Water sorption behavior of physically and chemically activated monolithic nitrogen doped carbon for adsorption cooling

Abstract

In this work, nitrogen doped resorcinol–melamine–formaldehyde (RMF) resins were synthesized, pyrolyzed and physically activated with CO₂. The influence of the activation time on the physicochemical properties and the water sorption behavior produced in this way was investigated. Furthermore, a comparison between physical activation with CO₂ and chemical activation with KOH is presented. Materials performance was validated in an adsorption chiller test setup with a temperature step from 90 °C → 50 °C. The CO₂ activated RMF carbon exhibits a maximal specific cooling power which is a factor of 1.7 higher in comparison to a commonly used, commercial silica gel reference material (430 W kg⁻¹ compared to 255 W kg⁻¹). This is surprising considering that the hydrophilicity of the CO₂ activated carbon is rather low. The superior performance of carbon based sorbents is attributed to originate from the superior thermal transport properties of monolithic carbons over commercial silica gels. At a more feasible temperature swing 60 °C → 30 °C, the RMF derived carbon yields a specific cooling power 3.2 times greater than that of the silica gel reference.

---

Introduction

Adsorption heat pumps (AHP) have recently gained growing interest by engineers and planners as they are considered an environmentally-friendly and cost-effective alternative to compressor-based heat pumps. In contrast to conventional vapor compression technology, where a mechanical compressor drives the refrigeration cycle, AHPs take advantage of the adsorption and desorption of a vapor onto highly porous substrates to provide heating. Cooling is obtained from the heat of evaporation which is continuously removed from the boiling adsorbate liquid. The fundamental limitations in today's AHP systems are primarily due to low thermal conductivity and poor mass transport properties of the solid adsorbent.¹ To address these issues, monolithic adsorbents have been proposed to enhance thermal transport properties.^2,3 Especially carbon based adsorbents have been investigated more recently as a new class of promising candidates.^4–6 An attractive method to produce carbon monoliths with different geometries is by mold-casting of organic precursor resins. However, commonly used resins such as resorcinol–formaldehyde, resorcinol–melamine–formaldehyde or resorcinol–urea–formaldehyde typically have long gelation times of up to several days.^7,8 Generally, longer gelation times lead to smaller particle-network type mesoporous structures, whereas fast gelation leads to a macroporous network. However, large adsorbate uptake capacities are primarily due to the gas adsorption in the micropores. In order to increase the microporosity responsible for high sorption capacities (preferably using water as an adsorbate) carbon monoliths must be physically or chemically activated. Alkali metal hydroxides are generally used as chemical activators to create microporosity and high surface areas in carbons derived from phenol–formaldehyde resins.⁹

In our previous research,³ pyrolyzed carbons derived from RMF resins were chemically activated by KOH. This treatment allows one to increase the micropore volume and the water sorption capacity of the activated carbon. However at the same time, the concentration of hydrophilic surface functional groups is tremendously decreased. Furthermore, the KOH activation is a very time consuming process and some remaining alkali metal hydroxides might cause corrosion inside a real AHP device if not properly rinsed before use. From this point of view, a further improvement in the activation procedure to increase the water vapor sorption capacity of the activated carbon is required hopefully leading to a faster and more straightforward practical preparation of carbon sorbents for AHPs. So far, many studies have been documented in the literature, comparing both chemical and physical activation treatments of various carbons. Teng et al.¹⁰ prepared porous carbons from phenol–formaldehyde resins with chemical and physical activation. It was observed that both activation methods can produce carbons with surface areas and pore volumes greater than 2000 m² g⁻¹ and 1.0 cm³ g⁻¹, respectively. Furthermore, carbons prepared from CO₂ activation feature a more compact surface than those produced form KOH activation. Contreras et al.¹¹ investigated the influence of chemical and physical activation of carbon xerogels prepared from resorcinol–formaldehyde resins. Chemical activation by KOH leads to the formation of narrow microporosity whereas physical activation by CO₂ creates a widening of the narrow micropores regardless of the process conditions. Moreover, carbons prepared by chemical activation have a four times higher concentration of oxygen surface functional groups compared to carbons prepared by physical activation.

In this work, we report for the first time the application of CO₂ activated monolithic carbons derived from nitrogen-doped RMF resins for adsorption cooling using water as refrigerant. The influence of physical activation by CO₂ on carbons derived from RMF resins is investigated. Additionally, a comparison between physical activation with CO₂ and chemical activation with KOH is also performed. The effect of the activation time on the surface chemistry, surface area and the water sorption behavior were examined. Best candidate materials were benchmarked against commercially available silica gel in a miniaturized AHP test stand.

Experimental

Materials

Formaldehyde was obtained from Sigma Aldrich as an aqueous solution (37 wt%; methanol stabilized). Resorcinol (≥99%), melamine (≥98%), sodium hydroxide (≥98%) and hydrochloric acid (37 wt%) were received from Sigma Aldrich. All solutions were prepared from ultrapure water (double distilled quality, >18 MΩ cm). Commercial silica gel beads (RD-type, bead size 1.8–2.0 mm, Fuji Silysia Chemical Ltd.) were used as reference adsorbents.

Preparation of the RMF resin

A typical procedure of RMF sol preparation is as follows. 20 ml distilled water and 15 ml formaldehyde were mixed in a beaker. The solution was stirred on a heating plate before 4.0 g melamine was added. During five minutes of stirring, the whitish opaque suspension was warmed up to 55 °C. Subsequently, 10 ml of 0.5 M sodium hydroxide solution were added under vigorous stirring. The suspension was stirred until a colorless clear solution was obtained. Then 3 ml concentrated hydrochloric acid and 7.0 g resorcinol was added to the solution which was kept stirring for around one minute until the resorcinol was fully dissolved. This RMF sol was poured into plastic cylinder molds which were placed in an oven at 55 °C for 24 hours. On top of the gel, a small amount of syneresis liquid was present. The gels appeared darker at the gel–air interface as a result of slight oxidation and the formation of a densified highly reflective skin.¹² The RMF resins were then taken out of the mold and dried in an oven at 55 °C for 24 hours.

Pyrolysis

The RMF resins were pyrolyzed in a tube furnace (Carbolite, STF 16/50 818P). The resins were placed in a quartz boat which was inserted into the center of a ceramic tube (Pythagoras, Haldenwanger MTC). The ceramic tube was purged for 2 hours with high purity nitrogen (99.995%) before heating to 900 °C at a rate of 5 °C min⁻¹. The dwell time at the pyrolysis temperature was 1 hour after which the setup was cooled down to room temperature at a rate of 5 °C min⁻¹. The nitrogen flow rate during the entire pyrolysis was 22.2 STP liters per hour.

Physical and chemical activation

The physical activation heat treatment of the pyrolyzed resins was conducted according to the following procedure: carbon monoliths were placed in a quartz boat and dried at 110 °C for 2 hours followed by dehydration at 250 °C for another 2 hours. Subsequently, the samples were activated at 800 °C for different activation times before the temperature was decreased again to room temperature. The heating/cooling rate was kept the same at 5 °C min⁻¹. The carbon dioxide (99.99%) flow rate during the entire activation process was 13.9 STP liters per hour.

For chemical activation with KOH, a procedure leading to the highest water uptake based on our previous study was selected as a reference chemical activation scenario.³ 2 g of potassium hydroxide and 10 ml of distilled water were dissolved in a beaker before 2 g of sample A-p900 were added. The samples were immersed with the KOH solution stirred at 55 °C and 100 rpm for 1 hour before they were removed from the beaker and dried in an oven at 55 °C.

KOH impregnated samples were then placed inside an alumina crucible and arranged in a quartz boat. The activation was started by first drying the wet monoliths at 110 °C for 2 hours followed by a dehydration at 400 °C for another 2 hours. Subsequently, the oven was heated to 800 °C for 1 hour for activation before the temperature was decreased again slowly to room temperature. The heating/cooling rates were kept the same at 5 °C min⁻¹. The nitrogen flow rate during the activation treatment was kept at 22.2 STP liters per hour. In order to fully remove the remaining KOH and salts formed during the activation, all samples were soaked in distilled water for a week. Each day, the aqueous wash solution was replaced with fresh distilled in order to remove all traces of KOH until the pH-value of the final wash solution was close to 7. The sample history and designation is presented below in Table 1.

Table 1 Table of experiments

Sample	Pyrolysis temperature [°C]	Activation temperature [°C]	Activation time [hours]
A
A-p900	900
A-p900-CO2-0.5 h	900	800	0.5
A-p900-CO2-1 h	900	800	1
A-p900-CO2-2 h	900	800	2
A-p900-CO2-3 h	900	800	3
A-p900-KOH-1 h	900	800	1

---

Characterization

X-ray photoelectron spectroscopy (XPS). For X-ray photoelectron spectroscopy (XPS) measurements, specimens were prepared by scratching a groove in the specimens before drying them inside a vacuum oven at 80 °C for 12 h. Immediately before mounting the specimen onto the sample holder and subsequent pump-down overnight, each sample was fractured in ambient air to generate a fresh surface with minimal contamination. XPS spectra were acquired on a Physical Electronics (PHI) Quantum 2000 photoelectron spectrometer using a monochromatic Al Kα source (hν = 1486.6 eV) and a hemispherical electron-energy analyzer equipped with a channel plate and a position-sensitive detector. Atomic concentrations were obtained from the different peak areas after Shirley background subtraction using the built in PHI sensitivity factors. The electron take-off angle was 45° and the analyzer was operated in the constant pass energy mode. The pass energy used for the detail spectra of the C 1s, O 1s and N 1s core levels was 46.95 eV to yield a total analyzer energy resolution of 0.95 eV (for Ag 3d electrons). The spectrometer was calibrated for the Au 4f_7/2 signal to be at 84.0 ± 0.1 eV and had a resolution step width of 0.2 eV. The area analyzed was typically 100 μm in diameter. Partial compensation of surface charging during data acquisition was obtained by simultaneous operation of electron- and an argon ion-neutralizer guns. Carbon samples were measured “as is” without any prior surface cleaning and the vacuum in the main chamber during analysis was 6.6 × 10⁻⁹ Torr. The N 1s envelopes were curve-fitted by mixed Gaussian–Lorentzian component profiles using the Multipak software from Ulvac-Phi Inc. The background subtraction was done by the Shirley baseline.

Wide angle X-ray diffraction (WAXD). Wide angle X-ray diffraction (WAXD) and small angle X-ray scattering (SAXS) measurements were carried out with a PANalytical X'Pert PRO system equipped with a Johansson monochromator (Cu K_α1, λ = 1.5406 Å) in Bragg–Brentano geometry. The step integration time was 424 seconds and 2θ was varied from 5° to 90° with a step size of 0.0167° for WAXD. For SAXS measurements, the step integration time was 700 seconds while 2θ was varied from 1° to 8° with a step size of 0.0167°. The diffracted/scattered X-rays were detected with an X'Celerator linear detector.

High-resolution transmission electron microscopy (HRTEM) images were recorded with a JEOL 2200FS TEM/STEM microscope at an accelerating voltage of 200 kV. Prior to the characterization, the carbon samples were ground and dispersed in methanol, before loading the sample onto a lacey carbon TEM grid.

N₂ adsorption and desorption. N₂ adsorption and desorption isotherms at 77 K were collected on a Micromeritics ASAP 2020 Surface Area and Porosity Analyzer. Prior to experiments, approximately 200 mg small carbon monolith pieces were degassed at 250 °C for 4.5 hours at a pressure of 1.3 × 10⁻² mbar. The total surface area (S_BET) of the materials was calculated by the Brunauer–Emmett–Teller (BET) method.¹⁴ The micropore volume (V_micro) was determined with Dubinin–Raduskhevich (DR) equation. The mesopore volume (V_Meso) was calculated by substracting the value of V_Micro from the amount of adsorbed at P/P₀ = 0.95.¹³ The average pore diameter (D_pore) was calculated by D_pore = 4V_pore/S_BET.

Dynamic vapor sorption. Dynamic vapor sorption (DVS) isotherms were recorded using an isothermal gravimetric DVS apparatus (TA Instruments VTI-SA+). A high-precision microbalance of accuracy 10 μg allows continuous monitoring of the amount of adsorbed water vapor. 9–11 mg of the sample material (granulate shape) was placed in the experiment chamber. In a first step, the sample was heated to 90 °C under nitrogen atmosphere during 1 hour and the resulting mass was taken as the dry reference mass. The sample was then cooled to 50 °C for the subsequent measurement. Adsorption/desorption cycles were performed at a constant temperature of 50 °C. The material is exposed to relative pressure (P/P_sat) steps between 0 and 0.6 in 0.05 intervals and between 0.6 and 0.8 in 0.1 intervals in a controlled flow of a mixture of nitrogen and water vapor while the sample mass is recorded every 60 seconds. The saturation pressure of water at 50 °C is 124 mbar (P/P_sat = 1). For each step, the pressure was kept constant until the rate of mass change became negligible (less than 0.08%/5 minutes) and the corresponding pressure and mass values were recorded until the entire adsorption and desorption isotherms were completed.

Skeleton density. The skeleton density (ρ_skeleton) was determined with an AccuPyc II 1340 helium pycnometer (Micromeritics, helium purity of 99.999%). Prior to the density measurements, the carbons were dried inside a vacuum oven at 80 °C for 12 h.

Thermal conductivity measurements using the laser-flash-apparatus (LFA). In order to determine the thermal conductivity (κ(T) = ρ(T)α(T)C_p(T)), the diffusivity α and the heat capacity C_p were measured via LFA (NETZSCH LFA 457). The envelope density (ρ) was determined from the geometry and the weight of the LFA samples (disks with a diameter of 10 mm and a thickness of 1 mm), which were cut from carbon monoliths using a diamond wire saw. The samples were coated with graphite spray in order to compensate for the surface roughness of the porous samples.¹⁵ For each sample six measurements were performed in vacuum atmosphere at a temperature of 50 °C. In between the measurements a relaxation time of five minutes was allowed for the sample to equilibrate. C_p values were determined by means of a comparative measurement using a Pyroceram 9609 reference sample. The uncertainty of the C_p measurement is in the range of 5%.

Electrical conductivity measurements. Electrical conductivity measurements were carried out at 50 °C in air by using an RZ200li unit (Ozawa Science). The carbon monoliths were cut into cuboids with dimensions of 2 mm × 2 mm × 12 mm.

Temperature swing adsorption (TSA). Temperature swing adsorption (TSA) tests were conducted on an adsorption test rig equipped with an evaporator, a flow sensor calibrated for water vapor, a thermal imaging camera (FLIR SC5000, resolution: 320 × 256 pixel) and a vacuum chamber. The carbon monoliths were cut into cylinders with a diameter of approximately 10 mm and a height of 5 mm. An epoxy resin (Araldit Rapid) was used to glue individual cylinders between two aluminum lamellae which were pressed onto a copper tube to form an adsorber heat exchanger. A reference measurement was carried out by filling silica beads between two aluminum lamellae separated by 4 mm and wrapping the lamellae with a light-weight steel wire mesh to fix the beads. The temperature of the heat transfer fluid was passing through the copper tube of the heat exchanger was controlled by two hydraulic circuits coupled to the adsorber heat exchanger by a 6/2-way-valve which was connected to two thermal baths with different temperatures. The evaporator temperature was kept at 20 °C for all measurements. Prior to the experiment, the test chamber was evacuated to a pressure below 1 mbar and the adsorber heat exchanger was heated to the maximum experimental temperature (90 °C or 60 °C) for drying of the material.

Results and discussion

In the following, we shall discuss the change in chemical, morphological and water uptake related properties as a function of the pyrolysis treatment. This allows for an optimized activation procedure to be selected and evaluated in terms of application relevant criteria.

Chemical identity

It is expected that the type and atomic concentration of the chemical species on the surface must have a significant influence on the water sorption behavior.¹⁶ XPS reveal that the surface atomic composition of the dried resin contains approximately one sixth of nitrogen and oxygen species each (Table 2) with the majority of two thirds being carbon. This means that every third atom in the depth analyzed (ca. 3 nm) is a heteroatom (not taking into account hydrogen atoms). Following pyrolysis at 900 °C, the concentration of oxygen and especially nitrogen surface functional groups is reduced. Chemical activation causes a decrease of these two types of chemical species. Moreno-Castilla et al.¹⁷ have also observed a decrease of the nitrogen and oxygen surface signatures during chemical activation, which is consistent with our present data. Physical activation with CO₂ leads to a slight decrease of nitrogen containing surface functional groups whereas the concentration of oxygen containing species is increasing. Molina-Sabio et al.¹⁸ reported similar results for the CO₂ activation of olive stones: the concentration of =CO and –CO₂ surface groups increases with longer activation time. One possible explanation might be the reaction mechanisms which are different for chemical and physical activation. For chemical activation with KOH, Otowa et al.¹⁹ suggests that potassium oxide is formed by dehydration of KOH. CO₂ which is formed by a combined water–gas/water gas shift reaction further forms potassium carbonate in the presence of potassium oxide (eqn (1) through (4)).

2KOH → K₂O + H₂O (dehydration) (1)

C + H₂O → CO + H₂ (water–gas reaction) (2)

CO + H₂O → CO₂ + H₂ (water–gas shift reaction) (3)

CO₂ + K₂O → K₂CO₃ (carbonate formation) (4)

Table 2 XPS results of the dried resin, the pyrolyzed and the activated carbons

Sample	C [at%]	N [at%]	O [at%]
A	65.54	17.25	17.20
A-p900	89.65	4.10	6.25
A-p900-CO2-0.5 h	91.40	4.06	4.54
A-p900-CO2-1 h	88.83	4.06	7.11
A-p900-CO2-2 h	87.93	3.92	8.16
A-p900-CO2-3 h	89.04	3.22	7.74
A-p900-1 : 1-KOH-1 h	96.13	0.44	3.43

---

For physical activation with CO₂, Marsh et al.²⁰ proposed that a free carbon active site (C_f) reacts with carbon dioxide and forms carbon monoxide and an oxygen surface complex (C(O), eqn (5)).

C_f + CO₂ → CO + C(O) (5)

Fig. 1 shows the N 1s XPS spectra of the samples before and after pyrolysis and activation. Insert shows a proposed structure of the carbon after pyrolysis with nitrogen atoms in different chemical environments. In the dried resin, the nitrogen peak shows one very broad main peak with a center of peak binding energy of 398.5 eV, which is attributed to a pyridine-type N with the nitrogen atom in a six-membered ring contributing with one p-electron to the aromatic π-system around 398.3 eV.^8,21 This pyridine type signature further overlaps with amine-type nitrogen species (C–NH₂) at 399.4 eV which are present in the melamine moieties.²¹ After pyrolysis at 900 °C, the nitrogen peak shows four chemical states present: the pyridinic-N (N-6), pyrrolic and/or pyridine-type nitrogen (N-5), quaternary nitrogen (N-Q) and pyridine nitrogen-oxide (N-X). The N 1s spectra has, analogous to references,^22,23 been separated into four different chemical states with binding energies of ≈398.9 (N-6), 400.5 (N-5), 401.4 (N-Q) and 403.2 eV (N-X). For the curve fitting, the relative distance of these four peaks had been kept constant in order to compensate for possible sample charging during analysis as well as for differences in spectrometer energy scale calibration. Only the pyrolyzed and activated carbons have been fitted, due to the fact that for the resin other chemical groups (e.g. amine) would have to be considered as well. During the physical activation with CO₂, the relative concentration of the chemical states is hardly influenced. However, chemical activation with KOH seems to decrease the concentration of pyrrolic and/or pyridone groups significantly.

Fig. 1 N 1s XPS spectra of the samples before and after pyrolysis and activation. The intensity of sample A is divided by a factor of 6. Insert shows a proposed structure of the carbon after pyrolysis.

Microstructural features

Fig. 2a summarizes the X-ray diffractograms of pyrolyzed and activated carbons for different periods. Residual sodium chloride which is present in the resin gives rise to very sharp lines and is marked with asterisks. Activation by CO₂ decreases the intensity of the (002) graphitic reflection. This is an indication for a slight destruction of the stacking order of the graphene sheets during physical activation. The scattering at low angles rises with longer activation time which depicts the evolution of porosity. Using the Scherrer equation, the domain size was calculated in dependence of the activation time which is shown in Table 3. The crystallite height L_c was derived from the (002) reflection at 24.5° while the crystallite length L_a was derived from the (100) reflection at 44°. The domains sizes are slightly increasing with longer activation times. No pronounced change in the ratio of aromatic to linear sp²-hybridized carbon could be observed by Raman spectroscopy (see ESI S2†). The Bragg equation, nλ = 2d sin θ, was used to calculate the distance d between two graphene planes for the first order diffraction peak (n = 1) based on the (002) reflection (Table 3), resulting in a typical spacing of 3.7 Å. This value is typical for disordered carbons, compared to the interplanar spacing (3.35 Å) for crystalline graphite.²⁴ Fig. 2b shows the small angle X-ray diffractograms of pyrolyzed and activated carbons with different activation times. For both, pyrolyzed and activated carbons, a broad “hump” is observed in the small angle scattering range (1° to 5°) which is more pronounced at longer activation times. At longer activation times, this signature is shifted slightly to smaller angles, which is an indication for the formation of additional micropores.²⁵ Assuming that the position of this peak maximum is linked to the pore size, we can obtain an estimate of the average pore size by using the Bragg-equation. Accordingly, during physical activation with CO₂, the average pore size seems to increase from 3.1 nm to 3.9 nm. A study by Teng et al.¹⁰ shows a similar behavior for physical activation of carbons derived from phenol–formaldehyde resins. A higher degree of carbon removal (burn-off) leads to a widening of the micropores.

Fig. 2 (a) Wide angle and (b) small angle X-ray diffractograms of the resin, the pyrolyzed and activated carbon. Diffraction peaks attributable to NaCl are marked with asterisks.

Table 3 Domain size and interplanar distance in dependence of the activation time

Sample	Lc [Å]	La [Å]	d002 [Å]
A-p900	9.4	14.7	3.7
A-p900-CO2-0.5 h	9.7	15.0	3.7
A-p900-CO2-1 h	9.7	15.8	3.7
A-p900-CO2-2 h	9.9	15.9	3.7
A-p900-CO2-3 h	10.2	17.0	3.7
A-p900-1 : 1-KOH-1 h	8.4	16.9	3.7

---

Fig. 3 shows high resolution transmission electron micrographs of sample A-p900 and A-p900-CO₂-3 h. The microstructure reveals short, linear features which might be identified as individual graphene sheets viewed edge-on. No long-range order can be discerned, however, there is local ordering in some regions involving 3–4 stacked graphene layers which is comparable to that of other forms of amorphous carbon.²⁶ The alignment of the graphene layers does not seem to be influenced by physical activation with CO₂. The micrograph confirm the findings of the XRD measurements: the calculated domain size along (002) lies between 8.4 and 10.2 Å which correlates approximately to the distance between 3 graphene layers. In addition, the lateral extension of the ordered domains was estimated to be between 14.7 and 17.0 Å from XRD (Table 3), which is in agreement with the TEM observations (Fig. 3).

Fig. 3 HRTEM micrographs of the sample (a) A-p900 and (b) A-p900-CO₂-3 h.

Fig. 4 illustrates the N₂ adsorption and desorption isotherms of the carbons activated for different periods of time. It is rather apparent that during physical activation with CO₂, the nitrogen sorption capacity increases with longer activation time. The N₂ isotherm of A-p900-CO₂-3 h shows a high nitrogen uptake at low relative pressures (P/P₀ < 0.1) and a plateau at high relative pressures which is typical for microporous materials (type I according to IUPAC classification^27,28). The influence of the activation time on the BET surface area is shown in Table 4. The resin pyrolyzed at 900 °C has a BET surface area of 410 m² g⁻¹ and a high fraction of micropores (90%). By physical activation with CO₂, the BET surface area further increases by a factor of 3 without destroying the monolithic character of the carbon. Hence, during physical activation with CO₂ both the microporosity and the concentration of oxygen containing surface functional groups are simultaneously increasing. The average pore diameter (D_pore) does not change significantly after physical and chemical activation. Furthermore, starting from the same resin, the 1 h chemical activation with KOH leads to a higher BET surface area and a higher volume when compared to a 1 h physical activation with CO₂. However, KOH activation is a very time consuming process due to the fact that the remaining KOH and salts formed during the activation have to get removed after chemical activation. These results are in good agreement with the findings of Teng et al.¹⁰ Both activation methods yield specific surface areas in the similar range.

Fig. 4 Low temperature nitrogen adsorption (open symbols)/desorption (closed symbols) isotherms.

Table 4 BET surface area, micropore volume and total volume of pyrolyzed and activated carbons

	SBET [m^2 g^−1]	Vpore [cm^3 g^−1]	Pore size distribution [%]		Dpore [nm]
			Micro	Meso
A-p900	410	0.22	90	10	2.1
A-p900-CO2-30 min	582	0.3	92	8	2.1
A-p900-CO2-1 h	709	0.37	92	8	2.1
A-p900-CO2-2 h	1032	0.5	94	6	1.9
A-p900-CO2-3 h	1207	0.61	94	6	2.0
A-p900-1 : 1-KOH-1 h	839	0.42	95	5	2.0
Reference silica gel	779	0.43	87	13	2.2

---

Water sorption isotherms

The influence of the activation time on the water uptake of the carbon materials is shown in Fig. 5a. Again, longer activation times lead to a steeper slope of the adsorption isotherm and a higher water sorption capacity given the increased specific surface area and micropore volumes. The water uptake of the carbon activated for 3 hours is shifted to higher relative pressure, which might be caused by larger pores or additional loss of nitrogen species.

Fig. 5 Water sorption of (a) RMF resins pyrolyzed and activated for different activation times and (b) samples used for TSA testing.

Fig. 5b illustrates the water sorption isotherms of all samples used for the TSA test as well as selected commercial activated carbons. Sample A-p900-CO₂-3 h exhibits a steeper slope and an even higher sorption capacity than the potassium hydroxide activated A-p900-1 : 1-KOH-1 h. The results are in agreement with the N₂ sorption measurements which exhibited a larger BET surface area and a higher micropore volume for the 3 h physically activated carbon than for the chemically activated carbon. The commercial silica gel, which was used as reference material, also adsorbs large amount of water at low relative pressure. However, the total water sorption capacity is lower for the silica gel compared than for the activated carbons. For the temperature swings investigated during the TSA experiments, the equivalent relative pressure can be estimated for different adsorbent temperatures as R.H. = 100 × P(T_e)/P_sat(T_ads), where P is the saturation pressure of water in the system given by the evaporator temperature T_e and P_sat is the saturation pressure of water at the adsorbent temperature T_ads. Based on the adsorption isotherms in Fig. 5b, the highest amount of water cycled in the temperature step 90 °C → 50 °C is expected from the silica gel. Within this temperature step it is able to cycle 8.3% of water (Table 5). The sample A-p900-CO₂-3 h shows by far the best performance for the temperature step 60 °C → 30 °C, with an overall water cycling capacity of 40.7% in this range.

Table 5 Amount of adsorbed water from 0.033 → 0.186 P/P_sat and 0.115 → 0.542 P/P_sat

Temperature step	90 °C → 50 °C (wt%)	60 °C → 30 °C (wt%)
A-p900-CO2-3 h	2.4	40.7
A-p900-1 : 1-KOH-1 h	4.1	30.1
A-p900	2.2	13.8
Reference silica gel	8.3	22.5

---

Dynamic heat transport and sorption properties

The effect of the activation time on the thermal and electrical conductivity is shown in Table 6. Due to the change of ρ the thermal conductivity was found to decrease steadily from 0.162 W m⁻¹ K⁻¹ after 0.5 hour of physical activation to 0.103 W m⁻¹ K⁻¹ after 3 hours of physical activation with CO₂. The gain of oxygen containing heteroatoms with increasing activation time results in an increase of the average molar mass of the remaining atoms M_mol,av. Therefore, a decrease of C_p might be understood as a gain of heavier elements within the limits of the Dulong–Petit law C_p = 3R/M_mol,av, whereas R is the gas constant. Furthermore, the electrical conductivity was found to decline from 7.67 S cm⁻¹ after 0.5 hour of physical activation to 2.82 S cm⁻¹ after 3 hours of physical activation with CO₂ and has a high correlation with the thermal conductivity. The electronic contribution to the thermal conductivity was calculated with the Wiedemann–Franz law κ_E = LTσ, whereas L is the Lorenz number which is equal to 2.44 × 10⁻⁸ V² K⁻², T the temperature and σ the electrical conductivity. For the physically activated carbons the electronic contribution lies between 2% and 4% of the total thermal conductivity.

Table 6 Thermal diffusivity and heat capacity measured by LFA

Sample	Envelope density ρ [g cm^−3]	Skeleton density ρS [g cm^−3]	Thermal diffusivity α [mm^2 s^−1]	Heat capacity Cp [J g^−1 K^−1]	Thermal conductivity κ [W (m^−1 K)]	Electrical conductivity σ [S cm^−1]	Electronic thermal conductivity κE [W m^−1 K^−1]	Lattice thermal conductivity κL [W m^−1 K^−1]
A-p900	0.579	1.856 ± 0.004	0.371 ± 0.019	0.60 ± 0.03	0.129 ± 0.013	5.21	4.11 × 10^−3	0.125
A-p900-CO2-0.5 h	0.484	1.903 ± 0.003	0.427 ± 0.021	0.782 ± 0.039	0.162 ± 0.017	7.67	6.05 × 10^−3	0.156
A-p900-CO2-1 h	0.424	1.859 ± 0.002	0.378 ± 0.019	0.785 ± 0.039	0.126 ± 0.013	4.01	3.16 × 10^−3	0.123
A-p900-CO2-2 h	0.413	1.932 ± 0.003	0.415 ± 0.021	0.754 ± 0.038	0.129 ± 0.013	5.05	3.99 × 10^−3	0.125
A-p900-CO2-3 h	0.398	1.882 ± 0.003	0.355 ± 0.018	0.730 ± 0.036	0.103 ± 0.011	2.82	2.22 × 10^−3	0.101
A-p900-1 : 1-KOH-1 h	0.438	1.834 ± 0.002	0.585 ± 0.029	0.63 ± 0.03	0.161 ± 0.016	10.77	8.49 × 10^−3	0.153

---

Chemically activated carbon has a 30% increased thermal conductivity compared to the physically activated carbon A-p900-CO₂-1 h. During chemical activation almost all heteroatoms are lost what causes a reduction of point defect scattering. The electrical conductivity of the chemically activated carbon is more than two times higher compared to the physically activated carbon A-p900-CO₂-1 h which has a similar density. According to Barroso-Bogeat et al.,²⁹ such significant changes on the electrical conductivity are likely connected with the formation of oxygen groups on the carbon surface rather than with the porosity development. To simulate the performance of the carbon sorbents in a heat-pump prototype, TSA tests were conducted under practically relevant conditions. Fig. 6a illustrates the transient cooling power response for a temperature step 90 °C → 50 °C. With a value of 123 W kg⁻¹ sample A-p900 shows a lower specific cooling power (SCP) over the entire cycle period than the reference silica gel which offers a maximal SCP of 255 W kg⁻¹.

Fig. 6 Specific cooling power at temperature step (a) 90 °C → 50 °C respectively (b) 60 °C → 30 °C. The temperature step was initiated after 60 seconds.

The carbon activated by KOH displays a fast increase and decline of the SCP which allows shortening the cycling time of the adsorption/desorption cycle. However, the maximal SCP of the sample A-p900-1 : 1-KOH-1 h (192 W kg⁻¹) is still lower than the silica gel reference and less than half of the sample A-p900-CO₂-3 h. At 430 W kg⁻¹, this 3 h CO₂ activated carbon outperforms the reference silica material by 68%. This is surprising since both chemically and physically activated carbons display virtually identical water sorption behavior at low relative pressure (<0.2 P/P_sat) but differ strongly in the SCP values. Even more surprising, the KOH activated carbon has the highest thermal conductivity of all carbons (<50% higher than sample A-p900-CO₂-3 h) but still shows less than half the SCP performance. This behavior is not understood at this point, however a possible explanation could lie in the almost three times higher the concentration of surface functional groups of sample A-p900-CO₂-3 h when compared with A-p900-1 : 1-KOH-1 h.

For a temperature step 60 °C → 30 °C, the transient SCP curves are presented in Fig. 6b. Both activated carbons exhibit a rapid power gain and drop with a maximum SCP value which is significantly higher than the silica gel reference material. The SCP of sample A-p900-CO₂-3 h exhibits a two-fold improvement over the maximal SCP of silica gel (779 W kg⁻¹ compared to 389 W kg⁻¹), which can be attributed to a combination of enhanced water cycling capacity in this operating window (Fig. 5) and improved heat & mass transport properties. This leads to the conclusion that CO₂ activated carbons have superior specific cooling power compared to their KOH activated analogs in the investigated cooling scenarios. More importantly, the CO₂ activated carbons are less sensitive to the temperature step/operating conditions than KOH activated ones and hence should be more universally applicable. Furthermore, the use of CO₂ rather than KOH enables to simplify and shorten the activation process from several days in total (including washing procedure) to a few hours. In order to further decrease the pyrolysis/activation treatment duration, the combined pyrolysis and physical activation in a single step will be the focus of a follow-up study.

The relative temperature change of the adsorbent during the thermal swing adsorption process was determined by infrared thermography. The intensity of the emitted radiation was normalized according I_n = (I_t − I_min)/(I_max − I_min) where I_n is the normalized intensity, I_t the intensity at time t, I_min the minimal measured intensity and I_max the maximal measured intensity during one temperature step. The normalized intensity is related to the temperature of the sample according to I_n ∝ εσT⁴, where ε is the emissivity of the sample, T is temperature and σ is the Stefan–Boltzmann constant. A comparison of the normalized temperature change in the center point of each of the three samples during adsorption is shown in Fig. 7. During both temperature steps, the carbon monoliths cool down faster than the silica gel. This can be explained by the monolithic structure of the carbon which eliminates thermal interfaces and provides a continuous thermal conduction path. The sample A-p900-CO₂-3 h takes longer to reach thermal equilibrium than A-p900-1 : 1-KOH-1 h for the 90 °C → 50 °C temperature step. This result agrees with the LFA measurements which indicated a 50% higher thermal conductivity of A-p900-1 : 1-KOH-1 h when compared to A-p900-CO₂-3 h. During the first minute following the temperature step from 60 °C → 30 °C, the sample A-p900-CO₂-3 h cools down as slowly as the silica gel which might be partially due to the reduced thermal conductivity and latent heat release. After 800 seconds, the CO₂ activated carbon reaches almost the same temperature intensity than the KOH activated one. In summary, carbon monoliths produced by pyrolysis and CO₂ activation have offer superior SCP over both, commercial silica gels and KOH activated carbons when benchmarked in 90 °C → 50 °C and 60 °C → 30 °C temperature swing adsorption scenarios.

Fig. 7 Thermal behavior of adsorbents during the TSA test by infrared thermography.

Conclusions

Nitrogen doped carbon monoliths with high water sorption capacity (>45 wt%) can be obtained by pyrolysis and CO₂ activation (800 °C, 3 hours), resulting in a significantly increased BET surface area of 1200 m² g⁻¹. During the CO₂ activation treatment, O and N containing surface functional groups are better preserved than with KOH activation. At the temperature step 60 °C → 30 °C, CO₂ activated RMF carbon showed a three-fold increase compared to a reference silica gel sorbent. These findings are the result of the enhanced water sorption capacity of CO₂ activated RMF carbon over silica gel sorbents combined with high O and N functional group densities. For a temperature step 90 °C → 50 °C, the CO₂ activated RMF carbon exhibits an increase of the specific peak cooling power by a factor of 1.7 in comparison to commercial silica gels, which comes as a surprise, considering that the sorption isotherms of the CO₂ activated carbon suggests low quantities of cyclable water. In conclusion, CO₂ activated resin based carbons represent a promising new class of potential sorbent materials for adsorption heat pumps/chillers.

Acknowledgements

The authors acknowledge the support of the Swiss National Science Foundation: National Research Programme (NRP70 under grant number 154008). Furthermore the following colleagues are acknowledged: Dr Bruno Michel for technical discussions and support, Hans-Jürgen Schindler for the access to the tube furnace, Dr Ulrich Müller, Dr Sascha Populoh and Elisabeth Michel for their assistance with materials characterization.

References

1. R. E. Critoph and Y. Zhong, Proc. Inst. Mech. Eng., Part E, 2005, 219, 285–300.
2. S. G. Wang, R. Z. Wang and X. R. Li, Renewable Energy, 2005, 30, 1425–1441.
3. L. Huber, P. Ruch, R. Hauert, G. Saucke, S. K. Matam, B. Michel and M. M. Koebel, RSC Adv., 2016, 6, 25267–25278.
4. R. E. Critoph and S. J. Metcalf, Appl. Therm. Eng., 2004, 24, 661–678.
5. Z. Tamainot-Telto, S. J. Metcalf, R. E. Critoph, Y. Zhong and R. Thorpe, Int. J. Refrig., 2009, 32, 1212–1229.
6. L. W. Wang, S. J. Metcalf, R. E. Critoph, R. Thorpe and Z. Tamainot-Telto, Carbon, 2012, 50, 977–986.
7. R. W. Pekala, C. T. Alviso, F. M. Kong and S. S. Hulsey, J. Non-Cryst. Solids, 1992, 145, 90–98.
8. H. F. Gorgulho, F. Gonçalves, M. F. R. Pereira and J. L. Figueiredo, Carbon, 2009, 47, 2032–2039.
9. Y. Yamashita and K. Ouchi, Carbon, 1982, 20, 41–45.
10. H. Teng and S.-C. Wang, Carbon, 2000, 38, 817–824.
11. M. S. Contreras, C. A. Páez, L. Zubizarreta, A. Léonard, S. Blacher, C. G. Olivera-Fuentes, A. Arenillas, J.-P. Pirard and N. Job, Carbon, 2010, 48, 3157–3168.
12. R. W. Pekala, J. Mater. Sci., 1989, 24, 3221–3227.
13. F. Rodríguez-Reinoso, M. Molina-Sabio and M. T. Gonzalez, Carbon, 1995, 33, 15–23.
14. S. Brunauer, P. H. Emmett and E. Teller, J. Am. Chem. Soc., 1938, 60, 309–319.
15. J. D. Merriman, A. G. Whittington, A. M. Hofmeister, P. I. Nabelek and K. Benn, Precambrian Res., 2013, 233, 358–372.
16. S. Shimooka, M. Yamazaki, T. Takewaki, E. Akashige, F. Ikehata, H. Kakiuchi, F. Watanabe, M. Kubota and H. Matsuda, Kagaku Kogaku Ronbunshu, 2006, 32, 528–534.
17. C. Moreno-Castilla, F. Carrasco-Marín, M. V. López-Ramón and M. A. Alvarez-Merino, Carbon, 2001, 39, 1415–1420.
18. M. Molina-Sabio, M. T. Gonzalez, F. Rodriguez-Reinoso and A. Sepúlveda-Escribano, Carbon, 1996, 34(4), 505–509.
19. T. Otowa, R. Tanibata and M. Itoh, Gas Sep. Purif., 1993, 7, 241–245.
20. H. Marsh and F. Rodríguez-Reinoso, Activated carbon, Elsevier, Oxford, 2006.
21. D. Feng, Z. Zhou and M. Bo, Polym. Degrad. Stab., 1995, 50, 65–70.
22. J. R. Pels, F. Kapteijn, J. a. Moulijn, Q. Zhu and K. M. Thomas, Carbon, 1995, 33, 1641–1653.
23. S. R. Kelemen, M. L. Gorbaty and P. J. Kwiatek, Energy Fuels, 1994, 8, 896–906.
24. H. Marsh, Activated Carbon, Elsevier Science Ltd, Oxford, 2006.
25. J. Yu, M. Guo, F. Muhammad, A. Wang, G. Yu, H. Ma and G. Zhu, Microporous Mesoporous Mater., 2014, 190, 117–127.
26. Y. Li, T. Ben, B. Zhang, Y. Fu and S. Qiu, Sci. Rep., 2013, 3, 2420.
27. K. S. W. Sing, D. H. Everett, R. A. W. Haul, L. Moscou, R. A. Pierotti, J. Rouquérol and T. Siemieniewska, Pure Appl. Chem., 1985, 57, 603–619.
28. K. M. Santhosh, J. Pérez-Ramírez, M. N. Debbagh, B. Smarsly, U. Bentrup and A. Brückner, Appl. Catal., B, 2006, 62, 244–254.
29. A. Barroso-Bogeat, M. Alexandre-Franco, C. Fernández-González and A. Macías-García, Microporous Mesoporous Mater., 2015, 209, 90–98.

---

Footnote

† Electronic supplementary information (ESI) available: Secondary electron SEM images, Raman spectroscopy and the deconvolution of the XPS spectra of the resin and the pyrolyzed and activated carbons. See DOI: 10.1039/c6ra18660b

---