Refractories and Industrial Ceramics

Vol. 45, No. 1, 2004

UDC 666.7.02

SILICON NITRIDE CERAMICS PREPARED BY VIBRATORY CASTING OF SELF-REINFORCED MIXTURES: TESTING FOR CRACK RESISTANCE G. D. Semchenko,1 Yu. M. Shmygarev,1 E. E. Starolat,1 V. V. Katin,1 and N. L. D’yakonenko1 Translated from Ogneupory i Tekhnicheskaya Keramika, No. 9, pp. 9 – 15, September, 2003. Technological factors capable of influencing the structure, physicomechanical properties, and crack resistance of silicon nitride ceramics intended for fabrication of component of complex shape are considered. The effect of a composite sintering aid Al2O3 + Y2O3 and an ethyl silicate binder on strength and crack resistance of raw and sintered ceramics subjected to hydrostatic compression is discussed. It is shown that the ceramics in question can be tested for crack resistance by the indentation method.

Conventional techniques for fabricating structural ceramics are little suited to molding components of irregular shape. Therefore of interest, both for practice and research, are sol-gel self-reinforced composites that could be used to fabricate, by means of a vibrocasting technology, components of complex configuration with superior physicomechanical properties and high crack resistance using powdered high-melting oxygen-free materials as precursors. A carbon-filled composite binder and a technology for its preparation have been developed for fabricating components of complex shape based on Si3N4 powder. An advantage of finely ground Si3N4 powder is its high specific surface which requires the use of a binder in large quantity (like with powders of ultradisperse silica or pulverized fused quartz) to prepare casting mixtures. Therefore it has been recommended to use ethyl silicate-based binders that are capable of forming a three-dimensional framework during the polymerization of polysilicic acids, that is, hydrolyzates prepared by hydrolysis of ethyl silicate using water in large quantity; water present in the hydrolyzate makes it possible to prepare a raw material with a sufficiently high porosity. A hydrostatic compression (HSC) technique was used to prepare a compacted raw material. It should be kept in mind that properties of Si3N4-based components, in particular crack resistance, are affected by various technological factors. Crack resistance (or fracture toughness) is characterized by the stress intensity factor K1c at the tip of a mode I crack (an opening mode crack) under plane deformation conditions. The value of K1c is a criterion for transition of a crack from the state of rest to a state of propagation. For ceramics, the crack resistance factor K1c is taken to signify the onset of 1

unstable crack propagation (catastrophic failure). The conventional and sufficiently accurate method for determining K1c is bending fracture of a notched specimen (a specimen with a preliminarily initiated artificial crack) [1]. A shortcoming of this method is that it requires destruction of the test specimen. In recent years, much effort has been aimed at developing nondestructive methods for determining the crack resistance. Among the techniques proposed, a method by which a crack of definite length is initiated by diamond indentation (using a Vickers pyramid) [2, 3] is considered one of the best suited. The indentation method using a Vickers diamond pyramid is technically simple and readily available; it allows the simultaneous determination of factor K1c and material hardness. A shortcoming of this method is that the factor K1c is calculated using semi-empirical expressions which, in the general case, are dependent on the load applied; furthermore, specific mechanisms of strain and fracture should be taken into account. This leads to a large spread in K1c values. In [4, 5], more accurate methods have been proposed. The determination of K1c by indentation is based on the property of a ceramic material in contact with a pointed indenter to develop cracks that propagate away from the edge of the impression in the plane of a thin section. Several types of crack may form in the ceramic material on its contact with an indenter (Fig. 1): radial (or median-radial), Palmqvist, and lateral cracks. The development of a crack under indentation is determined by elastoplastic fields generated in the material and by the material’s major characteristics: fracture toughness K1c , hardness H, and elastic modulus E. As is known [6], the length of a crack and the material toughness can be expressed by an empirical relationship. The relationship between K1c and crack dimensions (impression

Kharkov Polytechnical Institute National Technical University, Kharkov, Ukraine.

36 1083-4877/04/4501-0036 © 2004 Plenum Publishing Corporation

Silicon Nitride Ceramics Prepared By Vibratory Casting of Self-Reinforced Mixtures

37

c

c

l

2a

mc

lc

lc

mc

Fig. 1. Types of cracks generated in a brittle material on its contact with an indenter: a) radial (medianradial) cracks, mc; b ) Palmqvist cracks, mc; lateral cracks, lc; c is the radial crack length; l is the Palmqvist crack length; 2a is the impression diagonal.

of an indenter) is universal in character. In [7], an expression has been proposed: K1c =

H a æcö aç ÷ F èaø

3/ 2

æ EFö ç ÷ è H ø

0. 4

,

where a is the half-length of the impression diagonal, F is the constraint factor (about 3), c is the average length of a radial crack, and a is a constant (about 0.15). A property correlation for K1c determined by indentation method and by the standard bending method for a range of materials is shown in Fig. 2. The indentation method using the aforementioned relationship gives reliable values of K1c [8]. Departures of experimental data from this expression have been observed [9, 10] for small values of the c/a ratio when, under small loading, Palmqvist cracks, rather than radial cracks, are formed (Fig. 1). Here experimental data are best described by the expression K1c = 0.035

H a æ EFö ç ÷ F è H ø

2/ 5

ælö ç ÷ èaø

The ratio H/K1c has been proposed as an index of brittleness for materials [8]. This parameter characterizes the relative size of strain zone and ruptured zone near the impression. Both hardness and crack resistance vary with indenter load. It has been proposed in [4, 5] that a curve showing the ratio H/K1c as a function of the crack length or load applied be named a “damage curve.” It was argued that hardness and crack resistance are interdependent characteristics for any ceramic, and either of the two reflects the ability of a material to withstand surface damage. A typical damage curve is schematically illustrated in Fig. 3 [9]. Three portions are distinguished in the curve: portion I is associated with processes that take place within individual particles or grains. The transition to portion II characterizes the complete formation of main (transcrystallite)

– 1/ 2

,

where l is the length of a Palmqvist crack, and 0.25 £ l/a £ 2.5. For c/a > 2.5 (a region of radial cracks), experimental data are described by the formula K1c = 0.129

H a æ EFö ç ÷ F è H ø

2/ 5

æcö ç ÷ èaø

– 3/ 2

.

Later, a relationship for radial cracks was proposed [11]: K1c =

0.203Ha 2 c 3/ 2

.

According to data from [4], the departures for K1c determined using various relationships range from 2 – 3 to 16 – 20%.

Fig. 2. Property correlation for K1c derived by indentation method and standard bending method for various materials: 1 ) WC; 2 ) Si3N4 (hot-pressed); 3 ) SiC; 4, 5 ) Al2O3; 6 ) glass ceramics; 7 ) Si3N4 (reaction-sintered); 8 ) sapphire; 9, 10, 11 ) glass; 12 ) Si.

38

G. D. Semchenko et al. Hn ´ 10 – 3 m – 1/2 K1c

I

Hn , GPa II

a 1

2.8

III

2.6 2.4

2

2.2 60

100

120

140

K1c , MPa × m0.5

c, mm

0

80

5.5

Fig. 3. A schematic damage curve for structural ceramics tested by indentation.

160

a, mm

b 1

5.0 4.5

cracks at the corners of the impression; portion II is a “zone of stabilization.” This portion is associated with a controlled, proportional loading, deformation processes, and rupture by indentation. Portion III is associated with an avalanche-like degradation of the material, in fact, inability of the material to sustain fracture in the impression zone (in terms of Hn and K1c ). To reduce the spread in values it was proposed that the radial crack length be measured under conditions that protect the impression from air and moisture for 1 min starting from the instant of indentation (oil was proposed as the protective medium). If in the material tested for fracture toughness by indentation and exposed to air the value of K1c remained unchanged for 24 h, then the tests could be carried out without protection from the environment. To relieve surface stress in specimens intended for testing by indentation, the specimens are recommended to be annealed. For Si3N4 ceramics, the annealing is carried out at 673 – 773 K for 2 h, to avoid softening of the glassy phase. It should be kept in mind that in the Si3N4 ceramics tested for K1c , a relatively large zone of microscopic damage is retained in the test specimen relieved of the indenter load; therefore in conducting replicate testing, the surface layer of the test specimen should be ground off to a depth larger than the microdamage size (1 – 2 mm). The silicon nitride materials intended for testing were prepared by a thixotropic-mixture casting method. Metal molds of dedicated design were used to cast laboratory test specimens of complex shape (grinding bodies in the form of spheres and cylinders, casings for grinding mills, dosing nozzles for aluminum casting) that could be filled with a mixture based on an ethyl silicate bond during the vibratory compac-

4.0

2

3.5 3.0 100 200

300 400

500

Hn ´ 10 – 3 m – 0.5 K1c 2.5

c, mm c

2

2.3

1

2.1 1.9 1.7 1.5 100 200

300 400

500

c, mm

Fig. 4. Hardness Hn plotted as a function of the impression diagonal (a), and stress intensity factor K1c and brittleness index Hn /K1c plotted as a function of the crack length c (b, c) for Si3N4 specimens: 1 ) intact; 2 ) annealed at 700°C.

tion. The initial Si3N4 particle size was 1 – 2 mm. The specimens were annealed at 1550 °C under controlled regime conditions. The x-ray phase analysis of materials was carried out on DRON-3 and DRON-2 diffractometers using Cu and Fe radiation, respectively. MIM-7, MIM-8, MIN-8, and Neophot 21 toolmaker’s microscopes (´ 200 – 2000) were used to examine the structural materials. Thin sections were prepared using diamond polishing pastes and a GOI-type paste. Both etched and intact objects were examined. The microstructure was studied using a JEOL SUPERPROBE 733 instrument (a scanning electron and x-ray electron probe analyzer) at an

TABLE 1. Composition and Properties of Materials Tested Mixture composition No.

Sintering aids SiO2

Si3N4

12 80

Processing conditions

100 100

S-3

S-4

5 –

– 5

7 7

HSC pres- Vibrosure, MPa compaction

0 2.0

– +

Ceramic properties gap , g/cm3

P, %

sb , MPa

K1c , MPa × m0.5

Hn , MPa

2.35 2.90

15.0 2.7

197 225

2.6 5.7

350 980

Silicon Nitride Ceramics Prepared By Vibratory Casting of Self-Reinforced Mixtures

39

g, g/cm3 SAS

Si3N4 powder

Sintering aid Al2O3 + Y2O3

Ethyl silicate binder

2.8

2.4

Grinding Blending

2.0

Mixing

Vacuum degassing

1.8 0

Vibratory casting HSC Drying

Calcination Fig. 5. Flowchart for the synthesis of Si3N4 ceramics on a modified ethyl silicate bond using the vibratory casting technique.

acceleration voltage of 20 kV and a magnification of ´ 40 to ´ 6000. Prior to testing, the specimens were cleaned in acetone by sonication to remove the remains of polishing pastes from material pore; next, a thin gold layer was sputtered in vacuum to create a conducting surface film. Standard methods were used to determine the material properties [12]. The factor K1c was determined by indentation. Three to five series of experimental runs (7 to 11 test specimens) were carried out at a given confidence probability of 0.954; the material hardness was measured using a Vickers diamond pyramid. A standard Vickers TP-7r-1 hardness tester under loads of 50, 100, 150, 200, 250, 300, 350, and 400 N was used. The impressions in the material were sufficiently spaced to prevent stress fields and cracks from interfering with each other. The specimens for indentation tests were polished to surface finish class 12. The object of study was a high-hardness, brittle Si3N4 ceramic prepared from self-reinforced mixtures with an ethyl silicate bond. The composition and properties of the materials tested are given in Table 1. The crack resistance of the materials was determined using a modified Niihara formula K1c = 94.1108P c – 3/2, where P is the load, N; c is the crack length + impression half-diagonal, mm.

1

2

HSC pressure, MPa Fig. 6. Density of the raw material plotted as a function of the HSC pressure: 1 ) for specimens of series NS-3 (), ^, =); 2 ) for specimens of series NS-4 (&, p, :).

Hardness and crack resistance varied with indenter load. In [4, 5], the relationship Hn /K1c versus crack length + impression half-diagonal c was considered and three zones in the damage curve were distinguished. Accordingly, we have plotted K1c , Hn , and Hn /K1c as a function of c using our experimental results; relevant data are given in Fig. 4. As can be seen, three zones — an ascending branch, a plateau, and a descending branch — are clearly distinguished, of which the plateau is a stabilization zone according to [4, 5]. It is precisely the plateau values of K1c that provide a true characteristic of crack resistance [5]. Next, we have considered in what a way processing factors as implicated in the flowchart in Fig. 5 might influence the crack resistance of silicon nitride-based components. The conventional sintering aids Al2O3 and Y2O3 taken in varying ratios in their mixture Al2O3 + Y2O3 were used to improve sintering; their overall percentage did not exceed 10%. The ethyl silicate bond was always taken at the same concentration to prepare the mixture for test specimens. The hydrostatic compression pressure varied from 0 to 2 GPa. Density and porosity of the raw and sintered Si3N4 ceramics containing varying amounts of sintering aid S-3 plotted as a function of the HSC pressure are shown in Figs. 6 and 7. Under the action of HSC pressure, the porosity of both raw and sintered Si3N4 material tended to decrease. Heattreated specimens containing 10% of sintering aid S-3 and compressed under a pressure of 1 GPa exhibited zero porosity (Fig. 7). A petrographic study has shown that in the sintered ceramic specimens subjected to hydrostatic compression, the content of silicon oxynitride tended to increase and that of synthetic b-SiN to decrease. Pressure exerted an effect on the shape of pores. Under a pressure of 2 GPa, all the pores were exceptionally of spherical shape, which, as is well known, improves the strength of the sintered material owing to the

40

G. D. Semchenko et al. a

Hn , MPa

P, % 24

a 1 2

1000 20 16 600

12

3 8

2 200

4

1 0

g,

1

2

g/cm3

b

5

7

K1c , MPa × m0.5

3.2

10

b

2.8 6

2

2.4

1

2.0 4 0

1

2

HSC pressure, MPa Fig. 7. Density (a) and porosity (b ) of sintered Si3N4 specimens containing the composite sintering aid S-3 plotted as a function of the HSC pressure: 1 ) 10%; 2 ) 7%, and 3 ) 5% S-3.

2

5

sb , MPa

b

300

b 200

7

10

SiO2 concentration, %

a

a

100

Fig. 8. Bending strength plotted for Si3N4 specimens containing varying quantities of sintering aid S-3 [1 ) 7%; 2 ) 10%] subjected to HSC pressure of 0 (a) and 2 MPa (b ).

reduced level of stress concentrators. Raising the HSC pressure to 2 GPa causes a more uniform distribution of pores over the matrix, the crystallization of silicon oxynitride tends to increase, and the percentage of glassy phase tends to decrease. The Al2O3 + Y2O3 sintering aid when added to the mixture at concentrations less than 10% shows a less uniform distribution. However, when sintered, its phase distribution

Fig. 9. Hardness (a) and crack resistance (b ) plotted for silica nitride ceramics with varying amounts of SiO2 addition: 1 ) 5%; 2 ) 7%.

improves appreciably owing to the sizeable presence of b-SiC (10 – 15%). Subjected to hydrostatic compressions, all specimens displayed a uniform structure. Tests carried out with varying amounts and proportions of the composite sintering aid have shown that using compositions with a smaller concentration of Y2O3 makes it possible to prepare Si3N4 ceramics of higher density and higher strength omitting the compression of raw materials (Fig. 8). Likewise, in specimens with smaller amounts of Y2O3, the bending strength increases to 380 MPa, that is, by about 30%. Thus, the hydrostatic compression produces a better effect on specimens with a lesser concentration of Y2O3. Such specimens of the same initial density, when subjected to heat treatment, had a higher bending strength (Fig. 8). Further experiments have shown that increasing the concentration of both ultradisperse SiO2 and composite sintering aid makes it possible to prepare a silicon nitride material with a hardness of up to 1 GPa and K1c of about 6 MPa × m0.5 (Fig. 9). The bending strength and strength intensity factor

Silicon Nitride Ceramics Prepared By Vibratory Casting of Self-Reinforced Mixtures sb , MPa

K1c , MPa × m0.5

41 sb , MPa

Hn , MPa

400

1

300

1200

200

800

400

2 300

100

6

2 0

5

600 200 400 100

4

200 0

3

2

1 2 10

20

30

Open porosity, % Fig. 10. Crack resistance (1 ) and bending strength (2 ) plotted as a function of the open porosity of silicon nitride ceramics: ), ^) for specimens of NS-3 series; &, p) for specimens of NS-4 series.

K1c were also measured on specimens with different porosity (Fig. 10). It is seen in Fig. 10 that as open porosity is reduced to 3%, crack resistance of the material increases sharply. The crack resistance plotted against hardness and bending strength of the ceramic material is shown in Fig. 11. Silicon nitride materials with a composite sintering aid of 7% show high values of both hardness and K1c . Varying the Al2O3 /Y2O3 ratio and modifying properties of the sol-gel binder provides a route towards materials with K1c = 6 MPa × m0.5 or even higher. A specific feature of the Si3N4 cast ceramic with a high value of K1c is that its silicon nitride matrix is capable of self-reinforcement with the involvement of nanosize b-SiC particles and coiled whisker crystals b-SiC and a-Si3N4 that fill up round solitary pores [13]. Thus, through varying processing parameters of the molding mixture, one can control the structure and physicomechanical properties, hardness, and crack resistance of silicon nitride ceramics prepared by vibrocasting technology from self-reinforced mixtures with a modified ethyl silicate binder. This technique provides a route towards fabricating components of complex shape using ceramic materials with K1c of up to 6 MPa × m0.5 or even higher. REFERENCES 1. S. M. Barinov, “The crack strength of structural ceramics,” in: Advances in Science and Technology. Technology of Silicates and High-Melting Inorganic Materials, Vol. 1 [in Russian], VINITI, Moscow (1988), pp. 72 – 132.

3

4

5

6

K1c , MPa × m0.5 Fig. 11. Crack resistance (1 ) and bending strength (2 ) plotted as a function of the hardness: ), ^) for specimens of NS-3 series; &, p) for specimens of NS-4 series.

2. G. G. Gnesin, I. I. Osipov, G. D. Rontal’ et al., Ceramic Tool Materials [in Russian], Tekhnika, Kiev (1991). 3. D. K. Shetty, A. R. Roenfield, and W. H. Duckworth, “Indenter flow geometry and fracture toughness estimates for a glass-ceramic,” J. Am. Ceram. Soc., 68(10) 282 – 284 (1985). 4. G. A. Gogotsi and A. V. Basht, “Testing ceramics for hardness by the Vickers indentation method,” Probl. Prochn., No. 9, 49 – 54 (1990). 5. G. A. Gogotsi and A. V. Basht, “Hardness and crack resistance of structural ceramics,” Fiz. Khim. Mekh. Mater., 27(3), 12 – 18 (1991). 6. S. Palmqvist, “Occurrence of crack formation during Vickers indentation as a measure of the toughness of hard metals,” Arch. Eisenhuettenwes, 33(6), 629 – 633 (1962). 7. A. G. Evans and E. A. Charles, “Fracture toughness determination by indentation,” J. Am. Ceram. Soc., 59, 317 (1976). 8. D. G. Bhat, “Comment on elastic/plastic indentation damage in ceramics: the median/radial crack system,” J. Am. Ceram. Soc., 64(11), C165 – C166 (1981). 9. K. Niihara, M. R. Horena, and D. Ph. Hasselman, “Evaluation of K1c of brittle solids by the indentation method with low crack-to-indent ratio,” J. Mater. Sci. Lett., No. 1, 13 – 18 (1982). 10. K. Niihara, D. Ph. Hasselman, and R. Horena, “Further reply to Comment on elastic/plastic indentation damage in ceramics: the median/radial crack system,” J. Am. Ceram. Soc., 65(11), 116 – 118 (1982). 11. A. Niihara, in: 21st Conference on Basic Research of Ceramic Materials. Collection of Research Papers, Fukuda (1983), pp. 59 – 68. 12. Refractories and Refractory Components [in Russian], Izd. Standartov, Moscow (1975). 13. G. D. Semchenko, E. E. Starolat, and N. L. D’yakonenko, “Synthesis of b-SiC and a-Si3N4 whisker crystals from sol-gel compositions,” Steklo Keram., No. 6, 16 – 18 (1997).