Waste Glass Powder as Cement Replacement in Concrete

Hongjian Du, Kiang Hwee Tan

Journal of Advanced Concrete Technology, volume 12 ( 2014 ), pp. 468-477

Effects of PFA and GGBS on Early-Ages Engineering Properties of Portland Cement Systems
Xiangming Zhou , Joel R. Slater, Stuart E. Wavell, Olayinka Oladiran
Journal of Advanced Concrete Technology, volume 10 ( 2012 ), pp. 74-85

Estimation of strength, permeability and hydraulic diffusivity of pozzolana blended concrete through
pore size distribution
B. Kondraivendha , B. Sabet Divsholi , Susanto Teng
Journal of Advanced Concrete Technology, volume 11 ( 2013 ), pp. 230-237

Effect of nano-CaCO3 on compressive strength development of high volume fly ash mortars and
concretes
Steve W. M. Supit, Faiz U. A.Shaikh
Journal of Advanced Concrete Technology, volume 12 ( 2014 ), pp. 178-186
Journal of Advanced Concrete Technology Vol. 12, 468-477, November 2014 / Copyright © 2014 Japan Concrete Institute 468

Scientific paper

Waste Glass Powder as Cement Replacement in Concrete

Hongjian Du1* and Kiang Hwee Tan2

Received 25 August 2014, accepted 2 November 2014 doi:10.3151/jact.12.468

Abstract
The pozzolanic reactivity of waste glass powder was experimentally studied at cement replacement levels of 0, 15, 30,
45 and 60% by weight. Results revealed that the concrete compressive strength was not decreased by the cement substi-
tution after 28 days because of the pozzolanic reaction between glass powders and cement hydration products, if the
replacement is below 30%. Also, the resistance to chloride ion and water penetration continuously increases with in-
creasing glass powder content up to 60% cement replacement. At 60% replacement level, the electrical resistivity and
water penetration depth were reduced by 95% and 80%, respectively, while the compressive strength was maintained as
85%. These improvements in durability properties are due to the refined microstructures, particularly at the interfacial
transition zone. Pore size distribution was measured to confirm the refinement in the capillary pores, which partially
block the pathways for water and chloride ions. This study also demonstrates that high performance concrete (improved
strength and impermeability against chloride and water) could be achieved by using glass powder as 15% additive,
which contributes to the pozzolanic reaction instead of being inert fines for compact packing.

1. Introduction pansion in mortar containing glass particle as fine ag-

Due to the awareness on the need for environmental
protection, turning solid wastes or by-products into in-
gredients of concrete has drawn increasing attention.
Apart from material and energy conservation, reuse of
some solid wastes could result in better performances of
concrete in several areas. For example, use of waste
glass as fine aggregates provides concrete with higher
resistance to chloride penetration (Ling et al. 2011; Tan
and Du 2013; Du and Tan 2014a). Fine glass particles
(smaller than 75 µm) can exhibit pozzolanic reactivity,
thereby improving the microstructure of paste and the
strength and durability of concrete in the long term
(Wright et al. 2014; Du and Tan 2014a). The amorphous
silica in the glass would dissolve in an alkaline envi-
ronment such as due to OH- ions in the pore solution of
cement paste. Thereafter, it could react with calcium
hydroxide (CH) to form secondary calcium silicate hy-
drate (C-S-H), a process known as pozzolanic reaction
which can be expressed by CH + S + H → C-S-H. Pre-
vious literature shows that the pozzolanic reactivity of
glass particle increases due to the larger surface area
available for the reaction (Shayan and Xu 2004; Du and
Tan 2014b). At the same time, with finely ground glass
powder (GP), no alkali-silica reaction has been observed
(Jin et al. 2000). According to the authors’ previous
work (Du and Tan 2013; Du and Tan 2014b), ASR ex-
1
Research Fellow, Department of Civil and Environ-
mental Engineering, National University of Singapore,
Singapore.
*Corresponding author, E-mail: ceedh@nus.edu.sg
2
Professor, Department of Civil and Environmental
Engineering, National University of Singapore,
Singapore.
gregates would be negligible provided that the glass
particles were smaller than 0.3 mm. This finding is con-
sistent with the threshold particle size of 0.3 mm as re-
ported by Shayan and Xu (2004) and Rajabipour et al.
(2010).
Some studies have been conducted to characterize the
pozzolanic activity of finely ground GP in concrete
(Shao et al. 2000; Dyer and Dhir 2001; Shi et al. 2005;
Neithalath 2008; Schwarz et al. 2008; Jain and Neitha-
lath 2010; Matos and Sousa-Coutinho 2012; Khmiri et
al. 2013; Carsana et al. 2014; Kim et al. 2014; Mirza-
hosseiniand and Riding 2014; Du and Tan 2014c). De-
pending on its fineness, GP would generally exhibit
pozzolanic reaction at a slower rate compared with the
cement hydration. Thus, the replacement of cement by
GP might decrease strength at an early age but would
increase it at a later age. Most of the previous research
works emphasized on the pozzolanic reaction of GP on
the mechanical characteristics of mortar/concrete with a
fixed glass powder to the binder ratio. Shao et al. (2000)
found that concrete with 30% glass powder finer than
38 µm did not exhibit higher compressive strength until
after 90 days of curing. Shi et al. (2005) also observed
that mortar with 20% glass powder could give a higher
compressive strength from 28 days onwards, provided
that glass powders are close to or smaller than Portland
cement in size. A similar conclusion has been recently
reported by Khmiri et al. (2013). Shayan and Xu (2004)
reported that the use of glass powder (smaller than 10
µm) as cement replacement up to 40% can reduce the
ASR expansion for concrete. However, all the concrete
mixes with cement partially replaced by GP showed
continuously reduced compressive strength. Also, the
influence of GP as a supplementary cementitious mate-
rial on the hydration and the microstructure of paste has
H. Du and K. H. Tan / Journal of Advanced Concrete Technology Vol. 12, 468-477, 2014 469

rarely been reported (Shi et al. 2005). They found that Table 1 Chemical compositions of OPC and glass
curing temperature have a more obvious accelerating powder.
effect on the pozzolanic reactivity of glass powder than Composition, % OPC Glass powder
on that of fly ash. Both the hydration and pozzolanic SiO2 20.8 72.08
reaction greatly govern the structure formation and Al2O3 4.6 2.19
hence all properties like strength development and du- Fe2O3 2.8 0.22
rability. CaO 65.4 10.45
As reviewed above, the properties of concrete with MgO 1.3 0.72
glass powder as cement replacement in higher propor- SO3 2.2 —
Na2O 0.31 13.71
tion remains an unexplored area. Therefore, this re-
K2O 0.44 0.16
search aims to study the pozzolanic reaction of GP up to
TiO2 — 0.1
60% cement replacement and its influence on the micro-
Cr2O3 — 0.01
structure of cement paste, which has not been investi-
gated before. The hydration rate and rheological proper- Table 2 Mix proportion of concrete with different glass
ties of paste were also characterized. The chemical powder replacement levels.
composition and calcium hydroxide content of the hy- Content, kg/m3
drated products were analyzed at different ages. Also, Mix No. Glass Coarse
Water Cement Sand
the microstructure development of cement paste was Powder aggregate
determined by measuring the pore size distribution. At OPC 185 380 0 960 825
the same time, the compressive strength and resistance 15GP 185 323 57 955 825
to chloride and water penetration were investigated for 30GP 185 266 114 947 825
concrete with the same cement replacement levels as the 45GP 185 209 171 940 825
60GP 185 152 228 933 825
paste. The influence of pozzolanic reaction on the tran-
OPC+15GP 185 380 57 907 825
sition zone between aggregate particles and cement
paste is discussed. Based on the results, the optimum
cement replacement content by glass powder is sug- GP is amorphous and its X-ray diffraction (XRD) pat-
gested. tern is compared with OPC in Fig. 3. GP has a negligi-
ble water absorption capacity of 0.07%. Coarse aggre-
2. Experimental program gate with a maximum size of 10 mm was used in con-
crete. The fineness modulus of natural sand was 2.80.
2.1 Materials
The crushing process of recycled waste glass can be 2.2 Specimens
found in the authors’ previous work (Tan and Du 2013; The microstructure of cement paste with various con-
Du and Tan 2014a, 2014b). Waste bear bottles (soda tents of GP was investigated. The water to cement (w/c)
lime glass) were collected from a local recycler in Sin- ratio for the paste is 0.485 by weight. The mix propor-
gapore. To finely grind the sand-sized particles, a ball tion for concrete is listed in Table 2. Cement was re-
miller was used. The size distributions of ground GP placed by GP at 0, 15, 30, 45 and 60% by weight. An
and cement used in this study are shown in Fig. 1. Both additional mix (OPC+15GP) was cast in which the GP
cement and GP show the same median particle size of was used as an additive, instead of replacing cement at a
around 10 µm. The chemical compositions of GP and content of 15%. Compared to the use of cement re-
OPC are displayed in Table 1. The specific gravities of placement, the addition of GP in this mix leads to a
cement and GP are 3.15 and 2.53, respectively. The sur- lower water-to-binder (w/b) ratio of 0.42. This could
face appearance of GP and OPC are compared in Fig. 2. help examine the utilization of GP as additive in high
performance cementitious composites. For each paste
100
mix, three 50 mm cube specimens were prepared. For
each concrete mix, nine 100 mm cubes and five
Φ100×200 mm cylinders were cast for compressive
80
Percentage passed, %

strength and durability tests, respectively. Paste and

concrete were mixed in Hobart and pan mixers respec-
60 tively. All the specimens were compacted on a vibration
table and covered by plastic sheets to prevent moisture
40 loss. After demolding on the next day, the paste speci-
mens were cured in saturated lime water and concrete
20 samples were cured in water until the test age. The cur-
OPC ing method for paste is following ASTM C 109 (2005)
Glass powder
0 to prevent calcium leaching and curing method for con-
0.1 1 10 100 crete specimens is following ASTM C 39 (2005).
Size, μm
Fig. 1 Particle size distribution of OPC and glass powder.
H. Du and K. H. Tan / Journal of Advanced Concrete Technology Vol. 12, 468-477, 2014 470

(a) OPC particles.

(b) Finely ground glass powder.

Fig. 2 Particle morphologies using SEM.
F

A=alite, C3S
B,
A,

B=belite, C2S
Alu=aluminate, C3A
B
A,

F=ferrite, C4AF
B

A
A,

G=gypsum, CaSO4
P: Periclase, MgO
B
A,
u
Al

G
A,

A F A A A
u

G
Al

A
B
G

P F
F

G
A,

A AB A,B
B,

A,

A
F,

OPC

GP

10 15 20 25 30 35 40 45 50
2θ
Fig. 3 XRD patterns for OPC and GP.

2.3 Test methods
The rheological properties of cement paste containing
GP were determined by a coaxial-cylinder viscometer
named RotoVisco 1. After mixing the cement paste in
the Hobart mixer for about 3 minutes, a sample weigh-
ing about 150 g was taken out and placed in the outer
cylinder of the viscometer, followed by inserting the
inner cylinder immediately. The sensor then started to
rotate at different shear rate while the torque was re-
corded. Isothermal calorimetry was performed on tripli-
cate paste samples by using an eight-channel microcalo-
rimeter (TAM AIR).
Paste samples were taken from the center of 50 mm
cube specimens for the purpose of mercury intrusion
porosimetry (MIP), XRD and thermogravimetric analy-
sis (TGA). For MIP test, paste samples (around 1 cm in
cubic shape) were dried in dessicator at 50 ˚C for 1 day
before testing, at the age of 7 and 91 days. Micromerit-
ics AutoPore III with a maximum mercury pressure of
412.5 MPa was used to measure the pore size distribu-
tion. For XRD and TGA analyses, paste samples were
cured in 105 ˚C oven for 24 hours and then finely
ground to be less than 75 µm. Shimadzu XRD-6000
diffractometer was employed to qualitatively determine
the influence of GP on the chemical composition of the
hydrated production. The XRD scan was between 10˚
and 60˚ with a speed of 0.5˚/min. TGA was carried out
using LINSEIS L81-II. The powder sample was heated
from ambient temperature up to 950˚C at a rate of
10˚C/min while the weight loss was recorded during the
TGA test. The content of CH could be obtained from
two different intervals in the weight loss curve, corre-
sponding to the decomposition of Portlandite and calcite,
respectively.
H. Du and K. H. Tan / Journal of Advanced Concrete Technology Vol. 12, 468-477, 2014 471

For concrete specimens, compressive strength and τ = τ 0 + μγ (1)

rapid chloride penetration test (RCPT) were carried out
according to BS EN 12390-3 (2009) and ASTM C 1202 The values of τ0 and µ for pastes with various GP
(2005), respectively. RCPT is commonly criticized for contents are displayed in Fig. 4(b). Both the yield stress
its drawbacks including temperature rise due to the ap- (τ0) and viscosity (µ) decrease with higher cement re-
plied high voltage of 60 V (McGrath and Hooton 1999). placement level. Yield stress indicates the minimal
Thus, a reduced voltage of 20 V was applied for 24 strength to make the mixture flowable. The reduced
hours, to minimize the side effects of RCPT. At 28 days, yield stress implies that the inter force between cement
water penetration depth into concrete was determined and glass particles is less than that between cement and
using a water pressure of about 0.75 MPa for 7 days, in cement particles. With increasing GP contents, the parti-
accordance to BS EN 12390-8 (2009). The side and top cle density of cement is diluted and hence lesser interac-
surfaces were both coated with impermeable epoxy tion between cement and water, leading to a smaller
while the bottom surface was exposed to the water pres- yield stress and plastic viscosity. This could also be due
sure. After removed from the test rig, the concrete cyl- to the negligible water absorption and the smooth sur-
inders were axially split and the average depth water face of GP (as shown in Fig. 2). Previous studies also
front measured. indicate that the bond between cement paste and fine
glass particles was decreased due to the surface smooth-
3. Test results and discussion ness of glass powder (Taha and Nounu 2009; Ali and
Al-Tersawy 2012).
3.1 Rheological properties In this study, the cement has been replaced by weight
The shear stress (τ) was recorded with shear rate ( γ ) instead of by volume. As the specific gravity of glass
decreasing from 50 to 0.5 s-1 during the descending power is lower than cement, the solid-to-water ratio by
branch of the test loop. Test results are shown in Fig. volume is higher for GP blended paste compared with
4(a). Yield stress (τ0) and plastic viscosity (µ) of fresh pure cement paste. However, this adverse effect at a
cement paste are linearly related using Bingham model higher solid-to-water volume ratio is less pronounced
(Mindess et al. 2003). compared to the dilution of cement and smooth glass
surface, as mentioned earlier. The continuously de-
100 creased yield stress and plastic viscosity represent a
OPC better workability of paste mixture with higher content
15GP
30GP of GP. It is also noted that both the yield stress and vis-
80
cosity increase substantially for the mix with 15% addi-
Shear stress, τ (Pa)

45GP
60GP
OPC+15GP μ tional GP, attributed to the increased amount of solid-to-
60
1 water ratio.

40 3.2 Heat of hydration

The heat evolution rate and cumulative amount of heat
20 for pastes with varying contents of GP are shown in Fig.
5. It is clear that the maximum heat evolution rate and
0 the total heat generated reduced continuously with
0 10 20 30 40 50 higher OPC replacement level because of the dilution of
Rate of shear, γ (1/s) cement and the slower rate of pozzolanic reaction of GP.
(a) Relationship between shear stress and shear rate for cement The results in this study are consistent with previous
paste with varying contents of glass powder. findings (Dyer and Dhir 2001; Mirzahosseini and Rid-
ing 2014). The benefit of lower hydration heat is helpful
30 1.5
in preventing the temperature cracking, especially for
Shear yield stress, τ0 (Pa)

Plastic viscosity, μ (Pa•s)

25 structure members with large thickness and mass con-

1.2 crete.
τ0
20 In the presence of GP, the time to reach the peak hy-
0.9 dration rate is shortened, possibly because fine glass
15 powders can accelerate the cement hydration via the
0.6 adsorption of calcium ions from the liquid phase and
10 play as nucleation and growth sites for C-S-H and other
μ 0.3 hydrates. At the same time, the high content of alkalis
5
(Na2O) in GP may act as catalyst in the formation of
0 0.0 calcium silica hydrate at an early age (Jawed et al.
0 15 30 45 60 15%
75Addition 1978; Khmiri et al. 2013). The time corresponding to
GP content, % the peak rate reduces from 418 minutes for OPC paste
(b) Glass powder effect on the rheological parameters. to 377 minutes for paste with 60% glass powder paste.
Fig. 4 Test results of rheological properties of paste.
H. Du and K. H. Tan / Journal of Advanced Concrete Technology Vol. 12, 468-477, 2014 472

6
Rate of heat evolution, mW/g (a) OPC 300 (b) OPC
5 15GP

Cumulative heat, J/g

30GP 250 15GP
45GP OPC+15GP 30GP
4 60GP
200 45GP
OPC+15GP
3 150 60GP

2 100

1 50

0 0
0 12 24 36 48 60 72 0 12 24 36 48 60 72
Time, h Time, h
Fig. 5 (a) Rate of heat evolution, and (b) cumulative heat of cement paste with various replacement levels with glass
powder.

(a) 60% glass powder in which the CH seems to be insuffi-

91-day cient for glass powder to be fully reacted.
3.4 Ca(OH)2 content
28-day The CH content was calculated from the TG and DTG
CH

curves, as illustrated in Fig. 7(a). Derivative curve was

CH
CSH
CSH

CH
CH

7-day used to find the beginning and the end of the derivative
10 15 20 25 30 35 40 45 50 peak and measure the weight loss correspondingly (Ko-
2θ caba 2009). In the temperature range of about 425 and
(b) 550 ˚C, the amount of CH can be calculated by deter-
91-day mining the water loss during the CH decomposition:
Ca(OH)2→CaO+H2O. Also, the influence of carbona-
28-day tion on the CH content could be evaluated by the mass
loss in the temperature range of 650 and 800 ˚C, where
CH

CH

the calcite decomposes and release CO2:

CSH
CSH

CH
CH

7-day
CaCO3→CaO+CO2. Hence, the content of CH was cal-
10 15 20 25 30 35 40 45 50
2θ
culated taking into account molecular weight of each
component and were calculated as percentage of the dry
(c) weight at 105 ˚C.
91-day
As shown in Fig. 7(b), the CH content in the refer-
ence OPC paste remained at above 20% of the total
28-day weight after 28 days of curing, which is similar to pre-
vious findings (Jawed et al. 1978). With higher cement
CH

CSH

CSH

CH

substitution level, the CH content drops, especially at

CH

CH

7-day

10 15 20 25 30 35 40 45 50 later age such as 91 days. At the beginning, the reduc-

2θ tion is caused by the dilution of Portland cement in the
Fig. 6 XRD patterns of cement paste with different con- total cementitious material content. With longer curing
tents of glass: (a) OPC, (b) 30GP and (c) 60GP. and hydration time of 28 days, more CH is consumed in
the pozzolanic reaction of glass powder. The initial per-
3.3 XRD centage drop in CH content was higher for 30% GP
The XRD patterns for paste with glass powder at age of paste than in 60% GP paste. After 28 days, the reduction
7, 28 and 91 days are shown in Fig. 6. The compositions in CH content slowed down for 30% glass powder paste
of the hydration product for Portland cement paste are and stabilized at about 11%, indicating the pozzolanic
C-S-H and CH and their principal peaks do not change reaction was almost completed. In contrast, the CH con-
from 7 to 91 days. CH peaks tend to be weakened with tent in 60% GP paste continued to drop and was less
increasing glass powder and longer curing time, espe- than 4% at 91 days. This depletion in CH reveals that
cially for 60% glass powder paste at 91 days in which 60% might be the maximum practical glass powder con-
the CH peaks almost disappear. The XRD results di- tent since the CH content is insufficient for the further
rectly reveal that CH is consumed to form the additional pozzolanic reaction of glass powder. Dyer and Dhir
C-S-H in paste with glass powders. (2001) has reported that the amount of CH continuously
At 91 days, peaks corresponding to CH can be obvi- decreased with increasing GP replacement up to 40%, at
ously seen for OPC and 30GP paste, which means that the age of 28 days, but no information for higher GP
CH remains even after 91 days of pozzolanic reaction replacement.
with glass powder. However, it is not true for paste with
H. Du and K. H. Tan / Journal of Advanced Concrete Technology Vol. 12, 468-477, 2014 473

100 0.00 25

95 OPC
-0.02

DTG, dm/dT (mg/°C)

20
TG weight loss, %

90

% Ca(OH)2
85 -0.04 15
Portlandite 30GP
80 -0.06 10
75 Calcite
-0.08 60GP
5
70
(a) (b)
65 -0.10 0
200 300 400 500 600 700 800 900 0 7 28 56
91 84 91
91
Temperature, °C Curing time, days
Fig. 7 (a) TG and DTG curves for OPC paste at 28-day and (b) Portlandite content at different ages.

3.5 Microstructure 100

The pore size distribution of cement paste containing 7-day
80
glass powder at 7 and 91 days are shown in Fig. 8, from
which the change in the microstructure of paste can be 60

% of total intrusion volume

seen clearly particularly for paste with 30 and 60% glass
powder. The pore structures become more refined, as 40
indicated by the reduced threshold and critical pore di-
ameter, as well as the higher fraction of the micro-pores. 20
For paste with 60% glass powder, the threshold and
0
critical pore diameter was reduced from 488 and 140 100
nm at 7 days to 139 and 74 nm at 91 days, respectively. 90-day
Moreover, the volume fraction of capillary pores (10 µm 80
– 50 nm) in the total porosity (10 µm – 3.7 nm) was
60
lowered from 44.6% at 7 days to 30.7% at 91 days. Ac-
OPC
cording to Mindess et al. (2003), it is the capillary pores 40 30GP
that govern the permeability and diffusivity of paste. 60GP
OPC+15GP
Thus the MIP results reveal that the permeability of high 20
volume glass powder paste is likely to be reduced. Poz-
zolanic reaction between glass powder and CH is a 0
1E-3 0.01 0.1 1 10 100
slower chemical process compared with OPC hydration. Pore size, μm
Since glass powder is inert in the initial stage, the effec-
Fig. 8 MIP curves for paste containing different contents
tive w/c ratio is actually higher than that of the OPC
of glass powder at 7 and 91 days.
paste. This results in a more porous microstructure for
mix with glass powder as OPC replacement. However, 70
at later age, most of the glass powders in 30GP and part
Compressive strength, MPa

of it in 60 GP mix have reacted with lime to form the 60

secondary C-S-H. The pore structures are almost identi- 91-day
50
cal for each paste mix, regardless of glass powder con-
tent at 91 days. 40
28-day
30
3.6 Compressive strength of concrete
7-day
The compressive strengths of concrete with varying 20
contents of GP, at different ages, are shown in Fig. 9. At
7 days, the strength generally decreases with GP content 10
more than 15%. However, at 28 and 91 days, no reduc- 0
tion in strength was observed for concrete with 15 and 0 15 30 45 60 15%
75Addition
30% GP, due to the pozzolanic reaction between GP and GP content, %
cement hydration products. Instead, concrete with 15 Fig. 9 Compressive strength of concrete with glass pow-
and 30% GP possessed higher compressive strengths. ders at different ages.
This increase in later age strength was not observed for
concrete in which more than 30% cement is replaced by content. Therefore, there is an upper limit for cement
GP. The pozzolanic reaction requires the hydration replacement level, beyond which no further pozzolanic
products, CH, whose amount is governed by the cement reaction of GP can occur. In that case, GP can only play
H. Du and K. H. Tan / Journal of Advanced Concrete Technology Vol. 12, 468-477, 2014
the role of inert filler without being activated. The re-
sults in this study indicate that GP will exhibit obvious
pozzolanic reaction provided that the replacement level
is lower than 30%. Mix of OPC+15GP exhibited much
higher increase in compressive strength, that is 32%,
25% and 24% at 7, 28 and 91 days, respectively. It
would be readily for the added fine GP to react as poz-
zolans instead of being inert fillers.
The pozzolanic reaction improves not only the pore
structure in the bulk cement paste but also at the interfa-
cial transition zone (ITZ) between coarse aggregates
and cement paste. This ITZ governs the mechanical and
transport properties of concrete since it is more porous
compared to the bulk paste. CH content is also relatively
higher at ITZ which favors the pozzolanic reaction of
glass powder. The enhanced microstructure at ITZ has
contributed to the compressive strength with up to 30%
GP content.
3.7 RCPT result of concrete
The total charge passing the concrete with different
OPC replacement levels are shown in Fig. 10, at 7, 28
and 91 days. Due to the low voltage applied in this study,
the temperature rise during the 24 hours test was less
than 3 ˚C. The RCPT results show that the total charge
passed decreased significantly with increasing glass
powder content, regardless of the test age. RCPT is es-
sentially an indication of the electrical conductivity,
instead of a direct measurement for chloride permeabil-
ity. Therefore, the result depends on not only the pore
structure but also the chemistry of the pore solution,
such as ions (Ca2+, Na+, K+, OH-, etc) concentrations.
However, it is the pore structure and not the pore chem-
istry that govern the transport of ions in cement com-
posites. Therefore, it is also questionable to use RCPT
to assess the permeability of concrete with supplemen-
tary cementitious materials such as silica fume and fly
ash (Shi et al. 1998).
At 7 days, the total charge passed reduced almost
linearly with decreasing OPC content. The primary rea-
son could be the dilution of OPC in the concrete, as a
result of which there are fewer ions in the cement paste.
6000
Total charge passed, Coulombs
5000
4000
3000
2000
1000
0 0
0 15 30 45
GP content, %
Fig. 10 RCPT results of concrete with glass powders at
different ages.
474
With longer curing time, all the concrete mixtures ex-
hibited reduction in the total charge passed due to the
further hydration of OPC. The reduction caused by glass
powder became more distinct, especially for OPC re-
placement between 15 and 45%. This further reduction
is attributed to the pozzolanic reaction between GP and
cement paste, resulting in a more refined microstructure
of secondary C-S-H especially at the ITZ as well as a
lower concentration of ions. Glass powder has a similar
effect on the RCPT result whether it adds to or replaces
OPC as cementitious material, particularly in the long
term. At an early age, concrete mix with 15% GP addi-
tion has a higher resistance possibly because of the hy-
dration acceleration effect of GP, as discussed earlier.
3.8 Water penetration resistance of concrete
The water penetration depth into concrete with different
contents of GP is shown in Fig. 11. Consistent with the
RCPT results, the depth of water penetration continu-
ously decreases with higher glass powder content. Rela-
tive to the plain concrete, the water penetration depth
was reduced by 54, 65, 68 and 80% for cement re-
placement level of 15, 30, 45 and 60%, respectively.
Concrete with 15% GP additive showed much lower
water penetration depth compared to the plain concrete.
The refined pore structure, particular the ITZ is the main
reason for this reduced permeability. The reduction in
the total charge passed in the RCPT tests at 28 days was
60, 81, 89 and 95% for 15, 30, 45 and 60% of glass
powder content, respectively. Contrary to RCPT results,
the resistance to water penetration can accurately reflect
the pore structure. Hence, both accelerated and non-
accelerated tests demonstrate the better resistance
against chloride ions and water ingress, for concrete
containing glass powder as cement replacement and
additive.
3.9 Interfacial transition zone in concrete
The microstructures of paste at ITZ are compared for
plain concrete and 30% GP concrete in Fig. 12. It is
clear to find CH (plate crystals) and ettringite (needle
shape) in the paste of the OPC concrete. Large pores
30
Water penetration depth, mm
7-day
28-day
91-day
20
10
60 15%
75Addition 0 15 30 45 60 15%
75 Addition
GP content, %
Fig. 11 Water penetration depth into concrete with glass
powder at 28 days.
H. Du and K. H. Tan / Journal of Advanced Concrete Technology Vol. 12, 468-477, 2014
0.12
Cumulative pore volume, mL/g
0.10
0.08
0.06
0.04
0.02
0.00
1E-3 0.01 0.1
Pore size, μm
Fig. 13 MIP results for GP concrete at the age of 91
days.
exist between different hydration phases. In contrast, a
dense and more homogenous structure is found for con-
crete with 30% glass powder. The pozzolanic reaction
over 91 days turned crystal CH into amorphous C-S-H.
Large pores were much less noticeable. This change in
the chemical compositions (more C-S-H formed) and
pore size distribution (refined pore system) is the reason
for the better mechanical and durability performances
for concrete with glass powder as OPC substitution.
The pore size distribution of the microstructure was
also investigated for concrete including the ITZ, as
shown in Fig. 13. It is clear that with OPC partially re-
placed by glass powder, the pores are more refined, in-
dicated by the downwards shift of the curve. From the
figure, it can be seen that the critical pore diameter be-
475
CH
C-S-H
ettringite
(a) OPC concrete.
C-S-H
(b) 30GP concrete.
Fig. 12 SEM images of paste microstructures.
comes smaller for glass powder concrete. Therefore, the
OPC
30GP
effect of glass powder pozzolanic reaction on the con-
crete is distinct in comparison to that on paste. The rea-
60GP
OPC+15GP son is attributed to the existence of ITZ where the con-
centration of CH and water was higher and it was there-
fore easier to react with glass powder to form secondary
C-S-H, as mentioned above.
4. Conclusions
This study reported experimental results on the use of
recycled glass powder as supplementary cementitious
1 10 100 material in paste and concrete. Unlike previous work,
this study examined the effects of a higher glass powder
dosage of up to 60%. Based on the newly obtained re-
sults, the following conclusions can be drawn:
1. The rate and total heat generated during hydration
consistently decreased with higher GP content due to
the dilution of cement in the mix. Owning to its neg-
ligible water absorption capability, higher glass pow-
der content results in a smaller shear yield stress and
plastic viscosity of the paste.
2. Calcium hydroxide content decreases with glass
powder content and curing age, because of the re-
duced cement content in the initial stage and its con-
sumption by glass powder pozzolanic reaction at the
later stage. Calcium hydroxide was almost depleted
at 91 days when more than 30% cement was substi-
tuted by glass powder.
3. An optimum cement replacement of 30% by glass
powder was observed with respect to the develop-
ment of compressive strength of concrete after 7 days.
H. Du and K. H. Tan / Journal of Advanced Concrete Technology Vol. 12, 468-477, 2014
However, with respect to the resistance to water and
chloride penetration, 60% seems to be the optimum
replacement. Compared to the reference concrete mix,
the mix with 60% glass powder exhibited 75%
strength at 7 days and 85% strength at 91 days.
4. Resistance to chloride ion and water penetration re-
sistance were greatly improved by replacing cement
with glass powder, due to the refined microstructure
of paste, particular at the ITZ where calcium hydrox-
ide concentration is higher and more readily available
for pozzolanic reaction. With 60% cement replace-
ment, the electrical resistance and water penetration
depth of concrete decreases to 5% and 20% of the
reference concrete.
5. When used as additional supplementary cementitious
material at 15% level, glass powder can obviously
reduce the porosity and the pore size distribution.
Thus, large increases in compressive strength, resis-
tance to water and chloride penetration were ob-
served.
This study only investigated finely ground glass
powder as cement replacement or additive in concrete
based on soda lime glass, which is the most common
type of glass and accounts for 90% of manufactured
glass. It is a open question as to whether the results
could apply to other types of glass like lead glass and
borosilicate glass. The pozzolanic reactivity depends on
the chemical composition and properties of the glass. To
understand the pozzolanic performance of a specific
glass powder, it is recommended to (1) carry out chemi-
cal composition analysis to meet the minimum chemical
requirement for pozzolans; (2) determine the strength
activity index; and (3) investigate the different replace-
ment levels on mechanical and durability performances
on concrete.
Acknowledgements
The technical assistance of Ms. Li Wei and Mr. Ang
Beng Oon from the Structural Engineering and Material
Laboratory, National University of Singapore, in the
conduct of TGA and XRD tests, is gratefully acknowl-
edged.
References
Ali, E. E. and Al-Tersawy, S. H., (2012). “Recycled
glass as a partial replacement for fine aggregate in
self compacting concrete.” Construction and Building
Materials, 35, 785-791.
ASTM C 39/C 39M, (2005). “Standard test method for
compressive strength of cylindrical concrete speci-
mens.” Philadelphia: American Society of Testing and
Materials.
ASTM C 109/C 109M, (2005). “Standard test method
for compressive strength of hydraulic cement mortars
(using 2-in. or [50-mm] cube specimens).” Philadel-
phia: American Society of Testing and Materials.
ASTM C 1202, (2005). “Standard test method for
electrical indication of concrete’s ability to resist
476
chloride ion penetration.” Philadelphia: American
Society of Testing and Materials.
BS EN 12390-8, (2009). “Testing hardened concrete
part 8: depth of penetration of water under pressure.”
London: British Standards Institution, London.
BS EN 12390-3, (2009). “Testing hardened concrete
Part 3: compressive strength of test specimens.”
London: British Standard Institution.
Carsana, M., Frassoni, M. and Bertolini, L., (2014).
“Comparison of ground waste glass with other
supplementary cementitious materials.” Cement and
Concrete Composites, 45, 39-45.
Du, H. and Tan, K. H., (2013). “Use of waste glass as
sand in mortar: Part II - alkali-silica reaction and
mitigation methods.” Cement and Concrete Compo-
sites, 35(1), 118-126.
Du, H. and Tan, K. H., (2014a). “Concrete with recycled
glass as fine aggregates.” ACI Materials Journal,
111(1), 47-58.
Du, H. and Tan, K. H., (2014b). “Effect of particle size
on alkali-silica reaction in recycled glass mortars.”
Construction and Building Materials, 66, 275-285.
Du, H. and Tan, K. H., (2014c). “Transport properties of
concrete with glass powder as supplementary
cementitious material.” ACI Materials Journal, ac-
cepted for publication.
Dyer, T. D. and Dhir, R. K. (2001). “Chemical reactions
of glass cullet used as cement component.” Journal of
Materials in Civil Engineering, 13(6), 412-417.
Helmuth, R., (1987). “Fly ash in cement and concrete.”
Illinois: Portland Cement Association.
Jain, J. A. and Neithalath, N., (2010). “Chloride
transport in fly ash and glass powder modified
concretes-influence of test methods on micro-
structure.” Cement and Concrete Composites, 32(2),
148-156.
Jawed, I. and Skalny, J., (1978). “Alkalis in cement: a
review II. Effects of alkalis on hydration and
performance of Portland cement.” Cement and
Concrete Research, 8, 37-51.
Jin, W., Meyer, C. and Baxter, S., (2000) “Glascrete-
concrete with glass aggregate.” ACI Materials
Journal, 97(2), 208-213.
Khmiri, A., Chaabouni, M. and Samet B., (2013).
“Chemical behavior of ground waste glass when used
as partial cement replacement in mortars.” Construc-
tion and Building Materials, 44, 74-80.
Kim, J., Moon, J. H., Shim, J.W., Sim, J., Lee, H.G. and
Zi, G., (2014). “Durability properties of a concrete
with waste glass sludge exposed to freeze-and-thaw
condition and de-icing salt.” Construction and
Building Materials, 66, 398-402.
Kocaba, V., (2009). “Development and evaluation of
methods to follow microstructural development of
cementitious systems including slags.” PhD thesis.
Ecole Polytechnique Federale de Lausanne.
Ling, T. C., Poon, C. S. and Kou, S. C., (2011).
“Feasibility of using recycled glass in architectural
H. Du and K. H. Tan / Journal of Advanced Concrete Technology Vol. 12, 468-477, 2014
cement mortars.” Cement and Concrete Composites,
33(8), 848-854.
Matos, A. M. and Sousa-Coutinho, J., (2012).
“Durability of mortar using waste glass powder as
cement replacement.” Construction and Building
Materials, 36, 205-215.
McGrath, P. F. and Hooton, R. D., (1999). “Re-
evaluation of the AASHTO T259 90-day salt ponding
test.” Cement and Concrete Research, 29(8), 1239-
1248.
Mindess, S., Young, J. F. and Darwin, D., (2003).
“Concrete.” 2nd ed. New Jersey: Prentice Hall.
Mirzahosseini, M. and Riding, K. A., (2014). “Effect of
curing temperature and glass type on the pozzolanic
reactivity of glass powder.” Cement and Concrete
Research, 58, 103-111.
Neithalath, N., (2008) “Quantifying the effects of
hydration enhancement and dilution in cement pastes
containing coarse glass powder.” Advanced Concrete
Technology, 6(3), 397-408.
Rajabipour, F., Maraghechi, H. and Fischer, G., (2010).
“Investigating the alkali-silica reaction of recycled
glass aggregates in concrete materials.” Journal of
Materials in Civil Engineering, 22(12), 1201-1208.
Shayan, A. and Xu, A., (2004). “Value-added utilization
of waste glass in concrete.” Cement and Concrete
Research, 34(1), 81-89.
Schwarz, N., Cam, H. and Neithalath, N., (2008).
“Influence of a fine glass powder on the durability
477
characteristics of concrete and its comparison to fly
ash.” Cement and Concrete Composites, 30(6), 486-
496.
Shao, Y., Lefort, T., Moras, S. and Rodriguez, D.,
(2000). “Studies on concrete containing ground waste
glass.” Cement and Concrete Research, 30(1), 91-100.
Shi, C., Wu, Y., Riefler, C. and Wang, H., (2005).
“Characteristics and pozzolanic reactivity of glass
powders.” Cement and Concrete Research, 35(5),
987-993.
Shi, C., Stegemann, J. A. and Caldwell, R. J., (1998).
“Effect of supplementary cementing materials on the
specific conductivity of pore solution and its
implications on the rapid chloride permeability test
(AASHTO T277 and ASTM C1202) results.” ACI
Materials Journal, 95(4), 389-393.
Taha, B. and Nounu, G., (2009). “Utilizing waste
recycled glass as sand/cement replacement in
concrete.” Journal of Materials in Civil Engineering,
21(12): 709-721.
Tan, K. H. and Du, H., (2013). “Use of waste glass as
sand in mortar: Part I - fresh, mechanical and dura-
bility properties.” Cement and Concrete Composites,
35(1), 109-117.
Wright, J. R., Cartwright, C., Fura, D. and Rajabipour,
F., (2014). “Fresh and hardened properties of concrete
incorporating recycled glass as 100% sand
replacement.” Journal of Materials in Civil
Engineering, 26(10), 04014073.