Recommended Best Practices in Freeze Dryer Equipment Performance Qualification: 2022

Best practices for performing freeze dryer equipment qualification are recommended, focusing on identifying methods to quantify shelf thermal uniformity (also known as “shelf surface uniformity”), equipment capability, and performance metrics of the freeze dryer essential to the pharmaceutical Quality by Design paradigm. Specific guidelines for performing shelf temperature mapping, freeze dryer equipment limit testing (the capability curve), and condenser performance metrics have been provided. Concerning shelf temperature mapping and equipment capability measurements, the importance of paying attention to the test setup and the use of appropriate testing tools are stressed. In all the guidelines provided, much attention has been paid to identifying the balance between obtaining useful process knowledge, logistical challenges associated with testing in the production environment vs that at laboratory scale, and the frequency of the testing necessary to obtain such useful information. Furthermore, merits and demerits of thermal conditions maintained on the cooled surfaces of the freeze dryer condenser have been discussed identifying the specific influence of the condenser surface temperature on the process conditions using experimental data to support the guidelines. Finally, guidelines for systematic leak rate testing criteria for a freeze dryer are presented. These specific procedural recommendations are based on calculations, measurements, and experience to provide useful process and equipment knowledge.


Introduction
A few years ago, a big pharmaceutical company entered into a joint development agreement for a large molecule therapeutic agent with a small biotech company. The formulation was to be freeze-dried, and formulation development, as well as early process development, was carried out by development scientists at the biotech company. At the time, the development scientists had no clear idea of where scaleup and, ultimately, production scale operations would take place. It was known that the formulation itself was very robust; that is, it could be freeze-dried under very aggressive conditions without compromising any quality attribute of the product, such as cake appearance, reconstitution time, recovery of original activity, or residual moisture level. The development scientists at the biotech company tried to take advantage of the robustness of the formulation by recommending aggressive cycle conditions that resulted in a relatively short freeze-dry cycle. A contract manufacturer was chosen and a scale-up batch was prepared using the suggested cycle conditions. The scale-up batch did not go according to plan. Pressure in the freeze dryer chamber could not be controlled, but rather oscillated between alarm limits. When the pressure reached the upper alarm limit, the power to the heating elements in the heat transfer fluid was cut off. As a result, the sublimation rate decreased and the chamber pressure dropped to the lower limit. In response to the pressure drop, the heating elements turned on again, resulting in a pressure rise to the upper limit, and the cycle repeated. The failed scale-up run caused a significant delay in the product development timeline.
The underlying cause of the failed scale-up lot was a phenomenon known as choked flow [1,6,25,26,33], which we will not elaborate on here, but this event is a good example of the importance of understanding equipment capability. Any freeze dryer has performance attributes that limit the sublimation rate it will support-refrigeration capacity, condenser surface area, the upper temperature limit of the heat transfer fluid circulating through the shelves, system geometry such as vapor duct length and diameter, or the dynamics of vapor flow from the chamber to the condenser. It is important to know which factor, or factors, limits performance, as well as the maximum sublimation rate supported by a given freeze dryer. Understanding equipment capability or performance is a point of emphasis in the Quality by Design (QbD) regulatory paradigm [38], and has taken on increased visibility in light of the development of a graphical design space approach to optimization of the primary drying phase of a freeze-drying cycle [21]. One of the boundaries of the graphical design space is the equipment capability curve or the maximum sublimation rate supported by the equipment as a function of chamber pressure. This capability curve can vary widely between laboratory and production scale. Ideally, the optimized primary drying conditions should be based on the capability of the equipment intended to be used for the commercial manufacture of the drug product.
While the use of new process analytical technology has aided in freeze-drying process understanding, most processes are still run conservatively [20,23,24]. The conservative approach often taken is in part due to insufficient quantitative understanding of equipment capability. Unfortunately, the determination of the equipment capability curve is often not part of the traditional equipment qualification procedure. This limited equipment understanding has two negative consequences. First, drug products manufactured using an overly conservative freeze-dry cycle carry with them a "hidden cost" in the form of an excessively long process, and this cost stays with the product for its entire life cycle. Second, manufacturing management often has no idea of the true capacity of a freeze-drying plant because they do not know how long the freeze-dry cycles would be if the drug products were freeze-dried using more optimal conditions. Thus, thorough equipment qualification data should be the link between laboratory and manufacturing scale freezedrying operations, enabling seamless scale-up. Unfortunately, quite often, little time is set aside for performing thorough equipment characterization at the production scale. The performance qualification strategy should be designed to systematically identify characteristics of the equipment that not only test the limits of the equipment but also provide process relevant information using known, relevant product loads. Often, this exercise may be set aside for execution during equipment downtime which makes it even more expensive and, furthermore, limits the time needed for a systematic characterization.
The current work is aimed at summarizing the current best practices in performing pharmaceutical freeze-drying equipment performance qualification. The focus will be on disseminating a living guidance document that is updated periodically to incorporate the use of new technologies and challenges for understanding complex interactions between the freeze-dried product and the equipment used. The information is provided for use by development scientists, validation and operations engineers, technicians, and operators alike in the bio/pharmaceutical manufacturing industry, and the practices outlined in the documents are designed for voluntary use by anyone in the freeze-drying community. The recommendations presented here are based on experimental measurements supported by computational fluid dynamics and past experiences on (a) shelf temperature mapping requirements; (b) equipment capability testing; (c) condenser performance metrics; and (d) leak rate testing of the freeze dryer. Clean-in-place (CIP) systems, as well as sterilize-inplace (SIP) systems, are not in the scope of this document but may be covered in subsequent best practice documents.

Need for Standardization
Typical equipment testing strategy starts with design review/qualification and manufacturer internal testing. During this testing, non-conformance of required materials/finishes to surfaces, parts/assemblies, programmed sequence, and safety requirements is identified and corrected. This is followed by functional and documentation verification at the equipment manufacturer's factory, commonly referred to as factory acceptance testing (FAT). Subsequently, the verification shifts to the end-user site first with the installation qualification (IQ) for conformance to design specifications, where documentation, blueprints, and drawing compliance are verified. The operational (OQ) and performance qualification (PQ) testing conclude the equipment testing strategy. Here, the characteristics of the equipment are tested first for operational requirements at no load and then under known, relevant product loads. The end-toend process is time-consuming and can often take 4-5 years from supplier selection, procurement, and design review to installation with the end of performance qualification. It is estimated that 6-9 months could be saved in this end-toend timeline, for example, by eliminating the need for customized protocol development requiring multiple working sessions with documentation and subsequent review cycles. Industry-wide consensus on the qualification process, standardized through a focus on scientific merit, could prevent unnecessary costs and delays in the timeline.
Standardized testing will undeniably benefit the industry in eliminating performance testing that (a) provides little process relevant information such as maximum heating and cooling rates for a shelf under no load and (b) is not based on a scientific rationale, but carried out merely for comparisons with legacy equipment. It is not uncommon to follow a test protocol merely because it was once executed on older equipment and has since erroneously become a benchmark for comparisons. Furthermore, industry-wide standardized consensus can reduce time lost in the unnecessary debate over appropriate test protocols often encountered today which in itself is a lengthy process. Here, we focus on three key areas of freeze dryer performance qualifications, the shelf mapping, the condenser, and the equipment vapor transport capabilities. These three areas are selected because of their importance to provide process understanding and control strategies for validation of pharmaceutical lyophilization [36].

Shelf Temperature Mapping
The temperature-controlled shelves are the primary source of heating/cooling for product temperature control. A fundamental goal from every freeze-drying cycle is to ensure the uniform thermal history of the product in each dosage container (e.g., glass vial or syringe) from a lyophilization batch. In the foregoing, we will refer to an individual product dosage container as a vial for brevity. The dynamics of the freezing and sublimation/desorption process, combined with the boundary conditions of the coupled fluid/thermal heat transfer to a vial, make it extremely important to characterize and control uniformity across the entire batch which may contain tens of thousands of vials. The construction of the shelf, the lack of intimate contact between the vial and shelf surface, the location of the vial on the shelf, the shelf surface finish, inter-shelf spacing, and temperature of surroundings compared to the shelf surface all affect the batch homogeneity. Also, the configuration of the freeze dryer can affect batch homogeneity, such as an internal vs an external condenser and the location of the condenser (side, bottom, or rear of the chamber) [9]. Considerable work in the past has focused on understanding these effects. For example, Rambhatla and coworkers, quantified variation in heat and mass transfer due to the design characteristics of freeze dryers [29]. Data obtained from sublimation tests performed on laboratory scale, pilot, and production freeze dryers were used to evaluate various heat and mass transfer parameters. It was found that OQ data obtained from sublimation tests may be used to test the performance of freeze dryers under different conditions and, hence, provide the data needed to ensure the equivalence of freeze-drying cycles from one freeze dryer to another.
Quantifying shelf temperature uniformities, both during isothermal conditions and ramps, is crucial for understanding temperature variation across the batch. Non-uniform shelf surface temperatures can impact vial-to-vial variability from edge to center vials. Moreover, such tests can help identify and isolate anomalies in fabrication compared to the design specifications. Such performance testing should include the identification and location of hot and cold spots across the shelf stack. The interested reader can find a detailed discussion on the importance of understanding shelf temperature uniformity in Rambhatla et al. [29]; Barresi et al. [2]; and Fissore and Barresi [8].
Unfortunately, individuals sometimes are driven solely by the requirements specified in a user requirement specification (URS) and utilize that URS almost as a rule book for equipment qualification without understanding whether the requirements have any scientific rationale and, more importantly, hold any process relevance. The terminologies encountered in URS documents themselves take several forms as illustrated below: • Shelf probes shall be within X1 °C of each other across a single shelf • Shelf probes shall be within X2 °C of each other across the entire shelf stack • Shelf probes shall be within X3 °C of the average of the probes across a shelf/ all shelves Such differences, while they seem trivial, sometimes lead to unnecessary lengthy debates over what may have little process relevance. Some end users also combine shelf uniformity with shelf temperature control which should be avoided. One example of a process relevant test criteria would state "the shelf probes shall be within 2 °C of each other across the entire shelf stack at any instant of time." For an informative equipment qualification in terms of thermal performance, it is important to understand the major physical constraints within which the freeze dryer shelf operates: 1) The shelf acts as a heat exchanger where a heat transfer fluid, circulated internal to the shelf surface, exchanges heat with the metal surface on which the product is placed. This is typically done by maximizing the area for net heat exchange between the metal surface and the heat transfer fluid. Commonly used heat transfer fluids used in a freeze dryer include DOW Chemical Syltherm XLT, Bayer Baysilone KT3 and KT5, DOW Corning Xiameter PMX-200 5cst, etc. Figure 1 shows the velocity contours at mid-plane of a typical 5 cSt heat transfer fluid in its serpentine path from the shelf inlet to the outlet at a shelf inlet temperature of −40°C. The contours were generated using a coupled 3-dimensional fluid-thermal model with geometry created in GAMBIT and using an ANSYS solver. We see that there is a region of high velocity at the inlet and then relatively uniform velocity throughout the serpentine path across a shelf. 2) The heat transfer between the fluid and the shelf is governed by the conservation of energy as Q =ṁC p ΔT where Q is the net heat transfer rate in J/s, ṁ is the mass flow rate in kg/s, C p is the specific heat of the fluid in J/ (kg·K), and ΔT is the temperature difference between the inlet and outlet of the fluid in Kelvin. Thus, as the heat flux on the shelf increases (e.g., from the increased load in terms of the number of vials [26]), the inlet/ outlet temperature difference increases. The difference between the inlet and outlet temperatures also increases at lower shelf temperatures as illustrated later in the section. Figure 2 shows the shelf fluid setpoint, inlet, and outlet temperatures and illustrates the difference between inlet and outlet at different set temperatures at production scale as measured using a resistance temperature detector (RTD) placed in a thermal well at the inlet and outlet of the manifold to the shelf stack.
3) The heat flux to the shelf from the surroundings also varies from the edge of the shelf to the center. Since the walls of the chamber are typically warmer than the shelf during the freezing step, the walls increase the heat load along shelf edges. Thus, vials located in this region either during freezing or even during primary drying will tend to run warmer, commonly referred to as the "edge vial effect" [28]. Figure 3 shows the difference in the temperatures of the shelves. Varying the shelf fluid temperature from 30 to −40°C in 1 h, the wall temperature changes by < 10°C. Thus, the edges of the shelf would measure higher surface temperatures (red line) compared to the center locations (dark blue), where the view factor to the warmer walls is small. Thus, it is critical to quantify the variation in the heat transfer to the vials recognizing that such variations exist between the edge and center locations. 4) Freeze drying poses an interesting challenge in the selection of the heat transfer fluid. The range of temperatures within a freeze dryer can vary from above +121.5°C from steam sterilization cycles to −55°C during freezing. Often, a compromise is made in the heat transfer properties of the fluid and the viscosity-temperature coefficient. The viscosity of the heat transfer fluid increases as the temperature of the fluid is reduced and hence its flow rate drops for a given pump size at lower temperatures (a characteristic of the heat transfer fluid). Besides, at lower shelf temperatures, the net radiation from the surroundings is also higher because of the larger temperature difference between the walls and the door of a freeze dryer and the shelf. In Fig. 2, we see the temperature difference between the inlet and outlet temperatures for different shelf setpoint temperatures. The difference is less than 2°C at a −40°C shelf temperature setpoint, and it increases to about 2.5°C at a −50°C setpoint and can sometimes be even higher depending on the initial conditions of the shelves and the wall. Note: The ramp rate in steps 1, 5, 7, and 9 is 1°C/min, while that for step 3 is 0.3°C/min. Table I summarizes the max/min inlet/outlet temperatures in stages 1-10. Keeping these factors in mind, below are the recommended procedures for shelf temperature mapping as part of freeze-drying equipment qualification (summarized in Fig. 4).
It is recommended that the shelf mapping be done in two phases (Test 1 and Test 2): Test 1 is to ensure shelves are constructed properly and is performed under no product load and at reduced pressure.
Example Test 1 setup: • Use thermocouple feedthrough connections equipped with external thermocouple validators if the freeze dryer is not equipped with thermocouple ports.
One example is a Kaye validator, which allows up to 36 thermocouples with three sensor input modules. Typically, feedthrough connections allow connecting to a chamber port with up to 18 (or 36) thermocouples.   • It is crucial to understand the importance of choosing the correct temperature measurement tools in obtaining repeatable data. There are several techniques used in mounting thermocouples to the shelf surface as shown in Fig. 5. The easiest method for shelf temperature surface measurement is to dip the sensors in a thermal paste and to secure them to the shelf using aluminum tape. • Often, there is resistance to the use of thermal paste directly in contact with the shelf surface in a production environment. An alternate method would be to use a thermocouple holder made of copper/Brass or another high thermal conductivity material that contacts the shelf. The copper/Brass contacts allow for a uniform high thermal conductivity material in contact with the shelf and the thermocouple. The thermocouple itself is inserted inside the blocks after being dipped in silicone-based thermal paste (for example, OmegaTherm from Omega) for improved contact. Be aware that any such metal-metal interaction has the possibility to scratch surfaces if not handled with care. • In comparison, spring-loaded arrangements are often used whose height can be varied depending on the intershelf separation as shown in Fig. 5 (right) when one wants to avoid using thermal paste on the shelf. Logistically, they are a lot faster to mount and provide repeatable results by ensuring good contact under compression from the spring arrangement. The thermocouple itself is mounted in a small cavity in an aluminum or copper block at the base of the spring arrangement. • One disadvantage of the tall metal blocks is that they could receive additional radiation from the wall that could also influence measurement.
• Note of caution: It has been found that the use of thermocouples directly placed on the shelf under insulation and a fixed weight show similar results, as shown in Fig. 6, from a production freeze dryer. However, the copper block arrangements may provide more repeatable data, while the insulated thermocouples sometimes suffered from poor contact with the shelf even with the use of thermal paste. Note: The shelf was ramped in this test at ~5°C/min. • The authors also recommend using flat self-adhesive fast response thermocouples when possible, which have been found to provide excellent contact between shelf and thermocouple tip, especially when used with thermal paste. One disadvantage of this approach however is that they are harder to mount in production scale dryers in the center of the shelf. • Note 1: There have been numerous examples of damage caused to shelf stacks due to hardware left behind on the shelves from a previous experiment. It is recommended that a checklist procedure be adopted for accounting for all hardware used in such testing before and after each test. Similar caution should be exercised to avoid damage from glass fragments from broken vials. • Note 2: Measurements at atmospheric pressure and vacuum have their pros and cons: While at atmospheric pressure, the test mimics conditions representative of the freezing stage and highest heat flux between the shelf and the product, it also adds to variability in shelf measurements from the impact of non-repeatable convective heat transfer through the gas, in particular for large production freeze dryers. On the other hand, in a vacuum, there is the possibility for a net insulating effect due to the reduced heat conduction through the gas between the shelf surface and the thermocouple junction (compared to that at atmospheric pressure); some of this, however, can be overcome by using a thermal paste to ensure good contact. Although we have not made a systematic study of the influence of vacuum vs no vacuum on variability in shelf surface measurements, it seems that better uniformity is obtained under vacuum due to the above indicated reasons. • Note 3: While the freeze dryers typically operate in the range −55 to +121°C, the relevant range for conducting uniformity of shelf temperature mapping for most products lies in the range of −40 to 40°C. Note that the range suggested is an example. • As a minimum requirement, thermocouple positions should include corners and the center of some shelves, with more sampling points on at least three shelves (top, middle, and bottom of shelf rack), diagonal on the rest of the shelves in such a way that at least 3 thermocouples are used on each shelf covering the center and corner locations. In some cases, it may be necessary to run two feedthrough ports of 36 thermocouples each. It may be useful to consider mounting a few thermocouples on the under-side of the shelf to demonstrate comparable thermal profiles on the top and bottom surface of the shelf. Indeed, the thermocouple on the top of the shelf may sometimes be limiting given that the thermocouples may obstruct vial placement or even worse, knock vials over when handling large vial quantities in the production environment.
• Note 4: It is important to allow the chamber to reach ambient conditions if it is performed after a steam cycle and the authors recommend measuring wall and door temperature during the test.
Example Test 1 procedure (no load, and at reduced pressure): • We recommend performing this test at no-load conditions since the objective here is to test the construction of the shelf. • Place thermocouples in the best possible pattern as described above. • As one example of the proposed test, ramp from +20 to −40°C in 1 h and reduce pressure to at or below 20% of atmospheric pressure (200 mbar or 20,000 Pa). This has been shown to provide more repeatable results (unpublished data as described in note 2 earlier), typically performed during freeze dryer internal testing. One such sample data from a production dryer is shown in Fig. 7. • Data collection after 60 min or more from the time the inlet temperature setpoint was first reached (after evacuation) after post-ramp stabilization period. In the example shown in Fig. 7, we see that the post-ramp stabilization period is limited to about 30 min and all the probes reach a steady state in this duration. However, in the tuning of some freeze dryers, this stabilization may take longer, and hence, the recommendation to wait is 60 min prior to data collection. Note: The ramp rate is 1°C/min and the Fig. 7 Shelf uniformity test at no load on a production dryer including thermocouples from 4 corner positions and the center of the shelf across the shelf stack max and min of the probes after data collection begins are −39.8°C and −41.2°C respectively. • Use the same test to measure the control of the shelf fluid inlet temperature at −40°C, 0°C, and +40°C. • This is recommended as a one-time test, being logistically challenging in an aseptic facility and repeated with fewer sampling points after maintenance operations when the shelf circuit is exposed to the atmosphere. Such initial testing should be performed at the manufacturer's facility during factory acceptance testing.
Test 2 (under known load at 0.5 to 1 atm chamber pressure) is intended to provide process relevant data under a known product load.
• Here, we recommend repeating the above testing procedure under a known product load, preferably water in vials or, if not, a representative load (an equivalent surface area of total product (water) loaded into the freeze dryer representing a full load of vials in the freeze dryer) of ice in bottomless frames lined with plastic sheets and covering the entire shelf area at 0.5 to 1 atm chamber pressure. One should note that vials are preferred since they represent a process relevant load and thermal contact between the product and the shelf. The objective here is to test shelf temperature uniformity under typical full load freezing conditions. Example Test 2 setup: • Water in partially stoppered vials is preferred or bottomless frames lined with 2 mil (one-thousandth of an inch), medium-duty plastic sheets. • The frames are often made of plastic, PETG material to avoid scratching, or worse, damaging the shelf stack by accident during the testing. • Use vials or frames that cover the entire surface of the freeze dryer shelf. • If frames lined with plastic are used on production dryers with large shelf areas (for example, > 15 m 2 ), it may be better to use more than one frame per shelf for logistical reasons. Calculate the pre-determined volume to be used in each tray allowing 4 cm of water/ice on each shelf (or the maximum load supported by the condenser when combined with a freeze dryer equipment limit test as outlined later). Note: De-ionized water is recommended. • The water can be transferred using a peristaltic pump from the reservoir that has been measured gravimetrically for each shelf. Note: In small freeze dryers, such as laboratory dryers, the water can be pre-measured using graduated cylinders and transferred directly. • Thermocouples mounted on the shelf surface, in a fashion similar to Test 1, can be used to test uniformity under load. Because the placement of thermocouples in a loaded freeze dryer can be challenging, we suggest limiting the probe placement. For example, it may be challenging to mount a thermocouple in the middle of the shelf. However, it may be worth placing at least 1 thermocouple a few rows away from the shelf edge for comparison. Also, as described previously, thermocouples should be located in the 4 corners. With this test, heating and cooling rates under load can also be obtained. Note: When using surface thermocouples below plastic sheets on loaded shelves, it is important to insulate the thermocouple from the layer of water/ ice above it. Thermocouples could also be attached to the bottom of the shelves (covered with insulation to minimize radiation). • Thermocouples inside vials can be used to understand uniformity in product temperature at different ramp rates as described in the procedure below.
Example Test 2 procedure (under known load at 0.5 to 1 atm chamber pressure): • Repeat the process from Test 1 (summarized again below) with the following changes As an example of a proposed test, ramp from +20 to −40°C in 1 h and hold at atmospheric pressure. Data collection after 60 min or more from the time the set inlet temperature was first reached. Test uniformity on a gradual ramp up at 0°C, hold for at least 1 h, and again at +40°C with a similar hold (or at extremes to be used in the process). The same test can be used to understand uniformity in product and shelf surface temperature (when using vials) under different ramp rates. This test allows the user to understand spatial variation in the shelf surface and product temperature as a function of the ramp rate chosen. The recommended rates are 0.1°C/min, 0.2°C/min, 0.35°C/min, and 1°C/min at atmospheric pressure.
• While the freezing step is a critical stage in a freezedrying process, the focus of the testing here is to understand the influence of equipment design and characteristics on batch uniformity. Figure 8 shows the inlet-outlet shelf surface temperature difference under a load of 2 cm of water across the shelf area on a production dryer. We see that the temperature difference can be as high as 14°C on the shelf surface, while it is < 2°C under no load. The shelf acts as a heat exchanger; with the added load, there is a larger shelf temperature difference between the inlet and outlet. Note that while we see a 14°C in the inlet-outlet shelf temperature difference at full load, one can expect to see a temperature difference in the range of 5 to 15°C (for different load/ shelf sizes).
The following are the sources of uncertainty in shelf mapping.
1) Testing in vacuum vs testing at atmospheric pressure: While testing at atmospheric pressure provides the worst-case scenario in surface temperature variation, it also introduces repeatability errors, presumably from higher convective heat transfer at atmospheric pressure and in large production dryers (> 15 m 2 shelf area). Hence, here, we recommend performing tests at both vacuum (Test 1 under no load at or below 0.2 atm) and then at 0.5 to 1 atm chamber pressure (Test 2 with full load), where the worst-case conditions representative of maximum heat transfer during freezing under full load are also tested. 2) Measurement tools: It is extremely important to ensure good contact between a temperature probe and the shelf surface using thermal paste and insulation of the measuring probe. The pros and cons of different tools have been discussed earlier. 3) Ensure sufficient time is allowed for the chamber to cool to room temperature after a SIP cycle. 4) Allow sufficient equilibration time after the shelf setpoint temperature is reached; about 60 min is recommended.

Quantifying Dynamics of Vapor Flow with Minimum Controllable Chamber Pressure Testing
The objective of this test is to determine the minimum chamber pressure the freeze dryer can control as a function of the sublimation rate under full product load. The equipment limit can be influenced by several factors, among which two of the key factors are (a) the geometry of the freeze dryer hardware and (b) the capacity of the refrigeration system of the freeze dryer. The test is designed to combine the effects of these factors. The spacing between each shelf (inter-shelf separation) in a freeze dryer affects the local pressure above the sublimation front of the product in vials placed on each shelf [11,31]. Besides, the net heat transfer from the shelf to the vials/ product is affected by the emissivity of the shelves and the surroundings [27,29]. The mass transfer between chamber and condenser depends on the spacing between the shelves and the chamber walls, the spacing between the lower shelf and the duct, the length-to-diameter ratio of the duct, the duct location, and the presence of hardware such as cleanin-place/sterilize-in-place (CIP/SIP) systems [1], and finally the refrigeration system (discussed in more detail in the following section). It is evident that with these variables, two freeze dryers are unlikely to be similar. In fact, it is important to prove through systematic characterization of equipment capability, shelf temperature uniformity testing, application of first principle heat and mass calculations, etc. that they indeed show comparable performance characteristics. Yet, it is somehow not uncommon to find cycles transferred from one freeze dryer to another without sufficient equipment characterization.
The flow rate through the connecting duct develops by setting the chamber pressure and the resulting condenser pressure and/or coil temperature. Above a critical pressure ratio between the chamber and condenser, which primarily depends on the length-to-diameter ratio of the duct, local sonic velocity can be reached at the duct exit. The mass flow rate can no longer be increased without increasing chamber pressure and the flow in the duct is said to have choked [6,25,26,33]. Choked flow can lead to loss of chamber pressure control and an increase in heat transfer to the product. The choking point is often referred to as the equipment limit. The limit is freeze dryer dependent and it is recommended to perform this test before a cycle is transferred. Such characterization would be best performed at the manufacturing site or the least at the final installation site, before production. The following section contains the recommendations for performing the equipment limit testing. Note that there are differences between the minimum controllable chamber pressure test (described below) and the choked flow test [22], both of which provide the same equipment limit results [35]. Here, we recommend performing the minimum controllable chamber pressure curve test since it is the quicker of the two as described below.
The setup (for full product load): The setup is the same as Test 2 in shelf uniformity with the following changes: Thermocouples mounted similar to uniformity test (under the plastic sheets) can be used here to test uniformity under load, or for heating and cooling rates under load. Also, on 3 shelves, we recommend mounting 1 thermocouple each to measure the ice temperature at a known distance from the bottom of the tray Note: To ensure a low vacuum when using thermocouples, one should consider using the feedthrough with cured epoxy around the feedthrough junction. One example of feedthrough is from Conax Technologies (Buffalo NY). The authors recommend checking the lowest achievable pressure with the feedthrough connected before running the equipment limit test.
Procedure (summarized in Fig. 9): Fill trays on all shelves with about 4 cm of water (or the load corresponding to the maximum capacity on the condenser) and record the volume of water on all shelves (record any variation from the pre-determined volume). Note: 4 cm of water is recommended to allow combining this test with a condenser load test (defined as the maximum ice load a condenser can support without loss in pressure and temperature control) or maximum sublimation rate test corresponding to when the condenser temperature increases above −40°C. -40°C is typically considered the critical temperature above which the condenser may not efficiently trap the incoming water vapor and hence leading to the loss of pressure control. Freeze water by setting shelf to −40 to −50°C.
Evacuate the system to the lowest pressure possible by inputting a low vacuum setpoint such as 5 mTorr ( 0.0067 mbar or 0.67 Pa) (note that this pressure may not be achieved; it is merely to allow the chamber to evacuate to the lowest achievable pressure). Wait for stabilization, i.e., chamber pressure at lowest pressure (+/−2 mTorr or +/−0.27 Pa) for at least 30 min. Use about 20°C increments for shelf temperature from the freeze temperature to the max setpoint on the shelf. Shelf setpoints of −20°C, 0°C, 20°C, 40°C, and 50°C are recommended. Allow at least 60 min of stabilization time at each shelf temperature setpoint, or time necessary for the chamber Fig. 9 Equipment limit testing flowchart pressure to equilibrate for at least 15 min (when the pressure does not change more than +/−2 mTorr or 0.27 Pa). Monitor and collect condenser temperature, chamber and condenser pressure, and product thermocouple data (green and brown data points in Fig. 10).
The product load can also be created using automatic loading of vials filled with water or using trays with a bottom (not preferred as the bottom of some trays can warp and can cause variability) for equipment capability tests. The required graph contains the shelf temperature (magenta data points in Fig. 10) and the corresponding chamber pressure (blue data points in Fig. 10). For example, in Fig. 10, the minimum chamber pressure the freeze dryer can control is 160 μbar (16 Pa) corresponding to the sublimation rate at a 10°C shelf temperature on a fully loaded freeze dryer. Note: This test uses 10°C increments. It is possible that the sudden increase in ice temperature at t=32 h is likely from ice breaking and losing contact with the thermocouple measuring the ice temperature.

Option 1: Using gravimetric testing-one shelf temperature per test. Both vials and trays can be used in the experiment
Repeat the procedure from the minimum controllable chamber pressure testing; only this time, hold the shelf temperature at −10°C until about 25% of the mass of ice is lost. Note: It is recommended that the test be performed in such a way that the ice mass loss is less than or equal to 35% of the initial mass since at smaller residual ice content, the chances of losing contact between the vial/ tray surface and a planar ice surface due to the porous ice structure increase. After allowing the ice to melt, with the peristaltic pump earlier used to fill the trays, the water is pumped back from the trays to measure the remaining mass of water (take into account losses in the piping). Note: While it may be logistically easier in a production environment to allow the ice to melt, it may be quicker to directly weigh the ice rather than allowing it to melt. Repeat the above procedure for 2 more shelf temperatures (the curve should be linear until at least a sublimation rate of 1 kg/h/m 2 of product area is reached) to obtain the sublimation rate vs chamber pressure curve (from 3 data points). Calculate sublimation rate as weight loss per unit of time (h). Each test should not take more than 20 h.
These tests are time-consuming and can be avoided if the test shown in Fig. 10 could be used with mathematical relations discussed in the options below.

Option 2: Using tunable diode laser absorption spectroscopy (TDLAS)
The procedure from the minimum controllable chamber pressure curve can be used in conjunction with a TDLAS mounted in the connecting duct to obtain the real-time sublimation rate data [13]. The same can be used to obtain the sublimation rate vs chamber pressure curve. No extra calculations are required since TDLAS measures the sublimation rate directly. Note: The commercially available technique is limited to freeze dryers that have an external duct connecting the chamber and condenser.

Option 3: Using steady-state heat and mass transfer equations-multiple shelf temperatures sublimation test
Knowing the heat transfer coefficient of the plastic bag sheets, one can calculate the sublimation rate (kg/h) as [29]: where P is the chamber pressure in Torr and K v is the heat transfer coefficient of the plastic sheets in W.m −2 .K −1 .
Knowing the K v variation as a function of pressure and with the energy balance, one can calculate the sublimation rate as [27]: 10 Minimum controllable chamber pressure curve ΔH s is the heat of sublimation of ice, dm/dt is the sublimation rate, A p is the area of the product undergoing sublimation, K v is the heat transfer coefficient of the plastic bags, T s is the shelf surface temperature, and T ice is the ice temperature measured at the bottom of the ice slab.
The sublimation rate as a function of time can be integrated and compared to the actual weight loss measured gravimetrically.

Option 4: Using computational fluid dynamics
• Computational fluid dynamics has proven to be useful in the calculation of equipment limit and choked flow [10,19]; however, it is not in the scope of this paper to present a detailed discussion on the use of computational fluid dynamics in equipment limit testing. • Simple single species water vapor flow with an axisymmetric model for the freeze dryer or more complex 3-dimensional models can be used to provide the sublimation rate vs chamber pressure curve. Figure 11 shows a comparison of the 4 options for quantifying the sublimation rate vs chamber pressure curve; the sublimation rate for different chamber pressures agrees within 16% with gravimetric data.

Condenser Performance Metrics
The condenser performance together with the geometry of the freeze dryer/connecting duct determines the equipment capability. For example, a condenser not able to trap vapor ΔH s dm dt = K v A p T s − T ice efficiently will result in higher minimum controllable chamber pressure. Hence, it is extremely important to define the performance metrics that allow efficient vapor deposition on the cooled surfaces of the condenser. Here, we briefly provide the recommended condenser performance metrics and follow it with experimental evidence.
1. Ice buildup in the condenser tends to be more non-uniform at lower condenser surface temperatures. Specifically, at extremely low condenser temperatures, a larger fraction of the ice builds up close to the duct exit [17] which may adversely affect the overall vapor flow and condensing capacity. The non-uniformity of the ice buildup impacted the flow structure near the duct exit and the temperature variation across the coils. According to Ganguly and Alexeenko [9], the non-uniformity of the mass flux on the condenser coils impacts the resulting ice growth patterns. Indeed, the portion of the coils close to the duct exit receives the majority of the condensate as opposed to those far from the duct exit. Furthermore, at warmer coil surface temperatures, there is a chance for re-distribution of the ice preferentially towards cooler surfaces, and as explained further later, warmer coil surfaces have a lower sticking coefficient for ice allowing for wider re-distribution and, hence, more uniform ice growth patterns. 2. Condenser surface temperature of −60°C or lower is sufficient to condense water vapor even at maximum sublimation rates. 3. There is evidence of inefficient condenser area usage at condenser temperatures below −60°C as illustrated below.
The primary factors that limit vapor flow and, in turn, impact the minimum controllable chamber pressure are (a) the length-to-diameter ratio of the duct between the chamber and condenser, (b) the topology of the cooled surfaces and its relative position to the duct inlet, (c) the total area available for vapor condensation, and (d) the refrigeration capacity for ice accretion [9]. The layout and throughput required at the manufacturing facility together often dictate the arrangement of the freeze dryer. For example, while throughput at a manufacturing facility may require a 56 m 2 shelf area 2-story freeze dryer, another may operate at a lower capacity within a 5 m 2 shelf area in single floor installation. The presence of any significant obstruction/ hardware in the duct can reduce the maximum throughput [1]. However, unless one is operating at the equipment limit, the presence of such hardware typically has minimal impact on the critical process parameters. Furthermore, it was shown by Ganguly and Alexeenko [9] that the uniformity of vapor mass flux can be improved by eliminating the duct as in an integral condenser design. Traditionally, condensers were designed based on the premise that a colder condenser pumps more efficiently than a warmer condenser. Kobayashi for the first time showed using imaging that lower condenser temperature and presence of non-condensable gases increased nonuniformity of mass flux and hence ice accretion [17,18]. It was shown, for example, that at a temperature of −76°C, compared to that at −52°C, the ice accretion was preferentially towards the entrance to the condenser. However, until recently, there were no systematic measurements that had been performed for quantifying the effect of the condenser temperature on the ice growth uniformity and, more importantly, on the process conditions. Here, we present both computational data and measurements on the effect of condenser temperature on process conditions. Figure 12 shows the contours of normalized mass flux on a 15m 2 production freeze dryer. Case a represents a warm condenser where only 6% of the water vapor molecules impinging on the cooled condenser surface deposit, and case b represents a colder condenser surface where 30% of the water vapor molecules deposit on the surface. It is evident that on the colder condenser, the molecular mass flux is restricted to the region closest to the duct entrance into the condenser, while on a warmer condenser, the ice patterns re-distribute more uniformly. The following section discusses the results from measurements for varying condenser temperatures.
The surface temperature of a condenser as shown earlier affects the ice growth patterns. The deposition process of the water vapor molecules on the condenser surface is governed by the sticking coefficient. The sticking coefficient is the post-collisional probability that a water molecule deposits on the surface. The probability is found to be unity over a wide range of energies [3]. In Fig. 13, we demonstrate the impact of this on the vapor deposition uniformity and hence heat load on the coils of the condenser.
A test was performed by mounting resistance temperature detectors (RTDs) on the condenser surface and fully loading the freeze dryer with bottomless trays (similar to the equipment limit test described earlier). The water was first frozen and then sublimed at a shelf temperature of +50°C. After a quasi-steady state is reached, the condenser temperature setpoint is varied in steps (from −60 to −115°C). It was found that at a condenser temperature setpoint of −60°C, all the RTDs were within 2.5°C. The range of RTD readings is a useful measure of the vapor load (heat load) on each section of the condenser. However, when the setpoint is reduced to −90°C, the range increases to 40°C and even to 50°C when the setpoint is reduced further to −115°C. Also, when the setpoint is reduced, it can be seen that only the bottom coils get Fig. 12 Case a, normalized mass flux contours on a condenser with 6% sticking probability; case b, normalized mass flux contours on a condenser with 30% sticking probability showing more non-uniformity at colder condenser coil temperatures Thus, at lower temperatures, the effective area of the condenser reduces to the initial impact of the water vapor flux on the coil surfaces, making it more inefficient to operate at such low temperatures. Furthermore, in Fig. 14, we see that, even under aggressive drying equivalent to the equipment limit, there is no increase in chamber pressure until the condenser temperature is increased above −60°C (corresponding to the vapor pressure of about 8 mTorr (1.07 Pa) above the condensing surface). When the condenser temperature exceeds −60°C, as in at −40°C, the chamber pressure increases by 65 mTorr (8.67 Pa), but then when the condenser temperature is reduced, the chamber pressure drops back down to 55 mTorr (7.33 Pa), providing the repeatability. We conclude that the process conditions are not affected by the condenser temperature as long as it is maintained below −60°C even under extremely aggressive drying conditions at full load. As shown earlier, lower condenser temperatures can have a detrimental impact on the effective condensing area and ice growth uniformity. This is also illustrated in the comparison of the ice growth uniformity in Fig. 15 at −110°C (case a) and −60°C (case b). Case b shows a uniform ice buildup across all coils while case a shows about 2× larger ice thickness on sections of the coil closer to the duct entrance compared to those away from it.

Maximum Leak Rate of the Chamber/ Condenser
Although the vacuum integrity of freeze dryers is essential for GMP operations, there is currently no generally accepted scientific rationale for establishing acceptance criteria for such testing [7]. Sahni et al. [30] only recently presented a literature review and an extensive survey from fourteen pharmaceutical companies on the leak rate specifications commonly used in industry. Based on this information, recommendations were made for the lyophilizer leak rate test. The operational leak rate for routine manufacturing practice and additional guidance for large volume and older lyophilization equipment was provided. Often times, current acceptance criteria are based on equipment capability proposed by the manufacturer, or from data collected during the qualification of a new dryer. The targeted specification for the life of a dryer is often one cited by the Parenteral Society for a new dry and empty dryer, a leak rate of 2 × 10 −2 mbar-L/s or 2 Pa-L/s (Parenteral Society, 1995). However, no rationale was ever offered to justify this number.
A more scientific approach to establish leak rate acceptance criteria is to ask the question: What is the maximum volume of air that can leak into the freeze dryer without violating the grade A microbial specifications? Measurements and calculations, based on the ideal gas law, ∆pV = ∆n RT, can be performed to arrive at the maximum allowable pressure rise (leak rate) value that would maintain desired conditions. A method has been developed to measure a freeze dryer's void volume, consisting of the total volume of air present in chambers, condensers, and connecting ducts.  Fig. 15 Case a, ice growth patterns on the condenser coils at −110°C; case b, ice growth patterns on condenser coils at −60°C: view of the first coil pack exposed to the vapor flow from chamber to the condenser A worst-case bioburden value for air that could leak in is determined based on environmental monitoring data, this is multiplied by a factor of at least two for a safety margin, and then, the maximum allowable leak rate is calculated. The leak rate can then be converted to the maximum allowable pressure rise over the duration of a leak test [14]. The interested reader can find a sample calculation in the manuscript in the work published by Hardwick and coworkers.
The method is performed and evaluated under these assumptions: • Since positive pressure in the vial headspace protects the product during sublimation, the risk from a system leak is limited to the time between the end of primary drying and vial stoppering at cycle end (i.e., all of secondary drying). • The leak is worst-case, from a source that can admit microorganisms, located in the system part that is housed in the area with the lowest environmental classification. • The leak rate remains approximately constant over the time course of secondary drying.
The setup is summarized in Fig. 16: • An air IDEAL™ or similar impaction principle unit is used to measure the microbial load of air in the room with the lowest classification associated with the freeze dryer. Twice the standard deviation is added to the results, for a bioburden maximum with a 95% confidence level. The calculated limit is doubled for an extra margin of safety, and this is the value used as the maximum potential bioburden load. • A canister made from a closed section of stainless steel tubing is attached to a spare port on the chamber of the freeze dryer. The pressure inside the chamber is decreased and stabilized; then, a valve inside the canister is opened, allowing the air in the canister to enter the chamber. The change in internal chamber pressure is measured and used to calculate the void volume of the freeze dryer. • A barometer (placed locally or from the nearest airport) and thermometer are used to measure the pressure and temperature of the room air.  A-Measure pressure and temperature of the air, and use it to calculate number of moles of air in a canister of known volume.
B-By measuring the change in pressure of the system when the air in the canister is introduced, the number of moles of air that causes that pressure change can be used to calculate the void volume of the chamber/condenser. C-Calculate how much air at worst-case condition could enter the chamber/condenser and still keep it below the maximum allowable value for cfu/m 3 . Calculate how fast the airflow from the leak could be over the entire course of secondary drying to keep the total microbial load below the maximum allowable value for cfu/m 3 . The pressure rise result is then calculated from the flow rate and void volume of the dryer.

Future Considerations
The industry stands to benefit from establishing consensus standards and best practices for equipment performance qualification. The process starting from procurement to performance qualification today is time-consuming. While equipment manufacturers can aid in executing standardized equipment characterization protocols during FAT/SAT, the responsibility of aligning on these standards in testing strategy and requesting systematic characterization of the equipment we use is up to all of us in the industry. One good example of a standardized user requirement specification (URS) from an industry consortium, put to practice, is the BioPhorum: User Requirement Specification for Small, Flexible Filler Technology [4]. Our hope is that with the standardization of testing protocols, more equipment users are aware of the best practices and incorporate them into their respective business process.
Several topics not covered in this manuscript will require further investigation, and the intent of the authors here is to see many of these topics outlined briefly below discussed in the subsequent peer-reviewed guidance documents.
• Wireless data loggers and IR cameras (such as FLIR cameras) in shelf temperature mapping: Wireless data loggers and sensors have been proven to be useful tools for both product temperature monitoring and shelf temperature mapping [5,15,32]. Recent measurements by the authors demonstrated repeatable results using a simple setup. However, these advances will require additional testing under different conditions to review the pros and cons of wireless data loggers for shelf mapping exercises. Such tests will be useful in phasing out the use of traditional vacuum feedthrough-based setup. • CIP coverage requirements: Requirements in the freeze dryer coverage of clean-in-place systems are frequently debated with little process impact [37]. The community would benefit from establishing coverage criteria backed by a scientific rationale. • Good practice in preventative maintenance programs: Maintenance of equipment, systems, devices, and instrumentation needs to be addressed. Often, maintenance programs are based on conservative estimates of time-tofailure. A systematic discussion involving best practices in preventive maintenance programs will stand to reduce redundancies and lost time. • Silicone oil level detection and characterization: While there is growing interest in the use of mass spectrometry in the detection of trace concentrations of contaminants in the freeze dryer [12], there are currently no industry standards for critical levels of different contaminants in the freeze dryer. • Use of computational fluid dynamics: The community can greatly benefit from systematically and routinely using computational techniques to quantify and characterize equipment limits. They can help reduce the time and resource requirements needed for measurements at scale while augmenting with invaluable process relevant data [1,10,16,34].