Bioprocess and Biosystems Engineering

, Volume 36, Issue 11, pp 1787–1796

Influence of flow rate variation on the development of Escherichia coli biofilms


  • J. M. R. Moreira
    • LEPAE, Department of Chemical Engineering, Faculty of EngineeringUniversity of Porto
  • J. S. Teodósio
    • LEPAE, Department of Chemical Engineering, Faculty of EngineeringUniversity of Porto
  • F. C. Silva
    • CEFT, Department of Chemical Engineering, Faculty of EngineeringUniversity of Porto
  • M. Simões
    • LEPAE, Department of Chemical Engineering, Faculty of EngineeringUniversity of Porto
  • L. F. Melo
    • LEPAE, Department of Chemical Engineering, Faculty of EngineeringUniversity of Porto
    • LEPAE, Department of Chemical Engineering, Faculty of EngineeringUniversity of Porto
Original Paper

DOI: 10.1007/s00449-013-0954-y

Cite this article as:
Moreira, J.M.R., Teodósio, J.S., Silva, F.C. et al. Bioprocess Biosyst Eng (2013) 36: 1787. doi:10.1007/s00449-013-0954-y


This work investigates the effect of flow rate variation on mass transfer and on the development of Escherichia coli biofilms on a flow cell reactor under turbulent flow conditions. Computational fluid dynamics (CFD) was used to assess the applicability of this reactor for the simulation of industrial and biomedical biofilms and the numerical results were validated by streak photography. Two flow rates of 374 and 242 L h−1 (corresponding to Reynolds numbers of 6,720 and 4,350) were tested and wall shear stresses between 0.183 and 0.511 Pa were predicted in the flow cell reactor. External mass transfer coefficients of 1.38 × 10−5 and 9.64 × 10−6 m s−1 were obtained for the higher and lower flow rates, respectively. Biofilm formation was favored at the lowest flow rate because shear stress effects were more important than mass transfer limitations. This flow cell reactor generates wall shear stresses that are similar to those found in some industrial and biomedical settings, thus it is likely that the results obtained on this work can be used in the development of biofilm control strategies in both scenarios.


BiofilmsEscherichia coliComputational fluid dynamicsMass transferShear stress

List of symbols


Parameter related to the mechanical strength of the biofilm (T)


Nutrient concentration in the biofilm (M L−3)


Bulk liquid nutrient concentration (M L−3)


Nutrient concentration at the biofilm surface (M L−3)


Hydraulic diameter (L)


Molecular diffusivity of growth-limiting nutrient in water (7.0 × 10−10 m2 s−1 at 30 °C for glucose) (L2 T−1)


Effective diffusivity of growth-limiting nutrient in the biofilm (L2 T−1)


Biofilm production flux (Mbiofilm L−2 T−1)


External (liquid) mass transfer coefficient (L T−1)


Average biofilm thickness (L)


Mass of biofilm per unit surface area (Mbiofilm Lbiofilm−2)


Nutrient flux to the biofilm surface (M L2 T−1)


Flow rate used in each experiment (L3 T−1)


Reynolds number (ρ v d μ−1)


Schmidt number (μ ρ−1D−1)


Sherwood number (KmdD−1)


Time (T)


Flow velocity (L T−1)


Total system volume (L3)


Biofilm specific production rate (T−1)


Viscosity (M L−1 T−1)


Density (M L−3)


Residence time in whole system (V Q−1) (T)


Biofilms are composed of microorganisms attached to a surface forming a community surrounded by extracellular polymeric substances that they synthesize [1]. Such biopolymeric matrix protects their microbial inhabitants from harmful environmental factors. Cells within a biofilm are 100–1,000 times less sensitive to biocidal substances [2] and are more resistant to stress factors such as substrate limitation, pH changes, hydrodynamic conditions, temperature and osmolarity [35]. Biofilms can be used in processes that are beneficial for humans like bioremediation [6], in wastewater treatment or even for production of various chemicals [7]. However, in biomedical settings, biofilms are often detrimental [8, 9] and should be minimized. In industry, food spoilage by bioconversion [10, 11] and biofouling on heat exchangers [12] are examples of the detrimental effects of biofilms with significant economic impacts.

Escherichia coli is used as a model organism for studies in the fields of biotechnology and microbiology. This microorganism can be found in gut flora and is able to survive outside the body making it a good water quality indicator [13]. In the food industry, E. coli O 157:H7 was found attached to beef-contact surfaces in beef processing, acting as a source of cross contamination [14]. Moreover, it was verified that contamination by several microorganisms other than E. coli was enhanced by this event [15].

Hydrodynamic conditions are determinant in the process of cell adhesion to a surface [1619]. In the food industry, the efficiency of cleaning-in-place procedures depends on the shear stresses generated by the cleaning fluid on the equipment [20]. The shear stress is influenced by the flow rate and so is molecular transport. This is of particular relevance since it has been shown that, in most cases, access to nutrients controls biofilm development [21].

This work investigates the effect of the flow rate variation on the development of E. coli biofilms formed under turbulent flow conditions. Two flow conditions were simulated using computational fluid dynamics (CFD) enabling the estimation of wall shear stress, while mass transfer coefficients were calculated for the flow cell reactor. The applicability of this flow cell reactor for simulation of biofilm formation in industrial and biomedical settings was also investigated.

Theoretical approach

External mass transport

Mass transport of a compound to the interior of the biofilm occurs by convective external mass transfer, followed by diffusive mass transport. Thus, in equilibrium, convection and diffusion rates must be equal at the biofilm surface, with z being the normal direction to the biofilm surface [22]:
$$N = K_{\text{m}} \left( {C_{\text{b}} - C_{\text{s}} } \right) = D_{\text{f}} \left. {\frac{{{\text{d}}C}}{{{\text{d}}z}}} \right|_{{z = L_{\text{f}} }} $$
The transport of molecules from the bulk solution to the liquid–solid interface by the fluid flow is the first step in nutrient delivery [23]. Nutrient transport by the fluid flow in this work was characterized by the external mass transfer coefficient in dimensionless form by a correlation between the Sherwood number (Sh) as a function of the Reynolds number (Re) valid in the range between Re = 2,100 and 35,000, and Schmidt number (Sc) in the range between Sc = 0.6 and 3,000 [24]:
$$Sh = 0.023Re^{0.83} Sc^{{{\raise0.7ex\hbox{$1$} \!\mathord{\left/ {\vphantom {1 3}}\right.\kern-0pt} \!\lower0.7ex\hbox{$3$}}}} $$
With the Sherwood number estimated from (2), the external mass transfer coefficient can be calculated from the definition:
$$K_{\text{m}} = \left( {Sh D} \right)/d $$

To reach a high rate of nutrient transport (1), a high external mass transfer coefficient (Km) is needed which occurs at high flow rates (or velocities).

Model for biofilm development

The apparent reaction rate in a biofilm system depends on the transport of nutrients and on the biological activity of the biofilm. The change of biofilm mass over time is the result of two phenomena: the production of biomass by the microorganisms in the biofilm and their detachment [25], which can be expressed as:
$$\frac{{{\text{d}}m_{\text{f}} }}{{{\text{d}}t}} = J_{\text{p}} - bm_{\text{f}} $$
Equation (4) can be integrated, starting from t = 0 (mf = 0), resulting in:
$$m_{\text{f}} = \frac{{J_{\text{p}} }}{b}\left[ {1 - {\text{e}}^{ - bt} } \right] $$

This is a typical response of a first order system, and graphically it represents a curve that increases asymptotically to mf = Jp/b for long time values.

Knowing the biofilm production rate, by analogy with the microorganism specific growth rate, it is possible to determine the biofilm specific production rate by:
$$\mu_{\text{b}} = \frac{{J_{\text{P}} }}{{m_{\text{f}} }} $$

Biofilm specific production rate (μb) is defined by the mass of biofilm produced by the active layer per unit time and per unit mass of total biofilm. It is important to notice that biofilm mass results not only from the production of new microbial cells but also from the extracellular biopolymers [25].

Materials and methods

CFD Simulation

The Fluent CFD code (version 6.3.26, Fluent Inc.) was used in the numerical simulation of the flow field in the semi-circular flow cell reactor, using only the standard models offered in the package. The computational mesh used in the simulations was created using Gambit 2.3.26 mesh generator (Fluent Inc.) and has a total number of 1250472 hexahedral cells. The mesh covers the entry tube, the semi-circular section and the exit tube, with dimensions depicted in Fig. 1. In the numerical simulations, the momentum and mass conservation equations were solved using a second-order implicit time-marching algorithm. The coupling between pressure and velocity fields is enforced using the PISO algorithm (Pressure-Implicit with Splitting of Operators) developed by Issa [26]. The advective terms of the momentum equations were discretized using the third-order accurate QUICK scheme [27]. The k-ω turbulence model was used to model the sub-grid velocity fluctuations characteristic of turbulent flows. This model is based on two transport equations, one for the specific dissipation rate and another for the turbulent kinetic energy. The shear-stress transport (SST) version of the k-ω model [28] was used, which effectively merges the k-ε model [29] and the standard k-ω model [30]. This turbulence model usually leads to accurate simulations both in free stream and wall bounded flow.
Fig. 1

Schematic representation of the biofilm producing system. 1 Recirculating tank, 2 peristaltic pump, 3 centrifugal pump, 4 rotameter. Four zones are defined for the flow cell reactor (dimensions are reported in m). A Exit one, B sampling zone, C entry zone, D feed zone

Flow visualization

The flow patterns for Re = 4,350 and 6,720 were visualized using long-exposure streak photography as described in [31]. Spherical PVC particles were dispersed in the working fluid at a weight concentration of 100 ppm and then illuminated by a laser light sheet provided by a 635 nm laser diode equipped with a cylindrical lens (Vector, model 5200-20). The flow pattern images representing the pathlines (Fig. 2) were captured using a digital camera (Canon EOS 30D) equipped with a macro lens (Canon EF100 mm), which was positioned perpendicularly to the laser light sheet.
Fig. 2

Time instantaneous streamlines on the flow cell reactor for the two tested Re (in the middle). Flow visualization using tracer particles illuminated with a laser diode for each tested condition (in the edges). The flow direction is from bottom to top, and the views illustrated are through the planar wall. Three zones are represented (entry zone, sampling zone and exit zone)

Bacteria and culture conditions

Culture conditions were similar to those previously described by Teodósio et al. [32]. Briefly, a stock of Escherichia coli JM109(DE3) (kept at −80 °C) was inoculated in a media with 5.5 g L−1 glucose, 2.5 g L−1 peptone, 1.25 g L−1 yeast extract in phosphate buffer (1.88 g L−1 KH2PO4 and 2.60 g L−1 Na2HPO4) at pH 7.0 [32]. This starter culture was grown in a volume of 0.2 L of inoculation media (incubated overnight at 30 °C with orbital agitation at 120 rpm).

An intermediate tank containing 4 L of the referred media was then inoculated with the starter culture (incubated at 30 °C with orbital agitation at 100 rpm and aerated using an air pump with a flow rate of 108 L h−1). When the optical density values (OD), reached 1 (610 nm), the culture was used to inoculate the recirculating tank (at a flow rate of 1.4 L h−1) already containing 5 L of sterile water, under aerated conditions. System feeding started 18 h after inoculation at a flow rate of 0.025 L h−1 with a culture media consisting of 0.55 g L−1 glucose, 0.25 g L−1 peptone, 0.125 g L−1 yeast extract and phosphate buffer (0.188 g L−1 KH2PO4 and 0.26 g L−1 Na2HPO4), pH 7.0 [32].

Flow cell system

The flow cell system used to produce the biofilm was previously described in detail by Teodósio et al. [32]. The system (Fig. 1) is composed of a recirculating tank, peristaltic and centrifuge pumps and one vertical semi-circular flow cell reactor with a diameter of 30 mm (hydraulic diameter, d = 18.3 mm) with 10 removable coupons on its flat wall. This reactor allows individual sampling of each coupon without disturbing the biofilm formed on the others as reported by Simões et al. [33]. A recirculating water bath was used to keep the temperature constant at 30 °C. This was chosen because this E. coli strain is capable of forming high biofilm amounts at this temperature [34] and because this is a common temperature in industrial settings particularly in the food sector.

Two conditions were tested, Re = 6,720, corresponding to a flow rate of 374 L h−1 and Re = 4,350 corresponding to a flow rate of at 242 L h−1. Three independent experiments were performed for each condition. Biofilm formation was monitored for 9 days and sampling was performed as described by Teodósio et al. [32].

Sampling and analysis

Flow cell reactor sampling was carried out daily with the coupons being removed from the top to the bottom of the flow cell reactor. Biofilm thickness and wet weight were analyzed. A digital micrometer (VS-30H, Mitsubishi Kasei Corporation) was used to measure the biofilm thickness and biofilm wet weight was determined as previously described [32]. Average standard deviation on the triplicate sets was below 25 % for the biofilm thickness and wet weight.

OD and glucose concentration were determined in the recirculating tank using the methods described by Teodósio et al. [32]. A spectrophotometer (T80 UV/VIS Spectrometer/PG Instrument, Ltd) was used to measure the OD at 610 nm. Glucose quantification was performed by dinitrosalicylic colorimetric method (DNS), as described by Miller [35]. The purity of the culture was assessed by plating culture samples retrieved from the tank in solid medium to check for the presence of contaminants. Glucose consumption in the whole system was obtained by mass balance at a feed flow rate of 0.025 L h−1. Average standard deviations on the triplicate sets were below 22 % for the OD and below 17 % for the glucose consumption.

Statistical analysis

Paired t test analyses were performed to estimate whether or not there was a significant difference between the results originated from three independent experiments for each hydrodynamic condition. Each time point was evaluated individually using the three independent results obtained in one condition and the three individual results obtained on the other condition. Average results for each condition are presented and when a confidence level greater than 95 % was obtained (P < 0.05), these time points were marked with an *.


Numerical simulation of the flow

To assess the hydrodynamic properties of the flow cell reactor, two flow conditions (Re = 4,350 and 6,720) were simulated by CFD. Four zones were defined corresponding to the feed, entry, sampling and the exit zones (Fig. 1). In Fig. 2, the numerically predicted instantaneous streamlines are presented and significant vortices can be observed in the entry zone for both tested flow conditions. This effect was more pronounced for the higher Re. The fluid flow shows a quasilinear trajectory in the sampling zone and the numerical simulation did not indicate a significant flow perturbation in the exit zone. The streamlines that were computed from numerical simulation were compared to the pathlines of tracer particles visualized by streak photography (Fig. 2). These images show random particle trajectories in the entry zone and a linear trajectory in the sampling zone for both flow conditions. Again there was no indication of significant flow perturbation in the exit zone due to the sudden contraction as observed for the numerically computed flow streamlines. The residence time in the system is less than 90 s, while the duration of the experiments is 9 days. From these different time-scales, we can assume that the flow in both cases is steady and that the system is essentially well-mixed [32, 36].

Wall shear stress profiles were computed for both flow conditions (Fig. 3). The highest wall shear stress values were found in the feed zone. The entry zone is the area inside the flow cell reactor where the highest shear stress variations are found. For Re = 6,720, wall shear stress reached higher values up to the point where the first coupon is located, whereas for the lower Re the higher values were reached up to half of the entry zone length. In general, higher values were achieved for the higher flow rate. A nearly constant profile of wall shear stresses for both Re was achieved in the sampling zone. The transverse variation of the shear stress in the flat wall of the sampling zone was similar for both Re (13 %).
Fig. 3

Wall shear stress contour plots in the flat wall of the flow cell reactor for the two tested flow conditions (Re = 4,350 and 6,720). On this representation, the flow cell reactor height is compressed by a factor of five

Table 1 presents maximum and average wall shear stresses and velocities inside the flow cell reactor for both flow conditions simulated by CFD. These results show that the average wall shear stress for the higher Re was approximately twofold higher than for the lower Re for each studied zone. The maximum wall shear stress for the higher Re was approximately 2.5-fold higher in the entry zone and twofold higher in the sampling and exit zones than for the lower flow rate. The highest wall shear stress values were obtained in the entry zone of the flow cell reactor where values of 3.23 Pa (Re = 4,350) and 8.23 Pa (Re = 6,720) were computed. Average shear stress values of about 0.183 and 0.365 Pa were obtained on the sampling zone for the lower and higher Re, respectively, although a maximum shear stress of 0.511 Pa can be reached with the highest flow rate tested. Regarding the flow velocity, the average and maximum values were 1.5-fold higher for the highest Re. The maximum velocity values were obtained in the entry zone and in both cases these values were also about 8.6-fold higher than the average velocity on that zone. A larger variation was found on the entry zone in the maximum wall shear stress values (variations below 14 % were computed). For all other zones, variations were on average below 1 % for maximum velocities and wall shear stress.
Table 1

Average and maximum velocity and shear stress values calculated in each zone (entry, sampling and exit) of the flow cell reactor


Re = 4,350

Re = 6,720

Entry zone

Sampling zone

Exit zone

Entry zone

Sampling zone

Exit zone

Average wall shear stress/Pa







Maximum wall shear stress/Pa







Average velocity/(m s−1)







Maximum velocity/(m s−1)







Biofilm formation experiments

In the present work, Escherichia coli JM109(DE3) was used as a model organism for biofilm production because this strain has demonstrated a high biofilm producing capacity in this medium [34]. Two flow conditions were operated in the flow cell reactor (Re = 4,350 and 6,720) to study the effect of hydrodynamic conditions in biofilm development. Nutrient transport by the fluid flow was characterized in this work through the external mass transfer coefficient Km (Table 2), which was calculated using Eq. (3). The residence time in the whole system was also calculated.
Table 2

Calculated values for the external mass transfer coefficient using Eq. (3) and residence time in the system




Km/(m s−1)





9.64 × 10−6





1.38 × 10−5


Figure 4 depicts the average results obtained for biofilm and planktonic cells of three independent experiments for each hydrodynamic condition. The overall model given by (5) was applied to these data from which biofilm specific production rate, illustrated in Fig. 4c, was determined using Eq. (6).
Fig. 4

Time-course evolution of assayed parameters: a biofilm thickness, b biofilm wet weight, c biofilm specific production rate, d optical density in the recirculating tank, e glucose consumption in the system. Closed symbols higher flow rate (Re = 6,720), open symbols lower flow rate (Re = 4,350). These results are an average of those obtained from three independent experiments for each condition. An average standard deviation below 22 % was obtained for the optical density and below 17 % for the glucose consumption. Regarding the biofilm thickness and biofilm wet weight standard deviations below 25 % were obtained. Statistical analysis corresponding to each time point is represented with * for a confidence level greater than 95 % (P < 0.05)

The results in Table 2 show that Sc was equal for the two flow conditions (since the fluid and temperature are unchanged) but Sh and Km were different for the two tested flow conditions. For Re = 6,720, the Sh and Km (1.38 × 10−5 m s−1) values were higher when compared with Re = 4,350 (Km = 9.64 × 10−6 m s−1). The residence time in the system is inversely proportional to the flow rate and was 35 % lower at the higher Re.

Biofilm thickness (Fig. 4a) showed a slight increasing tendency for the higher Re during the experimental time. For the lower Re, a distinct behavior was observed, since a marked increase in biofilm thickness (on average 75 % higher values than for Re = 6,720) was obtained between days 3 and 7. The maximum biofilm thickness was reached on day 6 for Re = 4,350, and on day 8 for Re = 6,720 (67 % lower than the maximum value obtained for the less turbulent regime). In both cases, after the maximum values were reached the biofilm thickness decreased until the end of the experiment. Biofilm wet weight (Fig. 4b) had similar profiles to the thickness until day 6 for the lower Re and during the whole experimental time for the higher Re. The maximum value obtained for Re = 4,350 was 57 % higher than that obtained for Re = 6,720. The decrease in biofilm thickness for the lower Re after the maximum value was achieved was more severe than the correspondent decrease in wet weight for this condition.

Biofilm specific production rate (Fig. 4c) decreased over time and had its maximum on day 2 for both flow conditions. For Re = 4,350, the value was 0.22 h−1 and for Re = 6,720 the value was 0.07 h−1 (P < 0.05). Although a marked decrease was observed in biofilm specific production rate at the lower Re on day 3, this value was 46 % higher than that obtained for the higher Re (P < 0.05). On the following days, the biofilm-specific production rate decreased reaching an average value of 0.04 h−1 (P > 0.05) around day 6 for both biofilms.

Planktonic cell concentration (Fig. 4d) had a similar behavior (P > 0.05) for both flow conditions until day 4. From this day onwards, the OD increased for the higher Re and this only started to occur 1 day later and at a slower rate for the lower Re. Both planktonic concentrations reached a constant value at day 8. Higher planktonic cell concentrations were generally obtained from day 5 for the higher Re with statistically significant differences obtained between days 5 and 7.

Glucose consumption increased thorough the experiment (Fig. 4e) and with the exception of day 2, consumption profiles for both hydrodynamic conditions were statistically similar (P > 0.05).


Numerical simulation of the flow

In this work, the Fluent CFD code was used to simulate two flow conditions to assess if the generated shear stress fields were similar to those found during biofilm formation in industrial and biomedical settings. The predicted streamlines were in accordance with the pathlines observed in the experiments for the tracer particles, which is an indication that the CFD simulation is accurately predicting the flow behavior in the flow cell reactor. The numerical results show that for both flow conditions a constant shear stress field is obtained in the sampling zone and that a small transversal variation (13 %) is obtained in both cases. This is an important aspect on the flow cell reactor design and operation [36]. It has been reported [37] that unlike circular tubes, rectangular and square types of flow cells do not provide uniform shear stress in some conditions. In our case, it is very important that uniform shear stress is attained to obtain similar biofilms in the different coupons [32, 38]. Additionally, it was also observed that a smaller distance from the inlet was necessary for flow stabilization with the lower flow rate. For the highest flow rate, this distance encompasses the whole entry zone and this can be a limitation for the use of this flow cell reactor at higher flow rates. For higher Re, the entry length necessary for flow stabilization must be determined for each particular flow condition and eventually the bottom coupons may not be used on those experiments if comparable biofilms are to be obtained on the different coupons. Alternatively, a longer flow cell reactor must be constructed (with a higher entry length) if all ten coupons are necessary. The results also indicate that unlike the sudden expansion at the inlet, the sudden contraction at the exit does not cause significant perturbations on the upstream flow. Since the last coupon is located at 8 cm from the exit, it can be used at both flow rates investigated in this work and possibly at higher Re.

The bulk velocities attained with this experimental setup are in the range of those used in the dairy industry, namely on the heat exchangers used for milk pasteurization where fluid velocities of 0.65 m s−1 are often attained at the inlet and between 0 and 0.24 m s−1 in each plate [39]. In the human body, maximum velocity values ranging from 0.4 to 1 m s−1 are also found in the aorta [40].

Shear stresses ranging from 0.183 to 0.511 Pa were obtained in the sampling region of the flow cell reactor. Since shear stress is a major determinant in cell adhesion to a surface [1619], this is an indication that this flow cell reactor may be suitable for simulation of biofilm formation in diverse conditions such as those found in membrane processes used for wastewater remediation, where average shear stresses of 0.5 Pa can be obtained [41], rennet-induced coagulation of milk during cheese making, where values below 0.5 Pa are common [42], or in atherosclerosis development, where time-averaged wall shear stresses of 0.5 Pa were determined [43, 44]. Although our CFD simulations considered a “clean” system (without biofilms), the results presented in this work are of particular relevance to predict the onset of biofilms in different systems. This is of critical importance when analyzing the formation of unwanted biofilms which is the typical case in industrial and biomedical settings.

Biofilm formation

Under a particular set of flow conditions, nutrient transport (i.e. mass transfer) and wall shear stress are amongst the most important parameters that influence biofilm development [45].

For the two Re investigated numerically, corresponding to a 1.5-fold increase in the flow rate, Eq. (3) predicts a 1.4-fold increase in the external mass transfer for the higher Re. Therefore, one would expect a higher biofilm formation in this flow condition if nutrient transport from the bulk solution was limiting biofilm growth at a lower Re. Quite on the contrary, we have observed higher biofilm formation throughout most of the experimental time when a lower Re was used. There are a number of factors affecting this outcome and it is likely that their relative importance varies along biofilm development. Higher flow rates enable a higher external mass transfer which means that more nutrients are brought to the surface of the biofilm. However, mass transfer within the biofilm may also be important particularly when the biofilm thickness starts to increase [46]. On the other hand, as demonstrated by our CFD results, lower flow rates mean lower wall shear stresses and these may facilitate cell adhesion. Lower shear stresses may also lead to the build-up of less dense (less compact) biofilm layers [25] in the long term. It is possible that on the first experimental days, a balance occurs between shear forces and external nutrient transport effects. Thus, similar amounts of biofilm may be formed until day 3 in both conditions because, although nutrient mass transfer to the biofilm surface is favored at a higher Re, a lower shear stress with the lower Re may facilitate cell adhesion. From day 3 onwards, the amount of biofilm formed at the lower flow rate is clearly higher. This may happen because the shear stress is affecting biofilm cohesion at the higher Re or internal mass transport limitations become prevalent [46]. The second hypothesis does not seem plausible as thicknesses exceeding 1 mm are obtained at the lower flow rate. Also, the higher concentration of suspended cells coming from the higher Re biofilm suggests a more intense biomass detachment due to the stronger applied shear stress. On the other hand, we have shown on a previous study with the same flow cell reactor [32] that for Re = 6,000 biofilms with maximum thicknesses around 0.5 mm could be obtained. Thus, it is likely that shear stress and not internal mass transport is limiting biofilm development at the higher flow rate. After day 7, the biofilm obtained with the lower flow rate becomes much more compact (the decrease in thickness is much higher than in the wet weight). We have shown before [32] that biofilms can adapt their architecture to balance access to nutrients and internal cohesion. More compact biofilms are more resistant to shear forces, but less compact biofilms enable the formation of liquid channels that facilitate nutrient transport within the biofilm [47, 48]. It seems that in this case the biofilm build-up under lower Re favors a more compact structure towards the end of the experiment which guarantees cohesion in detriment of nutrient availability [32].

Although a higher flow rate does not favor biofilm development, probably because of the trade-off between mass transfer and shear stress, it favors planktonic cell growth as these cells are probably more sensitive to nutrient transport from the liquid than to shear stress (since they are carried with the flow, the relative velocity between liquid and cells and, therefore, the shear stress at the cell surface is small or almost nil).

Another important phenomenon associated with the shear forces that are applied in a biofilm is the increased production of exopolysaccharides (EPS) [45]. A high EPS production that can be enhanced by high shear forces has been observed during the early stages of biofilm formation [49]. This trend can affect the biofilm specific production rate. On the first 2 days of the experiment, a distinct behavior was observed in the biofilm specific production rate for the two tested Re. Biofilm growth results from the production of new microbial cells and EPS, the proportion depends on the hydrodynamic conditions as previously reported [21, 25]. For the higher Re, the EPS production was probably higher than the formation of new microbial cells resulting in a smaller fraction of the active layer. Contrariwise, almost all the biomass formed at a lower Re was active in the first days. After the initial phase, since only the cells in biofilm are able to produce new adhered biomass, the fraction of active layer will tend to decrease over time and the total adhered biomass will increase until the equilibrium is reached. Moreover, biofilms formed at lower velocities can have a higher inactive layer due to the depletion of substrate inside them [21].


The work carried out with the flow cell reactor here described is useful for the simulation of biofilm development in industrial and biomedical settings taking into consideration the shear stress values and the velocity profiles that were obtained. The data indicate that the reduced biofilm formation under the higher Re results from a balance between mass transfer and shear stress effects, the latter affecting not only the initial cell adhesion, but also the fraction of active biomass in the biofilm and the internal diffusivity. An interesting point is the fact that the final biofilm specific (i.e., per unit mass of biofilm) production rate tends to be independent of the imposed shear stress. Therefore, although the biofilm formed at lower Re has higher mass and thickness than the one obtained at a higher Re, it seems that the active biomass plays a similar role in both cases, leading to similar substrate consumption rates. These findings have potential application in biofouling control in industrial settings as well as in the design of biomedical devices where biofilm development must be minimized and effective penetration of biocides or antibiotics is a crucial objective.


The authors acknowledge the financial support provided by Operational Programme for Competitiveness Factors—COMPETE, European Fund for Regional Development—FEDER and by Portuguese Foundation for Science and Technology—FCT through Projects MIT-Pt/BS-BB/0082/2008 and PTDC/EBB-BIO/104940/2008. Manuel Moreira Alves (CEFT, Faculty of Engineering, University of Porto) is acknowledged for the numerical simulations and revision of the manuscript.

Copyright information

© Springer-Verlag Berlin Heidelberg 2013