Turbulence-assisted formation of bacterial cellulose
Sung-Ha Hong, Jia Yang, Mahdi Davoodianidalik, Horst Punzmann, Michael Shats, Hua Xia
TTurbulence-assisted formation of bacterial cellulose
Sung-Ha Hong, Jia Yang, Mahdi Davoodianidalik, Horst Punzmann, Michael Shats and Hua Xia a) Research School of Physics, The Australian National University, Canberra ACT 2601,Australia (Dated: 15 September 2020)
Bacterial cellulose is an important class of biomaterials which can be grown in well-controlled laboratory andindustrial conditions. The cellulose structure is affected by several biological, chemical and environmentalfactors, including hydrodynamic flows in bacterial suspensions. Turbulent flows leading to random motionof fluid elements may affect the structure of the extracellular polymeric matrix produced by bacteria. Herewe show that two-dimensional turbulence at the air-liquid interface generates chaotic rotation at a well-defined scale and random persistent stretching of the fluid elements. This leads to a remarkable change inthe structure of the bacterial cellulose morphology and to the formation of spherical cellulose beads whosesize is controlled by the turbulence forcing scale. In addition, turbulence affects the cellulose microstructure,including the porosity of the material, by random but persistent stretching of the fluid elements. The resultsoffer new approaches to engineering of the bacterial cellulose structure by controlling turbulence parameters.
I. INTRODUCTION
Bacterial biofilms represent the most abundant form oflife on Earth . Microorganisms form extracellular poly-meric matrices in a fluid environment, typically at theliquid-air, liquid-liquid and liquid-solid interfaces. Someof the biofilms, such as the bacterial cellulose (BC) , arevaluable biomaterials with great industrial and medicalpotential . The BC is produced by various bacteria,for example, by the genera Acetobacter, Acanthamoeba,and Achromobacter, at the liquid-air interface in a vari-ety of forms, such as cellulose sheets , or spherical cellu-lose beads (SCB) . The SCB is a desired cellulose formin many applications . Spherical beads make perfectdelivery vehicles for many functional materials, such asgraphene, magnetic and conducting metal particles andother chemical compounds which can be encapsulatedinto BC spheres .An attractive feature of the BC is the possibility tocreate cellulose with a desired fibre structure at bothmacroscopic and nanoscopic scales in the process of thematerial formation, rather than using post-productiontreatment methods. Along with chemical and biologicalprocesses which affect the biofilm development, hydro-dynamic flows in bacterial suspensions are also capableof shaping biofilms . It has been recently demonstratedthat when suspensions of E. coli are perturbed by surfacewaves, patterned biofilms at the liquid-solid interface aredeveloped .Surface waves can form ordered patterns, but they canalso lead to the generation of turbulent flows at the air-liquid interfaces . Such flows were shown to be verysimilar to quasi-two-dimensional (2D) turbulence whichcan be generated in laboratory . Turbulence greatlyenhances particle dispersion . Thus, it can signifi-cantly affect the growth of the extracellular polymericmatrix during the biofilm development and can change a) Electronic mail: [email protected] the resulting structure of the cellulose. Since the wave-driven fluid motion decays in the bulk of the liquid ,turbulent motion is the strongest in a relatively thin layer(a fraction of the wavelength) near the liquid-air inter-face. The BC produced at the air-liquid interfaces isparticularly suitable to test this idea. Since the microor-ganisms forming the BC typically are not motile, flowsin bacterial suspensions should not be affected by themotion of bacteria, as it occurs in suspensions of fastswimming micro-swimmers which can generate turbu-lence themselves .Turbulence is a strongly non-equilibrium state of aflow characterised by broad kinetic energy spectra andenhanced dispersion of fluid particles . Finite-size par-ticles in turbulence experience random walk while theyare also exposed to random torques causing rotationaldiffusion . At small scales, the rotational diffusion co-efficient associated with the Brownian motion stronglydecays with the increase of the particle size d : D rot ∝ T / ( µd ) (where µ and T are the fluid viscosity and itstemperature) and it is negligibly small in the millimetrerange of scales . However, in 2D turbulence, the parti-cle dispersion is governed by the scale comparable to theagitation (forcing) scale . Therefore, this scale mayaffect the characteristic scale of the bacterial fibre struc-ture. In addition, turbulent flows generate stretching ofthe fluid elements which can also affect the formation ofthe bacterial cellulose.Here we study the BC development by Gluconaceto-bacter xylinus at the liquid-air interface in a turbulentquasi-2D flow. We show that turbulence indeed gener-ates chaotic rotation in a fluid. The angular velocity ofturbulent rotation is the strongest at the scale just be-low the forcing scale of turbulence, while above this scaleit decays exponentially. Such turbulent rotation leadsto a dramatic effect on the formation of the G. xylinus
BC near the liquid surface. While in non-perturbed liq-uid, microorganisms form continuous sheets of the BC,in turbulent flows they form spherical beads whose sizeis proportional to the forcing scale of turbulence. Tur-bulence also effects the microscopic structure of the BC, a r X i v : . [ phy s i c s . b i o - ph ] S e p abd e
10 mm c
10 mm motion control amplifierPC shaker actuatortemperature control
Control sample Turbulent flow stacked6-wellcultureplates Incubator 1 Incubator 2 a cc e l e r a t i o n s i g n a l T T FIG. 1. (a) Schematic of the experimental setup. Bacterial suspensions of
G. xylinus are placed inside temperature controlledincubators 1 and 2. Incubator 1 houses control (non-vibrated) samples with a non-perturbed air-liquid interface (b). Samplesin the Incubator 2 are vertically vibrated at the prescribed frequency and acceleration. (c) The Faraday waves on the surfaceof the suspension generate turbulent flows. (d) The bacterial cellulose is developed in the control incubator in the form of acellulose sheet. (e) In the suspension exposed to turbulence, cellulose develops in the form of spherical beads. namely, the biofilm porosity due to chaotic stretching ofbacterial suspensions.
II. RESULTSA. Bacterial cellulose formation in quasi-2D turbulence
Effects of turbulence on the formation of the BC arestudied in bacterial suspension of
G. xylinus . Turbulenceis generated using parametrically excited waves (Faradaywaves) in vertically shaken containers (Fig. 1a and Meth-ods). Here vertical acceleration is characterised by a su-percriticality parameter α = ( a − a th ) /a th , where a is theamplitude of the sinusoidal acceleration of the containerand a th is the threshold acceleration for the parametricgeneration of the Faraday waves. Characteristics of thesurface flow generated by such waves closely resemble 2Dturbulence statistics generated using other methods, forexample electromagnetically driven turbulence in layersof electrolytes . By changing the vertical accelera-tion of the container, one can control the kinetic energyof turbulence, while the frequency of the vibration con-trols the energy injection scale L f (related to the Faradaywavelength λ ) . Here the kinetic energy of the hor-izontal fluid motion (per unit mass) is characterised bythe mean-squared fluctuating velocity at the fluid sur-face, (cid:10) u (cid:11) , which in these experiments is changed in therange of (cid:10) u (cid:11) = (0.5-20) × − m s − . This energyis proportional to the supercriticality parameter α (seeMethods), which ranges from 0.1 to 3 in these experi- ments. The forcing scale of turbulence is varied between2 and 5 mm which corresponds to the oscillation frequen-cies of f = (45 − G.xylinus are exposed to turbulent flows for 72 hours (seeMethods).We observe the formation of two forms of the bacterialcellulose. Flat cellulose sheets, or pellicles, are formed atthe surface of the suspension in the control samples (novibration) and at lower turbulence energies, at α < . α (and the level of turbulent velocity fluc-tuations) is increased, spherical beads of cellulose embed-ded within the biofilm develop in the suspension, Fig. 1e.The diameter of the spherical beads is reduced with thereduction in the turbulence forcing scale (increase in thevibration frequency). Figs. 2a,b illustrate SCBs formedat the frequencies of f = 60 Hz and 120 Hz respectivelywith a supercriticality of α = 1 .
5. The scanning elec-tron microscopy (SEM) images illustrate the beads inFig. 2c,d. The probability density functions of the beadsizes for these two cases are shown in Fig. 2e. Peaks ofthe PDFs at d b ≈ .
95 mm (120 Hz) and d b ≈ . L f = 2.5 and 4.4 mm. In the next section we discuss howthe turbulence forcing scale can affect the bead size.In addition to changing the macroscopic morphology ofthe cellulose, turbulence also substantially modifies mi-croscopic structure of the cellulose fibres. The fibre struc-ture of the cellulose formed in the presence of turbulentflows is rarefied, as seen in the SEM images of Figs. 3a-d. The analysis of the planar SEM images shows thatthe size and the area of the pores in cellulose, as well
10 mm a L f = 2.5 mm d b (mm) PDF ( d b ) L f = 4.4 mm b ec d L f = 4.4 mm L f = 2.5 mm μ m FIG. 2. Spherical cellulose beads formed in turbulent flows forced at the scale (a,c) L f = 4 . f s = 60 Hz), and (b,d) L f = 2 . f s = 120 Hz). (c,d) Cellulose bead images using SEM. (e) Probability density functions of the bacterial cellulosebeads formed at different L f . as their probability distributions change in turbulent en-vironment, Fig. 3e. In the samples formed in unper-turbed bacterial suspensions (control samples), the PDFof the pore diameters A decays exponentially as PDF(A) ∝ exp ( − A/A ), where A ≈ . µ m. In the sam-ples formed in turbulence, these PDFs are much broader,and the pore diameter distribution follows an exponen-tial scaling, PDF(A) ∝ exp ( − A/A ), where A ≈ . µ m.Changes in the microscopic structure of the BC are notcorrelated with the forcing scale of turbulence: similarPDFs of the pore size are observed in turbulence excitedat L f = 5 mm ( f = 45 Hz) and L f = 2 . f = 150Hz), Fig. 3e. The effect of turbulence on the fibre struc-ture does not have a threshold in the turbulence intensity,in contrast to the formation of the cellulose beads. Themicroscopic structure in turbulence changes regardless ofits macroscopic shape, both in the flat sheets, and in thespherical cellulose beads.To understand the effects of turbulence on the celluloseformation, we study two processes which can lead to (1)the formation of spherical beads, and (2) the changes inthe microscopic structure of the cellulose fibres, namely(1) turbulent rotation and (2) stretching of fluid elementsin turbulent flows. B. Turbulent rotation and stretching in quasi-2Dturbulence
It is recognised that the BC formation in agitated orshaken culture methods can lead to the production ofdifferent particle sizes and various shapes (spherical, el-lipsoidal, fibrous suspensions etc.) . The relationshipbetween the flow conditions and the resulting shape ofthe BC particles however is not established. The wave-driven turbulence provides an opportunity to study theBC development under well-controlled conditions sincethe statistical characteristics of such flows can be accu-rately characterised by a few key parameters: turbulenceforcing scale, and the turbulence kinetic energy whichis proportional to the supercriticality α (see above) .These parameters control the diffusive motion of fluidparticles at the liquid surface . In addition to dis-placement of particles in turbulence from their initialposition in a random-walk fashion, turbulent flows alsolead to random rotation of finite size objects due to thegeneration of turbulent torque . Such torque can leadto bending of the BC fibres and to the changes in themacroscopic structure of the cellulose. In quasi-2D tur-bulence, it is expected that the rotation is small at thescales smaller than the forcing scale since the flow con-sists of coherent bundles of fluid particles . However,if a particle is exposed to the forces produced by several Control sample Turbulent suspension da b c μ m 2 μ m e α =0.6
150 Hz, α =0.6 A ( μ m) PDF ( A ) FIG. 3. Microscopic structure of the bacterial cellulose (SEM). (a,b) Cellulose structure of the control (non-shaken) samples.(c,d) Structure of the cellulose formed in the turbulent bacterial suspension at the excitation frequency f = 45 Hz ( L f = 5mm) and acceleration a = 0 . g ( α = 0 . A in control samples(red circles) and in the samples vibrated at f = 45 Hz (blue squares) and f = 150 Hz (green diamonds) at the supercriticalityparameter α = 0 .
6. A green circle in (a) highlights a
G. xylinus microorganism. ca b t (s) )0.1 1 k / k f E k (m /s ) k -5/3 -5 -8 -7 -6 d d / Lf D θ (rad2/s) Lf = 4.4 mm d d θ α =1.5 α =0.75 α =1.5 α =0.75 Lf = 4.4 mm Lf = 4.4 mm α =1.5 FIG. 4. Turbulent rotation and angular displacement characteristics in 2D turbulence. (a) Kinetic energy spectra on thesurface of the liquid at two levels of the vertical acceleration α = 0 .
75 (green triangles) and α = 1 . d d = 3 mm (blue circles) and 8 mm (greentriangles). The forcing scale of turbulence is L f = 4 . α = 1 .
5. At short times, MSAD ∝ t , indicating theballistic regime. At longer times, MSAD= D θ t , indicating diffusive regime. (c) The rotational diffusion coefficients versus theratio d d /L f of the disk diameter over turbulence forcing scale. These are measured at two levels of the supercriticality α = 0 . α = 1 . of such bundles, one can expect that turbulent torqueswill lead to measurable fluctuations in the particle angu-lar velocities. The rotational diffusivity in such flows hasnot been studied before.Similarly to random translation of particles, the rota-tional diffusivity can be characterised by the mean squareangular displacement (MSAD) δθ ( t ) = (cid:80) t | ∆ θ ( t ) | ,where ∆ θ ( t ) = θ ( t + ∆ t ) − θ ( t ). The MSAD can betracked in time by following the orientational dynamicsof the disks floating on the surface of the liquid (picturedin the inset of Fig. 4(b), see Methods). No mean ro-tation is observed, as expected in an isotropic turbulentflow. In wave-driven, quasi-2D turbulence, kinetic en-ergy spectra are determined by the inverse energy cascaderange ( E k ∝ k − / at k ≤ k f ) and the forward enstrophycascade ( E k ∝ k − at k ≥ k f ) . Figure 4a shows themeasured spectra in these experiments. The MSAD cor-responding to these spectra show that turbulent rotationof the disks is described by the ballistic rotation regimeat short times, and diffusive rotation at longer times, Fig.4b: (cid:104) δθ (cid:105) ≈ V θ t , t (cid:28) T θ (1) (cid:104) δθ (cid:105) ≈ D θ t, t (cid:29) T θ (2)Here T θ is the autocorrelation time of the angular velocity T θ = 1 / (cid:104) ˜ ω (cid:105) (cid:82) ∞ (cid:104) ω ( t + t ) ω ( t ) (cid:105) dt , where ω = dθ/dt is theangular velocity of the disks, and (cid:104) ˜ ω (cid:105) is its variance.For the cellulose formation occurring on the time scalesof hours, we are interested in the long-time behaviour.We measure the rotational diffusion coefficient D θ at dif-ferent ratios of the disk diameter d d to the forcing scale L f . This coefficient, shown in Fig. 4c, is peaked at d d /L f < d d /L f . This behaviour is also observed in a different type ofquasi-2D turbulence, which is driven electromagnetically.The rate of diffusion is proportional to the turbulence ki-netic energy: the increase in the variance of the turbulentvelocity fluctuations (cid:104) u (cid:105) leads to the increase in D θ , Fig.4c. Such turbulent torques may lead to random bendingof the BC fibres resulting in the formation of the BCbeads whose size is correlated with the turbulence forc-ing scale L f .To understand how turbulence can affect the micro-scopic structure of the BC, we study the stretching ofthe fluid parcels . Two adjacent fluid particles remainclose to each other for a long time as they move within acoherent bundle. Occasionally these bundles split leadingto the stretching of the fluid parcels. Extreme stretchingevents have been found in numerical simulations of 3Dturbulence . However turbulent stretching has notbeen studied in 2D turbulence and it is not clear howthe probability of stretching events depends on the tur-bulence parameters.Stretching can be characterised by the relative sepa-ration of a pair of initially adjacent particles at times t and ( t + τ ): R ( t + τ ) /R ( t ), (Fig. 5a). Thelogarithm of stretching normalised by time τ gives thevalue of the finite-time Lyapunov exponent Λ( t, x ) =(1 /τ ) ln[ R ( t + τ ) /R ( t )] which characterises the rateof local deformation of an infinitesimal fluid element .Here x is the initial position of the pair of particles.Trajectories of the fluid particles are determined us-ing simulated trajectories derived from experimentallymeasured velocity fields (see Methods). The finite-timeLyapunov exponent field Λ( τ, x ) gives the spatial distri-bution of the stretching rates for a particular integration Λ/<Λ> α =0.7 α =1.5 α =2 PDF ( Λ ) Λ (1/s) -3 -2 -1 -3 -2 -1 P D F ( Λ / < Λ > ) α < Λ > ( / s ) dc a Λ (1/s)040
40 80 x (mm) y ( mm ) R ( t ) R ( t + τ ) b FIG. 5. Turbulent stretching of fluid parcels. (a) Schematic of the fluid particle separation. (b) Finite-time Lyapunovexponents Λ on the turbulent surface computed over time τ = 0 . T L . (c) Probability density functions of Λ / (cid:104) Λ (cid:105) (where (cid:104) Λ (cid:105) isthe spatially averaged quantity) at three different vertical accelerations α =0.7, 1.5 and 2 and the frequency of f = 60 Hz. Theinset shows the PDFs of non-normalised Λ. (d) Spatially averaged value of the Lyapunov exponents versus the supercriticalityparameter α at different excitation frequencies: f = 30 Hz (blue diamonds), f = 45 Hz (red triangles), and f = 60 Hz (greencircles). time τ . Of a particular interest here, is the probabilityof strong stretching events occurring on the time scalecomparable to the Lagrangian velocity correlation time T L = 1 / (cid:104) ˜ u (cid:105) (cid:82) ∞ (cid:104) u ( t + t ) u ( t ) (cid:105) dt , where u is the fluidvelocity along the Lagrangian trajectory, (cid:104) ˜ u (cid:105) is its vari-ance, and the angular brackets denote the ensemble av-eraging. Such short-time Lyapunov exponents have notbeen studied in 2D turbulence before.Figure 5b shows a snapshot of the stretching field com-puted for τ = 0 . T L . The ridges of high values of Λ con-stantly move and reconnect (see Supplementary Video1), such that over time, all points on the liquid surfaceexperience many stretching events. The probability den-sity functions of Λ are illustrated in the inset of Fig.5c for three values of vertical accelerations which cor-respond to the supercriticality parameter α from 0.7 to2. However, if Λ is normalised by its spatially averagedvalue (cid:104) Λ (cid:105) , the PDFs measured at different energies col-lapse onto a single function, Fig. 5c. The PDF of theLyapunov exponent can be fitted with a Weibull distri-bution, f ( x ) = ( k/β )( x/β ) ( k − e − ( x/β ) k , where k = 1 .
35 is the shape parameter and β = 1 . k = 1) andthe Rayleigh distribution ( k = 2).The collapse of the PDFs of the average-normalisedfinite-time Lyapunov exponents to this universal formis due to the fact that the spatially averaged Lyapunovexponents (cid:104) Λ (cid:105) increase approximately linearly with theincrease in the supercriticality α of the wave-driven tur-bulence, Fig. 5d. This trend is independent of the turbu-lence forcing scale (vibration frequency), and correlateswith the insensitivity of the PDF of the cellulose porearea to the forcing scale (Fig. 3e). III. DISCUSSION AND CONCLUSIONS
Turbulence generated by parametrically excited wavesin vertically shaken containers (Faraday waves) affectsthe formation of bacterial cellulose by
Gluconacetobacter xylinus microorganisms. Changes in the millimetre-rangemorphology of the biofilm are found in the samples ex-posed to quasi-2D turbulence: above some threshold tur-bulence energy corresponding to the supercriticality pa-rameter of α = 1 . L f , deter-mined by the frequency of the Faraday waves. The higherthe frequency, the smaller the beads (Fig.2). It is hypoth-esised that the turbulence-induced torque manifested aschaotic rotation of fluid elements, leads to the bendingof the cellulose fibres. Such a rotation is studied byanalysing the rotation of the circular disks at the liquidsurface. It is found that the maximum in the rotationaldiffusion coefficient corresponds to the ratio of the diskdiameter over the forcing scale of d d /L f ≈ .
7. For largerdisks (or smaller L f ), D θ decreases exponentially. Thecharacteristic size of the cellulose beads d b normalised by L f is in the range d b /L f = 0 . − .
4, Fig. 2e. Turbulentrotation results from the interaction between meanderingcoherent bundles of fluid particles whose widths isabout L f , therefore the fibre clusters whose size is muchless than L f do not experience strong torque. Thus itseems reasonable that the formation of the beads occurson the scales of the order of L f .The microscopic structure of the cellulose is also af-fected by turbulence: cellulose exposed to turbulent flowsis rarified in comparison with the control samples, Fig.3. We attribute this effect to the stretching of fluid el-ements by turbulence. Stretching is evaluated by mea-suring turbulent velocity fluctuations and by computingtrajectories of the fluid particles. The separation of ini-tially close particles is characterised by the finite-timeLyapunov exponents Λ. Spatially averaged (cid:104) Λ (cid:105) monoton-ically increases with the increase in the supercriticality α (Fig. 5d), while the probability density functions showexponential tails at larger stretching events (Fig. 5c).The mean stretching rate also appears to be indepen-dent on the forcing scale of turbulence. This behaviourcorrelates with exponential distribution of pore size inbacterial cellulose (Fig. 3e), which is also insensitive tothe the turbulence forcing scale. Thus turbulent stretch-ing is found to be a plausible mechanism responsible forthe modification of the BC microstructure at small scales.Summarising, we investigate two fundamental pro-cesses in wave-driven turbulence, turbulent rotation andstretching. It is shown that turbulence can be used as atool to produce biofilms with remarkably different macro-and microscopic structures. The results offer new meth-ods which allow to engineer bacterial cellulose properties(e.g. porosity, water holding capacity, etc.) by employingwell-controlled wave-driven turbulence. IV. METHODS
Turbulent flow generation
Faraday waves are generated on the surface of a liq- uid which is vertically vibrated at the frequency f s with sinusoidal acceleration a above a parametric ex-citation threshold ( a th ) , at a supercriticality param-eter α = ( a − a th ) /a th ≥
1. The flow generated onthe surface at high α exactly reproduces the statisticsof 2D turbulence . In the experiments, turbu-lence kinetic energy is proportional to the supercriticalityparameter . The kinetic energy spectrum has a form E k = C(cid:15) / k − / in the inverse energy cascade range.The energy injection rate (equal to the dissipation rate) (cid:15) = ( E ( k ) /C ) / k / where C = 6 is the Kolmogorovconstant and it increases with the increase in α as (cid:15) ∼ α .The forcing scale of the flow is half the wavelength ofthe Faraday waves, L f = 1 / λ . The experiments areconducted at f s = (30 − α = (0 . − L f = (2 −
13) mm.Floating disks of various sizes d d (normalised size d d /L f ranging from 0.4 to 7) are used to study the ro-tational torque in the flow. A white stripe is marked oneach disk to track its orientation θ ( θ ∈ [ − ◦ , +90 ◦ ]) .The angular velocities V θ are calculated based on the an-gular displacements between two adjacent image framesand averaged over three points in time. The statistics ofthe disk rotation is captured along the disk trajectoriesin turbulence. To minimise the disturbance to the flow,only one disk on the surface is used in each measure-ment. More than 600 independent realisations (movies)are recorded in every experiment to achieve statisticalconvergence.The velocity field is measured using a particle imagevelocimetry technique. To determine the FTLE fieldfrom the measured velocity fields, we simulate the par-ticle trajectories by integrating the velocity field from t = t and determine how much initially adjacent par-ticles are separated during the finite integration time τ using a 4th order Runge-Kutta integration method. TheFTLE field gives the spatial distribution of the stretch-ing rate within a flow for the integration time τ . Themean FTLE is obtained by spatially averaging over thefluid surface characterizing the mean stretching rate ofthe fluid parcels. Bacterial culture
The culture media of
G. xylinux (ATCC 700178) isprepared by adding 20 g of glucose, 5 g of peptone, 5 g ofyeast extract, 2.7 g of
N a HP O and 1.5g of citric acidin 1.0L of distilled water.
72 hour culture . To prepare the 72 hour culture, one100 µl frozen stock (stored in − ◦ C ) is transferred intoa 1.5 mL Eppendorf tube and centrifuged at 12,000 rpmfor 5 minutes to remove Dimethyl sulfoxide (used for thefrozen stock). The pellet is collected and resuspended in5mL of fresh media, incubated for 72 hours at 26 ◦ C in astationary condition.To collect the bacteria from the 72 hour culture, wegrind the cellulose formed at the 72 hour culture and col-lect the solution. The culture solution is also centrifugedat 12,000 rpm for 5 minutes to form the pellet. The pelletis then resuspending in 1 mL fresh media. Cellulose formation in turbulent flows . Six-well mi-croplates containing 8 ml of culture media are inoculatedwith 20 µl of 72 hour culture solution of G. xylinux as de-scribed above. The microplates are vibrated for 72 hoursat 26 ◦ C . The control samples are incubated for 72 hoursin a non-vibrated incubator. The layer thickness of theculture media is kept at h = 8 mm in all the experiments.Harvested bacterial cellulose is immersed in ethanol.Subsequently it is transferred to 0.1 M NaOH and boiledfor 40 min at 90 ◦ C. It is then neutralised in deionisedwater for 24 hours.
Sample preparation for SEM.
The fibrous structure ofthe cellulose is fixed using chemical fixative glutaralde-hyde. The samples are washed in deionised water to re-move fixatives, then dehydrated using critical-point dry-ing. After that, samples are mounted on the studs andsputter-coated with a thin layer of evaporated gold forthe SEM analysis.The software imageJ (particle analysis) is used to anal-ysis the fibre structure, such as the pore area, pore di-ameter, and the pore size distribution.
ACKNOWLEDGMENTS
This work was supported by the Australian Re-search Council Discovery Projects and Linkage Projectsfunding schemes (DP160100863, DP190100406 andLP160100477). H.X. acknowledges support fromthe Australian Research Council’s Future Fellowship(FT140100067). The authors acknowledge the techni-cal assistance of Centre for Advanced Microscopy (ANU)and the Australian National Fabrication Facility (ACTnode).Author contributions:H.X and M.S designed the project. S.H.H and M.D conducted the cellulose formation experiments. J. Yconducted the rotational diffusion experiments. H.P. de-signed the experimental setup. J.Y. and H.X. analysedthe experimental data. H.X and M.S wrote the paper.All authors reviewed the paper. Hall-Stoodley, L, Costerton J.W. & Stoodley, P. Bacterialbiofilms: from the natural environment to infectious diseases.
Nat. Rev. Microbiol. , 95-108 (2004). Jonas, R. & Farah, L. F. Production and application of microbialcellulose.
Polym. Degrad. Stabil. , 101-106 (1998). Campano, C. et al. Enhancement of the fermentation processand properties of bacterial cellulose: a review.
Cellulose , 57-91 (2016). Nguyen, P.Q. et al. Engineered Living Materials: Prospects andChallenges for Using Biological Systems to Direct the Assemblyof Smart Materials.
Adv. Mater. , 1704847 (2018). Svensson, A., Nicklasson, F., Harrah,T., Panilaitis, B., Kaplan,D.L., Brittberg, M. & Gatenholm, P. Bacterial cellulose as apotential scaffold for tissue engineering of cartilage,
Biomaterials , 419 (2005). Kucinska-Lipka, J., Gubanska, I. & Janik, H.Z. Bacterial cel-lulose in the field of wound healing and regenerative medicineof skin: recent trends and future prospectives.
Polym. Bull. ,2399-2419 (2015). Yamanaka, S. et al. The structure and mechanical properties ofsheets prepared from bacterial cellulose.
J. Mater. Sci. , 3141(1989). Y. Hu, J.M. Catchmark and E.A. Vogler. Factors impacting theformation of sphere-like bacterial cellulose particles and their bio-compatibility for human osteoblast growth.
Biomacromolecules , , 3444 (2013). Gemeiner, P., Benes, M,J. & Stamberg, J. Bead cellulose andits use in biochemistry and biotechnology.
Chem. Papers . ,805-848 (1989). Petersen, N. & Gatenholm, P. Bacterial cellulose-based materi-als and medical devices: current state and perspectives.
Appl.Microbiol. Biotechnol. , 1277-86 (2011). Picheth, G.F. et al. Bacterial cellulose in biomedical applications:A review.
Inter. J. Bio. Macromolecules . , 97-106 (2017). Iguchi M., Yamanaka S.& Budhiono A. Bacterial cellulose: amasterpiece of nature’s arts.
J. Mater. Sci. , 261-270 (2000). Drozd, r. et al. Functionalized magnetic bacterial cellulose beadsas carrier for Lecitase ultra immobilization.
Appl. Biochem.Biotechnol. , 176-193(2019). Kannan, A. et al. Dynamic switching enables efficient bacterialcolonization in flow.
Proc. Natl. Acad. Sci. U.S.A. , 5438-5443 (2018). Hong, S-H. et al. Surface waves control bacterial attachment andformation of biofilms in thin layers.
Sci. Adv. , eaaz9386(2020). von Kameke, A., Huhn, F., Fern´andez-Garc´ıa, G., Mu˜nuzuri,A. P. & P´erez-Mu˜nuzuri, V. Double cascade turbulence andRichardson dispersion in a horizontal fluid flow induced by Fara-day waves. Phys. Rev. Lett. , 074502 (2011). Francois N., Xia H., Punzmann H. & Shats M. Inverse energycascade and emergence of large coherent vortices in turbulencedriven by Faraday waves,
Phys. Rev. Lett. , 194501 (2013). Francois N., Xia H., Punzmann H., Ramsden S. & Shats M.,Three-dimensional fluid motion in Faraday waves: Creation ofvorticity and generation of two-dimensional turbulence,
Phys.Rev. X , 021021 (2014). Paret, J. & Tabeling, P. Experimental observation of the two-dimensional inverse energy cascade.
Phys. Rev. Lett. , 4162(1997). Xia, H., Punzmann, H., Falkovich, G. & Shats, M. Turbulence-condensate interaction in two dimensions.
Phys. Rev. Lett. ,194504 (2008). Xia, H., Byrne, D., Falkovich, G. & Shats, M. Upscale energytransfer in thick turbulent fluid layers.
Nat. Phys. , 321 (2011). Bodenschatz, E., Malinowski, S.P., Shaw, R.A. & Stratmann, F.Can we understand clouds without turbulence?
Science ,970-971 (2010). Barkley, D. et al. The rise of fully turbulent flow.
Nature ,550-553 (2015). Sreenivasan, K.R. Turbulent mixing: a perspective.
PNAS , ,116 (2019). Falkovich, G. Fluid mechanics. Cambridge University Press.ISBN-10: 1107129567 (2018). Colombi, R., Schluter, M. & von Kameke, A. Measurementsof sub-surface velocity fields in quasi-2D Faraday flow. arXiv:1911.11208v3 (2020). Dunkel, J., Heidenreich, S., Drescher, K., Wensink, H.H., Bar,M., & Goldstein, R.E., Fluid dynamics of bacterial turbulence.
Phys. Rev. Lett. , 228102 (2013). Toschi, F. & Sega, M. (Ed.). Flowing matter. Springer, (2019). Zeff, B.W. et al. Measuring intense rotation and dissipation inturbulent flows.
Nature , , 146(2003). Berg, H. C.Random Walks in Biology. Princeton UniversityPress, Princeton, NJ. 1983. Xia H., Francois N., Punzmann H. & Shats M. Lagrangianscale of particle dispersion in turbulence,
Nat. Commun. , 2013(2013). Xia H., Francois N., Punzmann H.& Shats M. Taylor particledispersion during transition to fully developed two-dimensionalturbulence,
Phys. Rev. Lett. , 104501 (2014). Xia, H. et al. Tunable diffusion in wave-driven two-dimensionalturbulence.
J. Fluid Mech. , 811 (2019). Brown, A.J. On an acetic ferment which forms cellulose.
J. Chem.Soc. , 432(1886) Faraday, M. On a peculiar class of acoustical figures; and oncertain forms assumed by a group of particles upon vibratingelastic surfaces.
Philos. T. R. Soc. A Wang, J., Tavakoli, J. & Tang, Y. Bacterial cellulose production,properties and applications with different culture methods: Areview.
Carbohydr. Polym , 63-76 (2019). Yang, J., Davoodianidalik, M., Xia, H., Punzmann H., Shats, M.and Francois, N. Passive swimming in hydrodynamic turbulenceat a fluid surface.
Phys. Rev. Fluids , , 104608 (2019). Yang, J., Francois, N., Punzmann, H., Shats, M. & Xia, H. Dif-fusion of ellipsoids in laboratory two-dimensional turbulent flow.
Phys. Fluids , , 085116 (2019) Xia, H., Francois, N., Faber, B., Punzmann, H. & Shats, M. Localanisotropy of laboratory two-dimensional turbulence affects pairdispersion.
Phys. Fluids , 025111(2019). Zimmermann, R, Gasteuil, Y, Bourgoin, M, Volk, R, Pumir, A,Pinton, J-F. Rotational intermittency and turbulence induced lift experienced by large particles in a turbulent flow.
Phys. Rev.Lett. , 154501(2011). Francois, N., Xia, H., Punzmann, H. &Shats, M. Rectification ofchaotic fluid motion in two-dimensional turbulence.
Phys. Rev.Fluids , 124602 (2018). Kraichnan, R. Inertial ranges in two-dimensional turbulence.
Phys. Fluids , 1417 (1967). Aref, H. Stirring by chaotic advection.
J. Fluid Mech. , 1-21(1984) Ottino, J.M. The kinematics of mixing: stretching, chaos andtransport (Cambridge University Press, Cambridge, 1989). Scatamacchia, R., Biferale, L. & Toschi, F. Extreme events inthe dispersions of two neighbouring particles under the influenceof fluid turbulence.
Phys. Rev. Lett. , 144501 (2012). Goto, S.& Vassilicos, J.C. Particle pair diffusion and persistentstreamline topology in two-dimensional turbulence.
New J. Phys. , 65 (2004). Rossi, L., Vassilicos, J.C. & Hardalupas, Y. Electromagneticallycontrolled multi-scale flows,
J. Fluid Mech. , 207-242 (2006). Boffetta, G. et al. Detecting barriers to transport: a review ofdifferent techniques.
Physica D ,159