- Email: [email protected]

Evaluation of thermal barrier coating exposed to different oxygen partial pressure environments by impedance spectroscopy Chunxia Zhang, Chungen Zhou, Shengkai Gong ⁎, Hefei Li, Huibin Xu School of Materials Science and Engineering, Beijing University of Aeronautics and Astronautics, Beijing, 100083, China Received 27 May 2005; accepted in revised form 28 November 2005 Available online 18 January 2006

Abstract In this investigation, the oxidation of two-layer thermal barrier coatings (TBCs) prepared by EB-PVD (electron beam physical vapor deposition) in different oxygen concentration environments at 1323 K was evaluated by impedance spectroscopy and microstructural observation (SEM). The higher partial pressure of oxygen accelerated the oxidation rate of bond coating and faster oxidation caused cracking at interface between the TGO and the bond coat. Impedance spectroscopy was used to evaluate the microstructural changes which occurred in the TBC system non-destructively. The relationship between the electrical conductivity of the YSZ layer and its microstructure including the parameters of microcracks was established. © 2005 Elsevier B.V. All rights reserved. Keywords: Thermal barrier coatings; Impedance spectroscopy; Microstructure; Thermal conductivity

1. Introduction Thermal barrier coatings (TBCs) consisting of Y2O3 stabilized ZrO2 as top coat and metallic bond coat are applied on superalloy components. The work environment of TBCs becomes severer due to the development of aeronautical techniques and the higher tolerating temperature and the higher performance were required for the high temperature components [1]. The TBC system is a complex solid system with porosity, multi-interface and multi-crystal. During service, the complexity increases as TGO (thermally grown oxide) formed at the YSZ/bond coat interface due to oxidation of the bond coat, and the nucleation and growth of cracks due to the stress levels caused by thermal conductivity mismatch with the bond coat, TGO and YSZ, can lead to delamination and spallation of TBCs and result in failure. Therefore, evaluation and assessment of the microstructure and physical properties of top ceramic layer, TGO layer, and interface of YSZ/TGO and TGO/ bond coat are significant and helpful for prediction of residual lifetime and proposing failure models.

⁎ Corresponding author. Tel.: +86 1082317117; fax: +86 1082338200. E-mail address: [email protected] (S. Gong). 0257-8972/$ - see front matter © 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.surfcoat.2005.11.136

Several non-destructive evaluation (NDE) techniques for assessment and quality control of TBCs have been developed including photoluminescence by analysing the frequency shift caused by TGO residual stress [2–4], acoustic emission during thermal fatigue and mechanical loading to identify crack mode and source [5,6], and infrared thermography which can give information about delamination directly [7]. Recently, impedance spectroscopy has been used to evaluate the TGO thickness and composition changes during oxidation and the effect of porosity of YSZ layer on TBCs system electronical properties [8–12]. However, a relationship between microstructure of topcoat and its electronical properties has not been set up. To evaluate the effect of oxygen concentration of environment on the oxidation of TBCs, this paper studied isothermal oxidation of TBCs in three kinds of environments, which are pure oxygen, mixed gas of 80 vol.% oxygen and 20 vol.% nitrogen, and air, respectively. This paper reports on impedance spectroscopy of thermal barrier coatings as function of isothermal oxidation and contains description about influence of microcrack on impedance spectroscopy. Theoretical development and experimental data analysis were proposed in this paper to set up a method to evaluate the variations of microstructure and thermal conductivity of top coat in TBCs by using of impedance spectroscopy.

C. Zhang et al. / Surface & Coatings Technology 201 (2006) 446–451

447

2. Experimental procedure

dispersion of electrode, in equivalent circuit, CPE is used instead of capacitance. The impedance of CPE is given by:

Specimens with double-side coating areas of 10 × 15 mm2 were prepared by EB-PVD with both a bond coat (∼80 μm thick) and a TBC top coat (∼150 μm thick) on Ni-based superalloy plates (3 mm thick). Specimens were oxidized in pure oxygen chamber at 1050 °C for 5, 10, 30, 50, 70, 90 and 100 h and oxidized in air and mixture of oxygen and nitrogen whose ratio is 4 : 1 for 100 h, respectively. For impedance measurement, the two-sides of TBC were coated with platinum gel (5 × 5 mm2) that served as electrodes. The platinum gel was sintered at 800 °C for 10 min. Impedance measurements were carried out at 400 °C using a Solartron SI 1260 Impedance/gain phase analyzer coupled with a Solartron 1296 Dielectric Interface. Spectra analysis was performed using Zview impedance analysis software (Scribner Associates, Inc., Southern Pines, NC). In the measurements, alternating current (AC) with an amplitude of 100 mV was employed and the AC frequency was in the range of 0.01 to 1 × 106 Hz. In order to study the relationship between microstructure of TBCs especially at interface and the impedance spectroscopy, the cross-section specimen was polished with a diamond paste prior to be examined using a scanning electron microscopy (SEM) coupled with energy dispersive X-ray spectroscopy. All samples were mounted in resin, grinded, and polished in water carefully to avoid cracks formed during sample preparation rather than formed during oxidation.

ZCPE ¼ A−1 ðjxÞ−n

3. Results and discussion 3.1. Impedance spectra of thermal barrier coatings after exposure in pure oxygen and TGO thickness analysis Typical Bode plots of impedance spectra measured for an isothermally oxidized specimen at 1323 K for 0, 5, 10, 30, 50, 70 and 90 h and the equivalent circuit for the measured system are presented in Fig. 1. Because the constant phase element (CPE) is more suitable to describe the behaviour of a non-ideal capacitance, caused by chemical inhomogeneity and the 7

O

o

YSZ

s

7

v

90

q

y Hz

Fig. 1. Impedance spectroscopy of TBCs after isothermal oxidation in pure oxygen environment and the equivalent circuit for TBCs.

ð1Þ

Where A is a constant that is independent of frequency, ω is angular frequency and j = (−1)1 / 2, and n is an exponential index which represents a dispersion of relaxation frequency. For a constant phase element, the exponent n is less than one. When this equation describes a capacitor, the constant A = 1 / C (the inverse of the capacitance) and the exponent n = 1. When this equation describes a resistor or inductor, the constant A = R, or L, and the index n = 0 or − 1 respectively. For the practical case, the index is usually more than 0 and less than one and indicates the dispersion of electrode. In the case shown in Fig. 1, the equivalent circuit consisting of two series parts represents TGO and YSZ, respectively, and in each part, there are a resistor and a CPE in parallel. For the typical Bode plots shown in Fig. 1, all plots are divided into two parts corresponding to the frequency range from 0.01 to 30 Hz and another corresponds to the frequency range from 30 to 106 Hz respectively. According to electrical and dielectric properties of Al2O3 and YSZ and the equivalent circuit, the contribution of TGO and YSZ to impedance spectra and equivalent circuit can be divided into two parts, one is for YSZ corresponding to high frequency range and the other is that for the TGO corresponding to low frequency range. Regarding Fig. 1, with increase of isothermal oxidation time in pure oxygen, there is an obvious increase in the impedance magnitude in the low frequency range (from 0.01 to 30 Hz), especially during the initial oxidation period as can be seen by comparing to the plots for sample in the as received case and after 5 h oxidation. However, when the oxidation time increased to 50 h, there is a slow decrease of the impedance magnitude in low frequency range (below 5 Hz). On initial oxidation, the thickness of oxidation layer increases rapidly. In pure oxygen, the oxidation rate is faster than in air, therefore, after 50 h, the oxidation rate becomes stable, that is why the thickness of TGO increases slowly with increasing oxidation time. Fig. 2 shows the SEM secondary electron images bond coat of TBC interface after oxidation at 1323 K in pure oxygen for 10, 20, 40, and 100 h, respectively. The thickness of TGO increases from about 1 to 3 μm continuously with increasing oxidation time from 10 to 40 h (see in Fig 2a, b, and c). But there is no significant increase in TGO thickness after 100 h oxidation (as shown in Fig. 2d) comparing to TBCs after 40 h oxidation. However, for the sample oxidized for 100 h, the loose scale shown in Fig. 2d indicates that there are transverse cracks along the interface of TGO/bond coat and some microstructure fractures in TGO layer. Wang et al. [9] proposed methods to calculate the thickness of TGO from the diameter of modulus spectra and Kazuhiro et al. [12] got the relationship between TGO thickness and the impedance fitting result according to equivalent circuits. Considering the dispersion of TBCs system and non-ideal electrode, constant phase element is the capacitance in equivalent circuit. On the other hand, during initial oxidation, as the main constitute of oxide is Al2O3, it is feasible to

448

C. Zhang et al. / Surface & Coatings Technology 201 (2006) 446–451

YSZ 8 SZ TGO TGO

(a)

BC

BC

5μm

10μm

(b) YSZ

YSZ

TGO

TGO

BC

BC

(d)

10μm

(c)

10μm

Fig. 2. SEM for the cross-sections in the TBCs oxidized in pure oxygen at 1050 °C for (a) 10 h, (b) 30 h, (c) 50 h and (d) 100 h.

calculate thickness according to the relationship between resistance and thickness as shown in the following equation: R ¼ qd=A

ð2Þ

Where R is the resistance, d is the thickness of electrode, A is the area of electrode and ρ is the resistivity of electrode material, and for Al2O3, its value of resistivity is a constant at the given temperature. A good linear relationship is shown in Fig. 3 by plotting resistance fitted according to the equivalent circuit for each plots (shown in Fig. 1) against square root of oxidation time. This indicates that the oxidation for times less than 100 h in pure oxygen follows a parabolic law (δ∼t1 / 2; t: time) and characterizing the thickness by resistance of TGO in fitting results is reliable.

3.2. Relation among thermal conductivity, electric conductivity and microcracks Fig. 4 shows the resistance of YSZ as function of oxidation time. Clearly, the resistance of the YSZ increases with oxidation time. The nearly linear relationship trend may be due to, 1. There is no sintering during exposure process which may increase the grain boundary conductivity of YSZ; 2. There may exist some microcracks in YSZ layer formed during heating and cooling process that could decrease the conductivity of YSZ; 3. Oxidation at high temperature especially in pure oxygen may decrease the oxygen vacancy concentration which may be the most important factor affecting the ionic conductivity of YSZ.

7

. x

Resistance of YSZ (Ohm)

. x

8. x

of

. x

R ss

. x . x .

8 /

H

/

Fig. 3. Fitted resistance of TGO according to equivalent circuit vs. square root of oxidation time in pure oxygen environment.

. x

. x

. x . x

. x . x

. x

. x

. x

. x . x

. x

Conductance of YSZ (S)

. x

.

.

8

t (Hr) Fig. 4. Resistance of YSZ as function of oxidation time in pure oxygen environment.

C. Zhang et al. / Surface & Coatings Technology 201 (2006) 446–451 .

According to the Wiedemann–Franz (WF) law the relationship between thermal conductivity (κ) and electrical conductivity (σ) for thin film [13] is

(a) β: .

.9 .9

where Γ is Lonze constant, kB is Boltmann constant, and e is electronic charge. For the given temperature, the ratio of thermal conductivity to electrical conductivity is constant. Therefore, the electrical conductivity can be used to characterize the thermal conductivity, and then the increase of resistance can be understood as a decrease in thermal conductivity. In addition, in Redondo and Beery's microcrack model [14], the effect of microcrack in the thin film on the thermal conductivity can be expressed by following equation: jn ðNslab −1Þð1−jc Þbg −1 ¼ 1þ ð4Þ 1−ð1−jc Þg j

β: .

/

ð3Þ

.8 .8

β: .

.7

Nsl b= .7

.

.

.

.

.

. .9

(b)

.8

β: .

.7 . . β: .

.

In Fig. 5(a)–(c), the ratios of the thermal conductivities versus γ for different numbers of slabs and the size of cracks are plotted according to the Eq. (5). These curves clearly show that the ratio of thermal conductivity for material with cracks to the bulk material decreases as the values of γ, Nslab, and β increase. As shown in Eq. (3), the ratio of thermal conductivity to electrical conductivity is constant at given temperature, therefore the electrical conductivity for material with microcracks can be expressed as, Nslab bg −1 r¼C 1þ ð6Þ 1−g Eq. (6) shows the dependence relationship of electrical conductivity with crack parameters. Comparing the curves

β: .

.

Nsl b=

. .

.

.

.

.

.

γ

.7

(c) . . .

/

It gives a quantitative estimation of the thermal conductivity of a film with microcracks, κn, to the thermal conductivity of the bulk material. The ratio is a function of the number of slabs, Nslab, the thermal conductivity of the cracks κc, the size of the cracks β, and the fraction of the cross-sectional area that they cover γ. For the case of YSZ thin film in thermal barrier coatings, with exposure time increasing, the microcracks formed and the amount of microcracks increased in the film. The number of slabs, Nslab, can be calculated by the number of cracks, then the number of slabs, Nslab, will increase with the increase of microcracks. On the other hand, the cracks will develop with exposure. Therefore the size of cracks is also a function of oxidation time and increase with time. With the increasing of amount and size of microcracks, the cross-sectional area that they cover, that is the parameter γ, increase too. Considering that the thermal conductivity of crack is much less than 1, and the number of slabs is not very large in the practical case, but it is much larger than 1, the Eq. (4) can be simplified as, jn Nslab bg −1 ¼ 1þ ð5Þ 1−g j

.

γ

/

j p2 kB2 ¼C¼ rT 3 e2

449

.

β: .

. . .

.

β: .

β: .

Nsl b= .

.

.

.

.

γ

Fig. 5. Ratio of the thermal conductivity vs. γ. (a) Nslab = 10; (b) Nslab = 100; (c) Nslab = 1000.

shown in Fig. 5 especially for Nslab = 1000 and β = 0.01 and the plot of electrical conductance of YSZ vs oxidation time, a similar trend is found. The similar trend demonstrates that the larger number of slabs and the larger the size of cracks, the curves were closely the actual situation, and this may be caused by the columnar structure deposited by EB-PVD. On the other hand, the similar trend shows that the decrease of electrical conductance is also caused by microcrack parameters such as Nslab, β, and γ. Therefore, the crack and thermal conductivity variation during oxidation can be estimated by testing impedance of YSZ. Fig. 6a–b show the CPE parameters of the TGO layer and YSZ top layer as function of oxidation time. As mentioned, CPE is used instead of capacitance due to the dispersion of electrode. The more homogenous the electrode is, the more

450

C. Zhang et al. / Surface & Coatings Technology 201 (2006) 446–451 .

(a)

. x

YSZ

. x

.

.

. x

TGO

n

A

. x

. x

BC

.

7

. x

10μm

(a)

.

8

o . x

8

. x

8

. x

8

. x

8

. x

8

. x

8

.

H

YSZ

.9

(b)

TGO .8

nYSZ

AYSZ

.8

BC

8

Fig. 7. SEM for the cross-sections in the TBCs oxidized 100 h in (a) air, (b) gas mixture of O2 : N2 = 4 : 1 for partial pressure.

H

Fig. 6. CPE parameters of (a) TGO layer; (b) YSZ layer exposed to pure oxygen environment.

uniform electrode is, then the value of index n is more closed to 1. For the TGO layer (Fig. 6a), the parameter A decreases with oxidation time, and this may be due to the increase in the thickness of alumina layer during oxidation. The index n increases slowly to 0.52. This indicates that the uniformity of TGO increases, but with the growth of mixed oxide, the inhomogeneity of TGO layer causes the index n much lower than 1. For the YSZ layer, the parameter A also shows decreasing trend with oxidation time, this maybe due to the development of microcracks in the YSZ layer formed during heating and cooling which can lower the dielectric constant of YSZ. The index n is basically consistent and larger than 0.8, which indicates that no significant microstructural or compositional change occurs during oxidation. 3.3. Influence of isothermal oxidation in different oxygen partial pressure on impedance spectra Fig. 7 shows the cross-section images of the bond coat–TBC interfaces after oxidation for 100 h at 1323 K in air and mixture that contains 80% oxygen and 20% nitrogen (see Fig. 7(a) and (b) respectively). Comparing to previous studies on oxidation of TBCs in pure oxygen and results shown in these experiments, the oxidation rate in pure oxygen is larger than in air. This can be confirmed by the thickness difference between the sample oxidized in pure oxygen (about 5 μm) and in air (about 4 μm) or

in mixed gas (about 4 μm, and the thickness is less than in pure oxygen and more than in air). On the other hand, the value is not numerically exact because all values were obtained from the average of values measured at different position. The TGO formed in air is dense and uniform. However, the TGO formed in oxygen is loose and contains some cracks in the layer. Meanwhile, the TGO scale oxidized in mixed gas has some cracks which are less than in pure oxygen. According to the oxidation kinetics, the oxidation rate is affected by the concentration of reactant because the compositions of TGO are Al2O3, NiO, CoO and (Ni,Co)(Cr,Al)2O4 which are all electronic hole semiconductors whose parabolic rate constants increase with increasing of oxygen partial pressures. x xy

x

:N o xy

= : .

Hz

. Hz

x x

Z"

o

.7

10μm

(b)

.7

.

Hz

x . 8Hz

x Hz

.

. x

. x

. x

8. x

x

7

Z' Fig. 8. Impedance spectroscopy of TBCs after 100 h static oxidation in different environments.

C. Zhang et al. / Surface & Coatings Technology 201 (2006) 446–451

Fig. 8 shows the Nyquist plots measured for samples oxidized in air, mixture, and pure oxygen, respectively. There are a complete semi-circle in high frequency range and an uncomplete semi-circle in low frequency range for all three plots. For the complete semi-circles, the height and diameter for the sample oxidized in pure oxygen are larger than for the samples oxidized in air and the gas mixture. Furthermore, the height and diameter for samples oxidized in air and the mixed gas are very similar. For the incomplete semi-circles, the height and diameter for the sample oxidized in air are smaller than that oxidized in pure oxygen and mixed gas. These are similar. As mentioned above, in the high frequency range, impedance can be attributed to YSZ, and at low frequency the impedance can be attributed to TGO. Therefore, the decrease of the conductivity of YSZ layer after exposure in pure oxygen for 100 h is more than in air and the gas mixture. The decrease of oxygen vacancy concentration in pure oxygen is more than in air and the gas mixture. This may explain the decrease in the conductivity of YSZ layer. For the height and diameter of incomplete semi-circles in the low frequency range, the parameters for the sample oxidized in air are smaller than for those oxidized in pure oxygen and mixed gas, with the two latter coming similar. This may due to the faster oxidation in higher oxygen partial pressure, resulting in the formation of a loose TGO scale containing more porosity and microcracks, as well as the delamination at YSZ/TGO layer, which would also decrease the conductivity of TGO. 4. Conclusion Isothermal oxidation of thermal barrier coatings in different environment including pure oxygen, air, and mixed gas containing 80 vol.% oxygen and 20 vol.% nitrogen is studied in this paper by impedance coupled with SEM. Oxidation in higher oxygen partial pressure, the oxidation rate is faster and

451

the TGO scale is looser and more porosity. With oxidation, impedance of sample increased and the resistance of TGO fitted according to equivalent circuit can be used to characterize the thickness of TGO. The parameters in equivalent circuit can be used to characterize the microstructure change of TBCs. And the expression of electrical and thermal conductivities of YSZ as function of microcracks parameters is obtained. The increase of cracks parameters such as the number of slabs, the size of cracks and the fraction of crack covered can be used to explain the decrease of YSZ electrical conductance, and to estimate the decrease of YSZ thermal conductivity during initial oxidation. References [1] R. Taylor, Surf. Coat. Technol. 50 (1992) 141. [2] D.R. Clark, R.J. Christensen, V. Tolpygo, Surf. Coat. Technol. 94–95 (1997) 89. [3] A. Selcuk, A. Atkinson, Acta Mater. 51 (2003) 535. [4] J.A. Nychka, D.R. Clark, S. Sridharan, Surf. Coat. Technol. 163–164 (2003) 87. [5] X.Q. Ma, M. Takemoto, Mater. Sci. Eng., A Struct. Mater.: Prop. Microstruct. Process. 308 (2001) 101. [6] A. Kucuk, C.C. Berndt, U. Senturk, R.S. Lima, Mater. Sci. Eng., A Struct. Mater.: Prop. Microstruct. Process. 284 (2000) 41. [7] S. Harada, Y. Ikeda, Y. Mizuta, Y. Sugita, A. Ito, Proceedings of the Int. Symposium on Thermographic NDT and E Techniques Symposium, 1995, p. 47. [8] S. Song, P. Xiao, Mater. Sci. Eng., B, Solid-State Mater. Adv. Technol. 97 (2003) 46. [9] X. Wang, J. Mei, P. Xiao, J. Eur. Ceram. Soc. 21 (2001) 855. [10] N.Q. Wu, K. Ogawa, M. Chyu, S.X. Mao, Thin Solid Films 457 (2004) 301. [11] P.S. Anderson, X. Wang, P. Xiao, Surf. Coat. Technol. 185 (2004) 106. [12] O. Kazuhiro, M. Dorian, S. Tetsuo, et al., NDT E Int. 32 (1999) 177. [13] G. Wiedemann, R. Franz, Ann. Phys. 89 (1853) 497. [14] A. Redondo, J.G. Beery, J. Appl. Phys. 60 (1986) 3882.