CEMENT and CONCRETE RESEARCH. Vol. 22, pp. 833-840, 1992. Printed in the USA. 0008-8846/92. $5.00+00.

1992 Pergamon Press Ltd.

A.C. IMPEDANCE SPECTROSCOPY (I): A New Equivalent Circuit Model for Hydrated Portland Cement Paste PING GU*, PING XIE*, J. J. BEAUDOIN**, R. BROUSSEAU** * Dept. of Civil Eng., University of Ottawa, Ottawa,Ont.,Canada, K 1N 6N5 ** Materials Laboratory, Institute for Research in Construction, National Research Council, Ottawa,Ont., Canada, K1A 0R6

(Communicated by Zhong-Wei Wu) (Received Oct. 21, 1991; in final form Feb. 3, 1992)

ABSTRACT A.C. impedance spectroscopy has been used to investigate the mechanism of hydration of portland cement paste, from 48 to 380 hours. Interpretation of a.c. measurements obtained over a wide frequency range was complimented by equivalent circuit modelling. The proposed equivalent circuit was chosen so that its RC parameters would physically represent microstructural elements of the cement paste. INTRODUCTION Hydrated cement paste is a moisture sensitive microporous material. The main cement hydration product is calcium silicate hydrate or C-S-H. The nature of C-S-H changes with hydration time and relative humidity. The stability of C-S-H depends on the variation of environmental conditions as evidenced by changes in morphology, interfacial regions, pore structure and intrinsic strength. These characteristics can be studied by means of impedance or conductance measurements. Previous applications of resistivity measurements to cement paste investigations include the pioneering work of Shimizu[1]. He observed an inflexion point in the resistance-time diagram for hydrating cement corresponding to the completion of setting under isothermal conditions. Calleja observed resistance peaks when portland cement is set under adiabatic conditions[2]. Hammond and Robson were among the first to apply their techniques to concrete[3]. The a.c. impedance technique has been applied recently to investigate the electrical properties of hydrated cement pastes[4-7]. New information related to microstructure and hydration mechanism was obtained. Impedance results on hydrated portland cement obtained by means of frequency scanning are among the earliest to be reported. In these studies, a single arc was observed in the high frequency range with a small part of second arc in the low frequency
833

834

Ping Gu et al.

Vol. 22, No. 5

region. The high-frequency arc was attributed to a bulk cement related effect whereas the low frequency arc was due to the electrode-cement interface effect. The Maxwell-Wayner effect model[8] and other complicated equivalent circuit models[4-7] were proposed in order to interpret the impedance spectra. The intrinsic properties of the cement system that lead to the observation of the high-frequency arc and the physical meaning of the proposed RC elements in the equivalent circuit models are however not clear. In this paper, a simple equivalent circuit derived from fundamental considerations of cement paste hydration is proposed. The uniqueness of this model is that a physical meaning has been assigned to each RC element to explain the impedance behavior of hydrated cement. NEW EQUIVALENT CIRCUIT MODEL FOR HYDRATED CEMENT PASTE Impedance spectra recorded over a wide range of the frequencies have provided new information and insight on cement paste microstructure and hydration. Interpretation of an impedance spectrum (IS) is however complex. It requires modelling with some equivalent circuit until the electrical response of the elemental microstructure of the cement paste is well simulated. The usefulness of the analyses is strongly dependent on how the electrical components are selected. Each RC component should have some physical meaning related to the microstructure of hydrating cement paste. solid-liquid unit cell 1st layer

rnth layer electrode

interfaces electrode

a.c. signal Figure 1 Layer model of hydrating portland cement paste used for equivalent circuit simulations. Hydrated cement paste is frequently viewed as an insulating porous matrix, its pores filled with a conductive electrolyte. Electrical conduction through paste samples is made possible by the net movement of charge carrying ions such as Ca ++, Na +, OH- and SO4= in the solution of its continuous micropore network. The microstructure and the chemistry of the pore solution of cement paste strongly affect its electrical impedance properties. Consider a cement paste specimen to which the a.c. signal is applied in Figure 1.The specimen is divided horizontally into m layers in the direction of the electric field. It is postulated that three basic elements: solid (unhydrated cement and hydration products), liquid (solution in the micro-pores or capillaries), and solid-liquid interfacial phases exist in each layer of the cement paste. It is also assumed that

Vol. 22, No. 5

AC IMPEDANCE SPECTROSCOPY,CEMENT PASTE

835

each divided layer contains n "solid-liquid interface unit cells" illustrated schematically in Figure 2a. The ionic charges align themselves along the interface in parallel with the direction of the electric field when the a.c. signal is applied. Similarly, the induced opposite charges line up in the solid side, like a charge plane, parallel to the solid-liquid interface. The resulting separation of charges simulates an electric capacitor capable of storing electric charge. The "unit cell" model shown in Fig.2a can be represented by an equivalent circuit, indicated in Fig.2b, in which the R s, R 1, Rin t and Cint terms are the resistance of solid, liquid and interface phases, and the capacitance of solid-liquid interface for a single "unit cell", respectively.
interlaces

(a)
liquid
Clnt C'int

(b)
CINT

(e)

Rs.L

i
R~r

Figure 2 (a). Schematic of the "solid-liquid interface unit cell" model; (b). The corresponding equivalent circuit for a single layer of cement paste; (c). The simplified equivalent circuit for one layer of cement paste. Since it is assumed that in each layer of cement paste there are n "unit cells", the impedance of that layer can be expressed by n unit equivalent circuits in series. This multiple equivalent circuit can eventually be simplified into the circuit displayed in Figure 2c, and the impedance of one layer of cement paste, Z L is: Zt. = Rs+L + 1/( 1/Rigr + jcx)Cxr~r) where the corresponding RC parameters are as follows:
RS,L= n(R,+Rt)

(1)

(2)

and
RI= 2nRi~t

(3) (4)

CI= Cm~n

836

Ping Gu et al.

Vol. 22, No. 5

The total impedance network of the whole block of cement paste is considered to be equal to that of m layers in parallel, i.e.: 1/'-Z~ot~l= 1/ZLt+ lf/-,L2+ ....... + 1/ZL~ If every layer is identical, 1/Z~ot~ 1= m/Z L
or

(5)

(6)

Zto,~l = ZL/m

(7)

Equation (7) indicates that the total impedance of a cement paste can be equivalent to a simple electrical circuit as exhibited in Fig.3. The RC parameters represent the total resistance or capacitance of a cement paste. Their physical meanings are defined as follows:

bulk cement effect

electrode effect

Ct(int)
Rt(int)

Cdl
Pet

Figure 3 A simplified equivalent circuit for hydrating cement paste.

Rl(s+l): The sum of the bulk solid and liquid phase resistances. Since the bulk solid resistance is always much larger than that of liquid phase in cement paste the contribution of the liquid phase resistance can be considered negligible.

Rt(int):

The solid-liquid interface resistance. It is noted that the structure of the interfacial

region between the solid surface and liquid is different from that of both bulk solid and liquid phases. Their conduction properties are, therefore, different. In the hydrated cement system, since the size of cement particles and corresponding hydrated products is minute(range of a few Hm in diameter), the contribution of the interface resistance to the cement electrical properties is likely to be significant. Ct(int): The solid-liquid interface capacitance. This component is attributed to the ions adsorbed chemically or physically in the interfacial area inside the macropores or capillaries, acting as tiny "double-layer capacitance units" when the electric field is applied. It is noted that each unit is not a pure capacitor, but rather a leaky one as it can be self-discharged, similar to a polarized metalsolution inter-phase observed in electrochemistry. The effect can also be due to the geometrical factors such as pore size and connectivity in cement paste, resulting in uneven charge distribution. It is also emphasized that only those tiny "double layer capacitors" normal to the electric field contribute to this apparent capacitance, with no contributions from those parallel to the field. Call: This is the electrode-cement system interface or so-called double layer capacitance, Cdl. The formation of Cdl is due to the existence of a boundary between the electrode and cement

Vol. 22, No. 5

AC IMPEDANCE SPECTROSCOPY, CEMENT PASTE

837

matrix in which a basic anisotropy in the force operating on the ionic particles causes a redistribution of the mobile charges and orientable dipoles.This redistribution is the structural basis of the potential difference across the inter-phase and resulting in the charge alignment at both the cement(with its electrolyte) and electrode surfaces. The inter-phase behaves as a capacitor. This cement-electrode capacitance effect always exists and its signal is usually picked up in the low frequency range. Rct: The charge transfer resistance. When charges are separated, a potential difference develops across the interphase so that to transfer a charge an energy barrier or so-called activation energy has to be overcome. Rct represents this required energy in an electrochemical reaction. In a nonelectric-polarization system, Rct is usually very large. In our study, the cement-electrode surface reactions are carefully avoided. It is also noted that depression of the high frequency arc can occur due to geometrical factors resulting in a spread of relaxation time of the ions. This has not been taken into account in the present equivalent circuit model. It will be discussed in a subsequent contribution. EXPERIMENTAL Materials: Type 10 portland cement was used. The chemical composition (wt.%) is as follows: SIO2=19.83; CAO=61.21; Fe203=3.20; A1203=4.18; MgO=4.09; SO3---3.93; Na20=0.45 and K20=0.82. Specimen Preparation: The fresh cement paste was mixed in a conventional Hobart mixer at a water/cement ratio 0.35, and placed into a 1.2 cm x 1.5 cm stainless steel cylindrical cell described previously[9]. The sample was left in a 100% relative humidity environment for 24 hours and subsequently immersed in lime solution for the remainder of the test period. Impedance Measurements: Impedance data was collected using a 1260 impedance gain-phase analyzer from Schlumberger Technologies. Measurements were made logarithmically down in frequency range from 20 Mhz to 1 Hz with 10 readings per decade. RESULTS AND DISCUSSIONS Experimental a.c. impedance data are plotted in the complex plane( Real vs. -Imaginary) in Figure 4 for the hydrating cement paste at various hydration time from 48 to 380 hours. At early hydration times(48 to 68 hours), only the electrode arc is visible. The high frequency arc, however, appears at an hydration time of 169 hours and continues to grows in diameter with time. The impedance behaviour of the cement paste at later hydration times is quite similar to those reported by other authors[4-7]. The spectra exhibit a typical high-frequency arc contributed by the solid-liquid interfacial capacitance, Ct(int), in the high frequency region and a "tail" or the second arc (at lower frequencies) corresponding to the effect of cement systemelectrode interfacial capacitance or so-called double layer capacitance, Cdl. Note that the two real axis intercepts, the bulk (at the high frequency end) and the electrode arc with the real axis, are the sum of the solid and liquid resistance, Rt(s+l) ' and the sum of Rt(s+l) and the interfacial resistance, Rt(int) ' respectively.

838

Ping Gu et al.

Vol. 22, No. 5

100

60

1
90 110 130 150
170

Figure 4 Real vs Imaginary component of impedance of hydrated cement paste, w/c=0.35, hydrated for various times.

70

Re~ (ohms)

The delayed appearance of the high frequency arc is evidence that the capacitive behaviour is related to microstructure changes during the cement hydration process of the tested sample. The sequential impedance spectra, Figure 4, are interpreted as follows. At early hydration times, the rigidity of the microstructure is too low to detect capacitive behavior due to this geometric effect. The proposed equivalent circuit is intended to simulate the experimental data through analysis of the impedance spectra(IS). Simulated and experimental data are plotted in Figure 5 in the form of Log(mod) vs. Log(Freq.) in which the experimental data is represented by circles
2.3

2.2

O~ O ...J

2.1

2.8

1.9

l_og(Fre. (Hz))

Figure 5 Simulated and experimental data plotted in Log(mod) vs. Log(Freq.). The circles are experimental data and solid lines are theoretical calculation. The numbers represent various hydration times: (1) 169; (2) 194; (3)286; and (4) 380 hours.

Vol. 22, No. 5

AC IMPEDANCE SPECTROSCOPY,CEMENT PASTE

839

and the solid lines are calculated according to our model. The modulus: IZl = [(Real) 2 + (Imaginary)2] 1/2. The simulated results of four hydration time, i.e. 169, 194, 286 and 380 hours as indicated in number 1 to 4, respectively. It is apparent that the simulation is excellent in the range of Log(Freq.)>5 where the high frequency arc appears. In this region reliable RC parameters can be acquired. In the region Log(Freq,)<5, the experimental data undulates upwards, as the contribution of cement systemelectrode double layer capacitive effect is evident. A list of all simulated RC parameters is given in Table 1. It can be seen that, with the increase of hydration time, (a) the sum resistance of solid and liquid, Rt(s+l) ' increases to a certain value and then approaches a constant; (b) the interfacial resistance, Rt(int),also increases as a function of time; and (c) the solid-liquid interfacial capacitance, Ct(int), decreases with hydration time. The capacitance values estimated from the top frequency of the bulk arc are also given in Table 1. Even though these estimated values are slightly larger than those from the simulation, the decreasing trend is still there. These observations are consistent with the previous work by McCarter et a113,4] and Scuderi et ai[5,9] but the explanations given here are different. Table 1. Variation in modelled and experimental by determined electrical parameters with increasinl; hydration time. Hydration Rt(-m0 Ctfm0 C_x, sam,t~ Top Freq. Rt0+l) (ohms) (nF) (nF) (MHz) Time(hours) (ohms) 48 80 68 87 169 99 13.5 7.2 12.0 0.978 104 194 15.0 5.0 10.8 0.978 111 286 25.0 3.2 4.2 1.506 380 113 30.0 3.0 3.5 1.506 The increase of the Rt(s+l) (high frequency arc) from 80 ohms to 99 ohms in the early hydration time is more rapid than at later hydration times i.e. 99 ohms to 113 ohms. This is because of the significant loss of the free water during the early days of the cement hydration process. However, once the degree of hydration reaches a certain level and the majority of free water is consumed, the microstructural effects begin to be observed. The change of Rt(s+l) is mainly due to the difference in resistivity values of the unhydrated cement and its hydration products, principally C-S-H. It is reasonable to expect that there is not much difference in resistivity between the solid reactants and the solid products; hence, the value of Rt(s+l) remains relatively constant. The increases of the arc diameter with time can be explained well with the proposed new model. It is useful to reflect on the physical meaning of the model parameters. The arc diameter represents the solid-liquid interfacial resistance,Rt(int). Once hydration proceeds, the mechanism involves dissolution of anhydrous cement compounds to their ionic constituents, formation of hydrates in the solution, and due to their low solubility, eventual precipitation of the hydrates from the supersaturated solution. These hydration reactions generate a tremendous amount of interfacial area which has impedance properties different from both bulk solid and liquid phases. Significant interfacial area is manifested by increasing values of the solid-liquid interfacial resistance, Rt(int). This point can also be explained by the equation (3) where n is the number of

840

Ping Gu et al.

Vol. 22, No. 5

"the solid-liquid unit cell". It is obvious that n increases with hydration time since more precipitates occur leading to more solid-liquid interfaces. The trend of decreasing capacitance readings with hydration time can also be explained by a similar argument. All the interfaces possess a capacitive component arranged in series.,The measured capacitance values decrease with n increasing according to equation (4). CONCLUSIONS A.C. impedance spectroscopy is a useful tool for following microstructural changes of a cement paste. The a.c. spectra have been analyzed with some success using equivalent electrical circuit models carefully chosen so the components have physical meaning with respect to the cement hydration process. Conclusions drawn from the modelling process include the following: (i) The significant decrease in free water content in the early hydration time causes the rapid increase of Rt(s+l). The solid resistance mainly contributes to Rt(s+l) in the later stages of the hydration process. (ii) The growth in diameter of the high frequency arc is directly related to the increase in the number of interfaces. Different impedance properties of the interface are derived from both solid and liquid phases during the hydration process. (iii) The terms Rt(int) and Ct(int) are highly dependent on the interface microstructure. They can be used to characterize qualitatively the degree of hydration and extent of microstructural change of cement paste. ACKNOWLEDGMENTS The authors would like to thank Mssrs. B. Myers and Gordon Chan for their assistance with the experimental apparatus. The financial support of NSERC and the Network of Centers of Excellence on High Performance Concrete is gratefully acknowledged.
REFERENCES

[1] [2] [3] [4] [5] [6] [7] [8]

Shimizu, Y., Concrete Cement Mill. Sect., 111 (1929). Calleja, J., J. Amer. Con. Inst., 23,525-535 (1952). Hammond, E. and Robson, T. D., The Engineer, 199, Jan.21, 78-80 (1955). McCarter, W. J., Garvin, S., and Bouzid, N., J. Mater. Sci. Lett., 7(10), 1056-1057 (1988). McCarter, W. J., and Brousseau, R., Cem. Conc. Res., 20 891-900 (1990). Brantervik, K., and Niklasson, G. A., Cem. Conc. Res., 21 469-508 (1991). Scuderi, C. A., Mason, T. O., and Jennings, H. M., J. Mater. Sci. Lett., 26 349-353 (1991) Bonanos, N., Steele, B. C. H., Butler, E. P., Johnson, W. B., Worrell, W. L., MacDonald, D. D., and McKubre, M. C. H., " Application of Impedance Spectroscopy", Chpt. 4, Ed., McDonald, J. R., Wiley & Sons, NY, 1987. [9] Christensen, B. J., Mason, T. O., and Jennings, H. M., J. Mater. Sci. Lett., in press.