Subsurface Evaluation of Ground Ceramics

Bi Zhang, Trevor D. Howes (2) University of Connecticut, USA Received on January 10,1995

Subsurface damage evaluations are performed using a combination of scanning electron microscopy and damage inspection techniques on silicon nitride, alumina, and silicon carbide ground by diamond wheels with various grit sizes. Two forms of subsurface damage are identified, material pulverization in the superficial layer forming a powder regime and void generation in the subsurface layer. Both forms of damage are assessed and characterized with respect to type of workpiece material and grit size of grinding wheel. The penetration depth of subsurface damage is shown to be predictable using an analytical model

b e : Ceramics. grinding, damage, diamond

1. INTRODUCTION

In a previous paper [l], the authors reported that pulverization is an important component of material removal mechanism in grinding of ceramics. In this paper, subsurface damage will be assessed, and shown to be predictable using experimental and analytical methods.

Subsurface damage can be a function of numerous factors including material properties, grinding conditions, grinding wheel parameters, and grinding wheel truing and dressing conditions. It was reported that the brittleness, which is defined as the ratio of hardness to toughness H I Kc of a material [2], was a governing factor affecting the penetration depth of subsurface damage. In 1977, Lawn and Evans [3] proposed a critical condition for the initiation of subsurface damage:

Lawn and Evans' critical condition was extended to a grinding situation by Bifano eta/. who advanced a model to correlate the critical grit depth of cut with material properties. The model is described by a simple power-law equation [4]:

4 = O.l5($(+g

Under a given grinding condition, subsurface damage is solely determined by material properties, predominantly by brittleness.

This paper proposes an analytical model for the prediction of subsurface damage of ceramics subjected to grinding. The model correlates subsurface damage to the type of workpiece material and grinding parameters. Three techniques [l] are used to assess the subsurface damage of ground hotpressed silicon nitride and alumina, and pressureless sintered silicon carbide. The results of damage prediction using the analytical model are verified by the experiments,

2 . DAMAGE PREDICTION MODEL

Grinding damage is induced by grinding forces, predominantly the normal force. The normal force on a single grit is proportional to the maximum grit depth of cut that can be expressed as [5]:

where a is the ratio of the normal to tangential grinding forces, and is a constant for a given grinding condition; KO and 8 are constants to be determined by the material properties of a workpiece; and a, is the maximum grit depth of cut that is given by the following equation:

I

(4)

where a,, d,, v,, and v, are wheel depth of cut, wheel diameter, workpiece speed, and wheel surface speed; C the grit surface density that is defined as the number of active points per unit area of the wheel surface, and r is the ratio of chip width to average undeformed chip thickness. On the other hand, a correlation between the subsurface damage depth and externally applied load can be obtained [6]:

F, = 6'" (5)

The exponent m is in the range of 112 S m S 3 1 2 for different loading conditions. By substituting equation (5) into equation (3). we then obtain

6 = (Ya,)'

in which v and Y are factors to be determined by material properties and grinding conditions. It is assumed that v is predominantly a function of material properties, and Y a function of grinding parameters. The model in the form of equation (6) needs to be verified by grinding experiments, and the factors \y and Y are to be determined by experiments.

3 . EXPERIMENTAL PROCEDURES

Silicon nitride, alumina, and silicon carbide were used in the experiments. Silicon nitride and alumina were prepared using the hot-press technique, while silicon carbide by the pressureless sintering technique. Table 1 lists the physical properties of these materials. They were sliced into bars with dimensions of 5x5~30 mm. Three rectangular surfaces of each specimen were polished with 15 pm diamond paste to remove possible damage induced by the previous procedures. The non- polished rectangular surface was bonded to the surface

Annals of the ClRP Vol. 44/1/1995 263

Si3N4 Density, s/cm3 3.28

Vickers hardness, GPa 17.9 Fracture toughness, MN/mm 6.6 Flexure strength, MPa 800 Brittleness, /mln 2,700

Grain size, pm 2-3

A1203 Sic 3.94 3.15 2-3 5 18.3 31.0 5.3 3.6 500 490

3,400 8,300

I pn I gritdmm2 I P SD80-N50M I 230 I 3.1 0.91

Mean Diamond wheel grit size,

Grit surface Maximum grii density,- depth of cut,

4. EXPERIMENTAL RESULTS

SD140-N75M SD400-N75M

a. Pulverization Grains in the surface layer of a ceramics can be pulverized into much finer ones (submicron sizes or smaller) by grinding through intergranular andor transgranular microcracking, and form a powder regime [l]. Fig.1 gives the cross-sectional views of a silicon nitride specimen ground by the SD80- N50M wheel. A powder regime was observed in the ground surface, and vanished from the ground surface as a resutt of etching. The powder regime readily dissolved by etching was approximately 4 pm in thickness, and was loosely connected, allowing a good penetration for etchant, and facilitating a quick dissociation.

The thickness of a powder regime varied along the ground surface due to the non-uniform distribution of abrasive grits on a wheel surface. The powder regime thickness was measured over a cross-section of 500 pm that was divided into 10 identical parts, each having a width of 50 pm. The maximum, the minimum, and the ten- point average thicknesses were obtained and are plotted

100 6.3 0.64 40 17.4 0.38

in Fig.2. The SD80-N50M wheel generated the largest amount of powder regime on silicon nitride and alumina which was followed by the SD140-N75M and SD400-N75M wheels. The experimental results reveal that a grinding wheel of a larger grit size generates more powder regime than does that of a smaller grit size. The thickness of the powder regime generated by the wheel with an average grit size of 230 pm was more than three times of that by the wheel with an average grit size of 40 pm. Accordingly, the number of active cutting points per unit area of the SD400-N75M wheel was 5-6 times that of the SD80-N50M wheel, resulting in a large difference in the grit depth of cut. The grit depth of cut for the SD400-N75M wheel was less than one-third of that for the SD8O-N50M diamond wheel. The larger the grit depth of cut, the more the powder regime, and vice versa.

Fig.1 Micrographs showing cross-sections of silicon nitride ground by the SD80-N50M wheel.

I - Maximum -Average - Minimum n

V 40 100 230

Mean Grit Size, prn

Fig.2 Thickness of powder regime vs mean grit size.

The amount of powder regime generated on silicon nitride was the largest, followed by alumina, and silicon carbide using the same grinding wheel. This suggests that brittleness play a significant role on grain pulverization, and retention on a workpiece. Because of the highest brittleness value, silicon carbide, exhibited chipping on its ground surfaces. The chipping level increased with the grit size of the diamond wheels tested. It was noticed that when chipping occurred it carried away much of the powder regime.

264

of a steel holder with araldite adhesive. The grinding wheels were metal bond SD80-N50M, SD140-N75M, and SD400-N75M wheels with grit mesh sizes of #60-80, #140- 170, and #325-400. For comparison purposes, the blocky diamond grits (MBG-II supplied by GE) were used for the diamond wheels. The grinding experiments were carried out on a precision surface grinding machine at a grinding wheel speed of 1,600 mlmin, a table speed of 1.4 mlmin, a wheel depth of cut of 15 pm. and a grinding width of 5 mm. The grinding coolant was soluble type (JIS W-2-2) 2% dilution. Truing of the wheels was performed by a cup- type diamond wheel SD140-Pi 00M-30, and dressing by an alumina stick WA150G1 OV. The truing was conducted at a grinding wheel speed of 400 mlmin. a truer speed of 60 rdmin, a table cross feed of 0.1 mm per truer revolution, and truing feedrates of 3 pmlstroke, 1 pmlstroke. The dressing was done at a grinding wheel speed of 400 mlmin. In order to stabilize the surface grits of the grinding wheels after the truing and dressing processes, each grinding wheel performed grinding on a zirconia block (5x1 15x80 mm) at a wheel depth of cut of 10 pm to a cumulative removal of 6 mm in thickness. After grinding tests, each specimen was inspected by the etching, taper polishing and fracture techniques for subsurface damage information [l].

Table 1 Physical Properties of the Ceramic Materials

Density, s/cm3 Grain size, pm Vickers hardness, GPa Fracture touahness. MN/mm

Si3N4 3.28 2-3 17.9 6.6

Flexure stre;gth, MPa Brittleness, /mln

- A1203 3.94 2-3 18.3 5.3 500

3,400 - 800

2,700

- SIC 3.1 5

5 31 .O 3.6 490

8,300

The grit surface density of the grinding wheels was measured by SEM to account the numbers of the wear flats generated by a grinding process. The value of r was assumed to be equal to 15 [A. Table 2 gives the values of the mean grit size and grit surface density and the maximum grit depth of cut of the diamond wheels.

b. Voids Grinding induced voids were found in the subsurface layers of all the ceramics. The void inspection was performed using SEM combined with the etching technique. Fig.3 shows a micrograph of a silicon nitride workpiece ground by the SD80-N50M wheel. The clustered voids align themselves in the grinding direction, forming void lines. Void clusters and scattered voids can also be seen. The densest distribution of voids were found to be adjacent to the powder regime layer of a ground surface. The size of a void was about 1-2 pm in diameter.

Fig.3 Voids induced in silicon nitride by the SD80-N50M wheel. The observation was made in the subsurface by 22 pm in depth.

A large number of voids was also observed in the subsurface of ground alumina, and silicon carbide. Both clustered and scattered voids were generated in silicon nitride, and alumina when the SD140-N75M and SD80- N50M wheels were used for grinding, however, only scattered voids were formed in these materials when the SD400-N75M wheel was used. Mainly scattered voids were found in the subsurface of silicon carbide regardless of the grinding wheels. These observations indicate that whether clustered or scattered voids form depends on the brittleness of workpiece materials, and the grinding parameters. A material with a high brittleness value. e.g. silicon carbide, tends to have more scattered voids, while a material having a low brittleness value, e m silicon - - nitride, tends to have more clustered voids.

I

r t

40 100

Mean Glit Size, prn 230

Fig.4 Depth of void distribution vs mean grit size.

The maximum depth of void distribution was quantitatively measured against the grit size of the grinding wheels used and is shown in Fig.4. The measurement was done over an area of 500x500 pm of a subsurface for each workpiece. Silicon nitride had the

largest penetration depth of voids among all the ceramics under the same grinding conditions. In using the SD80- N50M wheel, the depth of void distribution was as large as 40 pm in silicon nitride, and less in alumina, and silicon carbide. Wth a decrease in grit size, there was found to be a decrease in the depth of void penetration for silicon nitride and alumina, but very little decrease for silicon carbide. This result indicates that for a workpiece material with a large brittleness value, a coarse wheel does not cause a larger penetration depth of voids than a fine wheel does, however, this does not hold true for a workpiece material with a low brittleness value. The coarser the grit is, the deeper the void penetration will be for a material with a low brittleness value. The brittleness of these materials also plays an important role on penetration depth of subsurface damage.

c . Model Verification It should be noted that two factors y and Y are correlated with material properties, grinding wheel parameters, grinding, truing, and dressing conditions. Although the influence of grinding, truing and dressing conditions on the two factors is not known, these factors can be determined based on the experimental results shown in Figs 284:

Y = 200 and y = l/log(JJY/ K,)

where d = m112 is a constant. By substituting the factors y and Y into equation (a), the proposed model then become

(7)

Equation (7) correlates the damage depth with the brittleness of ceramics, and the grit depth of cut of a wheel. The relationship illustrates that a low level of damage depth is obtained in a material with a high brittleness value, while a high level of damage depth in a material with a low brittleness value. It should be made clear that other material properties such as grain size and multiphases should also give an influence on damage depth.

5 . DISCUSSION

A model has been proposed for predicting subsurface damage. Based on this model, variations of damage depth with the maximum grit depth of cut and brittleness of workpiece materials are plotted in Figs 586. The predicted damage depths show good agreement with those of the experiments. According to this model, a. damage depth exponentially increases with

increasing grit depth of cut. b. damage depth increases with a decrease of

material brittleness. c. as the grit depth of cut decreases, the difference in

damage depth among these materials tends to disappear.

Referring to the SEM observations concerning voids: a question remains why a more brittle material such as silicon carbide has a larger number of scattered voids with a relatively small penetration depth, and a less brittle material such as silicon nitride has a smaller number of scattered voids with a relatively large penetration depth. To answer this question, we may use equation (1) for critical damage size, and the following equation for the minimum load to propagate a critical damage [3],

Using equations (1) and (8) with the hardness and toughness values obtained from the ceramic materials used in the experiments, the values of the critical damage depth and critical loads to propagate subsurface damage

265

were obtained and are listed in Table 3 for the ceramics relative to silicon nitride. An interesting finding is that only a surprisingly small load is required to produce a critical damage in a material with a high brittleness value. For silicon carbide, a 2% load is needed to produce a critical damage of 10% in size with respect to silicon nitride. This is to say that for a given grinding condition, the population of critical damage would be much larger, but the depth of the damage would be much smaller in silicon carbide than in silicon nitride. Moreover, although the more brittle material shows less depth of damage, the number of cracks or voids is much greater than is the case for the less brittle material.

Eic- I 8,300 I 0.1

50 1 1

0.02

1 0 Si,Nq (experimental) I 0 AI,O, (experimental) 0 sic (experimental)

10

0 r l l l l l l l ( l ' l l l l ' ( l l l ' 0.0 0.2 0.4 0.6 0.8 1.0

Grit Depth of Cut, pm

Fig.5 Damage depth vs maximum grit depth of cut.

60 1 hg=O\gl 0 Si,N4 (experimental) I 0 AI,O, (experimental) 8 Sic (experimental)

5 50 40 k 0 . 6 4 k

G

3 a 20

n

$ 30

E

10

0 10 100 500

Briileness WK, , x 10 /m In

Fig.6 Damage depth ys brittleness of ceramic materials.

It should be pointed out that equations (1) and (8) were derived for predicting the median cracks formed in the process of Vickers indentation. Nevertheless, Veldkamp el a/. [8] demonstrated that the critical scratching load to initiate a crack complied with the same expression as equation (8) if the scratching hardness was substituted for the indentation hardness. The model for predicting critical scratching load developed by Veldkamp el a/. underlies the basic processes of deformation and

Table3 Relative Values of Critical Damage Sizes and Critical Loads to Propagate Subsurface Damage

A1;OA I 3,400 I 0.65 I 0.4

fracture in both static and scratching conditions.

Based on the discussion, it has become clear that for a given grinding condition, hardness and fracture toughness are the predominant factors in determining whether subsurface damage contains a large number of scattered voids with relatively small penetration depth as in the case of silicon carbide, or a small number of clustered voids with relatively large penetration depth as in the case of silicon nitride. The former may indicate that the energy given by an abrasive grit into a workpiece is dissipated in a distributed manner, while the latter in a concentrated manner. In the latter case, voids may be nucleated at a low level of dislocation pile-ups due to the high index level of brittleness of the material. As a consequence, this would relax the shear stress exerted by a grinding grit at a relatively short range, generating voids to a small penetration depth but a large number. In the latter case, the low index level of briileness of the ceramic material would bring a dislocation pile-up to a higher level before a void could be nucleated. The shear stress is not released until a certain level of stress concentration is reached, which results in a stress relaxation at a relatively long range. This process would result in clustered voids in a large penetration depth but a small number.

6 . CONCLUSIONS Subsurface damage has been evaluated on silicon

nitride, alumina, and silicon carbide ground by diamond wheels with various grit sizes. As a result, two types of Subsurface damage are identified, powder regime in the superficial layer, and voids in the subsurface of ground ceramics. Clustered voids are induced in the less brittle materials, while scattered voids occur in all the ceramic materials tested. A model is proposed for predicting grinding induced subsurface damage, in which damage depth is a function of brittleness of ceramics, and the maximum grit depth of cut. The model predicts that the penetration depth of subsurface damage increases with a decrease of material brittleness, and with an increase of the maximum grit depth of cut. The model shows good agreement with the experimental results.

REFERENCES

1. Zhang, B. and Howes, T.D., 1994. Material Removal Mechanisms in Grinding Ceramics, Annals of the

2. Lawn, B.R., 1993, Fracture of Brittle Solids - Second Edition, Cambridge University Press, 249-306.

3. Lawn, B.R., and A.G. Evans, 1977, A Model for Crack Initiation in ElastidPlastic Indentation Fields, Journal of Materials Science, 12:2195-2199.

4. Bifano, T., et a/., 1991, Ductile-Regime Grinding: A new Technology for Machining Brittle Materials, ASME Journal of Eng. for Industry, 1 13:184-189.

5. Kawamura, S., et a/., 1984. Grinding and Abrasive Machining, Kyoritsu Shuppan, Tokyo, 38-46.

6. Inasaki, I., 1987, Grinding of Hard and Brittle Materials, Annals of the CIRP, 36:l-9.

7. Kalpakjian, S.. 1989, Manufacturing Engineering and Technology, Addison-Wesley, 757-758.

8. Veldkamp, J.D.B. el a/., 1983, Crack Formation During Scratching of Brittle Materials, Fracture Mechanics of Ceramics, Vo1.5:273-301.

CIRP, 43:305-308.

266