Can we live Danckwerts' dream? Mixing Analysis in a Baffled Stirred Tank Reactor Based on 4D-Particle Tracking Experiments
Listen
Radio episode about this paper
Transcript
Introduction to the show: ident: Physics. The day in general physics: instruments, methods and results across the field.
Kai: We're looking at the paper: "Can we live Danckwerts' dream? Mixing Analysis in a Baffled Stirred Tank Reactor Based on 4D-Particle Tracking Experiments".
Mira: Here's the gist of the paper. We present an experimental investigation of mixing dynamics within a laboratory-scale 3-liter stirred tank reactor (STR) equipped with two Rushton turbines and three baffles Using time-resolved, four-dimensional particle tracking velocimetry, we successfully capture trajectories of up to 40,000 tracer particles in the full reactor volume despite obstructions by stirrer and baffles From these Lagrangian data, we analyze velocities, accelerations, and spatial dispersion, revealing anisotropic mixing By utilizing novel network-based analysis methods on the experimental particle trajectories, we identify coherent fluid compartments that exhibit strong internal mixing but weak exchange with neighboring compartments We uncover five distinct compartments acting as transport barriers, which have a high impact on substrate distribution in chemical and biochemical processes Our approach thus realizes and extends early thought experiments from Danckwerts and Levenspiel by providing detailed insight into the behavior of single fluid parcels and Lagrangian mixing withing chemical and biochemical reactors.
Kai: Experimental Setup The experiments were carried out in a three liter stirred tank reactor (STR) with three baffles and two Rushton turbines The STR is a handmade glass vessel with an inner diameter of dR = 130 mm and a Korbbogen head bottom.
Mira: The fluid is stirred with two Rushton turbines and diverted by three equidistant flow breaking baffles.
Kai: The vessel was equipped with two Rushton turbines powered by a MINISTAR 80 control.
Paper summary: Kai: To recap, this paper presents an experimental investigation of mixing dynamics inside a three liter stirred tank reactor. The core thesis is that using time-resolved four-dimensional particle tracking velocimetry allows researchers to capture trajectories for up to forty thousand tracer particles in the full reactor volume, even with obstructions like stirrers and baffles.
Mira: The claim they are making is that from this detailed Lagrangian data, they can analyze the velocities and accelerations and reveal anisotropic mixing patterns. Beyond just seeing where things move, they use novel network-based analysis on those particle trajectories to identify coherent fluid compartments.
Lev: What matters for us is that these compartments aren't just abstract mathematical constructs; they are physical volumes that exhibit strong internal mixing but poor exchange with neighboring compartments, and the paper identifies five of these distinct compartments that significantly impact substrate distribution in chemical and biochemical processes.
Kai: It really pushes the Danckwerts dream forward by showing that high-resolution particle tracking is now achievable in a lab setting for systems with real obstructions, which was previously too complex to measure at this detail.
Mira: This matters because it gives us a tool to understand how internal geometry dictates mixing efficiency, moving beyond simple assumptions about perfect bulk flow in stirred tanks.
Lev: For those of us building error correction systems, understanding these localized transport barriers is vital because the overhead and performance of our hardware depend on how effectively we can control substrate distribution within the system.
Kai: So in essence, this work realizes an early thought experiment by providing detailed insight into the behavior of single fluid parcels undergoing Lagrangian mixing within a chemical reactor environment.
Mira: It’s about understanding that flow isn't just one uniform thing; it’s a complex network of interacting volumes with different transport properties.
Lev: And that complexity is what we need to model correctly if we want to build something that actually performs well under non-ideal conditions, which this study seems to address by quantifying those barriers.
Conclusion: Kai: Looking at the full picture of this paper, "Can we live Danckwerts' dream? Mixing Analysis in a Baffled Stirred Tank Reactor Based on 4D-Particle Tracking Experiments," it’s about taking that old idea of tracking molecules and applying modern high-speed particle tracking to solve a practical problem in chemical reactors <ref:2607.26813#pg1,Mixing Analysis in a Baffled Stirred Tank Reactor Based on 4D-Particle>.
Mira: The authors did this by using specific hardware—a three liter tank with two Rushton turbines and three baffles—and they proved that with enough data, we can map out exactly how the fluid moves and where those mixing compartments are located.
Lev: So what this means practically is that we’ve gained a way to quantify transport limitations within a reactor before we ever have to build the full system, which is a big step for hardware design.
Kai: It shows that even in relatively simple lab setups, like this one, you can get detailed information about spatial dispersion and how it affects what happens chemically or biochemically inside.
Mira: It’s less about predicting every single molecule's path over years and more about understanding the characteristic behavior of fluid parcels within a defined space.
Lev: And for the real world, it means that when we design systems, we need to incorporate these localized mixing barriers into our models, not just assume everything is perfectly mixed everywhere.
Kai: It’s an important piece of evidence showing how experimental setups can reveal underlying complexities in fluid dynamics that theoretical models might miss if they aren't informed by this kind of detailed tracking data.
Mira: So the title itself speaks to the ambition—trying to bring that detailed, molecular-level view into a practical engineering context for mixing analysis.
Lev: And the implication is that we can start building better tools and models for chemical and biochemical systems because we now have a quantifiable way to measure those transport barriers.
Faculty of Sustainable Engineering, Hamburg University of Applied Science · Institute of Mathematics and its Didactics, Leuphana University Lünburg · Institute of Multiphase Flows, Hamburg University of Technology
physics.flu-dyn
Submitted: 2026-07-29
Updated: 2026-07-29
Comments: 22 pages, 15 figures, research data are made freely available, see https://doi.org/10.15480/882.17469
DOI: 10.1016/j.cej.2026.183917
License: http://creativecommons.org/licenses/by/4.0/
Importance score: 81/100
The gist: We present an experimental investigation of mixing dynamics within a laboratory-scale 3-liter stirred tank reactor (STR) equipped with two Rushton turbines and three baffles Using time-resolved…
Key concepts
- Four-Dimensional Particle Tracking Velocimetry (4D-PTV)
- This technique involves recording the position of many tracer particles in three dimensions over time. By tracking these paths across four dimensions (3 spatial coordinates plus time), researchers could accurately capture the movement of up to 40,000 particles within a complex reactor setup, even with obstructions like stirrers and baffles.
- Lagrangian Velocity
- This measures how fast an individual fluid parcel moves at any specific point in space and time. By calculating the change in position over time for every tracked particle, researchers could determine the velocity of each particle individually, which is crucial for understanding local fluid motion.
- Network-Based Mixing Analysis
- This method models the reactor as a network where particles are connected if they are close enough. By creating a weighted matrix based on these connections over time, researchers can identify 'coherent volumes'—groups of particles that move together and exchange fluids effectively, defining distinct mixing compartments.
Terminology
Summary
We present an experimental investigation of mixing dynamics within a laboratory-scale 3-liter stirred tank reactor (STR) equipped with two Rushton turbines and three baffles Using time-resolved, four-dimensional particle tracking velocimetry, we successfully capture trajectories of up to 40,000 tracer particles in the full reactor volume despite obstructions by stirrer and baffles From these Lagrangian data, we analyze velocities, accelerations, and spatial dispersion, revealing anisotropic mixing By utilizing novel network-based analysis methods on the experimental particle trajectories, we identify coherent fluid compartments that exhibit strong internal mixing but weak exchange with neighboring compartments We uncover five distinct compartments acting as transport barriers, which have a high impact on substrate distribution in chemical and biochemical processes Our approach thus realizes and extends early thought experiments from Danckwerts and Levenspiel by providing detailed insight into the behavior of single fluid parcels and Lagrangian mixing withing chemical and biochemical reactors. p. 0 Mixing Analysis in a Baffled Stirred Tank
Reactor Based on 4D-Particle Tracking Experiments
Eike Steuwea,∗
, Thanh Tung Thaib
, Christian Weilandc
, Anna Klünkerb
, Jan Hendrick Nissena
, Michael
Schlüterc
, Kathrin Padberg-Gehleb
, Alexandra von Kamekea,∗
aFaculty of Sustainable Engineering, Hamburg University of Applied Science
b
Institute of Mathematics and its Didactics, Leuphana University Lüneburg
c
Institute of Multiphase Flows, Hamburg University of Technology
Abstract
We present an experimental investigation of mixing dynamics within a laboratory-scale 3-liter stirred tank
reactor (STR) equipped with two Rushton turbines and three baffles. Using time-resolved, four-dimensional
particle tracking velocimetry, we successfully capture trajectories of up to 40,000 tracer particles in the full
reactor volume despite obstructions by stirrer and baffles, providing unprecedented time-resolved flow and
mixing information.Where the paper says it
Experimental Setup
The experiments were carried out in a three liter stirred tank reactor (STR) with three baffles and two Rushton turbines The STR is a handmade glass vessel with an inner diameter of dR = 130 mm and a Korbbogen head bottom. The fluid is stirred with two Rushton turbines and diverted by three equidistant flow breaking baffles. p. 2 25
Z
X
44°
45°
48°
44°
702
674
728
1311
766
LED 300
2x2 LED 35
4x LED 35
Illumination Unit
2,3°
5x phantom
veo 640 L
Y
Y
Z
X
Z
X
n27
9
10
36
7
76
n130
13
230
36
120°
3x
25
n8
water level 2x rushton turbine
3x baffle
korbbogen head
3
Figure 1: Technical drawing of the STR used and positioning of the cameras surrounding the STR during the measurement. The fluid is stirred with two Rushton turbines and diverted by three equidistant flow breaking baffles. Three 3D-printed baffles, made of black
polylactid acid (PLA), were arranged at 120 degree intervals along the reactor wall. The vessel
was equipped with two Rushton turbines powered
by a MINISTAR 80 control (IKA-Werke GmbH &
Co. For illumination, blue
high-power LEDs (one LaVision LED300 and eight
LaVision LED35s), with wavelengths of 445 nm,
were synchronized with the frame rate of the cameras. The STR was filled with VR = 3 L bidistilled
water at room temperature (22 ◦C), resulting in a
water level height of hR = 230 mm. For illumination, blue
high-power LEDs (one LaVision LED300 and eight
LaVision LED35s), with wavelengths of 445 nm,
were synchronized with the frame rate of the cameras. The STR was filled with VR = 3 L bidistilled
water at room temperature (22 ◦C), resulting in a
water level height of hR = 230 mm.Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Data Processing and Trajectory Reconstruction
The NP particle trajectories xP (t) ∈ R cubed, P = 1,..., NP are obtained at discrete times tk ∈ T by finite differencing along all particle trajectories in time. p. 5 The Lagrangian velocities vP (t) = dxP (t)
dt
∈ R
3
,
P = 1, . , NP for each particle are evaluated at
discrete times tk ∈ T by finite differencing along
all particle trajectories in time, using a difference
quotient of 4th order accuracy [26]
vP (tk) ≈
1
12δt · (xP (tk−2) − 8xP (tk−1)
+ 8xP (tk+1) − xP (tk+2)) ,
k = 3, . Varying the step size allows the algorithm to detect and track both fast- and slow-moving particles
within the investigated volume making the recording of highly dynamic flows possible. The resulting particle trajectories were exported from DaVis
as machine- and human-readable . The resulting particle trajectories were exported from DaVis
as machine- and human-readable . dat files and imported into MATLAB. The correct
position was determined by visually inspecting the
projection of the particles onto the three principal
coordinate planes. To further improve the quality
of the data set, a custom-made MATLAB algorithm
was used to reconnect interrupted particle trajectories. Through visual inspection of the trajectory
differences before and after reconnection, no falsely
reconnected trajectories were found. After reconnecting, the trajectories are each smoothed using
the MATLAB implementation of a Savitzky-Golay
filter [25] with a kernel size of 15 time steps and
an order of three.Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Lagrangian Velocity and Eulerian Field Derivation
The Lagrangian velocities vP (t) = dxP (t) dt ∈ R cubed, P = 1,..., NP for each particle are evaluated at discrete times tk ∈ T by finite differencing along all particle trajectories in time. p. 5 The Lagrangian velocities vP (t) = dxP (t)
dt
∈ R
3
,
P = 1, . , NP for each particle are evaluated at
discrete times tk ∈ T by finite differencing along
all particle trajectories in time, using a difference
quotient of 4th order accuracy [26]
vP (tk) ≈
1
12δt · (xP (tk−2) − 8xP (tk−1)
+ 8xP (tk+1) − xP (tk+2)) ,
k = 3, . From the Lagrangian particle velocity the Eulerian
velocity v(x, t) = (vx(x, t), vy(x, t), vz(x, t)) at the
particle positions xP (t) can be derived by
v(xP (t), t) = vP (t) . Thus, the Eulerian velocity field at a specific time
6 By evaluating
the scatteredInterpolant at evenly distributed
points, a Eulerian velocity field can be created for
an evenly distributed cartesian grid. Then the Eulerian acceleration field and spatial velocity field gradients from which insightful measures such as the
energy dissipation rate ϵ = 2 · TT · ηf/ρf can be
calculated from the square of the symmetric strain
rate tensor
TT =
∂vx
∂x 2
+
∂vy
∂y 2
+
∂vz
∂z 2
+
∂vx
∂y +
∂vy
∂x 2
+
∂vy
∂z +
∂vz
∂y 2
+
∂vz
∂x +
∂vx
∂z 2
2
and the dynamic viscosity ηf and density ρf of
the fluid under the measurement conditions [28].Where the paper says it
Where the paper says it
Where the paper says it
Network-Based Mixing Analysis
At each time instance tk an instantaneous binary adjacency matrix A(tk) ∈ 0, 1 NP is constructed with entries aij (tk) = (1 if xi(tk) − xj (tk)2 ≤ εr, i, j 0 otherwise. By summing up the instantaneous adjacency matrices A(tk) for all time steps, the weighted network matrix W ∈ NP×NP 0 with entries wij = X NT k=1 aij (tk) is obtained. p. 7 SciPy’s
sparse_distance_matrix implements such an ϵrneighborhood search based on KDTrees. By summing up the instantaneous adjacency matrices A(tk) for all time steps, the weighted network
matrix W ∈ N
NP×NP
0 with entries
wij =
X
NT
k=1
aij (tk)
is obtained. The entry wij is therefore large if the trajectories of the
two different particles i and j stay close for a long
time or if the two particles repeatedly meet in terms
of ϵr-close encounters. If these particles are viewed as fluid parcels, these
will have plenty of time for their liquids to interchange and to react, if initially carrying different reactants. Thus, they will rapidly approach
a similar level of reactant concentrations. We use
this characteristic to define coherent volumes, here
called compartments, that travel together later in
the text.Where the paper says it
Where the paper says it
Where the paper says it
Identification of Coherent Compartments
The eigenvalue problem D−1W u = λu is equivalent to the generalized eigenvalue problem considered in [35] and forms the basis for the identification of coherent clusters, as explained in the following. The sign structure of u2 for the eigenvalue λ2 < 1 can then be used to partition the underlying network into two clusters [35]. p. 8 We assume
that the network is connected and that λ1 = 1 is
a simple eigenvalue. The sign structure of u2 for
the eigenvalue λ2 < 1 can then be used to partition the underlying network into two clusters [35]. Thus, they will rapidly approach
a similar level of reactant concentrations. We use
this characteristic to define coherent volumes, here
called compartments, that travel together later in
the text. The RMS of the difference between
this extrapolation and the actual position is determined for all uninterrupted trajectories. The final
threshold used for the matching step is the 99th
percentile of the RMS values multiplied by a factor
of five. tillations in the particle images while keeping the
whole trajectory in its original position and shape. As shown in Figure 4a, due to the nature of the
VT-STB algorithm, fewer trajectories are found at
the beginning and end of the investigated time interval than on average for the other time steps. In the 4 mm slice through the
middle of the reactor shown in Figure 5a the stirrer
shaft, the two Rushton turbines and one baffle on
the left-hand side are visible as white voids, which
shows that no ghost particles were found there erroneously. Figure 5b shows a snapshot of the trajectories of all particles found in the entire STR over
the course of six stirrer rotations (∆t = 0. By summing up the instantaneous adjacency matrices A(tk) for all time steps, the weighted network
matrix W ∈ N
NP×NP
0 with entries
wij =
X
NT
k=1
aij (tk)
is obtained. This symmetric matrix implements
temporal information of how often a particle was
within the ϵr range of every other particle, thus encoding the number of total ϵr-close encounters over
the evaluated time interval ∆t = [t1, tNT
].Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.
Improvements for AI systems
-
Improved AI system can perform Lagrangian transport analysis by using network-based methods on experimental particle trajectories to identify coherent fluid compartments, such as
five distinct compartments acting as transport barriers.
-
The improved system can predict mixing, dissolution and reaction processes by utilizing the network-based analysis methods derived from tracking Lagrangian particles, which
have the potential to predict mixing, dissolution and reaction processes.
-
The AI can analyze velocity gradients and energy dissipation rates by evaluating Eulerian fields derived from Lagrangian trajectories, allowing it to
estimate the stresses that a microorganism experiences on its journey through the reactor.
-
The system can characterize flow inhomogeneities by identifying regions of high mixing efficiency and
macromixing hindering compartments
using spectral clustering based on the weighted node degree matrix W. -
The improved AI can develop feeding strategies to mitigate substrate inhomogeneities in biochemical reactors by specifying the
five coherent compartments
identified through spectral clustering, which are located between the baffles.
Abstract
We present an experimental investigation of mixing dynamics within a laboratory-scale 3-liter stirred tank reactor (STR) equipped with two Rushton turbines and three baffles. Using time-resolved, four-dimensional particle tracking velocimetry, we successfully capture trajectories of up to 40,000 tracer particles in the full reactor volume despite obstructions by stirrer and baffles, providing unprecedented time-resolved flow and mixing information. From these Lagrangian data, we analyze velocities, accelerations, and spatial dispersion, revealing anisotropic mixing. By utilizing novel network-based analysis methods on the experimental particle trajectories, we identify coherent fluid compartments that exhibit strong internal mixing but weak exchange with neighboring compartments. We uncover five distinct compartments acting as transport barriers, which have a high impact on substrate distribution in chemical and biochemical processes. Our approach thus realizes and extends early thought experiments from Danckwerts and Levenspiel by providing detailed insight into the behavior of single fluid parcels and Lagrangian mixing withing chemical and biochemical reactors, offering a valuable approach for evaluation and optimization of chemical and biochemical processes. The trajectory data are made freely available to serve as an experimental reference for further research.
Related papers
- Emergent Transfer of a Physics Foundation Model from Simulation to Laboratory Turbulence
- Early onset of secondary shear instability in Kelvin-Helmholtz braids at high Reynolds number
- Rotational covariance, continuous spectrum and non-modal growth of three-dimensional planar weak viscous shock layers
- Deep Learning-based Algebraic Reynolds Stress Closures for RANS Simulations of Turbulent Flows
- Neptuna: A Comprehensive Machine Learning Framework for Benchmarking Complex Multiphase Flows
- Size and shape of condensing steam jets and plumes