Define the balance boundary first
A complete gas balance identifies every inlet and outlet: main sparging, overlay, supplementary oxygen, carbon dioxide, sample withdrawal, purge flows and the actual exhaust path. A stream added downstream of the reactor affects the analyser without directly supplying the culture. Sampling before or after a condenser can give different volume fractions even when the dry fraction carries the same molar oxygen flow.
The suitable boundary depends on the question. Liquid supply involves transfer through bubbles and the free surface. Overall uptake can be assessed around the reactor including its gas space. Analyser correction considers the sample line separately. One flow diagram with explicit sub-boundaries connects these tasks without merging their meanings.
State sign conventions. Here OUR is positive oxygen consumption and CER is positive carbon dioxide formation. In a simplified steady overall balance, consumption equals inlet minus outlet and formation equals outlet minus inlet. During gas switching, inventory changes alter these equalities. This matters particularly when interpreting short peaks.
The gas-balance chapter introduces the basic relationships. This chapter concentrates on cases where an apparently correct calculation is misleading because of humidity, mismatched clocks or omitted gas inventories.
Converting volumetric flow to molar flow
A litre of gas is not a fixed amount of substance without temperature and pressure. The ideal-gas relation is ṅ = pQ/(RT), with absolute pressure p and absolute temperature T. Q must refer to the same state. If a flow has already been expressed at reference conditions, inserting current process pressure again applies a second, incorrect correction.
The teaching case explicitly uses dry gas at 273.15 K and 101325 Pa. With R = 8.314462618 J/(mol·K), the molar volume is 22.41397 L/mol. This is a selected definition, not a universal interpretation of every “normal litre” label. Another instrument may use a different reference temperature or pressure. Its documented configuration is what matters.
For several mass-flow controllers, first convert each channel to a common molar basis. Total flow and mixture composition then follow from sums of component flows. For an ideal gas mixture, volume percentages equal mole fractions; practical analysis must also establish whether the display is normalised on a dry or wet basis.
Small concentration differences can produce large relative uncertainty in a difference-based rate. The difference between two large oxygen flows is sensitive to shared calibration error, regardless of the number of spreadsheet decimals. Shared calibration sources can also create correlated errors. An uncertainty calculation should therefore represent the actual input quantities and their relationships.
Treating humidity and condensation consistently
A wet mole fraction yi,w includes water in its denominator. The dry fraction is yi,d = yi,w/(1 − yH₂O,w), or conversely yi,w = yi,d(1 − yH₂O,w). Dry total molar flow is ṅd = ṅw(1 − yH₂O,w). Multiplying consistent quantities preserves component flow.
For an independent numerical check, suppose oxygen is 18% on a dry basis and water is 5% of the wet gas. Oxygen is then 17.1% on the wet basis. The 0.9 percentage-point difference is not additional culture oxygen uptake; it is a changed denominator. Combining a wet outlet fraction with a dry inlet fraction can create an apparent biological rate.
A condenser removes water according to its actual operating state; its presence does not establish a completely dry analyser stream. Temperature, pressure and condensate removal determine remaining humidity and possible transients. Unintended condensation in a line can also alter transport and measurement stability. The correction must represent the complete analytical chain.
Liquid oxygen equilibrium depends on the relevant partial pressure. The NIST Henry data for oxygen explicitly state the convention and temperature basis. They do not supply a universal culture-medium solubility. Salt content and composition must match the property model being used.
The inert-gas balance and its assumptions
If a tracer gas is neither produced nor consumed during the considered period and its inventory is steady, ṅoutyI,out = ṅinyI,in. Hence outlet flow is ṅout = ṅinyI,in/yI,out. Both sides require the same dry or wet basis. A very small denominator can make the calculation unstable.
Nitrogen may serve as an effectively inert balance component in suitable aerobic cultures, but it is not inert in every biological process. Nitrogen switching, fixation, denitrification or significant dissolved-inventory changes need a different treatment. A remainder calculated as 100% minus O₂ minus CO₂ also contains every unmeasured gas. It equals nitrogen only under a justified mixture assumption.
The balance can diagnose inconsistencies. If a directly measured outlet flow differs from the inert-gas estimate, separately examine reference states, leakage, side streams, humidity and calibration. The discrepancy does not prove one particular defect. Documenting a supplementary gas stream may improve the balance more than introducing a more complicated biological model.
Do not apply the steady equation indiscriminately at every point during switching. Inlet minus outlet then equals the rate of change of inert-gas inventory. The quantity assumed constant in the steady model may become central during a gas change.
Dry-gas worked example with OUR, CER and RQ
The synthetic case supplies 5 L of liquid with 2 reference-L/min of a dry mixture containing 21% O₂ and 79% N₂. Inlet CO₂ is set to zero for this example. Dry outlet gas contains 18% O₂, 4% CO₂ and 78% N₂. All fractions are molar; other gases, leaks and inventory changes are excluded. Nitrogen is assumed inert.
Inlet flow is 0.0892301 mol/min. The inert-gas balance gives 0.0903740 mol/min at the outlet. Different dry totals are possible because oxygen consumption and carbon dioxide formation need not change total molar flow equally. Simply setting ṅout = ṅin would introduce another approximation.
Overall oxygen uptake is ṅin·0.21 − ṅout·0.18. Dividing by 5 L and converting minutes to hours gives OUR = 29.6518 mmol/(L·h). Similarly, ṅout·0.04 gives CER = 43.3795 mmol/(L·h). RQ = CER/OUR is 1.46296. This ratio is a balance result for the selected example, not unique identification of a metabolic pathway.
| Quantity | Value | Basis |
|---|---|---|
| Inlet | 0.0892301 mol/min | Dry reference flow |
| Outlet | 0.0903740 mol/min | Inert-gas estimate |
| OUR | 29.6518 mmol/(L·h) | Constant 5 L liquid volume |
| CER | 43.3795 mmol/(L·h) | Zero inlet CO₂ in this example |
| RQ | 1.46296 | Common molar time basis |
Gas switching makes inventories important
For a boundary including gas and liquid, oxygen consumption is rO₂ = ṅinyO₂,in − ṅoutyO₂,out − dNO₂/dt. Here r is total consumption; divide by the appropriate liquid volume for volumetric OUR. N includes oxygen inside the chosen boundary. During enrichment, inlet minus outlet may initially represent storage rather than consumption.
For carbon dioxide the signs differ: rCO₂ = ṅoutyCO₂,out − ṅinyCO₂,in + dNCO₂/dt, provided formed component and inventory are defined consistently. Dissolved CO₂ is also connected to bicarbonate and carbonate. A pH change can release CO₂ from existing dissolved inventory. An off-gas jump is therefore not automatically a simultaneous jump in biological formation.
An ideally mixed gas space at constant volume and constant actual volumetric flow has inert-step time constant Vg/Qg. A 0.5 L space at 2 L/min gives 15 s under matching state conditions. This is an ideal flushing example, not a universal reactor dead time. Bubbles, surface transfer, pressure control and additional line volumes can introduce further dynamics.
The original OUR study by Pappenreiter et al. distinguishes liquid accumulation from transfer. The overall balances above explicitly extend that reasoning to the selected gas-system boundary. A real dynamic analysis requires the relevant inventories to be measured or represented by a tested model.
Transport time and signal alignment
A sample line may introduce approximate dead time, a residence-time distribution or both. Dividing its volume by actual sample flow initially gives a mean transport time under the relevant flow assumption. It does not necessarily indicate when the first or last part of a gas change reaches the analyser. Add measurement-cell dynamics separately.
With changing sample flow, the time displacement changes. An ideal displacement path is then described by integrated volume: the volume transported between entry and exit must equal the path volume. A fixed shift of, for example, 20 s may be correct before a flow change and incorrect afterwards. The sample-flow reference state must also match the geometric line inventory.
A practical alignment uses a documented test event with known arrival along the chain. One shared control command is not a reference for every output. Separately assess valve movement, pressure or flow response, concentration arrival and analyser output. Only then align biologically related signals to a common event time.
Correlation between oscillating traces can suggest a displacement but does not prove physical identity. Gas supply and biological consumption may influence one another. An algorithm can therefore find a plausible delay between different causes. The section on sensor deconvolution explains why even mathematically clean alignment requires a physical justification.
Distinguishing common misinterpretations
A higher off-gas CO₂ fraction can reflect increased formation, lower total flow, changed inlet composition or release of dissolved inventory. A lower oxygen fraction can mean increased uptake, reduced air supply or changed dilution. The first question is which component flows and inventories changed. Percentage traces alone cannot answer it.
A calculated negative OUR is initially a balance observation. It may arise from incorrect time alignment during switching, mismatched humidity bases or wrong flow. Actual oxygen formation is also possible in appropriate phototrophic processes and requires a different biological interpretation. The sign is therefore neither an automatic sensor-fault criterion nor proof of an unusual metabolic state.
Interpret RQ only when its OUR denominator is sufficiently reliable. Near zero, small absolute errors create large ratio changes. The ratio also depends on the carbon streams, dissolved inventories and time interval included. Associating it with substrates and products requires an additional carbon balance.
As a final exercise, change only the humidity basis in the teaching case, then only outlet flow, and finally timing. Record the apparent rate created by each change and the independent observation that would reveal the error. This turns a formula into a diagnostic method: the calculation states both a result and the information still needed for a defensible decision.
A gas volume is incomplete without its state
The teaching case specifies dry reference flow of 2 L/min at 273.15 K and 101,325 Pa. The ideal-gas law gives approximately 0.0892301 mol/min. Exactly the same dry amount at 310.15 K and the same absolute pressure occupies about 2.27091 L/min. Molar flow has not increased; its volume at the other state is larger.
At 310.15 K and 150,000 Pa, the corresponding ideal actual volume flow is approximately 1.53400 L/min. Substituting this operating flow directly as reference litres creates a systematic quantity error. Conversely, an instrument reporting reference conditions is not directly reporting actual pipe-volume flow.
For the same molar quantity, Q2/Q1 = (T2/T1)(p1/p2)(Z2/Z1). Temperatures and pressures are absolute. The ordering of Z factors follows directly from Q = ṅZRT/p. This provides a simple sign check: increasing temperature increases volume under otherwise fixed conditions, whereas increasing pressure decreases it.
The label normal litres alone is insufficient for a reproducible handover because technical conventions differ. Explicitly state temperature, absolute pressure, composition and wet or dry basis. A consistent calculation can first convert everything to molar flow, then use desired volume units for display.
The quoted values are independent ideal conversions of the teaching-case amount. They include neither instrument accuracy nor a claim about suitability of actual gas equipment. Where real-gas effects matter, obtain Z from a source appropriate to composition. Defined reference conditions remain necessary.
Converting wet and dry fractions consistently
For an additional purely synthetic example, assume water vapour is 0.05 mol/mol of wet exhaust. Convert dry fractions 0.18 O₂, 0.04 CO₂ and 0.78 N₂ to wet fractions by multiplying by 0.95. They become 0.171, 0.038 and 0.741; adding 0.05 water vapour restores a total of one.
Total wet molar flow is correspondingly larger: ṅwet = ṅdry/(1 − ywater). A dry flow of 0.0903740 mol/min gives approximately 0.0951306 mol/min wet. Multiplying wet component fraction by wet total flow gives exactly the same component flow as dry fraction multiplied by dry flow. This invariance is a useful spreadsheet check.
A common error combines dry analyser composition with wet flow. Here it overestimates component flow by a factor 1/0.95. Even if the relative error is similar for exhaust components, resulting difference rates can be affected very differently. OUR subtracts two larger oxygen flows and may consequently be particularly sensitive.
The stipulated water-vapour fraction is not a saturation value at a named temperature. It was selected solely to make the balance transparent. Actual exhaust requires suitable humidity measurement or a justified equilibrium model with temperature and pressure. Condensation along the path can change those conditions over time.
A condenser also need not remove only water. For CO₂, dissolution, liquid drainage and chemical equilibria may matter. A precise carbon balance therefore checks which quantities actually cross the chosen boundary. Dry-gas accounting is a defined partial balance, not automatically a complete balance of all generated carbon dioxide.
Difference-rate uncertainty and plausibility checks
In the teaching case, oxygen input is approximately 0.0187383 mol/min and output 0.0162673 mol/min. Consumption is their difference, around 0.0024710 mol/min. A small absolute error in either larger flow can therefore create a substantial relative error in consumption. Displaying three decimals for composition does not establish correspondingly precise OUR.
For independently uncertain flows A and B, variance of the difference is u²(A−B) = u²(A) + u²(B). Correlated errors add −2cov(A,B). Shared calibration or computing outlet flow from the same analyser fractions can create dependencies. Uncertainty propagation follows the actual measurement and calculation chain.
As a simple check, an abiotic stationary condition without oxygen reaction should approach zero consumption after complete exchange. Persistent discrepancy may indicate incorrect basis, leakage, time alignment or calibration. During a gas change, however, a temporarily nonzero balance residual can also result from accumulation.
Check that composition sums consistently include all relevant components. Do not automatically normalise a sum differing from one without justification. A missing contribution may represent water vapour, another gas or systematic analyser error. Normalisation alone hides that information.
Finally, elevated RQ does not uniquely diagnose one metabolic pathway. Substrate composition, gas basis, accumulation and additional carbon flows need consideration. Calculate the quotient after assessing both rates; when oxygen consumption is very small, the quotient may become numerically unstable and scientifically uninformative.
Headspace inventory can dominate apparent oxygen uptake
A steady-state balance assumes that no relevant amount of oxygen accumulates or disappears from storage inside the chosen boundary. For a boundary enclosing headspace and liquid, the full balance is instead explicit: uptake = oxygen inflow − oxygen outflow − rate of change of gaseous inventory − rate of change of dissolved inventory. Uptake is positive when biological activity consumes oxygen. Transfer between gas and liquid is internal to this boundary and cancels when the two phase balances are added.
An original scale calculation uses an idealised, explicitly dry 10 L headspace at a constant 300 K and 101,325 Pa. With R = 8.314462618 J/(mol K), it contains 0.40622 mol of gas. If its dry oxygen fraction changes from 0.21 to 0.18 over one minute, its oxygen inventory decreases by 0.012187 mol, provided the total gas amount remains constant under the stated conditions. The mean storage term is therefore −12.187 mmol/min. This is an inventory calculation using selected teaching values, not a measurement from a bioreactor.
For comparison, the earlier steady-state teaching case with 5 L of liquid gives an oxygen uptake of approximately 2.471 mmol/min. The magnitude of the new headspace storage term is substantially larger. These two cases do not describe the same experiment; comparing them simply demonstrates why a small change in gas composition during switching does not automatically provide a steady-state metabolic rate. Equating the relevant flows without checking accumulation could even lead to a mistaken interpretation of the sign of an inferred rate.
Liquid inventory can matter as well. For example, if dissolved oxygen in 5 L of liquid falls by 0.10 mmol/L within a minute, its additional inventory change is −0.50 mmol/min. The total balance subtracts this negative storage term. Uptake is consequently 0.50 mmol/min greater than the difference between external oxygen flows when all other storage terms vanish. A falling concentration here means that consumption is partly supplied by oxygen previously stored in the liquid.
Analysing an actual gas switch therefore requires headspace volume, pressure and temperature histories, its mixing behaviour and composition measurements with compatible time references. Additional numerical precision cannot compensate for a large but poorly described headspace. When those data are unavailable, a clearly stated steady-state analysis window may be more defensible than a finely resolved transient rate. Whether the selected window is genuinely stationary must still be assessed from the inventories involved and the behaviour of the measurement chain.
Closing gas fractions is a check rather than an additional measurement
In the dry-gas teaching case, O₂, CO₂ and N₂ are assumed to account for the whole mixture, so their fractions sum to one. A real system may contain other components in the remaining fraction. If N₂ is calculated solely as 1 − yO₂ − yCO₂, the result is initially a remainder. It can be used as nitrogen only when knowledge of composition and the required accuracy justify that assignment. Closing a mathematical sum does not identify the chemical species in the remainder.
A measured nitrogen fraction is also statistically different from a remainder calculated from two other measurements. In the latter case, dependencies arise among the reported quantities. Changing the measured oxygen fraction then affects both the directly calculated oxygen flow and the total exhaust flow derived through the remaining fraction. An uncertainty analysis must retain this shared origin. Three apparently different columns in a table do not necessarily represent three independent pieces of information.
A useful check varies derived quantities together from the original measurement channels. For each assumed change in one channel, the remainder, inert-gas balance and rate are recalculated completely. This preserves their mathematical relationships. Increasing O₂ while holding a nitrogen fraction calculated from that same O₂ value fixed would be inconsistent. Even this straightforward calculation can reveal which measurement channel dominates a comparison between two small difference rates.
Finally, a gas leak outside the selected boundary differs from an unaccounted flow crossing it. What matters is whether an additional external material stream enters the balance. An inlet quantity inferred only from valve setpoints may fail to capture such a stream. A known boundary, measured flows, plausible composition and compatible timing therefore matter more than an individually closed sum. The flow lists used for compartment models apply the same accounting principle to a different problem.
Supplementary teaching calculations: Python calculation and results as CSV. The data description records assumptions, units and execution.
Terms in this chapter
33 terms explained in the glossary
- Absolute pressure
- Amount of substance
- Bioreactor
- Calibration
- Carbon dioxide evolution rate
- Condensate
- Correlation
- CSV file
- Dead time
- Deconvolution
- Denitrification
- Downstream processing (DSP)
- Dry matter and dry basis
- Dry-gas basis
- Event time
- Experimental factor
- Gas reference state
- Headspace
- Inert-gas balance
- Molar flow rate
- Oxygen uptake rate
- Partial pressure
- pH measurement
- Residence-time distribution
- Respiratory quotient
- Response variable
- Single-use bioreactor
- Substrate
- System boundary
- Time constant
- Variance
- Volumetric flow rate
- Water vapour