You are on page 1of 20

Accepted Manuscript

Modifications of basic-oxygen-furnace slag microstructure and their effect on the


rheology and the strength of alkali-activated binders

Pavel Leonardo Lopez Gonzalez, Rui M. Novais, Joao Labrincha, Bart Blanpain,
Yiannis Pontikes

PII: S0958-9465(18)30642-5
DOI: https://doi.org/10.1016/j.cemconcomp.2018.12.013
Reference: CECO 3205

To appear in: Cement and Concrete Composites

Received Date: 22 June 2018


Revised Date: 9 November 2018
Accepted Date: 17 December 2018

Please cite this article as: P.L. Lopez Gonzalez, R.M. Novais, J. Labrincha, B. Blanpain, Y. Pontikes,
Modifications of basic-oxygen-furnace slag microstructure and their effect on the rheology and
the strength of alkali-activated binders, Cement and Concrete Composites (2019), doi: https://
doi.org/10.1016/j.cemconcomp.2018.12.013.

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to
our customers we are providing this early version of the manuscript. The manuscript will undergo
copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please
note that during the production process errors may be discovered which could affect the content, and all
legal disclaimers that apply to the journal pertain.
ACCEPTED MANUSCRIPT

MODIFICATIONS OF BASIC-OXYGEN-FURNACE SLAG MICROSTRUCTURE AND THEIR EFFECT ON THE


RHEOLOGY AND THE STRENGTH OF ALKALI-ACTIVATED BINDERS

Corresponding author: Pavel Leonardo Lopez Gonzaleza

e-mail: pavelleonardo.lopezgonzalez@kuleuven.be

Co-author: Rui M. Novaisb

PT
Co-author: Joao Labrinchab

Co-author: Bart Blanpaina

RI
Co-author: Yiannis Pontikesa
a

SC
KU Leuven, Department of Materials Engineering, Kasteelpark Arenberg 44, 3001 Leuven, Belgium
b
Department of Materials and Ceramic Engineering / CICECO- Aveiro Institute of Materials, University of
Aveiro, Campus Universitário de Santiago, 3810-193 Aveiro, Portugal

U
AN
Abstract

Microstructure tailoring of metallurgical slags allows the production of alternative construction binders
with customized properties. In this study, the variations of rheology and strength of alkali-activated
M

basic-oxygen-furnace (BOF) slags are quantified. Two modifications of BOF slag were created adding
amounts of alumina and silica at high temperature (>1250°C). The additions, defined by thermodynamic
modeling, lowered the liquidus temperature facilitating the generation of amorphous when the slag was
D

fast cooled. The first modification (SAT1) incorporated 5 wt% silica and 11 wt% alumina, while the
second (SAT2) included around 13 wt% alumina. Both modifications generated a hybrid microstructure
TE

composed of cementitious and non-cementitious crystalline phases and an amorphous fraction. During
alkali activation using NaOH solutions of 0.25 M, rheological measurements on fresh paste using SAT2
registered plastic viscosity values 2.3 times higher than those of SAT1. The compressive strength after 28
EP

days for the binder developed from SAT2 slag was 10 to 30% stronger than the one from SAT1. These
binders showed similar crystalline reaction products but compositional differences in the amorphous gel
correlated to the initial slag modification. The detected differences in the binder properties are
significant enough to justify BOF-slag engineering as a way to deliver customized precursors for alkali
C

activation.
AC

Keywords: BOF slag, slag engineering, slag processing, slag valorization, alkali activation, alkali-activated
material.
ACCEPTED MANUSCRIPT

1. Introduction

Microstructure tailoring of metallurgical slags opens the possibility of creating new binder precursors
while providing materials to the construction sector with customized behavior and performance. At the
same time, environmental gains are achieved through the valorization of secondary materials and the
reduction of net CO2 emissions linked to the binder production when compared to traditional Portland
cement [1,2]. Basic oxygen furnace (BOF) slag generated during steel production exemplifies the case:

PT
considerable amounts of this material are available (an estimation of 120 to 240 Mt in 2016 was
calculated using the reported global production of steel via oxygen furnace [3] with a slag generation
rate of 100 to 200 kg per ton of steel [4]), and several studies have demonstrated that appropriate

RI
processing develops cementitious properties in the slag [4–10]. The current re-utilization of BOF slag
varies depending on the country of origin. Nevertheless, considerable amounts of slag are still
temporary stored and landfilled. In Europe a production of 18.4 million of tons of steel-making slag was

SC
reported for 2016 from which 46% was used in road construction, 15% reused in metallurgical processes
and 14% was disposed of (representing around 2.6 MTon) [11]. More noticeable, China, the biggest steel
producer, reports total re-utilization rates of only 29.5% leading to an accumulation of more than 300
MTon by 2016 [12].

U
The BOF-slag use as aggregates in concrete is restricted by the volumetric expansion associated with the
AN
hydration of the free-CaO and free-MgO present in the slag [13,14]. Thus, alternative valorization paths
have been studied for several years now. The research focused on the slag conversion into a volumetric
stable aggregate [15,16], later shifting to the modification of the slag into a latent hydraulic binder
M

compatible with traditional Portland-cement-based mixtures. The latter case includes studies of blends
[4], use of accelerating additives [7], thermal treatments [9,10], and compositional modifications [5,7,8].
The role of amorphous content and its link with reactivity has been also identified [5,17], results
D

obtained from hydration studies showed that modified BOF slag with an amorphous content of at least
60 wt% was reactive enough to be considered a binder by itself [5], raising the possibility of developing
TE

cementitious materials made solely by steel slags that could be activated by alkaline solutions [18].
More recently, the impact of alumina additions to BOF slag has been reported [19]. This treatment
lowers the liquidus temperature and modifies the viscosity of the melt [20] offering the possibility of
producing glassy slag when cooled fast enough, which is a consequence of the increase in the kinetic
EP

barrier to crystallization due to larger melt viscosities [21]. The modified BOF slag was shown to be
reactive and its potential use as a binder for tailoring cementitious product was pointed out [22].

The interest in alkali activation of slags and their use as cementitious binders has increased in recent
C

years [18,23,24]. However, more specific data linking the microstructure and the performance of the
activated binders is still needed. The objective of this study was to quantify the variation of strength and
AC

rheology of alkali-activated binders produced with two processed BOF-slags with modified compositions
and microstructures. The extent of such impacts can confirm the possibility of tailoring slags for specific
purposes as binders. The studied modifications consisted of additions of alumina and silica at high
temperature followed by fast cooling (granulation). This processing has also been reported as a way to
remove free lime from the BOF slag [19]. Both modified slags were fully characterized and activated
using NaOH solutions; additions of gypsum, cement, and plasticizer were also incorporated to control
setting time, strength development and workability, respectively. The rheology of fresh mixes of the
binder was studied and compared. For each modified slag, an optimal design of experiments (ODOE)
[25] was defined and used to fit a response surface model of the compressive strength. From the model
ACCEPTED MANUSCRIPT

the mixture that delivered the highest strength (>15 MPa) was selected, these binders were further
analyzed and compared.

2. Materials and methods

Two modifications of an industrial master BOF slags were investigated (SAT1 and SAT2). The additions
were selected based on thermodynamic modeling and aimed at the reduction of the liquidus

PT
temperature of the melt and the generation of amorphous content in the processed slags. FactSage 7.1
software was used to model the expected changes in the liquidus temperature and the mineralogy. The
Equilib module was employed including FactOxid and FactPS databases with the precipitate target phase
option activated for the FToxid-SLAGA solution. Liquidus temperature was plotted as a function of

RI
alumina and silica additions up to 14 and 12 wt% respectively. The assumed starting composition of the
master slag in the calculations is presented in Table 1, together with the actual slag composition, which
has been previously reported [19]. Liquidus and mineralogy calculations considered a Fe3+/Fe2+ ratio

SC
close to 2, based on information provided by the slag producer. The mineralogy of SAT1 and SAT2 was
modeled using the Scheil-Gulliver approach, considering that it provides more accurate predictions of
the mineralogy for industrial slags [26]. These solidification calculations used a simplified composition of

U
the slags and only components representing 1 wt% or more were taken into consideration.
AN
Table 1. Compositions of master and modified slags (in wt%). Coding: (R) - previously reported, (A) -
assumed, (M) – measured. Note: Except for Fe, the oxidation state of multivalent metals was assumed.
Total
Origin/Use CaO FeO Fe2O3 SiO2 Al2O3 MgO MnO TiO2 V2O5 ZrO2 Cr2O3 P2O5 NiO
Fe
M

Master BOF slag (R) 42-55 14-20 - - 12-18 0-3 0-5 0-8 0-1 0-0.5 - - 0-2 -
In liquidus modeling (A) 42.0 - 7.0 13.0 12.0 1.0 4.0 2.0 - - - - 1.0 -
From XRF on SAT1 (M) 38.5 15.4 - - 17.0 12.1 5.4 1.9 1.0 0.5 0.2 0.2 1.4 0.0
D

In Scheil-Gulliver solidification SAT1 (A) 38.9 - 6.7 14.6 17.1 12.1 5.4 1.9 1.0 - - - 1.4 -
From XRF on SAT2 (M) 38.9 18.2 - - 12.6 13.8 4.1 2.2 0.9 0.5 0.1 0.2 1.2 0.3
TE

In Scheil-Gulliver solidification SAT2 (A) 39.0 - 7.5 17.6 12.6 13.8 4.1 2.2 - - - - 1.2 -

The two slag modifications involved the addition of defined amounts of silica and alumina, the re-
EP

melting of the mix, and the fast cooling of the melt through granulation. For each, the parent-industrial
BOF slag was mixed with an alumina-silica- rich source in the appropriate proportions. The mixes were
melted at a pilot plan in a reactor of 1500 kg of capacity coupled with a plasma generator and a wall
C

cooling system that allowed the generation of freeze lining. The heating cycles for SAT1 and SAT2
modification lasted 56 and 27 hours respectively. Several tapped samples were collected during the
AC

heating to track the melt compositions. The mix of air with liquefied-petroleum-gas (LPG) feeding the
reactor was kept constant to 21 during most part of the heating, then modified to 23.5 in the last 30
minutes aiming at generating a Fe3+/Fe2+ ratio equals to 2 in the melt. Once targeted compositions were
achieved, the slag (718 kg in the first run and 672 kg in the second) was granulated using water jets on
the molten stream and poured into a water tank. The melt temperature before granulation (around
1300°C) was estimated by the reactor operator using data from the cooling system.

In the first part of the study, the modified slags were characterized. X-ray fluorescence spectrometry
(XRF), using a Panalytical PW2400 with 3 kW rhodium anode tube, was performed to establish the
ACCEPTED MANUSCRIPT

elemental composition. The X-ray diffractograms (XRD) were obtained with a D2 Phaser (Bruker)
measuring 2theta angles from 5 ° to 70 °, with a step size of 0.02 ° and 0.6 s per step employing a Cu
tube at a voltage of 30 kV and a current of 10 mA. Quantification of the mineralogical phases was made
with modified slag powder samples mixed with 10 wt% of crystalline ZnO, which were then milled for 5
minutes using a McCrone micronizing mill using hexane (purity >99 %) as grinding agent and corundum
beads as milling media. EVA and TOPAS-Academic (Rietveld method) software were used to identify and
quantify the mineralogy respectively. Identification of the microstructure was supplemented with back-

PT
scattered electron images and mapping obtained using an FEI XL30 FEG scanning electron microscope
with a Schottky type of gun coupled with an EDAX EDS detector. Electron probe microanalysis (EPMA)
using a field emission microprobe JXA-8530F (Jeol) with wavelength dispersive spectrometers (WDS) was

RI
employed to establish the composition of the amorphous regions present among crystalline structures
and in particles of apparent glassy nature. At least five-point analyses were performed in each region
and the average composition was reported. A material flow showing the fate of the compounds during

SC
processing was estimated based on the measured compositions and the quantified mineralogy of SAT1
and SAT2. Phases were grouped and their elemental proportions assumed as reported in Table 2 and 3.
To evaluate the representability, the estimated amorphous composition of each slag was compared with

U
the actual EPMA compositional measurements. Modified slags' infrared spectra for wavelengths ranging
from 400 until 1400 cm-1 were also acquired using a Fourier transform infrared spectrometer, model
AN
Alpha (Bruker), coupled with an attenuated total reflectance module ATR-FTIR.

The fresh behavior of the activated slags was studied through rheological measurements of pastes using
two different formulations using a HAAKE Viscotester iQ with cylinder sensors. Tests were carried out
M

using shear rates ranging from 0 to 160 s-1 during 90 s both for the up and the down curves. The
Bingham model was used to fit the data from the down curve to determine the slurries’ yield stress and
plastic viscosity, as described elsewhere [27]. Samples were tested immediately after mixing (time 0
D

min) and then after 5, 10, 15 and 20 min. Due to the fast setting of the slurries, the sample was kept
inside the rheometer throughout the measurements.
TE

For the study of mechanical strength, paste specimens with the mixture proportions shown in Table 4
were produced. The mixtures were selected employing JMPpro 12.1, a commercial statistical data
analysis software, and its DOE-Custom design module. The experimental design was run with I-
EP

optimality criteria algorithm, which minimizes the average variance of prediction of the model. For the
specimen preparation, finely-milled slag samples with an average specific surface around 3550 g/cm2
(measured with a Blaine apparatus) were alkali activated with different concentrations of NaOH
C

solutions ranging from 0 (only water) to 0.5 M. The mixtures included additions of 2 to 6 wt% gypsum
and 0 to 10 wt% cement CEMI. The impact of 0.2 wt% of a commercial plasticizer (Sika ViscoCrete 150-P
AC

powder), a polycarboxylate-based dispersant that causes steric hindrance at the binder particles, was
also evaluated. Paste specimens were hand mixed for 60 s and cast in 20x20x80 mm3 molds, then de-
molded after 48 h and cured in sealed plastic containers until the moment of testing. The liquid to
binder ratio was constant and equal to 0.4 for all mixtures. The compressive strength was measured
after 28 days using an Instron universal machine with a loading rate of 1 mm/min, each specimen was
first split into two halves using a 3-point bending set-up and the compressive strength was determined
on each resultant half. The specimens providing the highest strength were further characterized using
XRD, FTIR, BSE microscopy, EDS to map elemental distribution and WDS to establish compositional
microanalysis.
ACCEPTED MANUSCRIPT

In addition, the response surface models (RMS) for 28-day compressive strength for SAT1 and SAT2
were determined. To construct and evaluate the models, JMP software’s Fit-Model module from the
Analyze menu was used. The models considered 3 numerical factors (activating solution molarity
expressed as M, cement addition and gypsum addition as wt% added to solid binder) and one
categorical (plasticizer included or not).

3. Results and discussion

PT
3.1 Modified slags characterization

Figure 1 presents the modeled liquidus temperature when different amounts of silica and alumina are

RI
added to the master slag. The impact on the liquidus temperature of single additions (values across x-
and y-axis) is similar, reducing in average 39 °C and 37 °C per each additional wt% of alumina and silica
respectively. Combined additions can reach similar reductions in temperature provided equivalent wt%

SC
addition is reached (e.g. an addition of 8 wt% alumina or silica will have roughly the same impact of
adding 4 wt% silica and 4 wt% alumina). The studied modifications were selected to have similar
reductions of liquidus temperature to close to 1350 °C. An additional selection criterion, the amorphous

U
content generation, was approached considering a random network model for the glass forming ability
of the melt [28,29]. In this context, the additions would work as supplementary network formers (SiO2
AN
and Al2O3) altering the rheology of the melts and thus their amorphous content after fast cooling. The
chosen modifications, SAT1 and SAT2, contain (CaO+MgO)/ (SiO2+Al2O3) molar ratios of 2.0 and 2.3
respectively. Assuming that cooling rate is fast enough and equal for both modified slags, the slightly
lower amount of network modifiers in SAT1 implies a melt with higher viscosity and thus more prone to
M

form amorphous content. The studied compositional modifications are indicated in Figure 1.
D
TE
C EP
AC

Figure 1. Modeled liquidus temperature (in °C) as a function of different additions of SiO2 and Al2O3 to
the master slag.

The actual compositional modifications, determined by XRF analysis, are reported above in Table 1.
SAT1 had an addition of SiO2 that is 5 wt% higher than that of SAT2, while both slags present practically
ACCEPTED MANUSCRIPT

the same Al2O3 addition level (close to 11 wt%). The differences in the total Fe reported can be
associated with variations of the actual oxidation state of Fe present in each slag linked to O2 availability
during processing; the total Fe value reported in Table 1 was estimated averaging the quantification
results of analyses reporting Fe as FeO and as Fe2O3. The lower values of CaO of the modified slags with
respect to the reported range for the master slag can be attributed to interactions with the freeze lining
of the reactor vessel.

PT
Figure 2 presents the calculated solidification for SAT1 and SAT2. The predicted phases are common to
both slags. Spinel/Magnetite would precipitate first in SAT1, whereas dicalcium silicate would be first in
SAT2. Liquid slag, dicalcium silicate, and magnetite/spinel would represent the main portion for

RI
temperatures above 1225 °C. Below this temperature calcium aluminoferrite phases would appear,
reaching up to 15 and 35 wt% at 1100 °C for SAT1 and SAT2 respectively. The oxide solid solution would
develop at temperatures below 1150 °C and no higher than 10 wt% in SAT1 and 15 wt% in SAT2. The

SC
representability of the calculated microstructure was evaluated through comparison with the actual
phases identified in the XRD patterns, presented in Figure 3, and their correspondent quantification
reported in Table 2 and 3. For SAT1, the main precipitating phases are correctly predicted.
Quantitatively, the amounts of amorphous content, assumed to correspond to liquid slag, and the

U
magnetite/spinel fairly agree to those calculated for a temperature close to 1290 °C. Yet, for that
AN
temperature, the amount of dicalcium silicate was overestimated and no other phases were predicted,
which differed from the XRD findings. In the case of SAT2, the main phases are again correctly
predicted, amounts of liquid slag and calcium aluminoferrite fit the ones calculated for a temperature
close to 1210 °C but dicalcium silicate and magnetite/spinel phases are overestimated and oxide
M

solution underestimated.

(a) (b)
D
TE
C EP
AC

Figure 2. Solidification modeling of the modified slags using Scheil-Gulliver assumptions: (a) SAT1 and (b)
SAT2.
ACCEPTED MANUSCRIPT

PT
RI
SC
Figure 3. XRD patterns for modified slags SAT1 and SAT2.

U
The actual microstructure quantified from the XRD patterns in Figure 3 indicates that modified slags are
mainly composed of amorphous fractions: 65.9 and 55.5 wt% for SAT1 and SAT2 respectively. Figure 4
AN
shows the average amorphous composition for both slags measured on particles with no crystal
presence, and the one measured in the amorphous regions near crystalline phases (no less than 2 µm
distance). The detected differences are associated with the localized diffusion of elements from
amorphous regions (areas without evident crystalline features) into adjacent crystals.
M

(a) (b)
D
TE
C EP
AC

Figure 4. Comparison of EPMA-WDS compositional microanalysis of amorphous and mainly crystalline


regions for (a) SAT1 and (b) SAT2 (XRF composition is included for reference).

The remaining microstructure is composed of crystalline structures identified in Figure 5 and is


described as follows:
ACCEPTED MANUSCRIPT

• Dicalcium silicate refers to high-temperature forms of a phase close in composition to belite.


EDS analyses showed that, besides Ca and Si, it contains around 3 wt% of P and minor amounts
(~1 wt% or less) of Fe, Mg, Mn, Al, and Ti. This phase constitutes the second major feature after
the amorphous fraction being 15.5 and 23.0 wt% for SAT1 and SAT2 respectively.
• Magnetite and a spinel form with a composition close to galaxite with the empirical formula
(Mg0.142Mn0.815Fe2+0.043)(Fe3+0.227Al1.773)O4 were fitted for both slags. In SAT1 the pattern shows

PT
more intense peaks associated with the high presence of both spinel and magnetite (4.4 and 7.5
wt%) than in SAT2 (0.6 and 5.7 wt% ).
• Calcium aluminoferrite refers to a solid solution of formula Ca2(AlxFe1-x)O5 with varying amounts
of alumina (x was assumed from 0 to 0.48). Preliminary EPMA results on lab samples indicated

RI
an average composition close to Ca2(Al0.45Fe0.55)O5, which was used in mass flow calculations.
Rietvield quantification fitted C4AF (x=0.5) for both slags in the same proportion (4.1 wt%) but

SC
for SAT2 the end member C2F (x=0, 1.2 wt%) and CaFe2O4 (0.9 wt%) were also included
indicating a composition richer in Fe for SAT2.
• (Fe, Mg, Mn)O refers to a solid solution of different oxides. Preliminary EPMA analyses
indicated a broad compositional variation with Fe and Mg as main elements (representing up to

U
80 wt% of the total non-oxygen elements). The solution composition provided by the
thermodynamic calculations for 1300°C were used for the mass flow. Presence of this phase is
AN
lower in SAT1 (2.6 wt%) than in SAT2 (9.0 wt%).

(a) (b)
M
D
TE
EP

Figure 5. BSE-SEM images of crystalline regions of (a) SAT1 and (b) SAT2; 1: amorphous, 2: dicalcium
silicate (removed during SEM sample preparation), 3: Solid solution of Fe, Mg, and Mn oxides, 4:
C

Magnetite/Spinel
AC

The mass flows depicting the fate of the modified compositions into the quantified microstructures are
presented in Figure 6. Tables 2 and 3 contain the assumed proportions used and the estimated
amorphous compositions. The variation between the estimated and measured amorphous compositions
were never higher than 0.05 g/g except for the case of alumina in SAT2 amorphous where an
overestimation of +0.09 g/g occurs.
ACCEPTED MANUSCRIPT

(a)

PT
RI
U SC
AN
(b)
M
D
TE
C EP
AC

Figure 6. Calculated material flow for (a) SAT1 and (b) SAT2 representing the mineralogy developed
during slag modification. The initial compositions consider the master slag plus the selected additions.
ACCEPTED MANUSCRIPT

Table 2. Component mass proportions (mass of component/mass phase) for quantified microstructure
of SAT1 used in material flow with the correspondent estimated amorphous composition. The measured
composition and its difference with the estimate are reported in the last two columns.
Amorphous
Calcium Amorphous
Dicalcium Ferrite measured ∆ (estimated
Component Magnetite (Fe, Mg, Mn)O alumino estimated
silicate spinel composition - measured)
ferrite composition
(EPMA)

PT
CaO 0.65 0.45 0.41 0.37 0.04
FeO 0.31 0.01 0.25 0.06 0.07 -0.02
Fe2O3 0.69 0.32 0.20 0.36 0.09 0.14 -0.05
SiO2 0.35 0.18 0.18 0.00

RI
Al2O3 0.37 0.19 0.15 0.12 0.03
MgO 0.13 0.48 0.05 0.05 0.00
MnO 0.16 0.07 0.02 0.02 0.00

SC
P2O5 0.02 0.02 0.00
Other 0.03 0.01 0.02
Quantified
amount in wt% 15.5 7.5 4.4 2.6 4.1 65.9
(QXRD)

U
AN
Table 3. Component mass proportions (mass of component/mass phase) for quantified microstructure
of SAT2 used in material flow with the correspondent estimated amorphous composition. The measured
composition and its difference with the estimate are reported in the last two columns.
M

Amorphous
Calcium Amorphous ∆
Dicalcium Ferrite Calcium Dicalcium measured
Component Magnetite (Fe, Mg, Mn)O alumino estimated (estimated -
silicate spinel ferrite ferrite composition
ferrite composition measured)
(EPMA)
D

CaO 0.65 0.26 0.41 0.45 0.39 0.35 0.03


FeO 0.31 0.01 0.28 0.06 0.08 -0.02
Fe2O3 0.69 0.32 0.10 0.74 0.59 0.36 0.18 0.16 0.01
TE

SiO2 0.35 0.08 0.13 -0.05


Al2O3 0.37 0.19 0.23 0.14 0.09
MgO 0.13 0.42 0.00 0.05 -0.04
MnO 0.16 0.20 0.01 0.02 -0.01
EP

P2O5 0.02 0.02 0.01


Other 0.04 0.00 0.04
Quantified
amount in 23.0 5.7 0.6 9.0 0.9 1.2 4.1 55.5
wt% (QXRD)
C
AC

3.2 Comparison of alkali-activated binders

The rheological measurements of the slurries are presented in Figure 7. As expected, the slurries’ plastic
viscosity and yield stress are remarkably lower when using water in the compositions (Mixture ID 13 in
Table 4) in comparison with the use of sodium hydroxide (ID 11), this being particularly visible for SAT1.
Indeed, the plastic viscosity and yield stress for sample SAT1-13 remained almost constant, around 0.1
Pa.s and 10 Pa respectively, throughout the test, suggesting that it did not set/hardened at least during
the first 20 min. An increase in the plastic viscosity with time from 0.18 to 0.61 Pa.s was observed for the
composition SAT1-11 in the first 20 min, suggesting faster setting. These results are in good agreement
ACCEPTED MANUSCRIPT

with the compressive strength exhibited by these two compositions (see Table 4), which showed much
higher compressive strength for SAT1-11 in comparison with SAT1-13.

As for the SAT2 compositions similar trends were observed, with higher plastic viscosity and yield stress
being reached for the SAT2-11 in comparison with their SAT2-13 counterpart. Nevertheless, a slight
increase on both the plastic viscosity (from 0.17 to 0.32 Pa.s) and yield stress (from 17.0 to 23.5 Pa)
upon curing was observed for SAT2-13 which did not occur in the SAT1-13, again suggesting distinct

PT
setting and hydration behavior between the two studied slags. Composition SAT2-11 shows a much
faster and higher increase in the plastic viscosity in comparison with SAT1-11. In fact, immediately after
mixing (t = 0 min) the slurries’ viscosity differs, with SAT1-11 having roughly half of that observed for

RI
SAT2-11. The differences between the compositions were further intensified upon curing, the SAT2-11
reaching a viscosity of 1.4 Pa.s which is 2.3 times higher than that of SAT1-11. Once again these results
are in line with the higher mechanical performance exhibited by the SAT2 slag. The measured

SC
differences can be regarded as a sign of reactivity of the slag, indicating that SAT2 slag is more reactive
than SAT1.

U
AN
M
D
TE

Figure 7. Rheological properties of fresh mixtures containing SAT1 or SAT2 activated when mixed with
H2O (L/B=0.4) and 2 wt% gypsum (mixture ID 13) or with NaOH (0.25 M), 4 wt% gypsum and 5 wt%
EP

cement (mixture ID 11), expressed as (a) yield stress and (b) plastic viscosity.

For the mechanical strength study, the complete list of tested alkali-activated paste specimens, their
mixture proportioning, and measured compressive strength at 28 days are presented in Table 4. The
C

compressive strength of SAT2 specimens was higher than the one of SAT1 in 14 out of 17 mixtures
AC

(representing 82 % of tested specimens) with 6.4 MPa difference in average. Considering individual
samples, the strength difference ranged from 2.5 to -18.7MPa depending on the mixture used,
indicating a difference in behavior for each slag. In both slags, consistent maximum strength values were
obtained for mixtures 7 and 11, with a strength increase (of SAT2 specimens over SAT1 ones) of 10% for
the former and 30% for the latter. Paste specimens produced with mixture 11 were used for further
characterization of the binder.
ACCEPTED MANUSCRIPT

Table 4. 28-day compressive strength of paste specimens produced using SAT1 and SAT2 as precursors,
mixture details are included (green-dotted box: excluded data, orange-box: replicate measurement,
grey: negative strength difference).
SAT1 28d SAT2 28d
Addition of Additional ∆ strength
Activating Addition of Inclusion of compressive compressive
Gypsum SAT2 28d (SAT1-SAT2) Mixture ID
solution [M] CEM1 [wt%] plasticizer strength strength
[wt%] replicates [MPa]

PT
[MPa] [MPa]
0.00 2 10 No 11.6 9.8 13.4 0.0 3
0.00 2 5 Yes 12.3 14.4 - - 17
0.00 4 10 Yes 10.9 15.6 - -4.7 9

RI
0.00 4 0 No 2.7 21.4 - -18.7 13
0.00 6 0 Yes 1.3 6.0 8.1 -5.8 5
0.00 6 8.75 No 18.7 20.6

SC
- -1.9 7
0.25 2 0 Yes 11.5 17.7 - -6.2 14
0.25 2 5 No 11.8 20.0 - -8.2 20
0.25 4 5 Yes 15.1 22.5 - -7.4 6

U
0.25 4 5 Yes 17.4 20.0 - -2.6 11
0.25 4 10 No 16.7 14.2 - 2.5 12
AN
0.25 4 5 No 12.6 16.8 - -4.2 16
0.25 6 10 Yes 15.6 15.2 - 0.4 10
0.33 6 0 No 5.8 18.2 15.4 -11.0 2
0.50 2 10 Yes 12.3 15.3
M

- -3.0 15
0.50 2 0 No 2.0 10.3 - -8.3 18
0.50 4 5 No 10.4 12.8 13.7 -2.9 1
0.50 4 0 Yes 5.0 12.0
D

- -7.0 4
0.50 6 5 Yes 27.3 14.3 - - 8
TE

0.50 6 10 No 12.2 16.8 - -4.6 19

The fitted models presented in Figure 8 and 9 for paste specimens, using SAT1 and SAT2 as precursors,
EP

predict the existence of optimal values in the region explored with higher and more broadly distributed
strength values for SAT2 than for SAT1. RSM fittings are reliable with a coefficient of determination (R2)
of 0.96 and a root mean square error (RMSE) of 1.6 MPa for SAT1, and R2= 0.81 and RMSE=2.4 MPa for
SAT2. Mixture 8 in SAT1 and mixture 4 and 17 in SAT2 model were excluded to improve model fitting
C

after considering them as non-influential outliers based on calculated Cook's distance values. Close
inspection showed that their high strength readings were artificially generated by irregularities of the
AC

specimens edges, which impacted the actual surfaces loaded. Four additional SAT2 mixtures were
replicated confirming experimental reproducibility. In the RSM fitted for SAT1, cement addition and the
molarity of the activating solutions are the factors with the biggest influence on strength development,
whereas for SAT2 the interaction between the added gypsum and the inclusion of plasticizer played the
main role.
ACCEPTED MANUSCRIPT

(a) (b)

PT
RI
SC
Figure 8. RSM model for the 28-day compressive strength (MPa) of paste specimens produced using

U
SAT1 as precursor, (a) behavior for the addition of 6 wt% gypsum and (b) behavior once plasticizer is
AN
included and 2 wt% gypsum is added.

(a) (b)
M
D
TE
EP

Figure 9. RSM model for the 28-day compressive strength (MPa) of paste specimens produced using
C

SAT2 as precursor, (a) behavior for the addition of 6 wt% gypsum and (b) behavior once plasticizer is
included and 2 wt% gypsum is added.
AC

Regarding the binders produced, the XRD patterns are presented in Figure 10. The only crystalline
product detected was identified as katoite (Ca3Al2(SiO4)3-x(OH)4x with x=1.5-3), a component linked to the
hydration of calcium aluminum- ferrite and silicate-rich phases [30–34]; this is present in both slags with
peaks more developed in SAT2 than in SAT1. The remaining crystalline phases are linked to unreacted
Fe-rich phases, R-O for SAT1, and spinel/magnetite for SAT2. The comparison with initial XRD patterns of
precursor slags confirms the disappearance and thus consumption of reactive cementitious phases
(dicalcium silicate and calcium aluminoferrite).
ACCEPTED MANUSCRIPT

PT
RI
U SC
AN
Figure 10. XRD patterns of binders produced using mixture 11 for SAT1 and SAT2, only new crystalline
phase developed (katoite) is identified, initial patterns of modified slags are replicated for reference.

A comparison between the initial FTIR spectra of the modified slags, presented in Figure 11, shows a
M

broader shoulder in the region between 700 and 850 cm-1 for SAT2 possibly linked to the differences in
alumina-iron contents of the calcium aluminoferrite phases. Sharper more intense features in SAT1 can
also be attributed to the presence of more Si-O bonds in this sample due to its higher silicon content.
D

The main-spectra features fairly match the ones previously reported for pure high-temperature
polymorphs of dicalcium silicate (alpha’- and b-C2S) at 500, 515, 845 cm-1 and calcium-aluminum-rich
TE

phases at 894, 904 and 980 cm-1, normally associated with the shifting of the Si-O stretching band when
Ca and Al are present [35].

The patterns for the alkali-activated binders depict the splitting of the Si-O stretching mode into two
EP

components in the region between 850-950 cm-1. This splitting can be associated with the presence of
Na with shifts assigned to those SiO4 tetrahedra that contain non-bridging oxygens facing sodium atoms
after alkali activation [36]. The bands at 1140 cm-1 and in the region 600 to 670 cm-1 can be linked to the
gypsum present in the final mixture. The features around 1426 cm-1 were assigned to carbonation
C

bands, while those present at 3490 and 1645cm-1 were linked to stretching and bending vibrations of
AC

H2O. The features associated with reactive cementitious phases disappear in the activated patterns

(a) (b)
corroborating once more their consumption.
ACCEPTED MANUSCRIPT

PT
RI
SC
Figure 11. ATR-FTIR patterns for (a) SAT1 and (b)SAT2 and their correspondent binder (mixture 11)

SEM images presented in Figure 12 show the generation of a globular binder with unreacted Fe-rich
crystals. EDS mapping in Figures 13 and 14, for SAT1-11 and SAT2-11 respectively, evidences the

U
generation of an amorphous C(F,N)-A-S-H type binder in both cases. Low or null reactivity of Fe-rich
phases (e.g. R-O, magnetite) is confirmed by the region of concentrated Fe. Elemental compositions of
AN
the two binders obtained by WDS microanalysis are provided in Table 5. Test on the mean values using
a level of significance of 0.05 resulted in rejection of the null hypothesis (actual values for SAT1 equal to
those obtained for SAT2) for the silica (P-value= 0.036), and the alumina content (P-value= 0.027),
M

providing strong evidence of the lower silica and higher alumina presence in the SAT2-11 binding matrix
when compared to SAT1-11. WDS results detected the presence of Fe into the binder to some extent,
this suggests that the generated binder could belong to a CaO-FeO-SiO2 inorganic polymer type [37] or a
D

C-S-H type with Ca-Fe substitutions [38][39]. However, the exact identification of the type of binder is
outside the scope of the present work and is proposed for future research.
TE

(a) (b)
C EP
AC

Figure 12. BSE-SEM images of binder matrix generated by mixture (a) SAT1-11 and (b) SAT2-11
ACCEPTED MANUSCRIPT

Table 5. EPMA-WDS microanalysis of binding matrices generated for SAT1-11 and SAT2-11

Binder ID CaO SiO2 Total Fe Al2O3 MgO MnO TiO2 Na2O SO3 P2 O 5
SAT1-11 27.5 17.6 12.7 8.1 8.9 1.2 1.0 0.6 1.3 1.1
St. Dev. SAT1-11 1.5 0.6 1.9 2.5 5.3 0.3 4.6 0.2 0.7 0.4
SAT2-11 31.3 15.7 12.4 13.4 2.3 1.4 0.6 1.1 1.0 1.3
St. Dev. SAT2-11 0.3 1.2 2.2 2.9 2.8 0.3 1.1 3.2 0.5 0.1

PT
RI
U SC
AN
M
D

Figure 13. EDS elemental mapping of SAT1-11 binder (scale bar length is 5 µm).
TE
C EP
AC

Figure 14. EDS elemental mapping of SAT2-11 binder (scale bar length is 5 µm).
ACCEPTED MANUSCRIPT

4. Conclusions

In this study, the microstructural engineering of an industrial (non-hydraulic) BOF slag generated a
material with cementitious properties. Two slag modifications (SAT1 and SAT2), defined by
thermodynamic modeling, were produced using the addition of silica (5 wt% in SAT1 and 0.6 wt% in
SAT2) and alumina (11 wt% in SAT1 and 12.8 wt% in SAT2), the re-melting of the mix, and the fast
cooling of the melt through granulation. These modifications provided binder precursors with a hybrid

PT
mineralogy composed of cementitious- and non-cementitious-crystalline phases, and a reactive
amorphous fraction. The latter representing 65.5 wt% of SAT1 and 55.5 wt% of SAT2. Considering the
cementitious phases, lower amounts of dicalcium silicate were measured in SAT1 (15.5 wt%) than in

RI
SAT2 (23.0 wt%), whereas a similar quantity of calcium aluminoferrite phases (4.1 wt%) was identified in
both slags, this phase was richer in Fe for SAT2. The amorphous composition followed the overall
compositional change of the slag with localized variations in regions adjacent to crystals.

SC
The fresh and hardened properties of the modified slags varied under alkali activation. The viscosity of
the fresh mixes with SAT2 was almost twice that of those with SAT1, and the 28-day mechanical
compression of SAT2 was either 10 or 30 % higher than that of SAT1, depending on the mixture used for

U
the activation. The binder, generated using a 0.25 M NaOH solution, showed similar crystalline reaction
products for both slags indicating resemblances in the activation paths. Yet, the main reaction product,
AN
an amorphous C(F,N)-A-S-H binder, was richer in alumina for SAT2. The binder composition was also
affected by the contribution of crystalline phases consumed during activation.

The results are promising enough to promote a broader mapping of BOF modifications and the study of
M

other important properties (like durability, shrinkage, or leaching of problematic compounds). The
proposed slag processing can be easily integrated into current industrial practice: the additions can be
injected directly into the slag ladle (avoiding any re-melting step), and the fast cooling implemented
D

through granulation facilities at the slag yard. The collection of data from industrial tests will open the
door to the actual tailoring of slags, providing a binder catalog able to meet specific needs of the slag
TE

end-users, and bringing the steel industry a step closer to full slag valorization.
EP

Acknowledgments

This work was part of the activities executed for the grant 140514 of the Flemish Institute for the
Promotion of Innovation by Science and Technology (IWT). Pavel Lopez acknowledges the financial
C

support of the Colombian Administrative Department of Science, Technology, and Innovation


(COLCIENCIAS).
AC

References

[1] Monteiro PJM, Miller SA, Horvath A. Towards sustainable concrete. Nat Mater 2017;16(7):698–
9.
[2] Juenger MCG, Winnefeld F, Provis JL, Ideker JH. Advances in alternative cementitious binders.
Cement and Concrete Research 2011;41(12):1232–43.
[3] World steel association. World steel in figures 2017. [May 10, 2017].
ACCEPTED MANUSCRIPT

[4] Mahieux P-Y, Aubert J-E, Escadeillas G. Utilization of weathered basic oxygen furnace slag in the
production of hydraulic road binders. Construction and Building Materials 2009;23(2):742–7.
[5] Ionescu DV. The hydraulic potential of high iron bearing steel slags. The University of British
Columbia; 1999.
[6] Murphy JN, Meadowcroft TR, Barr PV. Enhancement of the Cementitious Properties of
Steelmaking Slag. Canadian Metallurgical Quarterly 2013;36(5):315–31.
[7] Belhadj E, Diliberto C, Lecomte A. Characterization and activation of Basic Oxygen Furnace slag.

PT
Cement and Concrete Composites 2012;34(1):34–40.
[8] Ferreira Neto JB, Fredericci C, Faria JOG, Chotoli FF, Ribeiro TR, Malynowskyj A et al.
Modification of Basic Oxygen Furnace Slag for Cement Manufacturing. Journal of Sustainable

RI
Metallurgy 2017.
[9] Gautier M, Poirier J, Bodénan F, Franceschini G, Véron E. Basic oxygen furnace (BOF) slag
cooling: Laboratory characteristics and prediction calculations. International Journal of Mineral

SC
Processing 2013;123:94–101.
[10] Reddy AS, Pradhan RK, Chandra S. Utilization of Basic Oxygen Furnace (BOF) slag in the
production of a hydraulic cement binder. International Journal of Mineral Processing

U
2006;79(2):98–105.
[11] Euroslag. Statistics 2016. [October 12, 2018]; Available from:
AN
http://www.euroslag.com/products/statistics/2016/.
[12] Guo J, Bao Y, Wang M. Steel slag in China: Treatment, recycling, and management. Waste
Manag 2018;78:318–30.
M

[13] Shi C. Steel Slag—Its Production, Processing, Characteristics, and Cementitious Properties.
Journal of Materials in Civil Engineering 2004;16(3):230–6.
[14] Shi C, Qian J. High performance cementing materials from industrial slags — a review.
D

Resources, Conservation and Recycling 2000;29(3):195–207.


[15] Juckes LM. The volume stability of modern steelmaking slags. Mineral Processing and Extractive
TE

Metallurgy 2013;112(3):177–97.
[16] Tossavainen M, Engstrom F, Yang Q, Menad N, Lidstrom Larsson M, Bjorkman B. Characteristics
of steel slag under different cooling conditions. Waste Manag 2007;27(10):1335–44.
[17] Duxson P, Provis JL. Designing Precursors for Geopolymer Cements. Journal of the American
EP

Ceramic Society 2008;91(12):3864–9.


[18] Provis JL, van Deventer JSJ. Alkali activated materials: State-of-the-art report, RILEM TC 224-
AAM / John L. Provis, Jannie S.J. van Deventer, editors. Dordrecht: Springer; 2014.
C

[19] Liu C, Guo M, Pandelaers L, Blanpain B, Huang S. Stabilization of Free Lime in BOF Slag by
Melting and Solidification in Air. Metall and Materi Trans B 2016;47(6):3237–40.
AC

[20] Zhuangzhuang Liu. Effects of solids on the rheology of heterogeneous silicate melts [PhD Thesis].
Leuven, Belgium: KU Leuven; 2017.
[21] Lopes M, Shelby JE. Introduction to Glass Science and Technology (2nd ed.). Royal Society of
Chemistry; 2005.
[22] Liu C, Huang S, Wollants P, Blanpain B, Guo M. Valorization of BOF Steel Slag by Reduction and
Phase Modification: Metal Recovery and Slag Valorization. Metall and Materi Trans B
2017;48(3):1602–12.
[23] Pacheco-Torgal F (ed.). Handbook of alkali-activated cements, mortars and concretes.
Cambridge, England: Elsevier; Woodhead Publishing; 2015.
ACCEPTED MANUSCRIPT

[24] Shi C, Krivenko PV, Della Roy. Alkali-activated cements and concretes. Boca Raton: Crc Press;
Taylor and Francis Group; op. 2015.
[25] Goos P, Jones B. Optimal Design of Experiments: A Case Study Approach. Wiley; 2011.
[26] Durinck D, Jones PT, Blanpain B, Wollants P, Mertens G, Elsen J. Slag Solidification Modeling
Using the Scheil-Gulliver Assumptions. J American Ceramic Society 2007;90(4):1177–85.
[27] Novais RM, Ascensão G, Buruberri LH, Senff L, Labrincha JA. Influence of blowing agent on the
fresh- and hardened-state properties of lightweight geopolymers. Materials & Design

PT
2016;108:551–9.
[28] Shelby JE. Introduction to glass science and technology. Cambridge: Royal Society of Chemistry;
1997.

RI
[29] Zarzycki J. Glasses and the vitreous state; 1991.
[30] Adhikari P, Dharmawardhana CC, Ching W-Y. Structure and properties of hydrogrossular mineral
series. J Am Ceram Soc 2017;100(9):4317–30.

SC
[31] Blanc P, Bourbon X, Lassin A, Gaucher EC. Chemical model for cement-based materials:
Thermodynamic data assessment for phases other than C–S–H. Cement and Concrete Research
2010;40(9):1360–74.

U
[32] Fernández-Jiménez A, Vázquez T, Palomo A. Effect of Sodium Silicate on Calcium Aluminate
Cement Hydration in Highly Alkaline Media: A Microstructural Characterization. Journal of the
AN
American Ceramic Society 2011;94(4):1297–303.
[33] Geiger CA, Dachs E, Benisek A. Thermodynamic behavior and properties of katoite
(hydrogrossular): A calorimetric study. American Mineralogist 2012;97(7):1252–5.
M

[34] Ectors D, Neubauer J, Goetz-Neunhoeffer F. The hydration of synthetic brownmillerite in


presence of low Ca-sulfate content and calcite monitored by quantitative in-situ-XRD and heat
flow calorimetry. Cement and Concrete Research 2013;54:61–8.
D

[35] Ghosh SN, Handoo SK. Infrared and Raman spectral studies in cement and concrete (review).
Cement and Concrete Research 1980;10(6):771–82.
TE

[36] Gervais F, Blin A, Massiot D, Coutures JP, Chopinet MH, Naudin F. Infrared reflectivity
spectroscopy of silicate glasses. Journal of Non-Crystalline Solids 1987;89(3):384–401.
[37] Peys A, White CE, Olds D, Rahier H, Blanpain B, Pontikes Y. Molecular structure of CaO–FeO x –
SiO 2 glassy slags and resultant inorganic polymer binders. J Am Ceram Soc 2018;101(12):5846–
EP

57.
[38] Richardson IG, Groves GW. The incorporation of minor and trace elements into calcium silicate
hydrate (C-S-H) gel in hardened cement pastes. Cement and Concrete Research 1993;23(1):131–
C

8.
[39] Le Callonnec C, Faucon P, Bonville P, Genand-Riondet N, Jacquinot JF. Superparamagnetic
AC

properties of a cement-derived synthetic Fe-substituted calcium silicate hydrate. Journal of


Magnetism and Magnetic Materials 1997;171(1-2):119–28.

You might also like