Living Droplets with Mesoscale Swimmers L. Malik, N. Sharadhi, M. Lamminmäki, R. A. Lara, V. Jokinen, M. Backholm Abstract We study the activity of “living” droplets, which confine 1 − 6 1-6 mesoswimmers in 3D using a superhydrophobic substrate. The swimmers induce oscillations of the droplets at their inherent resonant frequencies, regardless of swimmer size and number. In contrast, the droplet oscillation amplitude is strongly affected by crowding, which we successfully model with a new scaling law and show that crowding reduces the speed of the swimmers. These fundamental living matter physics results reveal mechanisms for bio-inspired droplet actuation with implications for mesoscale robotics, fluidics, and sensing. Droplets enclosing out-of-equilibrium living or artificial swimmers Elgeti et al. 2015 ; Bechinger et al. 2016 present a compelling system for exploring collective dynamics in active, living, and soft matter physics Klotsa 2019 ; Marchetti et al. 2013 . Previous studies on confined and/or crowded active matter, such as active nanoparticles Kokot et al. 2022 , motile colloids Bricard et al. 2013 ; Vutukuri et al. 2020 , active network of microtubules Sciortino et al. 2025 , acousto-magnetic microswimmers Ahmed et al. 2017 , robots Boudet et al. 2021 and robotic fish Boudet et al. 2025 , as well as living organisms, such as fish Ventéjou et al. 2024 ; Lafoux et al. 2024 , bacteria Wioland et al. 2013 ; Lushi et al. 2014 ; Soto and Golestanian 2014 ; Villalobos-Concha et al. 2025 , algae Ostapenko et al. 2018 ; Cammann et al. 2021 ; Mondal et al. 2021 ; Tainio et al. 2021 ; Wei et al. 2024 , spermatozoa Kantsler et al. 2013 , and theoretical microswimmers Soto and Golestanian 2014 ; Sprenger et al. 2020 ; Kawakami and Vlahovska 2025a ; Kawakami and Vlahovska 2025b , has revealed how concerted motion of internal swimmers can drive droplet shape fluctuations and self-propulsion, as well as organized fluid flows, especially within geometrically constrained environments. Most studies have focused on the effects of crowding and confinement in 2D using cylindrical disks or planar surfaces. Recently, the 3D confinement of microscopic swimmers has gained attention, with bacteria confined in droplets Villalobos-Concha et al. 2025 and 3D chambers Wei et al. 2024 as well as active particles Vutukuri et al. 2020 and microtubules Sciortino et al. 2025 deforming giant lipid vesicles. Studying the effect of 3D confinement is of great fundamental interest since most organisms naturally swim in 3D. Liquid droplets represent excellent spherical experimental microcompartments for probing 3D confinement. These can be achieved by using superhydrophobic surfaces, where liquid sessile droplets become nearly perfect spherical pearls Quéré 2005 . The mechanical response of sessile droplets on such surfaces has been described with classical resonance models, which relate the fundamental mode oscillation frequency to droplet volume, contact angle, and surface tension Noblin et al. 2004 ; McHale et al. 2009 ; Celestini and Kofman 2006 ; Sharp et al. 2011 ; Bostwick and Steen 2014 ; Chang et al. 2015 ; Sakakeeny et al. 2021 . Yet, these have exclusively focused on externally activated, non-living droplets, and there is no understanding of how the classical models generalize to droplets that are internally activated by living swimmers. We expect measurable effects, especially in the mesoscale regime where the size of the swimmer L is comparable to the droplet radius R . This physical regime is not only relevant to living and soft matter physics, but also supports the upscaling of droplet-based sensors Kaminski et al. 2016 and swimmer-driven micro-machines Hiratsuka et al. 2006 ; Leonardo et al. 2010 ; Oda et al. 2024 to the mesoscale. Furthermore, a wide range of living micro- to meso-organisms live as aeroplankton within atmospheric water droplets Smith 2013 , affecting the weather Christner et al. 2008 and public health through, for example, allergies and as transmission vectors Smith 2013 . How actively swimming meso-organisms influence the mechanics of such droplets remains unstudied. Figure 1: Experimental micrographs of living droplets supported by a superhydrophobic black silicon substrate, exhibiting an average contact angle \theta , with (a) N=1 and (b) N=6 Artemia swimmers of length L=490\pm 20\ \mu m. Images are shown for droplet volumes V=3-13.5\ \mu L, corresponding to radii R=895-1550\ \mu m and L/R=0.30-0.57 . Droplet deflections are measured along the x direction. Scale bar 200\ \mu m. Figure 2: Oscillatory signal, frequency response and characterization. (a) Deflection versus time plot shows a typical oscillatory signal for an experimental case of N=6 Artemia swimmers ( 48 h old) inside a droplet of volume \ V\sim 4\ \mu L. A (dashed line) denotes the root-mean-square of the time-varying amplitude. Inset shows a zoomed-in view of an oscillation envelope. (b) Fourier transform of the oscillatory signal in (a), showing two different responses: f_{\mathrm{1}} and f_{\mathrm{2}} which denote the resonant and the envelope frequencies, respectively. (c) f_{\mathrm{1}} plotted as a function of droplet volume for varying N and age of Artemia as well as a few experiments with a copepod ( N=1 ). The best experimental fit to the data ( f_{\mathrm{1}}=18.7\ V^{-0.5} ) is found to be in good agreement with the CK model ( f_{\mathrm{1}}=22.7\ V^{-0.5} ) Celestini and Kofman 2006 ; Sakakeeny et al. 2021 . Here, we investigate the effects of 3D confinement and crowding on the swimming of living mesoscale organisms trapped in nearly spherical droplets (see Fig. 1 and Movies S1 - S7 Sup ) on a superhydrophobic surface. Our experimental work is the first to focus on crowding and confinement in 3D at L\sim R , where the effect of the confinement will be very different than at the previously studied L\ll R regime. Due to the major optical challenges associated with observing individual organisms swimming inside a 3D droplet, we develop a new experiment to probe the intricate many-body system through the active oscillations of the living droplets. We primarily use Artemia nauplii as a model organism for their accessibility and well-characterized butterfly swimming mechanism Azra et al. 2022 ; Williams 1994 ; Lara et al. 2025 , and complement our study with adult copepods for their distinct intermittent swimming strategies characterized by powerful swimming bursts Svetlichny et al. 2020 . We incorporate classical resonance models Sakakeeny et al. 2021 ; Celestini and Kofman 2006 — originally developed for non-living droplets — for predicting oscillation frequencies in swimmer-laden droplets across varying swimmer lengths, numbers, and droplet volumes. To describe the amplitude of the droplet oscillations, we develop a new theoretical scaling law and find excellent agreement with experiments in the unconfined and uncrowded regime. Furthermore, we introduce a crowding–confinement factor that accounts for mesoscale swimmer interactions within the droplets, and report a measurable reduction in swimming velocity as the swimmer number density increases. Below, the experiments are briefly described, see Materials and Methods section in the Supplementary Information (SI) Sup for all technical details. Artemia sp. are hatched in saline solution under controlled conditions and collected via positive phototaxis for gentle transfer and accurate counting Williams 1994 ; Azra et al. 2022 ; Stappen et al. 2024 . Droplets of volume V=2-20\ \mu L corresponding to radii R=600-1750\ \mu m, containing N=1-6 Artemia nauplii either 24 or 48 h old with lengths L=490\pm 20\ \mu m and 710\pm 30\ \mu m, respectively, are used. A few experiments were also performed with an adult Acartia tonsa copepod of L=1120\pm 20 \mu m (see Fig. S1 for images of organisms Sup ). Nearly spherical droplets are created by using a superhydrophobic black silicon substrate coated with fluoropolymer Sainiemi et al. 2011 , yielding an average contact angle of \theta=163^{\circ}\pm 3^{\circ} and extremely low kinetic friction force Backholm et al. 2020 ; Backholm et al. 2024 . A vertical glass micropipette is used to assist in detecting the interface oscillations (see Fig. S2 Sup for details of analysis). The micropipette is manufactured using a micropipette puller and microforge as described in Backholm and Bäumchen 2019 , and its tip is positioned symmetrically just inside the droplet interface from the top for consistent lateral deflection measurements as shown in Fig. 1 . Micropipettes with the lengths of 2.5-2.8 cm are used in all experiments. Due to the low spring constant of the micropipettes ( 5.2\pm 0.1 nN/ \mu m for Artemia and 2.44\pm 0.03 nN/ \mu m for copepods; see Fig. S3 and SI for details on the calibration Sup ), the elastic contribution from these did not influence the droplet oscillations (see Movie S8 for droplet oscillations without a micropipette and Fig. S4 for comparison of oscillation data with and without a micropipette Sup ). The experiments are observed at 40–100 fps at 5x magnification with a side-view camera (FLIR, GS3-U3-23S6M-C, Integrated Imaging Solutions, Inc.) using back-illumination with an LED (BT100100-WHIIC, Edmunds Optics, UK). The entire setup (see Fig. S5 Sup ) is mounted on a vibration-isolated table (Halcyonics_i4large, Accurion) and shielded with a black hardboard box (Thorlabs). Image analysis quantifies droplet geometry and the micropipette deflection as a function of time. Representative micrographs for N=1 and N=6 swimmers ( 24 h old) are shown in Fig. 1 (Movies S1 - S7 Sup ), with L , R , \theta , and micropipette deflection x indicated. See Fig. S6 and Movie S9 for similar examples with 48 h swimmers Sup . The detection of droplet activity using the micropipette leads to a typical oscillatory signal as shown in Fig. 2 a (see Movie S10 Sup ). The root-mean-square of the time-varying amplitude, A , is hereafter referred to as the average experimental amplitude. The Fourier transform of the x-t signal shown in Fig. 2 b highlights the frequency response. Also see Fig. S7 and Movie S11 Sup , which present the oscillatory signal and its Fourier transform for the case of copepods. The swimming bursts of the copepod are evident in the deflection–time plot, showing large peaks separated by damped harmonic oscillations. This behaviour contrasts with Artemia ’s continuous swimming (see Fig. 2 a and Movie S10 Sup ), as Artemia drive oscillations with rhythmic appendage strokes, while adult copepods do so through burst-like intermittent swimming strokes. In contrast, encapsulated microscopic Paramecia exhibited no such effect (see Movie S12 Sup ), likely because their steady, symmetric ciliary beating produces smooth propulsion without creating the large momentum transfer required for droplet oscillations. In the case of Artemia , as shown in Fig. 2 b, we observe two distinct peaks: f_{\mathrm{1}} , associated with the fundamental mode of resonance of the oscillating droplet and f_{\mathrm{2}} associated with the formation of repetitive patterns which we refer to as “envelopes”, also shown in a zoomed-in view via the inset of Fig. 2 a. The lower frequency response, f_{\mathrm{2}} may be attributed to the occurrence of simultaneous or synchronized impact events, at times from multiple swimmers, collectively enhancing the droplet oscillatory motion. We find f_{\mathrm{1}} to be roughly two orders of magnitude higher than f_{\mathrm{2}} , but observe no connection between f_{\mathrm{2}} and the level of confinement or crowding. We do not investigate this mode of oscillation in more detail. We plot f_{\mathrm{1}} against V for all experiments with varying V , N , and L of Artemia and few cases of copepods, as shown in Fig. 2 c (see Fig. S8 for better distinction between the different experiments Sup ). We find f_{\mathrm{1}}\sim V^{-0.5} in good agreement with the CK model f_{\mathrm{1}}=\left[{\gamma\,\eta\,(1-\cos\theta)}/{(2\pi\,\rho_{d})}\right]^{1/2}\,V^{-0.5} , where \eta(\theta)=0.52\left(e^{0.99(1+\cos\theta)}-1\right) , \rho_{\mathrm{d}} is the droplet density, \gamma the surface tension, and \theta the contact angle Celestini and Kofman 2006 ; Sakakeeny et al. 2021 . This model assumes a Bond number \mathrm{Bo}=\rho_{\mathrm{d}}\ gR^{2}/\gamma=0 , where g is the gravitational acceleration. This is similar to our experimental conditions of \mathrm{Bo}\sim 0.05-0.4\ <1 . We observe a slight deviation from the scaling as \mathrm{Bo} approaches 1 , but no measurable deviations due to the crowding or confinement of the swimmers. Figure 3: Theoretical prediction of droplet amplitude and characterization of crowded swimmer velocity. (a) Schematic drawing of two Artemia swimmers ( L^{*}=0.4 ) moving with a velocity, U_{\mathrm{c}} and impacting the droplet interface while swimming within distance L from the periphery characterized by peripheral volume, V_{\mathrm{p}} . (b) Comparison of average experimental amplitude, A and theoretical amplitude, A_{\mathrm{t}} (Eq. 1 ) for the case of unconfined ( L^{*}<0.4 ) and uncrowded ( N^{*}<0.03 ) swimmers. The least crowded cases with N=1 are indicated. The solid line is A=1.8A_{\mathrm{t}} , indicating an excellent agreement between the experimental data and theory. (c) The crowded swimmer velocity, U_{c} normalized with the average swimmer velocity in the least crowded case, U_{\mathrm{0}} (average of N=1 and L^{*}<0.4 data, filled markers) plotted against the dimensionless swimmer density, N^{*} . The experimental data of robotic fish confined in 2D Boudet et al. 2025 is shown as a comparison in red. An empirical scaling of U_{\mathrm{c}}/U_{\mathrm{0}}\sim{N^{*}}^{-0.7} is found for our experiments as opposed to U_{\mathrm{c}}/U_{\mathrm{0}}\sim{N^{*}}^{-0.4} for the 2D robotic fish, indicated by black and red lines, respectively. The collapse of all experimental data for different N , L , and species is surprising. Copeopods and Artemia cover similar ranges of swimming frequencies ( 4-6 Hz and 3-10 Hz, respectively) but use very different swimming styles: copepods with a powerful beat–sink swimming mechanism and Artemia with a continuos butterfly flapping of its antennae. Yet, this does not translate to different droplet oscillation frequencies. The different swimming kinematics as well as swimmer number and size should strongly influence the internal fluid flow of the droplets. In an extended model by Sakakeeny et al. Sakakeeny et al. 2021 , a kinetic energy term for the internal flow is included. However, for superhydrophobic surfaces and very low Bo, the difference between the Sakakeeny Sakakeeny et al. 2021 and CK Celestini and Kofman 2006 model is negligible and the effects of the internal flow is thus minimal. Crowding and swimmer activity could be more pronounced in lower contact angle sessile droplets. Our results on high contact angle substrates demonstrate that the dominant oscillation mode corresponds to the inherent mechanical resonant frequency of the droplet, irrespective of the living activity. Interestingly, the primary effect of internal activity instead manifests through changes in the oscillation amplitude. To fully understand the activity of the living droplets, we introduce a theoretical scaling law to predict the droplet oscillation amplitude. We consider the impact probability and the collision mechanics of our swimmer-droplet system. The probability that an Artemia bumps into the droplet interface is given by the ratio of the peripheral volume V_{\mathrm{p}} to the total droplet volume V as shown in Fig. 3 a, P\sim V_{\mathrm{p}}/V\sim[V-(4\pi/3)(R-L)^{3})]/V , assuming a spherical droplet for simplicity. This simplifies to P\sim 3(L/R)-3(L/R)^{2}+(L/R)^{3} . Assuming momentum conservation during an impact between Artemia and the droplet yields: \rho_{\mathrm{d}}VU_{\mathrm{d}}\sim\rho_{\mathrm{a}}V_{\mathrm{a}}U_{\mathrm{a}} , where \rho_{\mathrm{d}} , \rho_{\mathrm{a}} and V , V_{\mathrm{a}} are the mass densities and volumes of the droplet and Artemia , respectively, and U_{\mathrm{d}} is the droplet velocity right after impact and U_{\mathrm{a}} the Artemia velocity right before impact. We model the volume of Artemia as an ellipsoid (Fig. S9 Sup ) and assume \rho_{\mathrm{d}}\sim\rho_{\mathrm{a}} . Considering f_{\mathrm{1}} to be the dominant frequency of oscillation, the droplet velocity can be approximated as U_{\mathrm{d}}\sim 4A_{\mathrm{0}}f_{\mathrm{1}} given that the interface would move a distance of \sim A_{\mathrm{0}} in a time span of \sim 1/4f_{\mathrm{1}} , where A_{\mathrm{0}} refers to the average amplitude contribution by one swimmer when colliding with the drop. This simplifies the momentum conservation equation to 4VA_{\mathrm{0}}f_{\mathrm{1}}\sim V_{\mathrm{a}}U_{\mathrm{a}} , which yields A_{\mathrm{0}}\sim U_{\mathrm{a}}V_{\mathrm{a}}/4Vf_{\mathrm{1}} . We finally make the first-order assumption that the total droplet oscillation amplitude caused by many swimmers scales with the drop amplitude caused by one swimmer, the probability of a droplet collision, as well as the number of swimmers in the droplet: A_{\mathrm{t}}\sim PA_{\mathrm{0}}N . This is given by A_{\mathrm{t}}\sim\dfrac{U_{\mathrm{a}}V_{\mathrm{a}}N}{4Vf_{\mathrm{1}}}\left[3\dfrac{L}{R}-3\left(\dfrac{L}{R}\right)^{2}+\left(\dfrac{L}{R}\right)^{3}\right]. (1) To compare A_{\mathrm{t}} with the average experimental amplitude A , we first consider only the unconfined and uncrowded swimming cases, defined by L^{*}<0.4 and N^{*}<0.03 , respectively, where L^{*}=L/R and N^{*}=NV_{\mathrm{a}}/V refer to the dimensionless swimmer length and dimensionless swimmer density. In this regime, we can assume that the Artemia are swimming freely, and use swimming velocities measured in Lara et al. 2025 on single Artemia without confinement: U_{\mathrm{a}}=1.2\pm 0.5 mm/s for L=500\pm 10 \mu m and U_{\mathrm{a}}=2.8\pm 0.9 mm/s for L=720\pm 10 \mu m. The comparison is presented as a parity plot in Fig. 3 b, showing an excellent agreement, with the least crowded condition ( N=1 ) highlighted. Crowding and confinement are known to reduce the speed of swimming Boudet et al. 2025 ; Mondal et al. 2021 . We define a dimensionless factor K_{\mathrm{c}} to capture the effects of crowding on the swimming velocity of Artemia , such that the modified velocity is U_{\mathrm{c}}\sim K_{\mathrm{c}}U_{\mathrm{a}} , where K_{\mathrm{c}} is assumed to depend on N^{*} . This allows us to empirically measure K_{\mathrm{c}}=A/A_{\mathrm{t}} from the deviation between the theoretical free-swimming model of A_{\mathrm{t}} and the experimentally measured A . We plot the resulting U_{\mathrm{c}}/U_{\mathrm{0}}=K_{\mathrm{c}}U_{\mathrm{a}}/U_{\mathrm{0}} as a function of dimensionless swimmer density, N^{*} in Fig. 3 c, where U_{\mathrm{0}}=K_{\mathrm{c,0}}U_{\mathrm{a}} is the average swimmer velocity in the least crowded condition ( K_{\mathrm{c,0}} is the slope of N=1 and L^{*}<0.4 data in Fig. 3 b). We find a significant reduction in the normalized swimmer velocity with increasing swimmer population inside the droplet. This is expected since Artemia stops and turns when encountering an obstacle Çarkoğlu et al. 2015 , which in this case is another organism or the droplet interface. Such a behaviour will drastically reduce the time-averaged swimming speed in crowded and confined environments. Our experimental and analytical approach quantitatively relates the crowding of swimmers inside the droplet to their collective motion, and consequently, the overall droplet dynamics. By assessing changes in oscillation amplitude, we infer reductions in swimmer velocity due to crowding, reflecting complex mesoscale interactions in 3D confinement. To verify the validity of our results, we compare our empirical findings with the direct swimming speed measurements on crowded robotic toy fish in 2D confinement by Boudet et al . Boudet et al. 2025 (red data in Fig. 3 c plotted as a function of area density NA_{\mathrm{fish}}/A_{\mathrm{c}} , where A_{\mathrm{fish}}=0.25\pi\cdot 7.6\cdot 1.9 cm 2 is the cross-sectional area of the fish and A_{\mathrm{c}}=\pi R_{\mathrm{c}}^{2} is the area of the 2D circular confinement of radius R_{\mathrm{c}}=0.15 m). Both datasets show a decrease in swimming speed with increased swimmer density. The difference in the slope of the data, which is empirically quantified by a power law fit of the form U_{\mathrm{c}}/U_{\mathrm{0}}\sim{N^{*}}^{-0.4} for the toy fish and U_{\mathrm{c}}/U_{\mathrm{0}}\sim{N^{*}}^{-0.7} for Artemia , may be attributed to several differences between the systems. Firstly, our experiment involves 3D confinement at L\sim\mathcal{O}(R) whereas the toy fish move in 2D confinement at L\ll R . Secondly, we have worked with a living organism which actively adapts and responds to its environment in a much more intricate manner than a robotic toy fish. Lastly, the difference in Reynolds number at the mesoscale of Artemia ( \mathrm{Re}=10^{0}-10^{1} Lara et al. 2025 ; Williams 1994 ) and inertial macroscale of the toy fish ( \mathrm{Re}=10^{2}-10^{3} Boudet et al. 2025 ) could also contribute through different hydrodynamic interaction mechanisms. In this work, we have explored the dynamic behaviour of living droplets containing actively swimming meso-organisms. We achieve nearly spherical, 3D confinement by depositing droplets on superhydrophobic substrates. Our experimental and theoretical investigation shows that droplet oscillation frequencies, surprisingly, closely follow classical resonance predictions, regardless of the level of confinement and crowding of the swimmers in the droplets. In contrast, we find that the amplitude of the droplet oscillation is strongly affected by the density of swimmers, and develop a new scaling law model to describe the physics of the system. We find excellent agreement between the data and theory for unconfined and uncrowded swimmers and use the new model to further understand the collective effects of the swimmers on the drop oscillations. We find that the speed of the swimmers drastically decreases in response to increased crowding, which is in agreement with previous studies on 2D-confined toy fish. This work offers a foundation for understanding how droplets containing confined mesoscale swimmers or active matter behave. Our findings point to new opportunities for fluid control and transport at small scales using living organisms, which may enable responsive systems for sensing and actuation. Future studies that explore internal flows could further expand the usefulness of these living droplet platforms. Data Availability The datasets used to plot graphs in the paper and examples of raw data files are shared on Zenodo Malik et al. 2025 .