Bouncing Barrier in Protoplanetary Dust Growth

• Bouncing barrier is a collisional regime in protoplanetary dust evolution where aggregates primarily rebound, halting sustained growth.
• Laboratory experiments and simulations reveal that factors like aggregate size, porosity, and charging significantly influence the threshold between sticking and bouncing.
• Overcoming the barrier via mechanisms such as sweep-up growth, small-dust feeding, or magnetic/triboelectric effects may catalyze planetesimal formation.

Searching arXiv for the key papers on the bouncing barrier in protoplanetary dust growth and closely related usages of the term. First, I’ll retrieve the charged-grain aggregation paper that explicitly frames growth “beyond the traditional bouncing barrier.” Now I’ll retrieve foundational and papers on the bouncing barrier in dust evolution, plus one paper using the term in a different field to distinguish meanings. The bouncing barrier is the collisional regime in protoplanetary dust evolution in which aggregates no longer grow efficiently by pairwise sticking, but instead mainly rebound, or stick only transiently and later detach, so that direct coagulation stalls before planetesimal scales are reached. In laboratory-calibrated disk models, this regime replaces the previously assumed growth–fragmentation cycle, produces long-lived quasi-steady size distributions, and stops growth by bouncing rather than fragmentation, with final Stokes numbers of roughly $10^{-4}$ and particle masses limited to approximately $1$ gram in the modeled 1 AU environments (Zsom et al., 2010). Subsequent work has shown that the barrier is not a single universal threshold: it depends on aggregate size, filling factor, internal structure, velocity statistics, and material physics, and it can be softened, shifted, or bypassed by mechanisms such as sweep-up growth, small-dust feeding, triboelectric charging, and magnetic dipole attraction (Booth et al., 2017).

1. Definition and emergence in dust-growth theory

In the dust-collision literature, the barrier occupies the intermediate regime between sticking and fragmentation. Low collision speeds produce sticking, intermediate speeds produce bouncing, and high speeds produce fragmentation; for compact silicate grains in the inner disk this regime is especially relevant near millimetre sizes, where similar-sized aggregates tend to bounce rather than merge (Dominik et al., 2023). The physical content of the term is therefore not merely “no growth,” but a change in the dominant collision outcome: collision energy is no longer dissipated in a way that yields durable contact.

A foundational formulation was given in a collision model based on laboratory experiments, in which dust evolution at 1 AU was followed for three gas densities using nine collision types, including Bouncing with compaction (B1) and Bouncing with mass transfer (B2) (Zsom et al., 2010). That model distinguished similar-sized and different-sized collisions through a critical mass ratio $r_m=10,\ 100,\ 1000$, and found that catastrophic fragmentation hardly occurs in the three disk models. Instead, repeated bouncing compacts aggregates, narrows the effective growth channels, and leads to long-lived, quasi-steady distributions. In that sense, the bouncing barrier is an emergent property of the collision map rather than an externally imposed cutoff.

This theoretical picture also clarified a common ambiguity. The barrier does not require that every collision be an elastic rebound. It is sufficient that the ensemble-averaged evolution cease to yield sustained net growth because sticking channels become too weak, too rare, or too fragile. Later laboratory work repeatedly confirmed that distinction.

2. Experimental and numerical evidence

A broad empirical record now supports the existence of a bouncing-dominated regime, while also showing that its location is preparation-dependent and material-dependent.

Study	System	Main finding
"Experimental Study on Bouncing Barriers in Protoplanetary Disks" (Kelling et al., 2014)	$\sim 100$ compact $1$ mm $\mathrm{SiO_2}$ aggregates, $>10^5$ collisions	Sticking occurs, but no temporal stable net growth
"Failed Growth at the Bouncing Barrier in Planetesimal Formation" (Kruss et al., 2016)	Compact mm quartz aggregates, 899 resolved collisions	$\sim 20\%$ sticking in individual collisions, but clusters detach and the ensemble returns to the initial size distribution
"Growing into and out of the bouncing barrier in planetesimal formation" (Kruss et al., 2017)	Self-consistent levitation experiments	The barrier emerges from smaller dust and can shift to larger size under continued small-dust feeding
"Size Dependence of the Bouncing Barrier in Protoplanetary Dust Growth" (Arakawa et al., 2023)	3D SSDEM head-on collisions of icy aggregates at fixed $\phi_{\rm agg}=0.4$	Sticking probability decreases systematically with aggregate radius
"Investigating the Bouncing Barrier with Collision Simulations of Compressed Dust Aggregates" (Oshiro et al., 5 Feb 2025)	3D $N$-body collisions of compressed icy aggregates	Bouncing appears above a threshold mass that scales as $1$0
"Observation of bottom-up formation for charged grain aggregates related to pre-planetary evolution beyond the bouncing barrier" (Jungmann et al., 2021)	Microgravity collisions of triboelectrically charged basalt grains	No bouncing collision was observed among hundreds of collisions in the explored parameter range

Long-duration laboratory studies established the ensemble-level character of the barrier. In one experiment, roughly 100 compact cylindrical $1$1 aggregates with mean equivalent radius $1$2, mean height $1$3, and filling factor $1$4 were observed for $1$5 s at collision speeds from below mm/s up to cm/s; more than $1$6 collisions occurred, nearly 2000 were analyzed in detail, and no temporal stable net growth of larger aggregates was observed even though sticking collision occur (Kelling et al., 2014). The same study reported that larger ensembles of aggregates sticking together are formed but were disassembled again during the further collisional evolution.

A closely related ensemble study of compact quartz aggregates with filling factor $1$7 found 181 sticking collisions and 460 bouncing collisions among 899 resolved impacts, giving a high probability of sticking for individual collisions of about $1$8, yet still no sustained growth (Kruss et al., 2016). Over 50 minutes, and after more than 1000 collisions per aggregate, the evolution of the ensemble always leads back to a distribution of individual aggregates of initial size. This directly motivated the interpretation that the barrier is robust against shape and size variations for small compact silicate aggregates in the tested regime.

The barrier also emerges self-consistently. Starting from micron-sized quartz dust without pre-compression, aggregates first grow by hit-and-stick, then compact, and finally evolve into a bouncing state with relative and absolute velocities of order $1$9 mm/s on average; in that final state the porosity is reported as $r_m=10,\ 100,\ 1000$0 for aggregates selected with $r_m=10,\ 100,\ 1000$1 (Kruss et al., 2017). This result removed the concern that the barrier might be an artifact of specially prepared compact samples.

3. Microphysical controls and quantitative descriptions

A central result of later work is that the barrier is controlled by more than a single porosity threshold. Molecular-dynamics simulations of sub-mm silicate aggregates constructed by different preparation methods showed that there is no unique relation between the average volume filling factor and the coordination number of the aggregate (Seizinger et al., 2013). In that study, realistic aggregates bounce only if their volume filling factor exceeds $r_m=10,\ 100,\ 1000$2 and collision velocities are below $r_m=10,\ 100,\ 1000$3 m/s, whereas artificially regular hexagonal-lattice aggregates were much more prone to bounce. The paper’s physical interpretation was energetic: bouncing requires enough impact energy to be stored elastically so that the contact can break again, whereas irregular aggregates dissipate more energy through rolling and sliding unless they are already highly compact.

Size dependence became explicit in later SSDEM calculations. For equal-mass head-on collisions of icy aggregates built from monomers of radius $r_m=10,\ 100,\ 1000$4, with aggregate radii $r_m=10,\ 100,\ 1000$5 and fixed initial filling factor $r_m=10,\ 100,\ 1000$6, the sticking probability decreases systematically with increasing aggregate radius (Arakawa et al., 2023). Using the largest-remnant metric $r_m=10,\ 100,\ 1000$7, that study found $r_m=10,\ 100,\ 1000$8 for $r_m=10,\ 100,\ 1000$9, but a drop to $\sim 100$0 at $\sim 100$1. The inferred conclusion was that the threshold filling factor for sticking versus bouncing, $\sim 100$2, decreases with aggregate size. This directly addresses the discrepancy between small numerical aggregates and much larger laboratory aggregates: the barrier is not a universal single-number threshold.

Recent collision simulations of compressed icy aggregates sharpened this quantitative picture (Oshiro et al., 5 Feb 2025). In those 3D $\sim 100$3-body calculations, aggregates with filling factors $\sim 100$4–$\sim 100$5 were built by compressing BCCA structures rather than by close-packing and particle extraction. Bouncing occurred above a threshold mass that decreases with impact velocity and obeys

$\sim 100$6

The best-fit values were $\sim 100$7 for $\sim 100$8, $\sim 100$9 for $1$0, and $1$1 for $1$2, giving a very steep scaling $1$3. The same work decomposed the collision into compression, transition, and stretching phases, and found that nearly $1$4 of the initial impact energy is dissipated during the initial compression phase, while over $1$5 of the remaining energy is dissipated during stretching, regardless of whether the outcome is sticking or bouncing. This makes clear that bouncing is a high-dissipation process; the distinction is not low dissipation versus high dissipation, but whether the recovered elastic energy after compression still exceeds the fracture cost of the contact network.

A semi-analytic model later cast this in statistical form (Arakawa et al., 25 Jun 2026). The collision is divided into a compression phase, described by an elastoplastic contact model, and a separation phase, treated as fracture of a stochastic bond network. The sticking probability is written as

$1$6

so that the $1$7 sticking–bouncing boundary is determined by $1$8. Because larger contact patches are more likely to contain a statistically weak bond, the model naturally predicts that larger aggregates bounce more readily. Comparison with DEM simulations was reported to work best for $1$9 and $\mathrm{SiO_2}$0.

4. Robustness, softness, and routes beyond the barrier

Although the barrier is robust, it is not identical to an absolute prohibition on growth. A recurring misconception is that “bouncing barrier” means all collisions bounce. Long-term experiments instead show that some collisions do stick, sometimes with clear bound rotation, but that the resulting clusters are weak and are later disassembled; one laboratory study therefore remarked that detachment barrier may even be more accurate than “bouncing barrier” for its parameter set (Kelling et al., 2014).

A second misconception is that the barrier must be a hard stop. Coagulation models that resolve the low-velocity tail of the collision-speed distribution show a softer picture (Booth et al., 2017). In a simple sticking–bouncing–fragmentation model, breakthrough by rare lucky collisions requires the mass ratio at which high velocity collisions transition to growth instead of fragmentation to be low, $\mathrm{SiO_2}$1. When the fragmentation threshold is allowed to depend on mass ratio, however, breakthrough occurs more readily, even if mass transfer is relatively inefficient. That work concluded that bouncing may only slow down growth, rather than preventing growth beyond a threshold barrier, although radial drift will usually prevent growth to arbitrarily large sizes.

The same softening appears in seeded-growth models. A continuum coagulation calculation using laboratory-motivated collision outcomes found that for the general dust population, bouncing collisions prevent growth above millimeter-sizes, but if a small number of cm-sized particles are introduced, they can act as a catalyst and start to sweep up the smaller particles (Windmark et al., 2012). At a distance of 3 AU, 100-meter-sized bodies are formed on a timescale of 1 Myr. In that specific sense, the barrier can be beneficial: it prevents the growth of too many large particles that would otherwise only fragment among each other, and creates a reservoir of small particles that can be swept up by larger bodies.

Laboratory feeding experiments reached a related conclusion from the opposite direction (Kruss et al., 2017). When small porous dust aggregates are added to an ensemble of larger bouncing aggregates, the small dust sticks efficiently to the large aggregates and temporarily increases the apparent stickiness between them, but the final number of large bouncing aggregates remains the same. The size of each aggregate nevertheless increases because the large aggregates capture the supplied dust. This suggests that in the presence of a dust reservoir aggregates grow into but also out of a bouncing barrier at larger size.

More radical escape routes rely on additional interaction physics. In microgravity experiments with basalt spheres of diameter $\mathrm{SiO_2}$2–$\mathrm{SiO_2}$3 and mass $\mathrm{SiO_2}$4 mg, triboelectric charging generated net charges of about $\mathrm{SiO_2}$5 to $\mathrm{SiO_2}$6 per grain and produced bottom-up formation of irregular aggregates (Jungmann et al., 2021). Charged grains continued to stick at velocities of $\mathrm{SiO_2}$7, whereas neutral basalt grains are only expected to stick below $\mathrm{SiO_2}$8, and no bouncing collision was observed among hundreds of collisions in the explored parameter range. The same study reported that some trajectories are consistent with a pure Coulomb potential,

$\mathrm{SiO_2}$9

while others deviate from pure Kepler-like motion, plausibly because of non-homogeneous surface charge distributions and higher-order multipoles.

Magnetic fields provide another modifier for specific materials. In levitation experiments with iron–quartz mixtures, all samples without magnetic fields grow into a bouncing barrier, but in homogeneous magnetic fields up to $>10^5$0 mT the iron-bearing aggregates form reversible chains (Kruss et al., 2018). The size of the largest entities increases by a factor of about 3, and for the iron:quartz $>10^5$1 mixture the critical field is $>10^5$2. This defines an iron-sensitive bouncing barrier: magnetic dipole attraction shifts the largest attainable aggregate size without creating permanent magnetic welding.

5. Consequences for disk evolution, streaming instability, pebble accretion, and observables

Once implemented in global dust-evolution calculations, the bouncing barrier changes both the aerodynamics and the observables of disks. In DustPy models with a Maxwell-Boltzmann collision-speed distribution, adding bouncing reduces in many cases the size of the typical or largest particles available in the disk, produces a very narrow, almost mono-disperse size distribution, and removes most micrometer-sized grains in the process (Dominik et al., 2023). In the fiducial comparison, particle radii become up to about a factor of 10 smaller, or $>10^5$3 smaller in mass, and the width of the size distribution in $>10^5$4 drops to about $>10^5$5, compared with about $>10^5$6 without bouncing. Because the barrier lowers the typical Stokes number and modifies settling, it changes the effectiveness of and timescales for the streaming instability and for pebble accretion. The same models predict strong observational consequences: the 10 and 20 $>10^5$7m silicate emission features can be strongly suppressed, and self-shadowing can generate a strong depression in scattered-light surface brightness between about 50 and 80 au. The authors explicitly noted, however, that the complete removal of small grains in their model is not consistent with observations and suggested incomplete vertical mixing or some level of erosion in collisions as possible remedies.

A complementary disk-scale argument links the barrier to long-lived mm-bright disks (Qian et al., 8 Jul 2025). There, bouncing is assumed to stall grain growth at a near-universal size of $>10^5$8, with a characteristic Stokes number $>10^5$9–$\sim 20\%$0. Because such grains are tightly coupled to the gas, radial drift is drastically slowed: the paper estimates $\sim 20\%$1 yrs at the fragmentation barrier but $\sim 20\%$2 Myrs at the bouncing barrier. This was proposed as an explanation for why disks in both Lupus and Upper Scorpius can remain bright and close to optically thick at mm wavelengths for millions of years. The same argument carries a planetary cost: these $\sim 20\%$3-scale grains offer poor prospects for processes like streaming instability or pebble accretion.

At the same time, earlier coagulation models already suggested that a bouncing-limited population may be useful for concentration mechanisms (Zsom et al., 2010). If direct sticking stalls near chondrule-sized particles with $\sim 20\%$4, turbulent concentration or gravitational collapse of dense clumps may become the relevant route onward. This does not remove the barrier; it repositions it as a bottleneck that shapes the size distribution presented to later planetesimal-formation processes.

6. Terminological extensions outside planet formation

Although the dominant modern use of bouncing barrier is in protoplanetary dust growth, the phrase and closely related constructions also appear in unrelated fields. In Lorentzian geometry and AdS/CFT, an extremal surface barrier is a codimension-1 splitting surface $\sim 20\%$5 such that extremal surfaces anchored on one side cannot be continuously deformed past it while remaining extremal (Engelhardt et al., 2013). A sufficient condition is nonpositive outward extrinsic curvature, $\sim 20\%$6, and the special case $\sim 20\%$7 yields a totally geodesic barrier: anchored extremal surfaces may approach it, or lie within it, but cannot cross it. In the authors’ formulation, the barrier “blocks” extremal surfaces from “bouncing” through it.

In droplet-impact dynamics, the barrier is energetic rather than collisional in the dust-aggregate sense (Bange et al., 2016). For a droplet impacting a superhydrophobic surface, bouncing occurs if the total droplet energy at the instance of maximum recoiling exceeds the initial surface and gravitational energy,

$\sim 20\%$8

whereas otherwise the droplet remains attached and oscillates. The term therefore refers to a minimum detachment condition rather than to a growth-limiting collisional regime.

These alternative usages do not alter the planet-formation meaning, but they matter for literature searches. In current astrophysical practice, the bouncing barrier refers specifically to the regime in which aggregate collisions become growth-limiting because durable sticking is replaced by rebound or by short-lived contacts that do not survive subsequent impacts.