Design and Testing of a Prototype Large-Scale Bag Freeze–Thaw System

Published on: 
BioPharm International, BioPharm International-11-02-2007, Volume 2007 Supplement, Issue 6

A solution for the problems of a "bag in a can" system would be a fully jacketed and insulated container, similar to a traditional freeze tank.


It is common practice to freeze protein bulks for long-term storage, allowing product life to be extended up to several years. For high-volume products such as antibodies, the batch size can be hundreds of liters. Freezing such large quantities in a timely manner has generally been handled using "freeze–thaw vessels", which are jacketed metal tanks with a cooling surface provided by internal coil and fin assemblies. With the increasing cost of high-grade steel alloys, it is desirable to explore disposable technology alternatives to costly freeze–thaw vessels. The following is a report on the design and testing of a prototype large-scale bag freezing system developed at Genentech.

Interest in disposable technologies for long-term frozen bulk storage has been on the rise. This is partly because of the high capital cost of large freeze–thaw vessels, the scarcity of higher alloys as a material of construction, differences in pressure vessel fabrication standards, and requirements from country to country. Disposable containers are typically available in presterilized and custom fitted configuration, alleviating the need for clean-in-place and steam-in-place steps required for traditional freeze–thaw vessels.

Genentech Inc.

Polymeric containers with capacity of up to 20 L can be frozen in conventional freezers, though many days may be required before the entire volume reaches the desired temperature. The Celsius, a commercial system offered by Integrated Biosystem Inc. (a division of Sartorius Stedim Biotech, Goettigen, Germany), allows rapid freezing (several hours) of material in flexible containers at up to 16.5 L per container with the ability to process up to six containers at once. However, the system requires specialized equipment in addition to the freeze–thaw skid and is not operationally optimal for large batch sizes.

The lack of a commercial large-scale disposable container system that could meet our needs prompted us to develop the technology in-house. The following reports our efforts at designing and testing the first prototype of a large scale freeze–thaw system based on disposable bags that would complement our current frozen bulk storage strategy. We discuss our rationale and design methodology, and present experimental results from the first at-scale prototype.

Prototype Design

Design Constraints and Initial Concepts

Disposable bags used as liquid bulk storage containers are readily available in volumes of 1,000 L and above. The technology is proven and, because of its high structural flexibility, is relatively resilient against mild impact from blunt objects. Unfortunately, these soft-wall containers are much less forgiving when the contents are frozen. Light bumps may compromise the integrity of the bag, and the puncture may not be discovered until the contents are thawed. Such difficulties must be taken into account when designing a freezable bag system. Our aim was to develop a system that is compatible with facilities already geared toward handling commercially available freeze–thaw tanks. We had four main design constraints:

1 Each individual container must hold at least 100 L. There is potential for large bulk sizes, possibly in excess of 1,000 L per batch, and the use of many small containers would cause increased handling, tracking, and testing.

2 The system must be compatible with current commercially available freeze–thaw skids. The system also must be able to function under the same conditions with operational times, equivalent (within a few hours) to those of current commercial freeze–thaw vessels.

3 The system must be compatible with the bulk transportation and storage infrastructure used by commercial freeze–thaw vessels. The footprint and vertical dimensions must not be larger than those of existing 300-L freeze tanks in order to accommodate storage freezers, air shipping containers, and entrances to existing facilities.

4 The system must cost less than current commercial freeze–thaw tanks.

A large-volume freestanding or minimally enclosed bag would be difficult to handle and offer insufficient protection against puncture. A solution to this problem is to use the bag as a liner inside a non-disposable container. This "bag in a can" approach offers puncture protection as well as blockage against vapor transmission though the bag wall. The basic incarnation of this type of container would be a simple, thin-walled metal shell which would be placed in a freezer and rely solely on convected cold air for freezing the contents. Based on computer simulations, however, the freezing time for such a device would be unacceptably long. A better solution would be to fully jacket and insulate the container, similar to a traditional freeze tank, and use the current freeze–thaw skids to provide the require cooling.

The cost of the jacketed bag holder would be significantly higher than a simple enclosure. This configuration is necessary to achieve a sufficient heat transfer rate to freeze the bag contents in an acceptable amount of time. An additional benefit of this design is that the jacketing assembly provides thermal insulation protection during transport. Furthermore, because the product does not contact the metal surface, the container or shell can be fabricated from a low grade, less costly, metal alloy.


Table 1. Characteristics of bag shapes depicted in Figure 1.

Various container shapes and dimensions were considered during the initial design phase of the project. Figure 1 depicts several such containers. The rationale behind our selection or dismissal of each candidate is summarized in Table 1. We conducted transient simulations of the freezing process for water in the basic shapes shown in Figure 1 using fluent computational fluid dynamic software with the solidification model. Figure 2 shows the results of the simulations in the form of remaining liquid fraction as a function of freezing time for each of the basic shapes. Also plotted for performance comparison purpose is the Fluent simulation result for the commercially available Integrated Biosystem Inc., (IBI) 300 L CryoVessel. In that case, the freezing time obtained experimentally was within 15% of the simulation value. These simulations represent the best possible cases because we did not account for resistance caused by the bag material or the contact resistance stemming from microscopically imperfect contact between the walls of the container and the bag.

Figure 1.

The results from the simulations show that for these basic shapes, the overall freezing performance follows the cooled surface area to volume ratio, as expected. The IBI 300 L vessel has a wetted surface area to volume ratio of 0.157 cm-1 but this value includes contributions from the internal fins, which are not active cooling surfaces. Although composite shapes, as shown in Figure 1e, provide more cooling surfaces, the increased complexity in manufacturing both the holder and the bags negate the benefits of these configurations.

Figure 2.

Prototype Bag Holder Design

To satisfy the design requirements, we selected a rectangular shape (Figure 1d) with dimensions that would accommodate a 100-L fill volume (the actual volume is larger to account for expansion of the ice). In the proposed system, three bags can fit into a three-cavity jacketed holder that provides the same capacity and occupies the same footprint as the IBI 300 L CryoVessel. Figure 3 shows the dimensions that were specified for the manufacture of the first prototype holder and bags. The bottom angle of five degrees allows for drainage of the bag contents. Each vertical wall was purposely designed with a one-degree outward tilt from bottom to top to ensure that the bag contents freeze from the bottom up. The dimensions were chosen with consideration to manufacturability and performance. Compromises had to be made to satisfy all constraints. For example, narrower cavities would yield faster completion of freezing but would be more difficult to fabricate.

Figure 3.

At-scale Protptype Testing

Physical Description

The prototype bag holder (Figure 4) was fabricated in accordance with the dimensions presented in Figure 3. Each cavity is independently jacketed and connected in parallel to a common cooling loop inlet and outlet. A drain opening is present at the lowest point in each compartment to permit fluid removal during cleaning. The final production unit should have a cover for each cavity that would provide additional moisture blockage as well as anchor points for the hoses and accessories from each bag. Additionally, a single large insulated lid would cover the top of the bag holder for thermal protection.

Figure 4.

Matching bags with a nominal capacity of 117 L each were manufactured by Thermo Fisher Scientific (Logan, UT), in the 3-D configuration from HyQ CX3-9 film.

Experimental Setup

Preliminary performance testing was conducted using an IBI CU5000 freeze–thaw skid. The freezing set point was –50 °C, and the thawing set point was 25 °C for 12 hours, then stepped down and held at 5 °C until stopped.

The bags were inserted into the cavities and gently inflated with compressed air to expand all the folds in the plastic film before filling the bag with liquid. The first set of experiments was conducted with deionized water at the 100-L fill level. It was discovered that the jacket surrounding each cavity did not extend sufficiently close to the top, resulting in long freeze and thaw times for the top 2–3 cm layer of liquid. This problem can be resolved in a subsequent prototype holder by designing deeper cavities to account for the gap in the jacketing. For the current prototype, all subsequent experiments were conducted with a 90-L fill volume.

The prototype bag provided a port fitted with a flexible weighted internal hose intended as the dip tube for fluid introduction and removal. A similar arrangement without the weight was to serve as the thermowell (in this case, the hose end was welded shut) that can accept a probe for temperature monitoring during the freezing and thawing operation.

After the bags were filled with liquid, we discovered that the flexible thermowell tended to coil up and did not provide adequate placement accuracy, which could severely affect temperature monitoring. Furthermore, a flexible thermowell would not allow for the removal or re-insertion of the temperature probe after the bag contents have been frozen. To a lesser extent, a similar problem affected the dip tube. In that case, the dip tube failed to reach to the bottom of the bag, leaving a considerable amount of residual liquid after emptying of the contents. Consequently, we decided to use rigid polycarbonate (PC) tubes for the dip tube and thermowell. A custom split bar assembly was used as a clamping mechanism to hold the PC tubes vertically. The split bars were secured to the metal bag holder by adhesive-backed industrial Velcro, which allowed minor adjustments in the position and easy removal when necessary. Figure 5 shows the setup with the rigid dip tube and thermowell. Recirculation mixing can be accomplished by pumping the liquid from the bottom of the bag by the dip tube and returning it near the top at the opposite side of the cavity, approximately 10–12 cm above the liquid level. The outlet for the return port is aimed toward the nearest wall to minimize potential foaming.

Figure 5.

The tops of the bags were removed and the sides were secured to the cavity walls using heavy-duty adhesive tape to monitor the progress of freezing in each cavity. A video camera connected to a computer was used to record images at one-minute intervals as a time-lapse movie during the freeze and thaw phases of each experiment. This allowed us to visually assess the completion of the freeze or thaw process accurately.

Results and Discussion

We conducted a series of freeze–thaw experiments in the prototype bag holder filled with water as well as a 25 mM citrate buffer containing 150 mM sodium chloride (NaCl). Sodium chloride solutions form a eutectic at approximately –21 °C; as a result, the NaCl solution exhibited very different freezing dynamics compared with pure water. Table 2 summarizes the results for a select number of runs, which are discussed in a subsequent section. The completion time for freezing or thawing was determined from the time-lapsed video data. In some instances, it was not possible to clearly observe all three cavities simultaneously; in those cases, no time data is given. Freeze time is defined as the time required to freeze all of the liquid; thaw time is the time required to melt all of the ice as visually determined from the time lapse movies. The freeze–thaw skid permits the recording of only one external temperature probe. The location of the probe varies from experiment to experiment and is indicated in Table 2.

Table 2. Experiment summary

A flow rate of 1 L/minute was used in all cases involving recirculation mixing during thaw. Similar conditions are adequate for post thaw mixing in freeze–thaw vessels. Only the cavity containing the temperature probe was connected to the pump. It should be noted that the recirculation time of three hours was chosen arbitrarily; the optimal duration will be determined later but is expected to be between three and five hours.

Experiment D, in which no bags were used, established that the heat transfer performance of all three cavities is reasonably comparable. The freezing of 3 x 90 L of water was completed in eight to nine hours (for this specific geometry, the simulated freeze time estimated using Fluent software is about nine hours), although the right-hand cavity was ~10% slower than the other two. For comparison, an IBI 300 L CryoVessel can completely freeze 300 L of water in 10–11 hours on the same freeze–thaw skid. Although in this experiment the bag holder is not operated in the intended configuration, this can be viewed as being the worst case scenario in terms of heat load on the freeze–thaw skid because there is no additional thermal resistance caused by the bag material or contact resistance. The results suggest that the heat transfer fluid flow is distributed fairly uniformly among the three cavities, although the right-hand cavity may be slightly more restricted.

A new set of bags was installed immediately preceding experiment A1. All of the remaining experiments, listed chronologically in Table 2, used the same set of bags. The heavy-duty adhesive tape used to anchor the top of each bag to the holder walls also provided some protection against moisture caused by condensation seeping between the bag and the holder walls. A thin water film present between the bag and cavity wall can significantly enhance the heat transfer rate by minimizing contact resistance. This can be seen by comparing the freeze time data for the center cavity in duplicate experiments A1 and A2. A closer inspection of the center cavity confirmed that large wall areas had been wetted. A comparison of the freeze time between the left and center cavities in experiment A1 also reveals that the performance varies from cavity to cavity, with the left cavity performing significantly worse than the center cavity (more than 16 hours versus 13.5 hours). In this instance, the discrepancy can be attributed to the poorer fit of the left side bag (no bag wall contact in some areas) compared to the center bag. The bags used in this study were manufactured by a manual process; hence, the dimensional tolerances are rather wide at ± 25 mm.

In experiment set B, the PC thermowells and dip tubes were added to each cavity. This addition did not have an effect on freezing performance. We also investigated the magnitude of heat transfer enhancement from having a wetted bag–cavity wall interface by intentionally introducing water in this area. In experiment B2, we injected water between the bag and the cavity wall of the center cavity. In this case, there was no noticeable decrease in freezing time as compared with B1, probably because that interface was already wetted from the condensation generated in previous freeze–thaw cycles. Injecting water in the interfacial region between the bag and wall of the left cavity, however, resulted in significant improvement in freezing time, as seen in experiment B3. The magnitude of this improvement is consistent with those seen for the center cavity in experiment set A. The thin layer of water reduced the contact resistance between the bag and cavity wall.

The temperature traces for experiment set B are shown in Figure 6 and give a measure of reproducibility from experiment to experiment. Note that the freeze or thaw time for the right -hand cavity does not vary appreciably, indicating that there was minimal moisture ingress between the bag and cavity wall throughout the studies. Hence, the right cavity serves as an internal control for the various experiments.

Figure 6.

In experiment sets B and C, the cavity monitored by the temperature probe also underwent recirculation mixing during the last three hours of thaw. Interestingly, the thaw times show little variation across all experiment sets, ranging from approximately nine to 10.5 hours, irrespective of cavity location, contact resistance, or recirculation. In IBI 300 L vessels, 300 L of water can be thawed in approximately eight to nine hours using the same freeze–thaw skid. Unlike the freezing process, controlled by conductive heat transfer where the thermal resistance increases as ice is formed, the thaw process is primarily dominated by natural convective heat transfer. As thaw proceeds, the heat transfer rate between the container walls and the contents does not change drastically.

In experiment set C, the contents of the center cavity were replaced with 90 L of buffer. The freezing time was greater than that of pure water by two to three hours (compared to experiments B and C, center cavity). This reflects the lower thermal conductivity of the solute containing ice and the freezing point depression from the cryoconcentrate. Similar results are observed for standard freeze–thaw vessels. Experiment C2 also included thermowell sleeves, constructed from the same material as the bags. A configuration involving thermowell sleeves integral to the bag instead of the original flexible tubing would allow the use of a rigid thermowell made from any convenient material without introducing another product contacting surface. This is shown in Figure 5. There were no significant differences in the measured temperature resulting from the addition of thermowell sleeves (data not shown).

Figure 7 shows a sequence of images taken during freezing and thawing from experiment C1. It shows that during thaw, a piece of ice is kept at the bottom of the cavity by the thermowell but eventually fractures and floats to the top. The ice motion is unpredictable and ice in close proximity to the heated wall will melt faster, as expected. This is likely a major contributor to the variability in thaw times.

Figure 7.

In Table 3, we have summarized several important technical issues that were identified during the testing of the prototype large freeze bag holder and offer some possible solutions.

Table 3. Prototype 1 issues and possible solutions


We demonstrated that it is possible to conduct a freeze–thaw operation in large rectangular disposable bags housed inside a jacketed container. The overall concept and dimensions of the system are compatible with the infrastructure (freeze–thaw skids, transportation methods, and storage facilities) typically used for current commercially available large-scale freeze–thaw vessels (e.g., IBI 300 L CryoVessel). Having a compatible system provides internal flexibility and facilitates potential technology transfer to external sites and contract manufacturers.

The exploratory studies revealed that the freezing step is likely to require a slightly longer cycle and show more variability in terms of freezing time as compared with that of the traditional freeze tanks. The bag holder is expected to be significantly cheaper than a freeze–thaw tank of comparable capacity because of the simpler design and use of standard stainless steel as opposed to higher grade alloys. In addition to capital costs, there are advantages such as quicker turnaround time caused by the elimination of the clean-in-place and steam-in-place steps, the lack of requirements for pressure vessel certification, and lower shipping weight. There remain some process consistency issues to be addressed. In-depth studies to examine the long-term stability, shipping, and durability of the bags will need to be performed to guarantee robustness of this technology before it can be implemented in production.


The authors would like to thank C. Hsu, S. Martin-Moe, and B. Wolk for their support and leadership. We would also like to thank M. Goodwin and B. Buchanan of Thermo Fisher and the prototype vendor for their help with the design and fabrication part of this project.

Philippe Lam is senior engineer and Samir Sane is principal engineer, both of Process R&D at Genentech Inc.,South San Francisco, CA. 650.225.1000,