Absorption heat pump cycles with NH3 – ionic liquid working pairs

(cid:129) Nine NH 3 /IL pairs are investigated in AHPs for building heating. (cid:129) COP of four NH 3 /IL pairs beats that of the NH 3 /H 2 O. (cid:129) Idealized NH 3 /IL mixture can reach a COP of 1.84. (cid:129) [Emim][SCN] is currently a feasible candidate to be used in AHPs with PHX.


Introduction
The Paris Agreement adopted by 195 countries in the 2015 Paris climate conference (COP 21) reset the global ambition: limiting the temperature rise from pre-industrial levels well below 2 K. Efforts responding to climate change are also accelerating the way the energy sector is developing [1].Heating and cooling, especially for buildings take up the majority of the energy consumption and the greenhouse gases emission.According to the European Commission, heating and cooling consumed 50% (22.85EJ) of the final energy consumption in the EU in 2012.45% of energy for heating and cooling in the EU was used in the residential sector, 37% in industry and 18% in services [2].In the US, 41% (42.2 EJ) of the primary energy in 2010 was consumed by the buildings sector, compared to 30% by the industrial sector and 29% by the transportation sector.Heating and cooling took 59% of the buildings energy consumption [3].As an increasingly significant energy consumer in the buildings sector, China is the largest energy-consuming economy in the world, and buildings energy used in China was the second-largest in the world after the US, representing nearly 16% of total global energy consumption in buildings in 2012 (more than 18 EJ) [4].
Absorption refrigeration and heat pump cycles are drawing considerable attention because they can take effective advantage of lowgrade heat from concentrating solar collectors or waste heat, providing opportunities for clean and sustainable energy utilizations [5][6][7][8].Working pairs H 2 O/LiBr and NH 3 /H 2 O have been widely used in certain applications in absorption systems, while many challenges do exist, such as crystallization possibilities of the H 2 O/LiBr pair and the difficulty in the separation of the NH 3 /H 2 O pair [9].Thus, the investigation of alternative solvents is still a relevant topic [10][11][12][13][14]9].
Ionic liquids (ILs), whose properties can be adjusted by the design of anion and cation combination for a task-specified purpose, have drawn considerable attention for their potential roles in replacing conventional absorbents used in absorption refrigeration and heat pump cycles in the past years.Researchers recognized the strengths of ILs in applications, such as high boiling point, good affinity with refrigerants, and high chemical and thermal stabilities [9].Nevertheless, there are also some challenges related to the technical feasibility and costs when introducing them, thus many efforts are still needed before the ILs are accepted in practice.
In order to preselect promising ILs to be used in absorption systems, many researchers did performance investigations.The majority of investigations were focused on performance predictions, in which the frequently studied refrigerants include H 2 O [15,16], hydrocarbons (HCs) [17], hydrofluorocarbons (HFC) [18,19] and CO 2 [20].Since NH 3 based absorption systems hold strengths such as sub-zero degree applications and free of air infiltration, research related to these mixtures is most relevant.Nevertheless, there is only limited work which has been reported.Yokozeki and Shiflett [21,22] measured solubility data for NH 3 with a set of ILs, and calculated the thermodynamic performance of these mixtures in a single-effect cycle.Kotenko [23] also developed thermodynamic simulations for absorption heat pumps (AHPs) with 4 NH 3 /ILs mixtures in Aspen Plus, and compared their performances with that of the NH 3 /H 2 O system.Their results showed that the efficiency of some of the investigated NH 3 /IL AHP processes, at specified operating conditions, was higher than that of conventional NH 3 /H 2 O systems.Chen et al. [24,25] investigated vapor-liquid equilibria (VLE) for metal ion-containing ionic liquid [bmim][Zn 2 Cl 5 ] with NH 3 , and compared the thermodynamic performance of this mixture with that of the NH 3 /NaSCN pair.The performance of the former system is better than that of the latter one when the generator temperature is high and the absorber and the condenser temperatures are low.Ruiz et al. [26] modeled NH 3 /IL absorption using COSMO-based Aspen simulations and analyzed cycle performance for conventional and task-specific ILs.
In these performance prediction studies of absorption systems, the enthalpy of the NH 3 /IL solution is always an essential thermodynamic property.Most researchers obtained this property by adding an excess enthalpy to the sum of enthalpies of the two pure components.The excess enthalpy could be obtained from the VLE data via a variety of models.Some researchers [27,15,28] used the non-random two-liquid (NRTL) activity coefficient model to predict it.However, Shiflett and Yokozeki [27] found that an accurate prediction of the mixing enthalpy with NRTL is very difficult, because the excess enthalpy is derived from the temperature derivative of the activity coefficient, and the temperature-dependency in any activity model is always in a purely empirical form.Therefore then they turned to a cubic equation of state (EOS) method for excess enthalpy predictions [21,22].Meanwhile, for the pure components part, one of the challenges is the heat capacity of the IL.In the work of Yokozeki and Shiflett [21,22], this part was obtained from a group function contribution method [29].However, Cai et al. [20] pointed out that this group function contribution method is not always accurate.In the prediction work of Chen et al. [25], experimental heat capacity data of ILs were employed in the enthalpy predictions.Therefore, according to the previous studies, the combination of EOS based method for the excess enthalpy with the experimental heat capacity data of ILs may provide a more accurate way for the performance prediction of AHPs.
Previous experimental studies were carried out either by substituting the working pairs in a traditional commercial system [30][31][32] with H 2 O/ILs pairs, or restricted to small scales systems [33].Apart from these studies, the understanding of ILs in more practical aspects, for example the heat and mass transfer aspects, is still limited.However, researches of IL-based working pairs is emerging recently taking more practical aspects into account.Meyer et al. [34] studied the combined heat and mass transfer phenomena of H 2 O/[emim][DEP] pairs in an absorption refrigeration system by using analytical functions.Ariyadi and Coronas [35] developed a measurement setup to study the absorption capacity of the NH 3 vapor in ILs in a pool type absorber.Wadekar [36] simulated the heat transfer behavior of IL [bmim][Tf 2 N] in different heat exchangers (HXs).The results showed that the heat transfer performance was not particularly attractive, but heat transfer enhancement technology can improve it effectively.Boman et al. [37] screened working pairs including the IL-based ones for a single-effect AHP based on both thermodynamic and heat transfer principles.The shell-and-tube HXs of IL-based AHP systems need more heat exchanger area due to the poor heat transfer performance of the ILs.Chugh et al. [38] implemented a membrane-based semi-open absorption system using IL for heating, dehumidification and cooling application.The experimental test achieved a heating coefficient of performance (COP) of 1.4.
In this paper, a thermodynamic model of single-effect AHPs is first proposed accompanied with an accurate method to estimate enthalpies of solutions.With this model, the performance behaviors of 9 commercialized ILs with NH 3 are investigated along with that of the conventional NH 3 /H 2 O pair.Considering previous studies, the 9 working pairs cover all the ILs which have sufficient published data (VLE data with NH 3 and pure heat capacities of ILs) for this calculation.The performance of four of these pairs in absorption systems is reported for the first time.The influence of the GEN temperatures on the circulation ratio (f) and on the COP is also studied.As one of the most original parts of this study, the developed NRTL and heat capacity models have been made generic, by integrating them with a genetic algorithm (GA) in the thermodynamic AHP model, to determine the maximum COP of the AHP cycle and explore how the thermodynamic properties of the ideal ILs should show.In addition, the heat transfer calculations for each heat exchanger of an AHP are carried out considering all heat exchangers are plate heat exchangers (PHXs).Based on that, the feasibility of applying IL in an AHP system is analyzed by investigating its economic performance.

Thermodynamic model of the cycle
Thermodynamic models of the single-effect AHP systems have been frequently reported in the literature, see for instance Kiss and Infante Ferreira [39].In this section, only the details required for the following steps of this paper will be discussed.
Fig. 1 depicts a schematic diagram of a single-effect absorption refrigeration/heat pump cycle.The system is mainly composed of an absorber (ABS), a generator (GEN), a condenser (CON), an evaporator (EVA), along with a solution heat exchanger (SHX), a pump and two throttle valves.
To qualitatively illustrate the temperature and pressure relationship of each state, the process is also plotted in a ln P − (−1/T) diagram in Fig. 2. In the ABS, the weak NH 3 /IL solution 5 (weak in the refrigerant, NH 3 ) absorbs the saturated pure NH 3 refrigerant vapor 1 from the EVA, and then it turns into strong solution 2. The heat Q abs is delivered to the heating system by the ABS.The outlet solution 2 from the ABS is then pumped to a high pressure level and enters the SHX as a cold flow.The outlet flow of the cold side, stream 4, then goes into the GEN, where the driving heat Q gen is input.With the heat input, strong solution 4 releases some refrigerant vapor 8, then becomes the poor solution 7 and enters the SHX.In the SHX, the weak solution 7 is cooled by the cold side to a state of 6 and then throttled to a low pressure level through a valve, before going back to the ABS.The superheated refrigerant vapor 8 from the GEN is condensed to a saturated pure liquid refrigerant in the CON, where the heat Q con is delivered to the heating system.After that, the saturated liquid refrigerant 9 expands to a low pressure level through a valve, and extracts heat Q eva from the surrounding in the EVA.The outlet vapor 1 goes back to the ABS and finishes the cycle.
In order to create an integrated model for the thermodynamic analysis of the absorption process with NH 3 /ILs pairs, several assumptions are made to simplify the calculations: • The system operates in a steady state.
• The heat losses, pressure losses and pumping work are neglected.
The throttling is an isenthalpic process.
• The operating pressures of the EVA and the ABS are the same, and similarly, the pressures of the GEN and the CON are also equal.
• The minimum temperature approach of the solution heat exchanger, SHX, is set to 5 K.
• The refrigerant stream is saturated liquid or saturated vapor at the outlet of the CON or the EVA, respectively.The solution is at equilibrium state when leaving the GEN.While the solution leaving the ABS is subcooled, with a subcooling of 3 K.
• Vapor leaving the GEN is pure NH 3 which has the same temperature as the inlet solution (Fig. 3(a)).
The AHP system with an NH 3 /H 2 O pair is also modeled in this study for reference purposes.For it, an additional column and a rectifier are needed to purify the outlet vapor, as shown in Fig. 3(b).The reflux ratio must be iterated until an acceptable purity of the vapor outlet is obtained.Thus, additional simplification is taken into account, • Vapor leaving the rectifier is pure enough (99.99% NH 3 ), and has a saturated temperature corresponding to the operating pressure (Fig. 3(b)).After the rectifier, it is treated as a pure fluid.
With the enthalpy of each state point, the heats exchanged in the EVA and the GEN are, (1) From the mass balance of the refrigerant in the solution, The performance parameters, i.e. circulation ratio ( f ) and COP can be calculated as, 2.2.Properties 2.2.1.Vapor-liquid equilibria for the NH 3 /ILs binary solutions Vapor-liquid equilibria describe the relationships between parameters P-T-x, which can be used to identified the state points in the cycle.NRTL models for the prediction of VLE of mixtures have been frequently reported in the literature, see for instance [39].In this section, only the details required for the following steps of this paper will be discussed.
For the NH 3 /IL system, due to the non-volatility of ILs, the equilibrium criterion is simplified as, here, P sat NH3 can be obtained from NIST Refprop [40].The activity coefficient γ NH3 can be obtained through the NRTL activity coefficient model after correlating VLE data, where, (8)

Enthalpies of the refrigerant and solutions
The enthalpy data of pure NH 3 are directly obtained from NISTs Refprop [40].For a real solution, the total enthalpy can be estimated using the following method, depending on its state.
For a saturated solution at an equilibrium condition T P , and w NH3 , the total enthalpy is, ) where the enthalpies of NH 3 are chosen at their saturated liquid states.
For the ILs, the enthalpies are calculated with the help of their pure heat capacities C p IL , The calculation of the excess enthalpy, h E , can be obtained using an   [21,22,41] employed a modified Redlich-Kwong (RK) type of cubic EOS to fit the vapor pressure data and to predict the excess enthalpy.This method is also used for the prediction of the mixing heat in the present work.The detailed approach has been reported in the mentioned references.The values of the critical temperature and pressure of ILs needed for the following calculations, along with their molecular weights, are listed in Table 1.
For subcooled solutions at condition T P , and w NH3 , its enthalpy can be obtained by subtracting the subcooled part from a corresponding saturated solution, In this study, the weighted average heat capacity of both components has been implemented to express C p sol .
This treatment has been verified for H 2 O/[mmim][DMP] with C p sol data in [15] showing that the relative deviation is always smaller than 4%.
If T of a stream is higher than T sat , part of the NH 3 in the solution will be boiled off.For this case, the total enthalpy can be expressed as, sol sat sol vap NH3 (13) where, h sat sol and h vap are the specific enthalpies for the saturated solution part and the vapor part, respectively.q is the quantity, which can be identified as,

Optimization problem
The properties of ILs can be adjusted by the design of anion and cation combinations for a task-specified purpose.However, because of the large number of anions and cations, the number of possible combinations is considerable.In this paper, the determination of screen criteria of task-specific ILs for AHPs will also be discussed.These criteria are identified via the optimization of the cycle performance.
In the optimization, the objective is the maximization of the COP which depends on solubility (concentration of weak and strong solutions) and enthalpy values as will be discussed in Section 3.5.Making use of an NRTL model, a total of eight parameters are identified which affects the attained COP value.They are α τ τ τ , , , (0) and τ 21 (1) of the NRTL model for VLE, c c , 0 1 in linear molar C p expression and the mo- lecular weight Mw.Once the COP reaches the optimal value, corresponding optimal variables can be determined for the optimum IL and mixture.Meanwhile, in order to obtain a practical and reasonable result, constraints of these optimal variables are needed.They are determined in terms of experimental data which are collected and discussed in Section 3.1.
The GA is better at finding global solutions than gradient-based solvers.GA selectively generates new candidate points to evaluate based upon a method that is similar to breeding between two "parents" to generate a "child".They are useful for problems that are highly nonlinear, such as the present problem.One of the concerns is the computational efficiency.To check a large amount of individuals from generation to generation is time-consuming.Luckily, the present optimization problem is not so CPU-intensive.In addition, the selection of individuals in the current generation is random, which can also lead to local minima.To overcome this drawback, instead of using GA for only one optimization, we try thousands of optimizations based on GA independently to remove local minima.The non-physical optimization results will be rejected.Finally, the optimum value for the objective function can be identified.The GA toolbox of Matlab has been used to identify the optimum combination of parameters.The effect of the settings for what concerns population size, elicit count, crossover fraction and generations has been investigated by studying the effect of variations and finally the settings proposed in the Matlab toolbox, respectively, 200, 10, 0.8 and 500, have been adopted.

Vapor liquid equilibria
With the experimental VLE data of binary NH 3 /ILs, the binary parameters, α τ τ τ , , , (0) and τ 21 (1) of the NRTL model (Eq.( 7) and ( 8)), can be correlated and will allow for the determination of the operating concentrations.The correlated results and accuracies are listed in Table 2.In this work, most of the data have been correlated with a rootmean-square deviation (RMSD) smaller than 5.62% as shown in Table 2.Only the model for the pair NH 3 /[omim][BF 4 ] showed slightly larger deviation: 8.7%.

Heat capacities
Experimental heat capacity (C p ) data of 61 ILs at 298.15 K, reviewed by Paulechka [46], are plotted in Fig. 4 in mole-based and mass-based units, respectively.It is quite interesting to see that the mole-based C p data are distributed in a linear trend with respect to the molecular weight.The mass-based C p data are centralized near 1.44 kJ − − kg K 1 1 in a nearly constant range between 1 and 2 kJ − − kg K 1 1 .This trend provides a general relationship between C p and Mw for ILs, which will be used in the property optimization as a constraint.Even though the relationship is not very accurate, it still can be helpful to identify how this property impacts on the performance of the mixture in the AHP.
For the same ILs involved in the NRTL correlation in Table 2, the mole-based C p values are also plotted as a function of temperature in    [47][48][49][50][51][52][53].The accuracies of all correlations are also listed.Notice that for most fluids the available data are limited to 100 °C, so that the relation is extrapolated when calculating values up to 130 °C.This may lead to larger errors than reported in Table 3.

Performance comparison
To compare the performances in detail, some calculated results are listed in Table 4 at a specific condition (T gen /T con /T abs /T eva = 120/45/ 45/10 °C).This operating conditions range is specified based on applications in floor heating.Apart from the solubility levels of the in-and outlet of the ABS and performance parameters f and COP, the conditions at the GEN inlet (state point 4) are also checked.Qualities q 4 are listed to show if there is vapor boiled-off before entering the GEN.The table also includes the results for the NH 3 /H 2 O system and optimum results obtained in Section 3.5.The single-effect AHP cycle with NH 3 / H 2 O as working fluids, reported by Kotenko [23], operates at similar conditions as applied in the current work.His predicted heating COP for a system with a rectifier and slightly lower evaporating temperature (T e = 5 °C) is around 1.59, which is quite close to the values obtained in the current work of 1.61.
Two facts resulting from the circulation ratio, f, can influence the cycle performance.One is its impact on the pumping power.A higher value of f means a larger mass flow rate through the pump (at the same flow of refrigerant stream), which can increase the power consumption of the solution pump.The second one is due to the relationship between f and the energy and mass balances for the GEN and for the ABS, which will be discussed in detail in Section 3.3.In all, a small f is preferable.Because of the significant difference in molecular weights between NH 3 and ILs, the mass concentrations of NH 3 /ILs pairs are much lower than that of the NH 3 /H 2 O system.As a result, the circulation ratios of NH 3 / ILs mixtures (29.2-112.3)are significantly higher than that of conventional NH 3  system could be improved by implementing advanced cycles, such as the generator/absorber heat exchanger (GAX) cycle, that would also increase the complexity and investment of the system.These promising results show the potential of NH 3 /ILs working pairs which can be executed with a simple cycle, making these pairs superior alternatives.
Previously Yokozeki and Shiflett [21,22] have compared the performance of some of the investigated mixtures in single-effect absorption refrigeration cycles.Their results are also included in Table 4 for reference.At their considered operations conditions, the NH 3 /H 2 O pair has been identified to perform better than the considered NH 3 /ILs pairs.This is different from the current work, in which the NH 3 /H 2 O pair is not identified as the superior one for a heat pump operation.Yokozeki and Shiflett [21,22] did not include the effect of the rectifier when calculating the performance of the NH 3 /H 2 O pair.The rectifier is essential to guarantee the purity of the produced refrigerant.This is the main reason why their COP of the NH 3 /H 2 O pair was overestimated.Besides, although the experimental VLE data of Yokozeki and Shiflett [21,22] have been used for these fluids in the present work, the NRTL parameters have been independently correlated.The fitted VLE behaviors are not identical but quite similar to the ones reported by Yokozeki and Shiflett [21,22].The difference in COP results for the NH 3 /IL pairs is mainly due to the method used to predict the specific heat of the working pairs.For example, Yokozeki and Shiflett [21,22] made use of the group contribution function method to predict these values, while in this study the reported experimental values have been used as discussed in Section 3.1.Problems encountered when using the group contribution function method for the prediction of the specific heat of ionic liquids have previously been reported by Cai et al. [20], as already  discussed in the introduction.

Contributions to the generation heat
Considering the energy balance of the 4 main devices, the COP of a heat pump system can also be expressed as, The EVA heat, Q eva , is equal for all fluids.Thus, the difference in COP results from the GEN heat, Q gen : A higher Q gen leads to a lower heat pump COP.
To make this clear, relevant expressions of the total enthalpies in Section 2.2 along with Eq. ( 4) and ( 14) are substituted into Eq.( 2).Taking into account the heat and mass balances of the GEN, the heat input in the GEN, Q gen , can be rewritten as, Q gen is split into 4 terms in this expression.The first term is the latent heat effect from the condensation of NH 3 vapor (while also includes the sensible heat associated with the superheated state, while very small).
The second one denotes the excess heat.The other two terms represent the contributions of sensible heat, in which, the term in the third line is the sensible heat change of the NH 3 component while the term in the last line is that of the IL component.
In order to analyze the performance, each of the above contributions to the GEN heat, for the operating condition T gen /T con /T abs / = T 120/45/45/10 eva °C, is depicted in Fig. 6 for all the NH 3 /ILs working pairs.
The values of T 4 are almost identical for all the cases, indicating that the latent heats per unit mass flow, − h h , are more or less the same for all cases.The difference in latent heat in Fig. 6 is due to the different mass flows, −fq 1 .Large values of f or q can lead to a low value of the latent heat contribution.The q has an obvious negative correlation with the latent heat contribution, which can be observed in Fig. 6, e.g.pairs with [omim][BF 4 ] and [bmim][PF 6 ].A higher q implies more vapor is boiled off before the flow enters the GEN.In this way, the heat duty of the GEN is reduced.The q value results mainly from the VLE properties of the NH 3 /ILs systems.
Again, due to the almost equal values of T 4 , the sensible heats per unit mass flow from NH 3 component are identical for all the mixtures.The difference in NH 3 sensible heats is mainly caused by the factor, 3 .Also the circulation ratio, f, has a stronger impact compared with ( ) . This is also true for the contribution of the sensible heat of the ILs: since the mass-based C p values of ILs and temperature differences between inlet and outlet T Δ are approximately the same, the sensible heat of the ILs per unit mass flow, C T Δ p , are similar for all the ILs.The difference in sensible heat contributions is mainly due to the required circulation ratio, f.Smaller f of pairs with [bmim][BF 4 ], [emim][Tf 2 N] and [emim][SCN] lead to smaller sensible heat contributions and correspondingly higher COPs.This indicates that the circulation ratio f is dominant in this case.The solubility difference at the in-and outlet of the ABS will determine the f values.
Since the VLE properties are usually studied with mole-based units, when considering mass-based properties, say f here, Mw also plays a role.For the NH 3 absorption system, a larger molecular weight of the absorbent leads to a smaller mass concentration change when the molar change is maintained.This also implies that smaller molecular weight of the absorbent is preferable in terms of performance.
The contribution of excess heat will be discussed in Section 3.4.

Influence of heat source temperature on the performance
To investigate the influence of heat source temperature on the performance, condensing temperature T con , absorbing temperature T abs and evaporating temperature T eva are set to be 45 °C, 45 °C and 10 °C, respectively, while the temperature of the heat source, T gen , varies in a range from 100 °C to 130 °C.Because all the experimental VLE conditions are lower than 130 °C, T gen is maintained below 130 °C in all calculations.

Working pairs
Heat pump performance * Refrigeration performance  Under these operating conditions, the circulation ratio f of the NH 3 / [hmim][BF 4 ] working pair always has a negative value, what means this pair cannot operate in an AHP which operates under the imposed conditions.Thus, Figs. 7 and 8 only show the circulation ratio f and COP variation, respectively, for the other 8 NH 3 /ILs pairs and NH 3 /H 2 O pair.All the working pairs, considered for the proposed AHP system, show similar trends.With increase of T f , gen first decreases rapidly and then reduces to a more constant value.On the contrary, the COPs increase first sharply and then rise to a relatively constant value.At the higher temperature range, the values of f and COP are quite close for most NH 3 /ILs pairs.f falls in the range of 20-60 and COPs reach 1.4-1.8.It looks like these trends will be maintained when T gen increases above 130 °C.

The performances of working pairs with [mmim][DMP], [bmim] [BF 4 ], [emim][Tf 2 N] and [emim]
[SCN] are quite outstanding just as shown in Section 3.2 at a constant condition.Even though the fs of these points are larger than that of NH 3 /H 2 O, when T gen is high enough, the COPs are still higher.In addition, the NH 3 /[mmim][DMP] pair has the highest COP.
As for the NH 3 /H 2 O pair, it holds the lowest f and relatively high COP values when compared with the NH 3 /ILs pairs.With an increase of T gen , the COP slightly rises, and after 105 °C, it becomes more or less constant.
Because an accurate prediction of the excess enthalpy is difficult as discussed in Section 1, and to assess the sensitivity of excess enthalpy on the performance, COPs calculated based on ideal solutions (without taking account the excess enthalpy) are presented in Fig. 9.
In these cases, the range of the COP is smaller than when the excess enthalpy is taken into account.Even though the COP values are now lower, the working pairs, which show a better performance than that of NH 3 /H 2 O pair in Fig. 8, still perform better when the excess enthalpy contribution is neglected.An accurate measurement is still needed for a better assessment of the excess enthalpy of NH 3 /IL pairs in AHPs.

Optimum performance and corresponding properties
In the optimization work, the molecular weight is assumed in a range 170-400 kg/kmol.Taking the previous correlation results into account, the upper and lower limits of the thermodynamic model parameters are summarized in Table 5.Based on these, and the linear relationships of C p with both molecular weight and temperature as discussed in Section 3.1, the constraints of the search domain have been defined.
The property optimization work is conducted under the same operating conditions as discussed in Section 3.2, i.e.T gen /T con /T abs / = T 120/45/45/10 eva °C, for the application of building's floor heating.
Using the GA method, the optimized performance and corresponding optimum variables are obtained and listed in Table 6.The maximum COP under above constraints could reach 1.836 and the circulation ratio 1.102.
The P-T-x diagram of the optimum IL/NH 3 mixture described by the parameters of the NRTL model is shown in Fig. 10.As a comparison, the P-T-x diagram of NH 3 /[emim][SCN] is also plotted.Generally, the vapor pressure of the optimum pair has a large deviation between the low and high temperature range.For the optimum working pair, there is a negative deviation effect from the Raoult's law at the low temperature range while a positive deviation applies at the high temperature range.This absorption capability difference causes a large difference of NH 3 concentration between in and outlet of the ABS, what will lead to a smaller f, and a higher COP.
The optimum molecular weight Mw is exactly its lowest limit, 170.For the same molar solubility difference, a lower value of Mw will lead to a higher value of mass solubility difference which will then lead to a lower f and a higher COP.Besides, since the molecular weight Mw has a linear ascending relationship with mole-based C p , correspondingly, a lower mole-based C p is observed for the optimum IL, which is depicted in the lowest position in Fig. 5.
The optimum properties including vapor pressure, C p and Mw allow us to screen the ideal ILs for AHP cycles.The challenge for future work is identifying ILs which show properties close to the Mw C , p and vapor pressure of the optimized ideal mixture.It is clear that a low molecular weight, low C p and large concentration difference between in-and outlet of the ABS are essential requirements.

Outlook of economic and technical feasibilities
In an economic analysis, the COP values are related to the operational costs of the AHP while the capital costs are related to the   7.
Boman et al. [37] have recently shown that the IL-based AHP systems need more heat exchanger area due to the poor heat transfer performance of the ILs, caused by their higher viscosity, lower thermal conductivity and heat capacity.In this section, PHXs are considered in the IL based absorption systems in the roles of GEN, ABS, SHX, CON and EVA.The PHX is selected mainly due to its compact size and good performance of heat and mass transfer.The compact design of PHXs keeps the system volume small so that a smaller amount of expensive working fluid is sufficient to fill the system.

Equipment sizing
The floor heating system for a building in a moderate climate area, for example, the Netherlands, is taken for the economic comparison of the AHPs.The heating load will generally not exceed 60 W/m 2 [59].A 3750 m 2 building will have a heating capacity of 225 kW and its yearly heating requirement will be, approximately, 337.5 MW h.
PHX with plate sizes 0.191 × 0.618 m (width × length) and plate spacing of 1.5 mm have been considered for all heat exchangers except for the SHX.The dimensions of the SHX were 0.390 × 0.990 m with an identical plate spacing.
The overall heat transfer coefficient for each heat exchanger has been obtained by summing the heat transfer resistances on both sides of the heat exchangers.The local heat transfer coefficient for single phase flows (both of external fluid and working fluid) has been predicted making use of the correlation proposed by Yan et al. [60].Similarly, for the evaporation processes the correlation proposed by Khan et al. [61] and for the condensation processes the correlation proposed by Thonon and Bontemps [62] have been used.For the GEN and ABS the smallest value obtained considering single phase flow or, respectively, evaporation and condensation have been adopted so that mass transfer resistance is taken into account.The required heat exchanger area follows from the heat exchanger duty, its overall heat transfer coefficient and the logarithmic mean temperature difference between the two fluids.
The properties of the external fluid (water) and of NH 3 have been obtained from Lemmon et al. [40].The properties of [emim][SCN] have been obtained from Freire et al. [57] (density and dynamic viscosity), Navarro et al. [53] (heat capacity) and Tenney et al. [63] (thermal conductivity).Table 8 shows the calculated areas for the different heat exchangers.For the SHX of the NH 3 /IL system two options have been considered: a minimum temperature approach of 5 K (option 1) and a minimum temperature approach of 16 K (option 2).In this case the COP drops from 1.73 (option 1) to 1.54 (option 2) while the required PHX area significantly reduces.

Economic analysis
For the economic calculation, a cost equation based on DACE [64] PHX costs has been applied to the areas reported in Table 8 with C phx expressed in k€ and A expressed in m 2 , for the SS316 PHX in the area range of 40 to 300 m 2 .The cost of SS316 shell-and-tube HXs is taken from the same source for the area range of 30 to 200 m 2 .The yearly HX costs take into account a (linear) depreciation time of 15 years.The price of natural gas for households and commercial consumers has been taken as 55 €/MW h [65].A boiler efficiency of 85% has been adopted, which has also been taken into account for the other AHP systems.The price of IL varies significantly depending both on production amounts and manufacturing technique.The prices of the [emim][SCN] listed in Table 7 are based on quotations in a lab scale production.It is also reported that, in large scale production, some ILs cost will reduce to 3.00 $/kg [66].Honeywell UOP [67] reports the use of ILs to produce highoctane motor fuels and claims it is a "cost-effective solution".This indicates that the application of ILs at an industrial scale does make their price economically more competitive.The values from the current vendors and the expected low value (3.00 $/kg) have been adopted in the present calculations.Table 9 shows the results of the economic comparison between the different AHP solutions and the use of a conventional boiler.All considered AHP options lead to both energy savings and costs savings, when the cost of IL is at a reasonable level (industrial scale production).The influence of the IL price is also shown.The NH 3 /IL AHP with a large minimum temperature approach in the SHX performs economically the best for current IL prices.
3.6.3.Other technical concerns STABILITY Most ILs have been reported as being stable as liquids over a very wide temperature range.It has also been reported for some NH 3 /IL mixtures that chemical reactions take place [24].Chemical reactions make the cycles less reversible.The long term operation of these fluids might be a concern but NH 3 /IL mixtures which do not undergo chemical reactions are expected to be capable of realizing a large number of operational cycles in a reliable way.
VOLATILITY There have also been concerns about the negligible vapor pressure of ionic liquids [68].Although being small, very small concentrations of IL vapor may leave the generator and enter the condenser.After a large number of cycles, IL may accumulate in the evaporator requiring additional actions to bring it back to the absorbent loop.Since no long term operation with these cycles has been reported, the practical performance of these mixtures still needs to be confirmed.

Conclusion
After a review of methods and a summary of available experimental properties, a thermodynamic model has been proposed to investigate single-effect AHPs with NH 3 /ILs working pairs as working fluids for the purpose of the floor heating of buildings.With this model, the performance of the AHPs has been calculated for all 8 feasible NH 3 /ILs pairs (one additional pair cannot operate under the considered conditions) and also for NH 3 /H 2 O. Additionally, a properties optimization work and economic analysis have been executed.Based on the work, the following conclusions could be drawn: • The circulation ratio f decreases and COP increases with an increase of the generator temperature (up to 130 °C).
• Under the considered conditions, the COP of the NH 3 /[mmim] [DMP] pair reaches the best performance (COP of 1.79), and along with NH 3 /[bmim][BF 4 ], NH 3 /[emim][Tf 2 N], NH 3 /[emim][SCN] all showing a higher COP than that of the NH 3 /H 2 O pair.Nevertheless, the circulation ratio f is significantly higher than that for the NH 3 / H 2 O pair.
• The analysis of the generator heat requirement revealed that, high vapor quality values at the inlet of the GEN resulted in a high COP because of a lower latent heat contribution.The influence of circulation ratio, f, is mainly associated with the two sensible contributions (the sensible heat of both components), and a low f would lead to a high COP.Neglecting the excess enthalpy, the performance changes, but the better working pairs still beat NH 3 /H 2 O pair in terms of COP.
• The optimum COP of this type of working pairs and for the condition considered can be expected to reach 1.84 as demonstrated by the property-optimization study.
• To realize an ideal performance, the optimum IL candidates should show high absorption capabilities, large solubility differences between in-and outlet of the generator, low molecular weights and low heat capacities.The optimization study shows its potential to assist in the selection of IL as absorbents.
• Large circulation ratio combined with worse heat transfer perfor- mance in the solution heat exchanger lead to large demand of heat transfer area, which additionally requires a large amount of expensive ILs.
• An economic feasibility analysis indicates that, when the [emim]   [SCN] would be produced at industrial scales, this NH 3 /IL AHPs lead to both significant energy (42%) and economic (29%) savings.

Fig. 3 .
Fig. 3. Two constructions of the generator.(a) The generator for working pairs using absorbents with a negligible vapor pressure at operating temperature, say NH 3 /ILs in this study.(b) The generator with purification section above it (distillation column and rectifier) for working pairs using absorbents with a non-negligible vapor pressure at operating temperature, say NH 3 /H 2 O in this study.

Fig. 4 .Fig. 5 .
Fig. 4. C p values of 61 ILs at 298.15 K (upper: mole-based, lower: mass-based).These 61 data points have been collected from experimental C p data reviewed by Paulechka [46].They are selected because they are the only ones having C p data at 298.15 K with un- certainties lower than 15%.

Fig. 7 .
Fig. 7. Circulation ratios, fs, with respect to generation temperature for the NH 3 /IL based working pairs in the single-effect AHP cycles at the condition of = T T T / / 45/45/10 con abs eva

Table 1
Molecular weight and critical data of the investigated ILs.
equation of state (EOS) and mixing rules.Yokozeki and Shiflett

Table 2
Correlated binary parameters in the NRTL model for the investigated working pairs.
Table 3 lists the correlated parameters of the linear expression and is based on the experimental C p data from literature

Table 3
Correlated parameters in C p Concentration data are all for the NH 3 component.The subscript 2 and 7 represent outlet conditions of the ABS and the GEN, respectively, which are locations with strong and weak solution flows.The subscript 4 represents the inlet conditions of the GEN.
To determine the sizes of the main components, which are the heat exchangers, the duty and overall heat transfer coefficient of each main heat exchanger are first estimated taking the local flow and fluid properties into account.A conventional NH 3 /H 2 O AHP with shell-and-tube HXs is here compared with an NH 3 /IL AHP.IL [emim][SCN] is selected as absorbent for the NH 3 /IL AHP since it is one of the best performing ILs and because its viscosity and current price are the lowest among the studied ILs as shown in Table

Table 5
Limits of optimization variables and properties in the optimization problem.

Table 6
Optimized performance & optimum variables of the working fluids in the single-effect AHP at a condition of T gen /T con /T abs /Teva = 120/45/45/10 °C.

Table 7
Viscosities and prices of the most promising commercialized ILs for single-effect AHPs in this study.

Table 8
Equipment size for both AHP systems, based on the same conditions applied in Table4.

Table 9
Yearly energy requirements and yearly capital costs of the different AHP systems in comparison to a conventional boiler.