What Is a Chiller? Types and How Chillers Work

Walk into the plant room of any large office tower, hospital, mall or university campus and you will find the chillers. They are the largest, loudest and most expensive machines in the building, and they usually account for the biggest single slice of the electricity bill. Yet in my experience of sitting with facilities teams while we set up asset registers, PM programmes and plant dashboards, the chiller is also the piece of equipment most people describe confidently and understand loosely. They know the readings to write down. They do not always know what the machine is physically doing between one reading and the next. This guide fixes that, at orientation depth, without assuming you are an engineer.

The message up front: a chiller does not make cold. It picks heat up out of a water loop, concentrates it, and dumps it somewhere else. Every component in the machine exists to serve that one transaction. The single most consequential decision about a chiller is where that heat gets dumped, which is the whole air-cooled versus water-cooled question, and that choice shapes the footprint, the plant room, the water bill, the maintenance load and the rest of the mechanical design.

1. What a chiller actually is

A chiller is a machine that produces chilled water. That is the entire job description. It takes water arriving from the building warmer, removes heat from it, and sends it back out colder, continuously, for as long as the building calls for cooling. That chilled water is piped to coils in air handling units, fan coil units and process equipment, where it absorbs heat, warms up, and returns to the chiller to have that heat taken out again. It is a closed loop that never stops circulating while the plant runs.

So a chiller system is really two circuits joined at the machine. On one side, a chilled water circuit that carries heat from all over the building back to the chiller. On the other side, a heat rejection path that carries that heat out of the building entirely, either straight into the outdoor air or into a condenser water loop feeding a cooling tower. The chiller sits in the middle, and its whole function is to make heat flow from the cool side to the warm side, which is the opposite of the direction heat naturally wants to travel. Doing that against nature is what the energy input pays for.

The terminology is worth pinning down early. A water chiller usually just means a chiller that produces chilled water, the normal building case. An industrial chiller typically means one serving a process rather than human comfort: injection moulding tools, laser equipment, medical imaging, data centre equipment, food and beverage production. The machine is recognisably the same animal, but industrial duty often demands tighter temperature control, different fluids, and a level of redundancy that comfort cooling does not. An air cooled chiller and a water cooled chiller are not different kinds of machine in principle, they simply reject their heat differently, and that difference is large enough that I have given it a section of its own.

2. Why buildings use chilled water instead of cooling air directly

A reasonable question from anyone new to the plant room: if the point is to cool the air people breathe, why not put a small cooling machine at every point of use, the way a domestic split unit works? Small buildings do exactly that, and for them it is the right answer. Large buildings move to chilled water for a handful of practical reasons.

• Water carries heat far better than air. For the same amount of heat moved, a water pipe is dramatically smaller than an air duct. In a tower with limited riser space and tight floor-to-floor heights, distributing cooling as water rather than as cold air is often the only thing that physically fits.
• Centralising the refrigeration is cheaper to run and easier to maintain. A few large machines in one plant room are more efficient at scale than dozens of small ones, and they concentrate the skilled maintenance work, the spares, the controls and the safety management in one place.
• Refrigerant stays in the plant room. Keeping the charge inside a small number of machines in a controlled, monitored space is far easier to manage, detect leaks on and service than piping refrigerant all over an occupied building.
• You can stage capacity to match load. Demand swings enormously between a hot afternoon and a mild night. With several chillers on a common header you bring machines on and off to follow the load, which a fixed collection of small units cannot do nearly as gracefully.
• One loop can serve many different jobs. The same chilled water feeds comfort coils, dehumidification, a server room, a kitchen cold room and a process load, each with its own control valve.

The trade is that you now own a distribution system: pumps, pipework, valves, insulation, air separation, water treatment and controls. That is real cost and real maintenance. The general pattern is that below a certain scale, packaged or split direct expansion equipment wins on simplicity, and above it chilled water wins on space, efficiency and manageability. Where the line falls depends on the building and the load profile, and it is genuinely a design question rather than a rule. If you are weighing centralised chilled water against a distributed refrigerant approach, see chiller versus VRF. For where the chiller sits in the wider building services picture, start with the HVAC fundamentals pillar.

3. The conceptual flip: chillers move heat, they do not make cold

This is the single most useful idea in the article, and it is worth slowing down for, because almost every subsequent confusion traces back to getting it wrong.

There is no such thing as a cold-generating device. Cold is not a substance you can manufacture and inject into water, it is simply the absence of heat. So the only way to make something colder is to take heat out of it and put that heat somewhere else. Every refrigeration machine ever built, from a domestic fridge to the largest centrifugal chiller, is a heat-moving device. It is a pump for heat.

Heat moves on its own from hot things to cold things, never the other way. A chiller needs heat to travel in the forbidden direction: out of chilled water that is already cool, and into outdoor air or condenser water that is already warm. To do that it uses a working fluid, the refrigerant, and one trick: changing the refrigerant's pressure changes the temperature at which it boils and condenses. At low pressure it boils colder than the chilled water, so heat flows naturally into it. At high pressure that same refrigerant condenses hotter than the outdoor air or condenser water, so heat flows naturally out of it. Nature does the heat transfer at both ends, willingly, downhill. The machine's only job is to shuttle the refrigerant between those two pressures.

One consequence is immediately practical: the heat a chiller rejects is always more than the heat it removed from the building. The energy the compressor puts in has to go somewhere too, and it leaves through the condenser along with the building's heat. That is why the heat rejection side of a chiller is bigger than you might expect, and why the condenser end of the plant is never an afterthought.

4. The vapour compression cycle, step by step

Almost every chiller you will meet runs on the vapour compression cycle. It has four steps and four matching pieces of equipment, running round and round continuously. Follow the refrigerant once around the loop and you have understood the machine.

Step 1: evaporation. Start in the evaporator, at the low pressure end. Cold liquid refrigerant sits on one side of a heat exchanger and the building's returning chilled water flows on the other. Because the pressure is low, the refrigerant boils below the chilled water temperature, so heat flows out of the water and into the refrigerant. That heat is used up turning liquid into vapour rather than raising its temperature, which is why boiling is so effective: a change of state absorbs a great deal of heat at a nearly constant temperature. The water leaves colder and goes off to the building. The refrigerant leaves as a cool low pressure vapour carrying the building's heat.

Step 2: compression. That vapour is drawn into the compressor, which squeezes it into a much smaller volume and pushes it out at high pressure. Compressing a gas raises its temperature, so the refrigerant leaves as a hot high pressure vapour. This is the step that consumes the energy and the step that makes the cycle possible, because it lifts the refrigerant above the temperature of the outdoor air or condenser water. Without the compressor there is nowhere for the heat to go.

Step 3: condensation. The hot high pressure vapour flows into the condenser, another heat exchanger, where it meets either outdoor air blown across finned tubes or condenser water returning from a cooling tower. Because the refrigerant is now hotter than whatever it faces, heat flows out of it naturally. As it gives up heat it condenses back to a liquid, still at high pressure. All the heat picked up from the building, plus the heat added by the compressor, leaves the machine at this point. That is why the condenser is the part of the chiller that most directly determines how the plant is designed and sited.

Step 4: expansion. The high pressure liquid then passes through the expansion device, a controlled restriction that drops it back to the low pressure of the evaporator. As the pressure falls, some of the liquid flashes to vapour and the mixture becomes very cold. It arrives in the evaporator ready to boil again, and the cycle repeats. The expansion device is also the metering valve that decides how much refrigerant is fed to the evaporator, which is why it matters far more than its small size suggests.

Notice the symmetry. Two heat exchangers, one where heat comes in and one where heat goes out. Two pressure-changing devices between them, one that raises pressure and one that lowers it. That is the whole machine. Everything else is support.

5. The main components, and the support systems people forget

The four cycle components are the headline, but a real chiller is more than four parts, and the ones people overlook are disproportionately the ones that cause callouts.

• Compressor. The heart of the machine and the largest energy consumer. It raises refrigerant pressure and therefore temperature. Its type defines much of the chiller's character, covered in its own section below.
• Condenser. The heat exchanger that rejects heat. In an air cooled machine, a large finned coil with fans drawing outdoor air across it. In a water cooled machine, typically a shell and tube vessel with condenser water through the tubes. Fouling here, dirty fins or scaled tubes, is one of the most common causes of poor performance.
• Expansion or metering device. Depending on the machine, refrigerant pressure reduction and flow metering may be handled by a thermostatic or electronic expansion valve, a fixed orifice, a float or other manufacturer-specific metering arrangement. It sets the pressure drop and meters refrigerant flow so the evaporator is neither starved nor flooded. Do not confuse this with the inlet guide vanes found on centrifugal compressors, which throttle refrigerant vapour entering the compressor for capacity control and are not the expansion device.
• Evaporator. Usually a shell and tube heat exchanger where chilled water gives up its heat to boiling refrigerant. This is the component whose performance the building feels most directly. For how these vessels transfer heat and why approach temperatures behave as they do, see the heat exchanger explainer.
• Oil system. Frequently ignored by non-specialists and frequently the reason a chiller trips. Most compressor types need lubrication, which means an oil sump, an oil pump, oil coolers, filters, oil pressure and temperature monitoring, and often a sump heater. Oil that migrates into the refrigerant circuit and coats heat exchanger tubes quietly degrades performance in a way no single reading announces. Some machines are designed to be oil free, which removes this whole subsystem and its failure modes.
• Controls. The onboard controller decides how much cooling to produce, stages or modulates capacity, manages compressor loading, protects the machine, and reports to the building management system. Modern chillers hold a lot of useful data internally that never reaches the BMS unless somebody deliberately integrates it. That gap is one of the most common findings in a plant data review, and it is covered in BMS in HVAC: controls, points and sequences.
• Safeties. High and low pressure cutouts, low chilled water temperature protection to prevent the evaporator freezing, flow switches on both water circuits, motor overload and phase protection, oil limits, and refrigerant leak detection in the plant room. These exist because a frozen evaporator or a compressor run without oil or flow is not a minor repair.

6. Air cooled versus water cooled: the decision that shapes everything

If you remember one distinction about chillers, make it this one. It is not a question of which machine is better. It is a question of where the heat goes, and the answer cascades into the building design, the plant room, the roof, the water strategy, the maintenance load and the operating cost.

An air cooled chiller rejects heat directly to outdoor air. Its condenser is a large finned coil with fans, and the machine usually sits outdoors, commonly on a roof or podium. There is no second water loop, no cooling tower, no condenser water pumps and no tower water treatment. It is self contained and the mechanical scope around it is far simpler.

A water cooled chiller rejects heat into a condenser water loop. That water carries the heat to a cooling tower, where it is rejected to atmosphere largely by evaporating a portion of the water. The machine itself is compact and lives indoors, but it now depends on a whole subsystem: condenser water pumps, pipework, a cooling tower, makeup water, blowdown, and water treatment to control scale, corrosion and biological growth. That dependency is not optional, which is why I always point people at the cooling tower explainer alongside this one. You cannot understand a water cooled plant by understanding only the chiller.

Why water cooled machines are generally the more energy efficient choice is directly traceable to the cycle above. A cooling tower can deliver condenser water below the outdoor air temperature, because evaporation cools it. An air cooled condenser never can. Lower condensing temperature means a smaller pressure lift for the compressor, and a smaller lift means less energy. That is the whole efficiency story in one sentence, and it is why large plants in hot climates so often go water cooled despite the extra complexity.

7. Compressor types and roughly where each fits

The compressor is what most distinguishes one chiller from another, and the four types you will encounter each earn their place in a different part of the capacity range. Treat the capacity bands below as broad and qualitative, because they overlap heavily and manufacturers keep pushing the boundaries.

The practical selection logic is less about the compressor badge and more about the load profile. A building whose demand spends most of the year well below design capacity is served better by a machine, or combination of machines, that is efficient at part load, and that often matters more in practice than the headline full load figure. Multiple scroll compressors staged in one machine, screw compressors on variable speed, and centrifugal machines with variable speed drives are all answers to the same question: how do I stay efficient when the building is not asking for everything I have? Ask any supplier how their proposal behaves across your actual load profile, not just at design conditions.

8. Absorption chillers: the thermally driven alternative

Everything so far has assumed a mechanically driven cycle: an electric motor turns a compressor. An absorption chiller reaches the same result using heat as the driving energy instead of shaft work.

The principle, simplified: instead of a compressor raising refrigerant pressure mechanically, an absorption machine uses a pair of fluids, a refrigerant and an absorbent, most familiarly water as the refrigerant with a salt solution as the absorbent. Refrigerant vapour leaving the evaporator is absorbed into the solution rather than compressed. A pump moves that liquid solution to higher pressure, which takes far less energy than compressing a vapour, and a heat input, the generator, boils the refrigerant back out of solution at high pressure. From there the cycle continues conventionally through a condenser and an expansion path to the evaporator. The compressor has effectively been replaced by an absorber, a solution pump and a heat source.

That makes absorption chillers interesting in specific circumstances rather than in general:

• You have cheap or waste heat. Exhaust heat from a generator or turbine, steam from a process or district network, or solar thermal. If the driving heat is essentially free or already paid for, converting it into cooling is compelling.
• Electrical supply is constrained or expensive. Where the incoming supply cannot support a large electrical cooling load, or where tariffs and demand charges are punishing, shifting cooling off the electrical system has real value.
• Combined heat and power or trigeneration schemes. Absorption cooling is what turns a CHP plant's rejected heat into something useful in summer, which is often what makes the overall scheme viable.
• Electrical peak shaving. Running absorption capacity during expensive peak periods and electric capacity otherwise is a legitimate strategy on the right tariff.

9. The chilled water circuit around the machine

A chiller is only as good as the loop it sits in, and a great deal of what gets blamed on chillers is actually a distribution problem. At orientation depth, here is what you need to know about the circuit.

Primary and secondary. In a classic primary secondary arrangement, dedicated primary pumps push a constant flow through each chiller's evaporator, keeping the machine in the flow band it was designed for, while separate secondary pumps handle the variable flow the building actually demands, usually on variable speed drives responding to differential pressure. A bypass or common header decouples the two so the building can throttle without starving the machine. Variable primary flow designs, where one set of variable speed pumps serves both duties within limits the chiller can accept, are common too and remove a set of pumps at the cost of tighter control requirements. Both are valid. What matters is knowing which one you have, because the operating rules differ.

Delta T. The temperature difference between chilled water returning from the building and chilled water leaving the chiller is the single most informative number in the plant, and it is where most distribution problems announce themselves. If return water is coming back cooler than designed, the delta T has collapsed, and the plant has to circulate much more water to deliver the same cooling. Pumps run harder, chillers may be forced to run at low load, and you can end up running an extra machine purely to satisfy flow rather than load. Low delta T syndrome is one of the most common and most expensive conditions in real chilled water plants, and its causes are usually out in the building: valves passing or stuck open, coils fouled or bypassed, incorrect three port arrangements, wrong setpoints, or a system modified repeatedly over the years without anyone revisiting the hydraulics.

What the load side looks like. The chilled water arrives at coils, most visibly inside air handling units and fan coil units, where it cools and dehumidifies air. Understanding what those terminal units are doing with the water helps enormously in diagnosing plant behaviour, and the AHU explainer covers that end of the system. A plant room reading and a terminal unit problem are often the same story told from opposite ends.

10. Refrigerants, treated carefully

Refrigerant is the working fluid that makes the cycle possible, and it is also the most regulated and fastest moving aspect of chiller selection. I am deliberately not going to tell you which refrigerants are acceptable, because that is exactly the kind of statement that ages badly and differs by jurisdiction.

What is stable enough to say plainly: refrigerant selection is driven by regulation as much as by engineering. Handling, recovery, leak detection, record keeping and technician competency requirements exist in most jurisdictions, and the specific rules, the substances they cover and the timelines they set differ by country and change over time. Equipment efficiency requirements are likewise regulated in many jurisdictions and revised periodically. The practical consequence is that a refrigerant decision made three years ago may not be the right decision on a project starting today in the same city, let alone in a different country.

So the advisory position I would take:

• Treat refrigerant as a live compliance question, not a settled specification. Confirm current requirements with the authority having jurisdiction and with the manufacturer at the time of the decision, in writing, rather than carrying forward a previous project's specification.
• Ask about the whole service life, not just handover. Ask what the manufacturer's position is on long term availability and service support, and what a future conversion would involve.
• Take safety classification seriously. Refrigerants differ in toxicity and flammability characteristics, and those characteristics drive plant room design, ventilation, leak detection and alarm requirements. Confirm this locally rather than assuming it from a previous project.
• Manage leaks as an operational discipline. A machine short of charge will not deliver its rated cooling and will run inefficiently while it disappoints you.
• Keep the records. Charge quantities, top ups, recovery, technician qualifications and leak test results. Whatever your jurisdiction requires, disciplined record keeping is what makes an inspection uneventful.

For authoritative starting points rather than secondhand summaries, go to the technical society literature and your national environmental regulator. In the United States the US EPA refrigerant management pages set out that jurisdiction's handling and certification regime, and ASHRAE is the usual reference point for refrigerant designation and safety classification practice. Outside those jurisdictions, go to your own regulator. Do not assume a rule from one country applies in another, or that a rule you learned five years ago is still current.

11. The readings that tell you a chiller's health

Now that the cycle makes sense, the log sheet stops being a ritual and becomes a diagnostic. A few families of reading carry most of the signal, and you can interpret them from first principles using nothing more than the four step cycle above.

• Approach temperatures. The approach is the gap between the refrigerant temperature in a heat exchanger and the water temperature leaving it, at the evaporator and again at the condenser. It is a direct measure of how well that heat exchanger is transferring heat. When an approach grows over time at comparable load and conditions, heat is struggling to cross the barrier, and the usual reasons are fouled or scaled tubes, oil coating the surfaces, air or non condensables on the condenser side, or a charge problem. Approach trends are the closest thing a chiller has to an early warning system, and they only work if you record them consistently enough to have a trend.
• Delta T across the evaporator, and across the building. Compare actual delta T against design. A collapsed delta T at the plant usually points outward into the distribution system rather than inward at the machine. This single comparison redirects more plant investigations than any other number on the log sheet.
• Current draw, alongside load and conditions. Motor current tells you how hard the compressor is working. On its own it means little. Read against the cooling being delivered and the conditions at the condenser, it tells you whether the machine is working harder than it should for the output, which is what an efficiency problem looks like from the log sheet. Rising current at the same load and conditions is a question worth asking.
• Condenser and evaporator pressures, and the gap between them. The pressure difference is the lift the compressor is being asked to achieve, and the lift is the energy. A high condensing pressure is the machine telling you the heat rejection side is not doing its job, whether that is a dirty coil, a struggling tower, poor airflow, recirculation of hot discharge air, or fouled tubes.
• Oil pressure, temperature and level. On machines with an oil system, these are both a health indicator and a protection function. Treat an oil alarm as a serious event rather than a nuisance trip to be reset.

The discipline that makes all of this work is boring and non negotiable: readings taken at consistent intervals, recorded against load and ambient conditions so they are comparable, and actually reviewed by somebody who will act. A log sheet filled in faithfully and filed unread is a cost with no benefit. Most sites already have somewhere structured to put that data, whether that is the BMS or a maintenance management system, and the generic point is that readings should sit close to where work orders are raised. Which product that is matters far less than whether the trend gets looked at.

I have deliberately not written a PM schedule here, because this is an equipment explainer and there is a better home for that. For the routine servicing regime, the log sheet practice and a liftable schedule you can adapt, go to boiler and chiller preventive maintenance. For what happens when you take these same readings continuously and analyse them at plant level rather than machine level, including sequencing, staging and plant efficiency, see chiller plant analytics. Those two articles carry the depth this one deliberately stops short of.

The idea to walk away with

A chiller is a heat pump with a job title. It picks heat up out of a water loop at low pressure, where the refrigerant boils below the water temperature, and puts that heat down at high pressure, where the refrigerant condenses above the outdoor air or condenser water. Compressor, condenser, expansion device, evaporator: four components, one transaction, endlessly repeated.

Hold that model and the rest becomes reasoning rather than memorising. Air cooled versus water cooled is a question about where the heat goes and what you are prepared to maintain. Compressor type is a question about scale and load profile. Absorption is a question about what energy you have available to drive the cycle. And the readings on the log sheet are the cycle reporting on itself: approach temperatures telling you whether heat can get across the heat exchangers, delta T telling you whether the distribution system is behaving, and current draw telling you what the compressor is paying for the lift.

Final thoughts

A common failure in chiller plant operation is not mechanical. It is that the people running the plant were trained on the procedure and never on the principle, so they can record a reading but not interpret it, and cannot tell a machine problem from a water side problem or a building problem. That gap costs real money in unnecessary callouts, misdirected investigations, premature replacement and energy quietly wasted for years.

If you are responsible for a chiller plant and you want one action from this article, make it this: take a month of your existing log sheets, plot the approach temperatures and the delta T against load, and see whether the trend tells you anything. In most plants it does, and the finding is usually something that was in the data all along. The plant was talking. Nobody had been taught the language.

Related reading: What is HVAC: meaning, systems and components, Chiller vs VRF: what is the difference, What is a cooling tower and how it works, What is a heat exchanger, What is an AHU, Boiler and chiller preventive maintenance, Chiller plant analytics, BMS in HVAC: controls, points and sequences.