ground cellulose pulp; this phenomenon being more pronounced at sub-stoichiometric (r), Set 2. According to literature, swelling of cellulose at mercerisation conditions i.e. [NaOH] ≥ 10%, causes a change in the morphology of the cellulose, transforming it into a highly swollen material (Na-cell) (Budtova and Navard 2016). Because of the morphology change, we reason that swelling facilitates a further diffusion of NaOH into more cellulose fibres, leading to a continued and controlled Na-Cell formation over time. As noted by us in a previous study, this controlled feedback of swelling → diffusion of NaOH → Na-Cell formation → swelling seems to be important for the DoA at low (r) (Reyes et al. 2017). It is known that, at mercerisation conditions, the cellulose swells better at low [NaOH] (Richter and Glidden 1940), and the diffusion of NaOH in aqueous media is fast (Gavillon and Budtova 2007). Therefore, the diffusion of NaOH into well-swollen cellulose at low (r) should benefit from low [NaOH] (Illustration S1, supplementary material). The data suggests that this diffusion is fast at [NaOH] ≤ 20%, when mercerisation is performed at (r) = 0.8 and at (r) = 1.3 (Fig. 4). Thus, it is possible that the NaOH diffuse through the well-swollen cellulose, which makes the diffusion too fast to notice on our timescale. At 30% and 40% [NaOH], however, the H2O content at (r) = 0.8 is not high enough to allow good swelling. Thus, a quick diffusion of NaOH through the system cannot occur, and the data shows that the measured DoA slowly increases with time. At (r) = 1.3 more H2O is present, and the diffusion through swollen cellulose is faster (c.f. Fig. 1). Hence no increase of DoA can be observed with time. This explains why the amount of water in the system, given by the combination of both (r) and [NaOH], has an influence on the apparent diffusion rate of NaOH, affecting the time it takes for the system to reach a steady-state. As was discussed above, the alkali adsorption onto cellulose surfaces could also affect the diffusion of the NaOH when (r) < 1.
In this study, the gradual increase of the DoA observed at some conditions, indicates a continuous Na-Cell formation. A plausible explanation is the diffusion of NaOH through poorly swollen cellulose fibres. This diffusion would be affected by the amount of water present in the system given by [NaOH] and/or (r). The amount of time required for the system to reach its equilibrium is thus influenced. At sub-stoichiometric conditions, such diffusion was clearly observed at 30% and 40% [NaOH]. Our data show a remarkable influence of (r) on DoA and its complex dependence on the [NaOH]. At (r) ≤ 1.3, the lowest DoA was found at (r) = 0.8 and 10% [NaOH], while the highest value was observed at (r) = 1.3 and 20% [NaOH]. Lastly, 13C CP/MAS NMR measurements allow us to identify the Na-Cell type formed at low (r). Our data shows that, at a fixed temperature, the [NaOH] plays an important role in determining the form of Na-cell, as it does when mercerisation is done at high (r). However, close to the transition area between Na-Cell I and Na-Cell II, (r) also seems to be of importance.
This research was supported by the Industrial Doctoral School (IDS) at Umeå University (UmU), AkzoNobel Functional Chemicals, Domsjö Fabriker AB, Bio4Energy, and SP Processum. We thank the Vibrational Spectroscopy and NMR platform at KBC (UmU) for experimental guidance and assistance.
- Atalla RH (1975) Raman spectral studies of polymorphy in cellulose. Part 1, Celluloses I and II. IPC Technical paper series no. 19Google Scholar
- Berglund L, Johansson KA, Sundberg K (2009) Process for the manufacture of methyl cellulose ether. US Patents 7,504,498 B2Google Scholar
- Borysiak S, Garbarczyk J (2003) Applying the WAXS method to estimate the supermolecular structure of cellulose fibers after mercerization. Fibers Text East Eur 11:44Google Scholar
- Boström P, Karlson L (2013) Associative water-soluble cellulose ethers. US Patents 8,487,089 B2Google Scholar
- Eriksson L, Byrne T, Johansson E, Trygg J, Vikström C (2013) Multi-and megavariate data analysis basic principles and applications, 3rd edn. Umetrics Academy, MalmöGoogle Scholar
- Mozdyniewicz DJ, Sixta H (2012) Carbohydrate degradation reactions during alkaline steeping of dissolving pulp-influence of air exclusion. Lenz Berich 90:103–107Google Scholar
- Philipp B, Fink H, Kunze J, Purz H (1983) Recent results on formation and structure of alkali cellulose [Cotton linters]. In: Proceed technical assoc of the pulp and paper industry (USA)Google Scholar
- Saxell H, Heiskanen I, Axrup L, Hensdal CL, Jokela V (2015) Method for the preparation of cellulose ethers with a high solids process, product obtained and uses of the product. US Patents 2015/0094464 A1Google Scholar
- Wilkes AG (2001) 3—the viscose process A2—Woodings, Calvin. In: Regenerated cellulose fibres. Woodhead Publishing, pp 37–61. http://dx.doi.org/10.1533/9781855737587.37
Open AccessThis 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.