Minimum fluidization velocity and reduction behavior of combusted iron powder in a fluidized bed

The fluidization and reduction behavior of micron-sized iron oxide powder, produced by iron combustion, is studied in a lab-scale cylindrical fluidized bed. The minimum fluidization velocity 𝑢 mf is found to stabilize above normalized static bed heights of 0.5 𝐻 ∕ 𝐷 (static bed height divided by the bed diameter). 𝑢 mf is measured as a function of temperature between 280 and 860 K for both H 2 and N 2 as fluidizing gas. The experimental results start to deviate from the Ergun correlation at temperatures above 560 K, both for N 2 and H 2 . A new correlation, taking the cohesive inter-particle solid bridge force into account, is proposed in this work to predict the minimum fluidization velocity at high temperature. Reduction experiments are carried out for a total time of 5 h at constant excess velocity with 50, 75 and 100 vol% of H 2 and temperatures between 623 and 823 K. Gradual defluidization occurs when the operating temperature exceeds 800 K. A maximum reduction degree of 61% is obtained at 807 K and 100 vol% H 2 .


Introduction
''Metal fuels'', metallic powders which are cyclically combusted and reduced, are promising carbon-neutral dense energy carriers for the future low-carbon economy [1][2][3][4][5][6][7].Efficient regeneration of the metal powder through chemical reduction with renewable energy is the key in determining the cycle efficiency.For iron powder, one of the most promising metal energy carriers, reduction using green hydrogen has the most potential [2].This process closely aligns with existing research on direct reduction of iron ores (DRI) with hydrogen for the steelmaking industry [8][9][10][11], as well as chemical looping combustion, which often employ iron powder as oxygen carrier [12][13][14].However, the steelmaking industry typically uses millimeter size pellets, in contrast to the micron-sized particles envisioned for the metal energy carrier cycle.Similarly, in chemical looping combustion, the metals are used to aid carbon capture and storage in fossil fuel combustion, while the metal energy carrier cycle uses renewable energy and thus is carbon-free.Our previous study investigated the intrinsic chemical kinetics of combusted iron powder with hydrogen using thermogravimetric analysis and a packed bed reactor [15].However, for larger quantities of powder, a fluidized bed is preferred above a packed bed reactor due to greater heat and mass transfer rates.
Fluidization of micron-sized iron(-oxide) powder at high temperature has, however, proven to be difficult due to the cohesiveness of this material [16][17][18][19][20][21][22][23].Zhong et al. [24][25][26] studied the fluidization behavior of iron and iron-oxide powders with different fluidizing gases.They observed sticking in beds with pure iron as well as in beds of iron oxide with both nitrogen and hydrogen as fluidizing gas.This sticking resulted in an increase in the minimum fluidization velocity at elevated temperatures.Secondly, they were able to determine a defluidization temperature of the bed.For nitrogen this temperature was around 923 K, while for hydrogen defluidization already occurred around 673 K.However, most of their experiments were performed using a non-isothermal method (fixed heating rate), which is different from typical industrial reactors (fixed temperature).They also used a relatively small fluidized bed reactor made of silica, and the possible effect of electrostatic forces was not considered.Mikami et al. [27] studied the mechanism of defluidization, concluding that it was due to particle-to-particle neck growth.They observed that the sticking started to affect the fluidization behavior from 773 K.However, they performed minimum fluidization experiments at room temperature after 1 h of fixed bed heat treatment (sintering) which might overpredict the influence of sintering in a real fluidized bed (where particles are in constant motion).They also used pre-reduced solid steel shot particles, which are different from combusted iron.Spreitzer et al. [28,29] investigated the fluidization and reduction behavior of different iron ores in the temperature range of 873-1073 K.They observed that fluidization was possible up to 1073 K, depending on the specific ore composition.

Solid
They limited their particle size range to 250-500 μm, significantly larger than those used in the metal fuel cycle (10-100 μm).Thus, the cohesiveness of iron/iron-oxide powder with increasing temperature poses a significant challenge in the design of the fluidization process for iron fuel regeneration.The minimum fluidization velocity  mf is one of the most important fundamental parameters in the analysis, characterization and design of fluidized beds [30].It determines the gas velocity above which fluidization starts to occur, and in combination with the excess velocity ( e =  0 −  mf , with  0 the superficial gas velocity), appears in most correlations describing the fluidization state.Many expressions exist for predicting  mf , mainly based on the equation derived by Ergun [31].However, it is also known that these correlations do not match experimental results both at high temperature as well as for small (Geldart C) particles [24,32,33].The cause of this is often attributed to cohesive forces, which are not taken into account in the Ergun equation.Since the minimum fluidization velocity and the related excess velocity determine properties such as solid mixing, bubble size and voidage, they also (in combination with the intrinsic chemical kinetics) determine the conversion that can be obtained in fluidized bed reactors.This also means that experiments trying to understand chemical reactions by varying gas composition and temperature, should optimally try to keep the excess velocity the same at different operating conditions.This can only be done with an accurate prediction of  mf .
Therefore, in this work,  mf of combusted iron powder is measured experimentally in a lab-scale cylindrical fluidized bed.In Section 2, the experimental setup and procedure are explained.The derivation of a new correlation for  mf at high temperature is also given, which incorporates the cohesive solid bridge force into the Ergun equation.In Section 3, the effect of static bed height, temperature and fluidizing gas composition on the minimum fluidization velocity is studied and results are compared with the Ergun equation and the correlation derived in Section 2. Finally, in Section 4, results of reduction experiments conducted at constant excess velocity are presented and discussed.

Materials
The powder used for this research is the product of high-purity iron powder (Rio Tinto ATOMET95), combusted with air in a pilot scale industrial burner.In this combustion process the particles are dispersed in air and injected into the burner.A non-premixed propane pilot flame is used to stabilize the flame.The combusted particles are captured in a long horizontal cooling section, followed by a cyclone.Powder (Pometon MT63) combusted using the same burner has been studied in great detail by Choisez et al. [34], who analyzed the microstructure of the combusted powder in order to better understand the combustion process.They found their powder to be mostly spherical, consisting of a mixture of magnetite and hematite.The powder used in this study is recovered from the cooling section of the burner.It is characterized using scanning electron microscopy (SEM), particle size analysis (PSA) and X-ray diffraction (XRD).To remove ultra-fine particles and large agglomerates for the fluidization experiments, the powder is first sieved between 32 and 100 μm, using a RETSCH AS 200 vibratory sieve shaker and LINKER woven wire mesh sieves.The sieving resulted in 15 wt% of the initial combusted powder to be removed.Fig. 1 shows SEM images (FEI ESEM Quanta 600 FEG, 30 kV, 3.0 spot size) of the sieved powder.The combusted particles are mainly spherical, but some of them are tulip shaped hollow shells, resulting from micro-explosions (see center image in Fig. 1 for an example), as already observed in previous studies [34][35][36][37][38].The particles show different surface roughnesses, most likely resulting from different combustion profiles.The bulk density of the powder is determined using a measuring cylinder to be 2912 kg/m 3 , resulting in a particle density of 4601 ± 58 kg/m 3 , assuming close random packing.The solid density, measured using a pycnometer (AccuPyc II 1340 V3.00), is 5267.4kg/m 3 .XRD analysis (Brukes D8 Advance A25-X1, Co target, 35 kV) suggests a composition of 60wt% magnetite (Fe 3 O 4 ) and 33wt% hematite (Fe 2 O 3 ) and 7wt% iron.Fig. 2 shows the (volume-based) particle size distribution and circularity data measured by dynamic image analysis (Bettersizer S3 Plus).The particles have a Sauter mean diameter  32 of 56.63 μm and are highly spherical.The Sauter mean diameter is used as the effective particle size in the minimum fluidization velocity correlations as it gives the same surface area for the same total bed volume [30].
The hydrogen and nitrogen gases used are provided by Linde Gas Benelux and have a purity of N5.0.

Fluidized bed reactor
The experiments are conducted in a 3D fluidized bed designed in an earlier project [39].The reactor is made of stainless steel (type 316) with an inner diameter of 80.9 mm and a height (above distributor) of 750 mm.Fig. 3 shows a schematic and photograph of the setup.A three-zone tubular furnace (Carbolite GZF 12/-/546) is used to heat up the fluidized bed and preheat the inlet gas.Three thermocouples are used to control each zone and one central thermocouple is inserted into the bed from above (TC Direct, type N).The central thermocouple is placed 45 mm above the distributor plate.The pressure drop over the bed (including the distributor plate) is measured using a differential pressure transmitter (Nöding PD40, 0-400 mbar).The gas flows are set using mass flow controllers (Bronkhorst EL-Flow F201-CV/AV), where a mixture of hydrogen and nitrogen can be used as fluidizing gas.A second independent nitrogen supply is used to dilute the effluent gas to below the lower explosion limit of hydrogen.The distributor plate consists of a quartz fiber membrane filter (Merck Millipore AQFA09050) sandwiched between two perforated stainless steel plates (807 orifices, 1.5 mm orifice diameter with triangular pitch).The acquisition frequency of the system is 1 Hz.

Experimental procedure
Before each experiment the sieved powder is inserted in the cold reactor.The filled bed is then fluidized to break down any internal structure, such as residual clusters and stratification. mf is determined at room temperature using nitrogen by incrementally increasing the superficial gas velocity until the pressure drop stays constant, and subsequently decreased back to zero.The reactor is then heated up (6 K/min) to the desired bed temperature while the powder is fluidized using nitrogen.After the desired temperature is reached, a similar  mf experiment (as explained above for room temperature) is carried out using the desired fluidizing gas composition (H 2 :N 2 ).After the experiments are performed the reactor is allowed to cool down naturally.Nitrogen is again provided as fluidizing gas.The quartz fiber membrane filter is replaced after each experiment.
For the reduction experiments, the powder is inserted in the cold reactor and is subsequently heated under fluidizing conditions.When the desired temperature is reached, the hydrogen and nitrogen gas flow rates are set in order to reach the desired excess gas velocity and composition.After 5 h the gas flow is switched back to nitrogen and the reactor is allowed to cool down naturally under fluidizing conditions.The powder is weighed before and afterwards using a digital scale (Mettler Toledo PG5002-S).

Data analysis method
The minimum fluidization velocity is determined in the standard fashion by the crossing point of straight lines fitted to the static regime and the fluidized regime using a decreasing gas velocity method.The static regime is fitted with a linear line with a free origin ( = ⋅ 0 +) to account for sensor hysteresis/offset, while a horizontal line ( = ) is used to fit the fluidized regime.A typical graph of the pressure drop versus the superficial gas velocity and the determination of  mf is shown in Fig. 4. The pressure drop  is normalized using the hydrostatic bed pressure: with  the cross sectional area of the reactor,  the powder bed mass and  the gravitational constant.The reason that the bed in fluidized state does not reach the hydrostatic pressure is most likely caused by the dead zone above the distributor plate (between the holes and in the wall-distributor plate edge).The static bed height  is calculated using the particle density mentioned in Section 2.1 and a void fraction  mf of 0.4.This static bed height is subsequently normalized to the bed diameter .The pressure drop over the distributor plate, measured using an empty bed, is subtracted from the measurement data.
The reduction degree  is calculated by dividing the mass loss obtained during a reduction experiment by the theoretical mass loss, based on a full conversion of the initial powder to 100 wt% iron:

High temperature fluidization model
Traditionally,  mf can be determined by matching the pressure drop of a gas through a packed bed (of height ) following from the Ergun equation [31], with the particle bed mass: in which  mf is the bed void fraction at minimum fluidization,  g is the dynamic viscosity of the gas mixture,  32 the Sauter mean diameter of the powder,  g the density of the gas mixture,  s the density of the solid and  the gravitational acceleration.Xu and Zhu [32] suggested an improved balance for cohesive particles by incorporating inter-particle forces in the packed bed force balance: in which the third term is due to the tensile stress of the particle bed, following the expression by Rumpf [40]: where  c is the sum of all inter-particle cohesive forces acting between two particles in a bed.Multiple different interparticle forces can exist during fluidization [41].In the case of metallic particles at high temperature, the dominant cohesive force is the formation of a solid bridge via the surface diffusion mechanism [42,43].We can use the equation derived by Kuczynski [44] to calculate the solid bridge force between two particles: in which  sb is the tensile strength of the neck (with radius ),  is the surface free energy,  is the lattice spacing,  is the Boltzmann constant,  the temperature,  is the curvature radius of the particle (which is 1 2  32 for smooth particles) and  c is the contact time. s is the solid state surface self-diffusion coefficient of the material, which can often be described using an Arrhenius expression: in which  0,s is the frequency factor,  a the activation energy and  the universal gas constant.The contact time during fluidization is related to the particle size and the characteristic velocity (≈  mf ) [19]: (11) in which  is a factor, between 0 and 1.Values of 0.1 and 0.15 have been used by other researchers [19,45].
The gas viscosity  g of a mixture of hydrogen and nitrogen is determined using the equation derived by Wilke [46]: with with  i is the molar fraction,  i the molar mass,  g,i the dynamic viscosity of a pure component.The density of a gas mixture is: The density and viscosity of pure hydrogen and nitrogen gas at different temperatures are determined using NIST polynomials [47], which are provided in Appendix A. Although the density change has a negligible effect on the calculation of  mf compared to the viscosity change for the powder size used, both are included in the analysis for completeness.
For material properties, values for magnetite Fe 3 O 4 are used, since (1) the particles mainly consist of magnetite, and (2) the solid state diffusivity of iron in hematite is known to be much lower than in magnetite [48].Therefore it can reasonably be assumed that the magnetite phase is the main cause of sintering.
For the lattice spacing  of magnetite, we can use the correlation from Gorton et al. [49]: results obtained by Hidaka et al. [50], by fitting a line through their data: For the bed voidage at minimum fluidization  mf a constant value of 0.4 is assumed, which is estimated based on the minimum fluidization velocity at room temperature.It also matches the voidage for dense packing of spherical particles.Although some correlations exist for the change of bed voidage with temperature, the results are inconclusive and material dependent [30].Furthermore, although  mf is present in the cohesive term (Eq.( 7)), the influence of the voidage on  mf still decreases with temperature (when the cohesive term becomes important).
A summary of all other properties used can be found in Table 1.
Due to the fact that  mf is present in the cohesive term (Eq.( 11)), no trivial explicit form for  mf exists and Eq. ( 6) is therefore solved numerically.

Influence of bed height
Fig. 5 shows  mf as a function of static bed height, determined at room temperature using nitrogen.It can be observed that  mf is mostly independent of bed height, which is in agreement with Eq. (3).However, at low static bed height,  mf seems to be slightly underestimated.The reason for this might be that shallow bed effects, such as channeling, occur.There also seems to be more variation between experiments at low static bed heights, which might indicate slight variations in the mean particle size between batches.The larger error bars at low H/D are due to the higher uncertainty on the linear slope for the static regime.For the  mf measurements at high temperature as well as the reduction experiments, a bed height of ∕ ≈ 0.65 (750 gram,  mf = 0.4) is used, as a compromise between accuracy and the limited availability of combusted powder.

Influence of temperature and gas composition
Fig. 6 shows examples of  n −  0 graphs of nitrogen and hydrogen around 300 and 700 K.It can be observed that a higher temperature leads to a steeper slope of the pressure drop, when increasing the gas flow from zero to above the minimum fluidization velocity ( mf,increasing ).This is similar to the results obtained for small, cohesive particles (Geldart C) [54].Fig. 7 shows  mf as a function of bed temperature for hydrogen and nitrogen, while predictions by the Ergun correlation (Eq.( 3)) are represented by the dashed lines.Predictions using the new correlation incorporating the solid bridge force (Eq.( 6)) are represented by solid lines.
From the experimental results we can observe that with increasing temperature, the minimum fluidization velocity initially decreases, due to the changes in viscosity of the gas.However, from a temperature of 560 K onward (marked by the dotted black line in Fig. 7), the minimum fluidization velocity actually increases with Similar observations were also Zhong et al. [24] from experiments using slightly larger iron powder (74-149 μm).We can see that the Ergun correlation, using a void fraction of 0.4, accurately predicts the minimum fluidization velocity of both pure nitrogen and pure hydrogen at temperatures up to 560 K.However, above this the Ergun correlation fails to predict the experiments.However, the new correlation, which takes into account the cohesive solid bridge force, does match the experimental results quite well.It should be noted that, due to the relatively large bed mass (750 g of iron oxide powder) and the relatively low H 2 gas flow rate, far less than 1% conversion is expected to occur within one minimum fluidization experiment.Therefore, the prediction using magnetite instead of iron as material, is still valid.
Using this new predictive model for the minimum fluidization velocity (Eq.( 6)), it is possible to perform experiments at equal excess velocities, and therefore equal fluidization behavior, by using: In the following section, results of reduction experiments at equal excess velocity are presented.

Reduction behavior
Isothermal reduction experiments are carried out between 623 K and 823 K, with hydrogen concentrations of 50, 75 and 100 vol% H 2 .All reduction experiments are performed at an excess velocity  e of 26 mm/s, equaling superficial gas velocities in the range of  0,N 2 ( ) = 5-9  mf ,N 2 ( ) and  0,H 2 ( ) = 3-5  mf ,H 2 ( ).The total reduction time of each experiment was set to 5 h.

Fluidization behavior
Two different types of fluidization behavior occurred during the experiments.Examples of the pressure drop and temperature profiles of the two types are displayed in Fig. 8.The filtered signal (red line) is produced using the 'ischange'-algorithm in MATLAB and is meant as a visual aid only.At temperatures below 800 K (Fig. 8(a)), the bed stays fluidized during the entire 5 h, with only a minor decrease of the pressure drop due to the reduction process (decrease in particle density).At the experiments performed above 800 K, independently of the hydrogen concentration, defluidization occurs.Fig. 8(b) shows  that the pressure drop steadily decreases until the bed is completely defluidized at t ≈ 150 min.The remaining pressure drop is due to the resistance of the agglomerated bed.This type of defluidization matches with the classification of ''gradual defluidization'' described by Lei et al. [19].They visually observed an agglomerate forming at the distributor plate which subsequently grew upwards.In their case the gradual defluidization occurred at relatively low superficial gas velocities and bed temperatures between 923 and 1023 K.The higher temperature at which gradual agglomeration occurred, compared to the one found in this work (823 K), might be caused by the larger particles they used (106-150 μm versus 32-100 μm in this work).
The reason that defluidization occurs, even though the superficial gas velocity was above the predicted minimum fluidization velocity, is due to particles sintering to the wall of the reactor and in the dead zone just above the distributor plate, where the local gas velocity is known to be much lower than the minimum fluidization velocity [19,27].In these locations the contact times between particles are significantly longer than calculated by Eq. ( 11), leading to sintering.

Reduction degree
Fig. 9(a) illustrates the reduction degree obtained after 5 h at various temperatures and hydrogen concentrations, calculated using Eq. ( 2).It can be observed that the reduction degree increases both linearly with temperature.Interestingly, the highest reduction for all three hydrogen concentrations is obtained at the experiments above 800 K, even though defluidization of the bed occurs.The highest reduction degree of 61% is reached at a temperature of 807 K and a hydrogen concentration of 100 vol%.Qualitative X-ray diffraction indicated that all reduced powders consisted of a ratio of magnetite and iron.Fig. 9(b) displays the H 2 utilization, defined as the amount of H 2 used for the reduction divided by the total amount of H 2 supplied.The utilization degree increases linearly with reduction temperature and is nearly independent of hydrogen concentration, indicating that reduction degree is linearly dependent on the hydrogen concentration.The small difference above 800 K is most likely caused by differences in defluidization time, which are slightly shorter at higher hydrogen concentrations.

Morphological observations
Figs. 10 and 11 show SEM images of the powder after reduction.Fig. 10 shows the effect of gradual defluidization on the powder.Only images for reduction at 75 vol% H 2 are shown, since the effect of hydrogen concentration on the agglomeration was negligible.Images for the other cases can be found in the supplementary We can clearly that 822 the powder consists mainly of agglomerates, while at 722 K hardly any agglomerates are formed.This difference in powder morphology matches the pressure drop profiles in Fig. 8. Above 800 K gradual defluidization occurs, caused by this agglomeration.Below 800 K the decrease in pressure drop was solely caused by the reduction process.
Fig. 11 shows a close-up of a particle after reduction.A dendritic iron structure can be observed, leading to a particle surface which is highly porous.This matches with results obtained in previous studies on iron oxide reduction [8,15,[55][56][57].Only the results at 807 K and 100 vol% H 2 are presented here, while the images of the other cases are provided in the supplementary material.Interestingly, the difference in surface morphology between the particles reduced above 700 K was negligible, even though the reduction degree obtained varied significantly ( = 0.2 − 0.6).This matches with the kinetic result obtained in a previous study, which indicated that the reaction progresses via a shrinking core principle (phase boundary controlled), where the conversion of iron oxide to iron starts at the surface and grows inwards, creating an oxide core with an iron shell surrounding it [15].

Conclusions
The fluidization and reduction behavior of combusted iron powder is investigated in a lab-scale fluidized bed.The minimum fluidization velocity is measured for different bed heights, temperatures and fluidizing gases: • The influence of the static bed height on the minimum fluidization velocity at room temperature is insignificant for normalized static bed heights ∕ above 0. 5 h reduction experiments at equal excess velocity were carried out between 623 and 823 K and at hydrogen concentrations of 50, 75 and 100 vol%: • Gradual defluidization occurs at temperatures above 800 K, resulting in the formation of many agglomerates.• A maximum reduction degree of 61% is obtained at 807 K and 100% hydrogen, even though defluidization occurs.• The surfaces of the particles became highly porous during reduction.Since the particle surface morphology was equal for all experiments, even though the obtained reduction degree varied, it suggests that the surface already fully reduces to iron in an early stage of the reduction process, confirming a shrinking core principle.
This work contributes to the design of high temperature fluidized beds using metallic powders (such as for chemical looping combustion).It also supports the research being carried out on green steelmaking (DRI), iron fuel regeneration and the metal fuel cycle.

Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Fig. 1 .Fig. 2 .
Fig. 1.Scanning electron microscope (SEM) images of the iron oxide powder.The red circle in the center image marks an example of a tulip shaped particle.(For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Fig. 3 .
Fig. 3. Schematic and photograph of the lab-scale high temperature fluidized bed reactor setup.DP = differential pressure transmitter, TC = thermocouple, MFC = mass flow controller.Photograph courtesy of Bart van Overbeeke.

Fig. 4 .
Fig. 4. Typical graph showing normalized pressure drop over superficial gas velocity.Error bars on the  mf display a 95% confidence interval.

Fig. 5 .
Fig. 5. Minimum fluidization velocity  mf for different normalized static bed heights, determined at room temperature using nitrogen.The error bars on the experiment are 95% confidence interval.The gray area defines the range of  mf , according to Ergun (Eq.(3)), due to variation in ambient temperature between the experiments.

Fig. 6 .
Fig. 6.Normalized pressure drop over superficial gas velocity for N 2 and H 2 for low and high temperature.

Fig. 7 .
Fig. 7. Minimum fluidization velocity as function of temperature for N 2 and H 2 .Vertical error bars show 95% confidence intervals, while horizontal error bars represent the temperature span that occurred during each experiment.The colored areas are the results of using different values for  in Eq. (11).(For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.) C.J.M.Hessels et al.

Fig. 8 .
Fig. 8. Examples of the two types of pressure drop and temperature profiles occurring during the reduction experiments.

Fig. 9 .
Fig. 9. (a) reduction degree and (b) H 2 utilization as a function of temperature and for three different hydrogen concentrations, obtained after 5 h reduction experiments.Horizontal error bars represent the temperature span, while vertical error bars are due to the error on the initial composition.

Fig. 10 .
Fig. 10.SEM images of the powder after reduction experiments showing agglomeration.
5. • A new model for predicting the minimum fluidization velocity at high temperature is derived.The proposed model takes into account the cohesive inter-particle solid bridge force.• The minimum fluidization velocity at elevated temperature is reported for nitrogen and hydrogen: (1) The measured minimum fluidization velocity starts to deviate from the Ergun correlation at temperatures above 560 K; (2) The high temperature behavior of the minimum fluidization velocity is well captured by the proposed model in this work, taking the solid bridge force into account.
(15)5)Literature on the tensile strength of magnetite is limited, especially at elevated temperature. Trefore, an estimate is made based on the C.J.M.Hessels et al.