Evaluation of the Effects of Fluidization Conditions on Hydrogen Reduction in Manganese Ore Fines

Abstract

Hydrogen prereduction of two manganese ores fines was investigated under varied operating conditions in a fluidized bed. The manganese ores used in this study are the Zambian ore and the South African Nchwaneng ore from the Kalahari region. The samples were milled and sized before they were characterized with regard to sphericity, Inductively Coupled Plasma Optical Emission Spectroscopy (ICP-OES) chemical analyses, X-ray diffraction (XRD) analyses and Scanning Electrons Microscope (SEM) analyses. Prereduction experiments were conducted in a laboratory scale fluidized bed with the parameters of interest being minimum fluidization velocity, terminal velocity, elutriation, average bed voidage, residence time, temperature, intrinsic ore properties and cohesive adhesion. Experiments for the determination of fluidization velocity and terminal velocity were conducted at both ambient temperature and elevated temperature ($500 °$C, $550 °$C, $600 °$C, $700 °$C, $800 °$C and $900 °$C), and for varied sample masses (100 g, 300 g and 700 g) and varied particle-size ranges (200–$300 \mathrm{\mu }$m, 300–$425 \mathrm{\mu }$m, 425–$500 \mathrm{\mu }$m and 500–$600 \mathrm{\mu }$m). The experimentally observed minimum fluidization velocities for particles size groupings of [+106–$200 \mathrm{\mu }$m], [+200–$300 \mathrm{\mu }$m], [+300–$425 \mathrm{\mu }$m], [+425–$500 \mathrm{\mu }$m] and [+500–$600 \mathrm{\mu }$m] as well as the mix (20 wt% of each) was comparable with the theoretical minimum fluidization velocity. The fluidized bed was heated to a desired temperature at a rate of $10 °$C/min under argon whilst logging the pressure drop across the bed with increasing temperature. The convectional cooling during the introduction of cold hydrogen as well as the net energy of endothermic and exothermic chemical reactions were observed to result in a temperature drop in the order of 100 to $250 °$C. Thermal mineral transformation under argon was observed to yield iron manganese oxide in the order of 15 to 30 wt/wt%. Prereduction was conducted using hydrogen gas at a desired temperature and terminal velocity. Reduction extent was observed to increase with the increasing temperature and residence time. Increasing reduction temperature beyond $700 °$C was not observed to improve reduction, whereas longer residence time (of up to 40 min) was observed to favor the formation of iron manganese oxide, iron manganese and manganosite. For hydrogen prereduction experiment conducted at $900 °$C, the reactor was observed to be brittle after the experiment. Cohesive adhesion was observed to be more pronounced at $900 °$C.

1. Introduction

Manganese is considered to be a strategic metal with a wide range of industrial applications that are crucial to our daily lives, for example, alloying, battery electrons, medicinal needles, chemicals, etc. [1,2,3,4,5].

The global decline in the availability of traditional lumpy manganese ores and the concomitant costs coupled with environmental regulations of agglomeration processes (e.g., sintering, pelletizing, and coking) and carbothermic smelting necessitate investigations towards alternative processes. In recent years, green prereduction and smelting processes for ores has gained more traction. For the companies generating sand-size fines and companies employing DC arc furnace technologies, green fluidized bed prereduction technologies could be attractive.

The thermodynamic requirement for the traditional carbothermic reduction in manganese ores is high temperature (temperatures greater than $800 °$C) for carbon gasification [6]. In pursuit of lowering energy consumption during manganese ore reduction, various workers have contrived several alternative methods, e.g., hydrogen reduction [7], carbon monoxide reduction [8,9,10], methane reduction [11], FeP slag reduction [12], aluminium reduction [13], sulphur reduction [14,15], ferromanganese silicide [16] and so forth.

The technologies commonly employed for alternative methods include pre-heating in a shaft furnace or rotary kiln with furnace off-gas; calcinations in a shaft furnace or rotary kiln; nodulizing; sintering ore fines or pellets in a shaft furnace, travelling grate (linear or carousel configurations), grate-kiln or steel belt sinter plant; etc. [1,17]. Gordon et al. (2018) [17] compared and ranked these technologies with regard to technical risk; product quality; ore sizing; need for ore pre-treatment; typical plant capacity; consumptions of fuel and electrical power and CAPEX.

High efficiency of heat and mass transfer and accelerated chemical reactions rates have been reported as some of the technical advantages of using fluidized bed technologies. Gou et al. (2020) [18] and Crawford et al. (1995) [19] conducted a study on a two-stage process that involved pre-reduction in ferromanganese sand concentrates (produced from tailings at BHP’s GEMCO mine) in a fluidized bed as a first stage. A hot gas consisting of CO and ${\mathrm{H}}_2$ was used to pre-reduce the sample in this stage. The second stage involved DC arc furnace smelting. Their investigation was demonstrated on the bench scale for the separate stages; however, their scope did not include investigating the effect of hydrogen pre-reduction on fluidized bed behavior.

The aim of the current study is to investigate the influence of various experimental conditions on the hydrogen reduction in two manganese ores in a fluidized bed. The primary conditions of interest are temperature, intrinsic properties of ores, sintering and residence time.

2. Materials and Methods

2.1. Materials

The two manganese ores (namely Zambian ore and Nchwaneng ore) used in this study were supplied to Mintek by Assmang Proprietary Limited, cato Ridge South Africa.

The Zambian ore originates from the NorthWestern part of the Zambia and the Nchwaneng ore originates from the Kalahari district of South Africa. The samples were crushed, milled and sized to varied particle-size ranges, namely 200–$300 \mathrm{\mu }$m, 300–$425 \mathrm{\mu }$m and 425–$500 \mathrm{\mu }$m.

2.2. Characterization

2.2.1. Structural Properties

Particle structural properties such as diameter, sphericity and convexity are important in determining the fluidization properties of particles. Particle shape descriptions were determined using the Olympus Streams Essential Image Processing software, version 2.4.2. Images acquired from the SEM were processed using the software to determine the particle mean diameter, sphericity, convexity and elongation of the particles in the samples. The observed results are summarized in

Table 1. Sphericity and other structural properties of raw Nchwaneng ore particles.

Sample	Area (µm²)	Min (Diameter) (µm)	Equivalent Circular Diameter (µm)	Phase ID	Sphericity	Elongation	Aspect Ratio	Max (Diameter) (µm)	Mean (Diameter) (µm)	Perimeter (µm)	Shape Factor	Convexity
−600 + 500 µm	17,645.45	37.89	53.79	1.00	0.77	1.31	1.38	76.19	51.30	225.37	2.23	1.00
	7342.47	17.85	25.94	1.00	0.73	712.42	1.91	36.45	24.31	113.72	2.73	0.99
	6518.40	17.19	28.56	1.00	0.71	708.52	1.50	46.22	26.02	122.37	3.22	1.00
−500 + 450 µm	7641.97	26.17	38.72	1.00	0.67	2039.33	1.50	56.48	36.31	155.35	2.39	1.00
	20,083.59	56.58	89.08	1.00	0.44	3900.72	1.94	128.97	83.59	380.89	2.17	0.99
	24,265.5	64.8	100.22	1.00	0.45	2637.19	1.93	141.01	94.87	425.94	2.01	0.99
−450 + 300 µm	5898.03	26.17	41.34	1.00	0.58	1.62	2.07	61.02	38.98	186.97	1.61	0.99
	5652.78	24.14	37.16	1.00	0.62	301.80	2.01	53.96	35.02	166.92	1.77	0.99
	5877.38	27.95	43.57	1.00	0.50	2393.35	2.33	63.79	41.01	198.36	1.96	0.99
−300 + 200 µm	7353.34	32.62	52.47	1.00	0.64	1042.72	1.67	78.74	49.28	229.44	1.39	1.00
	4456.58	17.97	32.01	1.00	0.67	2508.51	1.55	51.29	29.09	141.98	2.30	1.00
	4313.82	21.18	29.86	1.00	0.70	3761.76	1.37	40.87	28.40	116.79	2.67	1.00
−200 + 106 µm	1231.8	11.50	17.09	1.00	0.68	2242.2	1.42	22.97	16.08	65.87	2.71	1.00
	4058.7	27.41	42.72	1.00	0.55	1499.6	1.86	60.69	40.62	188.61	1.55	1.00
	1469.9	12.30	18.80	1.00	0.67	1679.4	1.47	26.27	17.64	77.76	2.55	1.00

.

These structural properties were used in evaluating fluidization parameters such as minimum fluidization velocity, terminal velocity, disengagement velocity, etc.

Micromeritics Accupyc II 1340 Gas Pycnometer was used to measure the samples’ skeletal density. The pycnometer measures the volume of the solid discrete particles and their inaccessible pores with accuracy of 0.2% and uses helium gas as a medium at an outlet pressure of 1.34 bars in a sample cell. The samples were first dried under vacuum at $105 °$C prior to loading them into a sample cell (filling it to 75% volume). The sample cell was placed in the sample compartment, flushed with helium gas and degassed automatically to a pressure of 10 µm Hg. Experiments were conducted in triplicates and averages of three independent results were used, with error calculations at 95% confidence level. The skeletal densities obtained for Zambian ore and Nchwaneng ore were $4.3817 {\mathrm{g/cm}}^3$ and $4.3918 {\mathrm{g/cm}}^3$, respectively.

2.2.2. Chemical Analyses

Both raw materials and prereduction products were subjected to chemical analyses, which follows Mintek internal methods (ICP_FEMN and XRF_Q_SCAN) and with reference material SARM16. The obtained results are summarized in

Table 2. Respective chemical analyses of raw Nchwaneng and Zambian manganese ores.

	ICP_FEMN							XRF_Q_S_SCAN
	${\mathrm{Al}}_2{\mathrm{O}}_3$	CaO	${\mathrm{Cr}}_2{\mathrm{O}}_3$	${\mathrm{Fe}}_2{\mathrm{O}}_3$	MgO	MnO	${\mathrm{SiO}}_2$	Al	Ca	Fe	K	P	Si	Ti
	%	%	%	%	%	%	%	%	%	%	%	%	%	%
Nchwaneng ore	0.28	7.16	<0.07	17.4	1.97	56.80	3.84	0.32	5.32	12.95	0.05	0.04	1.93	0.02
Zambian ore	2.28	0.37	<0.07	1.98	0.27	59.40	8.80	1.13	0.10	1.99	0.41	0.05	3.13	0.19

. It is important to highlight that oxides/elements with concentrations of parts per million are not reported in Table 2, and that thus the percentages do not add to 100 percent. For mineral-assay reconciliations, the inductive coupled plasma optical emission spectroscopy (ICP-OES) results were used to validate the mineral abundance data obtained from X-ray diffraction (Section 2.2.3).

2.2.3. X-Ray Diffraction

The received pulverized samples were micronized to achieve particle size below $20 \mathrm{\mu }$m to ensure the acquisition of more representative data for mineral and element quantification. The samples were subjected to the Bruker D8 diffractometer (supplied by Karlsruhe, Germany) equipped with Fe-filtered Co K$\alpha$ radiation, and ran over a 2 theta range of 5–80 degrees with a step size of 0.02 degrees 2 theta. Minerals were initially identified using Bruker EVA^® software, version 4.2.2, while the quantification of minerals was performed using TOPAS^® software, version 4.2. Table 2 results were used to reconcile and to validate the mineral abundance data in

Table 3. Respective mineral proportions in as-received Nchwaneng and Zambian manganese ores.

Mineral	Ideal Chemical Composition	Nchwaneng Ore	Zambia Ore
		%	%
Braunite	${\mathrm{Mn}}^{2+}{{\mathrm{Mn}}^{3+}}_6$${\mathrm{SiO}}_{12}$	41.36	ND
Hausmannite	${\mathrm{Mn}}^{2+}{{\mathrm{Mn}}^{3+}}_2$${\mathrm{O}}_4$	3.62	ND
Kutnohorite	${\mathrm{Al}}_2{\mathrm{Mg}}_5$(${\mathrm{Si}}_3{\mathrm{O}}_{10}$)(OH${)}_8$	3.49	ND
Bixbyite	${\mathrm{Mn}}^{3+}$1.${{\mathrm{5Fe}}^{3+}}_{0.5}$${\mathrm{O}}_3$	14.16	ND
Manganite	MnO	6.96	ND
Marokite	${{\mathrm{CaMn}}^{3+}}_2$${\mathrm{O}}_4$	2.7	ND
Rhodochrosite	${\mathrm{Mn}}_{2+}$(${\mathrm{CO}}_3$)	<1	ND
Romanechite	(Ba,${\mathrm{H}}_2$O${)}_2$(${\mathrm{Mn}}^{4+}$,${\mathrm{Mn}}^{3+}$${)}_5$${\mathrm{O}}_{10}$	ND	56.1
Pyrolusite	${\mathrm{MnO}}_2$	ND	21.44
Cryptomelane	K(${\mathrm{Mn}}^{4+}$,${\mathrm{Mn}}^{2+}$${)}_8$${\mathrm{O}}_{16}$	ND	5.91
Hollandite	Ba(${\mathrm{Mn}}^{4+}$,${\mathrm{Mn}}^{2+}$${)}_8$${\mathrm{O}}_{16}$	ND	2.19
Jacobsite	${\mathrm{Mn}}^{2+}{{\mathrm{Mn}}^{3+}}_2$${\mathrm{O}}_4$	ND	2.04
Coronadite	Pb(${\mathrm{Mn}}^{4+}$,${\mathrm{Mn}}^{2+}$${)}_8$${\mathrm{O}}_{16}$	ND	1.26
Galaxite	(Mn,Mg)(Al,${\mathrm{Fe}}^{3+}$${)}_2$${\mathrm{O}}_4$	ND	<1
Ankerite	Ca(${\mathrm{Fe}}^{2+}$,Mg)(${\mathrm{CO}}_3$${)}_2$	ND	1.26
Lizardite	${\mathrm{Mg}}_3$(${\mathrm{Si}}_2{\mathrm{O}}_5$)(OH${)}_4$	ND	<1
Hematite	${\mathrm{Fe}}_2{\mathrm{O}}_3$	13.68	<1
Dolomite	CaMg(${\mathrm{CO}}_3$${)}_2$	7.87	ND
Calcite	${\mathrm{CaCO}}_3$	5.61	ND
Quartz	${\mathrm{SiO}}_2$	ND	8.27

ND = Not detected.

.

2.2.4. Scanning Electron Microscope

The Zeiss EVO MA15 Scanning Electron Microscope, Munich, Germany equipped with backscattered electron imaging and energy dispersive X-ray spectrometer (EDS), Karlsruhe, Germany was used. Back-scattered electron images were used to present the textural properties, while the EDS was used to determine the semi-quantitative elemental composition of the mineral phases in the samples.

Similar to Section 2.2.3, the received pulverised samples were micronized to achieve particle size below $20 \mathrm{\mu }$m to ensure the acquisition of more representative data for mineral and element quantification. Preparation for SEM analysis included mounting the samples in an epoxy resin, which was ground and polished. These mounts were carbon coated. The obtained SEM analysis results for the Nchwaneng ore and Zambian ore are given in Figure 1,

Table 4. Mineral phases in spots shown in the back-scattered electron images (Figure 1) from SEM-EDS analyses of Nchwaneng manganese ore.

Spot	C	O	Mg	Si	P	Ca	Mn	Fe	Phase
Spot 1	ND	26.18	ND	ND	ND	1.38	72.44	ND	Hausmannite
Spot 2	ND	29.42	ND	ND	ND	14.32	56.26	ND	Marokite
Spot 3	ND	36.04	ND	ND	ND	21.75	42.21	ND	Marokite–calcite mix
Spot 4	ND	45.00	ND	ND	ND	44.73	9.07	ND	Mn–calcite
Spot 5	ND	22.73	ND	ND	ND	2.06	75.21	ND	Hausmannite
Spot 6	ND	24.73	ND	2.68	ND	4.91	67.68	ND	Braunite II
Spot 7	ND	32.72	2.77	ND	ND	2.92	57.57	4.02	Mn–oxide
Spot 8	ND	29.85	ND	5.07	ND	5.59	59.49	ND	Braunite
Spot 9	ND	25.07	4.64	ND	ND	14.31	55.98	ND	Marokite–Kutnohorite mix
Spot 10	ND	40.60	6.00	ND	ND	32.79	20.61	ND	Marokite of Kutnohorite/Mn–calcite
Spot 11	ND	38.79	22.49	14.28	ND	5.63	18.81	ND	Mn–serpentine
Spot 12	ND	26.50	1.99	0.09	ND	39.67	31.75	ND	Kutnohorite
Spot 13	ND	25.76	ND	4.66	ND	1.75	67.84	ND	Braunite

ND = Not detected.

, Figure 2 and

Table 5. Mineral phases in spots shown in the back-scattered electron images (Figure 2) from SEM-EDS analyses of Zambian manganese ore.

Spot	0	Al	Si	K	Ca	Mn	Fe	Ba	Ti	Phases
Spot 1	19.89	ND	ND	ND	ND	59.19	1.29	19.63	ND	Romanechite
Spot 2	20.08	ND	ND	ND	ND	60.77	0.87	18.28	ND	Romanechite
Spot 3	53.26	ND	46.74	ND	ND	ND	ND	ND	ND	Quartz
Spot 4	53.26	ND	46.74	ND	ND	ND	ND	ND	ND	Quartz
Spot 5	20.85	ND	ND	ND	ND	66.53	ND	12.62	ND	Romanechite
Spot 6	21.34	ND	ND	ND	ND	69.69	ND	8.96	ND	Romanechite
Spot 7	34.21	13.93	9.66	1.81	ND	32.83	ND	7.56	ND	Mica–Romanecite mixed spectra
Spot 8	20.38	ND	ND	ND	ND	63.57	ND	16.05	ND	Romanechite
Spot 9	36.12	12.48	13.88	5.93	ND	24.68	ND	6.91	ND	Mica–Romanechite mixed spectra
Spot 10	23.24	1.65	2.24	0.52	ND	61.22	ND	11.14	ND	Romanechite
Spot 11	20.46	ND	ND	ND	ND	64.08	ND	15.46	ND	Romanechite
Spot 12	23.66	2.71	2.38	0.52	ND	58.37	ND	12.36	ND	Romanechite
Spot 13	21.65	0.85	0.81	ND	ND	61.30	1.91	13.48	ND	Romanechite
Spot 14	25.11	3.87	4.00	1.18	ND	52.66	ND	13.19	ND	Romanechite
Spot 15	53.26	ND	46.74	ND	ND	ND	ND	ND	ND	Quartz
Spot 16	23.76	3.17	2.34	ND	ND	54.33	3.2	13.21	ND	Romanechite
Spot 17	20.17	ND	ND	ND	ND	59.70	2.58	17.55	ND	Romanechite
Spot 18	32.36	ND	ND	ND	ND	2.23	31.4	ND	33.98	Ilmenite
Spot 19	20.36	ND	ND	ND	ND	63.43	ND	16.21	ND	Romanechite
Spot 20	22.75	1.82	1.89	ND	ND	59.55	ND	13.98	ND	Romanechite
Spot 21	45.78	ND	35.36	ND	ND	17.87	0.99	ND	ND	Mn oxide + quartz mixed spectra
Spot 22	22.55	ND	ND	ND	ND	77.45	ND	ND	ND	Mn oxide
Spot 23	19.69	ND	ND	ND	ND	59.17	ND	21.14	ND	Romanechite
Spot 24	29.59	6.75	7.19	ND	3.29	45.12	ND	8.05	ND	Romanechite mixed with silicate
Spot 25	23.73	2.46	2.52	ND	ND	59.31	ND	11.98	ND	Romanechite mixed with silicate
Spot 26	20.29	ND	ND	ND	ND	63.01	ND	16.7	ND	Romanechite
Spot 27	20.11	ND	ND	ND	ND	61.80	ND	18.09	ND	Romanechite

ND = Not detected.

, respectively.

2.3. Fluidization

The laboratory-scale fluidization tests were conducted in externally heated quartz reactors. It is important to highlight that two reactor sizes were used in this study, namely the 40 mm (ID) reactor (mainly for Zambian ore experiments) and 80 mm (ID) reactor (mainly for Nchwaneng ore experiments). Schematic representation of the experimental setup is given in Figure 3 for the 80 mm reactor. The diameter of the lower section of the quartz reactor is 40 mm, and the height is 300 mm. The diameter of the extended freeboard section is 110 mm, and the height is 500 mm. The lower section of the reactor is fitted with a porous distribution plate that serves to support the sample particles that are not suspended (before or during fluidization) as well as to distribute the fluidization gas. The distributor plate consists of a quartz fiber membrane filter.

The reactor was connected to the crossover duct that leads to the first collection flask, and via the cyclone, to the second collection flask as well. A pressure gauge is used to measure the inlet pressure, whereas a pressure meter is used to measure the pressure on the outlet. A thermocouple (TC Direct, type K) tip is positioned above the gas distributor (mid-material height) for the temperature of the sample.

The general fluidization procedure includes loading the sample and the subsequent flushing of bed with argon to break down any residual clusters and stratification. The superficial gas velocity is increased gradually, until the pressure drop remains constant. For high-temperature tests, the reactor (with a sample) is heated at a rate of $10 °$C/min under argon flowing at a desired flow rate. When the desired temperature is reached, argon gas is shut off, and hydrogen gas is introduced at a desired flow velocity, where the reaction test is allowed to run for a desired residence time. After the reduction test, the reactor is cooled down to room temperature under argon gas flowing at a desired velocity. Entrained/elutriated particles are generally collected from the crossover flasks, weighed, and analyzed for particle size distribution.

Prior to the determination of the prereduction extent, it was important to determine the fluid dynamic properties such as Geldart particle classification, minimum fluidization velocity, terminal velocity, entrainment, elutriation, etc.

2.3.1. Geldart Particle Classification

The mean particle diameter, ${\mathrm{d}}_p$, and density (relative to the density of the fluidization medium (${\mathrm{H}}_2$)) can be used to classify the fines according to fluidization behavior. Such a map, developed by Geldart (1973) [20], plots density difference between particles and fluidization medium versus mean particle size. Geldart (1973) grouped material into groups A to D, according to their fluidization behavior. Contrary to the grouping by Geldart, Goossens (1998) [21] proposed the use of Archimedes number to define the boundaries of the groups. The classifications for both Nchwaneng (red marks) and Zambian (green marks) fine ores can be seen in Figure 4 for various mean diameters.

The respective groups are described below:

• Group A: Fine and aeratable solids which show a significant bed expansion above the minimum fluidization velocity and have a minimum bubbling velocity and thus good fluidization quality;
• Group B: Intermediate-sized particles which hardly show bed expansion; bubbles start to appear when fluidization is obtained;
• Group C: Very fine and cohesive materials which show the formation of channels and agglomerates rather than fluidization behavior and are therefore difficult to fluidize at all;
• Group D: Coarse particles which show bubbles rising and coalescence and form vertical channels rather than fluidization behavior.

2.3.2. Minimum Fluidization Velocity

Since the Geldart classification for both Nchwaneng and Zambian ores was Group B, the observation for minimum fluidization velocity can be based on the velocity at which bubbles start to form. Experiments to determine ${\mathrm{u}}_{mf}$ were video-recorded and selected tests were captured on high speed-camera (e.g., Figure 5). From the sample masses (300 g, 500 g, 700 g and 1000 g) and particle mixes tested it was observed that a sample mass of 700 g and particle mixes of (140 g [+106–$200 \mathrm{\mu }$m], 140 g [+200–$300 \mathrm{\mu }$m], 140 g [+300–$425 \mathrm{\mu }$m], 140 g [+425–$500 \mathrm{\mu }$m], 140 g [+500–$600 \mathrm{\mu }$m]) resuled in better visualization. Under the same operating conditions, 50% of the particles were observed to be fluidized at a superficial velocity of 2.67 × ${10}^{-2}$ m/s whereas 60% of particles were observed to be fluidized at a superficial velocity of 3.33 × ${10}^{-2}$ m/s. All particles were observed to be fluidized at a superficial velocity of 4 × ${10}^{-2}$ m/s, and increasing the velocity beyond that only led to particle elutriation and carry-over.

In addition to the Geldard classification approach, the approach of crossing point of straight lines fitted to the static regime and the fluidized regime using a decreasing gas velocity method was also followed [22,23]. The static regime is fitted with a linear line with a free origin to account for sensor hysteresis/offset while a horizontal line is used to fit the fluidized regime. The pressure drop can be normalized using the hydrostatic pressure drop according to Equation (1):

$$\mathsf{\Delta }{P}_n={\displaystyle \frac{\mathsf{\Delta }P·A}{mg}}$$

(1)

where A is the cross-sectional area of the reactor, m is the bed mass and g is the gravitational constant. A plot of observed normalized pressure drop versus superficial gas velocity is shown in Figure 6.

For a mass of 100 g, the Hessels approach resuled in a minimum fluidization velocity of 1 × ${10}^{-1}$ m/s.

Similarly, for the determination of the average bed void (from pressure drop measurements), the approach of Botterill et al. (1982) [23] was followed. The measured pressure drop was related to the average voidage by Equation (2),

$$\epsilon =1-{\displaystyle \frac{1}{\left({\rho }_p-{\rho }_g\right)g}}\left({\displaystyle \frac{\mathsf{\Delta }P}{\mathsf{\Delta }H}}\right)$$

(2)

The average void at minimum fluidization was then estimated by extrapolating the average bed void values as a function of gas velocity back to that at ${\mathrm{u}}_{mf}$.

3. Results and Discussion

3.1. Experimental Limitations

The most important limitations of fluidization experiments include:

• Samplability of both solids and gases during a test. This limitation affects the tracking of thermal and reaction mineral transformations as well as the determination of the reaction kinetics.
• Monitoring the mass change during the test. This impedes the thermogravimetric determination of the thermal decomposition and prereduction kinetics.
• Sticking of fine particles on the walls of the reactor, cross-over duct, and collection flask. This limitation obstructs the high-speed camera recording of the motion of particles during fluidization. The mass and particle size distribution of allutriated and carried-over particles cannot be accurately determined.
• Hydrogen embrittlement, which limits the operating temperature to below $900 °$C.
• Cohesive adhesion also obstructs the high-speed camera recording of the motion of particles during fluidization.

3.2. Minimum Fluidization Velocity

Ergun’s Equation (3), which relates the pressure drop along the height of the fluidized to various factors, e.g., fluidization medium properties (such as fluid flow rate, viscosity and density) and particle properties (such as particle size, shape, surface, closeness and orientation of packing), has been rearranged to give the minimum fludization velocity (Equations (6)–(8)).

$${\displaystyle \frac{\mathsf{\Delta }P}{H}}=K_1{\displaystyle \frac{{\left(1-\epsilon \right)}^2{\eta }_{f}u}{{\epsilon }^3{\varphi }^2{d}_p^2}}+K_t{\displaystyle \frac{\left(1-\epsilon \right){\rho }_f{u}^2}{{\epsilon }^3\varphi d_p}}$$

(3)

where $\mathsf{\Delta }P$ is the pressure drop, H is the height of the packed bed and $\varphi$ is the sphericity of the particles. The first term on the right-hand side of Equation (3) accounts for viscous losses, where the constant ${\mathrm{K}}_1$ is determined experimentally for any particle shape, e.g., for spherical shape ${\mathrm{K}}_1$ it has been reported to be 150. The second term accounts for inertial losses, where the constant ${\mathrm{K}}_t$ is also determined experimentally for any particle shape, e.g., ${\mathrm{K}}_t$ for spherical shape is 1.75 [24,25,26].

From Ergun’s Equation, it can be seen that pressure drop is a result of simultaneous kinetic and viscous energy losses and that minimum fluidization takes place when the weight of the bed is equal to the pressure loss according to Equation (4):

$${\displaystyle \frac{\mathsf{\Delta }P}{H}}=\left(1-\epsilon \right)\left({\rho }_p-{\rho }_f\right)g$$

(4)

Combining Equations (3) and (4) at a the point of incipient fluidization and manipulating (by multiplying each term by ${\displaystyle \frac{d_p^3{\rho }_f}{\left(1-\epsilon \right){\eta }_f^2}}$ ), Equation (5) is obtained.

$$\begin{array}{c}(1-{\epsilon }_{mf})({\rho }_p-{\rho }_f)g\times {\displaystyle \frac{d_p^3{\rho }_f}{(1-{\epsilon }_{mf}){\eta }_f^2}}\hfill \\ \hspace{1em}\hspace{1em}=K_1{\displaystyle \frac{{(1-{\epsilon }_{mf})}^2{\eta }_f{u}_{mf}}{{\epsilon }_{mf}^3{\varphi }^2{d}_p^2}}\times {\displaystyle \frac{d_p^3{\rho }_f}{(1-{\epsilon }_{mf}){\eta }_f^2}}\\ \hfill \hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}+K_t{\displaystyle \frac{(1-{\epsilon }_{mf}){\rho }_f{u}_{mf}^2}{{\epsilon }_{mf}^3\varphi d_p}}\times {\displaystyle \frac{d_p^3{\rho }_f}{(1-{\epsilon }_{mf}){\eta }_f^2}}\end{array}$$

(5)

Equation (5) can be rearranged (quadratic equation roots formula) to make ${\mathrm{u}}_{mf}$ the subject of the formula, as given in Equation (6)

$$u_{mf}={\displaystyle \frac{-\frac{K_1\left(1-{\epsilon }_{mf}\right)}{{\epsilon }_{mf}^3{\varphi }^2}·\frac{d_p{\rho }_f}{{\eta }_f}+\sqrt{{\left(\frac{K_1\left(1-{\epsilon }_{mf}\right)}{{\epsilon }_{mf}^3{\varphi }^2}·\frac{d_p{\rho }_f}{{\eta }_f}\right)}^2-4·\frac{K_t}{{\epsilon }_{mf}^3\varphi }{\left(\frac{d_p{\rho }_f}{{\eta }_f}\right)}^2·\left(-\frac{d_p^3{\rho }_f\left({\rho }_p-{\rho }_f\right)g}{{\eta }_f^2}\right)}}{2·\frac{K_t}{{\epsilon }_{mf}^3\varphi }{\left(\frac{d_p{\rho }_f}{{\eta }_f}\right)}^2}}$$

(6)

When the inertial term is negligible, Equation (6) can be rearranged to give the minimum fluidization velocity, $u_{mf}$, as

$$u_{mf}={\displaystyle \frac{1}{K_1}}·{\displaystyle \frac{{\epsilon }_{mf}^3{\varphi }^2}{\left(1-{\epsilon }_{mf}\right)}}·{\displaystyle \frac{\left({\rho }_p-{\rho }_f\right)g{d}_p^2}{{\eta }_f}}$$

(7)

The experimentally observed minimum fluidization velocities for particle size groupings of [+106–$200 \mathrm{\mu }$m], [+200–$300 \mathrm{\mu }$m], [+300–$425 \mathrm{\mu }$m], [+425–$500 \mathrm{\mu }$m] and [+500–$600 \mathrm{\mu }$m] as well the mix (20 wt% each) was comparable with the theoretical minimum fluidization velocity calculated using Ergun’s Equation,

Table 6. Comparison of minimum fluidization velocities of determined using respective approaches of Geldard, Hessels and Ergun.

	Effect of Particle Size						Effect of Sample Mass
Particle-Size Range ($\mathrm{\mu }$m)	+106–200	+200–300	+300–450	+450–500	+500–600	+106–600	Mass (g)	300 g	500 g	700 g
Sample Mass (g)	700	700	700	700	700	700	Particle-Size Range ($\mathrm{\mu }$m)	+300–450	+300–450	+300–450
Geldard Approach,$u_{mf}$ (m/s)	0.053	0.072	0.072	0.088	0.088	0.070	0.053	0.070	0.070
Hessels Approach,$u_{mf}$ (m/s)	-	-	-	-	-	0.092	-	-	-
Ergun Equation,$u_{mf}$ (m/s)	0.026	0.090	0.13	0.19	0.55	0.098	0.13	0.13	0.13

. The comparison was valid for minimum velocities determined following both Geldard and Hessels’ approaches.

To circumvent the difficulties of measuring ${\epsilon }_{mf}$ and $\phi$ and thus evaluating the voidage shape factor functions ${\displaystyle \frac{1-{\epsilon }_{mf}}{{\varphi }_s^2{\epsilon }_{mf}^3}}$ and ${\displaystyle \frac{1}{{\varphi }_s{\epsilon }_{mf}^3}}$, Wen and Yu (1966) [27] approximated them to be 11 and 14, respectively. Their approximations were validated with their experimental data and data from eleven other authors. Wen and Yu (1966) [27] developed the Wen-Yu Equation

$$u_{mf}={\displaystyle \frac{1}{1650}}·{\displaystyle \frac{({\rho }_p-{\rho }_f)g{d}_p^2}{{\eta }_f}}$$

(8)

The properties of both gases and manganese ore particles can be expected to change significantly under the operation temperatures of this study (500–$900 °$C); such properties include densities and viscosities, which are used in the determination of the minimum fluidization velocity. These properties were determined according to NIST polynomials [28].

The experiments aimed at understanding hydrogen prereduction of manganese ores at various temperatures maintain constant velocity. This can only be achieved with an accurately adjusting flow velocity below terminal/carry-over/elutriation value.

The plots of $\mathsf{\Delta }$${\mathrm{P}}_n$ – ${\mathrm{u}}_0$ have been reported to offer steeper slopes with the increase in temperature during the reduction in iron ores in a fluidized bed [22]. Haider and Levenspiel (1989) [29] attributed this decrease in minimum fluidization velocity (with increase in temperature) to change in viscosity of the gas. Their observation is consistent with Ergun’s Equation until at a temperature of $323 °$C, beyond which deviations were observed, where the minimum fluidization temperature increased with the increase in temperature. Zhong et al. (2012) [30] reported similar observation in their work.

In this study, the experimental and Ergun’s Equation minimum fluidization velocities were plotted against temperature to establish whether the temperature and reduction had any influence on the minimum fluidization velocity; Figure 7. It was observed that as temperature increased, elutriation and general particle carry-over took place at lower superficial velocities. The introduction of hydrogen at a superficial velocity of 6.6 × ${10}^{-2}$ m/s was also observed to result in a temperature drop of the order of 100 to $250 °$C, depending on the initial temperature, sample mass, hydrogen flowrate, etc. The detailed thermodynamics of such a temperature drop requires more attention.

3.3. Thermal Mineral Transformations Under Argon

Phase transformations are considered to occur due to the driving force that pushes the system towards thermodynamic equilibrium at a specific temperature in a given environment. The thermal mineral transformations observed in this study are comparable to those reported by other workers for other manganese ore samples, e.g., Sorensen et al. (2010) [31] (on Wessels ore and Groote Eylandt ore).

Figure 8 and

Table 7. Mineral phases of the products of Zambian ore heated under argon at $700 °$C and the product of Zambian ore reduced by hydrogen gas at $700 °$C.

		Relative Abundance (Mass %)
Compound Name	Formula	Thermal Mineral Transformations	Hydrogen Reduction
Iron Manganese Oxide	(FeO${)}_{0.198}$ (MnO${)}_{0.802}$	15–30	>50
Calcium Manganese Oxide	(MnO${)}_{0.877}$(CaO${)}_{0.123}$	ND	5–15
Leucite	${\mathrm{KAlSi}}_2{\mathrm{O}}_6$	5–15	ND
Ramsdellite	${\mathrm{MnO}}_2$	5–15	ND
Lithosite	${\mathrm{K}}_3$(${\mathrm{HAl}}_2{\mathrm{Si}}_4{\mathrm{O}}_{13}$)	5-15	ND
Spinel	${\mathrm{MgAl}}_2{\mathrm{O}}_4$	5–15	ND
Kaonite	Mn(${\mathrm{SiO}}_3$)	5–15	ND
Hausmannite	${\mathrm{Mn}}^{2+}{\mathrm{Mn}}_2^{3+}{\mathrm{O}}_4$	30–50	ND
Akageneite	${\mathrm{Fe}}^{3+}$O(OH)	ND	5–15
Quartz	${\mathrm{SiO}}_2$	ND	<5

Predominant (>50 mass%), major (30%–50%), intermediate (15%–30%), minor (5%–15%), trace (<5%) and not detected (ND).

reveal that when Zambian ore fines were subjected to a temperature of $700 °$C under argon environment, the major mineralogical phase observed (in the order of 30 to 50 wt/wt%) was hausemannite. Iron manganese oxide was observed to be in the order of 15 to 30 wt/wt%. Only minor concentrations of ramsdellite (in the order of 5 to 15 wt/wt%) were observed. The inverse is true for a case where the same Zambian ore fines are reduced under hydrogen at $700 °$C, where hausemannite was not detected and iron manganese oxide was predominant (>50 wt/wt%), whereas minor concentrations (in the order of 5 to 15 wt/wt%) of calcium manganese oxide were also observed.

3.4. Effect of Manganese Ore Intrinsic Properties on the Reducibility by Hydrogen Gas

The effects of initial physicochemical properties of both Zambian and Nchwaneng ores on their respective hydrogen reductions are investigated in this section. Zambian ore is dominated by romanechite (at 56.10 wt/wt%) and to a lesser extend pyrolusite (at 21.44 wt/wt%), while on the other hand Nchwaneng ore is dominated by braunite (at 41.36 wt/wt%) and to a lesser extend bixbyite and hematite at 14.16 and 13.68 wt/wt %, respectively. Upon hydrogen reduction in both ores at $700 °$C, Figure 9 and

Table 8. Mineral phases of the products of Nchwaneng ore and Zambian manganese ore that were reduced by hydrogen gas at $700 °$C.

		Relative Abundance (Mass %)
Compound Name	Formula	Nchwaneng	Zambian
Iron Manganese Oxide	(FeO${)}_{0.198}$(MnO${)}_{0.802}$	30–50	>50
Calcium Manganese Oxide	(MnO${)}_{0.877}$(CaO${)}_{0.123}$	5–15	5–15
Manganosite	MnO	15–30	ND
Iron Manganese	${\mathrm{Fe}}_{0.95}{\mathrm{Mn}}_{0.05}$	15–30	ND
Akageneite	${\mathrm{Fe}}^{3+}$O(OH)	ND	5–15
Quartz	${\mathrm{SiO}}_2$	ND	<5

Predominant (>50 mass %), major (30%–50%), intermediate (15%–30%), minor (5%–15%), trace (<5%) and not detected (ND).

reveal that the Zambian reduction product was dominated by iron manganese oxide (in order >50 wt/wt %) and iron manganese was not detected. This can be expected, since Zambian ore did not contain hematite. On the other hand, Nchwaneng hydrogen reduction product had 30–50 wt/wt % and 15–30 wt/wt % iron manganese oxide and iron manganese, respectively. It is also important to highlight that an intermediate concentration of MnO (in the order of 15–30 wt/wt % ) was only observed on the Nchwaneng hydrogen reduction product.

3.5. Effect of Residence Time of Manganese Ore on the Degree of Reduction by Hydrogen

In line with the observation reported in Section 3.2, the temporary temperature drop during the introduction of hydrogen may significantly influence the hydrogen reduction reaction kinetics of the manganese ore. Depending on the hydrogen flowrate, the bed temperature may take 2–3 min to recover the desired temperature. Longer residence time (of up to 40 min), as shown in Figure 10 and

Table 9. Mineral phases of the products of Nchwaneng manganese ore that was reduced by hydrogen gas over two respective residence times (10 and 40 min) at $800 °$C.

Compound Name	Formula	Relative Abundance (Mass %)
		10 min	40 min
Iron Manganese Oxide	(FeO${)}_{0.198}$(MnO${)}_{0.802}$	15–30	30–50
Calcium Manganese Oxide	(MnO${)}_{0.877}$(CaO${)}_{0.123}$	5–15	5–15
Manganosite	MnO	5–15	15–30
Iron Manganese	${\mathrm{Fe}}_{0.95}{\mathrm{Mn}}_{0.05}$	5–15	15–30
Hollandite	Ba(${\mathrm{Mn}}^{4+}$,${\mathrm{Mn}}^{2+}$${)}_8$${\mathrm{O}}_{16}$	5–15	ND
Hematite	${\mathrm{Fe}}_2{\mathrm{O}}_3$	5–15	ND
Bredigite	${\mathrm{Ca}}_7$Mg(${\mathrm{SiO}}_4$${)}_4$	15–30	ND

Predominant (>50 mass %), major (30%–50%), intermediate (15%–30%), minor (5%–15%), trace (<5%) and not detected (ND).

, has been observed to favor the formation of iron manganese oxide, iron manganese and manganosite. Minor concentrations of iron manganese and manganosite (in the order of 5 to 15 wt/wt %) were also observed on the test with residence time of 10 min; nonetheless, iron manganese oxide was in the order of 15 to 30 wt/wt % for the same test.

3.6. Effect of Reduction Temperature on the Degree of Hydrogen Reduction

The XRD results on the effect of temperature during hydrogen reduction in Zambian ore can be seen in Figure 11 and

Table 10. Mineral phases of the products of Zambian manganese ore that was reduced by hydrogen gas at $550 °$C, $600 °$C, $700 °$C and $800 °$C, respectively.

Compound Name	Formula	Relative Abundance (Mass %)
		550 $°$C	600 $°$C	700 $°$C	800 $°$C
Iron Manganese Oxide	(FeO${)}_{0.198}$(MnO${)}_{0.802}$	30–50	>50	>50	>50
Calcium Manganese Oxide	(MnO${)}_{0.877}$(CaO${)}_{0.123}$	ND	ND	5–15	ND
Lorenzenite	${\mathrm{Na}}_2$(${\mathrm{Ti}}_2{\mathrm{Si}}_2{\mathrm{O}}_9$)	ND	5–15	ND	ND
Sodium Manganese Iron Carbide Nitride	${\mathrm{Na}}_{1.8}{\mathrm{FeMnC}}_6{\mathrm{N}}_6$	5–15	ND	ND	ND
Srebrodolskite	${\mathrm{Ca}}_2{\mathrm{Fe}}_2{\mathrm{O}}_5$	5–15	ND	ND	ND
Andradite	${\mathrm{Ca}}_3{\mathrm{Fe}}_2$(${\mathrm{Si}}_{1.58}{\mathrm{Ti}}_{1.42}{\mathrm{O}}_{12}$)	5-15	ND	ND	ND
Forsterite	${\mathrm{Mg}}_2{\mathrm{SiO}}_4$	15–30	5–15	ND	ND
Akageneite	${\mathrm{Fe}}^{3+}$O(OH)	ND	5–15	5–15	ND
Ramsdellite	${\mathrm{MnO}}_2$	ND	ND	ND	5–15
Kurchatovite	${\mathrm{K}}_3$(${\mathrm{HAl}}_2{\mathrm{Si}}_4{\mathrm{O}}_{13}$)	ND	ND	ND	5–15
Barium Oxide Peroxide	Ba(${\mathrm{O}}_2$${)}_{0.972}$(O${)}_{0.027}$	ND	ND	ND	<5
Quartz	${\mathrm{SiO}}_2$	5–15	<5	<5	5–15

Predominant (>50 mass%), major (30%–50%), intermediate (15%–30%), minor (5%–15%), trace (<5%), and not detected (ND).

, whereas those for Nchwaneng ore are given in Figure 12 and

Table 11. Mineral phases of the products of Nchwaneng manganese ore that was reduced by hydrogen gas at $500 °$C and $700 °$C, respectively.

Compound Name	Formula	Relative Abundance (Mass %)
		500 $°$C	700 $°$C
Iron Manganese Oxide	(FeO${)}_{0.198}$(MnO${)}_{0.802}$	30–50	30–50
Calcium Manganese Oxide	(MnO${)}_{0.877}$(CaO${)}_{0.123}$	5–15	5–15
Manganosite	MnO	15–30	15–30
Iron Manganese	${\mathrm{Fe}}_{0.95}{\mathrm{Mn}}_{0.05}$	5–15	15–30
Quartz	${\mathrm{SiO}}_2$	5-15	ND

Predominant (>50 mass%), major (30%–50%), intermediate (15%–30%), minor (5%–15%), trace (<5%), and not detected (ND).

, respectively. For both ores, increasing the temperature beyond $700 °$C was not observed to improve reduction extent. At $700 °$C, and for both Zambian and Nchwaneng ores, a minor concentration (in the order of 5 to 15 wt/wt %) was observed. For the Zambian product, iron manganese oxide was predominant (>50 wt/wt %) for temperatures of 600, 700 and $800 °$C, whereas concentrations of 30–50 wt/wt % were observed for the Nchwaneng product at $700 °$C.

3.7. Effect of Thermal Physicochemical Property Changes on Prereduction

Manganese ore hydrogen reduction was not observed to increase when the operating temperature was increased from $700 °$C (

Table 12. Mineral phases on the spots on the back-scattered electron images (Figure 13) of the product of Nchwaneng manganese ore that was reduced by hydrogen gas at $700 °$C.

Spot	O	Mg	Al	Si	S	Na	K	P	Ca	Mn	Fe	Ba	Phase
1	25.4	ND	ND	ND	ND	ND	ND	ND	2.7	72.0	ND	ND	Impure Manganosite
2	26.2	ND	ND	ND	0.6	ND	ND	ND	3.8	69.4	ND	ND	Impure Manganosite
3	26.7	ND	ND	ND	ND	ND	ND	ND	2.4	70.9	ND	ND	Impure Manganosite
4	28.2	2.5	ND	3.2	ND	ND	ND	ND	11.2	43.9	11.1	ND	Calcium Manganese Oxide
5	44.4	32.3	ND	ND	ND	ND	ND	ND	9.9	7.4	6.0	ND	Magnesium Manganese Oxide
6	41.2	ND	ND	0.6	ND	ND	ND	17.5	40.7	ND	ND	ND	Calcium Phosphorus Oxide
7	24.2	ND	ND	ND	ND	ND	ND	ND	3.5	56.1	16.2	ND	Iron Manganese Oxide
8	28.0	ND	ND	ND	ND	ND	ND	ND	ND	72.0	ND	ND	Manganosite
9	ND	ND	ND	1.0	ND	ND	ND	ND	ND	3.7	95.4	ND	Iron Manganese
10	35.4	ND	ND	14.5	ND	ND	ND	ND	12.2	37.9	ND	ND	Mn Silicate Mix
11	26.8	ND	ND	0.9	ND	ND	ND	ND	15.5	56.8	ND	ND	Calcium Manganese Oxide
12	38.3	ND	0.9	15.1	0.8	5.2	1.0	ND	3.9	30.9	ND	3.8	Mn Silicate Mix
13	ND	ND	ND	ND	ND	ND	ND	ND	ND	3.9	96.1	ND	Iron Manganese
14	23.6	ND	ND	ND	ND	ND	ND	ND	ND	64.2	12.2	ND	Iron Manganese Oxide

ND = not detected

and Figure 13) to $800 °$C (

Table 13. Mineral phases on the spots on the back-scattered electron images (Figure 14) of the product of Nchwaneng manganese ore that was reduced by hydrogen gas at $800 °$C.

Spot	O	Mg	Al	Si	P	Ca	Mn	Fe	Phase
1	62.2	ND	1.6	15.4	1.0	1.8	8.6	9.3	Silicate Mix
2	24.8	ND	ND	ND	ND	10.9	64.3	ND	Calcium Manganese Oxide
3	26.8	2.3	ND	ND	ND	3.8	67.1	ND	Impure Manganosite
4	27.3	2.1	ND	1.2	ND	9.9	59.7	ND	Calcium Manganese Oxide
5	ND	ND	ND	ND	ND	ND	1.1	99.0	Iron Manganese
6	ND	ND	ND	ND	ND	ND	1.5	98.5	Iron Manganese
7	27.5	ND	ND	ND	ND	6.0	66.6	ND	Calcium Manganese Oxide
8	23.8	ND	ND	ND	ND	5.1	59.3	11.9	Calcium Manganese Oxide
9	ND	ND	ND	N	ND	ND	1.7	98.3	Iron Manganese

ND = not detected

and Figure 14). This observation can be attributed to the grain and crystal structural changes that were observed at $800 °$C as shown in Figure 14. The dominance of the silicate mix on the SEM analyses can be associated with sintering and thus a drop in porosity.

3.8. Cohesive Adhesion: Formation and Sticking of a Product Layer on the Walls of the Reactor

Prereduction experiments conducted at temperatures of 800 and $900 °$C were observed to result in the formation of a metallic layer on the walls of the reactor. This observation can be associated with the reduction in magnetite and hematite to wustite with sticky `iron-whiskers’ as well as the production of sub-micron fines. A similar observation was made by [22] while investigating the reduction in combusted iron using hydrogen (Table 14).

It has been reported by various workers (e.g., [22,30,32,33]) that minumum fluidization velocity derived from Ergun’s Equation (e.g., in Section 2.3.2) does not agree with experimental values at high temperatures and also for Geldart particles C. This is associated with the cohesive forces and sticking which is not catered for by Ergun’s Equation. In this study, agressive cohesive adhesion even on the walls of the reactor was observed at $900 °$C as shown in Figure 15.

Table 14. Mineral phases on the back-scattered electron images (Figure 16) of the product of Nchwaneng manganese ore that was reduced by hydrogen gas at $900 °$C.

Spot	O	Mg	Al	Si	Na	K	Ca	Mn	Fe	Phase
1	54.4	ND	ND	45.6	ND	ND	54.4	ND	ND	Quatrz
2	34.1	ND	ND	12.3	ND	ND	34.1	ND	ND	Silicate Mix
3	25.3	ND	ND	ND	ND	ND	25.3	ND	ND	Manganosite
4	31.9	3.6	ND	3.5	ND	ND	9.5	51.5	ND	Calcium Manganese Oxide
5	36.7	3.7	ND	0.7	ND	ND	0.8	58.0	ND	Magnesium Manganese Oxide
6	36.1	3.6	ND	15.4	ND	ND	17.8	27.2	ND	Silicate Mix
7	35.5	ND	ND	14.5	ND	ND	12.0	37.9	ND	Silicate Mix
8	ND	ND	ND	ND	ND	ND	ND	ND	100.0	Iron
9	ND	ND	ND	ND	ND	ND	ND	6.6	93.4	Iron Manganese
10	ND	ND	ND	ND	ND	ND	ND	ND	100.0	Iron
11	33.0	ND	ND	15.1	ND	ND	20.9	23.1	7.9	Silicate Mix
12	24.9	ND	ND	ND	ND	ND	ND	41.8	33.3	Iron Manganese Oxide
13	35.6	ND	4.9	12.0	8.9	0.5	4.4	23.4	10.4	Silicate Mix
14	26.9	ND	ND	ND	ND	ND	5.6	67.5	ND	Calcium Manganese Oxide
15	54.0	ND	ND	46.0	ND	ND	ND	ND	ND	Quatrz
16	24.7	ND	ND	ND	ND	ND	0.5	61.6	13.2	Iron Manganese Oxide
17	ND	ND	ND	ND	ND	ND	ND	2.1	97.9	Iron Manganese
18	ND	ND	ND	ND	ND	ND	ND	2.6	97.4	Iron Manganese
19	ND	ND	ND	ND	ND	ND	ND	7.7	92.3	Iron Manganese

ND = not detected

Table 14. Mineral phases on the back-scattered electron images (Figure 16) of the product of Nchwaneng manganese ore that was reduced by hydrogen gas at $900 °$C.

Spot	O	Mg	Al	Si	Na	K	Ca	Mn	Fe	Phase
1	54.4	ND	ND	45.6	ND	ND	54.4	ND	ND	Quatrz
2	34.1	ND	ND	12.3	ND	ND	34.1	ND	ND	Silicate Mix
3	25.3	ND	ND	ND	ND	ND	25.3	ND	ND	Manganosite
4	31.9	3.6	ND	3.5	ND	ND	9.5	51.5	ND	Calcium Manganese Oxide
5	36.7	3.7	ND	0.7	ND	ND	0.8	58.0	ND	Magnesium Manganese Oxide
6	36.1	3.6	ND	15.4	ND	ND	17.8	27.2	ND	Silicate Mix
7	35.5	ND	ND	14.5	ND	ND	12.0	37.9	ND	Silicate Mix
8	ND	ND	ND	ND	ND	ND	ND	ND	100.0	Iron
9	ND	ND	ND	ND	ND	ND	ND	6.6	93.4	Iron Manganese
10	ND	ND	ND	ND	ND	ND	ND	ND	100.0	Iron
11	33.0	ND	ND	15.1	ND	ND	20.9	23.1	7.9	Silicate Mix
12	24.9	ND	ND	ND	ND	ND	ND	41.8	33.3	Iron Manganese Oxide
13	35.6	ND	4.9	12.0	8.9	0.5	4.4	23.4	10.4	Silicate Mix
14	26.9	ND	ND	ND	ND	ND	5.6	67.5	ND	Calcium Manganese Oxide
15	54.0	ND	ND	46.0	ND	ND	ND	ND	ND	Quatrz
16	24.7	ND	ND	ND	ND	ND	0.5	61.6	13.2	Iron Manganese Oxide
17	ND	ND	ND	ND	ND	ND	ND	2.1	97.9	Iron Manganese
18	ND	ND	ND	ND	ND	ND	ND	2.6	97.4	Iron Manganese
19	ND	ND	ND	ND	ND	ND	ND	7.7	92.3	Iron Manganese

ND = not detected

4. Conclusions

This study evaluated the effects of fluidization conditions on the hydrogen reduction in Zambian and South African (Nchwaneng) manganese ore fines, providing insights into the temperature, residence time and intricic ore properties.

The experimentally observed minimum fluidization velocities for particle size groupings of [+106–$200 \mathrm{\mu }$m], [+200–$300 \mathrm{\mu }$m], [+300–$425 \mathrm{\mu }$m], [+425–$500 \mathrm{\mu }$m] and [+500–$600 \mathrm{\mu }$m] as well the mix (20 wt% each) were comparable with the theoretical minimum fluidization velocity calculated using Ergun’s Equation. The comparison was valid for minimum velocities determined following both Geldard and Hessels’ approaches.

Particle elutriation and carry-over were observed to intensify with increasing temperature at lower superficial velocities. This phenomenon suggests a reduction in particle density or increased drag forces at elevated temperatures, necessitating careful control of gas velocity to minimize material loss.

Process thermodynamics during hydrogen introduction revealed significant temperature drops ranging from 100 to $250 °$C. This cooling effect can be associated with the combined influence of endothermic and exothermic reactions, along with the thermal shock from cold hydrogen inflow. The observed thermal gradient underscores the need for optimized preheating strategies or controlled hydrogen introduction to maintain thermal stability during reduction processes.

Mineralogical transformations under inert argon conditions resulted in the formation of iron manganese oxides (15–30 wt%). This transformation indicates partial reduction and phase stabilization under non-reducing conditions.

The degree of prereduction was observed to increase with the increasing temperature and residence time. This observation can be associated with kinetic dependency. However, increasing the reduction temperature beyond $700 °$C did not yield significant improvements in reduction efficiency. This plateau effect suggests a thermodynamic limitation or potential sintering phenomena that inhibits further reduction at elevated temperatures.

Extending the residence time up to 40 min promoted the formation of reduced phases such as iron manganese oxide, iron manganese and manganosite. This observation emphasizes the importance of time-dependent diffusion processes and solid-state reactions in achieving deeper reduction extents.

Operational observations revealed reactor brittleness after hydrogen prereduction at $900 °$C. This observation can be associated with thermal stress and/or chemical degradation of reactor materials through hydrogen embrittlement. Additionally, cohesive adhesion was more pronounced at $900 °$C, possibly due to partial melting or enhanced sintering effects, which could impair fluidization quality and reactor longevity.