Professional Documents
Culture Documents
Teodorodasilva 2020
Teodorodasilva 2020
Teodorodasilva 2020
LWT
journal homepage: www.elsevier.com/locate/lwt
A R T I C L E I N F O A B S T R A C T
Keywords: The objective of this study was to compare the sonocrystallization process of a palm-based fat in batch to a
Sonocrystallization continuous set-up using a scraped surface heat exchanger (SSHE). The sample was crystallized at 32, 30, 28, and
Batch 26 ◦ C (Tc). High-intensity ultrasound (HIU, 20 kHz, 12.7 mm horn, 50% amplitude, 57W) was applied using 10s
Scraped surface heat exchanger
pulses. After crystallization, samples were stored (48 h/25 ◦ C) and analyzed for crystal microstructure, melting
Low saturated fat
behavior, oil binding capacity (OBC), hardness, and viscoelastic properties. HIU improved all physical properties
of the material when crystallized in batch at all Tc; however, SSHE at 26 ◦ C was the only condition improved by
HIU. G’ and hardness were the highest in sonicated SSHE samples crystallized at 26 ◦ C (141 kPa, 4.5N) and these
values were higher than the ones obtained in batch. OBC was also improved in sonicated SSHE samples (77%)
but in lower magnitude compared to the OBC obtained in batch (82%). Crystal size was not affected by HIU in
the SSHE, but a reduction in crystal size was observed due to HIU in batch for all Tc. Similar behavior was
observed for Tp, however, Ton and ΔH were not affected by HIU.
* Corresponding author. Department of Nutrition, Dietetics, and Food Sciences, Utah State University, 8700 Old Main Hill, Logan, UT, 84322-8700, USA.
** Corresponding author.
E-mail addresses: thais_lts@yahoo.com.br (T.L.T. da Silva), silvana.martini@usu.edu (S. Martini).
https://doi.org/10.1016/j.lwt.2020.110593
Received 12 August 2020; Received in revised form 23 October 2020; Accepted 14 November 2020
Available online 21 November 2020
0023-6438/© 2020 Elsevier Ltd. All rights reserved.
T.L.T. da Silva et al. LWT 138 (2021) 110593
2
T.L.T. da Silva et al. LWT 138 (2021) 110593
Fig. 2. Images of the crystalline network obtained from polarized light microscopy at 25 ◦ C after crystallization in batch or in a SSHE at different temperatures with
and without HIU and stored at 25 ◦ C for 48h. Magnification 10x.
adjustments to the flow rate and mobile phase composition were made 2, Lumenera Scientific, Ottawa, ON, Canada). Images were obtained and
as described below. The analysis was performed on a Waters HPLC used to characterize crystal morphology. The diameters of the crystals
system (Zellik, Belgium), equipped with two stainless steel Nova-Pak were measured using the software Image-Pro® Premier E 9.2 (Media
C18 columns (4 μm, 3.9 × 150mm) (Waters, Belgium). The mobile Cybernetics, USA). For each crystallization condition four images were
phase was an isocratic solvent mixture of acetone and acetonitrile obtained.
(62.5:37.5) with a flow rate of 1.2 mL/min; the injection volume was 20
μl. The samples (0.04g) were dissolved in methanol/chloroform (1.6 mL; 2.3.3. Oil binding capacity
1:1 v:v) and a differential refractometer was utilized for detection. The After storage at 25 ◦ C for 48h, 1g of the crystallized sample was
equivalent carbon number (ECN) was used to predict the elution order. placed in a pre-weighed microcentrifuge tube and centrifuged at 10,000
ECN = CN-2(DB), where CN is the total carbon number and DB is the rpm for 15 min at room temperature using a microcentrifuge (Ther
total number of double bonds in the fatty acids. Peaks were integrated by moscientific, Sorvall micro 17 centrifuge, Germany). After centrifuga
using the program Empower Pro with a generic Apex Track method and tion, the microcentrifuge tube was inverted upside down and left to
areas were correlated with the quantities of TAG in the sample. Peak drain for 3min to release the liquid oil not entrapped in the crystalline
areas below 4000 area counts were not considered. network. The oil binding capacity (OBC) was performed in quadrupli
cate and calculated according to equation (1).
2.3.2. Polarized light microscopy ( )
wbefore − wafter
After storage at 25 ◦ C for 48h a small portion of the crystallized OBC(%) = 100 − * 100 (1)
Wsample
samples were placed on a slide and covered with a cover-slide. The slides
were placed on a polarized light microscope (Olympus BX41 micro
where Wbefore is the initial weight of the tube and the sample, Wafter is the
scope, Olympus, Tokyo, Japan) equipped with a digital camera (Infinity
3
T.L.T. da Silva et al. LWT 138 (2021) 110593
Table 2 removed and cut into 1 cm height cylinders. Hardness was measured
Crystal size obtained from polarized light microscopy images reported as using a Texture Profile Analyzer (Model TA, XT Plus, Texture Technol
diameter (μm, mean ± standard deviation) at 25 ◦ C after crystallization in batch ogies Corp., Hamilton, MA, United States) with a two-step 25%
or in a SSHE at different temperatures with and without sonication and stored at compression test and a 5 cm cylinder probe. Five tubes were prepared
25 ◦ C for 48h. for hardness measurement for each crystallization run.
Temperature of crystallization (◦ C) Process condition Diameter (μm)
4
T.L.T. da Silva et al. LWT 138 (2021) 110593
Fig. 3. Macroscopic view of crystalline networks obtained at 25 ◦ C after crystallization in batch or in a SSHE at different temperatures with and without HIU and
stored at 25 ◦ C for 48h.
SSHE system with wall temperature of 0 ◦ C and 40 rpm agitation, sug non-sonicated ones, and sonicated samples crystallized in batch at 26,
gesting that the time, amount, duration, and crystallization system used 28, and 30 ◦ C looked harder than the ones crystallized at 32 ◦ C. These
for shear influences crystal size. The reduction in crystal size in the SSHE differences in visual hardness are attributed to the crystal sizes discussed
was due to the shaved agitation that promoted faster cooling, induced above. Samples crystallized in the SSHE without sonication were visibly
nucleation, and reduced crystal growth. No differences in crystal size or harder than samples crystallized without HIU in batch and they looked
morphology were observed due to sonication at all Tc in the SSHE, as as hard as the batch sonicated samples. After sonication, only SSHE
observed by (da Silva, Danthine et al., 2020b) this result does not samples crystallized at 26 and 28 ◦ C kept this visual appearance while
necessarily indicate that HIU was not effective, but that SSHE has a samples crystallized with HIU in the SSHE at 30 and 32 ◦ C were visibly
bigger effect on tailoring crystal size. softer, even though no differences were observed in the crystals
Samples crystallized in batch at all Tc without sonication had a soft morphology and size.
texture that flowed easily and did not maintain its shape when placed in
a watch-glass (Fig. 3). Big crystals usually hinder the formation of a well- 3.3. Oil binding capacity (OBC)
structured crystalline network due to the small contact area and the
weak interactions among them (Kanagaratnam et al., 2013). After son The non-sonicated samples crystallized in batch, except for the one
ication, all samples crystallized in batch were visibly harder than the crystallized at 32 ◦ C, showed lower OBC (p < 0.05; Fig. 4) than samples
5
T.L.T. da Silva et al. LWT 138 (2021) 110593
Fig. 4. Oil binding capacity (OBC) obtained at 25 ◦ C after crystallization in batch or in a SSHE at different temperatures with and without HIU and stored at 25 ◦ C
for 48h.
Fig. 5. Viscoelastic parameters G’ (A) and G” (B) obtained at 25 ◦ C after crystallization in batch or in a SSHE at different temperatures with and without HIU and
stored at 25 ◦ C for 48h.
crystallized in the SSHE. These results suggest that SSHE induced weaker interactions of the crystalline network (Kanagaratnam et al.,
nucleation and created smaller crystals that can entrap more liquid oil 2013).
when samples are crystallized between 26 and 30 ◦ C. In a previous study The lack of effect on OBC for the samples crystallized in the SSHE at
in a palm blend with high levels of SFA, samples with bigger crystals had 32 ◦ C could be due to the slow crystallization process of the sample (De
a poorer OBC due to the lower surface area of the bigger crystals and Graef et al., 2009). When the sample was crystallized at 32 ◦ C, the higher
6
T.L.T. da Silva et al. LWT 138 (2021) 110593
Fig. 6. Hardness (N) measured at 5 ◦ C after crystallization in batch or in a SSHE at different temperatures with and without HIU and stored at 5 ◦ C for 48h.
OBC for all Tc (p < 0.05). The highest values were found for samples
Table 3 crystallized at 26, 28, and 30 ◦ C and the lowest for samples crystallized
Melting parameters obtained from DSC curves at 25 ◦ C after crystallization in at 32 ◦ C. An increase in OBC was observed only in sonicated samples in
batch or in a SSHE at different temperatures with and without HIU and stored at
the SSHE at 26 ◦ C and this improvement was not as big as the one
25 ◦ C for 48h.
observed for samples crystallized in batch. When samples were soni
Temperature of Process Parameters cated in the SSHE at 28 ◦ C no change in the OBC was observed. However,
crystallization(◦ C) condition
Ton (◦ C) Tp (◦ C) ΔH (J/g) sonication at higher temperatures, 30 and 32 ◦ C, in the SSHE resulted in
26 Batch control 27.26 ± 39.17 ± 15.54 ±
a lower OBC compared to the non-sonicated samples. A previous study
0.35a 0.26 bc 2.53a in our group found that HIU improved OBC in low saturated fats because
Batch HIU 28.12 ± 37.72 ± 13.72 ± cavitation and bubbles created during sonication induced secondary
1.30a 0.48 cd 1.36a nucleation and crystallization of low-melting-point TAGs that would not
SSHE control 27.58 ± 37.43 ± 13.83 ±
crystallize without sonication (da Silva, Cooper et al., 2020a). The TAGs
0.36a 0.42d 1.67a
SSHE HIU 27.98 ± 37.85 ± 13.10 ± presented in high amounts due to sonication in (da Silva, Cooper et al.,
0.45a 0.22 cd 0.69a 2020a) were LLL, OLLn, LLO, PLL and POL, similar to the TAGs present
28 Batch control 27.76 ± 40.61 ± 14.82 ±
in this study (Table 1), suggesting that either they were better entrapped
0.14a 0.53 ab 1.26a or induced to crystallize. Shear that induced nucleation in the SSHE and
Batch HIU 27.54 ± 37.77 ± 14.32 ± assisted in the formation of the crystalline network can also break it, if
0.25a 0.34 cd 0.74a applied under inappropriate conditions. This could explain the small
SSHE control 27.87 ± 37.53 ± 13.57 ±
effect of SSHE on OBC values, where the shear experienced in the sample
0.18a 0.30d 0.93a
SSHE HIU 28.23 ± 37.27 ± 13.10 ± after crystallization improved the plasticity of the sample and might
0.53a 0.71d 0.85a have interfered with oil entrapment.
30 Batch control 28.03 ± 41.93 ± 9.68 ±
0.68a 1.65a 1.78b 3.4. Viscoelasticity
Batch HIU 28.23 ± 38.40 ± 15.24 ±
1.17a 1.44 cd 0.70a
SSHE control 27.78 ± 37.19 ± 14.08 ± All G′ values were higher than G” indicating an elastic behavior
0.43a 0.45d 0.44a (Fig. 5A–B). The highest G′ and G” were found for samples crystallized
SSHE HIU 27.94 ± 41.81 ± 14.45 ± with HIU at 26 ◦ C in the SSHE. For G”, no other condition besides 26 ◦ C
0.44a 0.92a 1.59a
with HIU in the SSHE showed the same high G” value. Nevertheless,
32 Batch control 27.73 ± 41.62 ± 13.38 ± sonication in batch or in the SSHE at 26 ◦ C showed statically similar G′
0.71a 0.94a 1.05a values (p > 0.05). At 26 ◦ C sonicated samples in batch or in the SSHE
Batch HIU 28.17 ± 37.34 ± 14.36 ±
0.16a 0.25d 1.11a
presented a higher G’ (p < 0.05), however the improvement in G′ due to
SSHE control 28.31 ± 42.13 ± 14.06 ± HIU in batch was not significantly different from the sample without
0.53a 0.82a 0.25a HIU in the SSHE. This was also true for G′ and G” at 28 ◦ C and G′ at 30 ◦ C
SSHE HIU 28.20 ± 41.15 ± 13.59 ± (p > 0.05). At 32 ◦ C, the sample crystallized in the SSHE without HIU
0.38a 0.30a 1.91a
had a higher G′ compared to the samples crystallized at the same Tc with
*Means followed by a common letter are not significantly different by the Tukey- HIU in batch. The high shear and efficient heat transfer in the SSHE
test at the 5% level of significance. explain the improvement in physical properties such as G’, G”, and OBC,
observed in samples crystallized using a SSHE compared to a batch
temperature, low driving force, and low residence time in the SSHE was system.
not enough to create a significant amount of nuclei before collecting the Similar to the tendency observed for OBC, G′ and G”, samples crys
sample from the SSHE. This low number of nuclei hampered the for tallized in batch were always improved with sonication, similar to re
mation of a strong network during storage such as the one formed at sults previously reported for other fat sources (da Silva, Cooper et al.,
lower temperatures. 2020a; Ye & Martini, 2015). When samples were crystallized at 30 and
Sonicated samples in batch showed a significant improvement in 32 ◦ C in the SSHE, sonication reduced sample viscoelasticity (p < 0.05).
Unlike tendencies observed in OBC, the significant reduction in G′ due to
7
T.L.T. da Silva et al. LWT 138 (2021) 110593
Fig. 7. Melting behavior obtained after crystallization in batch or in a SSHE at different temperatures with and without HIU and stored at 25 ◦ C for 48h.
sonication in the SSHE was already observed at 28 ◦ C. This behavior p > 0.05) but both sonication and SSHE crystallization affected the
could be attributed to excessive sonication for this crystallization con melting profiles (Fig. 7) and Tp (Table 3). All control samples in batch
dition, which also prove the better effect of HIU on inducing secondary showed one main melting peak and a small shoulder at lower temper
nucleation, since effective improvement of G’ was found only at 26 ◦ C, atures (grey filled arrow).
where primary nucleation was already established (da Silva and Martini, When HIU was applied in batch, two behaviors were observed: (1) at
2020b). lower temperatures (26 and 28 ◦ C) the shoulder became insignificant
and (2) a reduction in Tp to values between 37 and 38 ◦ C was observed.
3.5. Hardness At 30 and 32 ◦ C a reduction in Tp due to sonication was also observed
and Tp was like the ones obtained at 26 and 28 ◦ C. However, the melting
Similar to the results discussed for G’, the hardest samples were profile obtained at 30 ◦ C showed an additional shoulder (non-filled grey
obtained in the SSHE with HIU at 26 ◦ C, followed by the sonicated arrow) which was more evident at 32 ◦ C. It is possible that sonication
samples crystallized in batch (26–30 ◦ C) and the non-sonicated samples melted some of the growing crystals at 30 and 32 ◦ C and that nucleation
crystallized in the SSHE (Fig. 6). The softer samples were non-sonicated occurred in these samples when stored at 25 ◦ C and not during pro
and crystallized in batch. cessing. Nuclei formed during storage generated a crystalline network
For samples crystallized in batch, sonicated samples were always different from the one generated during processing which is evidenced
harder than non-sonicated ones at all Tc (p < 0.05), as previously found as shoulders in the melting profile.
for different fat sources (da Silva, Cooper et al., 2020a; Suzuki et al., The non-sonicated samples crystallized in the SSHE at 26 ◦ C showed
2010). For samples crystallized in the SSHE, harder samples were ob a melting profile similar to the non-sonicated in batch with a shoulder
tained with sonication at 26 ◦ C while softer samples were obtained at 28 around 32 ◦ C and a main peak at the same Tp as the sonicated sample in a
◦
C and no difference in hardness was obtained at 30 and 32 ◦ C. The lack batch, confirming that SSHE promoted similar changes in the sample in
of effect of sonication in hardness values in a continuous system was also a continuous system than HIU in batch system. No changes were
found for a milk fat (Gregersen et al., 2019). This might have happened observed in the melting profile or in Tp after sonication in the SSHE
for two reasons:(1) a slight increase in temperature due to sonication suggesting that even though cavitation might have increased sample
melted the few crystals formed in the SSHE at high temperatures (28–32 temperature as discussed above, nucleation formed in the SSHE at 26 ◦ C
◦
C). It is possible that when the sample was crystallized at lower tem was sufficient to allow HIU bubbles to induce secondary nucleation and
peratures, such as 26 ◦ C, more crystals were present and the slight in improve other physical properties without changing the amount of
crease in temperature generated during sonication was not sufficient to crystalline material (same ΔH and SFC) or melting behavior. Similar
melt the crystals; (2) insufficient nuclei were present in the sample at results were found using different HIU conditions in a SSHE for a
high temperatures and HIU was not able to induce secondary crystalli soybean-based fat (da Silva & Martini, 2019).
zation. In addition, storage at a lower temperature (5 ◦ C) resulted in a At 28 ◦ C the non-sonicated SSHE sample had an additional shoulder
higher supercooling that can hinder the effects generated during at higher temperatures 45 ◦ C (black arrow), even though this sample had
sonication. the same melting parameters (Ton, Tp and ΔH) as the sonicated ones in
batch. After sonication, this shoulder was less pronounced. Similar re
sults were observed by Kadamne, Ifeduba, Akoh, and Martini (2017a)
3.6. Melting behavior
where sonication promoted incorporation of high-melting point TAGs
into the crystalline matrix forming a more organized crystalline network
Melting enthalpies and Ton were not significantly different (Table 3,
8
T.L.T. da Silva et al. LWT 138 (2021) 110593
of a stearic acid-based interesterified sample. When the sample was Declaration of competing interest
crystallized at 28 ◦ C few nucleation sites were generated in the SSHE and
these shoulders represent nuclei and crystals formed during storage. The authors declare no conflict of interest, and that this paper was
During sonication, the presence of these peaks were not as pronounced, not submitted to any other journal besides LWT- Food Science and
suggesting that sonication helped form a more uniform crystalline Technology.
network compared to the non-sonicated samples.
Although samples showed the same Ton at 30 and 32 ◦ C, a broader Acknowledgements
melting profile with various shoulders was observed and Tp was similar
to non-sonicated samples in batch, which confirms our theory of static This project was supported and approved by the Utah Agricultural
crystallization under storage. Due to the similarity in the melting profile Experiment Station (USA) as project number #9293. Financial support
observed at 30 and 32 ◦ C we believe that almost all crystallization for for this research was provided by University Technology Acceleration
these temperatures occurred during storage and that the bubbles Grant (UTAG) program (Utah Science Technology and Research
generated during sonication were not able to induce crystallization (USTAR) initiative, USA) as grant number 18065UTAG001.
during storage.
References
4. Conclusion
AOCS. (2009). American oil chemists’ society (5th ed.). Official Methods and
Recommended Practices of the American Oil Chemists’ Society.
Crystallization in batch and in a continuous sheared system (SSHE) Chen, F., Zhang, H., Sun, X., Wang, X., & Xu, X. (2013). Effects of ultrasonic parameters
generated different crystalline materials. HIU can act as a crystallization on the crystallization behavior of palm oil. Journal of the American Oil Chemists’
modifier in both crystallization systems. In general, sonication induced Society, 90(7), 941–949. https://doi.org/10.1007/s11746-013-2243-y
DeMan, J. (1998). Functionality of palm oil. Journal of Food Lipids, 5, 159–170.
secondary nucleation by inducing more nucleation or formation of Graef, V. De, Puyvelde, P. Van, Goderis, B., & Dewettinck, K. (2009). Influence of shear
smaller nuclei. This effect forms a more uniform crystal network that is flow on polymorphic behavior and micro- structural development during palm oil
more elastic, harder, and capable of entrapping more liquid oil. When crystallization. European Journal of Lipid Science and Technology, 111, 290–302.
https://doi.org/10.1002/ejlt.200800181
samples are processed in batch a significant effect of HIU is observed in Gregersen, S. B., Frydenberg, R. P., Hammershøj, M., Dalsgaard, T. K., Andersen, U., &
all supercoolings tested, which can be attributed to the lower efficiency Wiking, L. (2019). Application of high intensity ultrasound to accelerate
of the batch process on inducing nucleation, and on the fact that HIU, crystallization of anhydrous milk fat and rapeseed oil blends. European Journal of
Lipid Science and Technology, 121(1), 1–9. https://doi.org/10.1002/ejlt.201800200
was always applied at the onset of crystallization, which is not always
Kadamne, J. V., Ifeduba, E. A., Akoh, C. C., & Martini, S. (2017a). Sonocrystallization of
the case in the SSHE. Nevertheless, when samples were crystallized in a interesterified fats with 20 and 30% C16:0 at sn-2 position. Journal of the American
SSHE, heat transfer is more efficient, and an improved crystallization is Oil Chemists’ Society, 94(1), 3–18. https://doi.org/10.1007/s11746-016-2914-6
observed even in the absence of sonication. However, if the crystalli Kadamne, J. V., Ifeduba, E. A., Akoh, C. C., & Martini, S. (2017b). Sonocrystallization of
interesterified fats with 20 and 30% of stearic acid at the sn-2 position and their
zation settings are sufficient to induce a significant number of nuclei in physical blends. Journal of the American Oil Chemists’ Society, 94(8), 1045–1062.
the SSHE before sonication, as observed at 26 ◦ C, HIU is efficient at https://doi.org/10.1007/s11746-017-3014-y
improving physical properties in the SSHE even further. Results from Sato, K., & Ueno, S. (2011). Crystallization, transformation and microstructures of
polymorphic fats in colloidal dispersion states. Current Opinion in Colloid and
this study show that the best values of OBC, G’, G”, and hardness were Interface Science, 16(5), 384–390. https://doi.org/10.1016/j.cocis.2011.06.004
obtained when samples were crystallized and sonicated at 26 ◦ C in the da Silva, T. L. T., Cooper, Z., Lee, J., Gibon, V., & Martini, S. (2020a). Tailoring
SSHE. The SSHE used is a good representation of what is used nowadays crystalline structure using high-intensity ultrasound to reduce oil migration in a low
saturated fat. Journal of American Oil Chemists’ Society, 97, 141–155. https://doi.org/
in industry for fat crystallization and the results found in the study 10.1002/aocs.12321
suggest that it is possible to combine sonication with a SSHE for a low- da Silva, T. L. T., Danthine, S., & Martini, S. (2020b). Effect of processing conditions as
saturated palm-based fat and obtain better physical properties than in a high-intensity ultrasound, agitation, and cooling temperature on the physical
properties of a low saturated fat. Journal of Food Science, Early View. https://doi.org/
batch lab-scale. 10.1111/1750-3841.15436
da Silva, T. L. T., & Martini, S. (2019). Crystallization of interesterified soybean oil using
CRediT authorship contribution statement a scraped surface heat exchanger with high intensity ultrasound. Journal of Food
Engineering, 263, 341–347. https://doi.org/10.1016/j.jfoodeng.2019.07.017. July.
da Silva, T. L. T., & Martini, S. (2020a). Incorporation of high intensity ultrasound (HIU)
Thais Lomonaco Teodoro da Silva: Conceptualization, Methodol to a scraped surface heat exchanger: Effect of HIU position. Journal of Food
ogy, Formal analysis, Investigation, Validation, Writing - review & Engineering, 274, 109824. https://doi.org/10.1016/j.jfoodeng.2019.109824
editing, was responsible for develop the conceptualization and meth da Silva, T. L. T., & Martini, S. (2020b). Sonocrystallization of a palm-based fat with low
level of saturation in a scraped surface heat exchanger. Journal of the American Oil
odology. Made the formal analysis, investigation and validation and also Chemists’ Society. https://doi.org/10.1002/aocs.12409. Early view.
wrote, revised & edited the manuscript. Sabine Danthine: Investiga Suzuki, A. H., Lee, J., Padilla, S. G., & Martini, S. (2010). Altering functional properties of
tion, Writing - review & editing, was responsible for investigation, fats using power ultrasound. Food Engineering and Physical Properties, 75(4),
208–214. https://doi.org/10.1111/j.1750-3841.2010.01572.x
revision and edition of the manuscript. Silvana Martini: Supervision, Ueno, S., Ristic, R. I., Higaki, K., & Sato, K. (2003). In situ studies of ultrasound-
was the Supervisor, Project administration, Funding acquisition, stimulated fat crystallization using synchrotron radiation. The Journal of Physical
Conceptualization, Validation, Writing - review & editing, responsible Chemistry B, 107(21), 4927–4935. https://doi.org/10.1021/jp027840f
Ye, Y., & Martini, S. (2015). Application of high-intensity ultrasound to palm oil in a
for project administration and funding acquisition, also participated on continuous system. Journal of Agricultural and Food Chemistry, 63(1), 319–327.
conceptualization, validation and manuscript writing and revising steps. https://doi.org/10.1021/jf505041s