Regina
Rusch‡
a,
Yasamin
Mohebi Satalsari‡
b,
Angel B.
Zuccolotto-Bernez
b,
Manuel A.
Escobedo-Sánchez
b and
Thomas
Franosch
a
aInstitut für Theoretische Physik, Universität Innsbruck, Technikerstraße 21-A, 6020 Innsbruck, Austria. E-mail: thomas.franosch@uibk.ac.at
bCondensed Matter Physics Laboratory, Heinrich Heine University, Universitätsstraße 1, 40225 Düsseldorf, Germany. E-mail: escobedo@hhu.de
First published on 12th May 2025
We investigate the dynamics of individual colloidal particles in a one-dimensional periodic potential using the intermediate scattering function (ISF) as a key observable. We elaborate a theoretical framework and derive formally exact analytical expressions for the ISF. We introduce and analyze a generalized ISF with two wave numbers to capture correlations in a periodic potential beyond the standard ISF. Relying on Bloch's theorem for periodic systems and, by solving the Smoluchowski equation for an overdamped Brownian particle in a cosine potential, we evaluate the ISF by numerically solving for the eigenfunctions and eigenvalues of the expression. We apply time-dependent perturbation theory to expand the ISF and extract low-order moments, including the mean-square displacement, the time-dependent diffusivity, and the non-Gaussian parameter. Our analytical results are validated through Brownian-dynamics simulations and experiments on 2D colloidal systems exposed to a light-induced periodic potential generated by two interacting laser beams.
In order to understand how colloidal particles move within these periodic potentials, researchers have analyzed the probability distribution, the first-passage time, and low-order moments, such as the mean-square displacement, the long-time diffusivity, and the non-Gaussian parameter.4,6,13,14 Although these observables provide valuable information, most theoretical studies have focused only on diffusion coefficients in the short- and long-time limits.15–17 More recent work has expanded on these studies by analytically exploring tilted washboard potentials,18,19 comparing theoretical predictions with experimental results,20 and investigating memory effects in such systems.21,22 These low-order moments are useful for identifying deviations from free Brownian motion, revealing effects such as trapping and non-Gaussianity.23,24 However, they offer only a limited perspective of the system dynamics, as they do not capture the full spatial-temporal evolution of particle motion. To achieve a more complete description, it is necessary to adopt a framework that incorporates both spatial and temporal correlations.
One such framework emerges naturally in Markovian systems, where the future state depends only on the present and not on the past. In these systems, all relevant dynamical information is contained in the propagator, which describes the probability of a particle transitioning between states over a given time interval. Although the propagator provides a full statistical description of particle motion, it is often challenging to access experimentally. A more practical and experimentally accessible alternative is the intermediate scattering function (ISF), which encodes both spatial and temporal correlations in particle motion. Unlike traditional low-order moments, the ISF offers a more comprehensive characterization of the system and can be directly measured using techniques such as dynamic light scattering,25,26 differential dynamic microscopy,27 and single-particle tracking.28
Furthermore, upon a small-wave-vector expansion of the ISF the low-order moments are recovered. Analytical derivations of the ISF have been achieved for various systems, including anisotropic active Brownian particles,23,29 Brownian circle swimmers in gravitational fields,30–32 anisotropically diffusing colloidal dimers,33 and run-and-tumble agents.34,35 Notably, experimental validations have been conducted for colloidal dimers and active colloids.
In this work, we extend the traditional ISF by introducing a generalized version that incorporates two wave numbers, allowing us to investigate the correlations in periodic systems in reciprocal space. The central question we address is twofold. First, we develop a theoretical framework to gain analytic insight into the dynamics of colloidal particles in general periodic potentials. Second, we compare our predictions with experimental results, exploring a spatio-temporal regime that has not yet been fully investigated.
To develop the theoretical framework, we solve the Smoluchowski equation for a single, overdamped Brownian particle in a cosine potential. Reformulating the problem as a Hermitian Schrödinger equation allows expressing the solutions in terms of eigenvalues and eigenfunctions.36 Taking advantage of the systems periodicity, we apply Bloch's theorem, which provides a systematic way to expand the ISF using time-dependent perturbation theory and extract key dynamical moments. These methods are based on previous work31,32,37 and have been successfully applied to bistable periodic potentials.38,39 To validate our theoretical predictions, we compared our theoretical framework with Brownian dynamics simulations.
Experimentally, we track individual colloidal particles confined between two walls in a two-dimensional configuration, subjected to a periodic potential created by the interference of two laser beams. By systematically varying the laser power, we control the potential amplitude and record particle trajectories under different conditions. The generalized ISF, along with its associated moments, is then extracted from these experimental data and directly compared with our theoretical predictions, enabling a precise evaluation of the Brownian motion description in periodic optical fields. By integrating theoretical modeling, experimental measurements, and computational simulations, this work aims to provide a comprehensive understanding of colloidal motion in structured environments. Our approach bridges the gap between fundamental stochastic dynamics and experimentally accessible observables, offering new insights into the interplay between Brownian motion and periodic potentials.
This work is organized as follows. In Section 2 the experimental materials and methods are described. In Section 3, we introduce the analytical model, and derive the theoretical framework and observables. Readers more interested in the results rather than the theoretical derivation may skip this section and jump directly to Section 4, where we present our findings, compare experimental results with theoretical predictions, and discuss their implications. Finally, we summarize our findings, concluding with an outlook on future research directions in this field in Section 5.
To numerically solve the equation of motion, as it will be explained in detail in the theory Section 3, we employed a spectral method. Specifically, we expanded the equation in its Fourier basis to obtain its operator form. Numerically, we truncated the expansions to form a finite-dimensional matrix eigenvalue problem. We used the scipy.linalg.eig function from the SciPy library44 to compute the eigenvalues and eigenvectors of the resulting matrix. The convergence of our results was verified by systematically increasing the truncation order until the eigenvalues stabilized to the desired precision.
The imaging system is an inverted bright-field microscope (Nikon Ti-E) with a 20× objective (Nikon Plan Flour, 0.5 NA). We illuminate the sample with a blue LED (Thorlabs M455L4) and capture images using a CMOS camera (Mako U-130B) at a resolution of 1280 × 1024 pixels, with a pixel pitch of 0.24 μm per px. To prevent laser light from reaching the camera sensor, we employed a second dichroic mirror (D2) that redirects the laser beams to a beam dump (BD), with any residual laser light being filtered out by a notch filter (NF). This optical arrangement allows us to simultaneously apply the periodic potential and observe particle dynamics with high precision.
This exceptional spatial resolution is crucial for accurately capturing the subtle dynamics of particles within the periodic potential, particularly when comparing experimental results with theoretical predictions. The observables are computed for each measurement and averaged over all the measurements. Results in Section 4, represent the mean values and the error bars their standard deviation.
U(x) = U1![]() | (1) |
ẋ(t) = DQ1u![]() | (2) |
Here, D denotes the short-time diffusion coefficient such that D/kBT is the mobility as derived by the Einstein relation. The stochastic term η(t) represents a centered Gaussian white noise with a delta-correlated variance 〈η(t)η(t′)〉 = 2Dδ(t − t′).
We identify three quantities which set the characteristic units of the system: The period L of the potential is the fundamental unit of length. The time a free particle needs to diffuse the distance of a period is L2/D and will be used as a time unit. Energies are compared to the thermal energy kBT. Therefore, the problem displays the dimensionless amplitude u as a single control parameter.
∂t![]() ![]() ![]() ![]() ![]() | (3) |
A stationary solution is provided by
![]() | (4) |
![]() | (5) |
Explicitly
![]() | (6) |
We replace our infinite system with a finite system by dividing it into N ∈ unit cells of length L and apply periodic boundary conditions. The limit N → ∞ will be performed at the end of the calculations.
In order to find non-trivial solutions of eqn (3) we first perform a separation ansatz
![]() | (7) |
![]() | (8) |
![]() | (9) |
Left and right eigenfunctions to different eigenvalues are orthonormal, satisfying
![]() | (10) |
Therefore, only the product of left and right eigenstates to identical eigenvalue is normalized. Furthermore, the eigenfunctions are complete, fulfilling the condition
![]() | (11) |
By comparing eqn (3) and (8) for the stationary state, we identify that the eigenfunction ψR0(x) to eigenvalue zero has to be proportional to the equilibrium distribution ψR0(x) ∝ peq(x). We choose
ψR0(x) = peq(x), ψL0(x)* = 1, | (12) |
The formal solution of the Smoluchowski, eqn (3), can be written as = eΩtδ(x − x0). We can insert the completeness relation, eqn (11), and apply the eigenvalue equation, eqn (8), and obtain
![]() | (13) |
![]() ![]() | (14) |
A straightforward calculation reveals that this yields in general to
![]() | (15) |
By specializing to the potential given in eqn (1), we find
![]() | (16) |
We note that the operator 0 is Hermitian with respect to the standard scalar product in eqn (9). Equivalently to the procedure of finding solutions of the Smoluchowski operator, to find the non-trivial solution of eqn (16) we perform again a separation ansatz
![]() | (17) |
![]() | (18) |
![]() | (19) |
As 0 is an Hermitian operator, the eigenvalues are real and two eigenfunctions with different eigenvalue are orthonormal, equivalently to eqn (10). The eigenfunctions are complete, fulfilling the completeness relation, similarly to eqn (11). The eigenfunction to eigenvalue zero can be easily found using eqn (12) and (19)
![]() | (20) |
Ψnq(x) = eiqxunq(x), | (21) |
![]() | (22) |
![]() | (23) |
![]() | (24) |
![]() | (25) |
![]() | (26) |
![]() | (27) |
![]() | (28) |
For the actual computation of the eigenfunctions we use the Fourier modes as orthonormal basis {|ν〉: ν ∈ } in
with real-space representation
. It is favorable to express the Bloch functions unq(x) in terms of their Fourier decomposition, and we express our eigenmodes in Dirac notation
![]() | (29) |
![]() | (30) |
This is possible because all Bloch functions are lattice periodic.
Fμν(q,t) := 〈e−i(q+Qμ)x(t)ei(q+Qν)x(0)〉, | (31) |
![]() | (32) |
Further, we note that the wave number q is identical in both exponentials and that the diagonal elements, Fμμ(q,t), correspond to the conventional ISF evaluated at wave vector q + Qμ. In translationally invariant systems only the diagonal elements are non-vanishing, since shifting the trajectory of a particle by an arbitrary displacement leads to an equally likely trajectory. For our case, the discrete symmetry is reflected in the sense that a common shift x(t) ↦ x(t) + R for all times by a lattice vector R ∈ Λ leaves the ISF invariant. The brackets 〈⋯〉 indicate an ensemble average.
Using the conditional probability density, the ISF of eqn (31) can be expressed as
![]() | (33) |
Here, we used that without restriction, the initial position of the particle can be chosen to be in a definite cell and is sampled from the equilibrium distribution for this single cell.
Reversely, we can compute the probability density by the backward Fourier transform
![]() | (34) |
The explicit form of eqn (33) in terms of the Bloch functions is obtained by inserting eqn (24) and simplifies to express the stationary solution in terms of the eigenfunctions of the Schrödinger operator, using eqn (25). After some algebra and rearranging the terms, we find the final expression
![]() | (35) |
To determine the functions unq(x) we introduce the operator q for which the eigenvalue equation
![]() | (36) |
![]() | (37) |
The matrix representation of the operator, is given by
![]() | (38) |
〈μ|δ![]() | (39) |
The matrix 0 is a Hermitian matrix and pentadiagonal in the Fourier basis, i.e., it has non-zero elements only on the main diagonal and the two diagonals above and below it. The q-dependent matrix is diagonal in the Fourier basis. The eigenvalue problem
q|unq〉 = −λnq|unq〉 is computed numerically by diagonalizing the (truncated) Hermitian matrix
![]() | (40) |
The time evolution of the ISF is encoded in the eigenvalues and eigenfunctions of the operator q. The eigenvalues λnq form continuous bands as the number of cells goes to infinity, N → ∞, see Fig. 2. All eigenvalues are non-negative, and the only zero eigenvalue is in the center of the Brillouin zone at the lowest band. Only the lowest bands are significantly affected by the potential U(x). For λ ≳ π2Du/L2 the bands correspond to a particle freely diffusing without underlying potential modulation. Albeit the bands come very close at the edges of the Brillouin zone, we checked numerically that they do not touch. For symmetric potentials the avoided crossing theorem does not apply and in principle bands can cross. For the simple cosine potential, one can actually show that all eigenvalues for q = 0 except for λ00 = 0 are twofold degenerate, see Appendix C.
![]() | ||
Fig. 2 Eigenvalues λnq of the operator ![]() |
Finally, the ISF in eqn (35) can be conveniently expressed in a spectral representation using the Fourier basis
![]() | (41) |
Of particular interest is the conventional ISF with wave vector in the first Brillouin zone, F(q,t) := F00(q,t), which simplifies upon exploiting the completeness of the Fourier basis
![]() | (42) |
Since all eigenvalues are strictly larger than zero, except for λ00 = 0 in the lowest band and in the center of the Brillouin zone, F(q,t) decays to zero for large time, t → 0 for q ≠ 0.
Fμν(q,t) = Fμν(q, −t) = Fνμ(q,t)*. | (43) |
For symmetric potentials U(−x) = U(x), space-inversion symmetry implies
Fμν(q,t) = Fμν(q,t)* = F−μ,−ν(−q,t). | (44) |
Fμν(π/L,t) = Fμ+1,ν+1(−π/L,t) = F−(μ+1),−(ν+1)(π/L,t). | (45) |
![]() | (46) |
From eqn (35) and , one infers that for q = 0 the ISF displays a non-vanishing long-time limit
Fμν(0,t → ∞) = 〈e−iQμx(t)〉〈eiQνx(0)〉. | (47) |
The factorization of the limit can be interpreted as the system being ergodic. For the simple cosine potential, eqn (1), the limit can be calculated explicitly
![]() | (48) |
An equivalent formal expression for the long-time limit follows from eqn (41)
![]() | (49) |
Since and employing the Jacobi–Anger expansion,50 the Fourier coefficients 〈μ|u00〉 can be calculated explicitly
![]() | (50) |
With Neumann's addition theorem for modified Bessel functions50 the sums in eqn (49) can be performed and we recover eqn (48). Note that eqn (46) and (48) represent static quantities that only depend on the potential amplitude.
![]() | (51) |
The long-time dynamics is diffusive, in particular, the long-time limit D∞ := D(t → ∞) > 0 is finite and defines the long-time diffusion constant. An analytic expression for arbitrary periodic potentials is known.15,16,18,22,51 In particular, for the simple cosine potential it evaluates to
![]() | (52) |
The exponential suppression of the diffusion constant reflects Kramers's rule for hopping over a potential barrier.36 A convenient measure to discuss deviations from simple diffusion is the non-Gaussian parameter47,52
![]() | (53) |
The derivation for extracting the moments from the ISF is similar to ref. 31 and 32 with only the essential steps presented here. The key idea is to apply perturbation theory for small wave numbers q and compare it to the Taylor series of the ISF, which yields the moments
![]() | (54) |
![]() | (55) |
Replacing the time-evolution operator in the integral on the right-hand side iteratively generates the Born series, see ref. 32. The main simplification steps are to make use of the fact that e0t|un0〉 = |un0〉 and also 〈un0|e
0t = 〈un0| and to insert complete basis sets, eqn (26), for q = 0. Occurring integrals can be formally evaluated and the terms are simplified to obtain the final result, which is similar to the result in ref. 32, but slightly changed for our operator and eigenvectors. We find the formal expression
![]() | (56) |
![]() | (57) |
Here, all sums over n, m, p include all bands and therefore formally the expression causes divisions by zero if a band index corresponds to the lowest band or two band indices correspond to the same band. In both cases the corresponding numerators also vanish. The appearance of the zero divisors can be avoided in the first place by handling these case separately before performing the integrals in the simplification steps. Here we follow a different route to keep the expressions simple by analytically continuing the expression for the case of zero numerators/denominators.
A further complication arises in the case of a simple cosine potential, since all eigenvalues λn0 are, additionally, twofold degenerate, except for the ground state causing additional zero divisors. However, as in degenerate perturbation theory, one can choose basis states such that the matrix elements of δ1q coupling different states to the same eigenvalues vanish. Since
1q anticommutes with space inversion, only states of different parity couple, however, because un0 are either even or odd, no zero divisors occur.
The low-order cumulants of the random variable Δx(t) are generated upon expanding the logarithm of the ISF in powers of the wave number q
![]() | (58) |
To order O(q2) we find the mean-square displacement as first nonvanishing cumulant
![]() | (59) |
![]() | (60) |
![]() | (61) |
For the fourth cumulant we need to collect terms of order O(q4) and we obtain
〈[Δx(t)]4〉 − 3〈[Δx(t)]2〉2 = ![]() ![]() | (62) |
For completeness, let us argue explicitly that all odd powers in q in the expansion of F(q,t), eqn (56), vanish. A term linear in q corresponding to a mean drift would involve the matrix element 〈u00|δ1q|u00〉 which is shown to vanish in eqn (60). This vanishing of the mean drift is, of course, a general property in equilibrium. For any symmetric potential all odd moments vanish. The expansion of F(q,t) in eqn (56) generates a chain 〈u00|δ
1q|un0〉 〈un0|…|up0〉 〈up0|δ
1q|u00〉 in eqn (57) of products of matrix elements. By parity and the property of the operator δ
1q a matrix element is non-vanishing only if the states are of different parity. Thus, for the chain to yield a non-vanishing contribution, the first intermediate state has to be odd, the second even, and so on. However, the last state is the ground state again which is even. Therefore, only even powers of q are generated.
We first focus on the behavior for q ≠ 0, where all ISF eventually relax to zero. For moderate potential amplitudes U1 ≈ kBT, the potential is not high enough to significantly inhibit hopping between different potential valleys, yielding a single-step relaxation. For larger amplitudes, a two-step process occurs. The particle initially freely diffuses with short-time diffusion coefficient D until the potential forces become dominant. For U1 ≳ kBT the motion occurs essentially at the bottom of the potential, which can be approximated by a harmonic well
![]() | (63) |
The particle then locally equilibrates on the time scale of the harmonic relaxation time, τ = (L2/4π2D)(kBT/U1), and the ISF saturates at a plateau value, see Fig. 3. The ISF within the harmonic approximation can be calculated explicitly, see Appendix E. For large barriers, Fig. 3(c), the harmonic approximation quantitatively describes the relaxation to the plateau value for wave numbers resolving smaller length scale than a period L. The relaxation from the plateau occurs on a much larger time scale provided by Kramers’ theory τK ∝ exp(2U1/kBT). Once the particle overcomes the barrier and reaches additional minima, the ISF eventually decays to zero. For small wave numbers and long times, the hydrodynamic regime is reached F00(q,t) = exp(−D∞q2t). In this regime, the wave numbers only resolve the motion over many periods at time scales much larger than Kramers’ escape time. Our analytical predictions as well as the simulation results within the Smoluchowski picture of a simple cosine potential show excellent agreement with the experimental results.
For the wave number in the center of the BZ, q = 0, the ISF does not decay to zero in the long-time limit t → ∞, rather approaches a finite value, as computed in eqn (48). In Appendix A we used this feature to calibrate the laser power of the experiment to the theoretical kBT value.
For wave numbers at the edge of the BZ q = π/L additional symmetries of the ISF, eqn (45), hold. For example, the curves for (μ, ν) = (1, −1) and (0, −2), or (1,0) and (−1, −2) are identical.
For q = 0 both the initial value and the long-time limit are non-zero. As soon as q ≠ 0 the curves decay to zero for long times. If |μ − ν| is odd, the initial value is negative and for even differences the initial value of the ISF is positive, eqn (46). And also according to eqn (48) for odd (or even) values, the long-time limit is negative (or positive, respectively). In contrast to the diagonal elements of the ISF, we find no longer strictly monotone behavior but minima and maxima, see eqn (32).
Results from the experimental data of the off-diagonal ISF closely follow the shape of the analytical predictions and simulation results at long times. However, a clear deviation is observed at short and intermediate times, as shown in Fig. 4. We attribute these discrepancies to experimental factors not implemented in the theoretical model. We consider the primary factor to be the spatial inhomogeneities in the periodicity and amplitude of the light-induced potential across the field of view, which slightly deviates from the ideal cosine form assumed in the theoretical model. As a result, individual particles effectively experience slightly different periodicities and amplitudes across the field of view. A detailed characterization of the amplitude and periodicity in the field of view is shown in Appendix B. Additionally, confinement effects due to the two-dimensional nature of the system and inertial effects not considered in the theoretical framework are factors that might also contribute to these discrepancies. Despite these differences, experimental results for the indices μ and ν with equal |μ − ν| collapse into a common intercept (short-time limit), capturing the expected phenomenology from the theoretical predictions given by eqn (46), though with a noticeable shift respect to theory.
Theoretically, each Fourier component of the ISF is defined in terms of a well-defined lattice wavenumber 2π/L. However, in experiments, achieving this level of precision is challenging because of variations in how individual particles interact with the optical field. To allow a more accurate comparison between theory and experiment, a normalization procedure is applied, an approach that is further examined in Fig. 5.
Fig. 5 illustrates the normalized ISF, Fμν(q,t)/Fμν(q,0), for different potential amplitudes U1/kBT, three different μ, ν combinations and, for all of them, qL = π. The choice of qL = π is particularly insightful, as it balances, for the measurements time window, sensitivity to both free diffusion and potential-induced localization, providing a clear distinction between different transport regimes. As it can be seen in Fig. 5, after normalization, the agreement between experimental results, the theoretical framework and computer simulations is remarkable, confirming that the model effectively captures full description of the system.
The most striking feature in the off-diagonal elements is the emergence of a maximum at intermediate times, τ ≪ t ≪ τK as seen in Fig. 5(b) and (c). For larger times, the curves decay to zero for q ≠ 0. From the harmonic approximation we anticipate the development of a plateau, whose value can be determined from eqn (77). If (Qμ + q)(Qν + q) < 0, the plateau corresponds to a maximum, which is nicely approached for large potential amplitudes, see Fig. 5(b) and (c). If (Qμ + q)(Qν + q) > 0, the curves look similar to the diagonal elements of the ISF, where a simple plateau emerges, see Fig. 5(a). The slowing down of the relaxation towards the plateau or maximum as the potential amplitude grows, is captured as well by τ ∝ 1/U1.
Finally, it is important to emphasize that, although experimental factors cause deviations at short and intermediate times, normalization effectively accounts for these variations, leading to excellent agreement between theory, simulations and experiments.
Furthermore, we analyze the non-Gaussianity of the particle displacements using the parameter defined in eqn (53). As expected, for higher barriers, the particle dynamics become increasingly non-Gaussian, see Fig. 6(c). We observe that both very small and very large amplitudes pose challenges in experiments. For small amplitudes, it is difficult to distinguish the dynamics from those of a free particle, as the external potential has little effect. Conversely, for very large amplitudes, the low diffusivity makes it challenging to sample a sufficient number of particles that successfully hop over a barrier within the experimental observation time. However, the experimental results show excellent agreement with both simulations and analytical predictions.
Based on the Smoluchowski equation reformulated in a Hermitian Schrödinger form, we found formal expressions in terms of a spectral-theoretical approach. The eigenfunctions were expressed in Bloch form, to make use of the periodic nature of the system. We found an analytic expression for the generalized ISF, Fμν(q,t). By using the time-dependent perturbation theory and Taylor expansion of the ISF we computed lower-order moments. In our system, without memory effects, the ISF effectively captures the full dynamics of colloidal particles in periodic potentials, revealing both short- and long-time diffusive behavior and trapping at intermediate times.
We performed experiments on 2D dilute colloidal suspensions subjected to a periodic potential generated by two interfering laser beams. Using particle tracking, we obtained particle trajectories and averaged them to extract relevant observables. The laser power was calibrated to its corresponding amplitude value using two theoretical predictions (eqn (48) and (52)). From our experimental data, we extracted the observables of interest and compared them to our analytical solutions and Brownian-dynamics simulations. We compared the results for various strengths of the amplitude of the potential and found excellent agreement between the theoretical description and experiments. The most sensitive observables were the off-diagonal elements of the generalized ISF, where slight differences in the experimental setup were amplified in the curves. It was crucial to ensure equivalent experimental conditions, minimizing variations in periodicity, potential amplitude, and confinement effects. To obtain good agreement with the analytical predictions, a normalization was necessary.
We have carried out a comprehensive test of the underlying fundamental dynamics of colloidal dynamics in a structured environment, combining theory, simulations, and experiments. By analyzing the generalized ISF, we identified new observables with distinct features, including a non-vanishing long-time limit. We provide a detailed theoretical framework and rationalize our findings through a harmonic approximation. Developing an explicit formula for the whole-time dependence of the ISF allowing for a comprehensive description of colloidal motion across all time scales. Furthermore, we introduced a new approach for calibrating the experiment using these observables, offering a reciprocal space alternative to conventional calibration methods. Comparing the results to a harmonic approximation, we confirm that the Brownian particle first diffuses freely, before it is temporarily trapped in the minima of the periodic cosine potential. For these times the dynamics are well approximated by a harmonic potential for large enough amplitudes, and only at longer times it hops over the potential barriers and once again exhibits diffusive behavior.
The analytical and experimental framework presented can be extended to more complex systems. Although this work focused on dilute suspensions, exploring more dense systems would allow us to study particle interactions and many-body effects. A possible other extension is the study of periodic lattices, where higher-dimensional effects and collective behavior become important. Investigating tilted washboard potentials could provide further insight into driven transport. Our framework is also applicable to a wide range of periodic systems beyond simple cosine potentials. Experimentally, the new observables could also be measured using differential dynamic microscopy.
Our alternative calibration method utilizes the generalized ISF, specifically its asymptotic behavior, which explicitly depends on the potential amplitude. For a simple cosine potential and q = 0, the long-time limit of the generalized ISF is analytically described by eqn (48), where the dependence on the potential amplitude, U1, is evident. Specifically, for μ = ν = 1, eqn (48) (F11(0,t)) evaluates the characteristic wave number imposed on the system by the periodic potential, i.e., Qμ = Qν = 2π/L. In the top panel of Fig. 7 we plot F11(0,t), which exhibits an opposite trend to diffusivity, with higher plateaus as laser power increases. It is important to note that this approach utilizes equilibrium correlation properties instead of transport characteristics, providing complementary insights into the system.
The calibration is performed for both methods by extracting plateau values as a function of laser power and solving eqn (48) and (52) for the diffusivity and generalized ISF methods, respectively. The calibration results from both methods are presented in Fig. 7(b), showing the relationship between laser power and dimensionless potential amplitude U1/kBT. The remarkable agreement between these independent approaches validates both the theoretical framework and the experimental implementation, as seen in the central panel of Fig. 6 and 7(a). This strong consistency confirms the theoretical predictions and demonstrates the reliability of this method for calibration purposes.
![]() | (64) |
Since
![]() | (65) |
This property is somewhat hard to see in the representation of the time-evolution operator in the Schrödinger representation in the Fourier basis 〈μ|0|ν〉. However, the property can be easily deduced, omitting the gauge transform in the first place, i.e. representing the dynamics in terms of the non-Hermitian matrix
![]() | (66) |
This matrix displays the symmetry 〈−μ|Ω|−ν〉 = 〈μ|Ω|ν〉*. The argument now follows the one of Appendix A of ref. 32. The matrix 〈μ|Ω|ν〉 displays a zero row for μ = 0 and splits into a part with entries for μ > 0, ν ≥ 0 and an identical one for μ < 0, ν ≤ 0.
The only entries preventing the matrix to split into blocks with positive/negative μ, ν are the matrix elements 〈±1|Ω|0〉. However, as 〈l0| = 〈0| is a left eigenvector to eigenvalue 0, all eigenvectors |rn0〉 to non-zero eigenvalues have a zero entry in their Fourier representation by orthogonality of eigenvectors 0 = 〈l00|rn0〉 = 〈0|rn0〉. Therefore, the blocks with both μ,ν positive do not communicate with the blocks with both indices negative. In particular, one can choose eigenvectors with 〈μ|rn〉 = 0 for μ ≷ 0 or symmetric and antisymmetric eigenfunctions to the twofold degenerate eigenvalue λn0 > 0.
![]() | (67) |
Here in the second equality we used that the Bloch function to wave vector zero at the lowest band is related to the equilibrium density . Furthermore, we observed that the integral vanishes for q ≠ q′ due to the periodicity of the Bloch functions.
We can make further progress by using the Fourier modes as basis functions, eqn (29). By expanding the Bloch functions we obtain
![]() | (68) |
Collecting terms, we find the expression for the ISF
![]() | (69) |
This relation is eqn (41) in the main text.
We also derive how the probability density can be obtained from the intermediate scattering function
![]() | (70) |
This is eqn (34) of the main text.
![]() | (71) |
![]() | (72) |
![]() | (73) |
![]() | (74) |
We can readily compute the generalized ISF by using the definition of the main text, eqn (33), and by extending the integrals to infinity
![]() | (75) |
Solving the integrals then yields the ISF
![]() | (76) |
We further calculate the ratio
![]() | (77) |
We also readily find the MSD
![]() | (78) |
and time-dependent diffusion coefficient
D(t) = De−t/τ. | (79) |
Footnotes |
† We dedicate this work to the memory of our friend, mentor, and colleague Stefan U. Egelhaaf, with sincere gratitude for his contributions and dedication to the field of soft matter. |
‡ R. R. and Y. S. contributed equally to this work. |
This journal is © The Royal Society of Chemistry 2025 |