The cocoa powder was bought in the supermarket (Van Houten Cocoa, manufacturer Kataoka & Co., Ltd.). All the edible liquids were bought in the supermarket and were used as received (Supplementary Table S1; Supplementary Figure S1). Doubly distilled water prepared using a deionized water system (Advantec MFS RFD240NA: GA25A-0715) was used. The ethanol (purity, 94.8–95.8 v/v%) was from Sigma-Aldrich Japan G.K.

The gas marbles were produced by following the protocol developed by Roux et al. (2022). Cocoa particles were sprinkled onto a flat air-liquid surface to create a dense particle raft. A precise volume of air was then injected beneath the surface using a U-shaped needle (4982 PD 18G, Kyowa Interface Science Co., Ltd.: inner diameter, 0.90 mm) connected to a syringe. The bubble trapped beneath the particles raft was then scooped up and gently rolled on the particle layer using a spatula to ensure full coverage. As the bubble rolled, the particles raft formed a closed air bubble structure. The gas marbles could then be easily removed using two spatulas and placed onto a solid surface for characterization.

The cocoa particles were characterized using first an optical microscope (BX53, Olympus Co.) equipped with a digital imaging system (Moticam 5.0 MP, Shimadzu Co. Ltd.). Then, scanning electron microscopy (SEM) studies were performed on the edible particles coated with a thin layer of gold using a TM4000II miniscope (Hitachi High-Tech Corporation). The volume-average diameter (D _v) of the edible particles dispersed in water was measured using a laser diffraction particle size analyzer (Mastersizer 2000, Malvern). The circularity was also estimated using the ImageJ software by enclosing the outer circle of the particle using freehand selections as described in reference (Schneider et al., 2012). From the cocoa particles characteristics, we calculated the dimensionless Bond number (B _o) defined using Equation 1. B o = Δ ρ   g   L 2 γ (1) where Δρ is the difference in density of the two phases (liquid and particle (density = 1.32 g.cm^-3) (Yagishita et al., 2026); g is the gravitational acceleration; L is the particle diameter and γ is the surface tension of the liquid.

The surface tensions of the liquids were measured using the Wilhelmy plate method with a precision surface tension analyzer (DY-300, Kyowa Interface Science Co., Ltd.). The obtained force-displacement profiles were processed and evaluated using the dedicated analysis software (Dynalyzer, Kyowa Interface Science Co., Ltd.), ensuring reliable and reproducible determination of the interfacial properties.

To measure the static three-phase liquid contact angle, 5 µL droplets were placed on pressed pellets prepared from dried cocoa particles. Measurements were recorded immediately after droplet deposition and again after 10 s (the typical timescale for gas-marble formation), at 25 °C, using a DropMaster DMs-401 system (Kyowa Interface Science Co., Ltd.) and the θ/2 (height–width) method. Error bars represent the standard deviation from three independent replicate measurements performed on separately prepared pellets for each powder.

The cocoa particles used in this study were commercially edible particles, and we first characterized their properties to gain insight into the gas marbles stabilization mechanisms. The D _v in water medium was measured to be 32 ± 20 µm (Figure 1a). The shape of the particles was determined by calculating the circularity value, which was around of 0.78 ± 0.08, using stereomicroscopy images (Figure 1b). Scanning electron microscopy (SEM) observation revealed that the particles had surface roughness with a few micrometer sizes (Figure 1c). The density was measured to be 1.32 ± 0.01 g cm^-3 using a helium pycnometer.

First, we investigated the maximum air-bubble volume that could be stabilized by cocoa particles to produce gas marbles using water (Figure 2). The initial air-bubble volume was varied from 5 μL to 5000 µL. Stable gas marbles were successfully formed for bubbles ranging from 5 µL to 2000 μL, corresponding to diameters of approximately 2.5–15 mm. To our knowledge, this is the first-time gas marbles of such large size have been obtained. Gas marbles exhibited spherical shape for low air-bubble volume from 5 to 500 μL, however for larger initial bubbles (1000–2000 µL), gas marbles were a little bit flattened. Moreover, for even larger initial air bubbles (between 3000–5000 µL) gas marbles could still be formed, but they broke during removal from the particle rafts at the air–water surface (Supplementary Figure S2). Gas marbles exhibited the high water loading for initial bubble volumes between 5 and 10 μL, with values of 67.2% ± 4.4% and 63.0% ± 4.5%, respectively. Water loading then decreased gradually as the initial bubble volume increased from 100 to 500 μL, and eventually reached an approximately stable regime around 54%–58% (Supplementary Figure S3). The observed increase in the water weight fraction for smaller gas marbles can be rationalized by considering the role of gravity during the formation process. For gas marbles with small diameters, gravitational effects are negligible compared with capillary forces, allowing the water film formed around the gas core to retain a relatively large thickness throughout fabrication. In contrast, for larger gas marbles, gravitational drainage becomes significant during formation, causing the liquid initially present at the upper region of the marble to flow downward toward the lower hemisphere. This gravity-induced redistribution of the liquid phase effectively reduces the average thickness of the water film in the resulting gas marble. Consequently, gas marbles of smaller size preserve a thicker water shell and therefore exhibit a higher water weight fraction than their larger counterparts.

All gas marbles, regardless of diameter, were relatively spherical immediately after formation, prior to drying (Figure 2). However, after 1 day of drying at room temperature, gas marbles produced from initial bubble volumes of 1,000 µL and above began to collapse and adopted a non-spherical shape. This result highlights that controlling the initial bubble volume is essential for maintaining the spherical shape of gas marbles during aging. We suppose that the water evaporation plays a key role in the stability of the gas marbles and the decrease in size, and initial water loading is important. The smallest gas marbles contained higher water loading than the larger ones, and they were less sensitive to drying. This contraction of the gas marbles is likely attributable to evaporation of the liquid phase from the particle shell surrounding the enclosed air, which progressively compacts the granular film, reduces the overall marble size, and ultimately leads to jamming at the surface (Stratford et al., 2005). But it is important to note that evaporation and water loading are not the only parameters playing a role here. If evaporation were the only mechanism driving morphological changes, one would expect the height and width of the marbles to decrease equally, since the shell surface was initially widely exposed (with only a small contact area between the shell and the supporting support). However, the photographs clearly demonstrate a significant influence of gravity playing a non-negligible role in these gas marbles systems. There was a vertical collapse of the biggest gas marbles leading to a change from spherical gas marbles to flattened ones after 1 day of drying, highlighting the effect of gravity on their external morphology. Consequently, the larger the gas marble, the more sensitive it is to gravitational deformation due to the greater mass of stabilizing particles comprising the shell.

Building on the results obtained with water, we next investigated the production of gas marbles using various edible liquids while keeping the initial air-bubble volume fixed at 20 µL. We selected five broad categories of edible liquids: seasoning liquids, dairy products, edible beverages (non-alcoholic), alcoholic beverages with different ethanol contents, and vegetable oils. This classification is arbitrary and is used solely to facilitate interpretation of the results (Figure 3; Table 1).

We successfully produced gas marbles for a wide range of edible liquids across these categories. Gas marbles were obtained with all the seasoning liquids, dairy products liquids, drink and alcoholic beverages with low ethanol content such as wine. However, no gas marbles could be formed with vegetable oils, and alcoholic beverages with a high ethanol content (specifically whisky at 37% ethanol and gin at 47.3%). All the gas marbles formed were spherical. Only the gas marbles produced with honey were relatively flattened, adopting a so-called “pancake” shape. This can be explained by the high viscosity of honey compared to the other edible liquids, which are mainly composed of water. The shape of gas marble changed from near-sphere to pancake by mechanical stress applied using a spatula during rolling on the particle raft. Due to the high viscosity, the pancake shape could not be near-sphere to decrease interfacial energy. Once the cocoa particles fully covered the pancake marble surface, the shape could be retained due to interfacial jamming effect (Bala Subramaniam et al., 2005; Stratford et al., 2005; Cui et al., 2013).

To understand why gas-marble formation was or was not possible depending on the edible liquid phase, we first measured the air–liquid surface tension of all liquids. We found that gas marbles could only be formed for liquids with a surface tension above 34 mN/m. Next, we examined the evolution of the three-phase liquid contact angle over time for each liquid. We focused on static contact-angle measurements to ensure consistent comparison across the diverse set of edible liquids, which exhibit a broad range of viscosities. This method provides a general indication of cocoa-particle wettability and their ability to stabilize an air–liquid surface under comparable conditions (Figure 4; Table 1). It is important to note that the reported values represent apparent, time-dependent contact angles measured on porous powder pellets. These values are used as a consistent and comparative metric of wettability, rather than as strictly defined advancing or receding contact angles. Vegetable oils such as canola oil and olive oil showed low initial contact angles (<60°), which decreased rapidly over time, reaching 20°–30° within a few tens of seconds. These liquids did not lead to gas-marble formation. Similarly, alcoholic beverages with high ethanol content (≥30%), such as whisky (37%) and gin (47.3%), also exhibited initial contact angles below 60° that quickly dropped to 20°–30°. These two liquids failed to produce gas marbles. In contrast, more favorable behavior was observed for all other edible liquids. Their initial three-phase contact angles ranged from 70° to 100°, decreasing more moderately to 60°–80° after 40 s. For example, coffee exhibited relatively stable contact angles, starting near 100° at t = 0 s and decreasing only slightly to ∼90° after 40 s. For these liquids, cocoa particles displayed intermediate wettability, which was sufficient to stabilize the air–liquid surface and allow gas-marble formation. A particular case was honey, which showed very high and time-stable contact angles (>140°) due to its high viscosity. Despite the impossibility of measuring its surface tension for the same reason, honey nonetheless supported gas-marble formation.

Overall, these results allow us to classify the edible liquids into two groups. The first group was composed with liquids forming gas marbles, characterized by surface tensions between 34 and 68 mN/m and three-phase contact angles between 60° and 140°. The second group was composed with liquids that do not form gas marbles, characterized by low surface tensions (<34 mN/m) and three-phase contact angles below 60°, which prevent stable particle adsorption and thus inhibit gas-marble formation.

The key parameter that predicts whether gas marbles can form is the three-phase liquid contact angle, which governs particle wetting behavior (Figure 4). This angle determines whether particles remain adsorbed at the air–liquid surface, where they can effectively encapsulate air bubbles or whether they become fully dispersed in the liquid phase, which makes gas-marble formation impossible. An intermediate wettability regime is therefore essential: only when particles are partially wetted, they become energetically trapped at the surface and self-assemble into stable granular films capable of stabilizing gas marbles (Timounay et al., 2015; Timounay, 2016; Timounay and Rouyer, 2017). Our results confirm previous findings obtained with various edible particles in water, which already demonstrated the central role of the three-phase liquid contact angle in determining gas-marble stability. Here, we further show that this parameter can be tuned by the surface tension of the continuous liquid phase, providing a direct link between liquid physicochemical properties and the ability to form gas marbles.

For food applications, a key prerequisite is stability under elevated temperatures. We therefore followed the thermal stability of all successfully produced gas marbles by subjecting them to a heat treatment at 80 °C for 1 week (Figure 5). Photographs taken before and after heating were compared to evaluate morphological changes. All gas marbles remained intact throughout the heat treatment, demonstrating excellent thermal resistance across all edible liquid phases tested. However, a systematic decrease in diameter was observed for every gas marble, regardless of the liquid used for their preparation. For instance, gas marbles produced with water showed a reduction in diameter from approximately 3.7 mm–2.5 mm, corresponding to a 33% decrease after 1 week at 80 °C. The shrinkage was even more pronounced for gas marbles made from alcoholic beverages. For example, Shochu (20% ethanol) produced gas marbles with an initial diameter of ∼3.45 mm, which decreased to ∼1.95 mm after 1 week, representing a 43% reduction. This contraction of the gas marbles is likely attributable to evaporation of the liquid phase from the particle shell surrounding the enclosed air, which gradually compacts the granular film and reduces the overall marble size, leading to jamming at the surface (Stratford et al., 2005). Despite this shrinkage, all gas marbles preserved their structural integrity during heating, indicating robust stability against high-temperature drying, independent of the liquid phase used. We also observed for few gas marbles a darkening after heating (e.g., honey or soy sauce). This darkening effect could be attributed to thermal or oxidative browning of the complex food-based liquids upon exposure to heat and air, and while it does not directly reflect a loss of marble integrity, it may indirectly influence long-term stability through changes in liquid composition and viscosity.

To better understand the peculiar behavior of gas marbles obtained from alcoholic beverages, we investigated a model system consisting of water–ethanol mixtures with ethanol contents ranging from 0 to 100 wt%. The water–ethanol system serves as a model liquid platform that validates the generality of the liquid-dependent criterion identified from the broader set of edible liquids. Gas marbles were successfully formed only for ethanol concentrations between 0 and 20 wt%, confirming the trends previously observed for alcoholic beverages of varying ethanol content (Figure 6). Above 20 wt% ethanol, gas-marble formation was no longer possible. We measured the surface tension and the three-phase liquid contact angle for each mixture. Gas marbles were obtained only when the liquid phase exhibited a surface tension above 34 mN/m and a three-phase contact angle greater than 60° at 0 s (Table 2; Supplementary Figure S4). These observations reinforce the conclusions drawn from the full set of edible liquids: surface tension–controlled wettability is the dominant physical parameter governing gas-marble formation. It is also important to note that ethanol evaporation can induce interfacial instabilities driven by surface tension gradients (Marangoni effects), which may affect the trapping and retention of particles at the surface. That is why we also closely examined the behavior of cocoa particles when spread at the air–liquid surface to form particle rafts (Figure 6). For ethanol concentrations between 30 wt% and 100 wt%, cocoa particles immediately sank into the liquid phase upon deposition at the air–liquid surface. As no particles remained at the surface, none were available to wrap around and stabilize the air bubble, making gas-marble formation impossible under these conditions. This sinking behavior became increasingly pronounced with higher ethanol content, consistent with the progressive decrease in surface tension. In contrast, at low ethanol content (≤20 wt%), the liquids displayed intermediate wettability. When cocoa particles were spread onto these air–liquid surfaces, the particles remained trapped at the surface and formed a stable granular layer. Air bubbles could then be effectively encapsulated by rolling them across this interfacial particle raft, leading to successful gas-marble formation. It is important to note that the Bond number (B _o) was calculated for all liquid phases, and in all cases we obtained B _o ≪ 1 (Table 2) (Protière, 2023). A B _o much smaller than unity normally implies that surface tension should dominate over gravitational forces, and therefore cocoa particles would be expected to remain at the air–liquid surface. However, this was not what we observed experimentally for ethanol-rich liquids. This discrepancy most likely arises from the polydispersity of the cocoa particles in size, shape, and density, making accurate quantitative determination of the B _o challenging (Vella et al., 2006; Vella, 2015).