Ammonia System Evacuation — How Low Do You Go?

Eric M. Smith, P.E., Vice President and Technical Director, International Institute Of Ammonia Refrigeration

Removal of non-condensables and excess water is essential when starting up new refrigeration systems or large retrofits. Non-condensables contribute to higher than necessary head pressure, and excessive water in a system can cause inefficiency and operational problems. In all refrigeration systems, non-condensables and water vapor are removed from systems by vacuum pumps prior to charging with refrigerant. Removing liquid water (dehydration), can be done by “pulling” a deep vacuum – enough to boil the standing water and remove the vapor through the vacuum pump. It is very critical for CO2 and synthetic refrigerant systems to remove as much water vapor as possible. This will help to prevent the formation of acids and artificially high head pressure. But moisture removal is not as critical in ammonia systems, because ammonia will absorb the water due to its high affinity to it. This means that oil in the system will not be prone to contamination. 

To be clear, non-condensables should be minimized in any system. So when evacuating ammonia systems, one might wonder how low of a vacuum is low enough, and how much water is too much. IIAR publications regarding evacuation and dehydration have been inconsistent through the years. The following analysis helped the IIAR Standards Committee settle on requirements and recommendations for evacuation and dehydration which were issued in an interpretation of IIAR 5-2019, Startup of Ammonia Refrigeration Systems. Briefly stated, the interpretation says a vacuum of 25”Hg is sufficient if the system is subsequently purged of non-condensables, and there is no reason to believe that excess liquid water is present in a system upon startup. This is a lesser vacuum than was required in IIAR 5-2019. The interpretation request and response can be found on the IIAR website. 

Water in a system will be addressed first. A small amount of water in an ammonia refrigeration system is not problematic. Ammonia refrigeration systems will tolerate it with little or no detriment to operation. This is but one great advantage of ammonia over other refrigerants, especially for larger scale systems where opening them for maintenance is common and evacuating them repeatedly would not be possible. Indeed, a trace amount of water in an ammonia system is known to mitigate stress corrosion cracking of carbon steel vessels and piping. It is recommended that ammonia refrigeration systems have at least 0.2% water content for this reason. But excessive water concentrations should also be avoided because it can cause problems with efficiency and sometimes operation. Water that has been absorbed into ammonia can raise the boiling point of the mixture, which could require the suction pressure be unnecessarily low to accommodate a given load, resulting in inefficiency. 

Additionally if the ammonia becomes too saturated, there could be problems with control valve operation, where “freezing” of the mixture could hinder valves’ internal functions. It is beyond the scope of this article to explore the properties of aqueous-ammonia (ammonium hydroxide), but IIAR 6 requires periodic testing of ammonia on some systems and recommends a maximum water content of 5%. IIAR 6 also has an informative appendix that provides further detail on the effects of water in a system, methods to test its concentration, and methods to remove it. 

Water can be present in a system for a number of reasons. It could be left over from equipment that was hydrotested, it could be introduced through leaking heat exchangers, it can accidentally be drawn into a system during maintenance procedures that use water to capture ammonia vapor. It can enter a system that is under construction and left open to the weather, and likely other reasons. In all circumstances, excessive water should be kept out of a system during construction, and any hydrotested equipment should be examined for standing water prior to installation. But water can also be present because water vapor in air has either entered a system during operation (especially on systems that operate in a vacuum) or was not removed before a system was charged and started. 

If ammonia is tolerant to some amount of water, it is worth considering how low a vacuum is necessary when removing water vapor and non-condensables from an ammonia system. A deeper vacuum will remove more water vapor (and more non-condensable gases). But a lesser vacuum is easier to achieve, and the consequences are not severe in most circumstances. Following are engineering calculations to quantify the effects of vacuum to remove water vapor (non-condensables will be addressed later). A theoretical “large” system is used in this example. It has an ammonia charge of 10,000 lbs and an internal volume of 819 ft3. 

The water vapor within a system will behave like an ideal gas, and thus the use of ideal gas laws and equations are applicable. It may be helpful to review some basic terminology used to describe pressure and vacuum pressure. 

Absolute Pressure (P abs) is the pressure relative to the zero pressure in the empty, air free space of the universe. It includes the pressure due to the mass of the atmosphere (P atm, atmospheric pressure) and is used in ideal gas calculations. 

Gage pressure (Pg ) is the difference between absolute pressure less atmospheric pressure. Pg = Pabs- Patm. Gage pressure is used most commonly when discussing refrigeration system pressures. 

Vacuum pressure is the pressure that exists below atmospheric pressure. A perfect vacuum would mean that the absolute pressure is zero. This is a theoretical condition that can only be approached in practice. As such, there is always some positive value for absolute pressure, whether gage pressure is positive or negative (vacuum condition). There are several common units used to describe vacuum gage pressure. The following table easily depicts these units. Inches of mercury column (“Hg) and microns are most commonly used when discussing refrigeration vacuum pressure.

Table from “”

Start with the system full of saturated air at 90 deg F, 90% relative humidity (RH) and at atmospheric pressure (14.7 psia) and determine the mass of water vapor in the air before vacuum:

First, determine the pressure of the water vapor:

And from steam tables we can find that:


So it can be seen that water vapor remaining in a system after a vacuum to 25”Hg is not a concern. But if there is reason to believe that standing water remains in the system after efforts have been made to remove it during construction and prior to evacuation, water can most often be boiled out by increasing the level of vacuum, thus lowering the boiling point of the water within the system. Of course ambient conditions must be above freezing, and indeed relatively warm for this to be accomplished, as will soon be explained. It is also worth noting that a system that is vacuumed too low, too quickly can cause any standing water to freeze, making dehydration by vacuum extremely difficult. An examination of steam tables helps to demonstrate these concepts. As mentioned earlier, it is recommended that a system have at least a 0.2% water content to help mitigate stress corrosion cracking. It is therefore not generally a good idea to use metallurgic grade ammonia (which is very “dry”) for a system’s initial charge, because, as has been demonstrated, an appropriately evacuated and carefully constructed system will not have much water in it before

Table from ASME Steam Tables, Compact Version.

Assume, for example, the system is put into a 25”Hg vacuum (gage pressure) or 125,000 microns. Referring to the first table of vacuum pressures and interpolating, we see that 25”Hg (gage) is
approximately 2.4225 psia. Reading the steam table, and again interpolating, it can be seen that the water will boil at 133℉. This is obviously not practical because ambient conditions will not be this high nearly anywhere on earth. Now assume ambient conditions are 50℉, like on a nice fall or spring day, and that the temperature of the piping, and thus the water inside it, will equalize. To “boil out the water”, a pressure of 0.17813 psia will be required. Referring to the vacuum table and interpolating, it is seen that a vacuum of 29.57”Hg (8,904 microns), will be required to turn the liquid water into vapor and be removed by the vacuum pump- a more difficult accomplishment than achieving 25”Hg (125,000 microns). This reinforces that it is best to take care that standing water is not in the system for any reason when it is constructed. CO2 and synthetic refrigerants must be very dry inside, so a vacuum of 500 microns is typically specified for such systems. This is a greater than 99.9% complete vacuum, and corresponds to 0.009665 psia, well below the saturation pressure of water at 32℉. This means that ambient temperatures must be sufficient to transfer enough heat to the water such that it will not freeze before the water vapor is removed. This also demonstrates how pulling a vacuum too low, too quickly, can freeze standing water and
make dehydration efforts futile. It is worth noting that sometimes systems have been externally heated to help drive out water while the system is being evacuated. But as stated, it is best to be
sure during construction that there is not standing water in the system. It can also be seen that pulling a vacuum to at least the saturation pressure of the water at the system’s (ambient) temperature can be used to indicate if there is water present. If the vacuum pressure “stalls” at some point, or rises after a predetermined level of vacuum has been reached, this indicates that
water is “boiling” off (presuming that there are no leaks in the system). And again, we see the advantage of using ammonia as a refrigerant, because as stated, some water in the system is
tolerable, and under normal circumstances a deep level of vacuum is not necessary. 

With the issues of water vapor and liquid water addressed, removal of non-condensables will next be examined. Non-condesable gas in a system will eventually make its way to the condenser,
“blocking” the flow of discharge gas into it. Stated another way, non-condesnables will consume the volume of the condenser designated for condensing the ammonia. This dramatically raises the head pressure, as will be demonstrated. Purging, either manually or automatically, will dispatch noncondensables, but it might be impossible to even start or continue to run a system that has
excessive non-condensable gas, which drives the need for evacuation before startup. Also, noncondensable gas will exist mostly in the form of air, which is mostly composed of nitrogen and
oxygen. Oxygen in a system can contribute to stress corrosion cracking, so it is important that its presence is minimized, head pressure issues notwithstanding.

To demonstrate the effect of vacuum and presence of non-condensables, an actual system with an ammonia charge of 5000 lbm and a volume of 534 ft3 is used. As with the examination of water
vapor, we start with the system full of air at 90℉, 90% relative humidity, and at 14.7 psia (atmospheric pressure).

Start by considering the system full of air at the conditions stated above.

The humidity ratio of water vapor to dry air is the same as in our first example, because the starting conditions are the same: 𝜔𝜔 = 0.028

Consider that the system is pulled into a 25”Hg (gage) vacuum, or about 125,000 microns.

Referring to the vacuum table and interpolating, this pressure equates to 2.4237 psia or 349.01

From the first example, we know the specific gas constant for the mixture of water vapor and air at this humidity ratio is 54.187

The mixture (dry air plus water vapor) temperature is again assumed to be ambient temperature and is: 90℉ + 460 = 550°R

The mass of the mixture at this vacuum pressure is determined by the ideal gas equation:

The mass of air remaining after vacuum can now be determined:

Knowing the mass of air in the system after vacuum, the pressure it imposes on a charged system can be determined using the ideal gas equation. The condenser volume on this system is 70 ft3, and it is assumed that the condenser is normally filled 1/3 full of liquid ammonia, leaving 2/3 of the volume holding ammonia vapor and non-condensables. The design condensing temperature is
95℉. The air in the system (non-condensables) will migrate to the condenser. For this calculation, it is assumed that the air temperature will be that of the ammonia.

It can be surmised, even if readers disagree somewhat with the assumptions of volume and/or temperature of the ammonia/air mixture, that air, the most common non-condensable will have a
significant impact on the system’s head pressure. This relates directly to energy consumption because compressors will have to work against this un-necessary addition of pressure. It is
therefore necessary to either purge the non-condensables, or if a method of purging (either automatically or manually) is not available, to evacuate the system to a much greater level of vacuum. This analysis also demonstrates the importance of an automatic purger for any system operating below atmospheric pressure, where a bad seal could introduce air into the system and cause un-necessarily high head pressure. When automatic purgers are not installed, owners or owner’s representatives might wish to witness purging, or otherwise require some type of documentation that purging has been accomplished. Because ambient conditions, loads, and condenser conditions can easily change, documentation of purging can be difficult. Likely the best way is to record the system head pressure, compressor loading, and ambient conditions just before and just after a manual purge. It is also worth noting that an additional manual purge might be needed after the system has been running for a while, so that non-condensables in far reaches of the system will have a chance to migrate to the condensers.

For additional considerations of vacuum, readers are referred to Marty Timm’s paper Designing Industrial Refrigeration Systems for Full Vacuum – Considerations presented at the IIAR 2021
conference. The paper discusses some matters of evacuation presented here, but also investigatesthe topic of vessel design for vacuum pressure.