Cement & Concrete Composites 53 (2014) 289–304

Contents lists available at ScienceDirect

Cement & Concrete Composites

journal homepage: www.elsevier .com/locate /cemconcomp

X-ray computed tomography proof of bacterial-based self-healingin concrete

http://dx.doi.org/10.1016/j.cemconcomp.2014.07.0140958-9465/� 2014 Elsevier Ltd. All rights reserved.

⇑ Corresponding author. Tel.: +32 9 264 55 22; fax: +32 9 264 5845.E-mail addresses: jianyun.wang@ugent.be (J. Wang), nele.debelie@ugent.be

(N. De Belie).1 Tel.: +32 9 264 55 22; fax: +32 9 264 5845.

Jianyun Wang a,b,c,1, Jan Dewanckele d, Veerle Cnudde d, Sandra Van Vlierberghe e, Willy Verstraete b,Nele De Belie a,⇑a Magnel Laboratory for Concrete Research, Faculty of Engineering and Architecture, Ghent University, Technologiepark Zwijnaarde 904, B-9052 Ghent, Belgiumb Laboratory of Microbial Ecology and Technology (LabMET), Faculty of Bioscience Engineering, Ghent University, Coupure Links 653, B-9000 Ghent, Belgiumc SIM vzw, Technologiepark 935, BE-9052 Zwijnaarde, Belgiumd Department of Geology and Soil Science – UGCT, Ghent University, Krijgslaan 281 S8, B-9052 Ghent, Belgiume Polymer Chemistry and Biomaterials Group, Faculty of Sciences, Ghent University, Krijgslaan 281 S4bis, B-9000 Ghent, Belgium

a r t i c l e i n f o a b s t r a c t

Article history:Received 5 November 2013Received in revised form 8 July 2014Accepted 14 July 2014Available online 21 July 2014

Keywords:Self-healingHydrogelBacteriaCrackCaCO3

X-ray computed tomography

Self-healing strategies are regarded as a promising solution to reduce the high maintenance and repaircost of concrete infrastructures. In the present work, a bacterial-based self-healing by use of hydrogelencapsulated bacterial spores (bio-hydrogels) was investigated. The crack closure behavior of the speci-mens with/without bio-hydrogels was studied quantitatively by light microscopy. To have a view of theself-healing inside the specimens, a high resolution X-ray computed microtomography (X-ray lCT) wasused. The total amount and the distribution of the healing products in the whole matrix were investi-gated. This study indicates that the specimens incorporated with bio-hydrogels had distinct improvedhealing efficiency compared to the reference ones with pure hydrogel only. The healing ratios in the spec-imens with bio-hydrogels were in the range from 70% to 100% for the cracks smaller than 0.3 mm, whichis more than 50% higher than for the ones with pure hydrogel; and the maximum crack bridging wasabout 0.5 mm (in 7 d), while pure hydrogels only allowed healing of cracks of about 0.18 mm. The totalvolume ratio of the healing product in the specimens with bio-hydrogels amounted to 2.2%, which wasabout 60% higher than for the ones with pure hydrogel (1.37%).

� 2014 Elsevier Ltd. All rights reserved.

1. Introduction

Self-healing is a potential solution to obtain a sustainable con-crete, since it would reduce the high maintenance and repair costsof concrete infrastructures [1]. Due to the limited autogenous heal-ing capacity of concrete itself, extra healing agents are needed toenhance the self-healing properties. Among the currently investi-gated self-healing strategies, the microbial-based strategy forself-healing concrete cracks is an emerging field. This strategyrelies on the microbial-induced carbonate precipitation process.Most bacteria can produce or induce the formation of calcium car-bonate under suitable conditions [2]. This biogenic precipitation isnatural, environmentally friendly, durable and compatible withbuilding materials. Due to these distinct features, microbial CaCO3

has gained more and more attention from researchers and

engineers and is being widely investigated in civil engineering,specifically for surface protection [3–9], cementation and consoli-dation of loose particles [10–14], and crack repair [15–21].

In order to obtain microbial based self-healing, carbonate pre-cipitating bacteria are added into concrete during the mixing pro-cess. When cracks appear, bacteria in the crack zone are expectedto be activated and precipitate CaCO3 to in-situ heal the cracks.Therefore, the bacteria used should survive the mixing process,remain viable but not active inside the concrete, and become activeto precipitate CaCO3 when cracks appear. Also, it is known that thecement-based matrix gradually becomes a denser structurebecause of the ongoing hydration. Most pores have a size less than0.5 lm; while the size of bacteria is in the range of 1–3 lm and thesize of the spores is around 1 lm. Hence, there is a chance that bac-teria would be squeezed and crushed when the pores becomesmaller. To solve this problem, encapsulation of bacteria beforethe addition is preferable. The encapsulation material shouldhave a ‘shell’ function to protect bacteria and have no hinderingeffect on bacterial carbonate precipitation when cracking occurs.Furthermore, it should be noted that for a realistic self-healing,

290 J. Wang et al. / Cement & Concrete Composites 53 (2014) 289–304

no additional human interference should be required. Becausewater is an essential element for bacterial activities, obtainingsufficient water with the lowest amount of interference is of cru-cial importance for realistic self-healing. So far, full submersionhas been mostly used in lab tests, which is not feasible in manypractical cases.

Regarding the above-mentioned requirements, we appliedhydrogel as the bacterial carrier in this research. Hydrogels arehydrophilic gels which have high water absorption and retentioncapacity. They are widely used in many fields, such as hygienic(especially in disposable diapers), agriculture (water retention insoil, controlled release of fertilizers, etc.) and pharmaceuticals.Hydrogels can also be used to immobilize cells due to their goodbiocompatibility and mass transport (nutrients and oxygen) prop-erties [22–24]. Therefore, on one hand, the hydrogel can protectthe bacteria during the mixing and hydration processes; on theother hand, the swollen hydrogel (water is absorbed from the sur-roundings after cracks appear) can be used as the water reservoirto support bacterial activities, hence facilitating the formation ofthe microbial CaCO3.

So far, the healing efficiency in the microbial-based system hasbeen evaluated either directly by visualization of crack filling bylight microscopy, or indirectly by measurement of the improve-ment in matrix properties (strength regain, water-tightness, etc.)after healing [15,17–21,25]. The healing ratio, which was definedas the ratio between the healed crack width and the initial crackwidth, is often used as a measure. Since the crack width variesalong the length, several crack locations are investigated per crack(e.g. at intervals of 0.4 mm in [19]). This is a first possibility fordirect quantitative evaluation of the healing efficiency in bacteriabased self-healing systems; in fact, it is a semi-quantitative evalu-ation because only part of the crack locations are investigated andthe microscopy analysis is only focused on the surface. Neverthe-less, an estimate of the overall self-healing efficiency can still beobtained from this method.

Essentially, healing efficiency relies on the amount and distribu-tion of the healing products formed. Therefore, direct quantifica-tion of the total amount and the distribution of the precipitatesin the whole specimen is of utmost significance. In this study, highresolution X-ray computed microtomography (X-ray lCT) wasused for this aim. X-ray lCT is a non-destructive technique, whichgenerates three-dimensional (3D) images by combining a series ofcross-sectional images. lCT analysis is based on measurements ofthe attenuation of X-rays from different positions of an object anddepends on the atomic number and density of the object [26]. Itprovides information (visualization and quantification) about theinternal structure of the matrix without sample preparation orchemical treatment. Recently, X-ray lCT has become a frequentlyused technique in materials research [27–30]. The obtained resolu-tion is of key importance to get high quality images. The X-raybeam geometry used in laboratories is commonly conical and bymoving the sample between the source and the detector, one canchoose an appropriate magnification. This has as a consequencethat larger objects will have a lower resolution and smaller sam-ples a higher resolution. In order to evaluate the impact of self-healing products inside mortar samples on a pore-scale level,mm-sized samples were investigated in this study.

The aim of this research was to demonstrate the feasibility ofapplying hydrogel immobilized carbonate precipitating bacterialspores to approach a realistic self-healing in concrete. The self-healing efficiency of the mortar specimens with and withouthydrogel encapsulated bacteria incorporated, was investigatedboth by light microscopy (semi-quantification) and 3D X-ray lCT(full quantification) to quantify the amount of precipitates.

2. Materials and methods

2.1. Materials

2.1.1. Bacterial strainThe bacterial strain used in this research was Bacillus sphaericus

LMG 22257 (Belgian Coordinated Collection of Microorganisms,Ghent). Cultivation of B. sphaericus spores was performed in MBSliquid medium [31], which contained MgSO4�7H2O (0.3 g/L),MnSO4 (0.02 g/L), Fe2(SO4)3 (0.02 g/L), ZnSO4�7H2O (0.02 g/L), CaCl2

(0.2 g/L), Tryptose (10 g/L) and Yeast extract (2 g/L). The pH was 7.4(if not, it was adjusted to a pH of 7.4 by using 1 M HCl or NaOH).Mature spores suspensions (109 spores/mL) were used as inoculum(1% by volume). The cultures were incubated at 28 �C on a shakerat 100 rpm for 14–28 days until more than 90% of the cells werespores. The spores were then harvested by centrifuging the culture(7000 rpm, 4 �C, Eppendorf MiniSpin, Hamburg, Germany) for7 min. The centrifuged spores were resuspended in a sterile salinesolution (8.5 g/L NaCl). Subsequently, the suspension of the sporeswas subjected to pasteurization to minimize the amount of vegeta-tive cells in the culture. The final concentration of the spores in thesuspension was about 109 spores/mL.

2.1.2. HydrogelThe hydrogel used was developed by the Polymer Chemistry

and Biomaterials Group of Ghent University (PBM-UGent). Com-mercial Pluronic�F-127 (Sigma Aldrich) was used which is a tri-block polymer of poly (ethylene oxide) and poly (propylene oxide)(i.e. PEO–PPO–PEO) and has molecular weight approximately12,500 daltons. The OH-groups at the end of the chain were mod-ified with methacrylate groups to create double bonds at the endand Pluronic�F-127 bis-methacrylate (Pluronic�-BMA) wasobtained. Upon UV irradiation, the photoinitiator Irgacure�2959(2-dimethoxy-2-phenyl-acetophenone, 224.3 g/mol) will form freeradicals which then initiate the polymerization by reacting withthe double bonds of Pluronic�-BMA [32]. Finally, a crosslinkedpolymer network was formed for water absorption and retentioninside.

2.2. Methods

2.2.1. Encapsulation of the bacterial spores into hydrogelThe bacterial spores were encapsulated into the hydrogel dur-

ing the process of crosslinking. The suspension of the spores wasfirst mixed with the 20% w/w polymer solution (Pluronic�-BMA).Then, the initiator was also added to the solution. The whole mix-ture was degassed and mixed for 5 min, and was subjected to UVradiation for 1 h after which a gel sheet formed. For each hydrogelsheet, 10 g polymer solution and 173.8 lL Irgacure 2959 solution(8 g/L) were used. 1 mL spores suspension (109 spores/mL) wasencapsulated in one hydrogel sheet. Hydrogels with or withoutencapsulated bacterial spores will be represented as HS or H,respectively.

The hydrogel sheets were then subjected to freeze grinding (IKAYellowline A10 Analytical Grinder) and freeze drying (ChristAlpha2-4 LSC, Germany) to obtain the dry powders.

2.2.2. Mortar specimensMortar specimens were made by using cement (CEM I 52.5N),

sand (DIN EN 196-1 Norm Sand) and tap water. The water tocement ratio (w/c) by mass was 0.5 and the sand to cement ratioby mass was 3. Four series of specimens were made. Group R arespecimens without any additions. Group N are specimens with allthe nutrients added, including the food for bacteria(yeast extract) and the deposition agents (urea and Ca(NO3)2�4H2O).

J. Wang et al. / Cement & Concrete Composites 53 (2014) 289–304 291

Mortar specimens with nutrients, and H or HS, were representedas m-H and m-HS, respectively. The addition dosage (in mass%versus cement mass) was 12.85% for nutrients (4% urea, 8%Ca(NO3)2�4H2O and 0.85% yeast extract) and 2% for hydrogels,respectively.

The R specimens were made according to the procedure in thestandard NBN EN 196-1 [33]. For the specimens with nutrientsadded, the nutrients were first dissolved into all of the tap water.During the mixing process, the nutrients solution instead of tapwater was added. In order to incorporate hydrogels into the spec-imens, dry hydrogel powders were first mixed with cement for 15 sat the speed of 140 rpm and then 5 s at 285 rpm. After that, thenutrients solution was added into the mixer and the subsequentsteps were performed according to the same standard procedure.Two kinds of specimens were made for each group. Three rein-forced prisms (30 mm � 30 mm � 360 mm, diameter of the rebarUr = 6 mm, length of the rebar Lr = 660 mm) and one cylinder(D = 78 mm, h = 22 mm) were made. The rebar was in the centreof the prism (parallel to the long side). The specimens were storedin an air-conditioned room (20 �C, >95% RH) after casting and afterdemoulding. The R and N specimens were demoulded 24 h aftercasting while the ones with hydrogels (H or HS) were demoulded48 h after casting due to the retarded hardening caused by theaddition of hydrogels.

2.2.3. Crack creation and incubationAfter 28 d, the prisms were subjected to a tensile test to create

multiple cracks. The reinforcement of the prism was clamped intoa test machine (Amsler 100, SZDU 230, Switzerland). The distancebetween the clamp and the side face of the specimen was 50 mm. Auniaxial tensile load was applied on the specimen at a speed of0.01 mm/s (by stroke control). The load was stopped at the pointdetermined by Eq. (1).

d ¼W�anþ ðr=EÞ�L ¼W�

anþ 1:23 mm ð1Þ

where d is the elongation (the stroke in the loading program); Wa isthe average crack width, which was set as 150 lm in this study; n isthe number of cracks observed on one surface; r is the average loadin the plastic stage of the reinforcement, which equals 560 N/mm2;E is the E-modulus of the reinforcement, which equals 210,000 N/mm2; and L is the distance between two clamps (it was 360 mm+ 50 mm � 2 = 460 mm in the study).

After unloading, about 9 to 20 cracks were formed in each prism(3 to 5 cracks on each surface). The rebar was cut off (about140 mm from each end) and the remaining part was wrapped withaluminum tape to prevent steel corrosion during the subsequentincubation.

The cracked prisms were subjected to three incubation condi-tions: (1) 60% RH; (2) 95% RH; (3) wet–dry cycles with water asthe submersion solution. All lasted 4 weeks at a temperature of20 �C. During the wet–dry cycles, the specimens were submergedin water for 1 h and exposed to air (60% RH) for 11 h.

2.2.4. Visualization of crack healing by light microscopyCrack filling in the specimens was investigated by means of

light microscopy (Leica S8 APO, Switzerland). After multiple crack-ing, 5–7 positions were marked on each crack. These positionswere distributed homogenously along the crack length. Therefore,there were 5–7 images to represent each crack. Images were takenwith the markers in the centre. For the specimens under the condi-tions 1 and 2, initial and final images after 4 weeks incubationwere taken while for the ones under condition 3, images weretaken once a week for 4 weeks to monitor the changes of crackwidths. In each image, two crack widths at the locations nearbythe marker were analysed by the Leica software (LAS). The crack

healing efficiency was evaluated by the crack width decreases.The final healing ratio (r) was calculated by Eq. (2).

r ¼ ðCwi � Cwf Þ=Cwi� �

� 100 ð2Þ

where Cwi is the initial crack width (lm) and Cwf is the final crackwidth measured after 4 weeks.

2.2.5. High Resolution X-ray lCT (HRXCT)2.2.5.1. Sample preparation. For X-ray lCT scanning, small samples(U = 8 mm, h = 10 mm) were drilled from the cylinders, whichwere also taken out from the air-conditioned room after 28 d. Eightto ten samples were taken from each kind of cylinder. After dril-ling, the samples were wrapped with a paper tape (tesa, Germany)on the side surface to avoid collapse during crack creation. Thecracks in the small cylinders were made manually by a screw jackas shown in Fig. 1. During the process of screwing, it was difficultto create cracks of a fixed width. For each group, one sample wasscanned before and after the wet–dry cycles.

In order to make sure no movement occurs in the sample itselfduring wet–dry cycles, the cracked cylinders were plugged into aplastic pipe (Fig. 2(b)). A short (2 s) heat treatment was appliedto the pipe in order to induce some shrinkage and provide a tightwrapping of the cylinder. Then, the samples were glued on plasticsticks and the initial scans were performed. After that, they weresubjected to the wet–dry cycles (condition 3), as shown in Fig. 2.The end of the stick was plugged into a hole in the cover of a falcontube, which allowed for the submerging of the sample in a con-trolled way. The cylinders at the other end of the sticks were incontact with water (wet stage) in the falcon tubes for 1 h(Fig. 2(a)) and then were taken out from the tubes to expose themto air for 11 h (up-side down, Fig. 2(b)).

2.2.5.2. Sample scanning. The X-ray lCT scans were performed atthe Centre for X-ray computed tomography at Ghent University,Belgium (UGCT, [34]). Both external and internal features of thesamples were visible. Consequently, the initial and final state ofthe same sample could be investigated. All transmission X-ray CTdevices are based on the same principles where the object of inter-est, in this case cylindrical mortar cores with a diameter of 8 mm,is positioned between an X-ray source and an X-ray detector. CTrequires a rotational motion of the sample relative to the source–detector system. For all samples, 1801 projections over 360� weretaken with an exposure time of 1250 ms for each projection. Analuminum filter (1 mm) was used to pre-harden the X-rays. TheX-ray tube provided a voltage of 130 kV with a tube current around107 lA. The obtained voxel size for the cylindrical cores was 7 lm.Taking this voxel size into account, features smaller than 7 lmwere not visible in the reconstructed volumes. Because the sam-ples had to be scanned before and after healing, the same acquisi-tion parameters (filter, exposure time, number of projections, etc.)were installed for each scan. After obtaining the raw projections,reconstructions of the samples, before and after treatments (incu-bation), were performed using Octopus [35]. In addition, the sameset of parameters (ring- and spot filter, correction for tilt and skewof the detector, etc.) were also taken into account for the recon-struction process.

2.2.5.3. Image processing. A problem arose when in the recon-structed images (given in grey scale) the healing products had tobe separated from the matrix of the sample. Due to a similar ele-ment composition between the matrix and the healing products(they have a comparable X-ray attenuation coefficient and thussimilar grey values), it is difficult to threshold the formed productson the reconstructed images. In order to obtain a 3D distributionand exact location of the healing products, digital image

Fig. 1. Manual creation of cracks in small cylinders (a: Screw jack with a small cylinder; b: cracked small cylinders).

Fig. 2. Wet–dry cycles for small cylinders (the arrow indicates the plastic pipe; a: submersion (wet stage); b: exposed to air (60% RH, dry stage)).

Layer 1Layer 2Layer 3Layer 4Layer 5

Healing products

35 µm

Fig. 3. Graphical representation of the calculation of the precipitation in function ofthe depth inside the sample.

292 J. Wang et al. / Cement & Concrete Composites 53 (2014) 289–304

subtraction was performed on the scans taken before and after thehealing, resulting in a differential volume that represented thehealing products.

In order to perform this, the two datasets (before and after heal-ing) for the samples R, m-H and m-HS were loaded into the 3D ren-dering program VG Studio Max [36]. Subsequently, the sampleswere manually registered (aligned), re-sliced and exported. Thisresulted in the same structural information on the same slice num-ber before and after the healing. After exporting the volumes, a dif-ferential volume was obtained by subtracting the original volumefrom the volume after healing (thus with new precipitations). Aftersubtracting, a bilateral filter was applied to remove the influence ofnoise and isolated single voxels were removed from the volume. Inthis way, only the structural changes became visible.

In a second step, the differential volumes of the R, m-H andm-HS sample were loaded in Morpho+ [37] for the quantification

of the precipitation. This program enables segmentation of the fea-tures of interest, in this case the healing products. A cylindricalregion of interest was taken (about 6 mm in diameter and 8 mmin height). The precipitation as a function of depth inside the sam-ple core was calculated by virtually eroding the surface of the sam-ple with a step size of 35 lm (Fig. 3). In each interval layer, thetotal amount of healing products was calculated. This was doneover a distance of 7 mm deep inside the samples (not until the bot-tom of the sample because the influences from cone beam artifactsand from gluing the sample to the carbon sticks could disturb theresults). The total precipitation inside the whole sample core wasalso calculated.

3. Results

3.1. Microscopic evaluation of crack healing efficiency

3.1.1. Crack healing under 60% RH and 95% RHNo significant crack healing occurred in the specimens which

were stored under the condition of 60% RH. As shown in Fig. 4,none of the specimens, with or without hydrogels, exhibitedhealed cracks. The crack widths did not decrease at all after4 weeks. In some cases, the measured final crack widths were evenlarger than the initial ones.

Similarly, no obvious crack healing was observed in the speci-mens which were incubated under 95% RH. As shown in Fig. 5, inall specimens, most of the final crack widths remained the sameas the initial ones; only a few cracks had slightly decreased crackwidths.

0

0.1

0.2

0.3

0.4

Cra

ck w

idth

aft

er 4

wee

ks (

mm

)

Initial crack width (mm)

Initial

4 weeks

R

0

0.1

0.2

0.3

0.4

Cra

ck w

idth

aft

er 4

wee

ks (

mm

)

Initial crack width (mm)

Initial

4 weeksN

0

0.1

0.2

0.3

0.4

Cra

ck w

idth

aft

er 4

wee

ks (

mm

)

Initial crack width (mm)

Initial

4 weeksm-H

0

0.1

0.2

0.3

0.4

0 0.1 0.2 0.3 0.4 0 0.1 0.2 0.3 0.4

0 0.1 0.2 0.3 0.4 0 0.1 0.2 0.3 0.4

Cra

ck w

idth

aft

er 4

wee

ks (

mm

)

Initial crack width (mm)

Initial

4 weeksm-HS

Fig. 4. Crack width changes in the specimens stored at 60% RH.

0

0.1

0.2

0.3

0.4

Cra

ck w

idth

aft

er 4

wee

ks (

mm

)

Initial crack width (mm)

Initial4 weeksR

0

0.1

0.2

0.3

0.4

0.5

Cra

ck w

idth

aft

er 4

wee

ks (

mm

)

Initial crack width (mm)

Initial4 weeksN

0

0.1

0.2

0.3

0.4

Cra

ck w

idth

aft

er 4

wee

ks (

mm

)

Initial crack width (mm)

Initial4 weeksm-H

0

0.1

0.2

0.3

0.4

0.5

0 0.1 0.2 0.3 0.4 0 0.1 0.2 0.3 0.4 0.5 0.6

0 0.1 0.2 0.3 0.4 0 0.1 0.2 0.3 0.4 0.5

Cra

ck w

idth

aft

er 4

wee

ks (

mm

)

Initial crack width (mm)

Initial4 weeksm-HS

Fig. 5. Crack width changes in the specimens stored at 95% RH.

J. Wang et al. / Cement & Concrete Composites 53 (2014) 289–304 293

3.1.2. Crack healing under wet–dry cyclesThe specimens subjected to the wet–dry cycles showed more

crack healing compared to those stored at 60% RH and 95% RH.The changes of crack widths in each specimen are shown in Figs. 6–9. The closer star dots are to the X-axis, the smaller the crack widthis, and hence, a higher healing ratio was obtained.

A limited amount of crack healing occurred in the specimens ofthe R series after 4 weeks in wet–dry cycles (Fig. 6). The final crackwidths were almost the same or only slightly smaller than the ini-tial crack widths. The specimen N had a somewhat better healingeffect (Fig. 7). More decrease in the crack widths was observed.Most of the crack width decrease occurred in the 1st week;

0

0.1

0.2

0.3

0.4

0.5

Cra

ck w

idth

aft

er 1

wee

k (m

m)

Initial crack width (mm)

Initial

1 week

0

0.1

0.2

0.3

0.4

0.5

Cra

ck w

idth

aft

er 2

wee

ks (

mm

)

Initial crack width (mm)

Initial

2 weeks

0

0.1

0.2

0.3

0.4

0.5

Cra

ck w

idth

aft

er 3

wee

ks (

mm

)

Initial crack width (mm)

Initial

3 weeks

0

0.1

0.2

0.3

0.4

0.5

0 0.1 0.2 0.3 0.4 0.5 0 0.1 0.2 0.3 0.4 0.5

0 0.1 0.2 0.3 0.4 0.5 0 0.1 0.2 0.3 0.4 0.5

Cra

ck w

idth

aft

er 4

wee

ks (

mm

)

Initial crack width (mm)

Initial

4 weeks

Fig. 6. Evolution of crack widths in the specimen of R during wet–dry cycles.

0

0.1

0.2

0.3

0.4

0.5

Cra

ck w

idth

aft

er 1

wee

ks (

mm

)

Initial crack width (mm)

Initial

1 week

0

0.1

0.2

0.3

0.4

0.5

Cra

ck w

idth

aft

er 2

wee

ks (

mm

)

Initial crack width (mm)

Initial

2 weeks

0

0.1

0.2

0.3

0.4

0.5

Cra

ck w

idth

aft

er 3

wee

ks (

mm

)

Initial crack width (mm)

Initial

3 weeks

0

0.1

0.2

0.3

0.4

0.5

0 0.1 0.2 0.3 0.4 0.5 0 0.1 0.2 0.3 0.4 0.5

0 0.1 0.2 0.3 0.4 0.5 0 0.1 0.2 0.3 0.4 0.5

Cra

ck w

idth

aft

er 4

wee

ks (

mm

)

Initial crack width (mm)

Initial

4 weeks

Fig. 7. Evolution of crack widths in the specimen of N during wet–dry cycles.

294 J. Wang et al. / Cement & Concrete Composites 53 (2014) 289–304

between the 1st and the 4th week, the crack width changes werenot obvious anymore. In the specimens R and N, none of the crackswere completely healed after 4 weeks of incubation in wet–drycycles.

The specimens with hydrogels embedded showed more crackhealing compared to the ones without hydrogels. In the specimenm-H (Fig. 8), the crack widths in the range of 0.02–0.2 mm had an

obvious decrease in the 1st week; some locations with crackwidths less than 0.05 mm were completely closed. After twoweeks, more locations along the cracks were completely bridged(see more star dots on the X-axis). The maximum crack widthbridged was about 0.18 mm. Among the investigated locations,about 24% of the locations were completely bridged. In the 3rdand 4th weeks, the crack widths kept decreasing and more star

0

0.1

0.2

0.3

0.4

0.5

Cra

ck w

idth

aft

er 1

wee

k (m

m)

Initial crack width (mm)

Initial

1 week

0

0.1

0.2

0.3

0.4

0.5

Cra

ck w

idth

aft

er 2

wee

ks (

mm

)

Initial crack width (mm)

Initial

2 weeks

0

0.1

0.2

0.3

0.4

0.5

Cra

ck w

idth

aft

er 3

wee

ks (

mm

)

Initial crack width (mm)

Initial

3 weeks

0

0.1

0.2

0.3

0.4

0.5

0 0.1 0.2 0.3 0.4 0.5 0 0.1 0.2 0.3 0.4 0.5

0 0.1 0.2 0.3 0.4 0.5 0 0.1 0.2 0.3 0.4 0.5

Cra

ck w

idth

aft

er 4

wee

ks (

mm

)

Initial crack width (mm)

Initial

Fig. 8. Evolution of crack widths in the specimen of m-H during wet–dry cycles.

0

0.2

0.4

0.6

0.8

Cra

ck w

idth

aft

er 1

wee

k (m

m)

Initial crack width (mm)

Initial

1 week

0

0.2

0.4

0.6

0.8

Cra

ck w

idth

aft

er 2

wee

ks (

mm

)

Initial crack width (mm)

Initial

2 weeks

0

0.2

0.4

0.6

0.8

Cra

ck w

idth

aft

er 3

wee

ks (

mm

)

Initial crack width (mm)

Initial

3 weeks

0

0.2

0.4

0.6

0.8

0 0.2 0.4 0.6 0.8 0 0.2 0.4 0.6 0.8

0 0.2 0.4 0.6 0.8 0 0.2 0.4 0.6 0.8

Cra

ck w

idth

aft

er 4

wee

ks (

mm

)

Initial crack width (mm)

Initial

4 weeks

Fig. 9. Evolution of crack widths in the specimen of m-HS under wet–dry cycles.

J. Wang et al. / Cement & Concrete Composites 53 (2014) 289–304 295

dots fell on the X-axis which means that more crack locations werecompletely closed. However, it can be seen that none of the largecracks (>0.2 mm) were healed after 4 weeks.

The specimens with the bio-hydrogels embedded (m-HS) had ahigher crack healing efficiency than the ones without bio-hydro-gels, in view of the healing rate and the maximum healed crack

width. As shown in Fig. 9, considerable crack healing alreadyoccurred in the 1st week. Most of the crack widths (0.02–0.4 mm) decreased considerably and some of the cracks werealready completely closed. The widest crack healed was about0.4 mm in the 1st week. After that, the crack widths kept decreas-ing and more and more cracks were completely healed. The

0

0.2

0.4

0.6

0.8

1Sl

ope

of th

e tr

end

line

RNm-Hm-HS

0 week 1 week 2 weeks 3 weeks 4 weeks

Fig. 10. Evolution of the slope-values of the trend lines in the graphs of Figs. 6–9.

296 J. Wang et al. / Cement & Concrete Composites 53 (2014) 289–304

maximum crack width bridged at selected positions along thecracks after 4 weeks was about 0.5 mm. Among all investigatedlocations, about 60% of the locations were completely bridged.

The slopes of the trend lines in the graphs (Figs. 6–9) can indi-cate the development of the crack healing. As more star dotsmoved down towards the X-axis, the slopes of the trend lines alsodecreased. The slope-value 1 means no crack healing, while aslope-value 0 indicates that all cracks are completely healed. Thechanges of the slope-values during the incubation period can thusreflect the development of crack healing. As shown in Fig. 10, theslope-values of the specimens R remained unchanged (equal to1) over 4 weeks, which indicated that no crack healing occurredin the specimen. For the specimens N and m-H, a similar decreaseof the slope-values from 1 to 0.9 occurred in the 1st week, but afterthat, only a limited decrease was observed in the specimen N whilethe slope-value of the specimen m-H continued to decrease toabout 0.85 (till the 2nd week). The results showed that most of

0d

Fig. 12. The maximum healed crack part in the s

0

20

40

60

80

100

Hea

ling

rati

o (%

)

0-50 50-100 100-150 150-200 20

Initial crack w

Fig. 11. Average crack healing ratio in different ranges of

the crack healing occurred in the first week for the specimen Nand in the first two weeks for the specimen m-H. In the 3rd and4th week, a limited amount of crack healing occurred. The speci-mens with bio-hydrogels exhibited a distinct decrease of theslope-values during the four weeks. A sharp decrease was noticedin the first two weeks, followed by a milder decrease in the 3rd and4th weeks. The results indicate that crack healing kept going onduring the whole incubation period and a considerable amountof the crack healing already occurred in the first two weeks. Fromthe aspect of the final slope-values, the sequence of the healingefficiency in the specimens after 4 weeks was m-HS > m-H > N > R.

The crack healing ratio varied depending on the type of thespecimens and the initial crack width. Overall, the healing ratiodecreased as the crack width increased. The specimens which wereincorporated with hydrogels, especially the bio-hydrogels, hadhigher healing ratios than the ones without hydrogels. As shownin Fig. 11, the specimens R and N had healing ratios in the rangeof 1–7% and 5–30% respectively, when the crack widths were inthe range of 0–0.25 mm. The healing ratio decreased to almost zeroat larger crack widths (0.25–0.7 mm). For the specimen m-H, thehealing ratio varied from 75% to 10% in the crack range of0–0.25 mm and less than 5% when the crack width was larger than0.25 mm. Higher healing ratios were obtained in the specimenswith bio-hydrogels. All cracks with a width less than 0.05 mmwere completely healed. In the crack width range of 0.05–0.3 mm, specimens of the bacterial series had a healing ratio of95–80%. In the large crack zone (0.3–0.7 mm), the average healingratio was 50–30%, while the ones without bio-hydrogels had heal-ing ratios less than 10%. Overall, the sequence for the overall heal-ing ratio was m-HS > m-H > N > R.

Figs. 12–15 show the parts of the cracks where maximum heal-ing occurred in each specimen subjected to wet–dry cycles. For thespecimens with hydrogels embedded, the evolution of the crackhealing process (from the initial stage up to 4 weeks) is also exhib-

28d

pecimen R (initial: 207 lm; final: 187 lm).

R

N

m-H

m-HS

0-250 250-300 300-350 350-400 400-700

idth ranges (µm)

crack widths of the specimens under wet–dry cycles.

0d 28d

Fig. 13. The maximum healed crack part in the specimen N (initial: 91 lm; final: 45 lm).

0d 7d 14d

21d 28d

Fig. 14. The evolution of the crack healing process (maximum healed crack part) in the specimen m-H (initial: 184 lm; final: 0 lm).

7d 14d

21d 28d

0d

Fig. 15. The evolution of the crack healing process (maximum healed crack part) in the specimen m-HS (initial: 507 lm; final: 0 lm).

J. Wang et al. / Cement & Concrete Composites 53 (2014) 289–304 297

ited. The maximum crack widths healed in the specimens m-H andm-HS were about 0.18 mm and 0.5 mm, respectively. It can be seenfrom Fig. 15 that a crack width of 0.5 mm was almost completelyhealed only after 7 d. The majority of the cracks was almostentirely healed after two weeks. However, the maximum healedpart of the cracks in the specimens R and N was less than 50 lm(Figs. 12 and 13).

The morphology of the precipitation formed in the cracks of thespecimen m-HS is shown in Fig. 16. The particles had cubic, rectan-gular and spherical shapes and their size was about 10 � 50 lm.The Energy Dispersive Spectrum (EDS) results (Fig. 16(d)) showedthat the particles were mainly composed of three elements: Ca, Cand O. The weight ratio among the elements closely matched withthat of CaCO3, which indicated that these particles were CaCO3.

Bacterial imprints

CaCO3

Hydrogel remains

a

b

Bacterial imprints

CaCO3

d

c

Fig. 16. Images of the precipitation formed in the crack of the specimen m-HS in wet–dry cycles (a–c: images of the precipitation; d: EDS of particles).

298 J. Wang et al. / Cement & Concrete Composites 53 (2014) 289–304

Hydrogel remains were found in between the particles. Besides,the imprints left by bacterial cells were also found on the surfacesof some particles.

3.2. X-ray CT image analysis of crack healing efficiency

According to the results from the light microscopical analysis, itwas seen that the specimens had almost no crack healing whenstored at 60% RH or 95% RH. Only the specimens under wet–drycycles had visible crack filling. Since there were only minor differ-ences in crack healing between the specimen R and N, only thesamples from R, m-H and m-HS were chosen for HRXCT analysis.

3.2.1. 2D imagesThe X-ray tomography images gave a direct visualization of the

precipitation on the surface and inside the specimens. By subtract-

Fig. 17. 2D horizontal slice taken from th

ing the 2D images that were scanned before the healing processfrom the ones scanned after healing, the formed precipitatesbecame visible. Results for the different samples are shown inFigs. 17–22. For each sample, a 2D slice from the volume beforeand after healing along the vertical direction (Z axis, taken in thesurface layer of the sample) and along the horizontal direction (Yaxis) is shown. In addition, the thresholded precipitation (indicatedas white in color) which is derived from the differential slices(before and after) is also shown. For the reference sample (R)(Figs. 17 and 18), no obvious difference was observed betweenthe before and the after images. Only a very slight precipitationis visible. In the case of the sample m-H (Figs. 19 and 20), moreprecipitation was formed. Similar information can be seen for thesample m-HS, which also shows a strong precipitation. The afterimages (Figs. 20 and 22) show an obvious precipitation layer. How-ever, the precipitation occurred mostly in the surface layer, not

e vertical direction of the sample R.

Fig. 18. 2D vertical slice taken from the Y direction of the R sample, the surface of the sample is shown at the bottom side of the image.

Fig. 19. 2D horizontal slice taken from the vertical direction of the sample m-H.

Fig. 20. 2D vertical slice taken from Y direction of the m-H sample, the surface of the sample is shown at the bottom side of the image.

J. Wang et al. / Cement & Concrete Composites 53 (2014) 289–304 299

deep inside the matrix. The images from the vertical directionshow that cracks at these depths were not completely filledwith the healing products. Because the 2D images shown inFigs. 17–22 were taken at different depths from the surface ofthe samples (though all in the surface layer), it is not possible tocompare them quantitatively. Furthermore, it is difficult to obtaina general overview of the 3D distribution of the precipitation from2D slices.

Therefore, in the next step, the 2D reconstructed slices wererendered in 3D to obtain a full 3D insight into the distribution ofthe healing products in the whole matrix.

3.2.2. 3D imagesAfter differentiation, the precipitation (shown in yellow) was

thresholded and rendered together with the original volume inVG Studio (Fig. 23). In this figure, images in the first column are

Fig. 22. 2D vertical slice of the Y direction of the m-HS sample, the surface of the sample is shown at the bottom side of the image.

Fig. 21. 2D horizontal slice taken from the vertical direction of the sample.

300 J. Wang et al. / Cement & Concrete Composites 53 (2014) 289–304

the outlook of samples plus the precipitation. The second columnshows the distribution of precipitates inside. The sample volumewas removed about two-thirds from the top to give a more com-prehensive visualization. The last column shows all precipitatesin the sample. Some precipitation can be visualized in the sampleR though only a small amount was found in the 2D images. A slightconcentration is visible at the surface of the sample. Further insidethe sample, the precipitation is randomly spread and is not onlyconcentrated in the crack. In the case of the sample m-H, a densedeposition was found at the surface of the sample. The amountsharply decreased in the sub-surface layer and inside the matrix.For the sample m-HS, it seems that more precipitation was formed.However, similar to the sample m-H, a thick layer is seen more onthe surface and less inside the sample in the deeper part. Despitethis, a large amount of deposition was found 2–3 mm below thesurface, which is deeper than that of the sample m-H. By usingthe software program Morpho+ [37], quantitative data of theformed precipitation was obtained which is shown in the nextSection 3.2.3

3.2.3. Quantification of healing productsThe total amount of the precipitates and its volume percentage

in the matrix (in the volume of interest, VOI) are shown in Table 1.Although no obvious precipitation was found in the 2D images

of the specimen R, still some amount of the precipitates was dis-tributed in the matrix. The volume ratio was about 0.21%. Thiscould be attributed to the autogenous healing in the sample itselfand/or the noise contained in the image. Much more precipitation(5.42 mm3) was formed in the sample m-H and the volume ratiowas about 1.37%. Compared to the specimen m-H, the specimen

m-HS had more precipitation distributed throughout the matrix.The volume percentage of the precipitates was about 2.2% andtheir amount was 8.29 mm3.

In order to yield information on the distribution of the healingproducts: variation of the precipitates as a function of the depthinside the sample, partial precipitation percentages were calcu-lated by eroding the surface of the sample (35 lm per intervallayer). The whole volume was separated into three zones. Zone 1indicates the surface layer, which is from the top surface to a depthof around 500 lm. Zone 2 is at the depth between 500 lm andaround 3500 lm, indicating a subsurface layer, while Zone 3 isdeep inside the sample, from 3500 lm to 7000 lm. For all speci-mens, the volume ratio of the precipitation per interval layer wasmuch higher in Zone 1 than that in Zone 2 and Zone 3 (Fig. 24).The total volume ratio of the precipitation in Zone 1, 2 and 3 wasestimated as 0.67%, 0.28% and 0.06% for the sample R, 8.69%,1.33% and 0.35% for m-H, 8.04%, 3.30% and 0.61% for m-HS, respec-tively. It can be seen that as the depth increased, the ratio sharplydecreased in Zone 1, while it fluctuated in Zones 2 and 3. Furtherinside the sample, less precipitation was formed.

4. Discussion

4.1. Crack healing behavior in mortar specimens

4.1.1. Enhanced healing efficiency in the specimens with (bio-)hydro-gels under wet–dry cycles

When exposed to wet–dry cycles, specimens with (bio-)hydro-gels showed much more crack healing than the ones withouthydrogels. Hydrogels can absorb water and store water to promote

Fig. 23. 3D rendered view of the spatial distribution of healing products (in yellow) in the sample R, m-H and m-HS after treatment (left: outlook of samples plus theprecipitation; middle: distribution of precipitates inside; right: the whole precipitates in the sample). (For interpretation of the references to colour in this figure legend, thereader is referred to the web version of this article.)

Table 1Overview of the total amount of the precipitation in the samples.

Total volume of theprecipitates (mm3)

Volume percentagein the samples (%)

R 0.81 0.21m-H 5.42 1.37m-HS 8.29 2.20

Fig. 24. Quantification of the precipitation in function of depth for the R, m-H andm-HS sample. Three different zones can be distinguished.

J. Wang et al. / Cement & Concrete Composites 53 (2014) 289–304 301

the autogenous healing processes. Due to this, the crack healingefficiency in the H specimens was about two times that of the spec-imens without hydrogels. In view of the healing ratio, healedcracks, and maximum healed crack width, the highest healing effi-ciency was obtained in the specimens with bio-hydrogels, in whichthe bacterial spores were encapsulated in the hydrogel matrix.Compared to the specimen m-H, the healing efficiency in thespecimen m-HS was greatly strengthened: the healing ratio inthe specimens of the bacteria series was 70–100% for crack widthsof 0.05–0.3 mm, which is more than 50% higher than that of non-bacterial series. The improved crack healing observed in the spec-imens of the bacterial series compared to the ones without bacteriawas due to the calcium carbonate precipitate formed by the encap-sulated bacteria. The immobilized spores germinated into activecells when the hydrogel supplied them a favorable microenviron-ment, sufficient water and nutrients. The active cells decomposedurea into carbonate ions and calcium carbonate formed with theexistence of Ca2+ in the hydrogel or from the mortar matrix. There-fore, more precipitation formed in the specimen m-HS.

This was further demonstrated by the 3D image analysis froml-CT. Based on the quantitative analysis of the total amount ofhealing products formed, it was found that the volume percentageof the healing products in sample m-HS was about 60% higher thanthat in sample m-H. Since the only difference between these twospecimens was the addition of H or HS, the increased amount ofprecipitation can only be caused by the encapsulated bacteria.Regarding the distribution of the precipitates in the whole matrix,

302 J. Wang et al. / Cement & Concrete Composites 53 (2014) 289–304

it was found that the precipitates were mostly concentrated in thesurface layer (0–500 lm deep) but rapidly decreased at a certaindepth when going from top to bottom (glued to the stick), thatis, in the subsurface layer and the deep part of the specimen. Thisis due to the fact that CO2 and oxygen (for bacteria) are more avail-able in the surface layer. Precipitates thus firstly form in the sur-face part. The surface precipitation and the swollen hydrogelslowed down or hindered the further penetration of water, CO2

and oxygen. Thus, less precipitation was distributed inside thematrix.

Nonetheless, it is noted that the volume ratio of the precipita-tion in the subsurface layer of the sample m-HS (3.3%) was abouttwo times that of the sample m-H (1.3%). The increased precipita-tion was assumed to be generated from the bacterial activities.When there is a limited amount of oxygen, nitrate (NO3

�) can beused as the electron acceptor by some bacteria. Therefore, bacteriacould still have activities (germination, growth, decomposition ofurea and precipitation of calcium carbonate) in the subsurfacelayer which had limited available oxygen but had nitrate distrib-uted in the matrix. This needs to be further investigated in subse-quent research.

4.1.2. Influence of the incubation conditionsWater is a crucial constituent for crack healing, both for autog-

enous healing and bio-assisted healing. In case of autogenous heal-ing, all main relevant processes, such as secondary hydration ofunhydrated cement, precipitation of CaCO3 and swelling of hydra-tion products, need water. In case of the bio-assisted healing, bac-terial spores need water to revive their activities to precipitateCaCO3. Therefore, healing would not occur if water is not presentin the cracking zone. Therefore, for the specimens stored at 60%RH in which there is no free water in the surroundings, both inthe specimens with and without (bio)hydrogels, no crack healingwas detected. The final crack widths after four weeks were evenlarger than the initial crack widths, which could be due to the addi-tional drying shrinkage. Similarly, no crack healing was visualizedin all specimens at the condition of 95% RH. The final crack widthsdid not become larger but remained almost the same as the initialones because of a higher humidity. In the specimens with(bio)hydrogels, although the hydrogels can absorb some moisturefrom the surroundings, this amount is quite low (about 0.6 g/gunder 95% RH conditions) which is not sufficient to support theautogenous healing and bacterial activities, and hence no healingoccurred. Healing can only happen with a higher moisture capturecapacity. Snoeck et al. noticed that significant crack healingoccurred under 95% RH incubation in specimens with commercialhydrogel powders embedded, which had a moisture uptake abilityof 1.5–1.7 g/g. The crack healing efficiency was comparable tospecimens without hydrogels in wet–dry cycles [38].

During the wet–dry cycles, more free water was supplied to thespecimens, resulting in crack healing. The specimens withouthydrogels (R and N) only had limited healing efficiency. It wasnoticed that the same type of specimens had much higher autoge-nous healing efficiency when they were subjected to wet–drycycles with 16 h in water and 8 h in air [39]. According to this, itis demonstrated again that sufficient water is extremely importantto obtain a significant amount of autogenous healing. The speci-mens with hydrogels, had much more self-healing efficiency inthe same wet–dry cycles. This is due to the fact that hydrogels inor near the crack surface can not only absorb water but also retainthe absorbed water during the drying stage for the autogenoushealing and bio-assisted healing. Therefore, the samples m-H andm-HS had more significant amounts of crack healing in wet–drycycles than in 60% RH and 95% RH. The specimens R and N canabsorb some water during the immersion stage. However, the

absorbed water evaporated rapidly when the specimens wereexposed to the air (in about 2 h). Therefore, no continuous watersupply was available for the ongoing reactions, and hence, thehealing effect was limited.

4.2. Considerations regarding the use of l-CT to characterize self-healing

Light microscopy is a widely applied technique to characterizecrack closure during self-healing. It can give both direct visualand quantitative data about crack changes (crack widths and crackarea). However, the obtained information is mainly focused on thesurface, while the healing inside the matrix is unknown. Of course,thin section microscopy or SEM measurements can be used to pro-vide more information about the deeper layers and provide dimen-sional insight into the crack zone. However, these techniques aremore destructive; and hence, further research on the same samplebecomes impossible. By the use of l-CT, the problem can be solved.Spatial information can be achieved in a non-destructive way.

With the aid of HRXCT, it was not only possible to visualize theprecipitation in full 3D but also to quantify the precipitation. Thelatter was achieved by thresholding the differential volumes andcalculating the amount of precipitation within a certain zone(every 35 lm) from the surface of the sample. Although HRXCThas some important advantages as compared to light microscopy,the technique has to be used carefully. An obvious limitation ofthe technique is the resolution, which depends on the largestdiameter of the sample. In this study, the voxel size of the obtainedimages was 7 lm, so features smaller than this resolution cannotbe quantified. In spite of constant developments in the design ofX-ray systems and software, small samples remain unavoidable ifhigh resolutions have to be obtained. In order to see precipitationbelow 7 lm in thickness, smaller samples must be taken. However,this may imply problems with representivity.

A second limitation of X-ray CT is the lack of chemical informa-tion. On the reconstructed images after the precipitation test, thenewly formed CaCO3 is very difficult to differentiate from thematrix. This lack of differentiation, which is necessary to properlysegment the zones of interest was tackled by working with a differ-ential volume. The technique of differential volumes is often usedin material research and in most cases the location of water orother liquids is determined [40]. It was also used earlier by ourresearch team to quantify cement paste dissolution by biogenicacids [41]. The alignment of the before and after precipitation vol-umes was performed manually with VGStudio, taking into accountthe displacement along the three main directions, rotation aroundthe same axis and change in magnification. This method was timeconsuming, however, recently, significant progress has been madein automatic image registration. The advantage of virtually erodingthe sample for the calculation of the precipitation as a function ofthe depth (see Fig. 3) is that the zones in which the calculation wasdone are always similar. On the other hand, if the calculationwould have been done on each slide, as a function of the depth,the volume in which the calculation was done would be differenton the top and inside the sample.

Furthermore, it was also noticed that, although almost no crackhealing was visualized in the specimen R (in wet–dry cycles) underlight microscopy, small amounts (about 0.2%) of precipitation werestill found distributed in the sample under CT scan. The reasoncould be that light microscopy has a lower resolution than l-CTand does not allow one to visualize the precipitation inside thematrix (from autogenic sources). This 0.2% precipitation may alsoinclude the background noise from the l-CT scan, which is almostimpossible to avoid and the amount is difficult to calculate. There-fore, the absolute accurate amount of autogenous healing in the

J. Wang et al. / Cement & Concrete Composites 53 (2014) 289–304 303

reference specimens is difficult to determine. Nevertheless, thetotal amount of healing in the specimen R can be used as the con-trol for the specimen m-H and m-HS since background noise isalways there.

It should be noted that, only one sample of each kind was usedfor l-CT analysis due to the large amount of time consuming workper sample. Therefore, considering representivity, the resultsshould be combined with light microscopy results which werefocused on the investigation of multiple cracks in the largerspecimens.

4.3. Considerations of using hydrogel immobilized spores for self-healing concrete

The improved healing efficiency by use of hydrogel encapsu-lated spores in mortar specimens proved that this bacterial basedself-healing system was feasible. Autogenic and bacterial-basedcrack healing mostly relies on the water supply. So far, in orderto provide enough water to cracked specimens, most researchersmade use of full immersion, which is probably feasible for someunderground or underwater structures, but not so realistic forthose structures in normal environments: dry conditions, plussome wet conditions from rain or snow, etc. For a realistic self-healing, human interference should be kept as limited as possible.Due to a good water absorption and retention property, hydrogelscan continuously provide sufficient water for crack healing, bothfrom autogenic and bacterial-based sources, which can greatlydecrease the water supply (amount and frequency). Some specialhydrogels can even take up moisture from the surroundings undernormal humidity (for instance 60% RH) for the healing behavior.This is highly promising for self-healing in practice because extrahuman interference would be completely left out. Actually, asmentioned before, it has been reported by Snoeck et al. [38] thatsome commercial hydrogels can take up moisture (1.6 times oftheir weight) at a high humidity (>95% RH) and the crack healingefficiency obtained was equivalent to that under wet–dry cycles.However, it was impossible to immobilize the bacterial sporesinside these commercial hydrogels because they were already incross-linked form. Therefore, the investigation on the moistureuptake capacity under normal humidity of the bio-hydrogels isongoing in our research group.

It was noticed from l-CT images (2D) that the addition ofhydrogels created a more porous microstructure. This is due tothe non-optimal compatibility between the hydrogels and con-crete. Therefore, improvement of the compatibility of the hydro-gels with concrete is an essential need for self-healing concrete,which can be realized by means of optimization and modificationof the physical and chemical properties of hydrogels. Higher poros-ity may provide more space for precipitation, thus more precipita-tion was formed inside the specimens with added hydrogels.However, the superiority of self-healing in the specimens withhydrogel encapsulated spores is still convincing. If only taking intoaccount the precipitation in the surface layer (since deeper layershave different porosity), the volume ratio of the precipitation inthe surface layer (versus the volume of the surface layer) wasabout 0.45% for R, 5% for m-H and 5.9% for m-HS. And the precip-itation in the surface layer was about 40%, 63% and 46% of the totalprecipitation formed in the samples R, m-H and m-HS. Regardingthis, it can be seen that more precipitation was formed in m-HSthan in m-H, both in the surface layer and in deeper layers. Besides,multiple cracks of a range of 0–0.5 mm were created in all types ofmortar specimens. So, for all specimens there is quite some spacefor precipitates formation in the cracks. However, from crack heal-ing results (microscopy), it can be seen that more significant crackhealing occurred in the specimens with hydrogel encapsulatedspores.

5. Conclusions

In this study, results from 2D light microscopy combined withX-ray lCT analysis have proven that bacterial-based self-healingof concrete, induced by hydrogel encapsulated spores, was feasible.The specimens with hydrogel encapsulated bacterial spores incor-porated had an improved self-healing efficiency both regardingcrack closure and the amount of precipitation. The healing ratioin the specimens of the bacteria series was in the range from 70%to 100% for the cracks smaller than 0.3 mm, which is more than50% higher than that of non-bacterial series. The maximum crackbridging was about 0.5 mm (in 7 d) in the specimens with bio-hydrogels, while it was about 0.18 mm in the ones with purehydrogels. The total volume ratio of the healing product amountedto 2.2% in the specimens with bio-hydrogels, which was about 60%higher than that for specimens with pure hydrogel (1.37%). Theincreased amount of healing product (CaCO3) was due to bacterialactivity. The specimens without hydrogels showed the lowesthealing efficiency.

The 3D X-ray lCT analysis also revealed that the healing prod-uct was mostly distributed in the surface layer and had a sharpdecrease in the subsurface layer and deep inside the sample. How-ever, the specimens with bio-hydrogels still had a significantamount of precipitation (3.3% v/v) distributed in the subsurfacelayer, while the ones with pure hydrogels only had about 1.3% ofprecipitation, and the reference sample only 0.28%.

Acknowledgements

The financial support from the Research Foundation Flanders(FWO-Vlaanderen, Grant No. G.0157.08), Ghent University (a BOFgrant), and the Strategic Initiative Materials (SIM, program SHE)is gratefully acknowledged. The authors would like to thank StijnSteuperaert (Polymer Chemistry and Biomaterials Group, GhentUniversity, and Strategic Initiative Materials SIM-Flanders) forthe assistance during hydrogel synthesis, and Manuel Dierick(Department of Physics and Astronomy – UGCT, Ghent University)for help with acquisition of the X-ray CT images and installation ofthe software to scan the samples.

References

[1] Van Breugel K. Is there a market for self-healing cement-based materials? In:Schmets AJM, van der S Zwaag, editors. Proceedings of the first internationalconference on self healing materials. Noordwijk aan Zee, The Netherlands, 18–20 April, 2007.

[2] Boquet E, Boronat A, Ramos-Cormenzana A. Production of calcite (calciumcarbonate) crystals by soil bacteria is a general phenomenon. Nature1973;246:527–9.

[3] Le Metayer-Levrel G, Castanier S, Orial G, Loubiere JF, Perthuisot JP.Applications of bacterial carbonatogenesis to the protection and regenerationof limestones in buildings and historic patrimony. Sediment Geol1999;126:25–34.

[4] Dick J, DeWindt W, De Graef B, Saveyn H, VanderMeeren P, DeBelie N, et al.Bio-deposition of a calcium carbonate layer on degraded limestone by Bacillusspecies. Biodegradation 2006;17:357–67.

[5] Tiano P, Cantisani E, Sutherland I, Paget JM. Biomediated reinforcement ofweathered calcareous stones. J Cult Herit 2006;7(1):49–55.

[6] De Muynck W, Debrouwer D, De Belie N, Verstraete W. Bacterial carbonateprecipitation improves the durability of cementitious materials. Cem ConcrRes 2008;38:1005–14.

[7] De Muynck W, Cox K, De Belie N, Verstraete W. Bacterial carbonateprecipitation as an alternative surface treatment for concrete. Constr BuildMater 2008;22:875–85.

[8] Qian C, Wang J, Wang R, Cheng L. Corrosion protection of cement-basedbuilding materials by surface deposition of CaCO3 by Bacillus pasteurii. MaterSci Eng C-Bio 2009;29:1273–80.

[9] De Muynck W, De Belie N, Verstraete W. Microbial carbonate precipitation inconstruction materials: a review. Ecol Eng 2010;36(2):118–36.

[10] Whiffin VS. Microbial CaCO3 precipitation for the production of biocement.PhD thesis. Perth: Murdoch University; 2004.

[11] Whiffin VS, Van Paassen LA, Harkes MP. Microbial carbonate precipitation as asoil improvement technique. Geomicrobiol J 2007;24:417–23.

304 J. Wang et al. / Cement & Concrete Composites 53 (2014) 289–304

[12] DeJong JT, Fritzges MB, Nusslein K. Microbially induced cementation tocontrol sand response to undrained shear. J Geotech Geoenviron2006;132:1381–92.

[13] DeJong JT, Mortensen BM, Martinez BC, Nelson DC. Bio-mediated soilimprovement. Ecol Eng 2010;36:197–210.

[14] Cheng L, Cord-Ruwisch R. In situ soil cementation with ureolytic bacteria bysurface percolation. Ecol Eng 2012;42:64–72.

[15] Ramachandran SK, Ramakrishnan V, Bang SS. Remediation of concrete usingmicro-organisms. ACI Mater J 2001;98:3–9.

[16] Bang SS, Galinat JK, Ramakrisshnan V. Calcite precipitation induced bypolyurethane-immobilized Bacillus pasteurii. Enzyme Microb Technol2001;28:404–9.

[17] Jonkers HM, Thijssen A, Muyzer G, Copuroglu O, Schlangen E. Application ofbacteria as self-healing agent for the development of sustainable concrete.Ecol Eng 2010;36:230–5.

[18] Van Tittelboom K, De Belie N, De Muynck W, Verstrate W. Use of bacteria torepair cracks in concrete. Cem Concr Res 2010;40:157–66.

[19] Wiktor V, Jonkers HM. Quantification of crack-healing in novel bacteria-basedself-healing concrete. Cem Concr Compos 2011;33:763–70.

[20] Wang J, Van Tittelboom K, De Belie N, Verstraete W. Use of silica gel orpolyurethane immobilized bacteria for self-healing concrete. Constr BuildMater 2012;26:532–40.

[21] Wang J, De Belie N, Verstraete W. Diatomaceous earth as a protective vehiclefor bacteria applied for self-healing concrete. J Ind Microbiol Biotechnol2012;39:567–77.

[22] Degiorgi CF, Pizarro RA, Smolko EE, Lora S, Carenza M. Hydrogels forimmobilization of bacteria used in the treatment of metal-contaminatedwastes. Radiat Phys Chem 2002;63:109–13.

[23] Gutierrez MC, Garcia-Carvajal ZY, Jobbagy M, Yuste L, Rojo F, Abrusci C, et al.Hydrogel scaffolds with immobilized bacteria for 3D cultures. Chem Mater2007;19:1968–73.

[24] Dong Y, Zhang Z, Deng Y, Wang Y. Immobilization of nitrifying bacteria inwaterborne polyurethane hydrogel for removal of ammonium nitrogen fromwastewater. In: 3rd International conference on bioinformatics andbiomedical engineering, Beijing; 2009. p. 1–11.

[25] Wiktor V, Jonkers HM. Determination of the crack self-healing capacity ofbacterial concrete. In: Grantham, Mechtcherine, Schneck, editors. Concretesolutions. London: Taylor & Francis Group; 2012. p. 331–4.

[26] Cnudde V, Jacobs P. Monitoring of weathering and conservation of buildingmaterials through non-destructive X-ray computed microtomography.Environ Geol 2004;46:477–85.

[27] Ketcham RA, Carlson WD. Acquisition, optimization and interpretation of X-ray computed tomographic imagery: applications to the geosciences. ComputGeosci 2001;27:381–400.

[28] Cnudde V, Boone M, Dewanckele J, Dierick M, Van Hoorebeke L, Jacobs P. 3Dcharacterization of sandstone by means of X-ray computed tomography.Geosphere 2011;7:54–61.

[29] Dewanckele J, De Kock T, Boone MA, Cnudde V, Brabant L, Boone MN, et al. 4Dimaging and quantification of pore structure modifications inside naturalbuilding stones by means of high resolution X-ray CT. Sci Total Environ2012;416:436–48.

[30] Cnudde V, Boone MN. High-resolution X-ray computed tomography ingeosciences: a review of the current technology and applications. Earth-SciRev 2013;123:1–17.

[31] Kalfon A, Charles JF, Bourgouin C, De Barjac H. Sporulation of Bacillus sphaericus2297: an electron microscope study of crystal-like inclusion biogenesis andtoxicity to mosquito larvae. J Gen Microbiol 1984;130:893–900.

[32] Vandenhaute M, Schelfhout J, Van Vlierberghe S, Mendes E, Dubruel P. Cross-linkable, thermo-responsive pluronic

�building blocks for biomedical

applications: sythesis and physico-chemical evaluation. Eur Polym J2014;53:125–38.

[33] NBN EN 196–1. Method of testing cement – Part 1: determination of strength;2005.

[34] Masschaele B, Cnudde V, Dierick M, Jacobs P, Van Hoorebeke L, VlassenbroeckJ. UGCT: new X-ray radiography and tomography facility. Nucl Instrum Meth A2007;580(1):266–9.

[35] Vlassenbroeck J, Dierick M, Masschaele B. Software tools for quantification ofX-ray microtomography. In: 10th International symposium on radiationphysics. Coimbra, Portugal: Univ Coimbra, Sept 17–22, 2006 (Source: NuclInstrum Meth A 2007; 580(1):442–445).

[36] VG Studio Max. <http://www.volumegraphics.com/en/products/vgstudio-max.html>.

[37] Brabant L, Vlassenbroeck J, De Witte Y, Cnudde V, Boone MN, Dewanckele J,et al. Three-dimensional analysis of high-resolution X-ray computedtomography data with morpho+. Microsc Microanal 2011;17:252–63.

[38] Snoeck D, Van Tittelboom K, Steuperaert S, Dubruel P, De Belie N. Self-healingcementitious materials by the combination of microfibres and superabsorbentpolymers. J Intell Mater Syst Struct 2012:1–12.

[39] Wang J. Self-healing concrete by means of immobilized carbonateprecipitating bacteria. PhD thesis, Ghent: Ghent Universtity; 2013.

[40] Ketcham RA, Iturrino GJ. Nondestructive high-resolution visualization andmeasurement of anisotropic effective porosity in complex lithologies usinghigh-resolution X-ray computed tomography. J Hydrol 2005;302:92–106.

[41] De Graef B, De Windt W, Dick J, Verstraete W, De Belie N. Cleaning of concretefouled by lichens with the aid of Thiobacilli. Mater Struct 2005;38(284): 875–82.