Dynamic Powder Sputtering

• Dynamic Powder Sputtering is a process where powders or granular targets experience time-dependent sputter yields due to feedback between ion bombardment, evolving composition, and particle motion.
• Agitated-particle coating techniques enable uniform thin-film deposits on powders (25–500 µm), while cohesive fine powders exhibit limited mobility due to agglomeration under vacuum.
• Loose-powder sputtering relies on porosity and local geometry, with Monte Carlo models and scaling laws revealing distinct angular distributions and re-deposition effects.

Dynamic powder sputtering denotes a class of non-steady sputtering phenomena in which powders or granular targets participate directly in the sputtering cycle, either because powders are actively agitated to expose all surfaces to an incoming sputter flux, because loose powders themselves are bombarded and emit directionally filtered ejecta, or because injected solid particles ablate, redeposit, and modify wall composition during plasma operation. In the underlying dynamic-sputtering sense, the sputter rate and composition of the emitted flux do not remain constant in time, but evolve due to feedback between ion bombardment, surface composition, defect chemistry, and re-deposition. Recent work therefore treats dynamic powder sputtering not as a single apparatus class, but as a coupled plasma–surface–granular problem spanning compound-surface sputtering, agitated-particle coating, porous-target emission, and wall-conditioning by injected powders (Zayachuk et al., 2017, Lourens et al., 2022, Verkercke et al., 10 Mar 2026).

1. Conceptual domain and defining regimes

Dynamic powder sputtering has two experimentally distinct but physically connected meanings in the recent literature. One is sputtering onto powders that are actively moved to expose all sides to the sputter flux. In a laboratory magnetron system, this has been implemented with a piezoelectric actuator driving uniaxial vertical oscillation of a multi-pit plate, so that particles jump, tumble, and reorient during deposition; the same platform supports up to nine separate particle batches in one run and can be used for combinatorial coating experiments (Lourens et al., 2022). The other is sputtering from loose powders or highly porous granular media, where the target is a three-dimensional network of grains and interconnected voids rather than a planar surface; in that case the escaping sputtering yield and angular distribution differ markedly from both flat slabs and rough but non-porous surfaces (Verkercke et al., 10 Mar 2026).

A broader dynamic interpretation emerges from compound-surface sputtering studies. Under Secondary Neutral Mass Spectrometry conditions, stoichiometric PbTe bombarded by Ar plasma at 50–550 eV exhibits aperiodic oscillations of Pb and Te sputtering, a huge preference of Te sputtering reaching more than two orders of magnitude at the beginning of sputtering, and a significant excess of Te integrated sputter yield over that of Pb for prolonged sputtering by low energy plasma at 50–160 eV (Zayachuk et al., 2017). Although that system is a bulk single crystal rather than a powder bed, it provides a direct template for dynamic sputtering as a process in which yields are explicit functions of time or fluence rather than constants.

These two meanings are complementary rather than contradictory. Agitated-particle coating emphasizes complete geometric exposure of powder surfaces to an external sputter source. Loose-powder ion sputtering emphasizes how porosity, shadowing, and local incidence geometry alter the escape of sputtered atoms. A plausible implication is that practical dynamic powder sputtering systems often combine both: evolving surface composition on each grain, and evolving geometric accessibility across the granular ensemble.

2. Time-dependent sputtering, preferential emission, and re-deposition

The canonical dynamic-sputtering mechanism is feedback between composition-dependent erosion and re-deposition. For multicomponent surfaces in the linear-cascade regime, the ratio of partial sputter yields is written as

$$\frac{Y_1}{Y_2} \simeq \frac{N_1}{N_2}\left(\frac{M_2}{M_1}\right)^{2m}\left(\frac{U_2}{U_1}\right)^{1-2m},$$

where $N_i$ is the surface atomic concentration, $M_i$ the atomic mass, $U_i$ the surface binding energy, and $m$ the exponent in the low-energy power cross section (Zayachuk et al., 2017). In PbTe, however, low-energy behavior is not captured by binding-energy arguments alone. At 50 eV, the Te preference exceeds two orders of magnitude initially, while at 350–550 eV Pb dominates only moderately and the difference between integrated yields is on the order of 10–12% (Zayachuk et al., 2017).

The distinguishing feature is the coexistence of preferential sputtering and stochastic re-deposition. The PbTe study attributes aperiodic oscillations to the superposition of direct sputtering and re-deposition of sputtered material back onto the surface as crystalline formations. SEM after 3200 s at 160 eV shows an array of crystallites, and EDX shows that both the crystallites and the underlying surface have compositions close to stoichiometric PbTe. Their density increases when the sputtering energy decreases, so re-deposition becomes more important at low energies. Because nucleation is random in time and space, the net removal rate can fluctuate strongly. The reported amplitude reaches about 80% deviation from the mean Te intensity at 50 eV and about 10% at 550 eV (Zayachuk et al., 2017).

Near-threshold sputtering introduces an additional electrostatic channel. In PbTe, interstitial Pb behaves as a double-charged donor, $\mathrm{Pb}_i^{2+}$, with zero activation energy, whereas interstitial Te is electrically neutral, $\mathrm{Te}_i^0$. In Ar plasma, the positively biased plasma relative to the negatively biased sample adds an electrostatic barrier for the escape of $\mathrm{Pb}_i^{2+}$, while neutral Te recoils do not experience that barrier. This explains why Te can dominate the sputtered flux at 50–160 eV even though $U_{\mathrm{Te}} > U_{\mathrm{Pb}}$ (Zayachuk et al., 2017).

For powder systems, the immediate significance is methodological. Dynamic powder sputtering cannot be reduced to a static yield lookup indexed only by target composition and ion energy. It is inherently a feedback problem involving preferential sputtering, evolving near-surface composition, stochastic re-deposition, and, in some materials, defect-charge-state asymmetries.

3. Agitated powders as sputter-deposition substrates

The most direct implementation of dynamic powder sputtering as a coating method uses mechanically agitated powders in a conventional magnetron chamber. A representative system employs an AJA International chamber of about 25 cm inner diameter, three 3.81 cm magnetron cathodes, Ar at 0.5 Pa, and a particle holder located about 110 mm beneath the cathodes. The central element is a vacuum-compatible PZT stack actuator coupled to an aluminum plate containing nine milled pits, each $5 \times 5\,\mathrm{mm^2}$ and 1 mm deep. A square-wave drive with a linear frequency sweep from 0.1 to 4.0 kHz and amplitudes between 7 and 20 V $N_i$0 induces particle jumping, detachment from the pit floor, random reorientation, and collisions with other particles and pit walls (Lourens et al., 2022).

Under these conditions, separate batches of polystyrene microspheres of 500 $N_i$1m monodisperse diameter, Fe alloy particles of 300 $N_i$2m mean size, and NaCl particles of 350 $N_i$3m mean size were simultaneously coated with homogeneous CrMnFeCoNi thin films from a single high-entropy-alloy target. By co-sputtering Cr$N_i$4Mn$N_i$5, Fe$N_i$6Co$N_i$7, and Ni targets, nine separate batches of 25 $N_i$8m Al particles were coated with a CrMnFeCoNi thin film with a composition gradient. Full or near-full coverage was obtained for particle types that jumped and tumbled freely, with standard deviations below 0.5 at.% per element for the 500–350 $N_i$9m particles coated from the single HEA target (Lourens et al., 2022).

The lower size limit of the platform is set by cohesion and sticking. For 5 $M_i$0m Al particles, movement is significantly reduced at 0.5 Pa, agglomerates stick to the oscillation plate after a few minutes of sputtering, and only exposed parts of the outermost particles are coated. The reported causes are loss of lubrication under vacuum, increased friction and adhesion, and a metallic film “gluing” effect at particle–particle and particle–plate contacts. Even at the maximum 20 V drive amplitude, the agitation energy is insufficient to break those cohesive agglomerates (Lourens et al., 2022).

This establishes a practical distinction within dynamic powder sputtering. For approximately 25–500 $M_i$1m, simple vertical oscillation in shallow pits is sufficient for uniform films and combinatorial screening. For cohesive powders below about 10 $M_i$2m, dynamic exposure is limited by agglomeration and by time-dependent immobilization during coating itself.

4. Loose-powder sputtering, porosity, and escaping ejecta

When the powder is the sputtered target rather than the substrate, the key variables are porosity, local incidence geometry, and shadowing. A multiscale Monte Carlo model for Kr$M_i$3 bombardment of Cu powders composed of 50–90 $M_i$4m spheres shows that sputtering from loose powders differs markedly from both flat slabs and rough surfaces. The main differences are that for incident angles $M_i$5 relative to the bulk normal the escaping sputtering yield is dominated by backward-directed ejecta for all ion energies, for $M_i$6 the yield peaks toward the ion-beam origin, the angular-distribution peak is half or less than that of a flat slab, and as ion energy increases no evolution occurs from primary to secondary knock-on sputtering in the ejecta angular distribution (Verkercke et al., 10 Mar 2026).

The model attributes these effects to interconnected voids. Ions penetrate the void network and strike underlying grains; sputtered atoms must then escape through a geometry in which shadowing is weakest toward the ion-beam origin. For a Cu powder with porosity $M_i$7, the escape percentage is about 45% at $M_i$8 and about 60–63% at $M_i$9, largely independent of ion energy between 1 and 20 keV. At 20 keV, the reported values are $U_i$0, $U_i$1, and $U_i$2 at normal incidence, while at $U_i$3 they are $U_i$4, $U_i$5, and $U_i$6 (Verkercke et al., 10 Mar 2026).

The resulting scaling law for the total escaping yield is

$U_i$7

with $U_i$8 for spherical grains (Verkercke et al., 10 Mar 2026). The angular probability density is represented as a finite spherical-harmonic expansion,

$U_i$9

and the absolute doubly differential escaping yield is then $m$0. For porosities $m$1 and $m$2, a fourth-order expansion gives $m$3 with mean absolute error below 8%, while a reduced set of dominant coefficients still gives $m$4 with mean absolute error below about 10% (Verkercke et al., 10 Mar 2026).

A related fusion-wall case replaces a static powder bed by real-time boron powder injection into the DIII-D scrape-off layer. There, the injected powder ablates near the separatrix and becomes a three-dimensional boron source, after which the resulting B ions and neutrals participate in erosion and redeposition on plasma-facing components. Mixed-material migration modeling predicts evolving B deposition patterns and the formation of mixed B-C layers or predominantly B coverage depending on the powder mass flow rate (Effenberg et al., 2024). The common principle is that powder geometry or powder-derived source geometry controls the emitted flux only in conjunction with subsequent transport, recycling, and local re-erosion.

5. Multiscale transport and plasma–surface modeling

Dynamic powder sputtering is intrinsically multiscale because the surface source term cannot be prescribed independently of gas-phase transport or of later wall interaction. A machine-learning plasma–surface interface has been proposed precisely for this class of problem. In a proof of concept for Ar projectiles bombarding a Ti–Al composite, a multilayer perceptron with 151 input nodes, three hidden layers of 1000 neurons, and 1800 output nodes learns the mapping from a discretized incident ion energy distribution to full, multi-species energy–angle distributions of reflected Ar and sputtered Al and Ti (Krüger et al., 2018). For a Gaussian incident distribution at 590 eV, the predicted distributions achieve $m$5 against a high-statistics TRIDYN reference, while the low-statistics TRIDYN data used for training give $m$6. The inference time is about 0.87 ms per query, which is fast enough for quasi-continuous use as an interface inside gas-phase transport solvers (Krüger et al., 2018).

That interface addresses a specific limitation of lookup tables or analytic fits in dynamic conditions: incident ion energy distributions, surface composition, and local plasma conditions can vary in time and space. The same study explicitly notes that the input can be extended to additional dimensions such as projectile species, incidence angle, target composition, and surface state. This suggests a direct route to dynamic powder sputtering models in which each grain or surface element receives its own local ion distribution and returns its own sputtered energy–angle distribution (Krüger et al., 2018).

Gas-phase redistribution further alters the source before deposition or escape is observed. In a multi-frequency capacitively coupled Al sputter source at 0.5 Pa, kinetic test-particle simulations show that sputtered Al transport is governed by the interaction of imposed and backscattered fluxes and can be understood in a one-dimensional picture away from electrode edges. In the three-dimensional geometry, 62.1% of sputtered Al is absorbed at the substrate, 14.7% returns to the target, and 23.2% reaches the chamber walls. Near electrode edges, the governing mechanism becomes backscattering from the outside chamber volume (Trieschmann et al., 2015). A plausible implication is that powder reactors operated in similarly collisional regimes will likewise exhibit nontrivial recycling paths and wall-loss fractions even when the primary sputter source is well controlled.

In fusion-wall conditioning, the integrated EMC3-EIRENE, DIS, and WallDYN3D framework makes the same point more explicitly. First-flight boron transport alone predicts that more than 95% of the total B flux reaches the inboard divertor target, but once erosion, reflection, and redeposition are included the evolving wall composition can reverse the initial asymmetry, and at $m$7 B/s the outer divertor B flux grows from about $m$8 to $m$9 over the modeled evolution (Effenberg et al., 2024). Dynamic powder sputtering therefore requires source modeling, transport modeling, and mixed-material surface evolution to be treated as a single coupled problem.

6. Applications, design constraints, and unresolved issues

The applications now span thin-film coating, accelerator-vacuum technology, fusion-wall conditioning, and additive-manufacturing-scale particle ejection. In vacuum technology, TiZrV and Pd magnetron sputtering inside long stainless-steel pipes shows how multicomponent sources and time-dependent process parameters determine film composition, roughness, and grain size. TiZrV films deposited from a twisted Ti–Zr–V wire cathode exhibit grain sizes of about 0.42–1.3 nm and roughness of about 2–4 nm, whereas Pd overlayers exhibit grain sizes of about 8.5–18.25 nm and roughness of about 17–19 nm; the Pd atomic percentage in TiZrV/Pd films is 86.84–87.56 according to XPS (Wang et al., 2015). In a related rate-analysis study, comparison of experiment with Sigmund-based simulation shows that, for Pd in that cylindrical device, the deposition rate $\mathrm{Pb}_i^{2+}$0 can be estimated by the sputtered depth $\mathrm{Pb}_i^{2+}$1, with the Yamamura-plus-Seah treatment giving the closest agreement in most cases (Wang et al., 2015).

Dynamic powder sputtering also appears in a broader mechanical sense when transient forcing ejects powders from free surfaces. A versatile SPH framework for additive-manufacturing processes shows that droplet impact in binder jetting or evaporation-induced recoil pressure in powder bed fusion leads to powder motion, distortion of the powder packing structure, and powder particle ejection. In binder jetting, increasing impact velocity from 1 to 50 m/s changes the response from gentle penetration and local rearrangement to violent splashing with binder–powder spatter. In powder bed fusion, recoil pressure produces keyhole depression, wave patterns on keyhole walls, powder compaction ahead of the moving melt pool, and dynamic powder ejection from melt-pool edges (Fuchs et al., 2022). Although this literature concerns droplet impact and laser melting rather than plasma bombardment, it isolates the same structural issue: powder ejection is controlled jointly by transient forcing, dissipation, and evolving interface geometry.

Several constraints remain recurrent across otherwise different platforms. Fine cohesive powders tend to agglomerate and immobilize during coating (Lourens et al., 2022). Flat-slab sputtering yields do not transfer directly to loose powders because porosity and void connectivity reweight both escape probability and angular emission (Verkercke et al., 10 Mar 2026). First-flight transport can give qualitatively wrong deposition patterns when mixed-material erosion and redeposition are omitted (Effenberg et al., 2024). Standard binding-energy-based sputtering theory can fail in near-threshold compound systems when defect charge states and plasma potential introduce species-dependent escape barriers (Zayachuk et al., 2017).

The unresolved issues are correspondingly specific. For loose-powder sputtering, the current scaling laws are most robust for porosities $\mathrm{Pb}_i^{2+}$2, spherical grains, and keV-range BCA-valid bombardment (Verkercke et al., 10 Mar 2026). For agitated-particle coating, the demonstrated robust size range is approximately 25–500 $\mathrm{Pb}_i^{2+}$3m in the reported plate-and-pit geometry (Lourens et al., 2022). For fusion-wall conditioning, predicted near-pure B layers at high injection rates have not yet been experimentally verified in the carbon environment of DIII-D, and the modeling indicates that full-wall geometry is needed for accurate experimental correlation (Effenberg et al., 2024).

Taken together, these results establish dynamic powder sputtering as a coupled problem of non-steady erosion, directional escape from granular geometries, active particle motion, and repeated recycling. The central lesson is not merely that powders can be sputtered or coated, but that in powder systems the emitted or deposited flux is determined by evolving feedback among target composition, porosity, particle mobility, gas-phase transport, and re-deposition.