Nanoplastics in Water: Artificial Intelligence-Assisted 4D
Physicochemical Characterization and Rapid In Situ Detection
Zi Wang, Devendra Pal, Abolghasem Pilechi, and Parisa A. Ariya*
Cite This: Environ. Sci. Technol. 2024, 58, 8919−8931 Read Online
ACCESS Metrics & More Article Recommendations *sı Supporting Information
ABSTRACT: For the first time, we present a much-needed technology for the in situ and real-time detection of nanoplastics in
aquatic systems. We show an artificial intelligence-assisted nanodigital in-line holographic microscopy (AI-assisted nano-DIHM) that
automatically classifies nano- and microplastics simultaneously from nonplastic particles within milliseconds in stationary and
dynamic natural waters, without sample preparation. AI-assisted nano-DIHM identifies 2 and 1% of waterborne particles as nano/
microplastics in Lake Ontario and the Saint Lawrence River, respectively. Nano-DIHM provides physicochemical properties of single
particles or clusters of nano/microplastics, including size, shape, optical phase, perimeter, surface area, roughness, and edge gradient.
It distinguishes nano/microplastics from mixtures of organics, inorganics, biological particles, and coated heterogeneous clusters.
This technology allows 4D tracking and 3D structural and spatial study of waterborne nano/microplastics. Independent transmission
electron microscopy, mass spectrometry, and nanoparticle tracking analysis validates nano-DIHM data. Complementary modeling
demonstrates nano- and microplastics have significantly distinct distribution patterns in water, which affect their transport and fate,
rendering nano-DIHM a powerful tool for accurate nano/microplastic life-cycle analysis and hotspot remediation.
KEYWORDS: four dimensional, real-time, in situ, nanoplastics, physicochemical characterization, deep neural network, predictive model,
life cycle
1. INTRODUCTION
Nanoplastics enter the environment through direct emissions
and secondary breakdowns.1 Microplastic degradation may
generate an increased number of nanoplastic particles, by 2 to
3 orders of magnitude.2 Due to their small sizes and light
weights, nanoplastics possess substantial potential for extensive
dispersion in natural surroundings.3,4 Nanoplastics exist in
snow, air, polar ice, and subtropical ocean gyres.3,5−7
Waterborne nano/microplastics can reshape the aqueous
carbon storage and contribute to global warming; while
airborne nano/microplastics can alter the Earth’s energy
budgets and affect climate change.8−11
Environmental nano- (<1 μm) and microplastics (<1 mm)
have emerged as significant global concerns.8,12,13 Particle- and
mass-based techniques allow particle imaging and chemical
composition analysis of nanoplastics, like surface-enhanced
Raman spectroscopy, near-edge X-ray absorption fine-structure
spectroscopy, thermal-desorption proton-transfer-reaction
mass spectrometry, and so forth (Note S1 and Table
S1).3,5,14−16 Yet, there is currently no technology for in situ
and real-time physicochemical processes of nano/microplastics
in natural waters, precluding accurate nano/microplastic life-
cycle analysis in the Earth’s aquatic systems.8,12,13,17,18
Nanoplastics possess distinct physicochemical characteristics
compared to microplastics, which can result in disparate fates
and toxicities.1,12 For instance, given their colloidal nature,
nanoplastics are more likely to be expelled from sea ice.19 Due
to their high surface-to-volume ratios, nanoplastics tend to
Received: December 11, 2023
Revised: March 3, 2024
Accepted: March 27, 2024
Published: May 6, 2024
Articlepubs.acs.org/est
© 2024 The Authors. Published by
American Chemical Society 8919
https://doi.org/10.1021/acs.est.3c10408
Environ. Sci. Technol. 2024, 58, 8919−8931
This article is licensed under CC-BY-NC-ND 4.0
Physicochemical Characterization and Rapid In Situ Detection
Zi Wang, Devendra Pal, Abolghasem Pilechi, and Parisa A. Ariya*
Cite This: Environ. Sci. Technol. 2024, 58, 8919−8931 Read Online
ACCESS Metrics & More Article Recommendations *sı Supporting Information
ABSTRACT: For the first time, we present a much-needed technology for the in situ and real-time detection of nanoplastics in
aquatic systems. We show an artificial intelligence-assisted nanodigital in-line holographic microscopy (AI-assisted nano-DIHM) that
automatically classifies nano- and microplastics simultaneously from nonplastic particles within milliseconds in stationary and
dynamic natural waters, without sample preparation. AI-assisted nano-DIHM identifies 2 and 1% of waterborne particles as nano/
microplastics in Lake Ontario and the Saint Lawrence River, respectively. Nano-DIHM provides physicochemical properties of single
particles or clusters of nano/microplastics, including size, shape, optical phase, perimeter, surface area, roughness, and edge gradient.
It distinguishes nano/microplastics from mixtures of organics, inorganics, biological particles, and coated heterogeneous clusters.
This technology allows 4D tracking and 3D structural and spatial study of waterborne nano/microplastics. Independent transmission
electron microscopy, mass spectrometry, and nanoparticle tracking analysis validates nano-DIHM data. Complementary modeling
demonstrates nano- and microplastics have significantly distinct distribution patterns in water, which affect their transport and fate,
rendering nano-DIHM a powerful tool for accurate nano/microplastic life-cycle analysis and hotspot remediation.
KEYWORDS: four dimensional, real-time, in situ, nanoplastics, physicochemical characterization, deep neural network, predictive model,
life cycle
1. INTRODUCTION
Nanoplastics enter the environment through direct emissions
and secondary breakdowns.1 Microplastic degradation may
generate an increased number of nanoplastic particles, by 2 to
3 orders of magnitude.2 Due to their small sizes and light
weights, nanoplastics possess substantial potential for extensive
dispersion in natural surroundings.3,4 Nanoplastics exist in
snow, air, polar ice, and subtropical ocean gyres.3,5−7
Waterborne nano/microplastics can reshape the aqueous
carbon storage and contribute to global warming; while
airborne nano/microplastics can alter the Earth’s energy
budgets and affect climate change.8−11
Environmental nano- (<1 μm) and microplastics (<1 mm)
have emerged as significant global concerns.8,12,13 Particle- and
mass-based techniques allow particle imaging and chemical
composition analysis of nanoplastics, like surface-enhanced
Raman spectroscopy, near-edge X-ray absorption fine-structure
spectroscopy, thermal-desorption proton-transfer-reaction
mass spectrometry, and so forth (Note S1 and Table
S1).3,5,14−16 Yet, there is currently no technology for in situ
and real-time physicochemical processes of nano/microplastics
in natural waters, precluding accurate nano/microplastic life-
cycle analysis in the Earth’s aquatic systems.8,12,13,17,18
Nanoplastics possess distinct physicochemical characteristics
compared to microplastics, which can result in disparate fates
and toxicities.1,12 For instance, given their colloidal nature,
nanoplastics are more likely to be expelled from sea ice.19 Due
to their high surface-to-volume ratios, nanoplastics tend to
Received: December 11, 2023
Revised: March 3, 2024
Accepted: March 27, 2024
Published: May 6, 2024
Articlepubs.acs.org/est
© 2024 The Authors. Published by
American Chemical Society 8919
https://doi.org/10.1021/acs.est.3c10408
Environ. Sci. Technol. 2024, 58, 8919−8931
This article is licensed under CC-BY-NC-ND 4.0
adsorb a greater amount of other contaminants.20,21
Furthermore, nanoplastics likely pose more significant threats
to living organisms than microplastics, as they can more easily
penetrate cellular membranes.1,22 Understanding the character-
istics, concentrations, mechanisms, and transport of nano-
plastics is essential for exposure and risk studies.2,23 However,
given the limited data, many current assessments are based on
the knowledge extrapolated from microplastic or nanomaterial
research, which sometimes may not be applied.2,24
Numerical models play a crucial role in predicting the fate
and transport of nano/microplastics, serving as one of the
primary approaches for researchers and decision makers to
understand the behaviors, extents, accumulation zones, and
potential sources of plastic pollution.25,26 The capability of the
models to generate reliable predictions is heavily reliant on
input data quality for model initialization and boundary
conditions. These include details about the concentrations and
characteristics of the modeled particles like size, shape,
sedimentation velocity, aggregation rate, and degradation
rate.27,28 Such information is needed for environmental
nano/microplastic life-cycle analysis, yet it is not available
until this study.27 We use AI-assisted nano-DIHM to acquire
physicochemical data of environmental nano/microplastics and
show by a modeling study the distinct transport pathways and
accumulation zones of nano- and microplastic particles in
waters. We address the United Nations Environment
Programme tackled urgency on in situ and real-time nano/
microplastic detection in aquatic systems to improve environ-
mental assessments and targeted remediation strategies.29,30
Herein, we develop a novel technology, artificial intelligence-
assisted nanodigital in-line holographic microscopy (AI-
assisted nano-DIHM), to fill this knowledge gap.28,31,32 The
principle of nano-DIHM is described in Methods and Note S2.
We present in the following: (1) nano-DIHM identifies nano/
microplastics and other materials based on intensity and
optical phase by manual reconstructions; (2) encompassing a
deep neural network, AI-assisted nano-DIHM automatically
characterize and classify nano- and microplastic particles in
environmental waters without sample preparation; (3) the
three-dimensional (3D) particle spatial/structural and four-
dimensional (4D) dynamic tracking capabilities of nano-
DIHM allow us to address the knowledge gap on real-time and
in situ studies of waterborne nano/microplastics; (4) a model
study demonstrating the distinct distributions of nano- and
microplastics in the natural waters suggests that real-time
nano/microplastic detection is necessary for accurate life-cycle
analysis and targeted remediation. AI-assisted nano-DIHM,
tailored specifically for nano- and microplastic particles, serves
as a powerful tool to tackle the emerging environmental
challenges posed by plastic pollution.
2. METHODS
2.1. Nanodigital In-Line Holographic Microscopy
(Nano-DIHM). In digital in-line holographic microscopy
(DIHM), an interference pattern (or hologram) is formed by
the interaction between a reference wave and the scattered
light from the analytes/objects. The hologram is captured by a
complementary metal-oxide semiconductor (CMOS)-photo-
sensitive matrix sensor. To extract particle morphological
characteristics and obtain 3D data from the recorded 2D
images, numerical algorithms based on the Kirchhoff−
Helmholtz transform are employed in this study.31,33,34
The amplitude A(r, t) and intensity I(r, t) of the hologram
are expressed as= +A r t A r t A r t( , ) ( , ) ( , )ref scat
(1)= *I r t A r t A r t( , ) ( , ) ( , )
(2)
where Aref(r, t) is the amplitude of the reference wave and
Ascat(r, t) is the amplitude of the wave scattered by the analyte.
The intensity expressed in eq 2 can be further expanded as= * + [ *
+ * ] + *
I r t A r t A r t A r t A r t
A r t A r t A r t A r t
( , ) ( , ) ( , ) ( , ) ( , )
( , ) ( , ) ( , ) ( , )
ref ref ref scat
scat ref scat scat
(3)
where the first term represents the intensity of the reference
waves; the second term, also known as the holographic
diffraction pattern, describes the superposition of the
interference between the incident reference wave and the
scattered wave; the third term depicts the interference between
the scattered waves, which also refers to the classical diffraction
pattern.33,34 The amplitude of the scattered hologram is
expressed as= | |
A r iA
r I r ik
r r s( ) ( ) exp( ) d
rr
r
scat ref
(4)
In this work, we used a 4Deep digital in-line holographic
microscope with Octopus software to record and manually
reconstruct holograms. We implement an additional con-
volution−deconvolution route in the coding of the software to
improve the resolution for detecting nanosized particles.31 The
particle size measured from manual reconstructions is
determined based on the full width at half-maximum of the
peak from the intensity crosscut.35 Additional information and
schematic illustration of the principle of DIHM are in Note S2
and Figure S1.
2.2. Detecting Nano-, Micro-, and Nonplastics using
Nano-DIHM in Stationary Mode. The following suspen-
sions are prepared: polyethylene (PE), polypropylene (PP),
polystyrene (PS), polyethylene terephthalate (PET), polyvinyl
chloride (PVC), and polyurethane (PUR) nanoplastic hydro-
sols; suspensions of 0.05 μm polystyrene latex (PSL),
microplastics, phytoplankton, magnetite, titanium dioxide,
oleic acid, and humic acid; a sample of oleic acid-coated PE
nanoplastics; and a mixture of PE nanoplastics, PS nano-
plastics, oleic acid, magnetite, and phytoplankton. See Note S3
for details.
Prepared suspensions are analyzed using nano-DIHM in
stationary mode. For each analyte, a micropipette is used to
transfer 50 μL of the suspension onto a precleaned quartz
microscopic slide. Immediately, a cover slide is applied,
spreading the sample into a thin film.32 For nanoplastic
analysis, a thin microscopic cover slide is used in place of the
conventional microscopic slide. This substitution minimizes
the laser-to-sample distance, enabling improved resolution in
the imaging process.31 For the PE pellet, the particle is directly
placed onto a microscopic slide without a cover slide. The
source-to-camera distance is occasionally adjusted, depending
on the size of the analyte and the desired field of view. For
nanoplastic samples, the source-to-camera distance is 6 mm.
For oleic acid, humic acid, phytoplankton, and the mixture of
nanoplastics and nonplastics, the distance is 15 mm. For all of
the microplastic particles, the distance is 19 mm. The values
Environmental Science & Technology pubs.acs.org/est Article
https://doi.org/10.1021/acs.est.3c10408
Environ. Sci. Technol. 2024, 58, 8919−8931
8920
Furthermore, nanoplastics likely pose more significant threats
to living organisms than microplastics, as they can more easily
penetrate cellular membranes.1,22 Understanding the character-
istics, concentrations, mechanisms, and transport of nano-
plastics is essential for exposure and risk studies.2,23 However,
given the limited data, many current assessments are based on
the knowledge extrapolated from microplastic or nanomaterial
research, which sometimes may not be applied.2,24
Numerical models play a crucial role in predicting the fate
and transport of nano/microplastics, serving as one of the
primary approaches for researchers and decision makers to
understand the behaviors, extents, accumulation zones, and
potential sources of plastic pollution.25,26 The capability of the
models to generate reliable predictions is heavily reliant on
input data quality for model initialization and boundary
conditions. These include details about the concentrations and
characteristics of the modeled particles like size, shape,
sedimentation velocity, aggregation rate, and degradation
rate.27,28 Such information is needed for environmental
nano/microplastic life-cycle analysis, yet it is not available
until this study.27 We use AI-assisted nano-DIHM to acquire
physicochemical data of environmental nano/microplastics and
show by a modeling study the distinct transport pathways and
accumulation zones of nano- and microplastic particles in
waters. We address the United Nations Environment
Programme tackled urgency on in situ and real-time nano/
microplastic detection in aquatic systems to improve environ-
mental assessments and targeted remediation strategies.29,30
Herein, we develop a novel technology, artificial intelligence-
assisted nanodigital in-line holographic microscopy (AI-
assisted nano-DIHM), to fill this knowledge gap.28,31,32 The
principle of nano-DIHM is described in Methods and Note S2.
We present in the following: (1) nano-DIHM identifies nano/
microplastics and other materials based on intensity and
optical phase by manual reconstructions; (2) encompassing a
deep neural network, AI-assisted nano-DIHM automatically
characterize and classify nano- and microplastic particles in
environmental waters without sample preparation; (3) the
three-dimensional (3D) particle spatial/structural and four-
dimensional (4D) dynamic tracking capabilities of nano-
DIHM allow us to address the knowledge gap on real-time and
in situ studies of waterborne nano/microplastics; (4) a model
study demonstrating the distinct distributions of nano- and
microplastics in the natural waters suggests that real-time
nano/microplastic detection is necessary for accurate life-cycle
analysis and targeted remediation. AI-assisted nano-DIHM,
tailored specifically for nano- and microplastic particles, serves
as a powerful tool to tackle the emerging environmental
challenges posed by plastic pollution.
2. METHODS
2.1. Nanodigital In-Line Holographic Microscopy
(Nano-DIHM). In digital in-line holographic microscopy
(DIHM), an interference pattern (or hologram) is formed by
the interaction between a reference wave and the scattered
light from the analytes/objects. The hologram is captured by a
complementary metal-oxide semiconductor (CMOS)-photo-
sensitive matrix sensor. To extract particle morphological
characteristics and obtain 3D data from the recorded 2D
images, numerical algorithms based on the Kirchhoff−
Helmholtz transform are employed in this study.31,33,34
The amplitude A(r, t) and intensity I(r, t) of the hologram
are expressed as= +A r t A r t A r t( , ) ( , ) ( , )ref scat
(1)= *I r t A r t A r t( , ) ( , ) ( , )
(2)
where Aref(r, t) is the amplitude of the reference wave and
Ascat(r, t) is the amplitude of the wave scattered by the analyte.
The intensity expressed in eq 2 can be further expanded as= * + [ *
+ * ] + *
I r t A r t A r t A r t A r t
A r t A r t A r t A r t
( , ) ( , ) ( , ) ( , ) ( , )
( , ) ( , ) ( , ) ( , )
ref ref ref scat
scat ref scat scat
(3)
where the first term represents the intensity of the reference
waves; the second term, also known as the holographic
diffraction pattern, describes the superposition of the
interference between the incident reference wave and the
scattered wave; the third term depicts the interference between
the scattered waves, which also refers to the classical diffraction
pattern.33,34 The amplitude of the scattered hologram is
expressed as= | |
A r iA
r I r ik
r r s( ) ( ) exp( ) d
rr
r
scat ref
(4)
In this work, we used a 4Deep digital in-line holographic
microscope with Octopus software to record and manually
reconstruct holograms. We implement an additional con-
volution−deconvolution route in the coding of the software to
improve the resolution for detecting nanosized particles.31 The
particle size measured from manual reconstructions is
determined based on the full width at half-maximum of the
peak from the intensity crosscut.35 Additional information and
schematic illustration of the principle of DIHM are in Note S2
and Figure S1.
2.2. Detecting Nano-, Micro-, and Nonplastics using
Nano-DIHM in Stationary Mode. The following suspen-
sions are prepared: polyethylene (PE), polypropylene (PP),
polystyrene (PS), polyethylene terephthalate (PET), polyvinyl
chloride (PVC), and polyurethane (PUR) nanoplastic hydro-
sols; suspensions of 0.05 μm polystyrene latex (PSL),
microplastics, phytoplankton, magnetite, titanium dioxide,
oleic acid, and humic acid; a sample of oleic acid-coated PE
nanoplastics; and a mixture of PE nanoplastics, PS nano-
plastics, oleic acid, magnetite, and phytoplankton. See Note S3
for details.
Prepared suspensions are analyzed using nano-DIHM in
stationary mode. For each analyte, a micropipette is used to
transfer 50 μL of the suspension onto a precleaned quartz
microscopic slide. Immediately, a cover slide is applied,
spreading the sample into a thin film.32 For nanoplastic
analysis, a thin microscopic cover slide is used in place of the
conventional microscopic slide. This substitution minimizes
the laser-to-sample distance, enabling improved resolution in
the imaging process.31 For the PE pellet, the particle is directly
placed onto a microscopic slide without a cover slide. The
source-to-camera distance is occasionally adjusted, depending
on the size of the analyte and the desired field of view. For
nanoplastic samples, the source-to-camera distance is 6 mm.
For oleic acid, humic acid, phytoplankton, and the mixture of
nanoplastics and nonplastics, the distance is 15 mm. For all of
the microplastic particles, the distance is 19 mm. The values
Environmental Science & Technology pubs.acs.org/est Article
https://doi.org/10.1021/acs.est.3c10408
Environ. Sci. Technol. 2024, 58, 8919−8931
8920