XXXI International Mineral Processing Congress 2024 Proceedings/Washington, DC/Sep 29–Oct 3 2287
accuracy of these values often depends on a single experi-
ence (Erbil 2014).
It follows that the wettability of a solid surface is a very
challenging parameter for the recovery of ores in flotation,
particularly during the attachment stage. If a surface is wet-
table by the solution present in the cell, the probability that
the particle in question will be recovered is very low. As
can be argued from Figure 2, the higher a solid surface has
a surface energy compared to the tension energy of a liq-
uid, the more the liquid tends to wet this surface at the
expense of their own mutual interactions. On the contrary,
where the tension energy of the liquid is higher, hydropho-
bic behavior is observed. Thus, to promote to any ore lib-
eration, it is mandatory to understand of the interaction
between mineral surface and water.
In Silico Computational Technology to Gain Insights
on Flotation Processes
In the flotation industry (Farrokhpay et al. 2020), one of
the challenges is to find inexpensive and environmentally
friendly materials that selectively adsorb upon surfaces and
interfaces. In this respect, a fundamental step consists of
acquiring a predictive understanding of the thermodynam-
ics for both molecular and dissociative adsorption on dif-
ferent substrates. To this aim, it is essential to be able to
compute the binding energy of different adsorption modes
with sufficient accuracy to predict the adhesion on the sur-
face at molecular level. Hence, discerning descriptors that
can accurately capture the nature of the molecule-surface
interaction is highly desirable.
Computer simulations, especially when extended,
provide access to detailed mechanistic information on
solvent configurations and may ascertain crucial dynami-
cal events over the adsorption process. Nowadays, in silico
computational technology allows the simulation of flota-
tion processes that can be engineered for specific reagents
under appropriate chemical and hydrodynamic conditions.
Computation throughput screening can prevent both
the high cost of experiments and the related risks to the
environment. Quantum chemistry may significantly help
in disclosing the microscopic features at the solid/liquid
interface as it can screen with high accuracy the selected
molecules of interest. This screening can be enhanced by
machine learning (ML) calculations, which can suggest
solidophilic reagents (He et al. 2022) able to improve the
flotation efficiency. Since modern atomistic simulations
are accurate but computationally expensive, active learning
force field (ALFF) techniques (Bernstein et al. 2019) may
open a new era in materials design, as they may be based
on quite simple descriptors—such as the electronegativity
of the substrate, the ionization potential of the molecule
or the contact angle on the surface, to name a few—for
both the substrate and the adsorbed molecule. Overall, ML
screening techniques may shed new light on cutting-edge
flotation reagents.
THEORETICAL METHODS
Molecular Simulations
Nowadays, theoretical modeling relies on classical molec-
ular dynamics (CMD) and ab initio molecular dynamics
(AIMD) simulations. The former requires a reasonable
time scale to satisfy the ergodic hypothesis, yet it may over-
come the limits of infinitesimal time and space during the
adsorption since based on the principles of Newtonian
mechanics. The latter explicitly evaluates the full electronic
structure using density functional theory (DFT), (Marx
and Hutter 2009) effectively gauging the reactive events at
the interface. However, despite the progress made for both
classical (Cisneros et al. 2016) and ab initio simulations
(Gillan et al. 2016), when it comes to interpret the struc-
tural properties of water/solid interactions provided by the
experiments, CMD fail quite often in providing the micro-
scopic interface picture (Jakub et al. 2021). On the other
hand, although first-principles techniques would allow for
extended simulations, their high computational cost repre-
sents a severe bottleneck in gauging complex interplay of
many phenomena at the interface, such as surface diffusion
or water exchange. Thus, AIMD simulations are limited to
short times of the order of a few picoseconds and are inher-
ently restricted to relatively small system sizes.
Active Machine Learning
First-principles-based potential energy surfaces may extend
the simulation time to nanoseconds and have been dem-
onstrated to be quite effective in predicting adsorption
energies (Ras et al. 2013), especially when active learn-
ing techniques to generate the training data set are used
(Bernstein et al. 2019). Although nowadays approaches
Figure 2. Representation of the contact angle between bubble
air, mineral solid and liquid. correspond to the solid-air,
solid-liquid and air-liquid tensions, respectively
accuracy of these values often depends on a single experi-
ence (Erbil 2014).
It follows that the wettability of a solid surface is a very
challenging parameter for the recovery of ores in flotation,
particularly during the attachment stage. If a surface is wet-
table by the solution present in the cell, the probability that
the particle in question will be recovered is very low. As
can be argued from Figure 2, the higher a solid surface has
a surface energy compared to the tension energy of a liq-
uid, the more the liquid tends to wet this surface at the
expense of their own mutual interactions. On the contrary,
where the tension energy of the liquid is higher, hydropho-
bic behavior is observed. Thus, to promote to any ore lib-
eration, it is mandatory to understand of the interaction
between mineral surface and water.
In Silico Computational Technology to Gain Insights
on Flotation Processes
In the flotation industry (Farrokhpay et al. 2020), one of
the challenges is to find inexpensive and environmentally
friendly materials that selectively adsorb upon surfaces and
interfaces. In this respect, a fundamental step consists of
acquiring a predictive understanding of the thermodynam-
ics for both molecular and dissociative adsorption on dif-
ferent substrates. To this aim, it is essential to be able to
compute the binding energy of different adsorption modes
with sufficient accuracy to predict the adhesion on the sur-
face at molecular level. Hence, discerning descriptors that
can accurately capture the nature of the molecule-surface
interaction is highly desirable.
Computer simulations, especially when extended,
provide access to detailed mechanistic information on
solvent configurations and may ascertain crucial dynami-
cal events over the adsorption process. Nowadays, in silico
computational technology allows the simulation of flota-
tion processes that can be engineered for specific reagents
under appropriate chemical and hydrodynamic conditions.
Computation throughput screening can prevent both
the high cost of experiments and the related risks to the
environment. Quantum chemistry may significantly help
in disclosing the microscopic features at the solid/liquid
interface as it can screen with high accuracy the selected
molecules of interest. This screening can be enhanced by
machine learning (ML) calculations, which can suggest
solidophilic reagents (He et al. 2022) able to improve the
flotation efficiency. Since modern atomistic simulations
are accurate but computationally expensive, active learning
force field (ALFF) techniques (Bernstein et al. 2019) may
open a new era in materials design, as they may be based
on quite simple descriptors—such as the electronegativity
of the substrate, the ionization potential of the molecule
or the contact angle on the surface, to name a few—for
both the substrate and the adsorbed molecule. Overall, ML
screening techniques may shed new light on cutting-edge
flotation reagents.
THEORETICAL METHODS
Molecular Simulations
Nowadays, theoretical modeling relies on classical molec-
ular dynamics (CMD) and ab initio molecular dynamics
(AIMD) simulations. The former requires a reasonable
time scale to satisfy the ergodic hypothesis, yet it may over-
come the limits of infinitesimal time and space during the
adsorption since based on the principles of Newtonian
mechanics. The latter explicitly evaluates the full electronic
structure using density functional theory (DFT), (Marx
and Hutter 2009) effectively gauging the reactive events at
the interface. However, despite the progress made for both
classical (Cisneros et al. 2016) and ab initio simulations
(Gillan et al. 2016), when it comes to interpret the struc-
tural properties of water/solid interactions provided by the
experiments, CMD fail quite often in providing the micro-
scopic interface picture (Jakub et al. 2021). On the other
hand, although first-principles techniques would allow for
extended simulations, their high computational cost repre-
sents a severe bottleneck in gauging complex interplay of
many phenomena at the interface, such as surface diffusion
or water exchange. Thus, AIMD simulations are limited to
short times of the order of a few picoseconds and are inher-
ently restricted to relatively small system sizes.
Active Machine Learning
First-principles-based potential energy surfaces may extend
the simulation time to nanoseconds and have been dem-
onstrated to be quite effective in predicting adsorption
energies (Ras et al. 2013), especially when active learn-
ing techniques to generate the training data set are used
(Bernstein et al. 2019). Although nowadays approaches
Figure 2. Representation of the contact angle between bubble
air, mineral solid and liquid. correspond to the solid-air,
solid-liquid and air-liquid tensions, respectively