Solar Cooling Using Ejectors


The ejector is a thermally driven compressor that operates in a heat pump refrigeration cycle. In a heat pump system, the ejector takes the place of the electrically driven compressor, but uses heat rather than electricity to produce the compression effect

Ejectors – A brief history

Ejectors have been in use prior to 1900 where they found use in evacuating air from leaky low pressure steam condensers. An ejector in this application acts as a vacuum pump, driven by low pressure steam which was readily available in such environments. The ejector’s role was characterised by steady state conditions and empirical design. Efficiency was not as important as reliability.

Within 20 years, ejectors found widespread use as vacuum pumps in industrial settings. It was a small step to form a vapour compression heat pump using the ejector as a heat driven compressor. Steam driven ejector heat pumps became common in air conditioning, particularly of hotels and ships during the early 20th century; wherever there was a ready supply of low pressure steam or a steam boiler. Ejector systems were found to be low cost, very reliable and maintenance free.

During the 1930s, Freon refrigerants were developed and vapour compression heat pumps based on these new refrigerants were far superior in performance to ejector systems. Ejector air conditioning fell from favour for 50 years until the Montreal protocol of 1987 highlighted a link between Freon use and atmospheric ozone depletion.

This rekindled an interest in ejector technology and at about this time, two important improvements in ejector design were made. Firstly, refrigerants other than water were tested and found to perform better. Secondly, researchers began to look at system integration issues and to compose systems incorporating solar energy and hybrid designs.

The modern era of ejector research combines supersonic thermodynamics, computational fluid dynamics and experimental work. Despite this effort, the inner workings of the apparently simple ejector are not fully understood, but are reasonably well modelled.

Researchers are able to design ejectors with confidence and there are industrial ejectors ranging in size form several hundred watts to huge multi-megawatt steam ejectors.

Description of the Ejector System

The ejector is a thermally driven compressor that operates in a heat pump refrigeration cycle. In a heat pump system, the ejector takes the place of the electrically driven compressor, but uses heat rather than electricity to produce the compression effect (figure 1).

The ejector has no moving parts and is simple and reliable which make it attractive for commercial production. However, the thermal efficiency of the ejector is low which implies that the ejector requires a large solar collector and large condenser to operate in a heat pump application. Thus the savings in electricity consumption must be compared with the additional cost of the solar collector. One is trading capital cost for operating cost, as with most solar systems.

A liquid pump is required to generate a pressure difference for the ejector heat pump to operate, but since liquid is being compressed, the amount of electricity required is relatively small. All other components in the heat pump circuit are conventional.

The ejector cycle consists of high and low temperature sub cycles. In the high temperature sub cycle, heat that is transferred to the ejector cycle from the heat source causes vapourisation of the ejector cycle working fluid in the generator at a temperature slightly above the saturation temperature of the refrigerant. Vapour then flows to the ejector where it is accelerated through a converging-diverging nozzle.

Since much of the vapour enthalpy is converted to kinetic energy, conservation of energy suggests that the vapour temperature and pressure will be very low. The low pressure at the exit of the nozzle which acts to draw vapour flow from the evaporator (figure 2,c).

The generator and evaporator flows then mix in the ejector and the combined flow undergoes a transverse compression shock (figure 2,f). Thus thermal compression replaces the electrical compressor in a conventional heat pump. Further compression takes place in the diffuser such that a subsonic stream emerging from the ejector then flows into the condenser (figure 2,h).

At the condensor, heat is rejected from the working fluid to the surroundings, resulting in a condensed refrigerant liquid at the condenser exit. The ejector needs to provide sufficient exit pressure such that the saturation temperature of the refrigerant at this point is greater than the condenser cooling medium, otherwise heat cannot be rejected and the cycle ceases to operate. This is the malfunction mode of the ejector, caused by excessive condensing backpressure. Malfunction can be overcome by suppling greater generator pressure and temperature.

Liquid refrigerant leaving the condenser is then divided into two streams; one enters the evaporator after a pressure reduction through the expansion valve, the other is routed back into the generator after undergoing a pressure increase through the refrigerant pump. The fluid is evaporated in the evaporator, absorbing heat from the air-conditioned environment, and then it is entrained back into the ejector completing the cycle.

Although figure 1 indicates a direct connection of the generator and solar array, there is usually an additional heat exchange circuit (figure 3) in the high pressure loop to eliminate the possibility of two phase refrigerant flow in the solar collector. For readers familiar with p-h diagrams, figure 3 clearly shows the high and low pressure ejector sub cycles.

Despite the complexity of internal operation, the ejector may still be considered to be a compressor and its performance may be defined conventionally by its compression ratio (Pc/Pe) and its isentropic efficiency. The ejector heat pump cycle still benefits from subcooling prior to evaporation and from minimising superheating through compression.

The advantages of simplicity and reliability of the ejector will be apparent. Additionally, the ejector mechanism offers freedom of choice of refrigerant and is not complicated by the need for compressor lubricant compatibility. Also, the ejector is tolerant of liquid slugging since both generator and evaporator ports are essentially open tubes.

Operational Characteristics

The operational characteristics of fixed geometry ejectors are somewhat different to conventional compressors. The ejector performance is typically very sensitive to gas properties and operation away from the design point.

Ejector performance is noted by a constant capacity region, a critical operating point and a malfunction region, for a given evaporation and condensing temperature (figure 4). Ideal operation of the ejector, indicated by maximum entrainment of the evaporator flow, is indicated by the knee of each curve in the figure. This point is very close to the malfunction condensing temperature where the entrainment falls to zero so that there is no cooling effect. Indeed the ejector is so sensitive to backpressure (itself related to ambient temperature), that complete malfunction occurs with several degrees of condensing temperature from the optimum operating point.

A second important observation is that an increase in generator (solar) temperature will allow continued operation at elevated condensing temperature but at the expense of COP (figure 5). This is because the mass flow of the choked primary choked nozzle decreases with increasing driving temperature and thus, there is less motive power to combat the increased condenser backpressure. This implies that a fixed geometry ejector will not be able to take advantage of high collector temperatures during periods of high insolation.

While most research papers expound the benefits of the coincidence of solar radiation and cooling demand, close examination shows that this is not the panacea it is made out to be. Solar radiation intensity on a horizontal plane peaks at noon. Thus the heat available to a solar driven ejector peaks at the same time. Due to the thermal inertia of the earth, ambient temperature peaks around mid afternoon, several hours after the peak in solar availability (figure 6). Furthermore, the inside of a building peaks in temperature several hours later as its own thermal inertia reacts to ambient temperature and, to a lesser extent, solar radiation. Thus the peak cooling demand of a building may occur late into the afternoon, well after the solar cooling system is able to deliver its rated cooling capacity. Frustratingly, the solar cooling system will have excess capacity in the morning when it may not be required.

Some reprieve may be gained by orienting the solar collector to the west such that the peak solar availability is less, but more coincident with demand. However, far greater effectiveness may be achieved by storing the coolth produced early in the day for later use. This has benefits in reducing the peak afternoon requirements (and thus collector size and cost), but also in allowing solar contribution to evening cooling. An annual computer simulation readily reveals the tradeoff between collector size, storage capacity and solar contribution to the cooling demand (figure 7), in this case for a small ejector system for a residence. The cost of storage and solar collectors should be overlaid on this data in order to make an ejector system as cost effective as it can be.

Key Research Challenges

Ejector based systems are characterised by their simplicity and high reliability, tolerance to a range of working fluids, but also poor performance and range of operating conditions. The performance issues must be addressed if ejectors are to become mainstream. In particular, the low thermal COP and poor off-design performance must be improved.

Performance is usually interpreted relative to cost. The cost of the ejector system is dominated by the cost of the heat source which can be high if solar collectors are required. Commonly temperatures around 70-100ºC are required to drive an ejector, making them suitable for use with non-concentrating solar collectors or waste heat from cogeneration systems. Successful ejector based systems will seek to maximise annual utilisation of the heat source by providing multiple services: summer space cooling, winter space heating and water heating. Such systems do not exist although the technology is available.

The third set of challenges relate to system integration. There are challenges in design and sizing of ejector system components such that the cooling system maximises the use of the (varying) solar collector output to produce the highest solar contribution to the cooling load. This has implications for including energy storage and smart control schemes into the system design. Little research has been performed in this area to date. Indeed, there has been little research into real-time dynamic control of ejector systems since most operate at steady state in a laboratory environment.

Ejector Modelling

There are a number of means to model the performance of an ejector as a component and an ejector system. The ejector is often modelled in isolation in research reported in the literature. This modelling is usually either directly based on thermodynamic compressible flow theory with minor corrections for non-ideal behaviour, or numerically derived using computational fluid dynamics.

Thermodynamic Modelling

For over fifty years, ejectors were empirically designed based on rules of thumb and experience with steam driven devices. Early attempts to thermodynamically model the steam ejector were carried out by Keenan (1950). Keenan was able to reasonably predict the ejector performance characteristics firstly for constant area mixing ejectors then also for constant pressure mixing ejectors. Further clarification of the mixing mechanism was provided by Munday and Bagster (1977). Perhaps the most important improvement in understanding was provided by Eames (1995) and Huang et al (1999) with the description of a one dimensional design methodology. These advances provided researchers with a means to design an ejector, including the effect of supersonic shock and three calibration constants such that model and experimental data generally agreed within about 10%.

Based on this approach, ejector design maps for fixed and variable geometry ejectors may be produced (figure 8a,b). Such images do not describe the operational characteristics, but are helpful in determining the size of the mixing chamber for double choking mode of ejector operation.

Since Huang’s method, there have been several modifications that give refined descriptions of ejector operation. The first is the Shock circle method (Zhu, 2007) which better accounts for the boundary layer secondary flow in the mixing chamber prior to mixing. This gives typical errors of less than 5% compared to published experimental data and would be considered the current state of conventional thermodynamic modelling of ejectors.

An unconventional approach to ejector design was suggested by Eames (2002) whereby the ejector was shaped to produce a constant acceleration in the mixed refrigerant flow. This eliminated the entropic shock in the ejector and so this design was able to produce greater compression effect for a given driving power. However, the design is only suitable for steady state operation at the design point since the very specific shape is does not allow correct operation at off-design conditions. Nevertheless, it is well suited to a cogeneration system whereby the conditions would be reasonably controlled.

However, the ejector is but one component in a solar cooling system. A proper understanding on such a system can be better obtained by modelling the dynamic behaviour of the ejector in response to changing operating conditions (ambient temperature, solar radiation) and in response to a control strategy. The ejector component has low thermal mass and short transport delays so that the dynamics of an ejector circuit are generally dominated by other components, particularly the solar collector.

Furthermore, the ejector system should be modelled over an entire season to evaluate the effect of component sizing and control strategies, especially if storage is included in the system design. Annual modelling is readily accomplished using a program such as TRNSYS which captures local weather data, dynamic loads and control responses at high time resolution.

Researchers widely acknowledge that thermodynamic modelling cannot precisely describe the mixing process occurring inside the ejector. Further advances in this area through calibration constants may provide improved matching to experimental data but are not likely to provide useful insights into ejector processes. Recent advances in numerical modelling appear to be the best approach in this case.

Numerical Modelling (CFD)

Computational Fluid Dynamics has matured over the last decade with the advance in hardware computational capability. This is allowing researchers to investigate the ejector processes in far greater detail including supersonic shock effects, real gas behaviour, metastable refrigerant states, boundary layer flow, flow separation and the like. Due to the complexity of highly turbulent supersonic compressible flow involving a real gas model, only highly developed CFD packages are suitable for ejector modelling. Most researchers use Fluent or ANSYS CFD.

Perhaps the most important choice in CFD analysis of ejectors is the selection of a turbulence model. The standard ?-e turbulence model has found to be inadequate for describing expanding supersonic jets and Bartosiewicz (2003) offers some recommendations. In particular, the hybrid ?-e-sst model seems to offer good results.

The results of CFD studies are now producing close agreement in entrainment ratios with experimental data, provided that the ejector is operating in double choking mode (figure 9). Despite this apparent success, models with differing turbulence handling may agree on design point entrainment while demonstrating very different flow phenomena within the ejector. Thus it may be too early to make useful deductions on the inner workings of the ejector.

Insights into real ejector flows are provided by advanced visualisation techniques involving transparent ejectors. Few researchers are involved in this activity and results are limited at this stage. This technique will become important as a means to verify CFD predictions.

Combined Cycles

Hybrid Cycles

Despite recent advances in understanding of ejector operation, the COP of a simple ejector cooling system remains stubbornly low. As a result, many researchers have proposed hybrid cooling systems incorporating combinations of ejectors and another cooling system. The most common hybrid systems are:

  • Multiple ejector hybrids
  • Mechanical vapour compression / ejector hybrids
  • Absorption / ejector hybrids

Multiple Ejector Hybrids

Several attempts have been made to combine several ejectors in a cooling system. The most obvious configuration of a double effect ejector, whereby the condensing heat from the topping cycle ejector drives the generator of the bottom cycle ejector, is yet to be demonstrated.

However, a compound ejector has been proposed by Dennis and Garzoli (2009). In this configuration, several ejectors effectively work in series to build vapour compression. This configuration was found to substantially relieve the constant capacity constraint at low condensing temperatures and provide impressive performance gains at high condensing temperatures when compared to a conventional ejector (figure 10). The two ejectors were found to require the same mixing chamber geometry and differ only in the primary nozzle dimension. Design is possible using simple rules of thumb.

The critical design parameters are the setting of the intermediate pressure between the high and low pressure ejectors and the division of solar motive energy between the two ejectors.

Yu et al (2006) proposed a two stage compound ejector whereby the second stage of compression was performed by a liquid jet pump ejector. This configuration is also performing well using HFC refrigerants.

There appears to be an advantage with compound ejectors compared to multiple effect cooling systems (eg absorption) in that an increase in generator temperature is not necessarily required for this design.

True multiple effect ejector systems are yet to be demonstrated. This would involve collecting heat from the diffuser of one ejector to evaporate vapour in the generator of a downstream ejector. As with other multiple effect cooling systems (eg absorption) this would require elevated collector temperatures.

Mechanical vapour compression / ejector hybrids

The use of multiple ejectors helps to address the off design performance constraints of ejectors. However, the issue of intermittency of heat source can be troublesome, particularly when an ejector is coupled to a solar collector for motive energy. Furthermore, in such cases, cooling services are commonly required into the evening due to thermal inertia of buildings and so some kind of backup or storage system is required. Since thermal storage is expensive, an elegant solution is to combine a mechanical vapour compression cooling system with an ejector.

An ejector can assist the mechanical system in a number of ways; by reducing the compression work of the mechanical compressor by directly pre-compressing vapour from the evaporator (Takeuchi, 2009) by cooling the condenser of the vapour compression unit (figure 12)(Sokolov, 1993) and by pre-cooling the liquid prior to its entry to the evaporator (Huang 2001). This design neatly takes care of the solar intermittency but does not address the off design limitations of ejectors.

In each of these cases, peak load on the electrical vapour compressor is reduced. This has benefits for the electricity grid and may allow the electrical compressor to be downsized. Also, the cooling system reverts to conventional electrical operation when solar input is not available so that the addition of the ejector is a peak lopping strategy.

Several cascade configurations involving ejectors have also been proposed. In this configuration, the ejector is driven by the heat of compression of the electrical compressor and acts to reduce the pressure ratio of the electrical compressor (Yu, 2006). Such configurations are attractive since no external motive force is required to drive the ejector and little additional complexity or cost is added to the refrigeration system in order to realize the savings.

Absorption / ejector hybrids

A number of researchers have attempted to use ejectors to boost absorption cycle chillers.

In this case, the ejector compressor works in parallel or series with the absorption compressor and is driven by the regeneration heat from the absorption cycle. In such cases, the COP can easily exceed 1.0 for a single stage absorption chiller but requires elevated generating temperatures.

This type of hybrid forgoes the advantages of ejectors – simplicity of implementation.

Ejector Research at ANU

At ANU, we are progressing knowledge in the field of solar ejector cooling systems in the following areas:

  1. Hybrid designs
  2. Systems integration
  3. Advanced predictive control systems
  4. Cool storage using warm ice
  5. Advanced high performance ejector design including compound and variable
    geometry designs
  6. CFD analysis of ejectors


Bartosiewicz Y., Aidoun Z., Desevaux P.,Mercadier Y., CFD experiments integration in the evaluation of six turbulence models for supersonic ejector modelling, Proceedings of Integrating CFD and Experiments, Glasgow, 2003.

Chunnamond, K & Aphornratana, S 2004, Ejectors: applications in refrigeration technology, Renewable and Sustainable Energy Reviews, vol. 8, pp. 129-55.

Dennis M., Garzoli K., Compound Ejectors with improved off-design performance, School of Engineering, The Australian National University, 2009.

Eames I., A new prescription for the design of supersonic jet pumps: the constant rate of momentum change method, Applied Thermal Engineering, Vol 22, pp121-131, 2002.

Eames, IW, Aphornratana, S & Haider, H 1995, ‘A theoretical and experimental study of a small-scale steam jet refrigerator’, International Journal of Refrigeration, vol. 18, no.6, pp. 378-86.

Huang B., Chang J., Wang C., Petrenko V., 1999, A 1-D analysis of ejector performance, International Journal of Refrigeration, vol. 22, pp. 354-64.

Huang B., Petrenko V., Chang J, Lin C., Hu S., A combined cycle refrigeration system using ejector cooling cycle as bottoming cycle, International Journal of Refrigeration 24 (2001) 391-399.

Keenan, H., Neumann, E.P., Lustwerk, F., An investigation of ejector design by analysis and experiment., J. Appl. Mech.Trans. ASME 72, 299–309.,1950.

Lewin S., Performance optimisation of a solar assisted air conditioning system, Honours Thesis, School of Engineering, The Australian National University, 2008.

Munday J., Bagster D., A new ejector theory applied to steam jet refrigeration, Ind Eng Chem, Process Des Dev, 1977, Vol 16 pp 442-449.

Sokolov M., Hershgal D., 1993, Optimal coupling and feasibility of a solar-powered year round ejector air conditioner, Solar Energy, vol. 50, no. 6, pp. 507-16.

Takeuchi H., World’s first ejector cycle for mobile refrigerators to stop global warming, Proceedings Eurotherm 85 International seminar on ejector/jet-pump technology and applications, Louvain La Neuve, Belgium, 2009.

Yu J., Chen H., Ren Y., Li Y., A new ejector refrigeration system with an additional jet pump, Applied Thermal Engineering 26 (2006) 312–319.

Zhu C., Wen L., Shock Circle method for ejector performance evaluation, Energy Conversion and Management, Vol 48, pp 2533-2541, 2007

Updated:  10 August 2021/Responsible Officer:  Dean, CECS/Page Contact:  CECS Marketing