Multiple propagating modes in random polycrystals
Listen
Radio episode about this paper
Transcript
Introduction to the show: ident: Condensed Matter. The day in condensed matter: materials, phases and the measurements under them.
Kai: I'm Kai.
Mira: And I'm Mira, here with Lev.
Kai: We're looking at the paper: "Multiple propagating modes in random polycrystals".
Title and authors: Kai: So we’ve talked about what the paper claims, and they are showing us that in these random polycrystals, you get multiple propagating modes when the wavelength gets small enough to see the microstructure.
Mira: It’s interesting how they frame it—it’s not just finding more math roots, but observing a physical reality based on energy loss.
Lev: The core of it is that the number of observable modes is set by their attenuation hierarchy, meaning only the modes that lose energy slowly enough to travel far can be seen.
Kai: And they do this by using grain-scale finite-element simulations to look at the coherent field directly in frequency versus wavenumber space.
Mira: So, if you were designing a measurement system for this material, what would you prioritize looking at first based on their findings?
Lev: You’d prioritize the attenuation hierarchy. You’d figure out which roots have a sufficiently low attenuation coefficient over many grain correlation lengths to actually propagate through the medium.
Kai: That leads into the next part: they show how this is achieved by injecting a plane wave and then watching how much coherence survives as it travels deep into the material.
Mira: It’s a direct demonstration of how forward multiple scattering through distinct longitudinal and transverse channels rebuilds any surviving coherent field beyond that first correlation length.
Lev: It shows the mechanism: the initial wave gets extinguished within about one correlation length, and what's left is carried by these specific modes that have survived that initial loss.
Kai: And this applies to both transverse excitations and longitudinal excitations, which is a lot to see in a single study of statistically isotropic polycrystals.
Mira: It really puts the theory into perspective because it moves us from saying "this should happen" to "this is what happens when you look at the data."
Lev: It’s about moving from a purely mathematical prediction to something experimentally verifiable through simulation.
The paper's summary: Kai: So, summarizing the whole thing, the paper explores how multiple scattering theory predicts that a single wave in a statistically isotropic polycrystal isn't really just one wave at all.
Mira: They are saying that the dispersion equation admits multiple roots for any polarization, but these extra modes have never been observed before.
Lev: And they claim that their grain-scale finite-element simulations resolve the coherent field and directly observe this predicted multiplicity for both transverse and longitudinal excitations.
Kai: The key finding is that a single mode at low frequencies gives way to two modes in the geometric regime as the wavelength approaches the grain correlation length.
Mira: This transition point, where it happens near ten MHz for iron, is where you start seeing those two distinct ridges in the measured spectra with different slopes and widths <ref:2610.24047#pg1>.
Lev: And they confirm that this separation is due to two coherent modes with their own distinct phase velocities and attenuation coefficients that are observable.
Kai: They also tie it all back to the fact that the third root always has a faster attenuation throughout, which means you just can’t observe it at any frequency.
Mira: So they conclude that the number of observable modes is set not by the total number of roots, but by their attenuation hierarchy.
Lev: It’s a very clear conclusion because it shifts the focus from counting mathematical solutions to analyzing how much energy those solutions can carry through the medium.
The paper's improvements: Kai: Now for what they suggest as improvements, they show that this observation is not just a fluke in one material, because they demonstrate it in two different cubic materials with common and extreme anisotropy.
Mira: That’s important because if it confirms it across different types of materials, then the multimodal structure becomes a generic property of statistically isotropic polycrystals.
Lev: It suggests that the mechanism is tied to the presence of these constituents rather than some specific material parameter we might be thinking about alone.
Kai: They also show that in the geometric regime, for example, in SrMgSn, mode one speeds up toward faster propagation paths while mode two becomes slower with its propagation dominated by a slow transverse channel.
Mira: That detail about the relative speeds is really interesting because it tells us something about which physical channels are dominant for which modes.
Lev: It gives us information about the relationship between the different scattering channels L, T1, and T2 in a way that's more direct than just looking at a single velocity.
Conclusion: Kai: So wrapping up this paper "Multiple propagating modes in random polycrystals," they confirm that grain-scale simulations resolve two transverse modes and two longitudinal modes when the wavelength is comparable to the grain correlation length.
Mira: The main implication is that we now have a way to predict which of these theoretically predicted roots will actually be observable based on their attenuation hierarchy.
Lev: So, for real hardware, it means we can design things that are robust against scattering by focusing on modes that have long enough propagation lengths to survive the disorder.
Kai: It’s about knowing which waves will persist when you put them into a random medium.
Mira: The paper suggests that this insight is useful not just for polycrystals, but also for other random media like electromagnetic waves in random dielectrics and seismic waves in heterogeneous Earth.
Lev: So it’s applying this principle to things that are much larger than individual grains, which is a big step up.
Kai: And we’re getting ready to look at what's next in the field with these kinds of observations on polycrystals.
Mira: It’s a strong foundation for understanding how disorder affects wave propagation across many different physical systems.
Lev: That’s the takeaway, applying this principle to things that are much larger than individual grains.
Department of Engineering and Design, University of Sussex · Department of Mechanical Engineering, Imperial College London · Department of Materials Science and Engineering, The Ohio State University
cond-mat.mtrl-sci, physics.app-ph
Submitted: 2026-10-16
Updated: 2026-10-16
Comments: 6 pages, 3 figures
Journal ref: Physical Review Letters 137, 217201 (2026)
DOI: 10.1103/dxwt-yhh9
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Importance score: 85/100
The gist: Multiple propagating modes in random polycrystals resolves two transverse modes and two longitudinal modes…
Key concepts
- Geometric Regime
- This regime occurs when the wavelength of the wave becomes comparable to the grain correlation length in a random material. In this specific condition, multiple coherent modes emerge that are resolvable and distinct from a single wave.
- Attenuation Hierarchy
- The number of observable propagating modes is dictated by how quickly they lose energy (attenuation). The root with the lowest attenuation is always observable, and subsequent roots only become visible when their attenuation becomes comparable to others in this geometric regime.
- Dyson Equation Roots
- Solving the Dyson equation for the mean Green's function yields three roots corresponding to three scattering channels (one longitudinal and two transverse). These mathematical solutions predict potential modes, but only those with low enough attenuation are physically observable.
Terminology
Summary
Multiple propagating modes in random polycrystals resolves two transverse modes and two longitudinal modes, emerging when the wavelength becomes comparable to the grain correlation length The coherent field is found to be inherently multimodal in the geometric regime for statistically isotropic polycrystals This observation is significant because it demonstrates that the number of propagating modes is set not by the number of roots but by their attenuation hierarchy
Observation of Multimodality
The study uses grain-scale finite-element simulations to directly observe the coherent field in the frequency–wavenumber plane, resolving two transverse modes and two longitudinal modes when the wavelength approaches the grain correlation length The simulations show that an injected plane wave is extinguished within about one correlation length, and any surviving coherent field beyond that is rebuilt by forward multiple scattering through distinct longitudinal and transverse channels The dispersion equation admits three roots at all frequencies, but only in the shortwavelength geometric regime do two of them attain comparable attenuation while retaining distinct phase velocities, becoming separately observable
Theoretical Framework and Comparison
The theoretical framework involves solving the Dyson equation for the mean Green’s function, which yields a dispersion equation with three roots corresponding to three scattering channels (L, T1, and T2) The mass operator sums these three scattering channels M → L, M → T1 and M → T2 For transverse propagation in iron, the three roots are overlaid on the measured spectra, with Root 1 following the measured ridge closely across the entire frequency range
Role of Anisotropy and Regime Transition
The study compares results between two representative cubic materials of common and extreme anisotropy to show that the multimodal structure is a generic property of statistically isotropic polycrystals In iron, the transition to the geometric regime occurs near 10 MHz, where two ridges emerge with distinctly different slopes and widths indicating two coherent modes For the strongly anisotropic SrMgSn, the transition occurs earlier at 2k0LaC ≃ 3, and in the geometric regime, mode 1 speeds up toward fast propagation paths while mode 2 is slowest with its propagation dominated by the slow transverse channel
Conclusion on Physical Validity
Two main conclusions are drawn: first, the multiple roots of the effective dispersion equation are physical, as an approximate theory correctly predicts which modes appear and where they separate Second, the number of observable modes is governed by the attenuation hierarchy of the roots, with one mode always observable and a second becoming observable in the geometric regime as attenuation for both becomes comparable The mechanism involves distinct forward-scattering channels carrying distinct coherent modes once wavelength resolves microstructure This result applies to other random media such as electromagnetic waves in random dielectrics and high-frequency seismic waves in the heterogeneous Earth
Summary
Grain-scale simulations resolve the coherent wavefield of statistically isotropic polycrystals into its constituent modes and reveal it to be multimodal in the geometric regime The single effective elastic wave description holds through Rayleigh and stochastic regimes, but it is in the geometric regime that remaining modes appear The number of observable modes is determined by the attenuation hierarchy of the roots.
--- Page 1 ---
Multiple propagating modes in random polycrystals Ming Huang, 1, 2, ∗ Stanislav I. Rokhlin, 3 and Michael J. p. 0 S. Lowe2 Department of Engineering and Design, University of Sussex, Brighton BN1 9RH, United Kingdom 2Department of Mechanical Engineering, Imperial College London, London SW7 2AZ, United Kingdom 3Department of Materials Science and Engineering, The Ohio State University, 1248 Arthur E. p. 0 Adams Drive, Columbus, Ohio 43221 USA Multiple-scattering theory predicts that the coherent elastic wave in a statistically isotropic polycrystal is not a single wave. p. 0 Its dispersion equation admits multiple roots of a given polarization, but these extra modes have never been observed Here, using grain-scale finite-element simulations of three-dimensional polycrystals, we resolve the coherent field in the frequency–wavenumber plane and directly observe the predicted multiplicity for both transverse and longitudinal excitations: a single mode at low frequencies gives way to two in the geometric regime as the wavelength approaches the grain correlation length The measured modes follow the roots of the dispersion equation, which always carries a third root whose attenuation exceeds that of the others It is this attenuation hierarchy, not the root count, that dictates how many modes propagate and can be observed Coherent wave propagation in random media is thus inherently multimodal in the geometric regime, which we reach here for the first time in polycrystals An elastic wave traversing a random medium loses energy from its coherent, ensemble-averaged component to the scattered field, and the surviving coherent wave is conventionally described by a single complex effective wavenumber per polarization [1–3] Yet a statistically homogeneous random medium can carry more than one wave of a given polarization Biot predicted this first for fluid-saturated porous solids, whose coupled solid–fluid motion supports fast and slow compressional waves [4, 5] confirmed ultrasonically later by Plona [6] Brillouinscattering measurements on colloidal suspensions then revealed two longitudinal modes once the wavelength became comparable to the sphere size [7] These were explained as quasimodes [8], and were later mapped in detail by pulsed ultrasonics [9, 10] Most recently, for random particulate composites, it has been proved that the ensemble-averaged field is a superposition of infinitely many effective waves rather than one wave [11, 12] and an independent variational treatment found that a layer of random composite supports exactly two mean plane waves [13] In all of these systems, however, the multiplicity is tied to the presence of two constituents. p. 0 Single-phase polycrystals pose the question in its purest form. p. 0 There is no second constituent, and the only disorder is the random orientation of elastically anisotropic grains For scalar waves in continuous random media, Frisch [14], following Tatarskii and Gertsenshtein [15] gave the mean Green’s function as two terms with distinct wavenumbers, whose attenuations and amplitudes become comparable at high frequency (see also [16]), and Calvet and Margerin [17] showed that the second root is demanded by causality, because a single mode with the asymptotic velocity and attenuation of the geometric limit cannot satisfy the Kramers–Kronig relations For elastic waves, the far-field approximation of Rokhlin et al. p. 0 [18], like the second-order theory of Stanke and Kino [19] yields two longitudinal and transverse roots In polycrystals, however, this multiplicity has remained purely theoretical: no additional mode has ever been observed, and it has not been known whether the extra roots are physical waves or artifacts of the approximations that produce them In this Letter we supply that observation for statistically isotropic polycrystals Grain-scale finite-element simulations, which solve the elastodynamic equations directly and invoke no scattering theory, with the coherent field examined in the frequency–wavenumber plane, resolve two transverse modes and two longitudinal modes, emerging when the wavelength becomes comparable to the grain correlation length The simulations also show the mechanism directly: the injected plane wave is extinguished within about one correlation length, and the coherent field beyond is rebuilt by forward multiple scattering, whose distinct longitudinal and transverse channels support distinct modes Comparison with theory then completes the picture The dispersion equation admits three roots at all frequencies, but only in the shortwavelength geometric regime do two of them attain comparable attenuation while retaining distinct phase velocities, becoming separately observable The third root attenuates faster throughout and cannot be observed at any frequency This means that the number of propagating modes is set not by the number of roots but by their attenuation hierarchy We demonstrate these findings in two representative cubic materials of common and extreme anisotropy, and the material choice only shifts the frequency at which the geometric regime occurs
--- Page 2 ---
FIG. 1. Finite-element simulation of a plane transverse wave pulse in statistically isotropic polycrystalline iron; the pulse is excited on the z = 0 plane. p. 1 (a) Computational domain cut open at x = 6 mm, with the grains colored arbitrarily and the removed half outlined in gray; color on the cut face is the displacement ux at t = 0. p. 170 µs in a single realization at 16 MHz excitation (2k0TaC = 14. p. 137), and the black rectangle marks the window of (b) ux on the same plane at the four times shown, overlaid with the grain boundaries on that plane. p. 1 (b) ux on the same plane at t = 0. p. 016 µs, overlaid with the grain boundaries on that plane. p. 1 (c) Coherent field ux(t, z), averaged over the x–y cross section at each depth z and over 20 realizations; the strip above shows the waveforms at z = 0. p. 125 mm ≃ aC and at z = 1.2 mm ≃ 5aC, the latter magnified six times, marked by dashed lines and arrows of the same colors in the field plot (d) As (c) after high-pass filtering above 5. p. 13 MHz. Magnitude of the two-dimensional Fourier transform of (d), normalized to its maximum; points at right are the section at f = 16 MHz, marked by the dashed line and arrow in the map, and the red line is the Lorentzian fit, with the numbers labeling the two modes
--- Page 3 ---
transverse wave pulse for which λ0T = 0.20 mm is smaller than average grain and 2k0TaC = 14. p. 137 The computational domain, shown in Fig. p. 1 1(a), is a 12 × 12 × 5 mm3 cuboid containing grains generated by Voronoi tessellation, with mean grain size d = 0. p. 15 mm and two-point correlation length aC = 0. p. 123 mm [21] The domain is discretized into a uniform structured mesh of 369 million elements, achieving a numerical error of about 0. p. 11% [22] The crystallographic orientations of the grains are assigned uniformly at random, creating a statistically isotropic aggregate The grains are lossless, with no intrinsic dissipation, so all attenuation arises from scattering A three-cycle, 16 MHz toneburst transverse force polarized along x is applied uniformly over the z = 0 face and boundary conditions on the four lateral surfaces restrain their y and z motions to enforce plane-wave propagation along z The model is solved with the GPU-accelerated solver Pogo [23] Figure 1(a) shows the wavefield ux on the mid-x plane at t = 0. p. 17 µs, and Fig. p. 1 1(b) tracks its evolution The wavefront is smooth at t = 0. p. 116 µs, breaks up on the grain scale after roughly one correlation length at t = 0. p. 134 µs, and converts fully to speckle at t ≥ 0. p. 170 µs This means that the injected wave is extinguished at about one correlation length, and whatever coherence survives at depth must be carried by forward scattering To extract that coherent component, the displacement is averaged over the x–y cross section at each of 71 depths from z = 0 to 3. p. 15 mm in 0.05 mm steps excluding planes near the back face at z = 5 mm to avoid reflections and then averaged over 20 independent realizations with re-randomized orientations which is sufficient for convergence The resulting transverse coherent field ux(t, z) is shown in Fig. p. 1 1(c); the uy and uz components are three orders of magnitude smaller and carry no coherent part The coherent wave shows one streak near the source at z = 0, while two streaks of different slopes appear beyond z ≃ aC consistent with the extinction of the injected wave at one correlation length Because attenuation grows with frequency, the two streaks are mostly masked by the surviving low-frequency components beyond z ≃ 5aC but emerge clearly after high-pass filtering in Fig. p. 1 1(d) Here and throughout, the filter is a zerophase minimum-order IIR high pass with 60 dB stopband attenuation and passband edge fc(N − 2)/N the lower bandwidth limit of a N-cycle toneburst with center frequency fc A two-dimensional Fourier transform of ux(t, z) then maps the coherent field into the frequency–wavenumber plane in Fig. p. 2 1(e) where ridges of high amplitude trace the dispersion relations of the propagating coherent modes Near 10 MHz, where 2k0TaC ≃ 9 and λ0T falls below ≃ 1. p. 24aC, marking the onset of the geometric regime, two ridges emerge with distinctly different slopes and widths indicating two coherent modes with their own phase velocities and attenuation coefficients The two modes are unambiguously resolved, constituting a direct observation of multiple coherent modes in a single-phase random medium
--- Page 4 ---
For comparison we solve the Dyson equation for the mean Green’s function in the first-order smoothing approximation, which yields the dispersion equation for the complex effective wavenumber k(ω) ω2 − k2V20M − mM(k, ω) = 0 (1) for excitation mode M where ω = 2πf, V0M is the Voigtaveraged velocity of mode M and k = kp with p representing the propagation direction Other reference media can be used [27] we take the Voigt average for which the theory agrees well with simulations [28, 29] The mass operator, or self-energy, mM sums the three scattering channels M → L, M → T1 and M → T2 where L denotes the longitudinal wave and T1 and T2 the two transverse waves It is this channel structure that gives Eq. (1) three roots [30] Each root carries a phase velocity VM = ω/Re k and an attenuation coefficient αM = Im k The three roots for transverse propagation in iron are overlaid on the measured spectra in Fig. p. 2 2(a), which combines the 16 MHz simulation with an additional 8 MHz simulation on a longer, 12 × 12 × 10 mm3 domain where the coherent signals are taken over z = 0–4. p. 25 mm and all else is unchanged Root 1 follows the measured ridge closely across the entire frequency range and root 2 emerges above 2k0TaC ≃ 9 in the geometric regime The third root has no counterpart in the simulation To compare quantitatively we extract each mode from the simulation spectra Mode i is a pole of the mean Green’s function at complex wavenumber ki + jαi (with j squared = −1) so it propagates as e jkiz−αiz and contributes to the power spectrum a Lorentzian centered on ki with half-width αi Neglecting the interference between wellseparated modes, we fit P2 i=1 Ai/((k − ki) squared + α 2i) + b at each frequency, as illustrated in Fig. p. 2 1(e) and obtain the measured points in Figs. p. 2 2(b) and 2(c) Both modes are resolved in single realizations, whose cross-sectional average spans several hundred grains between realizations their fitted velocities vary by less than 1% and their attenuations by about 6% Theory and simulation agree closely The phase velocities of both modes are reproduced across the entire simulated frequency range, and the attenuation coefficients agree wherever a single mode propagates In the geometric regime the measured attenuation deviates from the prediction The deviation stems not from the Lorentzian fitting but from the theory, which is used here at 2αaC ≳ 1, beyond weak scattering
--- Page 5 ---
Longitudinal excitation completes the picture The longitudinal wavelength is longer at a given frequency, and we therefore select SrMgSn, a cubic polycrystal of very strong anisotropy with c11 = 7, c12 = 4 and c44 = 25 GPa, ρ = 3472.8 kg/m3 A = 16.67 and AU = 17.67 so that the geometric regime is reached at computationally accessible frequencies [29] Simulations at 2, 4 and 8 MHz center frequency reuse the 12 × 12 × 10 mm3 domain for 2 and 4 MHz and the 12×12×5 mm3 domain for 8 MHz with the polarization now along z (namely force applied in z direction, Fig. p. 2 1(a)) and symmetry boundary conditions on the lateral surfaces Figure 3(a) shows the resulting spectra with the three theoretical roots overlaid, and the same phenomenon appears, with a single ridge at low frequencies and two ridges abo… p. 2
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
-
Simulation-based mode identification for random polycrystals can be used to predict which modes are physically observable based on their attenuation hierarchy, as
the number of propagating modes is set not by the number of roots but by their attenuation hierarchy.
-
AI systems can perform direct observation of multiple coherent modes in single-phase random media by analyzing frequency-wavenumber plane data from grain-scale finite element simulations.
-
The improved system can determine that
the extra roots are not artifacts of the approximation
because the theory and simulation agree across different materials, indicating thatthe multimodal structure is a generic property of statistically isotropic polycrystals.
-
AI can distinguish between observable and unobservable modes by identifying those whose attenuation permits propagation over many microstructural lengths, answering the question:
of the many mathematically admissible effective waves, those that survive are the ones whose attenuation permits propagation over many microstructural lengths.
Related papers
- Structural contribution to light-induced gap suppression in Ta 2 NiSe 5
- Anatomy of fast current-induced skyrmion motion in synthetic antiferromagnets
- Magnetic control of an exciton-polariton condensate in a van der Waals magnet
- High-pressure elastic properties of GeO2 polymorphs up to 120 GPa
- Decoupling Strain-Rate Sensitivity and Deformation Length Scale Effects in Neutron-Irradiated Tungsten: A Coupled Nano-Indentation, HR-EBSD and Crystal Plasticity Study
- Magnetism modulation in Co3Sn2S2 by current-assisted domain wall motion