Application of a novel green collector for iron oxide
removal in glass sand through froth flotation
Filipe Miguel Serrano Balagueiras
Dissertação para obtenção do grau de Mestre em
Engenharia Geológica e de Minas
Orientadora: Professora Doutora Maria Teresa da Cruz Carvalho
Júri
Presidente: Professor Doutor António Jorge Gonçalves de Sousa
Orientadora: Professora Doutora Maria Teresa da Cruz Carvalho Vogal: Engenheira Liliana Marques Alonso
Junho de 2018
Declaro que o presente documento é um trabalho original da minha autoria e que
cumpre todos os requisites do Código de Conduta e Boas Práticas da Universidade
de Lisboa.
Abstract
Froth flotation is one of the most used methods for mineral processing around the world, the
versatility of the ores it can separate and the very fine particle sizes make it a common process
worldwide for many non-energy extractive industries. But it often requires the addition of a chemical mix
in the ore pulp that most usually has negative impacts on the waste water of the flotation process, being
sometimes very difficult to treat or stabilize.
In the last decades the mining activities in Europe stalled and one of the many reasons was the
environmental impact of the industry. Nowadays the need for producing its own geological raw materials
has to embrace the growing environmental awareness of the people, companies and states. New mining
techniques, water and soil treatment, enhancing ore recoveries and reducing the impact of many
processing plants methods have to be made to create if not a better World, a better Europe.
It is on the reduction of mineral processing impact that the present work has its emphasis. The
laboratory work and the present report was developed in collaboration with an international company
operating in Portugal, the Instituto Superior Técnico (IST) University and Oulo University, Finland. It was
funded by FCT for the CELMIN project, an ERA-min project for the utilisation of green chemicals in non-
energy extractive industries.
The work developed from March to May, 2015 in the company’s facilities had the objective of testing
different traditional, “non-green” collectors, used for removing iron oxide and other heavy minerals
from glass sand through reverse froth flotation and a novel “green” collector. The processing plant has
used different kinds of “traditional” collectors from the anionic oxydrilic family to float the iron oxides
throughout the years. To create a comparable point of view with the novel “green” collector, three
different traditional collectors were tested in a laboratory froth flotation device, changing one variable
at a time (OVAT).
The University of Oulo created various Nanofibrillated Celluloses able to collect pure hematite
(iron oxide) and the tests revealed a better performance for n-butylamine Nanofibrillated Celluloses with
samples sent to the company’s facility in Portugal. Preliminary tests using the NFC “green” collector
were made in laboratory to understand the necessary concentrations of collector in the glass sand pulp
as well as the pH and frother concentration changing one variable at a time (OVAT).
The results of the preliminary tests with NFC provided data to create a three level full factorial
Design of Experiments (DOE), defining the variable factors and their range. The studied factors were
the Collector concentration (g/t), the pH of the pulp and the Conditioning time of the collector in minutes.
And the response variables were the iron oxide grade in ppm in the glass sand (sunken product), the
iron oxide recovery (%) in the floated product and the mass pull (%) in the floated product.
The “traditional” collectors still performed better and within specifications for the iron oxide grade
in glass sand while the “green” NFC collector trials revealed it can reduce the iron oxide grade to half
the feed grade. Being insufficient to create a glass sand with less than 130 ppm of iron oxide.
Resumo
A flutuação por espumas é um dos métodos mais utilizados em processamento mineral por
todo o mundo. A versatilidade dos minérios que esta técnica permite separar e a capacidade para
separação de partículas muito finas tornam-no num processo utilizado mundialmente em grande parte
das indústrias extractivas não energéticas. No entanto, este processo requer, com frequência, o uso de
químicos na polpa, tendo geralmente, impactos negativos na água de processo usada na flutuação por
espumas, podendo esta água ser de difícil tratamento ou estabilização.
Nas últimas décadas a actividade mineira na Europa tem abrandado, sendo que uma das
muitas razões para tal, será a preocupação ambiental das indústrias, incluindo a industria extractiva.
Hoje em dia, a necessidade de produzir os seus próprios recursos geológicos e minerais tem de ter em
conta a crescente consciência ambiental das populações, indústrias e nações europeias. A resposta
estará em novas técnicas de produção mineira, novos métodos de tratamento de águas e solos,
aumento de recuperações de recursos geológicos e minerais e na redução do impacto ambiental dos
rejeitos de lavarias têm de ser alcançados para se conseguir manter o actual estilo de vida, mantendo
a Europa como um continente competitivo.
O foco do presente trabalho tem o seu ênfase na redução do impacto das operações de
processamento mineral, mais específicamente no processamento através de flutuação por espumas.
O trabalho laboratorial e a presente tese foram desenvolvidas em colaboração com uma companhia
internacional, com operações em Portugal, com o Instituto Superior Técnico (IST) e com a Universidade
de Oulo, Finlândia. A Faculdade de Ciências e Tecnologias (FCT) fundou o projecto CELMIN, um
projecto ERA-min para a utilização de reagentes “verdes” nas indústrias extractivas não energéticas.
O trabalho desenvolvido entre Março e Maio de 2015 nas instalações da companhia teve como
objectivo testar diferentes colectores, “tradicionalmente” usados e um novo colector “verde”, na
remoção de minerais pesados, com ênfase na remoção de óxidos de ferro, para produção de areia de
vidro, através de flutuação inversa por espumas. A instalação de processamento da companhia já
utilizou vários colectores “tradicionais” ao longo dos anos da mesma família de colectores aniónicos
oxidrílicos para flutuar as partículas de óxidos de ferro. Deste modo, foi desenvolvido trabalho
laboratorial para criar um ponto de comparação entre três colectores “tradicionais” e um novo colector
“verde”.
A Universidade de Oulo criou vários tipos de celulose nanofibrilada, capaz de adsorver
partículas puras de hematite (iron oxide), sendo que os testes em microflutuação revelaram melhores
performances com a celulose nanofibrilada com o isômero n-butilamina. Esta celulose foi enviada para
Portugal para ser testada com a areia vidreira da companhia através de flutuação laboratorial. Testes
preliminares foram realizados com este tipo de colector “verde” (NFC) com o objectivo de perceber as
concenctrações necessárias de colector na polpa, bem como as condições de pH e concentração de
espumante, mudando uma variável de cada vez (OVAT).
Os resultados dos testes preliminares com NFC forneceram dados que permitissem a criação
de um plano factorial completo de experiências 33(DOE), definindo os factores variáveis e a sua
amplitude. Estes foram a concentração de colector (g/t), o pH da polpa e o tempo de condicionamento
da polpa (minutos). As variáveis resposta foram o teor de óxido de ferro no afundado (ppm), a
recuperação de óxido de ferro no flutuado (%) e o rendimento em peso do flutuado (%).
Infelizmente, os colectores “tradicionais” apresentaram melhores performances, dentro das
especificações da empresa, para o teor de óxido de ferro na areia vidreira. O colector NFC, apesar de
reduzir o teor de óxidos de ferro na areia vidreira para metade do teor de alimentação, nunca se revelou
capaz de baixar o teor em óxidos de ferro na areia abaixo dos 130 ppm, especificação da companhia.
Na minha opinião, o colector NFC, tal como é, pode ser mais eficaz em areias com granulometria mais
finas, apesar de o seu uso para areia vidreira não ter tanto valor.
Index
Abstract ............................................................................................................................................. iv
Resumo .............................................................................................................................................. v
1. Introduction .................................................................................................................................1
1.1. Glass sand ..........................................................................................................................1
1.2. Froth Flotation .....................................................................................................................1
1.3. “Green” Flotation .................................................................................................................2
1.4. Thesis Organization .............................................................................................................3
Chapter 2: Froth Flotation – a brief overview ...................................................................................3
Chapter 3: Silica Sands ...................................................................................................................3
Chapter 4: Material and Methods.....................................................................................................3
Chapter 5: Results and Discussion ..................................................................................................4
Chapter 6: Conclusion .....................................................................................................................4
Chapter 7: Further Work ..................................................................................................................4
2. Froth flotation – a brief overview ..................................................................................................5
2.1. Three phases of froth flotation ..............................................................................................6
2.1.1. Mineral phase ..............................................................................................................6
2.1.2. Liquid phase ................................................................................................................7
2.1.3. Air phase .....................................................................................................................8
2.1.4. Understanding the interphases .....................................................................................8
2.2. Chemical reagents in flotation ............................................................................................ 11
2.2.1. Frothers ..................................................................................................................... 11
2.2.2. Collectors ................................................................................................................... 12
2.2.3. Modifiers .................................................................................................................... 14
3. Silica sands ............................................................................................................................... 15
3.1. Glass Sand ........................................................................................................................ 16
3.1.1. Glass sand deposits ................................................................................................... 16
3.1.2. Chemical specifications .............................................................................................. 17
3.1.3. Particle size specifications.......................................................................................... 18
3.2. Processing......................................................................................................................... 19
3.3. Environmental impact of glass sand production .................................................................. 19
3.4. Froth flotation in industrial sands ........................................................................................ 20
3.4.1. Collectors in industrial sand ........................................................................................ 20
4. Material and Methods ................................................................................................................ 23
4.1. Sample .............................................................................................................................. 23
4.2. Flotation reagents used in laboratory ................................................................................. 26
4.2.1. Tested Collectors ....................................................................................................... 27
4.2.2. Tested Frothers.......................................................................................................... 27
4.2.3. Tested Modifiers ........................................................................................................ 27
4.3. Laboratorial Equipment ...................................................................................................... 28
4.4. Laboratorial Test Procedures ............................................................................................. 30
4.4.1. Experimental plan ...................................................................................................... 30
5. Results and Discussion ............................................................................................................. 35
5.1. Iron oxide analysis ............................................................................................................. 35
5.2. Analysis of results .............................................................................................................. 36
5.3. NFC preliminary tests ........................................................................................................ 40
5.4. NFC Design of Experiments (DOE) .................................................................................... 41
5.4.1. Manipulated factors influence in the flotation of iron oxide .......................................... 42
5.4.1.1. Pearson correlation coefficient................................................................................ 44
5.4.1.2. ANOVA - Iron oxide grade ...................................................................................... 46
5.4.1.3. ANOVA - Iron oxide Recovery ................................................................................ 50
5.4.1.4. ANOVA - Weight Pull ............................................................................................. 54
5.5. Optimization of Laboratory Froth Flotation .......................................................................... 60
6. Conclusion ................................................................................................................................ 62
7. Future work ............................................................................................................................... 64
References ....................................................................................................................................... 65
ANNEXES .......................................................................................................................................... I
Annex I ........................................................................................................................................... I
Annex II .......................................................................................................................................... I
Annex III .........................................................................................................................................II
Index of Equations
Equation 1 - Interphase between tensile forces and air bubble/mineral surface contact angle in equilibrium ...... 9 Equation 2 - Work of adhesion between solid and air......................................................................................... 9 Equation 3 - Surface charge of an oxide mineral .............................................................................................. 10 Equation 4 - Iron oxide recovery (%) in floated material ................................................................................... 23 Equation 5 – Weight pull (%) in floated material .............................................................................................. 23 Equation 6 - Pearson correlation coefficient formula........................................................................................ 45 Equation 7 - Final iron oxide equation of actual factor for reduced 2FI response surface................................... 47 Equation 8 - Final iron oxide recovery (%) equation of actual factor for reduced 2FI response surface ............... 51 Equation 9 - Final weight pull (%) equation of actual factor for tranformed and reduced quadratic response
surface............................................................................................................................................................ 56
Index of Tables
Table 1- Mineral classification by polarity (Wills & Napier-Munn, 2006). ............................................................ 7 Table 2 - Chemical analyses of some European and North American typical glass sands, adapted from (McLaws,
1971). ............................................................................................................................................................. 18 Table 3 - One Variable At a Time (OVAT) traditional collectors variable factors and ranges. ............................. 31 Table 4 - Number of trials with traditional collectors when one variable was modified. In Annex IV there is a list
of all the tests and conditions used. ................................................................................................................. 32 Table 5 - Preliminary tests with NFC collector (61 g/t) to understand impact of frother type and concentration
and pH influence. ............................................................................................................................................ 32 Table 6 - Variable and constant factors throughout the NFC Design of Experiments (DOE). .............................. 33 Table 7 - 3 level full factorial Design of Experiments......................................................................................... 34 Table 8 – Analysis of uncertainty in 10 readings for the same sunken product of NFC trial not using Jones
Sampler. ......................................................................................................................................................... 35 Table 9 - Pearson correlation coefficients matrix between independent variables and responses. ..................... 45 Table 10 - ANOVA for iron oxide grade (ppm) Response Surface 2FI model. ...................................................... 46 Table 11 - ANOVA for iron oxide grade (ppm) Reduced Response Surface 2FI model. ........................................ 47 Table 12 - ANOVA for iron oxide Recovery (%) Response Surface 2FI model. ..................................................... 51 Table 13 - ANOVA for Iron oxide Recovery (%) Reduced Response Surface 2FI model. ....................................... 51 Table 14 - ANOVA for Weight Pull(%) Response Surface Quadratic model. ....................................................... 54 Table 15 - ANOVA for Weight Pull (%) Reduced Response Surface Quadratic model.......................................... 55 Table 16 - ANOVA for Weight Pull (%) Reduced Response Surface Quadratic model after Box-Cox transformation
with k = 0,06. .................................................................................................................................................. 56 Table 17 - Variable constraints for numerical optimization. ............................................................................. 60 Table 18 - Solutions for process optimization. .................................................................................................. 60 Table 19 - Estimated answer for both optimal solutions and estimated confidence intervals. ........................... 61
Index of Figures
Figure 1 - Froth flotation principles (Wills & Napier-Munn, 2006). ...................................................................... 5 Figure 2- Contact angle between air bubble and mineral surface in an aqueous medium, (Wills & Napier-Munn,
2006). ............................................................................................................................................................... 8 Figure 3 - Electrical double layer model of a mineral surface in a liquid solution (Kelly & Spottiswood, 1982). ... 10 Figure 4- Frother action (air-water interphase), adapted from (Wills & Napier-Munn, 2006). ........................... 12 Figure 5- Collector categories, (adapted from Wills & Napier-Munn, 2006). ..................................................... 12 Figure 6- Ionic collector adsorption on mineral surface, (Wills & Napier-Munn, 2006) ...................................... 13 Figure 7- Adsorption of anionic collector on a positively charged alumina particle. (Karimi et al. 2008) ............ 13 Figure 8 - Great Britain Silica sand supply chain in 2007 (British Geological Survey, 2009). ............................... 16 Figure 9 - Generic flowsheet of "glass sand" processing facility (adapted from Alonso, 2014)…………………………..19 Figure 10 - Iso-butylamine NFC sample seen with TEM, (Laitinen et al, 2014). .................................................. 21 Figure 11- Amine groups added to NFC, (Laitinen et al., 2014). ........................................................................ 22 Figure 12 - Studied industrial reverse froth flotation flowsheet. ....................................................................... 24 Figure 13 - Fine glass sand. ............................................................................................................................. 24 Figure 14 - XRF gangue mineral analysis of laboratory feed and industrial annual average feed. ...................... 25 Figure 15 - Particle size distribution for laboratory flotation feed. Retained weight % and cumulative passing %. . Figure 16 - Three “traditional” collectors tested in laboratory .......................................................................... 27 Figure 17 - Denver D12 Laboratory Flotation machine and 3000 cm3.stainless steel container .......................... 28 Figure 18 - pH meter. ...................................................................................................................................... 28 Figure 19 - Ecocell oven. .................................................................................................................................. 29 Figure 20 - Jones sampler. ............................................................................................................................... 29 Figure 21 - Minipal 4 Spectrometer by Panalytical. .......................................................................................... 29 Figure 22 - pH and frother influence (2,2 g/t) on the iron oxide grade using Trad 1 collector with a concentration
of 268 g/t........................................................................................................................................................ 37 Figure 23 -pH and frother influence in iron oxide grade in the sunken product using Trad 2 collector with a
concentration of 192 g/t. ................................................................................................................................ 37 Figure 24 - Trad 3 concentration influence in iron oxide grade in the sunken product with Teepol frother and pH
of 7. ................................................................................................................................................................ 38 Figure 25 - pH and frother influence in iron oxide grade in the sunken product using Trad 3 collector with a
concentration of 200 g/t. ................................................................................................................................ 38 Figure 26 – Iron oxide grade in the sunken product of the tests performed with the three different traditional
collectors in different conditions of pulp pH. .................................................................................................... 39 Figure 27 – Iron oxide grade in the sunken sand best results for different traditional collectors in different
concentrations with a pulp of pH 7. ................................................................................................................. 40 Figure 28 – Iron oxide grade in the sunken product using 61 g/t of NFC collector concentration for different pH
and frother concentration. .............................................................................................................................. 40 Figure 29 – Iron oxide grade in the sunken product with different NFC collector concentration with pH 7 and 1,5
g/t MIBC frother. ............................................................................................................................................ 41 Figure 30 - Manipulated factors and related responses for each trial of the design of experiments. .................. 42 Figure 31 - Iron oxide grade in the sunken product relating the NFC concentration and the pH for a conditioning
time of a) 3 minutes, b) 5 minutes and c) 7 minutes. ........................................................................................ 42 Figure 32 - Iron oxide recovery (%) in the floated product relating the NFC concentration and the pH for a
conditioning time of a) 3 minutes, b) 5 minutes and c) 7 minutes. .................................................................... 43 Figure 33 - Weight pull (%) of the floated product relating the NFC concentration and the pH for a conditioning
time of a) 3 minutes, b) 5 minutes and c) 7 minutes. ........................................................................................ 44 Figure 34 - Negative correlation between iron oxide grade in the sunken product and the iron oxide recovery in
the sunken product (-0.967 Pearson coefficient). ............................................................................................. 46
Figure 35 - Surface of reduced 2FI for iron oxide grade according to collector conc. and conditioning time for 8,5
pH. ................................................................................................................................................................. 47 Figure 36 - Normal Plot of Residuals for iron oxide grade (ppm) Reduced Response Surface. ............................ 48 Figure 37- Residuals versus Predict iron oxide grade values.............................................................................. 48 Figure 38 - Residuals versus Random Run Number. .......................................................................................... 49 Figure 39 - Interaction between Collector concentration (Cc) and Conditioning Time (Ct) ................................. 49 Figure 40 - Surface of reduced 2FI model for iron oxide Recovery (%) according to collector concentration and
conditioning time for 10 pH. ............................................................................................................................ 52 Figure 41 - Normal Plot of Residuals for iron oxide Recovery (%) Reduced Response Surface. ............................ 52 Figure 42 - Residuals versus Predict iron oxide Recovery (%) values. ................................................................. 53 Figure 43 - Residuals versus Random Run Number. .......................................................................................... 53 Figure 44 - Interaction between Collector concentration (Cc) and Conditioning Time (Ct) ................................. 54 Figure 45 - Surface of reduced quadratic model for Weight pull (%) according to collector concentration and pH
for conditioning time of 5 minutes. .................................................................................................................. 56 Figure 46 - Normal Plot of Residuals for Weight pull (%) Reduced Response Surface. ........................................ 57 Figure 47 - Residuals versus Predict Weight recovery (%) values....................................................................... 57 Figure 48 - Residuals versus Random Run Number. .......................................................................................... 58 Figure 49 - Desirability response surface for maximum iron oxide recovery in the floated product and minimum
grade in the sunken product for a conditioning time of 3 minutes. ................................................................... 61 Figure 50 - Samples of floated and sunken optimal trials. Solution 1 (right) and Solution 2 (left). ...................... 61
1
1. Introduction
1.1. Glass sand
Glass sand has usually a large percentage of silica (SiO2), the desired mineral, and other oxides,
such as aluminium, iron and titanium oxides, commonly considered as contaminants, with a restricted
grade on the final product to meet client specifications. The emphasis of this thesis is the removal of
iron (III) oxide (Fe2O3). The smelting of a glass sand with high iron oxide grade produce
nonhomogeneous glass with visual and physical deficiencies (Chammas. E, et al, 2001). Iron oxide
removal is then of maximum importance for the glass sand producer.
Glass sand producers often remove iron oxide contaminants recurring to spirals or, if very low
iron oxide grade is required, with froth flotation. Iron oxides and other heavy minerals (contaminants)
are usually removed through reverse froth flotation, leaving a purified sunken glass sand.
The present thesis was carried out with a sample from an industrial sand processing facility.
The company had the interest in trying different collectors for their reverse froth flotation circuit with
similar properties than the one already in use and also to try a novel green collector (NFC) with different
composition for the same purpose. For confidentiality issues, the industrial facility will be referred further
on as “case study”.
In this “case study” the final product could not have more than 130 ppm of iron oxide. The pulp
was conditioned with an oxydrilic anionic collector (traditional) and a promoter with frothing properties
at a neutral pH achieved by addition of a basic reagent.
With this thesis the main goal was to evaluate the capability of a new green collector (NFC)
made from nanofibrilated celluloses to remove iron oxide particles from the glass sand, achieving a
saleable product. The NFC collector capability of removing iron oxide from the purified sand was also
compared with the oxydrilic anionic collector with proven industrial results and other two similar
collectors.
1.2. Froth Flotation
Froth flotation process was invented in the beginning of the XX century (Wills & Napier-Munn,
2006), with further developments, it became one of the most versatile and productive mineral processing
techniques for low grade or small particle size ores and polymetallic sulphide ores. This mineral
processing method is based on physical (density, size and shape) and chemical differences in the
surface of mineral particles.
2
Usually, chemical reagents are added to the pulp to create an environment capable of
separating two different minerals or groups of minerals. The froth flotation process may concentrate
directly the desired species, by making them float, or reversely, where the non-desired mineral species
float being removed with the froth.
This process often requires the addition of collector, frother and modifiers. The pulp is
conditioned with a collector at a specified pH to guarantee that it reacts with desired particle surfaces
making particles to float hydrophobic.
The size and percentage of free particles, the complexity of the ore to be processed and another
multitude of variables may be adjusted to achieve an effective separation of mineral species. The variety
of chemical reagents in the market and their concentration and interaction in the pulp as well as impeller
speed and air flow rate all make froth flotation as one of the most complex mineral processing methods.
A first stage of flotation requires that collectors and modifiers are able to interact with the mineral
particles in the pulp with no air flow added, to prevent any flotation, it is called conditioning time. After
conditioning, the flotation begins, usually with the addition of frother and air. The particle surface
selectively modified by the collector will be rendered hydrophobic and by making contact with an air
bubble the particle will be carried to the pulp surface once their weight is won. The frother is added to
reduce air/water superficial tension, strengthening the air bubbles and easing the floated particles to
remain in the froth, where they can be paddled out.
1.3. “Green” Flotation
The concern with sustainability and social welfare reflects on the regular testing of greener and
more efficient processing techniques and chemicals created the partnership of the “case study”
company with the IST-ID in the CELMIN project. This study was born from this partnership with the goal
to study better alternatives, in this case, regarding a greener collector with low environmental impact for
iron oxide removal, a Nanofibrillated Celluloses (NFC) based collector produced by a project partner,
Department of Fibre & Particle Engineering of Oulo University.
The work was developed at the “case study” company facility. The aim was to find out the ability
of this new reagent to substitute traditional, more pollutant, collectors of iron oxide minerals present in
sands. For that a comprehensive set of batch flotation tests were planned and executed. The results
were evaluated and described in the present document.
The present thesis intends to show the availability of the NFC to substitute traditionally used
oxydrilic anionic collectors for iron oxide removal from glass sands and if possible improve the NFC
collector. The laboratorial work and the sample analysis were made in the “case study” facilities and the
traditional collectors and the modifiers were provided by the “case study” company an.
3
This implied that a time frame was given for the whole laboratorial work, which traditional
collectors to test, as well as the timing for each collector to test. Therefore the collectors were not tested
at the same length, the analysis were time consuming and should not interfere with “case study”
production, as well as the laboratory use.
1.4. Thesis Organization
After this introductory chapter, where a brief statement of the purposes of this study is
presented, the thesis comprehends five more chapters:
Chapter 2: Froth Flotation – a brief overview
The importance and complexity of this mineral processing technique is referred here. A more
emphasized description of the three different phases present in froth flotation and their interaction is
given. The objective of adding different chemical reagents in froth flotation process and their differences
can also be seen in this chapter.
Chapter 3: Silica Sands
The importance of sand in the present world market and its multitude of applications as well as
the industrial sand definition is approached in this chapter. The silica sand as a category of sand with
high grade in silica and the properties that make it a “glass sand” and the processing methods required
worldwide to provide glass industry with its most important raw material is also referred in this chapter.
Chapter 4: Material and Methods
This chapter has the characterization of the sample and its sampling points. Further onto the
chapter the reagents used in the batch flotation experiments are described according to their purpose.
In the end of the chapter the laboratory equipment and the experimental procedure is defined as one
variable at a time (OVAT) preliminary tests with traditional collectors and NFC collector and the outline
of the factorial plane of experiments with NFC is also defined.
4
Chapter 5: Results and Discussion
The results of batch experiments with traditional collectors with one variable change is
presented as well as with NFC. Further on the models to minimize iron oxide grade and recovery and
to maximize weight recovery based on the factorial plane of experiments with an NFC collector are
shown. Through the models, the optimal conditions of the process are achieved.
Chapter 6: Conclusion
In this chapter the conclusions of the study are described regarding the results of both the
traditional collectors and the NFC collector batch experiments.
Chapter 7: Further Work
Further work is proposed to optimize the flotation process using the tested collectors and the
novel green reagent possible use in the “case study” facilities.
5
2. Froth flotation – a brief overview
Froth flotation is one of the most important processing techniques used for concentration or
purification of minerals. It had its first industrial success in Australia at the beginning of the XX century
(Wills & Napier-Munn, 2006) and has since became one of the more important and versatile mineral
processing methods. The evolution of this method allowed the processing of low grade ores, fine ores
and the production of concentrates from a complex ore, for example, by using selective froth flotation in
massive sulphide ores.
This processing method relies on different physical and chemical properties of particles and
particle surfaces in interphases with water and air. The physical separation of particles occurs due to
molecular, interatomic and gravity forces between the particle surface and the surrounding environment,
with the addition of chemicals to confer floatability to particles with certain surface characteristics.
In a theoretical situation where all particles have the same size and density, the separation is
achieved by the different wettability of the particle surface (Figure 1). It is worth to notice that the terms
hydrophobicity and floatability are intertwined but they refer to different properties. While hydrophobicity
is a thermodynamic characteristic, floatability is a kinetic characteristic, incorporating other particle
properties affecting responsiveness to flotation (Wills & Napier-Munn, 2006).
Figure 1 - Froth flotation principles (Wills & Napier-Munn, 2006).
Even though the principles of froth flotation are pretty straight forward, the truth is that it is a
very complex process. The action of the collector in a pulp with certain pH level, different minerals,
composition of the processing water, oxidation of particles or agglomeration of floated particles making
6
them sink are just some of the difficulties that may be faced in a froth flotation industrial process, making
it one the most studied mineral processes.
2.1. Three phases of froth flotation
Froth flotation is characterised by a three phase system consisting of solids, air and water with
interaction in their interphases. In all the interphases exists surface energy that is crucial to the behaviour
of the flotation, each of the phases are exposed to transformations in the pulp. The surface energy is
responsible for the solubility, adsorption or non-adsorption of reagents in the interphases (Bulatovic,
2007).
2.1.1. Mineral phase
This is obviously a very complex phase not only because it may consist of different minerals but
because they suffer many changes. Therefore, different ions may be released from the particle surface
to the liquid, adding complexity to this phase.
Mineral particles are variable in shape and size and a particle is very often made of different
minerals due to liberation or inclusions. Consequently, particles physical and chemical properties can
have subtle differences due to heterogeneity, so theories regarding mineral floatability can only be an
approach to real froth flotation.
To optimize the froth flotation process not only the desired species requires full knowledge but
also the other non-desired mineral species. The crystalline network energy of the mineral determines
its capability of being hydrated, so in a general approach, a more stable crystalline network has less
polarity, having less potential in the solid-water interphase thus being more hydrophobic and easier to
float.
Despite all the heterogeneities above mentioned some mineral classifications have been made
according to their aptitude to float. In Table 1 the minerals are classified into 5 groups by their polarity
magnitude, increasing the polarity from left to right. Sulphides (Group 1) present less polarity, being
more hydrophobic and at the other end iron oxides (Group 4) and silicates (group 5) having more
polarity, being more hydrophilic.
7
Table 1- Mineral classification by polarity (Wills & Napier-Munn, 2006).
2.1.2. Liquid phase
Liquid phase is where the physical separation of the particles occurs and the properties of this
phase have a significant influence in the particle surface physical and chemical properties. Not only
because of the added reagents but also because dissolution of minerals liberates ions and the process
water has impurities and a given temperature, changing the ionic composition of the liquid phase. With
the liberation of ions from mineral particle surfaces to the liquid phase the surface becomes electrically
charged.
The formation of hydration sheaths around newly arrived ions to the liquid means that the energy
of the bond between ions and water dipoles is greater than between water dipoles. This energy of
hydration relies not only on the ion valence but also on the polarity, temperature and others. It is believed
that dissolution occurs when the hydration energy is greater than the lattice energy (Bulatovic, 2007).
8
2.1.3. Air phase
In froth flotation, air phase occurs with the aeration of the pulp occurring as air bubbles in the
flotation device. The air bubbles attach to the hydrophobic minerals surfaces, conferring a certain degree
of floatability to the mineral particle, capable or not of carrying it onto the surface. The air phase is also
responsible for creating a froth on the surface of the liquid that can sustain hydrophobic minerals until
removal. The dissolved oxygen provided by the aeration of the pulp also helps to achieve selective
flotation.
2.1.4. Understanding the interphases
The complexity of the interactions occurring in the 3 interphases isn’t yet thoroughly known even
though their fundamentals have been very well described almost a hundred years ago by Irving
Langmuir (Langmuir, 1920). He stated that the tendency of the particles to attach themselves to the
bubbles of the froth is measured by the contact angle formed between the oily surface of the bubble and
the contaminated surface of the solid. The selective action by which substances, like galena, are
separated from quartz is dependent upon the contact angle formed by the oiled surface rather than by
any selective tendency for the oil to be taken up by some minerals more than by others. In a reference
to the difference of the contact angle of hydrophobic (galena) and hydrophilic (quartz) with the particles
and the air bubble when covered by an oil.
The contact angle of the air bubbles with the mineral particle surface can be considered as a
measure of wettability or hydrophilic properties of the particle surface. The understanding of the
hydrophilic or inversely the hydrophobicity of the mineral particles surface in the pulp are of paramount
importance for a froth flotation process to be successful. What Langmuir, 1920 observed was that the
contact angle of the air bubbles with the surface of the mineral was really a measure of its hydrophobicity
(Figure 2).
Figure 2- Contact angle between air bubble and mineral surface in an aqueous medium, (Wills & Napier-Munn,
2006).
9
In Figure 2 the forces responsible for the separation or adsorption of the air bubble are pictured.
Each one of the three interphases having a respective surface energy, ϒs/a (solid/air), ϒw/a (water/air)
and ϒs/w (solid/water) with θ being the contact angle between the air bubble and the mineral surface.
The interphase free energy in equilibrium is given by the Young equation as in Equation 1.
ϒs/a = ϒs/w + ϒw/a × 𝑐𝑜𝑠 𝜃 Equation 1
The work of adhesion (Ws/a) between the solid and the air bubble is the amount of force required
to break the interphase and is described in Equation 2. A wettable or hydrophilic surface as a resulting
contact angle tending to 0º, inversely a hydrophobic surface as a contact angle tending to 180º.
Ws/a = ϒw/a × (1 − 𝑐𝑜𝑠 𝜃) Equation 2
In the eighties of the last century, the electrical double layer model was described (Fuerstenau,
1982) defining that a separation of electrical charge occurs within the mineral/water interphase (Figure
3), with a positive charge layer and a negative charge layer, giving electrical neutrality to the system.
The double layers could extend to one or both phases being a system of electrokinetic and
electrochemical energies.
10
Figure 3 - Electrical double layer model of a mineral surface in a liquid solution (Kelly & Spottiswood, 1982).
The activity of the potential determining ions corresponding to a null surface charge is called
Point of Zero Charge (PZC), being the conditions that lead to a PZC of critical importance since the
adsorption of the collector is defined by the signal and value of the surface charge (Durão, Cortez, &
Carvalho, 2002).
𝜎 = Ϝ(г𝐻+ − г𝑂𝐻−) Equation 3
The surface charge (𝜎) is determined by the adsorption density (г) of potential determining ions
at the mineral surface, in many cases like the oxides, the H+ (cation) and OH- (anion). The
electrochemical potential changes with the activity of the potential determining ions as in the lower part
of Figure 3.
11
The inner layer attracts counter ions of the solution by Coulomb or electrostatic forces bonding
with the potential determining ions, this is called Stern layer. Adjacent to the Stern layer there is a diffuse
layer of counter ions, the Gouy layer. The movement of the solution creates a shear plane between both
layers. Ions in the Stern layer remain anchored to mineral surface and ions in the diffuse layer departs
creating a new potential between the Stern layer and the new diffuse layer. This new environment
creates a new potential between both phases called electrokinetic potential or zeta potential (ζ), which
defines the electrical composition of the mineral surface.
2.2. Chemical reagents in flotation
“Without reagents flotation wouldn’t exist and without flotation the mining industry wouldn’t exist
as we know it.” (Bulatovic, 2007)
Nowadays a wide range of reagents allow, through direct or inverse froth flotation, the
concentration of almost all mineral species. These can be frothers, collectors and modifiers which can
act as pH regulators, activating or depressing agents. They act in the mineral particles surfaces, in the
diffuse layer, in the pulp or in films around small parts of the surface of the particle (Bulatovic, 2007).
Only some sulphide minerals, coal and plastic polymers can float effectively without any added
chemicals, although these can be used to increase selectivity or to increase flotation speed.
2.2.1. Frothers
Frothers reduce the superficial tension in the air-water interphase, increasing the duration of the
air bubbles during agitation of the pulp. By diminishing the volume of the air bubbles a better dispersion
is achieved and the air bubble raises up slower, increasing the residence time, increasing the probability
of the air bubble to collide with a particle. The kinetics of the flotation process is increased.
The frothers are made of organic heteropolar molecules with a hydrophilic polar group and a
non-polar, hydrophobic group (Figure 4). In industrial practice it is common to use alcohols, fatty acids
or amines, being the more common radicals of the polar group the hydroxides, carbonyls, amines and
carboxyls.
12
Figure 4- Frother action (air-water interphase), adapted from (Wills & Napier-Munn, 2006).
2.2.2. Collectors
A vast spectrum of collectors are used nowadays in industrial mineral processing facilities and
they are categorised according to the active ion and the molecular structure (Figure 5). The adsorption
of the collector in the mineral particle is made with the polar group of the reagent and the non-polar
group ensures the connection with air bubbles. Ionic collectors can be cationic or anionic, with a free
radical ensuring the solubility of the collector (Figure 6). Non-ionising collectors are practically insoluble
and make thin layers around the mineral particles rendering them hydrophobic. Some collectors possess
a cationic and an anionic function depending on the working pH of the pulp, these are called amphoteric
collectors.
Figure 5- Collector categories, (adapted from Wills & Napier-Munn, 2006).
Collectors
Non-ionic
Hydrocarbons and derivates
Ionic
CationicAmines,
N5+
Anionic
OxyhydrylSulpho acid or
organic
Carboxylic Sulphates Sulphonates
Sulphydhryls
Bivalent sulphur
Xanthates Dithiophosphate
Frother
molecule
13
The addition of collectors intends to assure the adhesion of mineral particles to the air bubbles
due to their hydrophobicity. The most important group is the ionising collectors which are widely used,
where the compounds dissociate into ions in the water. They are comprised by a non-polar hydrocarbon
group with a radical with water repellent properties and a polar group.
Figure 6- Ionic collector adsorption on mineral surface, (Wills & Napier-Munn, 2006)
The influence of the collector in a selective process of flotation depends on the association of
the polar group with the mineral surface. The non-polar part of the collector is responsible for the power
of the collector, increasing with longer hydrocarbon chains.
Collectors are usually used in small concentration because an excess of concentration may
originate multi-layers of collector around the mineral particles, forming ad-micelles, with the polar group
facing the water the hydrophobicity is reduced (Figure 7). To enhance the flotation system, the use of
more than one collector is usual, with a selective collector added in the beginning of the circuit to float
very hydrophobic particles and after that a more powerful collector with longer hydrocarbon chains to
recover less hydrophobic mineral particles.
Figure 7- Adsorption of anionic collector on a positively charged alumina particle. (Karimi et al. 2008)
14
2.2.3. Modifiers
The selectivity of the froth flotation process in industrial practice is rarely achieved without the
use of any modifying agent. They are classified as activators, depressants and pH modifiers although
this should not be looked at as a strict classification since the role of each reagent depend upon the
particle composition of the pulp and its conditions.
These can have impact on the process by altering the minerals surface, increasing or
decreasing collector adsorption on some particles. By removing layers of the mineral surface these can
be rendered hydrophilic, depressing it. Altering the pH of the pulp modifies the collector adsorption to
the surface of the mineral, having an activator or depressor capabilibty (Bulatovic, 2007).
Activators or depressants role is to increase or decrease, respectively, the collector adsorption
in the mineral surface. Their use might have several purposes depending upon its addition point in the
flotation circuit, the quantity, the pulp pH and its constituents, minerals and reagents.
Regulators of pH have a fundamental role in froth flotation processes because the ions H+ and
OH- interact with the particle surface and the collector and other reagents adsorption to the mineral
surface. Generally common acid is used when the pulp requires acid pH values, although it is preferable
to work with alkaline pulps due to more stable conditions of the collector and lesser corrosion of the
cells. To increase alkalinity lime or sodium hydroxide are commonly used, with some plants using
ammonia as well (Wills & Napier-Munn, 2006).
The balance of the pH level of the pulp and reagent concentrations in selective froth flotation
process is very subtle and hard to fully understand. Thorough experimentation has to be performed in
order to optimize selectivity in a complex ore pulp.
15
3. Silica sands
A sand deposit is made of unconsolidated or loosely consolidated silica particles, the silicon
compounds altogether constitute about 28% of th earth’s crust, thus being very abundant (Sundarajan
et al, 2009). It is a remarkably stable mineral with high chemical and mechanical resistance due to its
composition where an oxygen atom is shared between two silicon atoms in a three-dimensional
structure.
Silica sands have a widespread distribution worldwide and their physical, chemical and thermal
resistance allied with a low price make them one of the most used non-metallic minerals with an
extensive range of application. The United States Geological Survey (U.S.G.S) estimated that in 2010
the world production was 121 Mt of industrial sand (U.S. Geological Survey, 2012), with the United
States (USA) being the major producer, followed by Italy and then Germany, with this three countries
representing 43.86 % of world production (~0.57 Mt) , although China’s production is not pictured in that
study.
Industrial sands and gravels are a very abundant resource, characterised by having a high silica
(SiO2) grade. The quantity of uses it has is astonishing, for example as abrasives, for filtration, foundry,
glassmaking and hydraulic fracturing (U.S. Geological Survey, 2012) as seen on Figure 8. Typically,
several grades of sand are produced from one site either by selective extraction or processing. The
extraction and processing of silica sand generally involves the production of only small amounts of
waste. Including the sale of by-products an average of 90% the reserve yields saleable product (British
Geological Survey, 2009) with the British market posing as an example.
16
Figure 8 - Great Britain Silica sand supply chain in 2007 (British Geological Survey, 2009).
3.1. Glass Sand
The glass making industry most consumed raw material is “glass sand”. It is used to
manufacture colourless glass containers, flint glass (float, sheet and rolled glass) and coloured glass
containers. Those applications generally require a minimum of 98% silica content, a narrow particle size
distribution assuming a low percentage of fines and coarse grains which cause difficulties in the smelting
and refining process (Lines & Echt, 2004).
The glass industry and the glass sand producers need to have a high level of interaction
because though glass sand deposits can be found around the world, the size and characteristics of the
sand deposits and processing level vary. Therefore, glass industry has to adapt to their geographic area
and set an achievable set of physical and chemical specifications for the regional glass sand producers.
3.1.1. Glass sand deposits
Silica sand deposits are characterised by having been repeatedly eroded by fluvial, coastal
marine and aeolian processes and recycled mature sediments. High-quality sands are usually found in
the proximities of peat and emersion surfaces where infiltration water filled with dissolved organic carbon
leaches the alkalis of the soil and reduces the iron to Fe+ causing its dissolution. Further percolation of
17
water through the sand allows the removal of impurities. These processes cause multiple mechanical
and chemical attacks leaving deposits constituted mostly by quartz and most durable heavy minerals
(Pohl, 2011).
Silica sands are produced from loosely consolidated sands and weakly cemented sandstones.
Although sand and sandstone deposits have a geographically wide distribution only a small part of them
are suitable for producing silica sand whether from their physical or chemical characteristics. Each sand
deposit will have differences in purity, particle size, thickness and homogeneity and they all require
some kind of processing to produce a saleable silica sand. The ease of removing impurities and the
inherent particle size of the deposit are the defining factors for a sand deposit to be exploited as silica
sand without losing much of the sand. The market specifications and processing costs create a
restriction to the economic availability of a silica sand deposit to be mined.
3.1.2. Chemical specifications
The wide variety of applications for glass sand require different arrangements of mineralogical
composition. They universally require high silica content (SiO2). The major contaminants being alumina
(Al2O3), affecting the viscosity and density of the glass, titanium dioxide (TiO2), alkalis, such as lime
(CaO), soda (Na2O) and potash (K2O) which affect melting temperatures and especially iron oxides
(Fe2O3) with different restrictions depending on the use it requires.
SiO2 grade in glass sands typically ranges from 98.5% to 99%, Al2O3 from 0.2% to 1.6%, TiO2
under 0.3% and CaO+MgO with a lower grade than 0.05% (Pohl, 2011). The usual maximum Fe2O3
grade allowed in flat glass is 0.04%, flint glass 0.3%, amber containers 0.18% and for fiberglass 0.3%
(Lines & Echt, 2004), with each manufacturer having different requirements (Table 2).
Elements present in some oxidized minerals such as nickel (Ni), copper (Cu), cobalt (Co) and
chromium (Cr) have a colourant effect, as well as iron oxides. Refractory minerals such as andalusite,
zircon, chrome, rutile, staurolite among others which have the highest melting temperatures in glass
sands, create solid inclusions that are translated in a glass product with less physical resistance to
shocks and temperature. The specifications for refractory minerals are related with their grade and with
their particle size, approached in the next sub-chapter.
18
Table 2 - Chemical analyses of some European and North American typical glass sands, adapted from (McLaws,
1971).
Location
Chemical Constituents (%)
SiO2 Fe2O3 Al2O3 CaO MgO Na2O K2O L.O.I.
Ottawa, Illinois 99.61 0.02 0.16 0.05 0.03 - - 0.08
Valley,
Washington 99.60 0.03 0.29 - - 0.02 0.05 0.04
Selkirk, Manitoba 99.70 <0.06 0.1 tr tr tr tr -
Norfolk, England 99.25 0.04 0.59 0.11 0.03 - - 0.25
Fontainebleau,
France 99.80 0.006 0.13 tr - - - 0.18
Belgium 99.12 0.07 0.43 0.34 0.11 - - 0.22
3.1.3. Particle size specifications.
The particle size distribution is the most important physical specification affecting the amount of
energy required to melt the batch material, with uniform grain size granting a more efficient melting and
avoiding segregation. The shape of the grains also affects the melting process, with angular grains being
easier to melt due to their higher surface area than well rounded grains.
Each glass producer has an optimal size requirement but commonly the sand size distribution
for glass making usually ranges from 0.1 to 0.3 mm with a finer distribution for fibre glass manufacturing
with 90% of the grains smaller than 0.45mm (Lines & Echt, 2004). A coarser feed is reflected in an
incomplete melting of the coarser grains and a lower output of the furnace. Finer feed creates more dust
outside the furnace and generate problems in the refractories and heat exchangers of the furnace. The
refractory heavy minerals larger than 0.25mm are a big problem producing poor quality glass.
19
3.2. Processing
Silica sand processing has varying degrees of complexity depending on the raw material and
their final application. It is aimed at improving both the physical and chemical properties of the sand to
meet the user specifications. It often begins with a screen classification and a variable number of stages
of attrition depending on the degree of cementation and wet or dry screening to achieve the desired
particle size distribution.
For the production of colourless glass sand, more processing is necessary to remove
contaminating impurities, either from the sand and/or from the surface of the individual sand grains with
the material regularly grinded by ball, autogenous or rod mills and hydro-classified. If the sand has
contaminants like mica, feldspar or iron bearing minerals, froth flotation and gravity separation using
spiral classifiers are used (Figure 9). High intensity wet magnetic separation is being increasingly used
to remove iron-bearing impurities (British Geological Survey, 2009).
3.3. Environmental impact of glass sand production
Silica sand deposits for the production of glass sand exploitation depends on the degree of
cementation of the silica deposit, it may be completely loose, as sand, or very cohesive, as rock. Thus
the cementation and water level will weight on the chosen exploitation and processing method.
Dredging Screening
Clay and very fine sand
Attrition
Fine fraction
Coarse fraction
Hydrocyclone
Hydro classifier
Spiral classification
Conditioning Tank Flotation cells
Tailings
Tailings
Glass sand
Figure 9 - Generic flowsheet of "glass sand" processing facility (adapted from Alonso, 2014).
20
Particulate matter is emitted during many mining operations, loading, hauling, ripping or blasting
and processing operations, such as conveying, screening, crushing and storing. However, during
handling the material is wet or moist and the emissions are negligible. Most of the dust can be controlled
by barriers in the direction of main winds so that it settles down and in haul roads dust suppressant in
the soil or paving can be very effective (U.S. Environmental Protection Agency, 1995).
Mining operations despite being known for acidic processing water production can also extend
to alkaline or very alkaline conditions, like glass sand processing waters. Acid water principal anion is
sulphate with major cations being aluminium, iron and manganese. In alkaline water sulphate or
bicarbonate are the principal anions and calcium, magnesium and sodium are major cations.
Process tailings consist of solids discharged with process water, usually into a tailings dam.
Superficial and pore water also end up in the tailings dam mixing with waste water which has generally
high concentration of process chemicals (Lottermoser, 2007).
Tailings water typically is decanted for reuse and pumped back to the plant, diminishing the
amount of contaminants in tailing dams. Nevertheless, a part of the waste water remains in the tailing
dam with chemicals like organic collectors that may complex metals, sulfuric acid, collectors interacting
with tailing solids and a loss of oxygen available, which can be troublesome for fauna and flora if a
leakage occurs.
3.4. Froth flotation in industrial sands
Flotation circuits are often used in industrial sand processing facilities, especially if the desired
product is highly purified silica. A reverse froth flotation process is often implemented to deal with some
of the main contaminants, like iron bearing minerals and other heavy minerals.
3.4.1. Collectors in industrial sand
The description of all the families of collectors is a long and complex theme that has been the
subject of many scientific books. In this thesis a description of the oxyhydryl carboxilates collector
“family” traditionally used for removing iron bearing minerals from sand using froth flotation is given. An
additional “novel green collector” was studied to test its ability to replace the traditional collectors in the
industrial glass sand industry.
21
3.4.1.1. Oxyhydryl carboxylates
Oxyhydryl Carboxylates are generally used as collectors for flotation of oxides, silicates and
carbonate materials. These collectors group is mainly composed by oleic acids, sodium oleates,
synthetic fatty acids, tall oils and some oxidized petroleum derivatives. They are usually manufactured
from animal fat or vegetable oils with each producer assembling a different mix of various acids
(Bulatovic, 2007). These are the most used oxyhydryl collectors in industrial practice although their
selectivity heavily relies on pulp preparation, pH and depressants use.
Most of the fatty acids are mixtures of different acids from vegetable or animal origin which are
later distilled, these are known as tall oils. The percentage of each acid and the origin of the fat oils have
a great influence on the power of the collector and its selectivity. The tall oils are commonly used in the
flotation of phosphates, silicates and lithium minerals where the small size of the particles do not allow
gravity concentration processes to be successful.
3.4.1.2. Nanofibrillated celluloses (NFC)
At Oulu University in the Fibre and Particle Engineering laboratories an innovative collector was
created from cellulosic materials which can be chemically modified to have cationic or anionic properties
and hydrophobic behaviour. In Figure 10 a Transmission Electron Microscopy image of a NFC with an
embodied iso-butylamine group is shown.
Figure 10 - Iso-butylamine NFC sample seen with TEM, (Laitinen et al, 2014).
22
The NFC were produced in bench-scale from birch kraft pulp using periodiate oxidation to
produce dialdehyde cellulose and subsequently aminated. With the addition of charged groups to the
celluloses backbone these can act as polymeric collector with selective performances for certain mineral
types. The tested NFC was attached with an n-butylamine group has in Figure 11 (Laitinen et al, 2014).
Figure 11- Amine groups added to NFC, (Laitinen et al., 2014).
An NFC collector was created to enhance the floatability of iron oxides from a sample of the
sand used in the laboratorial studies presented further on this work. A microflotation study was carried
in Oulu University on its ability to float Hematite (~96% feed grade) in neutral or low alkalinity ranges
with some success (Laitinen et al., 2014).
At the “case study” facility where the work was realised an oxydrilic anionic collector was used
for the reverse froth flotation of glass sand with a feed with 100% of the particles under 1mm. The
flotation circuit aimed at floating the contaminants, heavy minerals with emphasis on iron oxide. The
glass sand product specifications regarding the iron oxide grade could not surpass 130 ppm of this
oxide.
23
4. Material and Methods
A sample of the “case study” glass sand was collected to perform froth flotation laboratory trials
with different reagents. The first batch laboratory tests were performed simulating the “case study” froth
flotation circuit with Trad1, an oxydrilic anionic collector. Two additional oxydrilic anionic collectors were
tested, Trad 2 and Trad3. Operating parameters were changed one variable at a time (OVAT), pH of
the pulp, frother type and concentration and collector and type concentration.
The traditional collectors operating parameters were compared regarding their capability in
reducing the iron oxide grade in the sunken product. The optimal performances of each Trad collector
should have been reached using kinetic analysis of the froth flotation process, regrettably these where
impossible to realize at the “case study” facility since the floated product hardly reached the weight
required for analysis. A factorial plan of analysis for each collector would have provided better insight of
their behaviour with the glass sand sample and guaranteeing the repeatability of trials should have been
made, but the time to perform all the trials was limited.
The NFC collector was tested in a first stage changing one operating parameter at a time
(OVAT), the type and concentration of the frother, the pH of the pulp and the collector concentration.
Once again, kinetic trials and repeatability of the OVAT trials should have been performed. A factorial
plan of experiments was designed to understand the influence of the NFC concentration, the pH of the
pulp and the conditioning time in the iron oxide grade (ppm) in the sunken product, the iron oxide
recovery (%) (Equation 4) in the floated product and the weight pull (w%) (Equation 5).
𝛤Fe2O3(%) = (1 − (
(𝐷𝑟𝑦 𝐹𝑒𝑒𝑑 𝑠𝑎𝑛𝑑 (𝑔𝑟𝑎𝑚𝑠)−𝐹𝑙𝑜𝑎𝑡𝑒𝑑 𝑚𝑎𝑠𝑠 (𝑔𝑟𝑎𝑚𝑠))×𝑆𝑢𝑛𝑘𝑒𝑛 Fe2O3(𝑊𝑡 %)
𝐷𝑟𝑦 𝐹𝑒𝑒𝑑 𝑠𝑎𝑛𝑑 (𝑔𝑟𝑎𝑚𝑠)×𝐹𝑒𝑒𝑑 Fe2O3(𝑊𝑡 %))) × 100 Equation 4
𝑊𝑝(%) =𝐹𝑙𝑜𝑎𝑡𝑒𝑑 𝑚𝑎𝑠𝑠 (𝑔𝑟𝑎𝑚𝑠)
𝐷𝑟𝑦 𝐹𝑒𝑒𝑑 𝑠𝑎𝑛𝑑 (𝑔𝑟𝑎𝑚𝑠)× 100 Equation 5
4.1. Sample
The general flowsheet of the processing facility studied is similar to the one given in Chapter
3.2. Dredged sand is screened with the underflow, below 1,6 mm, feeding a hidrocyclone that separates
sand from kaolin clay. The sand is subject to an attrition process and further hydro classified removing
remaining kaolin residues and very fine sand (< 90 µm) in the overflow. The overflow is treated in a
separate circuit. The underflow is further hydro classified to separate coarse sand and fine sand with
average particle sizes of 470 µm and 360 µm, respectively.
24
Both product flows of a second hydro classification suffer a gravity separation process through
spirals, which reduces the heavy minerals grade to half, mainly heavy minerals. The spiralled and
purified sand then feeds a froth flotation circuit (Figure 12).
The flotation circuit begins with a hydro classification of the spiralled sand to remove water from
the slurry. The hydro classification underflow, which was collected for this study, then feeds the first
conditioning tank where the chemical reagents such as frother, collector and caustic soda are added.
These then proceed to a second conditioning tank to increase the conditioning time of the pulp.
The glass sand that feeds the industrial flotation circuit was sampled and stored in moist
conditions into two 250 kg barrels to be used in the laboratory trials (Figure 13). The sand mainly
constituent mineral is silica with the main gangue minerals being iron oxide, alumina, titanium, potassium
and calcium oxides. Before each test the moist sand samples were drained remaining with about 20%
water.
Figure 13 - Fine glass sand.
Figure 12 - Studied industrial reverse froth flotation flowsheet.
Collected sample
25
The chemical analysis of the laboratory flotation feed was obtained using XRF (X-Ray
Fluorescence Spectrometer) and the mean composition of industrial flotation feed, obtained by the same
analytical method, the chemical analysis of both feeds is plotted in Figure 14 regarding the main gangue
minerals. It is to say that the silica (SiO2) grade is 99,64% with a Fe2O3 grade of 230 ppm in the
laboratory feed and 99,43% SiO2 grade and Fe2O3 grade of 550 ppm in the industrial feed.
The tests were carried out with the finer fraction of glass sand produced on the processing
facility. The particle size distribution was determined by sieving using a set of nine sieves (from 1000
µm down to 63 µm). The main two fractions (85,7%) were composed by particles between 500 and 250
µm shown in Figure 15. 80% of the particles were smaller than 452 µm (P80% = 452 µm) and around 13%
of the material was under 250 µm.
0
0,05
0,1
0,15
0,2
0,25
0,3
0,35
Fe₂O₃ Al₂O₃ TiO₂ K₂O CaO MgO Na₂O
Gan
gue
Min
eral
s gr
ade
(Wt%
))
Analysed Gangue Minerals
Laboratory
Industrial
Figure 14 - XRF gangue mineral analysis of laboratory feed and industrial annual average feed.
26
4.2. Flotation reagents used in laboratory
The experiments can be divided in two main phases. Changing one variable at a time (OVAT)
the traditional collectors performance in floating iron oxide particles was tested. Then laboratory trials
(preliminary) with NFC collector were also made using an OVAT approach. The traditional collectors
and the NFC preliminary trials were evaluated according to the iron oxide grade in the sunken product.
The OVAT trials helped understand the efficiency of the traditional collectors and provided
comparison to the pioneer NFC collector. The NFC collector concentration and the alkalinity of the pulp
had the biggest impact in the preliminary batch flotation results regarding the iron oxide grade in sunken
product.
The results achieved using an OVAT approach were taken into account in the construction of a
second phase, where a Design of Experiments (DOE) was made to understand the impact of changing
three factors, the NFC collector concentration, the pH of the pulp and the conditioning time in the iron
oxide recovery, the weight pull and the iron oxide grade in the sunken sand.
Through the DOE the factors influence on the froth flotation was analysed. Each factor had an
ANOVA and response surfaces were modelled accordingly for the iron oxide recovery in the floated
product and the iron oxide grade in the sunken sand.
Figure 15 - Particle size distribution for laboratory flotation feed. Retained weight % and cumulative passing %.
0
10
20
30
40
50
60
70
80
90
100
10 100 1000
Wei
ght
(%)
Sieve mesh size (µm)
Retained
CumulativePassing
27
4.2.1. Tested Collectors
Historically different collectors, frothers, promoters and pH conditions have been used in this
facility to “clean” the sand. The collectors traditionally used in the plant are carboxylates, from the
oxhydryl family. The conditioned pulp feeds Sala International AB cell banks with 22.5 m3 capacity
where the contaminants, heavy minerals, are floated leaving a fine purified sand with an average particle
size of 360µm.
In this work three anionic oxyhydryl carboxylate collectors (the traditionally used in the
processing plant) and the novel “green” collector (a nanofibrillated cellulose (NFC)) incorporated with
amine groups were used.
Although the traditional collectors can be catalogued as carboxylates, their composition and
aspect vary (Figure 16). A previously studied and optimized collector, Trad 1 (on the left), a similar
collector Trad 2 (in the middle) and the more thoroughly studied traditional collector Trad3 (on the right)
were used during this work.
Figure 16 - Three “traditional” collectors tested in laboratory
As referred above the NFC collector was designed to float iron oxide minerals from the original
sand feed of the studied facility on neutral to alkaline pH range. A sample of this collector was used to
perform the first batch laboratory trials.
4.2.2. Tested Frothers
Four different frothers were used in the laboratory trials. The frothers used in the laboratory tests
with traditional collectors were Teepol HB7, a sodium alkyl sulphate frother and a carboxylate promoter
with frothing properties CYTEC Aero 845N p. With the NFC collector Teepol HB7 was also used, as well
as MIBC and Dowfroth 250 C.
4.2.3. Tested Modifiers
Trad 1 Trad 2 Trad 3
28
Besides the promoter with frothing properties described in the previous subchapter a
solution of caustic soda was used to increase the alkalinity of the pulp.
4.3. Laboratorial Equipment
All batch experiments were carried out in a Denver D12 laboratorial flotation machine with a
stainless steel container with 3000 cm3 capacity (Figure 17), with natural air suction and manual
regulation of impeller rotation speed from 1000 to 2500 rpm. Sand and water were weighted in 1000 ml
beakers and the addition of chemical reagents was made using a 5 ml graduated cylinder.
To measure the pH of the pulp a Crison Instruments SA, pH 2000 was used coupled with an
electrode by Hanna Instruments model HI1312 (Figure 18). The material used to remove the froth was
a stainless steel paddle, a metallic board and a wash bottle to rinse the paddle in every paddling.
Figure 17 - Denver D12 Laboratory Flotation machine and 3000
cm3.stainless steel container
29
For the drying of sunken samples, a spoon and a dish were used to sample and store the glass
sand and an EcoCell oven made by MMM Medcenter Einrichtungen GmbH, Model LSISB2V/EC 55
(Figure 19) was used to dry the samples at 110 º C. The floated samples were drained using a filter
paper in a funnel and dried in a hot plate.
A Jones sampler (Figure 20) was used in the DOE analysis for dividing the glass sand in equal
parts.The analyses of iron oxide were made using an Energy Dispersive X-Ray Fluorescence, Minipal
4 spectrometer model by Panalytical (Figure 21).
Figure 18 - pH meter.
Figure 19 - Ecocell oven.
30
4.4. Laboratorial Test Procedures
4.4.1. Experimental plan
The laboratory reverse froth flotation of iron oxide made with the “case study” sand sample had
the following procedure:
1) Collect sample and drain it, take out the exceeding material until 1800g was reached;
2) Put the 1800g of drained sample in the stainless steel container and add 800mL of
water;
3) Turn on the impeller of the laboratory froth flotation at 1000 rpm with the air valve closed;
4) Add the desired quantity of collector and promoter if needed;
5) Add Sodium hydroxide (NaOH) to the pulp and measure the pH of the pulp, keep
monitoring the pH change in the pulp until the desired level was reached, starting the
chronometer to mark the conditioning time;
6) Ending the conditioning time, the air valve was opened;
7) Cautiously collect the froth with a paddle onto a heat resistant bin during the flotation
time (8 minutes);
8) Rinse the paddle onto the heat resistant bin at each paddling;
9) The floated product was then poured onto a funnel lined with filter paper which was
heated onto an hot plate and weighted;
10) The sunken product was carefully drained, sampled and inserted onto an oven to dry at
110ºC. The dry sunken product was again sampled and stored for further analysis.
Figure 21 - Minipal 4 Spectrometer by
Panalytical.
Figure 20 - Jones sampler.
31
4.4.1.1. Traditional collectors
The three traditionally collectors were tested to different extents. Trad 1 collector, had been
optimized for this same sand feed in laboratory batch trials and a concentration of 268 g/t was found to
be optimal for the removal of iron oxides (Alonso, 2014). Trad 2 had historic records of usage in the
processing plant within a range of concentrations with Teepol frother and Trad 3 had never been used
with this sand feed and was tested more thoroughly, so the range of concentrations tested was wider.
For the three different collectors, different pH conditions with 2 different frothing agents were
tried, Teepol and A845N (Table 3). The tested pH range varied from 6.5 (natural pH) to 10. It is to note
that with Trad 1 collector, the Teepol frother was mainly used to simulate the “case study” flotation
while the A845N was tested also to search for a suitable alternative frother. With Trad 2 only one trial
was performed with Teepol frother since the “case study” company already had historic records of its
usage with Teepol, so more emphasis was made on Trad2/A845N laboratory trials.
These tests were planned changing one variable at a time (OVAT) to understand which
variables and their ranges, collector concentration (g/t), type and concentration of frothing agent and
pulp pH performed better in reducing the iron oxide grade in the sunken product.
Table 3 - One Variable At a Time (OVAT) traditional collectors variable factors and ranges.
Collector Frother
pH
Type Concentration (g/t) Type Concentration (g/t)
Trad1 268
Teepol 2,2
6,5-7-8-9-12
A845N 2,2-3,4 -23,5
Trad2 192-384
Teepol 3,7
6,5-7-8-9-10
A845N 5,3-10,6
Trad3 50-100-200-268-300-400
Teepol 5,25
6,5-9
A845N 2,2-3,4-5,25-64
Trad1 Trad2 Trad3
32
Table 4 - Number of trials with traditional collectors when one variable was modified. In Annex IV there is a list of
all the tests and conditions used.
4.4.1.2. Nanofibrillated Celluloses (NFC) preliminary tests
In the present work, the first batch laboratory trials were performed in order to understand which
factors could have greater impact in floating iron oxide from sand feed.
The first preliminary tests carried out with the NFC collector were made to define the better
concentration of Teepol and MIBC frothers as well as pH level, as seen on Table 5. A low concentration
of collector (61 g/t) was chosen and the Teepol and MIBC frothers were tested. Having found that lower
grade in iron oxide in the sunken product was achieved at pH 7 and MIBC frother concentration of 1,5
g/t, a series of preliminary trials were designed to test for the working ranges of the NFC collector
concentration in the pulp.
Table 5 - Preliminary tests with NFC collector (61 g/t) to understand impact of frother type and concentration and
pH influence.
pH Frother
NAME Type Concentration (g/t)
C1 7 Teepol 0,5
C2 7 Teepol 1,5
C3 7 Teepol 2,5
C4 7 Teepol 3,5
C5 9 Teepol 0,5
C6 9 Teepol 1,5
C7 9 Teepol 2,5
Teepol A845N Teepol A845N Teepol A845N
5 6 1 5 11 11 39
11 6 22 TOTAL
33
C8 9 Teepol 3,5
MC1 7 MIBC 0,5
MC2 7 MIBC 1,5
MC3 7 MIBC 2,5
MC4 7 MIBC 3,5
MC5 7 MIBC 5
4.4.1.3. Nanofibrillated Celluloses (NFC) Design of Experiments (DOE)
The manipulated factors chosen to create a 3 level full factorial DOE are summarized in
Table 6, as well as the constant factors used throughout the experiment. The chosen factors to be
manipulated were the NFC collector concentration, the pH level and the pulp conditioning time.
The basic 3 level full factorial design requires 33=27 trials and 5 additional replicates of the mid-
level were made (120 g/t, 8.5 pH, 5 minutes conditioning) as seen on Table 7. The trials were performed
in a random order.
Table 6 - Variable and constant factors throughout the NFC Design of Experiments (DOE).
Factor min mean max units
Manipulated
pH 7 8,5 10
Collector 40 120 200 (g/t)
Conditioning
Time 3 5 7 minutes
Constant %Sp 35
Flotation Time 8 minutes
34
Impeller Speed 1000 rpm
Frother 1,5 (g/t)
For the DOE the flotation time, the frother concentration, the air flow, the sample and its
moisture content were constant throughout the experiment. However, the sample quantity and its
moisture content had some variations since it was not possible to guarantee the exact same drainage
before each flotation test.
Table 7 - 3 level full factorial Design of Experiments.
FACTORS FACTORS
Experiment Collector
(g/t) pH
Conditioning
time (min)
Experiment Collector
(g/t) pH
Conditioning
time (min)
1 40 7 3 17 200 8,5 5
2 40 8,5 3 18 200 10 5
3 40 10 3 19 40 7 7
4 120 7 3 20 40 8,5 7
5 120 8,5 3 21 40 10 7
6 120 10 3 22 120 7 7
7 200 7 3 23 120 8,5 7
8 200 8,5 3 24 120 10 7
9 200 10 3 25 200 7 7
10 40 7 5 26 200 8,5 7
11 40 8,5 5 27 200 10 7
12 40 10 5 28 120 8,5 5
13 120 7 5 29 120 8,5 5
14 120 8,5 5 30 120 8,5 5
15 120 10 5 31 120 8,5 5
16 200 7 5 32 120 8,5 5
35
5. Results and Discussion
5.1. Iron oxide analysis
A sample of the sunken pulp from each trial was analysed for its iron oxide (Fe2O3) grade with
a Minipal 4 spectrometer. It is an expedite analysis method where a dry glass sand sample of around
nine grams is inserted in a capsule with a Mylar Thin-Film by Chemplex in the bottom. The sample is
compressed and introduced in the Minipal 4 spectrometer for reading during 120 seconds.
The sunken pulp from traditional collector trials and NFC preliminary tests were sampled to
analysis only by mixing the sand in the bag and three tea spoons of sand were inserted in the capsule.
The analysis of results of traditional collector analysis had neglectable variability for the same sunken
sample.
The same wasn’t confirmed with sunken samples of NFC trials, where huge variability was
recorded by analysing multiple times the same sample (Table 8). For that reason, it was chosen to
sample the DOE sunken flotation products with a Jones sampler for analysis.
Table 8 – Analysis of uncertainty in 10 readings for the same sunken product of NFC trial not using Jones
Sampler.
Spectrometer variance analysis (Fe₂O₃ ppm)
Reading 1 158
36
Reading 2 277
Reading 3 307
Reading 4 219
Reading 5 267
Reading 6 151
Reading 7 284
Reading 8 215
Reading 9 187
Reading 10 339
Mean 240
Amplitude (max - min) 188
5.2. Analysis of results
The tests made with the three traditional collectors (Trad 1, Trad 2 and Trad 3) were evaluated
in terms of the grade in iron oxide in the sunken product. The first tests were made with Trad 1 collector
with a concentration of 268g/t, the pH of the pulp was changed for each different frothing (Figure 22).
A slight increase in iron oxide grade in the sunken product was observed when the pH rises until
9. With a very alkaline pulp of 12 pH, the iron oxide minerals concentrate in the sunken product. Both
frothers had similar results with Teepol HB 7 performing slightly better.
37
Figure 22 - pH and frother influence (2,2 g/t) on the iron oxide grade using Trad 1 collector with a concentration of
268 g/t.
The results achieved with Trad 2 and Aero 845N p revealed that the iron oxide grade in the
sunken product increases with the pH (Figure 23) (with one exception at pH 8 which should have been
confirmed). One trial was made with double concentration of the collector (384 g/t) which lead to good
results (123 ppm of iron oxide) but the froth was very sticky and had some agglomeration of heavy
minerals which sunk due to their weight. A trial was made to simulate previously used “case study
conditions” with Teepol frother at natural pH conditions with good result in terms of iron oxide grade in
the sunken product (plotted in Figure 24).
Figure 23 -pH and frother influence in iron oxide grade in the sunken product using Trad 2 collector with a
concentration of 192 g/t.
The first batch laboratory trials performed with Trad 3 collectors were made using Teepol frother
and a pH in the pulp of 7, changing only the collector concentration as seen in Figure 24. The lower
concentration resulted in higher iron oxide grade in the sunken product with the best result achieved
0
100
200
300
400
500
600
700
800
6 7 8 9 10 11 12 13
Fe2
O3
(pp
m)
pH
Teepol HB 7
Aero845N p
100
110
120
130
140
150
160
170
6 6,5 7 7,5 8 8,5 9 9,5 10 10,5
Fe2
O3
(pp
m)
pH
Teepol HB 7
A845N p
38
using a concentration of 200 g/t. An inconsistency can be found with collector concentration of 268 g/t,
where the iron oxide grade is more than 20 ppm higher than the test made with 300 g/t. These trials
results were the base for the following tests made with this collector.
Figure 24 - Trad 3 concentration influence in iron oxide grade in the sunken product with Teepol frother and pH of 7.
Plotting the achieved results related with the pH of the pulp using two different frothing agents
(Figure 25) it is observed that the iron oxide grade in the sunken product is lower for 7 pH. More alkaline
pulps lead to higher iron oxide grades in the sunken product. This behaviour is the same with both
frothers.
Figure 25 - pH and frother influence in iron oxide grade in the sunken product using Trad 3 collector with a
concentration of 200 g/t.
The analysis of traditional collectors trials (all results can be seen in Annex IV), revealed that
their concentration had a significant effect in the iron oxide performance of the froth flotation. It was
100
110
120
130
140
150
160
170
180
0 50 100 150 200 250 300 350 400 450
Fe2
O3
pp
m
Trad 3 Concentration (g/t)
100
110
120
130
140
150
160
170
180
190
6 7 8 9 10
Fe2
O3
(p
pm
)
pH
Teepol HB 7
A845N p
39
shown that with similar concentrations of each tested collector the average iron oxide grade in the
sunken product increases with the pH of the pulp (Figure 26).
Therefore, neutral pH conditions will have more emphasis henceforth. There was no clear
evidence that one frother provided better results over the other with either traditional collector. The Trad
1 and Trad 2 collectors performances were similar at neutral pH conditions, while Trad 3 performed
worse for the pH range tested.
Figure 26 – Iron oxide grade in the sunken product of the tests performed with the three different traditional
collectors in different conditions of pulp pH.
Figure 27 shows a plot of the trials where the iron oxide grade was significantly decreased with
pH 6,5 and 7 regardless of frother concentration. Both the Trad3 and the Trad2 seems to lead to better
results around 200 g/t with pH 7, while the Trad1 had better results than the others when a slightly larger
concentration (268 g/t) was used with preference to pulp of pH 7 as well.
The best result achieved with Trad1 was 115 ppm of iron oxide with a collector concentration
of 268 g/t, a pH of 7 and using Teepol with a concentration of 2,2 g/t. Trad2 best result was 119 ppm of
iron oxide with a collector concentration of 192 g/t, pH 7 and using Teepol with a concentration of 3,7
g/t. Trad 3 best result was 120 ppm using a collector concentration of 200g/t and a pH of 7 using A 845N
p with a concentration of 5,25 g/t.
100
120
140
160
180
200
6,5 7 8 9 10
Fe2
O3
(pp
m)
pH
Trad 1
Trad 2
Trad 3
40
Figure 27 – Iron oxide grade in the sunken sand best results for different traditional collectors in different
concentrations with a pulp of pH 7.
5.3. NFC preliminary tests
The preliminary tests for pH (7 and 9) and frother (type and concentration) with 61 g/t of NFC
collector are plotted in Figure 28. The plotted trial runs revealed a better performance for 1,5 g/t
concentration of frother and a pulp with pH 7, with Teepol having 6 ppm less iron oxide grade in the
sunken product than with the same concentration of MIBC. A pulp conditioned with a pH of 7 led to
better performances for trials with more than 1,5 g/t of either frother.
Figure 28 – Iron oxide grade in the sunken product using 61 g/t of NFC collector concentration for different pH and
frother concentration.
Knowing that a pH 7 pulp and frother concentration of 1,5 g/t are more adequate to provide low
iron oxide in the sunken product, different concentration of NFC collector was tested to determine the
working range of the NFC collector. Looking at Figure 29 it is observed that iron oxide grade in the
114
115
116
117
118
119
120
121
150 170 190 210 230 250 270 290
Fe2
O3
pp
m
Collector Concentration (g/t)
Trad 1
Trad2
Trad3
100
150
200
250
300
350
0 0,5 1 1,5 2 2,5 3 3,5 4
Fe₂O
₃p
pm
Frother g/t
pH 7,Teepol
pH 9,Teepol
pH 7,MIBC
41
sunken product was under 200 ppm for concentration of NFC between 50 g/t and 200 g/t. It is to note
that with very high concentration of NFC (1000g/t) the silica particles floated so easily that the iron oxide
ended up concentrated in the sunken product, the opposite of the desired in this study.
Figure 29 – Iron oxide grade in the sunken product with different NFC collector concentration with pH 7 and 1,5
g/t MIBC frother.
5.4. NFC Design of Experiments (DOE)
The full design of experiments (DOE) with their manipulated factors and respective responses
is listed in Figure 30. Design Expert 9.1 software was used to analyse correlations, ANOVA, surface
responses between variables and finally to optimize the manipulated factors to achieve a minimum
grade of iron oxide on the glass sand and a maximum recovery of the same mineral in the floated
product.
100
150
200
250
300
350
400
450
500
0 100 200 300 400 500 600 700 800 900 1000
Fe₂O
₃ p
pm
Collector Concentration (g/t)
42
Figure 30 - Manipulated factors and related responses for each trial of the design of experiments.
5.4.1. Manipulated factors influence in the flotation of iron oxide
A first approach to the design of experiments results was made plotting the iron oxide grade in
the sunken product, the recovery of iron oxide and the weight pull percentage per flotation trial. The
following graphics are made for a constant conditioning time, relating the NFC concentration and the pH
of the pulp for each of the response variables.
In Figure 31, the response in analyse is the iron oxide grade in the sunken product, which is
pretended to be the minimal possible. The lowest value (144 ppm) was achieved with a 40 g/t NFC
concentration, a pH of 10 and a conditioning time of 3 minutes, while the highest value (170 ppm) was
achieved with a 120 g/t NFC concentration, a pH of 7 and a conditioning time of 7 minutes.
For lower conditioning times the trend is to increase the iron oxide grade with an increase of
NFC concentration, regardless of pH. The opposite happens with higher conditioning time, where the
trend is to decrease the iron oxide grade with higher NFC concentration. In general, with a conditioning
time of 5 minutes more results lower than 150 ppm were achieved.
130
140
150
160
170
40 120 200iro
n o
xid
e gr
ade
pp
m
NFC concentration (g/t)
pH 7
pH 8,5
pH 10 130
140
150
160
170
40 120 200iro
n o
xid
e gr
ade
pp
m
NFC concentration (g/t)
pH 7
pH 8,5
pH 10
43
a) b)
c)
In Figure 32, the response in analyse is the iron oxide recovery (%) in the floated product, which
is pretended to be the highest possible. The highest value (38,8 %) was achieved with a 200 g/t NFC
concentration, a pH of 8,5 and a conditioning time of 5 minutes, while the lowest value (27,4 %) was
achieved with a 120 g/t NFC concentration, a pH of 7 and a conditioning time of 10 minutes.
The recovery of iron oxide is greater with low NFC concentration with a pH of 10. With this
alkalinity there is a trend to decrease recovery with higher NFC concentrations except for a higher
conditioning time, where the opposite occurs. In general, with a conditioning time of 5 minutes more
results with recoveries higher than 35% were achieved.
a) b)
130
140
150
160
170
40 120 200iro
n o
xid
e gr
ade
pp
m
NFC concentration (g/t)
pH 7
pH 8,5
pH 10
Figure 31 - Iron oxide grade in the sunken product relating the NFC concentration and the pH for a conditioning time of a) 3 minutes, b) 5 minutes and c) 7 minutes.
25
30
35
40
40 120 200
iro
n o
xid
e r
eco
very
(%)
NFC concentration (g/t)
pH 7
pH 8,5
pH 1025
30
35
40
40 120 200
iro
n o
xid
e r
eco
very
(%)
NFC concentration (g/t)
pH 7
pH 8,5
pH 10
25
30
35
40
40 120 200
iro
n o
xid
e re
cove
ry (%
)
NFC concentration (g/t)
pH 7
pH 8,5
pH 10
44
c)
In Figure 33, the response in analyse is the weight pull (%) of the floated product, which cannot
be analysed per se, it can be seen, if related with the other two response variables, as a measure of the
selectivity of the floatation process. The highest value (4,3 %) was achieved with a 200 g/t NFC
concentration, a pH of 8,5 and a conditioning time of 5 minutes, while the lowest value (0,002 %) was
achieved with a 40 g/t NFC concentration, a pH of 7 and a conditioning time of 3 minutes.
Low concentrations of NFC result in low weight pull. The weight pull (%) of floated product is
greater with higher NFC concentration for pH of 7 and 8,5, except for pH 10. In general, with a
conditioning time of 7 minutes less weight pull (%) was achieved.
a) b)
c)
5.4.1.1. Pearson correlation coefficient
The Pearson correlation coefficients is used to quantify correlations between two independent
variables (Equation 4). A perfect positive correlation between two variables has a Pearson coefficient of
Figure 32 - Iron oxide recovery (%) in the floated product relating the NFC concentration and the pH for a conditioning time of a) 3 minutes, b) 5 minutes and c) 7 minutes.
0
1
2
3
4
5
40 120 200
Wei
ght
pu
ll (%
)
NFC concentration (g/t)
pH 7
pH 8,5
pH 10
0
1
2
3
4
5
40 120 200
Wei
ght
pu
ll (%
)
NFC concentration (g/t)
pH 7
pH 8,5
pH 100
1
2
3
4
5
40 120 200
Wei
ght
pu
ll (%
)
NFC concentration (g/t)
pH 7
pH 8,5
pH 10
Figure 33 - Weight pull (%) of the floated product relating the NFC concentration and the pH for a conditioning time of a) 3 minutes, b) 5 minutes and c) 7 minutes.
45
1, a perfect negative correlation as a coefficient of -1, a Pearson coefficient of 0 means that the two
variables are not linearly dependent.
Equation 6
The Pearson correlation coefficients between the variable factors and responses of the DOE
are plotted in Table 9 with an intensity of colour indicating the higher correlation, red for positive and
blue for negative.
A negative correlation of -0,967 was observed between iron oxide grade in the sunken product
and its recovery in the floated product (Figure 34). The strong negative correlation reinforces the idea
that the NFC collector has selectivity within the range of the DOE since the iron oxide (Fe₂O₃) grade in
the sunken sand is lowering and its recovery is being made in the froth.
The weight pull is positively related with the collector concentration of 0,67, which reveals that
there is a possibility that by only increasing the collector concentration the mass pull will be greater,
which may affect the selectivity of the collector.
Table 9 - Pearson correlation coefficients matrix between independent variables and responses.
Pearson
Coefficients Run
A:Collector
Conc. B:pH
C:Cond.
Time % Weight Fe2O3 Recovery
Run 1
A:Collector Conc. 0.027 1
B:pH -0.384 0,000 1
C:Cond. Time 0.361 0,000 0,000 1
% Weight 0,068 0,676 -0,264 -0,110 1
Fe2O3 -0,222 0,215 -0,334 0,083 0,122 1
Recovery 0,229 -0,036 0,293 -0,118 0,123 -0,967 1
46
Figure 34 - Negative correlation between iron oxide grade in the sunken product and the iron oxide recovery in
the sunken product (-0.967 Pearson coefficient).
5.4.1.2. ANOVA - Iron oxide grade
The ANOVA analysis for the iron oxide grade allowed the creation of a two factorial interaction
(2FI) response surface with a significant model (p < 0,05) and statistical significance in pH and collector
concentration (Cc) x conditioning time (Ct) factor (Table 10). Another ANOVA was made with only the
significant factors (Table 11).
Table 10 - ANOVA for iron oxide grade (ppm) Response Surface 2FI model.
Source Sum of
Squares
df Mean
Square
F
Value
p-value
Prob > F
Model 643,53 6 107,25 2,91 0,027 significant
Cc - Collector Conc. 72,00 1 72,00 1,96 0,174
pH 174,22 1 174,22 4,73 0,039
Ct - Cond. Time 10,89 1 10,89 0,3 0,591
Cc * pH 8,33 1 8,33 0,23 0,638
Cc * Ct 341,33 1 341,33 9,27 0,005
Design-Expert® Software
Correlation: -0.927
140 145 150 155 160 165 170
26
28
30
32
34
36
38
40
Fe2O3 (ppm)
Re
co
ve
ry (
%)
47
pH * Ct 36,75 1 36,75 1,00 0,327
Residual 920,44 25 36,82
Lack of Fit 809,61 20 40,48 1,826 0,262 not significant
Pure Error 110,83 5 22,17
Cor Total 1563,97 31
Table 11 - ANOVA for iron oxide grade (ppm) Reduced Response Surface 2FI model.
Source Sum of
Squares df
Mean
Square
F
Value
p-value
Prob > F
Model 598,44 4 149,61 4,18 0,009 significant
Cc 72,00 1 72,00 2,01 0,167
pH 174,22 1 174,22 4,87 0,036
Ct 10,89 1 10,89 0,30 0,586
Cc * Ct 341,33 1 341,33 9,55 0,005
Residual 965,52 27 35,76
Lack of Fit 854,69 22 38,85 1,75 0,278 not significant
Pure Error 110,83 5 22,17
Cor Total 1563,97 31
It is observed that the reduced model had a non-significant lack of fit and there is only a 0,9%
probability that a 4,18 F value is due to noise. The final equation in terms of Actual Factors is:
𝐹𝑒2𝑂3(𝑝𝑝𝑚) = 148,15 + 0,19 × 𝐶𝑐 − 2.07 × 𝑝𝐻 + 0,49 × 𝐶𝑡 − 0.03 × 𝐶𝑐 × 𝐶𝑡 Equation 7
48
To validate the response surface model (Figure 35) is shown the Normal Plot of Residuals
(Figure 36) to check for normality of residuals, Predicted values vs Residuals (Figure 37) to check for
constant error and Residuals vs Run Number (Figure 38) to check for error independence. Everything
indicates that the 2FI reduced model is valid.
Design-Expert® SoftwareFe2O3
Color points by value ofFe2O3 :
170
144
Externally Studentized Residuals
No
rm
al
% P
ro
ba
bil
ity
Normal Plot of Residuals
-2.00 -1.00 0.00 1.00 2.00 3.00
1
5
10
20
30
50
70
80
90
95
99
Figure 36 - Normal Plot of Residuals for iron oxide grade (ppm) Reduced Response Surface.
Figure 35 - Surface of reduced 2FI for iron oxide grade according to collector conc. and conditioning time for 8,5 pH.
Design-Expert® Software
Factor Coding: Actual
Grade
Design points above predicted value
Design points below predicted value
170
144
X1 = A: cc
X2 = C: ct
Actual Factor
B: pH = 8.5
3
4
5
6
7
40
80
120
160
200
140
145
150
155
160
165
170
Gra
de
A: cc
C: ct
49
Figure 38 - Residuals versus Random Run Number.
Design-Expert® SoftwareFe2O3
Color points by value ofFe2O3 :
170
144
Run Number
Ex
tern
all
y S
tud
en
tiz
ed
Re
sid
ua
ls
Residuals vs. Run
-4.00
-2.00
0.00
2.00
4.00
1 6 11 16 21 26 31
Design-Expert® SoftwareFe2O3
Color points by value ofFe2O3 :
170
144
Predicted
Ex
tern
all
y S
tud
en
tiz
ed
Re
sid
ua
lsResiduals vs. Predicted
-4.00
-2.00
0.00
2.00
4.00
145 150 155 160 165
Figure 37- Residuals versus Predict iron oxide grade values.
50
The iron oxide grade model predicts an interaction between concentration of collector (𝐶𝑐) and
conditioning time (𝐶𝑡), as shown in Figure 39. With an increase in both factors the grade of the iron oxide
in the sunken material will decrease, which is a possible solution.
5.4.1.3. ANOVA - Iron oxide Recovery
The ANOVA analysis for the iron oxide recovery allowed the creation of a two factorial
interaction (2FI) response surface with a not significant model (p > 0,05) (Table 12). Another ANOVA
was made with only the manipulated factors and the significant interaction Cc*Ct with a significant model
(p < 0,05) (Table 13).
Source Sum of Squares df Mean
Square
F
Value
p-value
PROB> F
Model 95,00 6 15,83 2,21 0,076 not significant
Cc - Collector Conc. 0,35 1 0,35 0,05 0,828
pH 23,58 1 23,58 3,29 0,082
Ct - Cond. Time 3,83 1 3,83 0,53 0,472
Cc * pH 5,07 1 5,07 0,71 0,408
Cc * Ct 53,34 1 53,34 7,44 0,012
pH * Ct 8,84 1 8,84 1,23 0,277
Design-Expert® Software
Factor Coding: Actual
Grade
X1 = A: cc
X2 = C: ct
Actual Factor
B: pH = 8.5
C- 3
C+ 7
A: cc
C: ct
40 80 120 160 200
Gra
de
140
145
150
155
160
165
170
Interaction
Figure 39 - Interaction between Collector concentration (Cc) and Conditioning Time (Ct)
51
Table 12 - ANOVA for iron oxide Recovery (%) Response Surface 2FI model.
Table 13 - ANOVA for Iron oxide Recovery (%) Reduced Response Surface 2FI model.
It is observed that the reduced model had a non-significant lack of fit and there is only a 0,04%
probability that a 2,84 F value is due to noise. The final equation in terms of Actual Factors is:
𝛤𝐹𝑒2𝑂3(%) = 36,6 − 0.068 × 𝐶𝑐 + 0.763 × 𝑝𝐻 − 1,812 × 𝐶𝑡 + 0.013 × 𝐶𝑐 × 𝐶𝑡 Equation 8
To validate the reduced response surface model (Figure 40) shows the Normal Plot of Residuals
(Figure 41), Predicted values vs Residuals (Figure 42) and Residuals vs Run Number (Figure 43).
Everything indicates that the 2FI reduced model is valid.
Residual 179,15 25 7,17
Lack of Fit 159,99 20 8,00 2,09 0,212 not significant
Pure Error 19,15 5 3,83
Cor Total 274,15 31
Source Sum of
Squares
df Mean
Square
F Value p-value
PROB> F
Model 81,09 4 20,27 2,84 0,044 significant
Cc - Collector Conc. 0,35 1 0,35 0,05 0,828
pH 23,58 1 23,58 3,30 0,081
Ct - Cond. Time 3,83 1 3,83 0,54 0,471
Cc * pH 53,34 1 53,34 7,46 0,011
Residual 193,06 27 7,15
Lack of Fit 173,90 22 7,90 2,06 0, 216 not significant
Pure Error 19,15 5 3,83
Cor Total 274,16 31
52
Figure 40 - Surface of reduced 2FI model for iron oxide Recovery (%) according to collector concentration and
conditioning time for 10 pH.
Figure 41 - Normal Plot of Residuals for iron oxide Recovery (%) Reduced Response Surface.
Design-Expert® Software
Rec
Color points by value of
Rec:
38.8
27.4
Externally Studentized Residuals
Norm
al %
Pro
babili
ty
Normal Plot of Residuals
-3.00 -2.00 -1.00 0.00 1.00 2.00 3.00
1
5
10
20
30
50
70
80
90
95
99
Design-Expert® Software
Factor Coding: Actual
Rec
Design points above predicted value
Design points below predicted value
38.8
27.4
X1 = A: cc
X2 = C: ct
Actual Factor
B: pH = 8.5
3
4
5
6
7
40
80
120
160
200
26
28
30
32
34
36
38
40
Re
c
A: cc
C: ct
53
Figure 42 - Residuals versus Predict iron oxide Recovery (%) values.
Figure 43 - Residuals versus Random Run Number.
The iron oxide recovery model predicts an interaction between concentration of collector (𝐶𝑐)
and conditioning time (𝐶𝑡), as shown in Figure 44. With an increase in both factors the recovery of the
iron oxide will increase. The interaction provided by the iron oxide grade model (Figure 39) is the inverse
Design-Expert® Software
Rec
Color points by value of
Rec:
38.8
27.4
Predicted
Exte
rnally
Stu
dentized R
esid
uals
Residuals vs. Predicted
-4.00
-2.00
0.00
2.00
4.00
30 32 34 36 38
3.53229
-3.53229
0
Design-Expert® Software
Rec
Color points by value of
Rec:
38.8
27.4
Run Number
Exte
rnally
Stu
dentized R
esid
uals
Residuals vs. Run
-4.00
-2.00
0.00
2.00
4.00
1 6 11 16 21 26 31
3.53229
-3.53229
0
54
of the interaction shown here, corroborating the inversely proportional correlation between grade and
recovery presented in the pearson correlation table.
Figure 44 - Interaction between Collector concentration (Cc) and Conditioning Time (Ct)
5.4.1.4. ANOVA - Weight Pull
The ANOVA analysis for the weight pull allowed the creation of a quadratic response surface with a
significant model (p < 0,05) and a significant Lack of fit (Table 14). Another ANOVA was made with only
the manipulated factors and the significant factor interactions with a significant model (p < 0,05) and a
significant Lack of fit (Table 15).
Design-Expert® Software
Factor Coding: Actual
Rec
X1 = A: cc
X2 = C: ct
Actual Factor
B: pH = 8.5
C- 3
C+ 7
A: cc
C: ct
40 80 120 160 200
Rec
26
28
30
32
34
36
38
40
Interaction
Table 14 - ANOVA for Weight Pull(%) Response Surface Quadratic model.
Source Sum of Squares df Mean
Square
F
Value
p-value
PROB> F
Model 30,41 9 3,38 10,30 <0,0001 significant
Cc - Collector Conc. 17,21 1 17,21 52,46 <0,0001
pH 2,63 1 2,63 8,00 0,0098
Ct - Cond. Time 0,45 1 0,45 1,38 0,2530
CC * pH 3,69 1 3,69 10,03 0,0045
Cc * Ct 0,6 1 0,6 1,84 0,1885
pH * Ct 0,3 1 0,3 0,91 0,3512
Cc2 4,54 1 4,54 13,83 0,0012
pH2 0,87 1 0,87 2,64 0,1186
55
Table 15 - ANOVA for Weight Pull (%) Reduced Response Surface Quadratic model.
It is observed that the reduced model had a significant p-value but also a significant lack of fit.
A Box-Cox transformation was applied with a natural log constant of 0,06. The ANOVA is shown in Table
16.
Ct2 1,67 1 1,67 5,08 0,0345
Residual 7,22 22 0,33
Lack of Fit 7,01 17 0,41 9,85 0.0095 significant
Pure Error 0,21 5 0,042
Cor Total 37,63 31
Source Sum of
Squares df
Mean
Square
F
Value
p-value
Prob > F
Model 28,65 6 4,77 13,28 <0,0001 significant
Cc - Collector Conc. 17,21 1 17,21 47,89 <0,0001
pH 2,63 1 2,63 7,31 0,0122
Ct - Cond. Time 0,45 1 0,45 1,26 0,2727
Cc*pH 3,29 1 3,29 9,15 0,0057
Cc2 3,95 1 3,95 11,00 0,0028
Ct2 2,24 1 2,24 6,24 0,0194
Residual 8,99 25 0,36
Lack of Fit 8,78 20 0,44 10,49 0,0081 significant
Pure Error 0,21 5 0,042
Cor Total 37,63 31
56
Table 16 - ANOVA for Weight Pull (%) Reduced Response Surface Quadratic model after Box-Cox transformation with k = 0,06.
The reduced response quadratic model for weight pull after transformation is significant (p <
0,05) and had a non-significant lack of fit, there is only a 0,01% probability that a 36,30 F value is due
to noise. The final equation in terms of Actual Factors is:
log(𝑊(%) + 0,06)
= −24,899 + 0.058 × 𝐶𝑐 + 4,276 × 𝑝𝐻 + 1,201 × 𝐶𝑡 − 4,763𝐸(−003) × 𝐶𝑐 × 𝑝𝐻
− 0.234 × 𝑝𝐻2 − 0.126 × 𝐶𝑡2
Equation 9
Source Sum of Squares df Mean
Square
F
Value
p-value
PROB> F
Model 48,63 6 8,10 36,30 <0,0001 significant
Cc - Collector Conc. 36,21 1 36,21 162,16 <0,0001
pH 3,11 1 3,11 13,92 0,0010
Ct - Cond. Time 0,22 1 0,22 0,99 0,3282
CC * pH 3,92 1 3,92 17,55 0,0003
pH2 2,06 1 2,06 9,24 0,0055
Ct2 1,88 1 1,88 8,41 0,0077
Residual 5,58 25 0,22
Lack of Fit 4,83 20 0,24 1,61 0.3157 not-significant
Pure Error 0,75 5 0,15
Cor Total 54,21 31
57
To validate the reduced response surface model is shown the Normal Plot of Residuals (Figure
46), Predicted values vs Residuals (Figure 47) and Residuals vs Run Number (Figure 48). Everything
indicates that the transformed quadratic reduced model is valid.
Design-Expert® Software
Factor Coding: Actual
Original Scale
W
Design points above predicted value
Design points below predicted value
4.345
0.002
X1 = A: cc
X2 = B: pH
Actual Factor
C: ct = 5
7
7.6
8.2
8.8
9.4
1040
80
120
160
200
0
1
2
3
4
5
W
A: cc
B: pH
Figure 45 - Surface of reduced quadratic model for Weight pull (%) according to collector concentration and pH for conditioning time of 5 minutes.
58
Figure 46 - Normal Plot of Residuals for Weight pull (%) Reduced Response Surface.
Design-Expert® Software
Ln(W + 0.06)
Color points by value of
Ln(W + 0.06):
1.483
-2.781
Externally Studentized Residuals
Norm
al %
Pro
babili
ty
Normal Plot of Residuals
-3.00 -2.00 -1.00 0.00 1.00 2.00 3.00
1
5
10
20
30
50
70
80
90
95
99
Figure 47 - Residuals versus Predict Weight recovery (%) values.
Design-Expert® Software
Ln(W + 0.06)
Color points by value of
Ln(W + 0.06):
1.483
-2.781
Predicted
Exte
rnally
Stu
dentized R
esid
uals
Residuals vs. Predicted
-4.00
-2.00
0.00
2.00
4.00
-4 -3 -2 -1 0 1 2
3.56648
-3.56648
0
59
Figure 48 - Residuals versus Random Run Number.
Design-Expert® Software
Ln(W + 0.06)
Color points by value of
Ln(W + 0.06):
1.483
-2.781
Run Number
Exte
rnally
Stu
dentized R
esid
uals
Residuals vs. Run
-4.00
-2.00
0.00
2.00
4.00
1 6 11 16 21 26 31
3.56648
-3.56648
0
60
5.5. Optimization of Laboratory Froth Flotation
The analysis of the process is concluded with the testing of the best combinations of the studied
factor levels in order to minimize the iron oxide grade in glass sand as well as maximizing the recovery
of iron oxide in the floated product. The software Design-Expert 9 has a tool for numeric optimization, it
search the design of experiments for setups that fulfil the optimization requirements (Table 17) recurring
to the iron oxide grade and iron oxide recovery models created through the ANOVA.
Table 17 - Variable constraints for numerical optimization.
Name Goal Lower Limit Upper Limit
Cc - Collector Conc. is in range 40 200
pH is in range 7 10
Ct - Cond. Time is in range 3 7
Fe2O3 minimize 144 150
Recovery maximize 27,36 38,83
The two better solutions shown in Table 18 vary on the collector concentration and the
conditioning time of the pulp, with various solutions similar to solution 1 with 94,1% desirabilty. Solution
2 has only 78,3% desirability. The desirability response surface is shown in Figure 49 for a minimum
iron oxide grade in sunken sand and maximum recovery of iron oxide in the floated product.
Table 18 - Solutions for process optimization.
Figure 50 shows the samples of sunken and floated products of a laboratorial froth flotation trial
with solution 1 conditions on the right and the sunken and floated products of a solution 2 conditions
trial on the left. They have clear differences in the appearance of the floated product, with the iron oxides
and other heavy minerals being black. The estimation of results provided by the better solution is
summarized in Table 17 with the respective 95% low and high confidence intervals. Using a NFC
concentration of 40 g/t, a pH of 10 and a conditioning time of 3 minutes the model predicts a mean result
of 147 ppm of Fe2O3 in the sunken sand product.
Number Collector Conc. pH Cond. Time Fe2O3 Recovery Desirability
1 40 10 3 147 37,181 0,941
2 200 10 7 149 36,444 0,783
61
Figure 49 - Desirability response surface for maximum iron oxide recovery in the floated product and minimum
grade in the sunken product for a conditioning time of 3 minutes.
Figure 50 - Samples of floated and sunken optimal trials. Solution 1 (right) and Solution 2 (left).
Table 19 - Estimated answer for both optimal solutions and estimated confidence intervals.
Response Mean Median Std. Dev SE Mean 95% CI low 95% CI high
Fe2O3 147 147 6,01 2,48 141,95 152,10
Recovery 37,18 37,18 2,20 1,01 35,12 39,24
62
6. Conclusion
The work developed from March to May, 2015 in the company’s facility had the objective of
testing three different “non-green” collectors, traditionally used for removing iron oxide and other heavy
minerals from glass sand through reverse froth flotation. This created a comparison point for a novel
“green collector” which was tested for the first time through batch laboratory froth flotation with the
same sand feed.
The froth flotation laboratory results of both traditional collectors and “green” NFC collector
provided new insight on the reverse flotation of “glass sand” to remove iron oxide and other heavy
minerals which would otherwise contaminate the final product, decreasing its economic value. The
stipulated iron oxide grade in a purified glass sand should be lower than 130 ppm to be considered a
saleable product.
The traditional collectors codenamed Trad1, Trad2 and Trad3 were tested to different extents
using an OVAT approach. These were tested with two different frothers, Teepol and A845N and a set
of pH in the pulp ranging from 6,5 to 12.
Trad1 collector was previously studied for this “case study” glass sand and in this thesis
laboratorial work, consistently good results were achieved of under 130 ppm of iron oxide grade in the
sunken product for natural pH conditions of 6,5-7 with Teepol frother and 268 g/t of Trad1 concentration
in the pulp.
Trad2 collector was known in the “case study” flotation circuit history for achieving good results
with Teepol and the best result with this collector was achieved with the above mentioned frother (119
ppm) at a natural pH of 6,8 and Trad2 concentration of 192 g/t. A Trad2 trial performed with A845N
frother in related conditions (pH 6,5) achieved a similar result of 121 ppm of iron oxide grade in the
sunken product.
Trad3 performance was not tested with this glass sand before so a more wide-ranging
approach was used, being the best result achieved with 200 g/t of Trad3 concentration, using A845N
frother at a pH of 7 (120 ppm). A couple of results were achieved with an iron oxide grade under 130
ppm using Teepol frother at a pH of 7, with different Trad3 concentrations.
The company has the aim for continuous improvement of processes and the environmental
awareness justified completely the laboratory work by considering alternatives in the form of other
traditional collectors than the currently used and regarding the environmental concern, a novel
Nanofibrillated n-butylamine Celluloses (NFC) as an alternative to traditional collectors that can
consume great part of the oxygen in the industrial effluent waters.
The “green” NFC collector was tested in two phases, since it was only tested with this “case
study” glass sand in microflotation trials, preliminary batch laboratory trials were made for this thesis in
an OVAT approach. These preliminary tests tried to identify the working range of Teepol, MIBC and
63
Dowfroth frothers, the NFC concentration working ranges and the pH of the pulp suitable for the flotation
of iron oxide particles. From the preliminary test of trials a conclusion was made that for the different
frothers a 1,5 g/t concentration provided good results and that the NFC concentration and the pH of the
pulp as well as the conditioning time should have more emphasis henceforth.
The second phase of the study of the NFC collector to remove iron oxide particles from the glass
sand through reverse froth flotation should be to perform a 3-level full factorial design of experiments
(DoE). The defined variable factors were the NFC concentration, ranging from 40 g/t to 200 g/t, the pH
of the pulp, ranging from (7 to 10) and the conditioning time, ranging from 3 to 7 minutes. The defined
responses to be analysed were the iron oxide grade in the sunken product (ppm), the recovery of iron
oxide particles in the floated product (%) and the weight pull of the floatation trial (%).
The execution of the DoE and its further analyse and modelling allowed the conclusion that the
NFC, with the range of variable factors presented in this work, should be performed at a pH of 10 with
a NFC of 40 g/t with 3 minutes of conditioning of the pulp (in this conditions 147 ppm of iron oxide grade
was achieved).
The NFC collector provided worst results than the traditional ones with difficulties reaching iron
oxide grades lower than 150 ppm in the glass sand. Once the specifications for the finest quality glass
sand processed in the facilities has an iron oxide grade lower than 130 ppm the NFC collector is not
effective.
64
7. Future work
This study of the reverse froth flotation behaviour of a fine sand to produce a high standard
product for the glassmaking industry supplied some insight of the capabilities of carboxylate collectors
(traditional collectors) removing the main contaminant, iron oxide minerals. Nonetheless, a Design of
Experiments (DOE) for both the Trad 2 and Trad 3 collector would be useful. Trad 2 had similar results
under certain conditions, if an optimal point close to the Trad 1 is achieved with batch laboratory flotation,
further economic studies might prove beneficial if applied to the processing plant.
Reproducibility of the traditional collector trials and also the NFC preliminary trials should be
performed in future works to avoid any misguided conclusion, the sampling of the glass sand should be
performed more cautiously, using quartering techniques and using a jones sampler and an analytical
method of analyse of the iron oxide grade in the floated product should be performed in all trials, since
the X-Ray Fluorescence method used had a curve for a specific range of values and misguided
conclusions might have been taken.
The NFC collector design of experiments was performed with a range of conditioning time
between 3 and 7 minutes when the preliminary tests were all made with 5 minutes of conditioning.
Further studies of conditioning time with this collector is suggested. Other variables may be put to test,
like the air flow rate and the impeller speed since the collector sometimes produced a non-beneficial
froth for flotation.
Observations during the laboratory tests gave the perception that the NFC collector can produce
better results for a glass sand of lower particle size distribution since the size of the iron oxide minerals
recovered with the froth was much smaller using the NFC than any other traditional collector. Some
batch milling of the fine sand that fed these trials with further batch laboratory flotation might prove that
point.
65
References Alonso, L. M. (2014). Optimization of Flotation for the Reduction of Heavy Minerals and Iron Content
on Silica Sand. Lisboa: Instituto Superior Técnico.
Arr Maz. (2013). Safety Data Sheet - Custofloat E 229. Mulberry, Florida, U.S.A.
British Geological Survey. (2009). Silica Sand. Mineral Planning Factsheet, 1-10.
Bulatovic, S. M. (2007). Elsevier Science & Technology Books.
Bulatovic, S. M. (2007). Handbook of Flotation Reagents - Chemistry, Theory and Practice: Flotation
of Sulfide Ores. Elsevier Science & Technology Books.
Chammas, E., Panias, D., Taxiarchou, M., Anastassakis, G., & Paspaliaris, I. (2001). Removal of Iron
and other Major impurities from silica sand for the production of high added value materials. IX
Balkan Mineral Processing Congress, (pp. 289-295). Istanbul.
CYTEC. (2002). Mining Chemicals Handbook.
Durão, F., Cortez, L., & Carvalho, M. T. (2002). Flutuação por Espumas. Lisboa: CVRM - Centro de
Geosistemas.
Forchem Oy. (2010). For15 Product Datasheet. Rauma, Finland.
Fuerstenau, D. (1982). Mineral-Water Interfaphases and the Electrical Double Layer Principles of
Flotation. IMME, South African Institute of Mining & Metallurgy.
Johann Haltermann Ltd. (n.d.). Technical Data & Safety Bulletin - Methyl Isobutyl Carbinol (MIBC).
Karimi, M. A., Goudi, A., & Zali, S. (2008). Sodium dodecyl sulfate-coated alumina and C18 cartridge
for the separation od preconcentration of cationic surfactants prior to their quatitation by
spectrophotometric method. Journal of the Brazilian Chemical Society, vol.19, nr. 8.
Kelly, E., & Spottiswood, D. (1982). Introduction to mineral processing. New York.
Kogel, J. E. (2006). Industrial Minerals & Rocks: Commodities, Markets, and Uses. SME.
Laitinen, O., Kemppainen, K., Ammala, A., Sirvio, J. A., & Liimatainen, H. (2014, 12 10). Use of
Chemically Modified Nanocelluloses in Flotation of Hematite. Industrial & Engineering
Chemistry Research, pp. 20092-20098.
Langmuir, I. (1920). The Mechanism of the Surface Phenomena of Flotation. Transactions of the
Faraday Society, vol. 20, pp. 138-144.
Lines, M., & Echt, A. (2004). Silica sand supply demand in the Asia-Pacific glass market. Retrieved
from http://www.kdsolution.com/pdf_upload/technical_20061003124838.pdf
Lottermoser, B. G. (2007). Mine Wastes - Characterization, Treatment, Environmental Impacts.
Cairns, Queensland: Springer.
McLaws, I. (1971). Uses and specifications of silica sand. Edmonton: Research Council of Alberta.
Pohl, W. L. (2011). Economic Geology: Principles and Practice. John Wiley & Sons.
Sundararajan, M., Ramaswmy, S., & Raghavan, P. (2009). Evaluation for the Beneficiability of White
Silica Sands from the Overburden of Lignite Mine situated in Rajpardi district of Gujarat, India.
Journal of Minerals & Materials Characterization & Engineering, Vol.8, No 9, 701-713.
Teepol Products. (2013). Safety Data Sheets - Teepol HB 7. Orpington Kent.
U.S. Environmental Protection Agency. (1995, November). Chapter 11.9.1 - Sand and Gravel
Processing. Emission Factors & AP 42, I(Fifth Edition), 11.19.1-8.
66
U.S. Geological Survey. (2012). SILICA by Thomas P. Dolley. U.S. GEOLOGICAL SURVEY
MINERALS YEARBOOK—2010, 66.1-66.16.
Wills, B. A., & Napier-Munn, T. (2006). Mineral Processing Technology - An Introduction to the
Practical Aspects of Ore Treatment and Mineral. Elsevier Science & Technology Books.
I
ANNEXES
Annex I
Particle size distribution for finer glass sand, Industrial records and analysis made to the
laboratory froth flotation feed.
Fine fraction (Industrial) Fine fraction (Laboratorial)
Particle Size (µm)
Retained (%)
Cumulative retained (%)
Cumulative Passing (%)
Retained (%) Cumulative retained (%)
Cumulative Passing
(%)
>1000 0 0 100 0 0 100
]1000-710] 0,1 0,1 99,9 0 0 100
]710-500] 1,3 1,4 98,6 1,5 1,5 98,5
]500-355] 38,1 39,5 60,5 41,4 42,9 57,1
]355-250] 44,9 84,4 15,6 44,3 87,2 12,8
]250-180] 13,3 97,7 2,3 11,8 99 1
]180-125] 2,2 99,9 0,1 0,8 99,8 0,2
]125-90] 0,1 100 0 0,2 100 0
]90-63] 0 100 0 0 100 0
<63 0 100 0 0 100 0
Annex II
Mineralogical composition of industrial (average) and laboratory froth flotation feed using X-Ray
Spectrometry.
Mineralogical Composition
Fine Fraction (Industrial) Fine Fraction (Laboratorial)
Wt % ppm Wt % ppm
SiO₂ 99,43 994300 99,637 996370
Fe₂O₃ 0,055 550 0,023 230
Al₂O₃ 0,324 3240 0,213 2130
TiO₂ 0,086 860 0,020 200
K₂O 0,012 120 0,011 110
CaO 0,004 40 0,004 40
MgO 0,003 30 0,003 30
Na₂O 0,001 10 0 0
Loss on ignition 0,085 0,089
Total 100 100
II
Annex III
Record of all laboratory froth flotation trials and the concentration, dilution and added quantity of
frother and collector, pH value and iron oxide grade (wt. %) in glass sand. Due to the extension
of the table it is presented on the following page.
III
Collector Frother
NAME Type Concentration
(g/t) Quantity
(ml) pH Type
Concentration (g/t)
Quantity (ml)
Fe2O3 (wt. %)
E1 Trad1 (1:10) 268 4,1 7 Teepol (1:400)
2,2 1,2 0,0116
E2 Trad1 (1:10) 268 4,1 7 Teepol (1:400)
2,2 1,2 0,0115
E3 Trad1 (1:10) 268 4,1 12 Teepol (1:400)
2,2 1,2 0,0747
E4 Trad1 (1:10) 268 4,1 8 Teepol (1:400)
2,2 1,2 0,0135
E5 Trad1 (1:10) 268 4,1 9 Teepol (1:400)
2,2 1,2 0,0130
A1 Trad1 (1:10) 268 4,1 6,5 *A845N p
(1:25) 24 0,8 0,0107
A2 Trad1 (1:10) 268 4,1 6,5 *A845N p (1:250)
2,2 0,8 0,0122
A3 Trad1 (1:10) 268 4,1 8 *A845N p (1:250)
2,2 0,8 0,0135
C1 Celmin (1,5%) 61 5,7 7 Teepol (1:400)
0,5 0,3 0,0278
C2 Celmin (1,5%) 61 5,7 7 Teepol (1:400)
1,5 0,8 0,0120
C3 Celmin (1,5%) 61 5,7 7 Teepol (1:400)
2,5 1,4 0,0140
C4 Celmin (1,5%) 61 5,7 7 Teepol (1:400)
3,5 1,9 0,0139
C5 Celmin (1,5%) 61 5,7 9 Teepol (1:400)
0,5 0,3 0,0254
C6 Celmin (1,5%) 61 5,7 9 Teepol (1:400)
1,5 0,8 0,0149
C7 Celmin (1,5%) 61 5,7 9 Teepol (1:400)
2,5 1,4 0,0205
C8 Celmin (1,5%) 61 5,7 9 Teepol (1:400)
3,5 1,9 0,0197
I1 Trad2 (1:10) 192 2,8 6,5 *A845N p (1:250)
5,3 1,8 0,0121
I2 Trad2 (1:10) 192 2,8 8 *A845N p (1:250)
5,3 1,8 0,0146
I3 Trad2 (1:10) 192 2,8 9 *A845N p (1:250)
5,3 1,8 0,0140
I4 Trad2 (1:10) 192 2,8 10 *A845N p (1:250)
5,3 1,8 0,0165
I5 Trad2 (1:10) 384 5,6 6,5 *A845N p (1:250)
11 3,6 0,0123
IT1 Trad2
(1:10)/845(1:250) 192/5,3 2,8/1,8 6,5
Teepol (1:400)
3,7 2 0,0159
IT2 Trad2
(1:10)/845(1:250) 192/5,3 2,8/1,8 8
Teepol (1:400)
3,7 2 0,0140
IT3 Trad2
(1:10)/845(1:250) 192/0 2,8/0 6,8
Teepol (1:400)
3,7 2 0,0119
IT4 Trad2
(1:10)/845(1:250) 0/5,3 0/1,8 6,7
Teepol (1:400)
3,7 2 0,1301
MC1 Celmin (1,5%) 61 5,7 7 MIBC
(1:200) 0,5 0,2 0,0317
MC2 Celmin (1,5%) 61 5,7 7 MIBC
(1:200) 1,5 0,5 0,0126
IV
MC3 Celmin (1,5%) 61 5,7 7 MIBC
(1:200) 2,5 0,9 0,0146
MC5 Celmin (1,5%) 61 5,7 7 MIBC
(1:200) 5 1,8 0,0280
MC4 Celmin (1,5%) 61 5,7 7 MIBC
(1:200) 3,5 1,2 0,0900
CX1 Celmin (1,5%) 5 0,5 7 MIBC
(1:200) 1,5 0,5 0,0233
CX2 Celmin (1,5%) 10 0,9 7 MIBC
(1:200) 1,5 0,5 0,0228
CX3 Celmin (1,5%) 100 9,3 7 MIBC
(1:200) 1,5 0,5 0,0155
CX4 Celmin (1,5%) 200 18,7 7 MIBC
(1:200) 1,5 0,5 0,0182
CX5 Celmin (1,5%) 1000 93,3 7 MIBC
(1:200) 1,5 0,5 0,0479
F1 Trad3 (1:10) 50 0,8 7 Teepol (1:400)
5,3 2,8 0,0173
F2 Trad3 (1:10) 100 1,5 7 Teepol (1:400)
5,3 2,8 0,0155
F3 Trad3 (1:10) 200 3 7 Teepol (1:400)
5,3 2,8 0,0124
F4 Trad3 (1:10) 300 4,5 7 Teepol (1:400)
5,3 2,8 0,0128
F5 Trad3 (1:10) 400 6 7 Teepol (1:400)
5,3 2,8 0,0405
F5+T Trad3 (1:10) 400 6 7 Teepol (1:400)
5,3 2,8 0,0130
F6 Trad3 (1:10) 50 0,8 9 Teepol (1:400)
5,3 2,8 0,0169
F8 Trad3 (1:10) 200 3 9 Teepol (1:400)
5,3 2,8 0,0150
F10 Trad3 (1:10) 400 6 9 Teepol (1:400)
5,3 2,8 0,0137
CX6 Celmin (1,5%) 120 11,2 7 Nothing 0 0 0,0432
C9 Celmin (1,5%) 61 5,7 7 Nothing 0 0 0,1805
CX7 Celmin (1,5%) 200 18,7 7 Nothing 0 0 0,0512
F11 Trad3 (1:10) 200 3 7 *A845N p (1:250)
2,2 0,8 0,0136
F12 Trad3 (1:10) 200 3 7 *A845N p (1:250)
3,4 1,3 0,0122
F13 Trad3 (1:10) 200 3 7 *A845N p
(1:25) 64 2,2 0,0154
F14 Trad3 (1:10) 200 3 9 *A845N p
(1:25) 64 2,2 0,0163
F15 Trad3 (1:10) 200 3 7 *A845N p (1:250)
5,3 1,9 0,0120
F16 Trad3 (1:10) 200 3 9 *A845N p (1:250)
5,3 1,9 0,0158
MC6 Celmin (1,5%) 80 7,5 7 MIBC(1:200) 1,5 0,5 0,0670
V
F17 Trad3 (1:10) 200 3 8 *A845N p (1:250)
5,3 1,9 0,0182
F18 Trad3 (1:10) 50 0,76 7 *A845N p (1:250)
5,3 1,9 0,0155
F19 Trad3 (1:10) 400 6,05 7 *A845N p (1:250)
5,3 1,9 0,0131
F20 Trad3 (1:10) 200 3 6,5 *A845N p (1:250)
5,3 1,9 0,0129
A4 Trad1 (1:10) 268 4,1 7 *A845N p (1:250)
2,2 0,8 0,0117
A5 Trad1 (1:10) 268 4,1 7 *A845N p (1:250)
3,4 1,2 0,0116
F21 Trad3 (1:10) 268 4,02 7 *A845N p (1:250)
5,3 1,9 0,0122
F22 Trad3 (1:10) 268 4,02 7 Teepol (1:400)
5,3 2,8 0,0154
IST 1 Trad1 (1:10) 268 4,1 9,3 *A845N p
(1:50) 20 1,4 0,0154
A6 Trad1 (1:10) 268 4,1 7 *A845N p (1:250)
3,4 1,2 0,0116
F23 Trad3(1:10) 268 4 7 *A845N p (1:250)
5,3 1,9 0,0166
F24 Trad3(1:10) 268 4 7 Teepol (1:400)
5,3 2,8 0,0553
C10 Celmin (1,5%) 61 5,7 7 Dowfroth
250 C 1,5 1,1 0,0611
C11 Celmin (1,5%) 61 5,7 7 Dowfroth
250 C 1,5 1,1 0,0230
C12 Celmin (1,5%) 61 5,7 Dowfroth
250 C 1,5 1,1 0,0172
CIST Celmin (1,5%) 61 5,7 6,8 MIBC 1 drop ? 0,0201
D 1 Celmin (1,5%) 40 3,7 7 Dowfroth
250 C (1:500)
1,5 1,1 0,0151
D 2 Celmin (1,5%) 40 3,7 8,5 Dowfroth
250 C (1:500)
1,5 1,1 0,0150
D 3 Celmin (1,5%) 40 3,7 10 Dowfroth
250 C (1:500)
1,5 1,1 0,0144
D 4 Celmin (1,5%) 120 11,1 7 Dowfroth
250 C (1:500)
1,5 1,1 0,0156
D 5 Celmin (1,5%) 120 11,1 8,5 Dowfroth
250 C (1:500)
1,5 1,1 0,0166
D 6 Celmin (1,5%) 120 11,1 10 Dowfroth
250 C (1:500)
1,5 1,1 0,0151
D 7 Celmin (1,5%) 200 18,7 7 Dowfroth
250 C (1:500)
1,5 1,1 0,0165
D 8 Celmin (1,5%) 200 18,7 8,5 Dowfroth
250 C (1:500)
1,5 1,1 0,0165
D 9 Celmin (1,5%) 200 18,7 10 Dowfroth
250 C (1:500)
1,5 1,1 0,0156
VI
D 10 Celmin (1,5%) 40 3,7 7 Dowfroth
250 C (1:500)
1,5 1,1 0,0154
D 11 Celmin (1,5%) 40 3,7 8,5 Dowfroth
250 C (1:500)
1,5 1,1 0,0146
D 12 Celmin (1,5%) 40 3,7 10 Dowfroth
250 C (1:500)
1,5 1,1 0,0145
D 13 Celmin (1,5%) 120 11,1 7 Dowfroth
250 C (1:500)
1,5 1,1 0,0150
D 14 Celmin (1,5%) 120 11,1 8,5 Dowfroth
250 C (1:500)
1,5 1,1 0,0158
D 15 Celmin (1,5%) 120 11,1 10 Dowfroth
250 C (1:500)
1,5 1,1 0,0154
D 16 Celmin (1,5%) 200 18,7 7 Dowfroth
250 C (1:500)
1,5 1,1 0,0151
D 17 Celmin (1,5%) 200 18,7 8,5 Dowfroth
250 C (1:500)
1,5 1,1 0,0149
D 18 Celmin (1,5%) 200 18,7 10 Dowfroth
250 C (1:500)
1,5 1,1 0,0163
D 19 Celmin (1,5%) 40 3,7 7 Dowfroth
250 C (1:500)
1,5 1,1 0,0162
D 20 Celmin (1,5%) 40 3,7 8,5 Dowfroth
250 C (1:500)
1,5 1,1 0,0168
D 21 Celmin (1,5%) 40 3,7 10 Dowfroth
250 C (1:500)
1,5 1,1 0,0155
D 22 Celmin (1,5%) 120 11,1 7 Dowfroth
250 C (1:500)
1,5 1,1 0,0170
D 23 Celmin (1,5%) 120 11,1 8,5 Dowfroth
250 C (1:500)
1,5 1,1 0,0150
D 24 Celmin (1,5%) 120 11,1 10 Dowfroth
250 C (1:500)
1,5 1,1 0,0151
D 25 Celmin (1,5%) 200 18,7 7 Dowfroth
250 C (1:500)
1,5 1,1 0,0162
D 26 Celmin (1,5%) 200 18,7 8,5 Dowfroth
250 C (1:500)
1,5 1,1 0,0154
D 27 Celmin (1,5%) 200 18,7 10 Dowfroth
250 C (1:500)
1,5 1,1 0,0146
D28 Celmin (1,5%) 120 11,1 8,5 Dowfroth
250 C (1:500)
1,5 1,1 0,0151
D29 Celmin (1,5%) 120 11,1 8,5 Dowfroth
250 C (1:500)
1,5 1,1 0,0156
VII
D30 Celmin (1,5%) 120 11,1 8,5 Dowfroth
250 C (1:500)
1,5 1,1 0,0154
D31 Celmin (1,5%) 120 11,1 8,5 Dowfroth
250 C (1:500)
1,5 1,1 0,0165
D32 Celmin (1,5%) 120 11,1 8,5 Dowfroth
250 C (1:500)
1,5 1,1 0,0157