From Shake Flask to Stirred Tank
Hydrodynamic Matching Across Scales

Cells that grow well in a shake flask can change shape or lose viability in the stirred tank, because the flow around them is different. CFD (computational fluid dynamics) shows what the cells experience in the flask, so the reactor can give them the same conditions.

Book a Demo Contact Our Team

01 · Challenge

Unknown Hydromechanical Stress in Early Process Development

Most processes start in shake flasks, and nobody measures or controls the hydromechanical stress (shear and turbulent forces from the moving liquid) that the cells feel there. Shear-sensitive cultures such as Streptomyces spp. or NK (natural killer) cells notice the difference. If the stirred tank stresses them more or less than the flask did, their morphology or viable cell density (living cells per volume) changes.

A simulation shows where in the liquid the stress peaks and how it changes over time, which one average value for the whole flask cannot tell you.

Rendered shake flask with the moving liquid cut open; the cross-section is colored by shear rate from 10 to 1000 per second
Shear rate on a cross-section through the moving liquid in a shake flask at 180 rpm. © SimVantage

02 · Approach

Flask Simulation, Validation and Stress Matching at Scale

The simulation computes the shape of the rotating liquid in the flask from the flow itself, without assuming it. We only use its predictions after checking it against two published data sets, the line where the liquid touches the flask wall (Azizan & Büchs 2017) and the power input at different viscosities (Büchs et al. 1999). The model reproduces both without tuning, so we can trust its local flow data when we carry it over to the reactor.

Validation

Plot of simulated liquid height around the flask wall for five wall distances against reference measurements, 40 mL, 250 rpm
Where the liquid meets the flask wall, simulated and measured, 40 mL, 250 rpm (Azizan & Büchs 2017).
Plot of simulated power input over time for 1 and 16 mPa s converging to the reference values of Büchs et al.
Simulated power input settling at the reference values, 40 mL, 180 rpm, 1 and 16 mPa·s (Büchs et al. 1999).

03 · Scale-Up

Four Steps from Flask to Reactor

The flask tells you how the cells should grow, and the flask simulation tells you what flow they grew in. In-silico scale-up (scale-up carried out in simulation) then searches for reactor operating points that recreate that flow, and lab runs in the stirred tank check the result. Few teams scale up this way yet, so the projects deliberately tested several scale-up criteria against each other.

We used this workflow with TU Wien and TU Graz to scale up NK-cell expansion.

In Collaboration with

TU Graz TU Wien
Shake flask. Shear rate in the moving liquid on the orbital shaker, slow motion.
Stirred tank. Shear rate in the target reactor, simulated with SimVantage Bolt. The two videos use different color scales.
  1. 01

    Lab

    Flask Experiments

    Cells grown in the shake flask set the target values for viable cell density and cytotoxicity.

    Schematic of a 250 mL shake flask on an orbital shaker
  2. 02

    CFD

    Flask Simulation

    Free-surface CFD shows where shear and energy dissipation peak in the flask and how they spread through the liquid.

    Simulated liquid surface in a shake flask, colored by the distance of the interface to the flask wall
  3. 03

    CFD

    In-Silico Scale-Up

    The reactor simulation runs at different operating points until one matches the flow the cells saw in the flask.

    Simulated stirred-tank reactor with two impellers and baffles, liquid colored by velocity magnitude
  4. 04

    Lab

    Confirmation Runs

    Cultivations in the stirred tank at that operating point check whether viable cell density and cytotoxicity match the flask.

    Schematic plot of viable cell density over process time for an optimal and a suboptimal operating point

04 · Publication

Reactor Simulation Based on a Characterized Shake Flask

With RWTH Aachen (AVT), we scaled three Streptomyces species from shake flask to stirred tank. The stress in this shake flask was already known, so it went straight into the reactor simulation and no flask simulation was needed. When the reactor ran at the same average energy dissipation rate (how fast turbulent energy turns into heat, a measure of hydromechanical stress), the morphology changed. Cells that grew as compact pellets in the flask grew as loose mycelium in the stirred tank. The maximum value was no help either, because in our CFD runs from 100 to 1000 rpm it jumped around too much. Its 95th percentile, ε95%, is far more stable, so we used it to pick the operating point. There, all three species kept a morphology close to the flask, and oxygen transfer followed the same curve as in the flask for the first 12 to 15 h.

The study appeared as Brauneck et al., AIChE Journal 2026, which proposes ε95% as a new scale-up parameter for filamentous cultivations.

In Collaboration with

RWTH Aachen University
Figure from the AIChE Journal paper: oxygen transfer rate curves and morphology images for three Streptomyces species in shake flask and stirred-tank reactor
Oxygen transfer and morphology across scales, three Streptomyces species (Brauneck et al., 2026).
ε95% New simulation-based scale-up parameter for filamentous cultivations
3 species Streptomyces species that kept their flask morphology and oxygen transfer in the stirred tank
12–15 h Matching oxygen transfer in flask and stirred tank at the start of the cultivation

Facing a similar question? Let's talk.

Tell us about your reactor and your bottleneck. We will show you what simulation can quantify before you change anything in steel.

Book a Demo Contact Our Team All Case Studies