A Magnetically Recoverable Fe3O4–NH2–Pd Sorbent for Capture of Mercury from Coal Derived Fuel Gas

A sort of magnetical material named Fe3O4–NH2–Pd was prepared by loading varying amounts of immobilizing Pd on the surface of the magnetic Fe3O4–NH2 microspheres. This magnetical material was used firstly for capturing Hg° from coal derived fuel gas based on its recoverability. The experimental results showed that the loading Pd on the amine-functionalized magnetite nanoparticles can greatly improve the efficiency of removing Hg° at a high temperature range between 200 and 300 °C. The magnetic Fe3O4–NH2–Pd sorbent with 5% Pd loaded exhibited significantly high activity and stability in capturing Hg°, affording over 93% capture efficiency at 200 °C for more than 8 hrs. Compared to the Fe3O4–NH2 sorbent that converted the Hg° as HgS, this Fe3O4–NH2–Pd sorbent can remove the Hg° by forming Pd-Hg amalgam and HgS. In addition, the experimental tests indicated that the as-synthesized Fe3O4–NH2–Pd sorbent still showed stable magnetic properties after two regeneration cycles in removing Hg°, which provided the opportunity for preparing a recyclable sorbent which can be easily separated and recovered for Hg° removal.

Because of its practical importance, this field has also attracted considerably theoretical interests. Sasmaz et al. studied the adsorption of Hg on Pd binary alloys and overlays using PW91 functionally 25 . They found that Pd has the highest mercury binding energy in comparison to other noble metals. Lim et al. found that the number of vacancies surrounding the three-fold hollow site could affect the adsorption of Hg on Au surface 26 . It has been found that the adsorption performance of the sorbents can be improved by doping second metal or adding promoter to the sorbents. It was found that the addition of the small amounts of Au, Ag and Cu to the Pd could increase the overall mercury binding energy to the Pd surface 27 . DFT calculations were also carried out to investigate the adsorption of mercury and its compounds on the V 2 O 5 -WO 3 -TiO 2 , it is found that the ternary systems (V 2 O 5 -WO 3 -TiO 2 ) showed a higher reactivity compared with the binary systems (V 2 O 5 -TiO 2 or WO 3 -TiO 2 ) 28 . Therefore, it is expected that the ideal mercury removal sorbents can be prepared by introducing the magnetic nanoparticles to the Pd sorbent to form a bimetallic or alloy sorbent.
The objectives of this work are to develop a recyclable sorbent for Hg removal. To achieve the goals, The metallic Pd and a magnetic material were selected as the active metal and support to prepare these sorbents. The performance of these sorbents in removing Hg° from the simulated fuel gas was investigated using a laboratory-scale fixed-bed reactor. These include: (1) the effects of temperature on the removal of Hg° of the as-synthesized Fe 3 O 4 -NH 2 and Fe 3 O 4 -NH 2 -Pd sorbents; (2) the regeneration performance of the Fe 3 O 4 -NH 2 -Pd sorbent. The fresh and used Fe 3 O 4 -NH 2 and Fe 3 O 4 -NH 2 -Pd sorbents were characterized by transmission electron microscopy (TEM), scanning electron microscopy (SEM), X-ray powder diffraction (XRD) and X-ray photoelectron spectroscopy (XPS). The mechanisms of Hg° removal over these sorbents were clarified based on the experimental results.

Results and Discussion
Effect of Pd loading content on Hg° removal over Fe 3 O 4 -NH 2 -Pd sorbents. A Figure 3 summarizes the experimental tested mercury capacity (MC) and the theoretical mercury capacity (MC T ) of the Fe 3 O 4 -NH 2 -Pd sorbents. Clearly, it can be seen that the mercury capacities decreased with temperature. Compared to the theoretical mercury capacities, it can be seen that the MC obtained at three different temperatures were all lower than corresponding MC T . For instance, at 200 °C, the experimental mercury capacity was 30.1 μg/g, while the theoretical value is 35.0 μg/g. This is because the experimental mercury capacity (MC) was based on the measurements of the pyrolysis accessories of the mercury analyzer that can detect all the mercury species. By contrast, the theoretical mercury capacity (MC T ) was calculated on the basis of the curve of the Hg° removal efficiency that was generated based on the detected Hg° by the online mercury analyzer. The online mercury analyzer can only detect the Hg° species in the gaseous phase because of its measurement limitation. Therefore, the difference between MC and MC T was considered the escaping oxidation state of mercury that were produced during removing Hg°. Such small difference, such as less than 5 μg/g, indicated that the amount of the escaping oxidation state of mercury was very limited and the majority of the mercury were absorbed as Hg° by the Fe 3 O 4 -NH 2 -Pd sorbent.
Performance of the regenerated Hg° sorbent. In practice, the utilization of this mercury sorbent could be improved if it can be recycled by regeneration. Therefore, the performance of the multi-regenerated sorbent was included in this work. It has been found that the used magnetic materials can be easily regenerated by the external magnetic force. The used sorbent had been regenerated at 300 °C under N 2 for 2 hrs before it was used for the second Hg° removal test. The results are shown in Fig. 4. It can be seen that, for the first regenerated sorbent, the Hg° removal efficiency dropped from 96% to only 72% after 250 min. By contrast, it dropped linearly to 56% for the second regenerated sorbent with the same period of time. The unknown strategy of how to improve the performance of the regenerated Fe 3 O 4 -NH 2 -Pd sorbent will be an angle for the future work.   Stabilization of the Hg in the sorbent phase. The stabilization of the Hg in the sorbent was tested based on the following experiments. Firstly, the Hg content of the used Fe 3 O 4 -NH 2 -Pd sorbent after two days washing by water is 30.3 μg/g, corresponding to the data of 30.1 μg/g in Fig. 3. Also, the Hg content in the sorbents after being stored for more than one year at normal pressure and temperature is 29.7 μg/g, therefore, the Hg in the adsorbed phase can be considered to be stable.    Fe 3 O 4 -NH 2 -Pd sorbents before and after the Hg° removal measurements. Figure 6 indicates that a strong peak at 583 cm −1 was assigned to the vibration of Fe-O bonds, which demonstrated the existence of the iron oxides. The peaks at 1070 cm −1 , 1624 cm −1 and 3440 cm −1 correspond to C-N stretching vibration, N-H deformation vibration and N-H stretching vibration 29 , indicating the existence of the -NH 2 group on the Fe 3 O 4 -NH 2 sorbent. Therefore, it can be confirmed that the magnetic nanocrystals had been functionalized with amino groups in the synthetic process. There was no distinct variation after immobilization of palladium on the Fe 3 O 4 -NH 2 surface except that the peak intensity became slightly weak. The FTIR spectrum of the used Fe 3 O 4 -NH 2 (line c) and Fe 3 O 4 -NH 2 -Pd (line d) sorbents were shown. It was found that the signals (curves) did not show significantly change after the Hg° removal except for the peak intensity. The weak peak intensity of the Fe 3 O 4 -NH 2 -Pd after two regenerations (line e) indicated that the structure of Fe 3 O 4 -NH 2 -Pd changed partly. This is considered one of the reasons of the lower Hg removal efficiencies for the regenerated sorbent.

Characterization of Fe
Analysis of SEM and TEM. The morphology and crystallography of the as-synthesized sorbents were characterized by SEM, and the structure was clearly revealed by TEM images. Figure 7a and b indicate the Fe 3 O 4 -NH 2 and Fe 3 O 4 -NH 2 -Pd particles were regular sphere and the diameters of those spheres were less than 100 nm. There was no Pd nanoparticles identified according to the SEM images, which provides a clear evidence that the Pd nanoparticles had successfully loaded onto the surface of the Fe 3 O 4 -NH 2 nanoparticles.
The TEM image in Fig. 7c shows that the synthesized Fe 3 O 4 -NH 2 nanoparticles were nearly monodisperse with an average diameter of 90 nm. Figure 7d shows that the Fe 3 O 4 -NH 2 -Pd showed the uniform TEM micrographs. The small Pd nanoparticles were highly dispersed around the surface of the magnetite. The overall morphology and the size of these particles did not vary obviously after the palladium was attached onto the surface of the magnetic nanoparticles. Besides, it can also be concluded that the particle size distribution of the Pd was centered within 7 nm. Figure 7e showed that the morphologies of the Fe 3 O 4 -NH 2 -Pd particles after two regenerations varied significantly compared to the fresh sorbent. It can be seen that the looser and larger aggregate structures were formed after two regeneration for the Fe 3 O 4 -NH 2 -Pd sorbents. Some Fe 3 O 4 particles were crushed and the Pd particles were partly aggregated, this may be the second reason that leads to the low Hg° removal efficiency for the Fe 3 O 4 -NH 2 -Pd sorbents after two regenerations.      9. The peaks corresponding to oxygen, iron, palladium and carbon were found in the survey spectra (Fig. 9a). Especially, the peaks assigned to sulfur were also observed in the used Fe 3 O 4 -NH 2 -Pd (Fig. 9a). It indicated that some sulfur species were absorbed on the surface of the Fe 3 O 4 -NH 2 -Pd.
XPS spectra of Pd 3d were showed in Fig. 9b. The binding energy of 335.75 eV (Pd 3d 5/2 ) and 341 eV (Pd 3d 3/2 ) for both the fresh and used Fe 3 O 4 -NH 2 -Pd were assigned to Pd°3 1 . According to the literature 32 , the binding energy of Pd° (Pd 3d 5/2 ) was 335.1 eV. However, it is found in this study that the Pd 3d 5/2 peaks shifted to higher values as a result of the presence of the surrounding positively charged ammonium groups. It implied the existing of -NH 2 group can stable Pd on the surface of the Fe 3 O 4 -NH 2 nanoparticles. The unreacted Pd° and Pd-Hg amalgam formed during the Hg° removal can be found in the used sample 8 .
The binding energy of 337.4 eV, 336.9 eV (Pd 3d 5/2 ) and 342.7 eV, 342.2 eV (Pd 3d 3/2 ) could assign to Pd 2+ . The Pd 3d 3/2 peaks also shifted to higher values due to the presence of ammonium groups 33 . Pd 2+ species in the fresh sorbent could be assigned to PdO that were from the oxidation of the Pd° by the lattice oxygen of Fe 3 O 4 . PdO can react with H 2 S to form PdS. Pd 2+ on the surface of the used and the regenerated sorbents may be assigned to PdS.
The intensity of the XPS spectrum peak reflects the content of the surface atom 34 . Table 1 showed the key elements contents on the surface for the fresh and used Fe 3 O 4 -NH 2 -Pd based on the results of the XPS spectra. The ratio of Pd°/Pd on the surface of the fresh Fe 3 O 4 -NH 2 -Pd was 71.17%, indicating that the elemental Pd was the main composition of the fresh sorbent. It was deduced from the Pd 2+ /Pd ratio on the surface of the used Fe 3 O 4 -NH 2 -Pd sorbents (79.08%) that the Pd 2+ species was the main compositions after the removal of Hg°. The ratio of the Pd°/Pd on the surface for the used Fe 3 O 4 -NH 2 -Pd decreased to 20.92%, accordingly. It indicated that the elemental Pd on the surface of the fresh sorbent was greatly oxidized to Pd 2+ species such as PdO or PdS during the Hg° removal. It was found in Fig. 9b that only Pd 2+ of Pd existed in Fe 3 O 4 -NH 2 -Pd after three regeneration cycles while the Hg° removal efficiency sharply decreased for the Fe 3 O 4 -NH 2 -Pd after two regeneration cycles (showed in Fig. 4c), it also demonstrated that Pd° was the main active component.
Some variations of the relative abundance between Fe 2+ and Fe 3+ species before and after removing of Hg° were elucidated by Fe 2p XPS spectra in Fig. 9c. The Fe 2p peak consisted of octahedral Fe 2+ (710 eV), octahedral Fe 3+ (712 eV), and tetrahedral Fe 3+ (714 eV) peaks 35 . These values were very close to those of magnetite (Fe 3 O 4 ) reported in the literature 35 Figure 9d shows the XPS spectra of S 2p on the surface of the used Fe 3 O 4 -NH 2 -Pd sorbent. The binding energy of 161.6 eV and 162.9 eV can be assigned to PdS 36 and HgS 37 , respectively. The binding energy of 164.0 eV belonged to elemental S 38 . This result proved that H 2 S could react with lattice oxygen and Fe 3+ in Fe 3 O 4 to form FeS, elemental S and H 2 O. Afterwards, the S can react with Hg to form HgS which is considered the Hg removal reaction. However, the active temperature range of the reactions to produce HgS was very limited, being efficient only in the range of 60-140 °C 39 since elemental S was volatized with the increasing temperature. This can explain that the high efficiency in Hg° removal over Fe 3 O 4 -NH 2 sorbent at 100 °C and low efficiency at 150 °C and 200 °C. However, for Fe 3 O 4 -NH 2 -Pd sorbent, the active temperature range of the Hg° removal was enlarged because of the loading of Pd. Pd and Hg can react to generate Pd-Hg amalgam (showed in Fig. 9b), resulting in the removal of Hg°. The content of sulfur on the surface of the used Fe 3 O 4 -NH 2 -Pd sorbent was 0.58%, indicating that there are abundant sulfur species such as elemental S, HgS, PdS and FeS after of the Hg° removal. Figure 9e presents the XPS spectra of Hg 4 f on the surface of the used Fe 3 O 4 -NH 2 -Pd sorbent. Because the binding energy of Hg° (Hg 4e 7/2 ) was around 99.9 eV 40 , it can be inferred that the binding energy of 99.7 eV can be assigned to Pd-Hg amalgam. Also, that of 102.5 eV was consistent with Hg 2+ compounds (HgS) 37 .   41 and we do some modification on it. Typically, a solution of 6.5 g 1,6-hexanediamine, 2.0 g anhydrous sodium acetate and 1.2 g FeCl 3 ·6H 2 O as a ferric source in 35 mL ethylene glycol was stirred at 50 °C to give a transparent solution. The solution was then transferred into a teflon-lined autoclave and then kept at 200 °C for 6 hrs. The magnetic nanoparticles were then rinsed with water and ethanol for several times to effectively remove the solvent and unbound 1, 6-hexanediamine, and then dried under vacuum at room temperature to obtain a black powder for further use. During each rinsing step, the nanoparticles were separated from the supernatant by using a magnetic force.

Mechanism of the Hg° removal of Fe 3 O 4 -NH
The Fe 3 O 4 -NH 2 -Pd nanoparticles were prepared according to the reported method 29 with some modification. 0.5 g of synthesized Fe 3 O 4 -NH 2 nanoparticles was placed in a 50 mL ethanol solution and then treated with ultrasonic for 0.5 hrs. This black suspension solution was mixed with 3.0 mM of a Pd(NO 3 ) 2 solution for 1 hrs with ultrasonic. Then the sorbent was reduced by an excess 0.1 M KBH 4 aqueous solution. It was slowly dropped into the above mixture with stirring. The solid sorbent was separated by magnet and was washed by deionized water after 2 hrs of reduction. The sorbent was dried at 45 °C under vacuum. The Pd loading amounts in the sorbent ranged from 0 to 10 wt %. Finally, all samples were pressed for tableting and then sieved at 40-60 mesh.
The performance of Hg 0 removal over the as-synthesized Fe 3 O 4 -NH 2 and Fe 3 O 4 -NH 2 -Pd sorbents was detected by a fixed-bed reactor. It includes four parts: Hg generation, gas mixture, a reactor and an online mercury analyzer (Lumex RA-915 M + Zeeman, Lumex-Marketing JSC, Russia) as the detection system. Hg vapor was generated by a Hg permeation tube (Valco Instruments Company Inc., America) immersed in a constant water bath maintained at 35 ± 0.5 °C. Hg vapor was brought into the evaluation system using ultrahigh purity N 2 as a carrier. The flow rate through the U tube was accurately controlled by a mass-flow controller. The simulated fuel gas consisted of 10 vol.% H 2 , 20 vol.% CO, 300 ppm H 2 S, 45 ± 3 μg/m 3 Hg 0 vapor, balancing gas N 2 (470 ml/ min) and carrier gas N 2 (500 ml/min).
A total of 0.50 g sorbent was placed in the horizontal quartz reactor (5.0 mm of the inner diameter) and then packed with quartz wool to support the sorbent layer and avoid its loss. Subsequently, the simulated fuel gas was switched into the reactor at the desired temperature. Hg vapor concentrations at the inlet and outlet of the reactor containing sorbents were monitored using a Lumex mercury analyzer. The removal efficiency (η) of Hg 0 was used to evaluate the performance of the sorbents for the capture of Hg 0 from coal derived fuel gas. η is calculated by the following formula: where C 0 and C 1 , μg/m 3 or ppm, are the concentrations of Hg 0 in the feed and effluent gases, respectively. Mercury content of the sorbent after evaluating is defined as mercury capacity (MC) and it can be directly detected by the pyrolysis accessories (PYRO-915+) of mercury analyzer. All the mercury species can be detected by the pyrolysis accessories of mercury analyzer. Theoretical adsorption mercury capacity (MC T ) of the sorbents can be calculated by the curve of the Hg 0 removal efficiency, the formulas is showed as bellow: T i i i 0 where η i is the mercury removal efficiency at t i which is the adsorption time in the i reactive time (min), C 0 is the initial concentration (μg/m 3 ) of Hg 0 in the feed gas, Q i is the flow rate of coal derived fuel gas, G is the weight of sorbent in the reactor (g).
After the mercury removal test, the used sorbents were regenerated by heating to 300 °C in pure N 2 carrier gas for 2 hrs. Several capture-regeneration cycles were conducted to evaluate the regeneration performance of Fe 3 O 4 -NH 2 -Pd sorbent.
The fresh and used Fe 3 O 4 -NH 2 and Fe 3 O 4 -NH 2 -Pd sorbents were determined by Fourier transform infrared (FTIR, Bruker Vertex 70) and the scan range is from 4000 cm −1 to 400 cm −1 .
Morphology and particle dispersion of the as-synthesized Fe 3 O 4 -NH 2 and Fe 3 O 4 -NH 2 -Pd were investigated by scanning electron microscopy (SEM) (Cam scan MV2300). The chemical compositions of the synthesized nanostructures were measured by EDS performed of SEM. Transmission electron microscopy (TEM) images were obtained on a Hitachi H-800 transmission electron microscope with an accelerating voltage of 200 kV.
X-ray diffraction (XRD) was employed to investigate the crystal structures of the synthesized sorbents. The instrument was a Rigaku Miniflex 600 diffractometer, fitted with a nickel-filtered Cu K α radiation source operating at a voltage of 40 kV and 100 mA. The scan rate was 8°/min in the range from 15° to 80°.
X-ray photoelectron spectroscopy (XPS) surface analysis was conducted to determine the elemental speciation and concentration on the surface of Fe 3 O 4 -NH 2 -Pd sorbents, using an ESCALAB 250 spectrometer (VG Scientific Ltd., UK) equipped with an Al Kα source (1486.6 eV, 150 W). Energy calibration was performed using C 1 s peak at 284.6 eV. No smoothing routine of data was applied to analyze the results.