Normal Fermentation. Figure 1 shows a normal industrial batch fermentation as described above. Similar patterns were seen in other runs, with some differences of rate. The free glucose is seen to decrease smoothly from an initial level of about 60 g/1. This level reflects the balance between glucose production by the glucoamylase and its consumption by the yeast. Ammonia and FAN decrease rapidly; both are largely consumed by 8 h and exhausted by 12 h (Figure la). A portion of the FAN is not utilizable. The increase in viable cell count is essentially complete by 9h (Figure lb), the rate of glucose utilization slows at his point, and the general pattern is consistent with nitrogen limitation. The initial cell viability is low, typically 60­70%; this increases to the 88-93% range by 4 h and remains nearly constant within that range through 40 h (viability data are not plotted on the graph). Fermentation, as reflected by ethanol production, was 82% complete at 20 h and 94% complete at 30h. Essentially all of the starch was converted to glucose and fermented by 37 h. Glycerol production in the industrial fermentation appears to be largely growth-

Figure la. Substrate Levels, Control Run

log viable —♦— glycerol —0— ethanol/10 —■— C02, mM count

Figure lb. Product Levels, Control Run

и g/l. h —■ — g/vc. h

Figure lc. Glycerol Productivity, Control Run

associated; the productivity on a per-cell basis drops abruptly after 8 hours (Figure lc), at the same time the N sources are exhausted. The high starting glycerol level is due to the recycle of thin stillage and glycerol produced in the yeast propagator; there is a further increase in glycerol concentration in the beer well and still (not shown), possibly due to breakdown of cell components. The dissolved carbon dioxide concentration in a control run is shown in figure lb; it increased from 27 mM at the start to 35 mM at the time of peak fermentation, then settled back to about 29 mM. These concentrations correspond to partial pressures of 1.5, 1.7, and 1.3 atm (solubility calculated separately for the estimated medium composition at each point), confirming that supersaturation is significant in this system.

Laboratory Fermentations. We set up a series of laboratory fermentations to test the effects of CO2. The conditions of inoculum, pH (4.0), temperature (32.5 °С), ammonium and FAN and the initial glucose concentration were initially set to simulate the industrial process, and the glucose feed rate was set to simulate the rate of release of free glucose by glucoamylase in the industrial process. In one run, a low level of oxygen was added as air along with 1.5 atm CO2. The results of this series of experiments are summarized in Table II. Peak viable cell counts (5×10? to 10^ cells/ml) were lower than the levels seen in the plant but peak viability was typically 95-96%. The limiting factor or factors were not identified. Ammonia was not depleted and FAN was not reduced to the levels seen in the plant, so usable N was not limiting. Unlike the industrial fermentation, the lab runs showed decreasing viability after the peak cell count was reached.

Carbon Dioxide Effects. Carbon dioxide at 3.5 atm was somewhat inhibitory to cell growth (Table II). The peak cell count was decreased relative to the lower CO2 levels, and the rate of cell death was slightly increased compared with the runs at 0.5 and 2.5 atm (cell death in the 1.5 atm run was biphasic and cannot be readily compared with the others). Peak viable count was also somewhat reduced at 2.5 atm compared with the lower CO2 levels. An increase in cell size was noticed at

2.5 and 3.5 atm, but no unusual budding was observed. The fermentation rate decreased with increasing carbon dioxide concentrations (Table III); this effect was confined to roughly the last half of the fermentation. There were substantial differences between treatments at 30 and 45 h, but by 65 h ethanol had reached comparable levels in all the runs except possibly the run at 3.5. The ethanol production measured for the 3.5 atm run does not account for all the glucose apparently consumed; we have no ready explanation for this discrepancy but the fermentation rate is still depressed when measured as glucose consumption.

Percentage completion was calculated as the increase in alcohol concentration through the indicated time divided by the total increase in alcohol concentration when all glucose was consumed. *glucose feed was started late, at 20h, in the 2.5 atm experiment, contributing to the low completion at 30 h.

Glycerol and Ethanol Production Kinetics. Both glycerol and ethanol accumulated most rapidly while the cell number was still increasing; volumetric productivity declined after the peak cell number was reached (Figure 2). Although this would suggest that the production of both glycerol and ethanol was largely growth associated, a more detailed analysis disputes that conclusion. When the glycerol productivity is calculated on a per-viable-cell basis, most of the runs show a similar pattern: Initially high per-cell productivity, about 7xl0"^g/(cell. hour) declines within the first 6 hours to a plateau which persists through the remainder of the growth phase, stationary phase, and death phase (Figure 2a-c). Since the cell number is still increasing at 6 h, the higher glycerol productivity in the early periods is more specifically related to rapid growth than to growth per se. The bulk of the glycerol accumulation occurs later when the per-cell accumulation rate is constant, and cell growth has ceased; thus it is not growth-associated. The glycerol accumulation rate in this period is greater than that in the comparable period of the industrial fermentation. Overall glycerol yield was greatest at 0.5 atm CO2 and least at 3.5 atm CO2 among the non-aerated runs. An exception to this pattern was the air-supplemented run. Besides eliminating the initial spike air supplementation decreased the per-cell productivity during the plateau phase. However, the cell count in this run was enough higher to counteract this decrease so that the overall glycerol yield was little different than the control. The total oxygen uptake

time, h
g/l-hr	g/vohr
Figure 2a. Glycerol Productivity, 3.5 Atm C02. g/l-hr —■— g/vc-hr
Figure 2b. Glycerol Productivity, 1.5 Atm C02.

g/l. h, 3.5 atm —0— g/lh, 0.5 atm —■— g/vc. h, 3.5 —°— g/vc. h, 0.5 atm atm

Figure 2d. Ethanol Productivity

amounted to 5.8 mmol/1 based on the flow rate and composition of the gas streams entering and leaving the fermentor.

Ethanol productivity did not follow as simple a pattern and declined more slowly (Figure 2d), but also showed more rapid accumulation in the early part of the run and a plateau of decreased but continuing per-cell productivity as the run progressed (Figure 2d). The changes in per-cell productivity during the run were influenced by the CO2 level. Ethanol production was judged not to be strongly growth-associated.

In-plant Experiments. We attempted to manipulate the fermentation in the plant, first by carrying out a fermentation under partial vacuum. Due to the limited capacity of the Roots blower used as vacuum pump, we could only achieve appreciable vacuum after the rate of CO2 release had slowed. We ultimately reduced the headspace pressure to about 0.6 atm absolute, but had no impact on the level of dissolved carbon dioxide during the critical middle part of the run. The fermentation timecourse and the glycerol yield were similar to the control run (data not shown).

We also carried out an air-supplemented run in the plant. The fermentor was aerated at 20 cfm through sintered metal spargers, and the headspace gas was recirculated at about 40 cfm through separate perforated-pipe spargers. Figure 3c shows that there was a small impact of CO2 levels, as shown by the reversal of the CO2 trace when the sparge was shut off. However, the lowered CO2 concentration was still at the level seen in the control run and there was no improvement in the fermentation kinetics. As in the lab run, aeration led to a slightly higher cell count and altered the pattern of glycerol accumulation. However this time the early growth-associated glycerol production was not eliminated, though it was reduced on a per-cell basis (Figure 3c). Later in the run, second wave of glycerol production occurred which brought the overall glycerol yield up to a level comparable to the control run.