1. THE DEVELOPMENT OFA REAGENT
ENVIRONMENT FOR A SILICA-MAGNETITE
FROTH FLOTATION SYSTEM
UROP Project
July-September 2015
Anna Caklais and Charlie Singer

2. Project Aims
• Develop a simplified bench-scale system with which to
test design modifications to laboratory-scale flotation
tanks.
• Understand the principles of froth flotation of silica
through examining the fundamental principles of air
pressure, pH, and reagent dosages.
• Advance understanding from a single silica species
system through investigating the effects of varying
reagent concentrations on a silica-magnetite froth flotation
system.
• Analyse data to determine an optimum reagent
environment that would enable separation of the two
species.

3. Planning
• Determined the effects of pH, frother concentration and air speed
variation on a single silica species system.
- Allowed application to two species system
• Investigation into two projects:
- Charlie : Silica-barite
- Anna: Silica-magnetite
• Problems encountered:
- Magnetic separation not applicable to barite
- Attempted dissolving the barite using sodium carbonate -
unsuccessful on time scale
- Attempted using density contrasts between silica and
barite to deduce the relative proportions in overflow samples -
however, could not physically separate the two species

4. Planning
• Sample sizing of silica and magnetite to determine feed
grain size distributions.
0
10
20
30
40
50
60
70
80
90
100
-38 +38 -45 +45 -53 +53 -63 +63 -75 +75 -
90
+90 -
106
+106
Mass(g)
Particle Size (µm)
Silica Feed Size Distribution
+44 -88 +30 - 62 50:50
0
5
10
15
20
25
30
35
40
45
50
<38 38-45 45-53 53-63 63-75 75-90 90-106 >106
Mass(g)
Particle Size (µm)
Feed Size Distribution
Magnetite Feed Silica Feed Feed Used

5. Planning
• Undertook literature searches - scoping work by Filippov. L. O.
et al (2010) was used in the selection of reagents and their
concentrations.
• Reverse cationic flotation – dominant method used in research
and industry whereby the collector adopts a positive charge to
float the gangue mineral (silica)
Reagents:
• Collector: Dodecylamine at 98%-purity
• Activator: 1-Tridecanol
• Depressor: Modified cornstarch
• For each reagent a maximum, optimum and minimum
concentration was selected for investigation.

6. Risk Assessments
• Completed personal hazard assessments
• Updated COSHH files with reagents and materials
• Completed SOP forms with experimental procedure

7. Experimental design
• Experimental design:
• Fractional factorial programme was employed using an augmented
Box –Behnken test.
• Series of 15 representative experiments constructed to vary
collector, activator, and depressor dosages top understand effects
on both grade and recovery.
Experimental Methods:
Chemical -1 0 1
Dodecylamine 30 g/t; 0.027 g 45 g/t; 0.0405 g 60 g/t; 0.054 g
1-Tridecanol 5 g/t; 0.0045 g 15 g/t; 0.0135 g 25 g/t; 0.0225 g
Starch 500 g/t; 0.45 g 750 g/t; 0.675 g 1000 g/t; 0.9 g

8. Flotation methods
• Flotation undertaken using a Denver Cell:
• 60:40 ratio silica: magnetite:
• 270 g +30-62; 270 g +44-88; 360 g magnetite
• 2.1 L deionised water
• 30% solids
Method:
• Mixing of solids and deionised water for 2 minutes and pH reading taken
• Activator and depressor added – 5 minutes conditioning
• pH reading taken and collector added to slurry and allowed to condition for
10 minutes to allow absorption to the silica
• No PH modifier added – varied between 9.8 and 10.2
• Frother incorporated into mixture for 1 minute
• Air and impeller speed maintained constant
• Fast overflow collector for 30 seconds, and slow overflow for 90 seconds
• Wet masses measured
• Oven dried to determine dry masses.

9. Sampling methods
• Dry masses recorded
• Each sample placed on top of a stack of sieves and shaken for 15
minutes:
• 106 μm
• 90 μm
• 75 μm
• 63 μm
• 53 μm
• 45 μm
• 38 μm
- Samples re-weighed and masses recorded.
- Horseshoe magnets were used to separate magnetite grains from
the silica – each placed in separate sample bags.
- Remaining silica weighed to calculate grade and recovery.

10. Experiment
Run 1 (%) Run 2 (%)
Total flow grade
Total flow
recovery Total flow grade
Total flow
recovery
1 [-1] [-1] [0] 74.222 9.478
2 [-1] [1] [0] 83.350 15.654 80.589 24.555
3 [1] [-1] [0] 85.717 42.846 92.205 27.710
4 [1] [1] [0] 83.371 43.956 77.728 37.115
5 [-1] [0] [-1] 87.824 37.341 81.249 23.774
6 [-1] [0] [-1] 72.273 8.311 78.714 19.338
7 [1] [0] [-1] 83.787 38.535 83.548 47.291
8 [1] [0] [1] 89.024 50.787 81.379 48.814
9 [0] [-1] [-1] 84.669 47.111 76.037 40.810
10 [0] [-1] [1] 86.618 47.134 87.185 41.034
11 [0] [1] [-1] 88.056 32.091 86.406 30.691
12 [0] [1] [1] 76.413 17.099 77.865 20.449
13 [0] [0] [0] 77.499 38.310 77.650 51.606
14 [0] [0] [0] 87.351 46.555 79.949 46.785
15 [0] [0] [0] 84.374 37.718 81.471 39.969
Grade =
Mass of silica obtained in sieve x 100
Total mass of overflow solids
Recovery =
Mass of silica retained on sieve x 100
Initial mass of silica in tank pulp

11. Results from First Linear Regression
Variable Effect Standard Error t-value p-value
Intercept 84.061 1.009 83.315 9.62E-25
x1 C 3.8752 1.5959 2.4282 0.025882
x2 A -1.0464 1.4119 -0.74116 0.46816
x3 D -3.1492 1.2846 -2.4516 0.024662
x4 O -1.9506 1.6695 -1.1684 0.25788
x1:x2 C:A -3.8893 1.5165 -2.5646 0.019493
x1:x3 C:D 1.1614 1.7323 0.67046 0.51107
x1:x4 C:O -1.4321 2.3777 -0.6023 0.55448
x2:x3 A:D -3.2478 1.5482 -2.0978 0.050308
x2:x4 A:O 1.2101 2.0556 0.58869 0.56339
x3:x4 D:O 4.2758 3.0187 1.4164 0.17372
Model: y ~ 1 + x1 + x2 + x3 + x4 + (x1*x2) + (x1*x3) + (x1*x4) + (x2*x3) + (x2*x4) + (x3*x4)
R-squared: 0.67 Adjusted R-Squared: 0.486 p-value : 0.00827

12. Variable Effect Standard Error t-value p-value
Intercept 84.138 0.91242 92.214 7.04E-29
x1 C 3.5794 1.0379 3.4487 0.002406
x2 A -0.63544 0.97449 -0.65207 0.52143
x3 D -3.3619 1.1273 -2.9823 0.007103
x4 O -2.4212 1.5483 -1.5638 0.13281
x1:x2 C:A -3.8591 1.4464 -2.668 0.014392
x2:x3 A:D -2.9149 1.4548 -2.0037 0.058174
x3:x4 D:O 5.6073 2.5611 2.1894 0.040005
Model: y ~ 1 + x1 + x2 + x3 + x4 + (x1*x2) + (x2*x3) + (x3*x4)
R-squared: 0.64 Adjusted R-Squared: 0.52 p-value : 0.00128
First Step-wise Linear Regression

14. Variable Effect Standard Error t-value p-value
Intercept 83.534 1.0045 83.16 6.13E-28
x1 C 3.9178 1.651 2.373 0.027269
x2 A -0.062962 1.3645 -0.04614 0.96363
x3 D -3.0283 1.3274 -2.2814 0.033063
x4 O -1.6489 1.5826 -1.0419 0.30931
x1:x2 C:A -4.032 1.5608 -2.5833 0.017336
x1:x3 C:D 1.3275 1.7882 0.74236 0.46609
x1:x4 C:D -1.3725 2.4499 -0.56024 0.58125
x2:x3 A:D -3.3235 1.6012 -2.0757 0.050395
x2:x4 A:O 0.29558 1.9716 0.14992 0.88226
x3:x4 D:O 3.9729 3.1174 1.2744 0.21643
Model: y ~ 1 + x1 + x2 + x3 + x4 + (x1*x2) + (x1*x3) + (x1*x4) + (x2*x3) + (x2*x4) + (x3*x4)
R-squared: 0.613
Adjusted R-Squared:
0.429 p-value: 0.00977
Linear Regression with Additional 1-Tridecanol Data

15. Variable Effect
Standard
Error
t-value p-value
Intercept 83.657 0.90964 91.967 4.25E-32
x1 C 3.5646 1.0718 3.3259 0.002828
x2 A -0.030364 0.93536 -0.03246 0.97437
x3 D -3.2728 1.1649 -2.8094 0.009714
x4 O -2.018 1.4733 -1.3697 0.18346
x1:x2 C:A -4.0698 1.4878 -2.7354 0.011528
x2:x3 A:D -2.9536 1.503 -1.9651 0.06108
x3:x4 D:O 5.4524 2.6348 2.0694 0.049444
y ~ 1 + x1 + x2 + x3 + x4 + (x1*x2) + (x2*x3) + (x3*x4)
R-squared: 0.587 Adjusted R-Squared 0.467 p-value = 0.00156
Step-wise Linear Regression with Additional 1-Tridecanol Data

18. Non-Linear Regression