Magnetic roasting tests on Cartersville manganese ores (preliminary report)

DEPARTMENT of NATURAL RESOURCES
ZAcK D. CRAVEY, Commissioner
DIVISION OF MINES, MINING AND GEOLOGY
425 State Capitol ATIANTA, GEORGIA
GARLAND PEYTON, Director

MAGNETIC ROASTING TESTS on
CARTERSVILLE MANGANESE ORES
(Preliminary Report)
AUTHORS
L. L. McMURRAY, Assistant Mining Engineer
s. D. BROADHURST AND M. T. PAWEL
In Collaboration with H. S. RANKIN, Senior Mining Engineer
Under the General Qirection of W. HARRY VAUGHAN, Chief, &gional Products Research Division
Tennessee Valley Authority
Published in cooperation with the Tennessee Valley Authority

INFORMATION CIRCULAR 13

DECEMBER, 1941

DEPARTMENT of NATURAL RESOURCES
ZAcK D. CRAVEY, Commissioner
DIVISION OF MINES, MINING AND GEOLOGY
42 5 State Capitol ATLANTA, GEORGIA
GARLAND PEYTON, Director

MAGNETIC ROASTING TESTS
on
CARTERSVILLE MANGANESE ORES
(Preliminary Report)
AUTHORS
L. L. McMuRRAY, Assistant Mining Engineer
s. D. BROADHURST AND M. T. PAWEL
In Collaboration with H. S. RANKIN, Senior Mining Engineer
Under the General Direction of W. HARRY VAUGHAN, Chief, Regional Products Research Division
Tennessee Valley Auth&rity
Published in cooperation with the Tennessee Valley Authority

INFORMATION CIRCULAR 13

DECEMBER, 1941

I
FOREWORD
This preliminary report concerns some research and tests in connection with magnetic roasting of Cartersville (Ga.) manganese ores. This work was done at the request of the Georgia Division of Mines, Mining and Geology, and was perfonned in cooperation with them at the Minerals Testing Laboratory of the Tennessee Valley Authority at Norris, Tennessee, between the months of Kay and November, 1941. It is believed that the results submitted herewith may be useful in connection with the urgent need for additional supplies of manganese during the present national emergency.
This experimental work does not represent an exhaustive stuQy of the problem of the separation of manganese from manganese-iron ores. Instead, it represents a brief practical study of the problem of improving the quality of manganese from low grade manganese concentrates from the Cartersville district. It is reasonable to expect that further research into the subject would result in further improvements. However, it is believed that the results to date may warrant serious consideration and investigation by commercial interests. Moreover, it appears hardly likely that further research would result in diminishing the value of the work covered by this report.
While this report does not present figures representing operating costs based on actual tests of full-sized commercial equipment, it does include sane data from which competent operators may fairly estimate such costs.
Therefore, it is expected that those who are interested in the

II
subject matter will draw their own conclusions concerning the value of the procedures outlined and the possible commercial application of the process.
We wish to acknowledge the interest and assistance which has been received from the Division of Kines, Mining and Geology of the Georgia Department of Natural Resources and from several manganese producers in the Cartersville, Georgia, area.

l
Preliminary Report
MAGNETIC ROASTING TESTS ON CARTERSVILLE ~~GPJf.ESE ORE
ABSTRACT OF RESULTS
These tests were carried out on washed concentrate ore from the Cartersville (Ga.) district rhich had the following approximate analysis: manganese 21.4 per cent, iron 27.0 per cent, insoluble 11.2 per cent. The results of magnetic roasting and subsequent magnetic separation indicated that 61 per cent of the roasted ore can be separated into manganese and iron concentrates.
The combined roast~d product constituted 85.7 per cent by weight of the original washed concentrate ore and analyzed as follows: manganese 25.42 per cent, iron 31.41 per cent, insoluble 12.0 per cent.
B,y closely controlling the jigging of the manganese concentrate, we obtained the following three fractions:
(1) Manganese concentrate after jigging 32.0 per cent by weight
{loss in jigging 3.7 per cent by weight). This concentrate analyzed as follows: manganese 45.10 per cent, iron 6.41 per cent, insoluble material 12.0 per cent.
(2) Iron concentrate 25.3 per cent by weight. This concentrate analyzed as follows: manganese 8.74 per cent, iron 55.1 per cent, insoluble material 8.1 per cent.
(3) Middlings 39.0 per cent by weight. This concentrate analyzed as follows: manganese 23.3 per cent, iron 34.16 per cent, insoluble material 12.2 per cent.
All of the products separated are believed to be marketable under

2
present conditions. It must be understood that the foregoing represents the result
of tests under one set of conditions. In order to evaluate the work cov1ered by this report in tenns of commercial application, the report must be studied in its entirety.
Commercial Application The results represent conditions of roasting, magnetic separation
and jigging of the manganese concentrate considered optimum, but not beyond the limits of ordinary mill operations. Commercially, close control of the several operations is imperative. '" Moreover, there are certain features whit::h require particular attention. Among these are:
(1) The roasting furnace must have an air-tight reducing chamber. (2) Reasonably accurate control of the temperature and reducing agent are required. (3) The rate of processing must be carefully regulated by the ore discharge in order to allow sufficient time for heating and reducing the ore. (4) Means must be provided for cooling the roasted ore without contact with the atmosphere. This may be done by quenching in water. (5) Magnetic separators with controlled magnetic intensities must be provided, by the use of which the production of the middling product may be regulated. (6) The jigging of the feed to the furnace or the manganese
concentrate must be closely controlled. By this operation, silica and
iron, contained in the insoluble particles, are removed from the manganese concentrate.

3
Estimated Costs and Profits It is estimated that the costs of magnetic roasting with a
furnace of two tons per hour capacity should not exceed $1.50 per ton. It is not unreasonable, under present market conditions, to
expect a profit of $2.00 per ton of ore processed by the method outlined
in this report, assu!Jl:ing a steady processing of 90 per cent capacity in
a furnace designed for the production of two tons per hour.
Remarks This preliminary report will include some details and data by
the examination of which the foregoing has been estimated.
Substantiating Tests 2!!. Commercial Size Equipment The large scale tests on the manganese-iron ores were confined
to the roasting step and only small-scale laboratory magnetic separation testing was applied. In order to check the results of the magnetic separation on commercial size equipment, roasted ore samples were sent to various manufacturers for testing on magnetic separation equipment. These results, while not canplete, indicate that the laboratory magnetic separations given herein can be equaled or improved on commercial separators.

4
INTRODUCTION PJf.D DISCUSSION
Ores of manganese containing iron in excessive amounts are not suitable for use in the production of ferro manganese. Many of the manganese ore~ deposits in the Cartersville district of Georgia contain iron as an ingredient which may be present in quantities up to 27 per cent. Economic separation of this iron from the manganese has been studied, and the principal results of this work, to date, are reduced in this report.
Careful review of previous work performed Qy the United States Bureau of Mines indicated that magnetic roasting,with subsequent magnetic separation of the synthetic magnetite, offered promise of success. Preliminary small scale tests were made, and favorable separations of the manganese
and iron, b.1 the method suggested, were indicated.
reduction As a result of these studies, a pilot/roasting furnace, with a capacity of 500 pounds of ore per hour, was constructed. This furnace was of t he vertical shaft type, and was designed to provide complete control over the various phases of the roasting process. The furnace operating results indicate that 97 per cent of the iron in the material fed into the fUITtace was made magnetic. I!lving thus made the iron magnetic, the method of magnetic separation of the iron from the maganese in the product discharged from the pilot furnace was undertaken. In this operation it was discovered that this separation is by no means complete, e.nd that a large percente.ge of the manganese minerals are removed with the synthetic magnetite. However, the separation is partially complete, and it is possible that it is sufficiently

5
complete to warrant the use of the process commercie.lly. Apparently the manganese, at least to some degree, develops magnetic characteristics during the roasting, and several reasons for this condition of the manganese minerals are discussed in this report.
This report describes the work to date on magnPtic roasting of Cartersville, Georgia, manganese-iron ores v.rith a view to concentrating both the iron and manganese by subsequent magnetic separation. This
Vlork has been performed during the period from May to November, 1941, at
the Minerals Testing Laboratory of the Tennessee Valley Authority at Norris, Tennessee.
Little or no separation of the two minerals in the crude state of pyrolusite (or psllomelane) and hematite (or limonite) is possible by magnetic means until magnetic roasting is performed. This roasting
3 converts the iron oxide hematite (Fe2o .n~O) to magnetite (Fe3o4 , composed of .31.0% FeO and 69.0% Fe2o3 (1) ) The following table shows the relative
magnetic strength of various minerals which might be associated with the Cartersville manganese ores in the crude state or after being roasted.

6
TABLE I
PEL.ATIVE MAGNETIC FORCE OF MINERALS VHITCH MIGHT BE M.ADE IN MAGNETIC ROASTING I!LAlfGJlJ'<TESE ORE (a)

Mineral
Iron M:e.gnetite Manganese Hausmannite Hematite Limonite Pyrolusite Manganite Manganosite

Formula
Fe
Fe.304
Mn
Mn.304
Fe2o3 3 Fe2o .2~0
Mn02
3 r.~o .F2o
MnO

Relative Force
100.00 40.18 8.9.3
0.88 *
1.32 0.84 0.71 0.52
0.65 *

(a) Table from Catalogue No. 77, Dings Magnetic Manufacturing Company, Page 8.

* Calculated from data given in International Critical Tables.

These data are in specific susceptibilities; vlith Mn02 given as 69.0 and

ll.m3o4 as 55.8 (both as 21 C).

Thus:

69.0 x 0 71 = 55.8

0.88.

7

Chemical Reactions

Theoretically, the reactions, involved in roasting under various

conditions of reducing agent and temperatures, are represented by the follow-

ing equations: 1.

===.. co

2 Fe3o4 (Magnetite)

(i.e. 2(3(F1.e0O%.F6e29.o03%)) t co2

t

3.

Mno2

(psilomelane}

(pyrolueite)

t . CO ----;- MnO (Manganosite)

MnO

-f HzO

5. 3 Mn02 -+ 2 co

6. Fe2o3 -+ co

+ 7. 3 FeO

~0

s. FeO ... co

9. MnO ... co

. 10. 2 co

~
02

ll. 3 FeO t C02

Mn304

-+

( :&usmannit e }

2 FeO

-+

(FeiTous OXide)

Fe3o4

.

. Fe

co2

"' Mn

C02

2 co2

Fe o
34

t

co

2C02 C02

Discussion of Reactions
If equations 1, 2, 3, 4, and 5 are allowed to proceed to the right,
magnetite and either manganosite or hausmannite are produced, the specific

8
oxio_e depending upon roasting conditions. To allovr these equations to proceed to the right, two primary conditions are necessary:
(1) Sufficient temperature for the reaction to take place. (2) Sufficient vapor pressure of reducing agent. The most satisfactory temperature limits were found to lie between 500 C.
and 1000 C. At convenient ore sizing and temperatures below 500 C.,
reducing periods are too long to be practical. At temperatures above
1000 c., sintering may be encountered, which is an undesirable condition.
For all practical considerations, we have found the reduction of the iron o:x:i.des to magnetite may be accomplished at a temperature of 600 C. in one half hour, with the ore particles sized to 7/8 inch or smaller. Even though shorter reducing periods can be realized by higher temperatures than 600 C., the decrease in time is not wholly warranted due to added costs of operation and more difficult problems of furnace design.
To maintain sufficient vapor pressure of reducing agent, it has been found practical to perform the reduction in a closed chamber and allow
the gaseous products of reduction, co2 and H20, to escape. This may be done
by passing the reducing gas over the ore. In addition, after the reduction has taken place, means to keep the reaction from reversing are necessary. This can be done by quenching the ore in water or passing the ore through a cool inert atmosphere.
Means to control the reaction are also desirable. The reduction of hematite to magnetite, if continued, may produce ferrous oxide. The formation of this compound represents a loss of reducing agent and should be

9
avoided. If the reaction continues after ferrous oxide is produced, metallic iron may be made, as shown in equation 8. The reduction to metallic iron, however, re~uires a more intense reducing action than that obtained using a reducing gas and temperatures around 6o0 C. For the production of sponge iron solid carbon plus a reducing gas and prolonged heating at approximately 950 C have been found necessary (2). Likewise, the production of metallic manganese as shovm in equation 9 is not possible under the conditions employed.
To control the roast to yield magnetite of a composition approaching the reaction in
the theoretical, conditions were controlled to utiliz~equation 7. The reaction in equation ll also allowed some oxidation of the ferrous iron, but design of the reducing furnace prevented this reaction from being very effective. The use of steam in connection with such roasting was suggested by Davis (3) and proved very effective in the Cartersville ore roasting tests. The steam required was obtained by quenching the hot ore upon discharge from the furnace.
~Sizing
The ore size is an important variable in roasting technique. Fine ore generally permits shorter reducing periods than coarse ore, but the Cartersville ores used in these tests were sufficiently porous to allow rapid and complete reduction at jig sizes. The use of coarse ore overcomes several difficulties in roasting and also produces finished concentrates not requiring sintering for furnace use. Particle liberation was found to be relatively complete at sizes of -1/4 inch, but was never entirely complete,

10

even on grinding to -200 mesh. It appears ths.t some iron (3 per cent
to 4 per cent) is intimately associated with the manganese and could not
be liberat,ed at any practical sizing.

Reducing .P.gent

As a reducing agent, a gas is almost a necessity. Reduction

can be obtained by intimately mixing carbon and ore, but this solid to

solid reaction is poor. Generally, reduction ~ solid c~rbon is caused
~ the presence of some air, thus forming co, which is a very effective

reducing agent. Many types of reducing agents may be used, but their

effectiveness depends wholly upon their available carbon monoxide or hydrogen.

Producer gas and coke oven gas are particularly good as they are mixtures

of both carbon monoxide and qydrogen. Efdrocarbon gases such as propane

and methane are compounds of hydrogen and ce.rbon, and are not satisfactory

reducing agents until th~ have been converted to carbon monoxide and

eydrogen. Hot iron oxide is a fairly good catalytic agent for breaking

down these compounds. Superheated steam converts the carbon to carbon

monoxide: C .f. H:20

.. ~ .f. CO. Thus, the steam aids effectively in

producing a reducing agent as well as controlling the roasting. Ordinary

fuel oil may be gasified at a temperature of about 375 C, and the gas

tms produced is a mixture of eydrocarbons and can be made an effective

reducing agent by the means described. The low cost of fuel oil, combined

with its ease of handling and measuring, made it attractive for use in the

pilot-furnace.

11
The roasting furnace used in these tests, involving the scheme described above, proved efficient and practical, making over 95 per cent of the iron in the ore into magnetite.
ROASTING
Furnace Design end Operation After small batch preliminary testing had indicated promise for
this method of treating the Cartersville ore, a pilot furnace for continuous roasting was constructed. Since it was not possible to control the roast in the preliminary tests, the results of magnetic separation wereof doubtful value and the process could not be evaluated Q1 these small tests. As a shaft-type furnace bad proved effective and economical in magnetic roasting
of Mesabi hematite (3), this type was adopted for roasting these ores.
Description and Operation of Furnace The accompanying sketches show the principles of construction of
the pilot furnace erected at the laboratory. The essential details consist
of a shaft or cupola made of circle fire brick surrounded qy a metal shell
built in halves and fastened around the brickvmrk. A separate fire box of regular firebrick, in which the gaseous products from the oil burner were completely burned, adjoined the cupola. By means of an exhaust fan these hot gases were drawn through a bed of ore eighteen inches thick. The heat intake and exbaust pipes were of extra heavy cast iron pipe in the form of a 11T" with holes along the bottom of both for spreading the gases throughout the cupola cross-section to provide uniform heating. A skirt was provided

12

over the exhaust pipe to prevent small ore particles from being drawn

out through the exhaust. The reducing chamber was immediately below as a gas or in the form of gaseous
the heating zone. The reducing agent/ crude oil was introduced into
a l-inch cast iron ring, which distributed the reducing agent over the

furnace area. A quenching hopper was located below the cupola proper,
and the open end was sealed Q1 water to prevent air from entering the
furnace. A rotary disc ore discharge was used to provide a uniform dis-

charge from the furnace. As the normal furnace capacity of crude -7/8

inch ore was 500 pounds per hour, an elevator was built to facilitate ore

handling. Thermocouples were installed to permit accurate temperature

measurements in the cupola. In operation, two defects in'the original furnace were noted:

cupola.

{1) Objectionable air leaks developed in the single brick

(2) The reducing chamber should have been much longer, thus

providing a greater furnace capacity at very little additional expense.

To prevent the air leaks the furnace was rebuilt after a few

tests. A layer of asbestos refractory cement was plastered on the out-

side of the cupola, and the sheet iron shell was firmly fastened and all leaks pl~ged in the shell openings. This provided a good seal, but a

double brick furnace with interspersed impervious cement is recommended.

When the furnace was properly regulated, one men could carry out

the operation easily. Fuel consumption and other pertinent data are given

later in this report.

13
Heating Unit The beat capacity required varied greatly with the ore moisture
content, and thus a flexible heating unit was necessary. A Johnson Series 520 oil burner proved satisfactory. The capacity of this burner is from 0.65 gallons/hour to 7.5 gallons/hour, with an air blower furnishing air at from 100 to 350 cubic feet/minute at 10 ounces to 16 ounces pressure. This blower was arranged in series with a rheostat so that the air volume was variable at will. With a total moisture content of 10 per cent, the heat requirement was calculated to be 340 Btu/potmd; if the moisture content increased to 18 per cent, 445 Btu were required. The heat value of the fuel oil used was 18,000 Btu/pound so that with 10 per cent moisture, 9.5 pounds of oil/hour was required. In normal operation this amount was close to ll.6 pounds/hour due to radiation and other items not used in the calculations. The high coefficient of heat transfer of the ore should be noted. The fact that the heat of the gases could be dissipated in passing through 18 inches of ore in no small way affects the cost of such a roasting mheme.
Exhaust Fan Due to resistance of gas flow, the horsepower required for the
exhaust fan increases rapidly as the height of ore column is increased. The height of this ore column was determined from actual test data given by Davis (3) in heating Mesabi iron ores. The specific heats of the manganese and iron oxides are so nearly the same, and with tmiform size of

the ore, these values could be used directly. The hot gases entered at a temperature of approximately 950 C, and the exhaust temperature remained relatively near 90 C.
As no test date. were available to determine the fan capacity required for drawing the correct amount of hot gases through the ore, a variable speed for the fan was provided and a gate valve was inserted in the exhaust line. By these means the correct amount of hot gases could be dra1m through the charge, and at the same time a suitable be.lance between the oil burner blower and the exhaust fan could be maintained.
Reducin.. Oil The amount of reducing oil used was regulated by a needle valve
which had been calibrated. The needle valve reading was checked qy direct
measurement in the oil storage reservoir. This oil was introduced into the furnace in the raw state or vaporized. Generally a vacuum of 1/4 inch of water existed at the reducing oil entry level. Introduction of the raw oil absorbed heat during vaporization, but as the subsequent reaction was exothennic, very little total heat was lost by using the crude oil direct.
Regulation of ~ Discharge The ore discharge was closely regulated by the ore discharge
disc. The ore discharge rate determined the time the ore was in the heating and reducing zones. The quantity of ore discharged also determined the amount of steam made in quenching a.lthough the ratio of steam produced and ore discharged remained constant. By a study of the FeO-Fe304-H20 equilibrium

15

conditions, Davis (3) determined that a temperature of 60o C it was

desirable to have in excess of SO per cent by VIeight of steam in the

atmosphere in the reducing chambel". OUr calculations indicate the.t this

amount of steam was present at all times. A sample calculation will shOlr

the following:

At a feed rate of 500 pounds/hour, 42S pounds/hour were discharged,

the difference being loss of moisture (50 pounds at 10 per cent

water) and oxygen (22 pounds). With a heat capacity of 20S

Btu/pound of ore (assuming specific heat of MnO ~nd Fe3o4ae 0.20 : Q

(1112 F - 70 F) x 0.20 x 1 pound 20S Btu/pound). The heat

. (lJJ20p- : 600 C)

.

= required was 20S x 42S pounds S9,000 Btu. Of this amount of heat

11,500 Btu were required to heat 2200 pounds of water 5 F every

hour in the quenching tank (2200 x 5 :a 11,500 Btu.). 89,200 -
= 11,500 77,770 Btu available for steam production. To superheat 7 = steam tolll2 F, 1531 Btu. per pound are required. 77,770 1531

50.7 pounds water. Generally it was shovn1 that the reducing oil
= used amounted to 3.0 pounds/hour. Thus 50.7 ~ 3.0 53.7. + = 50.7 53.7 94.5 per cent steam.

Losses due to radiation of heat by the ore were not included,

buft is obvious that an excess of steam was available. This steam aided

the thermal be.lance of the unit as the heat of this rising steam aided in

beating the ore, but, as previously indicated, its main 1JllrPOSe was to crack

the gases and prevent the formation of ferrous oxide.

16
TESTS
Preliminary Tests
The Bureau of Mines, Report of Investigations 2936 (4), gives
preliminary results of small batch tests made on roasting Cartersville ores and then magnetically separating the roasted product. The results of treating one Cartersville jig concentrate is shown in Table II.
This ore was roasted in a batch revolving drum for thirty minutes. The nature of reducing gas was not disclosed. The ore was a clean jig CJncentrate between one inch and one-half inch in size. After roasting, the magnetic portion was removed with a hand magnet. While these test results are very good, it appears that this ore must have been quite different in character from the ore samples used in our tests. The results of the Bureau were limited in scope, and as no further work has been published by the Bureau in connection with these tests, performed in 1929, it may be assumed that the ore used in these tests was not representative of that now available in the field. The ore samples taken in this district for the roasting tests shown in this report have not allowed such complete separation as that shown in Table II.
Batch testing on Cartersville ore covered in this report may be show in Table III, ;hich is typical of the results obtained. This ore was roasted in a crucible with 10 per cent -200 mesh coke at 600 C for one hour. 'rhe ore was crushed and passed through 1/4 inch screen. The products representing various magnetic susceptibilities were kept separate and combined for final products. The magnetic separator waB a Stearn's

17 T.ABLE II
ROASTING OF CARTERSVILLE ORE *

Product

Per Cent Weight

Analysis, Per Cent

Per Cent total

Mn

Fe

Insol. Mn

Fe Insol.

Unroasted Cone.

1~gnetic Portion

.38.7

Nomag. Portion

61.~

Composite (roasted) 100.00

.32.8 17.9 7.1.:,. 12.4 45~5 12.2 51.7 2.3 5.0 .36.5 19.0 7.$

1.3.2 86.6 100.0

92.6 60.7 7.4 .39.3 100.0 100.0

* The Bureau of Mines, Report of Investigations 29.36 (4)

18
TABLE III

BATCH ROASTING TEST OF CARTERSVILLE MANGANESE

Product

- - - Per Cent Analysis, Per Cent Per cent Total Ma:giier-

Weight M!!.

Fe Insol. M!!. Fe Insol.Resistence

Mag. Cone. #1 18.6 1.10 66.91 7.56
Mag. Cone. #2 28.3 3.96 54.19 11.42

71 000 ohms
51 000 ohms

Mag. Cone. #3 32.1 39.68 23.23 14-54

2 1 000 ohms

Mag. Cone. #4 14-4 45.80 6.08 15.04

500 ohms

Tails

6.6 44.10 1.93 18.50

Combined Cone. 46.9 2.82 59.3

5.5 76.6

Combined Midds 32.1 39.68 23.2

54.0 20.5

Combined Tails 21.0 45.20 4.S

40.5 2.9

Heads (Calc. } 100.0 23.53 36.4

100.0 100.0

19

Type "KS" d.c. separator (dry material required) and the resistance inserted in the magnet circuit was taken as the standard method of determining the magnet strengths. By assuming these standards of magnet strengths the tests could be compared as the same resistance produced the same magnet strength each time. In the batch using coke as the reducing agent, the roast was more intensely reducing than in the pilot furnace, and the results are not very reliable.

Pilot Plant Tests

In the pilot roasting furnace many test runs were made, and after

the furnace was properly adjusted, all gs.ve very similar results. In order

to avoid unnecessary duplication, only Test 12, typical of other results,

is given in full.

Conditions of Test:

Test Number 12

Sample Number, Feed-Sample No. 8a.. -7/8 inch Cartersville Ore,

Jig Concentrate (See Screen .Analysis)

Feed rate-

500 pounds per hour

Solids removed from furnace--428 pounds per hour

Time of test--

12 hours. Samples collected over one-half

hour periods after furnace running smoothly.

12 samples collected.

Temperature range of ore--585 C to 630 C.

Reducing fuel oil, average--3.0 pounds per hour

Heating fuel oil, average--1.0 gallon ger hour

Temperature intake gas., average-960 C.

Temperature exhaust gas., average 93 C.

Vacuum in reducing chamber--1/4 inch water

Vacuum in steam chamber--1/8 inch water

Temperature rise in quenching water-5F, measured in first three

hours after discharge of hot ore.

20

TABLE IV

SCREEN ANALYSIS OF PRODUCTS IN FURNACE; TEST NO. 1.2

Screen Analysis of Feed

-7/8" .J.3/4" 9.0

-3/4TI .f.l/2" 16.1

-1/2" U/4" 43.7

-l/4TI .J.lO mesh 17.4

-10 mesh .J.20 mesh 7.6

-20 mesh +48 mesh 4.3

-48 mesh

1.9

100.0

Screen Analysis of Discharge

-7/8" -3/4" -1/~" -1/4" -10 mesh -20 mesh
-48 mesh

.J.3/4"
f.l/2"
.J.l/4" flO mesh .J.20 mesh
+48 mesh

4.7 18.4 50.0 16.6
4,..55
0.3
100.0

Screen Analysis--Feed to Magnetic Separator
Furnace Discharge Crushed to -4 Mesh

-4 mesh
-10 mesh -20 mesh -48 mesh

++ 10 mesh 20 mesh i 48 mesh

6o.5 21.9
10.8 6.8
100.0

21
TABLE V
RESULTS ON SEPARATION OF FURNACE DISCHARGE
SAMPLES cmmiNED AND CRUSHED TO -4 MESH
SEPARATED DRY ON "KS" MAGNETIC SEPARATOR

Product

J.QW_ Anal;:Lsis 1 Per Cent

Weight Mn

Fe Insol.

Per Cent Total Ma.~et

Mn

Fe Insol.Re~istance

Magnetic #1

25.3

Magnetic #2

39.0

Magnetic #3

17.6

Nonmagnetic

18.1

Heads (Calc. )

100.0

Heads (Analysed) 100.0

Combined Fe Cone. 25.3

Combined Middling 56.6

Combined Mn Cone. 18.1

8.74 23.30 42.22 43.79 26.68 25.42 8.74 29.20 43.79

55.10 8.1 34.16 12.2 12.95 14.8
8.27 15.6 31.04 12.2 31.41 12.0 55.10 8.1 27.58 13.0
8.27 15.6

8.3 45.2 16.8 5,000 ohms 34.1 43.0 39.0 2,000 ohms
27.7 7.4 21.1 ;oo ohms
29.9 4-4 23.1 100.0 100.0 100.0 100.0 100.0 100.0
8.3 45.2 16.8 5,000 ohms 61.8 40.4 60.1 500 ohms 29.9 44 23.1

22 TABLE VI
RESULTS OBTAINED BY COMBINING MAGNETIC PRODUCT #3 AND
NONMAGNETIC PRODUCT SHOWN IN TABLE V

Product

~Cent
Weight

Analysis, Per Cent ~Cent Total Magnet

MD.

Fe Insol. Mn Fe Insol. Resistance

Combined Fe Cone. Combined Mdds Combined Mn Cone. Heads (Calc.)

25.3 39.0 35.7 100.0

8.74 23.30 43.10 26.68

55.10 8.1 34.16 12.2 10.60 15.2 31.04 12.2

8.3 45.2 16.8 34.1 430 39.0 57.6 11.8 44.2 100.0 100.0100.0

5,000 ohms

23

Table V shows that in Test No. 12, 45 per cent of the iron in the

furne.ce discharge was recovered in an iron concentrate analyzing:

55.10% Fe

8.74% Mn

8.1% Insoluble

Also that 29.9 per cent of the manganese in the furnace discharge was recovered

ina product analyzing:

43.79% Mn

8.2'7% Fe

15.6% Insoluble

By combining magnetic concentrate No. 3 with the non-me.gnetic. portion, t.he results

would be as in Table VI, where the manganese concentrate has ananalysis of:

43.10% Mn

10.6()% Fe

15.2% Insoluble; with a recovery

of 57.6% of the mangs.nese in the ore.

The total iron made magnetic {at the magnet strength existing when

500 ohms resistance was inserted in the magnet circuit) was 95.6 per cent. The

remaining 4.~ per cent iron, not magnetic, remained in the manganese concentrate.

This iron has been found to be in the insoluble portion of the ore, and due to

the impervious nature of this cherty insoluble material, it cannot be properly

roasted in the pilot furnace. Tests have established that for each unit of

insoluble material, present in the non-magnetic portion, 0.266 per cent iron

oxide cannot be magnetically roasted. Applying this knowledge to the above results

it shows that 97.5 per cent of the iron available for roasting was made magnetic.

Thus the poor separation results obtained cannot be chargeable to poor furnace

roasting.

It should be noted that in the batch tests (Table III) the iron

portion in the manganese concentrate lla.s much lower due, no doubt, to the

more intense roast available with solid carbon and limited oxygen so that the

iron ass()Ciated. with the insolubles was roasted to the magnetic oxide. To determine if' this insoluble portion of the manganese concentrate
could be effectively removed by jigging, a sample of ore -7/8 inch in size was jigged in the laboratory with an 8-inch by 12-inch jig. The results are tabulated in Table VII.
These jigging results represent close regulation of the jig and of course this was necessary. in order to perform the task of removing only a
small amotmt of tai 1ings from such a rich feed. App~ these results to the total recovery, the results of this
process in separating the manganese from the iron are shown in Table IDI.

25
TABLE VII
JIGGING TEST ON MANGP..NESE CONCENTRATE

Product

.f!!: .9!m!

Analzsi~ 1 Per Cent

Weight !f!t

~

Inso1.

Per Cent Total

!g.

~

In~o1.

Concentrate Tails (Calc.)
Beads {Calc.)

89.7
10.3 100.0

45.10 32.91 43.10

6.41
47.2
10.60

12.0 47.1 15.2

92.2 7.8
100.0

54.2 68.9 45.8 31.1 100.0 100.0

26
TABLE VIII TOTAL RECOVERY AND GRADE OF MANGANESE FROM PILOT FURNACE

Product

Per Cent Weight

Analzsis 1 ~Per Cent

Mn

Fe

Insol.

Per Cent Total

Mn

Fe

Insol.

Feed to Furnace 100.0

Roasted Product 85.7

Magnetic Separation:

Fe Concentrate 25.3

Middlings

.39.0

Mn Concentrate .35.7

Jigging: Jig Mn Concentrate
Jig Mn Tails

30.6 5.1

21.40 25.42

27.05 31.41

11.18 12.0

100.0 101.8

1-00.0 100.0 99.5 92 .3*

8.74 55.10 8.1 23.30 .34.16 12.2 43.10 10.6C 15.2

8.3 45. 2 16.8 .34.1 43. 0 39.0 57.6 11.8 44.2

45.10 6.41 12.0

5.3.1

6.4 .34.0

32.91 47.2 47.1

4-5

5.4 10.2

* Theoretically this should total 100 per cent. The discrepancy is
due to error in weighing and sampling the ..many products to make up the furnace feed. For per cent totals of separations subsequent to roasting, 100 per cent
was used.

27

OPERATING COSTS

Operating costs from which a commercial operator could draw

any conclusions with reasonable certainty are not available. However, the

following data are submitted, together with some comments, by the inter-

pretation of which anyone familiar with the operations of roasting furnaces

may compile reasonably accurate estimates which may apply to any specific

condition.

The cost of roasting in the laboratory furnace, not including

the magnetic separation, crushing, and jigging, follow:

Labor (one man @ $5.00/8 hour shift) Fuel Oil (heating and reducing 1.6
gallon/hour @ 9)
Power (2-1/4 H. P.) Miscellaneous

$2.50 per ton

14
.04

n
"

" n

.04 It n

Total

$2.72 per ton

The low pilot furnace capacity of two tons per eight hour shift

makes the labor item badly out of proportion as compared with a furnace

of larger capacity. Since one man can operate with equal ease a furnace

with a capacity of two tons per hour, the labor cost would ?e reduced to

32 per ton.
In a report of the operation of the Mesabi Magnetic Roasting

Furne.ce (.3), which was of shaft-type construction, the following costs

are given:

28

Number of operatiQg days (September, 1935) 30

Opera,ting hours

692

Lost time, per cent

3.89

Tons of feed

5,802

Total gallons of oil

55,182

Number of men per shift

1.3

Electric power per ton or feed,

kilowatt hours

19,0SO

Gallons oil per ton of feed

9-5

Electric power per ton or feed,

kilowatt hours

3-3

Cost per ton of feed:
ou

$0.555

Power

0.050

Labor

0.072

Total

$0.677

An operation at Anzau, South Manchuria, described in the same

report, where 120 tons of iron oxides per day were magnetically roasted

in a shaft-type furnace, showed the following costs:

Cost per ton: Combustion fUel-pulverized coal
Redaction fuel-coke oven gas Power and electricit.Y

Total

$0.634

Both of the above furnaces are of a type applicable for roasting Cartersville

manganese-iron ores. The magnetic roasting of iron ores is not essentially

different from the roasting of manganese-iron ores, provided both lave the

general p~sical characteristics required for shaft fUrnace operation.

Therefore, these costs can be considered as representative of comparable

operations on the Cartersville ore. Operating costs will vary with the

capacity of the fUrnace. Computations based on the above large scale

29 commercial operations and the small pilot~furnace tests indicate that for a roasting furnace of two tons per hour capacity, the cost of operating would be less than $1.50 per ton.

Additional Magnetic Separation Tests

In an effort to improve the mapnetic separations ,

a large

number of small scale laboratory separators were made using various ore

sizings and alternating current as well as direct current separators.

The results of some of these tests are given and discussed in this section of the report.

The failure of the two metal oxides to completely separate is

difficult to explain. The question of locked grains has been studied

and the results show that good liberation is obtained even in sizes as
coarse as one-half inch with 95 per cent liberation at -4 mesh. Co~plete liberation is not attained by grinding to very fine sizes and this 4 or 5 per cent of iron in the manganese is thought to be intimately associated

with the manganese and not capable of being liberated at any practical ore size.

Wet Tube Separator Tests
A separation of the roast of pilot-furnace Test No. 4 (very
similar to Test No. 12), which was ground to pass through a 200 mesh screen, is shown in Table IX. This separation was made in a modified Davis Tube Wet Separator (5). This separator consisted of an inclined tube filled with water with the poles of a strong magnet resting against the

30
TABLE IX
WET TUBE SEPARATION OF ROASTED ORE NO. 4

Product

Per Cent

Analzsis 1 Per Cent

Per Cent Total

Weight ~

Fe

Insol. Mn

Fe

Insol.

Magnetic No. 1 12.4

Magnetic No. 2 1.4.3

Magnetic No. 3 23.8

Magnetic No. 4 1.4.2

Nonmagnetic

35.3

Combined Mag. Cone. 12.4

Middlings

38.1

Combined Nonmag. 49.5

Heads (Calc.) 100.0

13.63 21.13 38.38 43.88 43.88

47.07 35.71 18.94 10.82 14-34

10.02 14-24 19.78 25.58 15.52

13.63 32.2 43.8 35.8

47.07 25.2 13.3 22.02

10.02 17.6 18.2 17.10

4.8 35.5 6o.7 100.0

26.5 43.5 30.0 100.0

7.8 39.0 53.2 100.0

31
sides of the tube. The ore was made into a slurry and poured through
the water-filled tube with the magnet at different intensities. In
settling through the magnetic zone, the magnetic particles stick to the sides of the tube and the non-magnetic material settles to the bottom of the tube and is drawn off. By using sufficient magnet strength, the tube can be rotated, allowing the particles to roll over one another, liberating entrapped non-magnetic particles. Washing can also be performed. In this way clean separations are possible. This separator has been used with remarkable success qy many investigators and is especially valuable in testing fine ores.
Note in Table IX that no greater selection in the separation is possible even in this fine size. It was also found that dry separations in the "KS" Separator at fine sizes did not improve the separation over
dry separations at -4 mesh. Therefore, it appears that close sizing of the
product does not improve the separations appreciably.
Alternating Current Separator Test Two primary causes for the relatively poor separations have been
advanced as a result of study of our test results: (1) The high permeability and high coercive force of the manganese
minerals causes the manganese to be removed in the magnetic fractions. (2) The high permeability of the synthetic magnetite causes
induced magnetism in the manganese and it is removed along with the iron. In many instances an alternating current separator bas given
improved results over a direct current separator when separating minerals

32
with high permeabilities (6), (7), and (8). To determine if an alternating current separator would allow
improved results, a laboratory model was constructed. This separator consisted of an alternating current relay with a laminated core in series with a variable transformer which was in turn in series with a 120-volt, 60-cycle line. The separator permitted sufficiently accurate control to permit regulating the magnet intensity and the distance of the ore from the magnet, thus removing products of decreasing susceptibility. Table X shows the results of a typical separation with this alternating currentseparator.
Roasted ore, -4 mesh, from furnace Test No. l2 was used as the feed. Three
magnetic fractions were removed and the products were analyzed and combined without a middling product to give the results shown in Table X.
To determine the influence of ~cle changes in using an alternating current magnet, the laboratory magnet and variable transformer were placed in series with a rotary converter and the results are tabulated in Table XI. No improvement in separation was apparent.
Variations in Reducing Roast to Improve Separation A study of the possibilities of improving the separation by a
more intense reducing roast with production of sponge iron was made. In
. this test -4 mesh ore was roasted with 35 per cent of -200 mesh coke in a
m.t.lffle furnace. The charge was heated for three hours at 9500 C. The results obtained by the magnetic separation of this product are shown in Table XII.

33
TABLE X A.C. MAGNETIC SEPARATION

Product

~Cent
We!ght

Ana1~sis 1 Per Cent

Ma

l!.

Inso1.

Per Cent Total

Ma

Fe

Inso1.

Magnetic Nonmagnetic Heads (Calc.)

59.9 40.1 100.0

15.47 42-34 26.3

42.98 10.47 30.00

12.12 18.78 1.4.8

35.3 64.7 100.0

86.0 49.0 14.0 50.9 100.0 100.0

34 TABLE XI
EFFECT OF CYCLE CHANGES ON A. C. SEPARATION

Product
A. Mag. Cone. No 1 Mag. Cone. No. 2 Mag. CoJo.c. No. 3 Tails
B. Mag. Cone. No. 1 Mag. Cone. No. 2 Mag. Cone. No. 3 Mag. Cone. No. 4 Tails
C. Mag. Con.c. No. 1 Mag. Cone. No. 2 Mag. Cone. No. 3 Taila

Qzcles Voltage Per Cent Weight

21

20

21

45

21

40

60

55

60

70

60

100

60 no

114

55

114

70

114

75

18.9 20.0 14-9 46.2 22.5 19.5 12.0 17.2 28.8 20.9 16.4 17.2 45-5

Anal~si~ 2 fer Cent

Mn

~ Inso1.

5.28 6.60 24.42 38.50 5.38 10.08 25.42 37.41 40.88 6.16 6.93 23.42 38.72

54.97 7.46 52.51 9.84 30.08 14.S2 9.30 18.S2 57.3 7.96 50.1 ll.36 31.4 13.98 16.5 16.66 7.4 20.40 55.52 7.26 51.97 9.78 31.73 14.20 9.85 20.20

35
TABLE XII
MAGNETIC SEPAP~TION OF SPONGE IRON ROAST

Product

Per Cent Weight

Anal;y:sis 1 Per Cent Per Cent Total Ma~et
Mn Fe. Insol. MB. E!t Insol. Resistance

Magnetic No. 1 Magnetic No. 2 Magnetic No. 3 Magnetic No. 4 Nonmagiletic Heads (Calc. )

5.3 29.8 15.7 22.5 26.7 100.0

4-03 6.94 27.33 43.12 47.26 28.9

S2.79 9.04 0-7 12.5 2.6 10,000 ohms 72.18 14.15 7.2 61.3 21.9 5,000 ohms 32.51 21.50 14.8 145 17.7 2,000 ohme 13.78 19.50 33.5 8.9 22.9 500 obns 3.58 23.86 43.8 2.9 34.9 35.1 19.2 100.0 100.0 100.0

36
Inspection of Table XII indicates that an appreciable amount of metallic iron was made. The amount of metallic iron here cannot be determined accurately due to the fact that analysis for metallic iron, ferrous oxide, and ferric oxide are confused by the oxidization effect of the manganese oxide present. In some attempts to make such an analysis, it was found that the amount of oxidization during volumetric analysis was dependent upon the amount and type of manganese oxide present in the sample. Metallic manganese had little or no effect. A study of other methods of analysis for these products revealed length;y and somewhat doubtful successes, so that further attempts were abandoned.
A calculation of analytical percentages does reveal the metallic iron present. Calculating the manganese present as MnO and leaving the iron as metallic iron gives: Concentrate No. 1, 97.03; Concentrate No. 2, 95.28; Concentrate Mo. 3, 89.31. Assuming that the undetermined constituent of the concentrates is oxygen, it follows that the first two concentrates al'e almost wholly metallic iron, and the other concentrate probably contains an appreciable percentage of metallic iron. The charge is probably "over roasted." This intense reduction allowed no better selection in the magnetic separation than when only magnetite is made. The concentration was essentially the removal of oxygen from the iron. To proouce such a product commercially would be very expensive and would offer many difficulties.
Magnetic Separation Teets ~ Hand-Picked Sample An interesting test was made ~ band picking manganese minerals
from crude ore Sample No. Sa (feed to pilot-furnace Test No. 12) which were
entirely free from iron viewed under the microscope. This sample was crushed

37 to -4 mesh and resampled. This material was given a roast with 10 per cent -200 mesh coke in a muffle furnace for a time of one hour. The results upon magnetic separation are shown in Table XIII.
Note in Table XIII how the iron content of the manganese decreases
with decreased magnetic susceptibility. This test indica.tes that 40 per cent of the manganese minerals are "magnetic. n Grinding this sample to -200 mesh and separating in the Davis Tube at very high magnet intensity
gave the results shown in Table XIV.
In Table XIV note that much more manganese was removed e.t the high
magnet intensity than in Table XIII. Thus it ap!)ea.rs that only the very
purest manganese minerals are "non-magnetic." However, caution should be exercised in stating that this iron content of the manganese alone is the cause of poor manganese-iron separations. It must be remembered that the liberated magnetite has a relative magnetic strength of at least forty times the relative magnetic strength of pure mangenese minerals (see Table I). The influence of this unliberated magnetite contained in the manganese minerals (without bringing other forces into play, such as permeability and coercive force) cannot influence the magnetism of the manganese greater than the extent of its iron content. To approximate the influence of the magnetite content of the manganese minerals, we can assume
that the manganese containing 18.75 per cent Fe, or 25.8 per cent Fe3o4(of
which only 5.2 per cent exists as shown in Table nii, Magnetic Concentrate No. 1) would have a relative magnetic strength of:
40.18 x 25.$% or 10.3 (see Table I, where relative magnetic strength of magnetite is given as 40.18 and MnO as 0.65)
10.3 plus (74.2% x 0.65) or 10.8

.38 TABLE XIII A. C. SEPARATION OF HAND-PICKJ!D :MANGANESE l..J!NERALS

Product

Cycles Vo1ta.rut Per Cent Weight

Analzsis 1 Per Cent

Mn

Fe

Insol.

Mag. Cone. No. 1 60

100

Mag. Cone. No. 2 60

70

Mag. Cone. No. 3 60

50

Mag. Cone. No. 4 60

40

Non-magnetic

5.2 41.72 18.73 6.92

9.1 48.72 11.02 7.'36

11.7 54.60

6.06 6.02

13.9 57.12

4-a. 5-54

60.1 59 .36

1.65 5.38



39
TABLE XIV
WET SEPARATION OF -200 MESH HAND-PICKED MANGANESE MINERALS

Product

~Cent

.Analysis, Per Cent

Mg,

~

Insol.

Per Cent Total

Mn

Fe

Insol.

Magnetic Non-magnetic

75.0 53-33 5.53 5.79

93.7 90.7

25.0

61.05

l.ll

1.so

27.6

6.3

40 This value of relative magnetic strength is still only one-fourth as much as pure magnetite. If, however, the magnetite contained 75.2 per cent unliberated pure manganous oxide, its relative magnetic strength would be approxims.tely the same as the manganese oxide containing 25.8 per cent magnetite. It hardly seems probable that this is the cause of such poor seps.rations. As we have expressed, they are more due probably to the high permeability of the iron or the manganese minerals themselves. To determine this fact is quite out of the scope of our laboratory investigations.
Various other schemes have been tried for improving the separations. Regrinding of the middling gives no improvement. Schemes involving the use of both alternating current and direct current magnets intermittently have not indicated a promising trend. To obtain improved results over those obtained at present, appears impossible until more information can be gathered so that the cause of poor separations will be pinned down to one known condition and a practical means adopted to overcome this condition. For instance, if it were definitely determined that the high permeability of the manganese minerals were the cause of the failure to make a clean separation, a study to either reduce this high permeability, inherent in the
roasted product, or to nullify its effect b.1 using a specially designed
magnetic separator might solve the problem. In reganl to the reduction of the effect of high permeability, it
may be stated that the effect of temperature of roasting influences this condition and a study of temperature versus permeability might be enlightening.
In regard to the type of magnetic separator adaptable, R. s. Dean

41
and c. W. Davis (8) discuss the use of magnetic separators involving
separation b.Y magnetic remanence alone. Because of the variable existing in such a stu.dy, a practical solution appears doubtful.

FOOTNOTE REFERENCES
1. W. E. Ford, Dana's Textbook of Mineralogy, page 494
2. C. E. Williams, E. P. Barrett, and B. M. Larsen, Production of Sponge Iron, U. S. Bureau of Mines, Bulletin 270, 1927
3. E. W. Davis, Magnetic Roasting of Iron Ore, Bulletin No. 13, University of Minnesota, Mines Experiraent Station, 1937
4. }i'. D. DeVaney and w. H. Coghill, Beneficiation of Oxidized Mangenese Ores by' Magnetic Separation of Roasted Jig Concentrates. U. s. Bureau
of Mines, R. I. 2936, 1929
5 Magnetic Mineralogy, Dings Magnetic Sepa.rator Company, page 6
6. T. J. Martin, The Magnetic Concentration of Certain Natural and
Artificial Manganese Oxides. Transactions American Electro-Chemical Society, Vol. 57, (1930)
7. J.,. M. Gaudin, Principles of Mineral Dressing, page 456
8. R. s. Dean and C. w. Davis, Magnetic Concentration of Ores, A.I.M.E.
Transactions, Vol. 112 (1934)
ADDITIONAL REFERENCES
U. S. Bureau of Mines R. I. 6768, page 37
W. H. Smith, Low Temperature Reduction of Ore, Transaction A.I.M.E. 91:419 (1930)
C. W. Davis, Alternating-Current Magnetic Separation of Iron Ores. u. s.
Bureau of Mines, Department of Investigations 3229:37 (1934)
R. B. Sosman and J. C. Rostetter, The Ferrous Iron Content and Magnetic Susceptibility of Some Articifial and Natural Oxides of Iron, A.I.M.F.. 58:409-33 (1917)
s. G. Frantz and G. W. Jarman, Magnetic Beneficia.tion of Nonmetallics
Transactions of A.I.M.E. 102:122 (1932)
0. Lee, B. w. Gandred and F. D. DeVaney, Magnetic Concentration of Iron Ores
oi.. Alabama. U. S. Bureau of Mines Bulletin 278 (1927)

43
ADDITIONAL REFERENCES (continued)
S. Norton and s. LeFeure, The Magnetic Concentration of Law Grade Iron
Ores, Transactions of A.I.M.E. 56:892-916 (1917)
0. C. Ralston, Iron Oxide Reduction Equilibria, U. s. Bureau of llines
Bulletin 296, 1929
C. o. Hawk, P. L. Golden, H. H. Storch, and A. c. Fieldner, Conversion
of MetlE.ne to Carbon Monoxide and }Vdrogen, Industrial and Engineering Chemistry 28;23-27 (19.32)


'i
..

EXHAUST PORT

- - - -WATI!."' LEVI!.L- - - - - - - - - - - - QUENCHING lOX
VERTICAL SECTION

" iJ

6

f!' (}

Q
CJ 0
0
0 0 a

q 0
./> <J IJ
0

"'
.0 " II

-. - <)-

<7

<>

d

VERTICAL SECTION 2

= SCALE r" I'
DRAWING REDUCED TO APPROXIMATELY 1/2

HEAT INTAKE
1
j
REDUCING OIL RESERVOIR
MANGANESE CONCENTRATION
MAGNETIC ROASTING FURNACE
TENNESSEE VALLEY AUTHORITY
COMMERCE DEPARTMENT
IIECOWW[NDED