Mechanistic and Experimental Study of the CuxO@C Nanocomposite Derived from Cu3(BTC)2 for SO2 Removal

A tunable and efficient strategy was adopted to synthesize highly porous nano-structured CuO−carbonized composites (CuxO@C) using Cu3(BTC)2 as a sacrificial template. The as-synthesized CuO nanocomposites exhibited hollow octahedral structures, a large surface area (89.837 m2 g−1) and a high proportion of Cu2O active sites distributed on a carbon frame. Based on DFT calculations, both the Cu atoms on the surface (CuS) and oxygen vacancy (OV) exhibited strong chemical reactivity. On the perfect CuO (111), the CuS transferred charge to O atoms on the surface and SO2 molecules. A strong adsorption energy (−1.41 eV) indicated the existence of the chemisorption process. On the oxygen-deficient CuO (111), the O2 preferably adsorbed on OV and then formed SO3 by bonding with SO2, followed by the cleavage of the O−O bond. Furthermore, the CuO nanocomposites exhibited an excellent ratio of S/Cu in SO2 removal experiments compared with CuO nanoparticles produced by coprecipitation.


Introduction
Sulfur dioxide (SO 2 ), as a major pollutant gas, is mainly derived from fossil fuel combustion for electrical energy in large industries, such as coal-fired power plants [1]. The emitted SO 2 in flue gas can directly cause acid rain by the reaction with oxygen and water vapor, which is seriously harmful to the ecology and human health [2]. Thereby, the SO 2 purification from flue gas has enormous significance for ecological and environmental improvement. Currently, the SO 2 separation method using metal oxides as a desulfurizer, which can effectively reduce the content of SO 2 in flue gas by catalytic oxidation, has gradually become the chief technology of SO 2 removal [3,4].
Among them, the method by the reaction of calcium oxide (CaO) with SO 2 and O 2 to form CaSO 3 / CaSO 4 is extensively applied in industrial applications. However, this product after SO 2 uptake is almost irreversible due to the high energy consumption for regeneration caused by a high decomposition temperature [5,6]. The commonly used metal oxide desulfurizer for the sulfide removal also include V 2 O 5 , CeO 2 , MnO x , Fe 2 O 3 , ZrO 2 and CuO [7][8][9][10][11][12][13].
CuO possesses the advantages of high selectivity for SO 2 and renewability, resulting in an excellent performance in SO 2 removal compared with the other metal oxides [14][15][16].
To reduce the amount of metal and economic cost, CuO is usually supported on porous materials and then on the active sites exposed in the flue gas [17][18][19][20][21][22][23]. Tseng et al. reported that copper-oxide-supported activated carbon has outstanding performance in SO 2 adsorption and regeneration [24]. The introduction of CuO contributes to the enhancement of SO 2 adsorption and oxidation reaction [19]. Yuan et al. found that SO 2 can be catalytically oxidized to SO 3 on the copper sites in the existence of O 2 and then form sulfates, which causes a gradual deactivation of copper-loaded activated coke [18]. In addition, the effect of the oxidation state of Cu on SO 2 adsorption has been reported in the existing literature. In the report of Soldemo et al., the unsaturated copper atoms at the Cu 2 O (111) surface facilitated the SO 2 to SO 3 transformation [25]. Galtayries et al. presented two different paths and products for SO 2 adsorption on Cu 2 O as the number of oxygen vacancies increases [26]. Moreover, Baxter et al. also obtained data on the Co−adsorption of O 2 and SO 2 on Cu 2 O, which can lead to the formation of CuO and the adsorption of SO 3 (ads) [27]. Hammershøi et al. reported DFT calculations results indicating that both SO 2 and SO 3 can form stable species with Cu 2+ and Cu + , while the Cu + site and Cu 2+ site preferably bond to SO 2 and SO 3 , respectively [28].
Moreover, MOF materials can also be applied as templates to in situ synthesize MOF−based metals/metal oxides/metal-doped carbon composites through pyrolysis under controllable conditions [33][34][35][36][37][38][39]. The principle of this transformation is considered to be the release of small molecules from the precursors, which lead to the formation of porously hollow nanocomposites. The internal relationship between the structure and activity of CuO composites derived from Cu 3 (BTC) 2 was investigated in the existing literature [38].
In the report of Wang et al., Cu 2 O/CuO polyhedral materials with the shape of octahedral, truncated octahedral and cubic were successfully obtained by controlling the morphology of the Cu 3 (BTC) 2 precursors, in which the octahedral Cu 2 O/CuO exhibited excellent gas-sensing performance resulting from a large specific surface area and high ratio of adsorbed O on the surface [39]. In addition, this synthesized strategy that uses the Cu 3 (BTC) 2 as a template has been reported in many applications related to the field of adsorption [40,41] and catalysis [34,35]. Peng et al. found that octahedral CuO composites could be synthesized by pyrolysis of Cu 3 (BTC) 2 under the N 2 atmosphere, which showed a superior catalytic activity for NO x at a low temperature [34].
Vantran et al. synthesized a Cu x O−supported carbon matrix composite (Cu/Cu 2 O/ CuO@C) through the thermolysis of the Cu 3 (BTC) 2 template, which exhibited a high removal efficiency over chloramphenicol antibiotic [35]. These studies towards derivatives effectively extend the potential applications of Cu−MOFs and reconfirm that the texture properties of Cu x O@C composites are promising in the fields related to catalysis and gas separation. However, there are few studies on flue gas desulfurization (FGD) by using the pyrolysis product of Cu 3 (BTC) 2 as a desulfurizer. Therefore, the study of MOF−derived composites for SO 2 purification is promising to further broaden the application scope of MOF.
Herein, the outcomes of previous studies prompted us to synthesize the hierarchically porous CuO composites supported on carbon by pyrolyzing Cu 3 (BTC) 2 precursors and use these as a novel catalyst for FGD. The pyrolysis products of Cu 3 (BTC) 2 are characterized, and its SO 2 removal performance is tested and compared with that of CuO nanoparticles prepared by coprecipitation. The theoretical calculation method was also adopted to further investigate the reaction mechanisms over the perfect and oxygen-deficient CuO (111) surface through the state density, Bader charge transfer and charge density difference; thus, a possible adsorption microscopic mechanism was postulated.
In summary, the main aim of this study is to provide an adjustable synthesis method of CuO nanocomposites using MOFs for highly efficient SO 2 removal and to study the microscopic mechanism behind the reaction. Moreover, this work broadens the potential applications of MOF materials. The Cu 3 (BTC) 2 crystals show the morphologically octahedral structure with a size of about 1 µm ( Figure S1), which is consistent with the work described in previous literature [36]. Figure 1a-e shows the TEM images of synthesized CuO composites, which describe the interior structure of the thermal decomposition products. As shown in Figure 1a, the CuO−NC derived from Cu 3 (BTC) 2 almost completely maintains the well-defined octahedral structure of Cu 3 (BTC) 2 , while the numbers of loose pores are distributed within each octahedral unit, forming a hollow structure. However, the CuO−NP produced by coprecipitation showed a compact aggregation of many interconnected particles ranging from 50 to 100 nm. There was one group of lattice fringes in the HR−TEM image of CuO−NP (Figure 1e). Figure 1f shows the XRD patterns of the Cu 3 (BTC) 2 precursors and nano−CuOs. All peaks of the synthesized Cu 3 (BTC) 2 matched well with that of simulated Cu 3 (BTC) 2 . The diffraction peaks of CuO−NC could be assigned to CuO (JCPDS#04−0836) and Cu 2 O (JCPDS#05−0667), while those of CuO−NP could be exclusively attributed to CuO.

Results and Discussion
These results were consistent with HR−TEM analysis, indicating that using the Cu 3 (BTC) 2 as the template led to the formation of multiple oxides of copper in pyrolysis products. The thermal decomposition temperature of 300 • C for Cu 3 (BTC) 2 precursors in the air was determined based on the TG measurements. According to the TGA curves in Figure 1f, the weight loss process of Cu 3 (BTC) 2 was divided into five stages. The first sharp weight loss stage (stage I) was attributed to the removal of water and solvent molecules (CH 3 OH) from the surface and pores of the Cu 3 (BTC) 2 precursors.
The second theoretical weight loss stage was observed between 288 and 320 • C (stage III and IV) was caused by the pyrolysis of the organic ligands and oxidation in Cu 3 (BTC) 2 [34]. Notably, a decrease in the weight loss rate was observed at temperatures ranging between 292 and 320 • C, which could be related to the oxidation of Cu 2 O. When the temperature was higher than 320 • C (stage V), the second platform was obtained, indicating that the Cu 2 O may be oxidized entirely to CuO [34,42]. Therefore, the products at 330 • C were an intermediate presence of Cu 2 O along with CuO, which is consistent with the literature [38,42,43]. In addition, the existence of carbon also provided a favorable atmosphere of reducing for the formation of Cu 2 O in this work.

Porous
Structures of Nano−CuOs N 2 adsorption-desorption isotherms at 77 K were measured for examining the pore characteristics of the nano−CuOs. As shown in Figure 2a, in the CuO−NC prepared by Cu 3 (BTC) 2 , the N 2 adsorption capacity was less at the low P/P 0 region, while the amount of adsorbed N 2 rapidly increased once the value of P/P 0 reached 0.7. The curves of CuO−NC exhibited a typical type−IV sorption isotherm with the H3−type hysteresis loop between the relative pressure 0.45 and 1.0, which could be explained by the presence of slit-like pores that resulted from the aggregation of nanocrystalline particles [44], whereas the CuO−NP prepared by coprecipitation exhibited a low amount of N 2 adsorption (≤1.2 cm 3 g −1 ) with the variation of P/P 0 .  Figure 2b shows the pore size distributions of CuO−NC and CuO−NP, which were obtained by the BJH method. The CuO−NC shows five pore size distributions with a sharp peak at 2.2 nm as well as three broad peaks at 4.8, 11.3 and 18.6 nm. The pores with small aperture could be derived from the octahedral frames and shells, while that with the large aperture may originate from the gap between octahedral units. The detailed data were summarized in Table S1.
The specific surface areas of the two nano−CuOs are 89.837 and 1.502 m 2 g −1 , corresponding to the CuO−NC and CuO−NP, respectively, where the value of the porous octahedral CuO−NC is higher than that of many nano−Cu 2 Os or nano−CuOs reported in previous literature [34,36,42,45]. In addition, the CuO−NC showed a larger micropore volume (0.096 cm 3 g −1 ) and mesopore volume (0.160 cm 3 g −1 ) when compared with those of CuO−NP (0.000097 and 0.000998 cm 3 g −1 ). This suggested that using the Cu 3 (BTC) 2 precursors as templates facilitated the formation of the hierarchical pore structure of calcined products.

Chemical Compositions of Nano−CuOs
XPS analysis was conducted to examine the chemical compositions on the surface of as-synthesized CuO−NC and CuO−NP. Figure 3a displays the Cu 2p XPS spectra, where four sub-peaks were deconvoluted from the corresponding curve of the CuO−NC. The two main peaks could be observed at the position of 932.6 and 952.6 eV, which corresponded to the 2p3/2 and 2p1/2 peaks of Cu + , respectively [46]. The other two major peaks at the higher binding energy positions (933.7 and 953.6 eV) mainly originated from Cu 2+ [47]. Moreover, the typical satellite peaks accompanied by the electronic shake-up were also observed in the Cu2p XPS spectra of the CuO−NC, which reconfirmed the existence of Cu 2+ [31,36]. Thereby, these results proved the coexistence of Cu 2 O and CuO in the CuO−NC, where the relative proportion of Cu + was about 59.5% (Table 1). In contrast, in Figure 3a, only two main peaks corresponding to Cu 2+ appeared in the Cu 2p XPS spectra of the CuO−NP, indicating that the products may belong to a pure phase, which was consistent with the results of the HR−TEM and XRD analysis. As shown in Figure 3b, the two nano−CuOs exhibited similar O1s XPS spectra that were resolved into three fitting peaks at 530.3 ± 0.1, 531.5 ± 0.1 and 532.3 ± 0.2 eV, respectively, corresponding to the lattice oxygen (O L ), oxygen vacancy (O V ) and adsorbed oxygen (O C ) peaks [38]. The O C and O V components represent the hydroxyl (OH − ), chemically adsorbed and dissociated oxygen, such as O 2− , O 2− and O − , which is usually related to the ability to adsorb oxygen [38,39]. Table 1 summarizes the O element proportion of the surface chemical compositions according to the fitting results of the XPS spectra.
The CuO−NC displayed a larger relative proportion of O C than that of CuO−NP. In Figure 3c, the Raman spectroscopy analysis of CuO−NC and CuO−NP revealed that three similar peaks in both of two nano−CuOs were located at 298, 330 and 620 cm −1 , which were assigned to the characteristic peaks of CuO in the Raman active modes of Ag, B 1g and B 2g , respectively [48].
In particular, three other peaks at 204 cm −1 , 534 cm −1 and 585 cm −1 were observed in the CuO−NC, which was believed to be contributed by the Cu 2 O phase [49]. The EPR analysis of CuO−NC and CuO−NP was conducted to study the oxidation states of the metal and surface chemisorbed oxygen. As shown in the EPR spectra (Figure 3d), a sharp peak of the g-value (g = 2.003) was discovered in the CuO−NC samples, which was the characteristic signal of the paramagnetic materials containing superoxide ions (O 2− ) [50][51][52]. Based on the relevant literature, it was also related to the existence of oxygen vacancies [50]. To date, there have been very few studies regarding the fundamental understanding of the SO 2 adsorption mechanism on the CuO surface. It is also difficult to characterize the atomic-scale adsorption configurations through experiments. Owing to the lack of information on this research, we employ atomic simulations to investigate the SO 2 adsorption mechanism at the CuO surfaces.
Maimaiti et al. reported that the CuO (111) is the most stable surface by predicting the computed surface energies of different CuO surfaces using DFT simulations (E CuO (111) < E CuO (011) < E CuO (110) < E CuO (010) < E CuO (100) ) [53]. Therefore, the relatively stable (111) facet was adopted as the surface termination, which also was reconcilable with the XRD analysis of the experiment section (Figure 1f). Figure 4 shows the optimized surface slab models of CuO surface. The unit cell of simulated CuO−NC is shown in Figure 4a. The periodic surface was 11.6 Å × 12.6 Å × 24 Å with a vacuum slab of 16 Å in thickness, which can separate the layer from its periodic images. As shown in Figure 4b, the atoms of the bottom six layers (two Cu layers and four O layers) were frozen in their equilibrium positions within the geometry optimizations, whereas all other atoms were allowed to structurally relax. There are four different types of surface atoms on the clean CuO (111) surface, which are separately attributed to unsaturated (Cu 3c ) and saturated (Cu 4c ) copper atoms, unsaturated (O 3c ) and saturated (O 4c ) oxygen atoms.
Among them, the bond length of bulk Cu−O is 1.941 Å, which is much longer than that of Cu 3c −O 3c (1.797 Å) and Cu 3c −O 4c (1.888 Å) and shorter than that of Cu 4c −O 4c (1.966 Å) and Cu 4c −O 3c (1.919 Å). Furthermore, the SO 2 adsorption was simulated on one side of the exposed surfaces of the slab, where the atoms of the top six layers and adsorbates were fully relaxed. The adsorption energies between all possible adsorption sites as well as the structures with each adsorbate were calculated; however, only the maximum was used for discussion.

Adsorption Configuration of SO 2 Molecules over Perfect CuO (111)
The most stable SO 2 adsorption position was identified, and the calculated adsorption energy was −1.41 eV (Figure 4c). In Figure 4d, the S atom of SO 2 molecule forms bonds with O 3c atom on the surface with the bond length of 1.74 Å, while two O atoms form bonds with Cu 3c atoms and the bond lengths are 1.96 and 2.09 Å. After SO 2 chemical adsorption on the surface, the bond angle of SO 2 molecules changed from 119.308 • to 107.571 • . The bond length of S−O bonds of SO 2 molecule became 1.612 Å and 1.462 Å from two equal 1.448 Å, which increased compared with before adsorption.
The adsorption energy is −1.41 eV, indicating that the chemisorption between SO 2 and CuO (111) is very stable [54,55]. Furthermore, the charge density difference (CDD) is beneficial to directly exhibit the charge accumulation and consumption of SO 2 molecules after adsorption on the surface of CuO, which can reflect the adsorption strength according to the amount of charge transfer. In this paper, CDD is defined as the charge density of the SO 2 −CuO system after adsorption minus that of the CuO surfaces and the SO 2 molecules (CDD = C (CuO−SO 2 ) − C (CuO) − C (SO 2 )).  (Figure 4h).
In addition to the change of adsorption structure, studying the properties of electrons by the analysis of the density of states (DOS) is also conducive to understanding the bonding mechanism of adsorption. Figure 5 shows the DOS of SO 2 molecules before and after adsorption on the surfaces of CuO. The DOS of SO 2 molecules before adsorption possesses the peaks with energy of −22, −19, −9, −6, −5, −2.5, −1.85 and −1.25 eV (Figure 5a), while those of SO 2 after adsorption are shifted to the deep energy level.
The value of DOS peaks decreased, which could result from the transfer of charge between the SO 2 molecule and surface. Notably, a new continuous density peak of SO 2 molecules after adsorption was discovered at −9 to −1 eV (Figure 5b), which corresponded to the contiguous density peak of Cu atoms after adsorption (Figure 5c), indicating the existence of a strong interaction between SO 2 molecules and Cu atoms on the surface.
In addition, Table 2 summarizes the results of the charge transfer between SO 2 molecules and CuO surfaces in the reaction process, which are favorable to understanding the mechanism of the reaction. These results demonstrated that the Cu S atoms lose a total of 0.0182 e, while the O S , S m and O m atoms gained 0.047 e, 0.025 e and 0.137 e, respectively, indicating that the charges were mainly transferred from the Cu S to the O S and SO 2 molecules, which is consistent with the CCD analysis in the previous section.

Adsorption Configuration of SO 2 Molecules on the Oxygen-Vacancy CuO (111)
The DFT−calculated results regarding the adsorption configuration and electrontransfer mechanism on the perfect CuO (111) surface indicate that the Cu S sites are the favorable active sites for SO 2 . Following the experimental evidence on the existence of oxygen vacancy, we further investigated the adsorption configurations on oxygen vacancy on the CuO (111) surface and their energetics by the DFT calculations to understand the role of the vacancy of atomic oxygen. Figure 6a shows the oxygen vacancy formed by removing the O 3c .  According to the report of Rajendran et al., the O 3c vacancy was preferably formed as compared to O 4c [56]. Figure 6b illustrates Figure 6c,d after the optimization. For understanding the reaction mechanism of SO 2 removal, the adsorption energies of SO 2 on the O 2 −adsorbed oxygen-vacancy CuO (111) surface (E c ) and the SO 3 adsorbed on the perfect CuO (111) surface (E d ) were calculated, where the difference of value (E d − E c = −2.77 eV) exhibited that the stability of SO 3 adsorption on the CuO (111) surface was superior to that of SO 2 . The O 2* in intermediate stages (Figure 6b) is likely to be readily bonded to free SO 2 molecules, followed by the cleavage of the O 1* −O 2* bond. Subsequently, the free SO 3 molecules are adsorbed on the perfect CuO (111) surface and further form sulfates.

SO 2 Removal Performance of Nano−CuOs
FGD tests were conducted in order to evaluate the SO 2 removal performance of the two nano−CuOs and confirm the SO 2 removal mechanism obtained by DFT calculations. The influence of operating temperatures on the desulfurization was investigated by exposing the CuO−NC and CuO−NP under a range of temperatures between 100 and 250 • C. As shown in Figure 7a, the SO 2 removal efficiency of the two CuOs decreased over time, which was resulted from gradual inactivation of CuO sites by reacting with SO 2 . The S/Cu ratios and SO 2 capacity of the two nano−CuOs shown in Figure 7b were calculated by integration of the measured decrease in SO 2 concentrations of outlet (Figure 7a).
The CuO−NC demonstrated an excellent SO 2 removal performance at the range of 100 to 250 • C compared to CuO−NP, which could be explained by different desulfurization mechanisms. At the low temperature of 100 • C, there was a wide margin between the S/Cu ratio of CuO−NC and that of CuO−NP, which could be ascribed to the larger specific surface area and pores volume of CuO produced by Cu 3 (BTC) 2 (Figure 2), resulting in an enhanced physisorption at this temperature.
In contrast, with the increasing reaction temperature from 100 to 150 • C, a weakened physisorption appeared, while the chemical action of CuO sites with SO 2 was almost absent, which led to a similarly low SO 2 uptake capacity in both of two nano−CuOs. The differential of the SO 2 uptake between two CuOs at 200 • C became more pronounced due to the prominent role of chemisorption, where the octahedral CuO−NC possessed larger S−uptakes, which may be related to the existence of Cu + . Hammershoi et al. reported that SO 2 is bound more strongly to Cu + than Cu 2+ by DFT calculations [28].
The outstanding SO 2 capacity could also be attributed to the support of carbon skeleton for Cu active sites in the hollow CuO−NC, which was beneficial in terms of both the SO 2 reaction and diffusion. However, the S/Cu ratio of CuO−NC slightly decreased as the temperature further rose to 250 • C, which may be rationalized by the faster chemical reaction rate at higher temperatures. On the one hand, Cu + was rapidly oxidized to Cu 2+ , which narrowed the differentials of SO 2 removal performance, resulting from the chemical compositions, between the CuO−NC and CuO−NP. On the other hand, the CuO and SO 2 converted to CuSO 4 faster, which may cause the deactivation of desulfurizer due to the covering by rapidly formed sulfates [58]. Furthermore, a comparison of the desulfurization performance of CuO−NC at 200 • C in the presence and absence of O 2 (6 vol.%) and H 2 O(g) (6 vol.%) is shown in Figure 7c. The results revealed that both of S/Cu and SO 2 capacity significantly increased in the presence of O 2 , which can be explained by the oxidation of SO 2 and the formation process of sulfates (SO 2 + 1 2 O 2 → SO 3 ; SO 3 + CuO → CuSO 4 ) [18,58]. This process reconfirmed the adsorption mechanism of SO 2 molecules on the oxygen vacancy CuO (111) surface ( Figure 6). In addition, based on the presence of O 2 , the addition of H 2 O(g) further facilitated the SO 2 uptake, which contributed to the formation of sulfurous acid and reaction with CuO (SO 2 + H 2 O → H 2 SO 3 ; H 2 SO 3 + CuO + 1 2 O 2 → CuSO 4 + H 2 O). The SO 2 removal performance and test parameters of earlier reported CuO−based composites desulfurizers from literature were summarized in Table 3. In contrast, the best-performed sample CuO−NC of this work showed a smaller surface area but a larger removal capacity (157.63 mg g −1 ), which can be attributed to the more effective interaction between the exposed Cu sites and SO 2 .

Materials
Copper nitrate trihydrate (Cu(NO 3 ) 2 ·3H 2 O), sodium hydroxide (NaOH) and methanol (CH 3 OH) were purchased from Sinopharm Chemical Reagent Co., Ltd. (Shanghai, China). 1,3,5−benzenetricarboxylic acid (BTC) was purchased from J&K Scientific Co., Ltd. (Beijing, China). All of chemical reagents and solvents in this work were analytical grades without further purification. Figure 8 demonstrates the synthesis process of nano−CuOs. A synthetic route with high yield was adopted to fabricate octahedral Cu 3 (BTC) 2 [39]. First, Cu(NO 3 ) 2 ·3H 2 O and BTC with the mole ratio of 3:2 were separately dissolved in 25 mL methanol and then fully stirred for 10 min. The obtained solution was placed still at room temperature for 150 min after ultrasonication for 20 min. Then, the products were washed repeatedly with methanol and put inside the oven at 80 • C for 12 h, followed by marking as Cu 3 (BTC) 2 . Finally, the CuO nanocomposites were obtained by the pyrolysis of as-prepared Cu 3 (BTC) 2 under an air atmosphere at 300 • C for 120 min and then named as "CuO−NC".

Synthesis of Nano−CuOs
In addition, the prepared CuO nanoparticles, as comparative samples, were synthesized by the coprecipitation method. 2.45 g of Cu(NO 3 ) 2 ·3H 2 O and 4.30 g of NaOH were dissolved in 50 mL of methanol. The black products were collected by centrifugation as well as repetitive washing with methanol and then was dried in the oven at 80 • C for 12 h. Then, the resulting sample was calcined at the same thermal treatment conditions with that of CuO−NC, which was labelled as "CuO−NP".
The specific surface area was analyzed by the BET equation, which was measured by an accelerated surface area and porosity system (ASPS 2020 plus, Micromeritics, Norcross, GA, USA). The size distribution and volume of microporous and mesoporous were calculated by the t−plot method and the BJH (Barrett−Joyner−Halenda) method, respectively, and the total pore volume was given by adsorption amount at a relative pressure of P/P 0 = 0.99. The surface chemical composition of CuO nanocomposites was analyzed by XPS (ASIX Supra, Kratos Analytical, Manchester, UK). The oxidation states of metal oxide were investigated at room temperature by EPR (A200, Bruker, Billerica, MA, USA) using the X−band.
The chemical bonds of the CuO nanocomposites were examined by a laser confocal Raman spectrometer (JY−T64000, Jobin−Yvon, Longjumeau, France).  Figure S3 shows a flow chart of the FGD simulated instrument. In a typical FGD test, the CuO nanocomposites were loaded into a quartz tube (6 mm i.d.), where all of samples were weighted with the equal amount (0.15 ± 0.01 g). Subsequently, the quartz tube loading with sample was dried in the oven at 80 • C for 12 h, followed by installing into the FGD simulated instrument and purging with N 2 (120 mL min −1 ) for 30 min.

FDG Activity Test
Finally, the reaction gases with a total flow rate of 120 mL min −1 (2000 ppm SO 2 , 6 vol.% O 2 and 6 vol.% H 2 O (g), N 2 as balance gas) were introduced after the temperature was up to setpoint at a ramping rate of 5 • C min −1 , where the gas hourly space velocity (GHSV) was set to 16,900 h −1 . Moreover, the inlet content of SO 2 was measured by the mass flow meters, while the data of outlet SO 2 were collected by a flue gas analyser (Pro350, Testo, Titisee-Neustadt, Germany).
The items of SO 2 removal capacity (Q so 2 , mg g −1 ) and the ratio of S to Cu (S/Cu) were used to evaluate desulfurization performance of samples where the Q so 2 of samples was calculated by the data that were correspondence with the duration before the SO 2 removal efficiency drops to the threshold of 50%. The expressions of the SO 2 removal efficiency (η, %) and capacity were given by the following equations where C in and C out correspond to the inlet and outlet concentrations of SO 2 (ppm), respectively; F stands for the total flow of simulated flue gas (mL min −1 ); M r represents the relative molecular mass of SO 2 (g mol −1 ); m and V m are the standard the sample mass (g) and molar volume (L mol −1 ) at setpoint conditions, respectively.

Computational Method
All of the spin-polarized density functional theory (DFT) calculations were conducted using the Vienna Ab−Initio Simulation Package1-3 [60,61] (Version 5.4.4). The ion-electron interaction of Cu, O and S in the periodic boundary condition was performed by the standard VASP projector augmented-wave [62] (PAW) potentials. The valence electron configuration of Cu, O and S in the calculation corresponded to 3d 10 4s 1 , 2s 2 2p 4 and 3s 2 3p 4 , respectively. In this work, the Perdew-Burke-Ernzerhof [63] (PBE) exchangecorrelation functional was selected as the generalized gradient approximation for the exchange-correlation functional.
The DFT−D3 method proposed by Grimme [64] was used for van der Waals force corrections, which consisted in adding a pair wise potential to the DFT total energy after each self-consistent cycle. Based on the on-site Coulombic repulsion between the d electrons, the Hubbard U corrections were performed to transition metal d-electrons and the values of U-J parameters for Cu atoms (7 eV). Through benchmark calculations, the plane-wave basis kinetic-energy cut-off adopted was 450 eV, and its corresponding Brillouin zone was sampled by a Monkhorst-Pack [65] k-point grid of 1 × 1 × 1. The self-consistent calculation converged to 10 −5 eV and executed structural optimization by minimizing the forces on all the atoms until they were less than 0.01 eV Å −1 .

Adsorption Energy
The expression of adsorption energy (∆E ads ) can be given as where E surface+adsorbate , E molecule and E surface represent the total energy of adsorbate-substrate system, the total energy of bare surface and the total energy of gas-phase adsorbates, respectively. In particular, a negative value refers to an exothermic binding, which means a more stable state for the adsorbed species on the CuO (111) surface.

Conclusions
In conclusion, this work systematically investigated the mechanisms of SO 2 removal over as-synthesized CuO nanocomposite surfaces using experimental and theoretical methods. Octahedral CuO nanocomposites with a carbon frame were fabricated by the pyrolysis of Cu 3 (BTC) 2 at 300 • C under an air atmosphere. The as-synthesized composites possessed a hollow structure and high surface area (89.837 m 2 g −1 ) as well as multivalent copper oxides due to the presence of an MOF−derived carbon octahedra frame.
In addition, the effects of the synthetic method, reaction temperature and atmosphere on the SO 2 adsorption over CuO nanocomposites were investigated by experiments, where the CuO nanocomposites exhibited a superior SO 2 removal performance compared with the CuO nanoparticles produced by coprecipitation. Moreover, the charge transfer and density of state of the CuO (111) surface and SO 2 gas molecules before and after adsorption were studied based on the experimental results and DFT calculations.
On the perfect CuO (111) surface, the most stable SO 2 adsorption configuration was obtained over the CuO surface by the electron transfer mechanism between Cu atoms on the surface and SO 2 molecules. On the oxygen-deficient CuO (111) surface, O 2 molecules were adsorbed on the oxygen vacancy sites and bonded with free SO 2 molecules to form SO 3 . Based on the above results, this work provides a novel strategy to produce highly effective CuO nanocomposites as catalysts by MOFs for desulfurization and to further validate the mechanisms of SO 2 chemisorption with the DFT calculations.

Conflicts of Interest:
The authors declare no conflict of interest.