Crack nucleation in ultrathin blown silicate glasses

Journal of Non-Crystalline Solids 80 (1986) 481-486 North-Holland, Amsterdam

CRACK NUCLEATION GLASSES

IN ULTRATHIN

481

BLOWN SILICATE

T. T O N I N A T O 1, R. D A L L ' I G N A 1, V. G O T T A R D I 2, R. D A L M A S C H I O 2, S. G A T T A z and G. S C A R I N C I 2 t Stazione Sperimentale del Vetro, Murano-Venezia, Italy 2Istituto di Chimica lndustriale, Facolta di lngegneria, Universit~ di Padova, Italy

Ultrathin blown glasses, owing to the very high cooling rate, have an open structure, characterized by lower hardness and higher deformability, and no tempering residual stresses. In these glasses, due to their high densificability, the threshold load for crack nucleation is higher than in air-cooled plates, and it is also above the values estimated by present theoretical formulations.

1. Introduction Investigations into stress and fracture mechanics in glass have been intensified in the past few years [1-6] by the application of microhardhess techniques and microscope analysis of the cracks produced on indentation. While a n u m b e r of studies have explained the growth and propagation processes of different kinds of cracks (median, lateral or radial) in several materials, research p e r f o r m e d on the problem of their nucleation has not led to any definite conclusion [7-9]. In this paper, the authors seek to assess the influence of eventual structural changes occurring in a glass of given composition on the formation of cracks, particularly on deformability and on the critical threshold load. T h e simplest way to obtain structural changes is by varying the cooling rate, for instance by tempering, but it is well known that this procedure leads to the formation of a field of residual stresses in the material, namely of compressive stress on the surface and tensile stress in the core. In order to evaluate the role of structural changes in affecting the p a r a m e t e r s c o n n e c t e d to deformation and fracture, therefore, it is necessary to separate the effects due to structure f r o m the concomitant effects due to the field of residual stresses. T o fulfill this requirement, glass samples were p r e p a r e d by vigorously blowing a glass drop with a metal pipe. Owing to the high cooling rate, the samples thus obtained exhibit an open structure, but, due 0022-3093/86/$03.50 ~) Elsevier Science Publishers B.V. (North-Holland Physics Publishing Division)

482

T. Toninato et al. / Crack nucleation in ultrathin glasses

to their extremely low thickness (less than 100 microns) they solidify without allowing any considerable thermal gradient to arise in the wall, so that these samples should be practically free from residual stresses. For blown glass samples of 7 microns thickness, Charles [10] recorded a density variation of about 0.57% as compared to thick glasses cooled in a lehr. The higher fictive temperature achieved by heat treatment increases free volume and leads to a more open and uniform arrangement of the silica network. With a view to emphasizing the structural effects due to this reduction in density, the foregoing samples were compared with annealed blown specimens (in which structure compaction occurs) and with plate samples, both annealed and cooled in calm air after casting. All of the samples had the following weight composition: SIO2---73%, Na20 = 17%, C a O = 6 % , B203 = 2%, msaO5 = 1%, other oxides = 1%. The samples were prepared in the shape of plates of 5 mm thickness cooled in air (sample no. 1), or annealed from 550°C (sample no. 2), and in the shape of thin blown pieces ~85 microns thick (not annealed - sample no. 3 - or annealed from 550°C sample no. 4 - ).

2. Experimental Since microindentation with a Vickers indenter requires fiat, rigid surfaces, the extremely light and flexible blown samples had to be previously mounted on microscope slides using a suitable adhesive. A number of experiments were carried out with different kinds of adhesives in order to limit the eventual negative effects caused by the high surface tension of the adhesive that tended to produce a tensile stress in the sample side to be indented, thus making the crack length vary from one point to another. A substantial crack length uniformity, which means absence of stress, was finally achieved by using square blown glass samples of reduced size (5 mm) and appropriate thickness (> 50 ~m). Beside the foregoing considerations, the thicknesses adopted for the blown samples were selected taking into account that in the case of too low thicknesses the impression depth would be of the same order of magnitude as the sample thickness itself, whereas for too thick samples the cooling rate would be lower than required. The indentations were made in air, under constant temperature and humidity conditions, with a fall and contact time of the indenter of 15 s; the crack length was measured by an optical transmission microscope about half an hour later. The number of cracks starting from the corners of each impression was also evaluated. Loads ranging between 0.25 and 5.00N were applied, performing 10 impressions for each selected load. On the basis of the measurements carried out, the P / c 3/2 ratio, which is proportional to the critical stress

T. Toninato et al. / Crack nucleation in ultrathin glasses

483

Table 1 Samples

P/c 3/2 (MPa m 1/2)

n v (GPa)

1 2 3 4

11.677 9.692 10.032 9.658

5.52 5.53 5.09 5.54

+ ± ± ±

0.425 0.603 0.425 0.647

+ ± ± ±

0.0197 0.1615 0.049 0.1772

intensity factor, was calculated by the weighted standard deviation method, whereas the Vickers microhardness was evaluated by a weighted average calculation. The results obtained are shown in table ! and represented in figs. l and 2 with their standard deviations.

5.7

5.8

Hv

5.4

5.3

5.2

5.1

5.0

,;

I

2 Samples

Fig. 1. Vickers hardness.

i

3

i

4

484

T. Toninato et al. / Crack nucleation in ultrathin glasses

P / c ~' MPa

. ,m

12.0

11.0

t t

10.0

9.0

8.0

I

1

I

2

I

3

I

4

Samples Fig. 2. P/c 3/2 ratio, proportional to the critical stress intensity factor for all samples except sample no. 1.

Finally, the threshold load required for radial crack nucleation was determined experimentally and calculated theoretically. According to Dabbs and Lawn [11], the threshold load can be defined as the load value at which cracks appear during indentation with a probability of 50%. T h e best graphical representations of this p h e n o m e n o n are the so-called "S-shaped curves" [12] obtained by plotting the crack formation percentage, calculated with respect to a theoretical maximum of four per each impression, as a function of the applied load. These curves were shown to be most useful in characterizing glass surfaces resulting from different manufacturing processes or submitted to different mechanical, chemical or heat treatments [13]. T h e S-shaped curves obtained experimentally for the four kinds of samples examined are reported in fig. 3: the threshold load corresponds to the point on each curve at which the percentage of crack formation is 50%. These data, which are reported in table 2, were compared with the corresponding values calculated by means of the equation proposed by Lawn and Evans, into which /30 was introduced as a constant [7]. Table 2 also gives the value of the other critical parameter, the crack

485

T. Toninato et al. [ Crack nucleation in ultra&in glasses

% crack formation

/ / 7 / I / // / *

9O

70.

50,

/

10

,,a

fl

/

/

/®

/®

®i,l ®

!

i1,1I"

/,

7/

I / "1.0

f

/

--'"-IF" 30,

/

//

I

...,/ /

/

/

.1/

2".O

Fig. 3. S-shaped curves for the different samples.

Table 2 Samples

(~oKic/Hv)3[.~oKk c~ P*(N)

([3oKlc/Hv) 2 ~ C*(/xm)

P'exp. (N)

1

a}

~)

1.85

2 3 4

52.2 76.7 51.2

3.07 3.88 3.04

0.70 3.50 0.85

~) These values are not here because, for tempered glasses, the relation is P / c 3/2= /3oK,c[l + O'a(crtoc)'/:/g,c] [14].

length, which was also calculated by the model proposed by Lawn and Evans.

3. Discussion and conclusions

Fig. 1 clearly indicates that, while the hardness values for air-cooled and annealed plate samples are nearly equal, the same value is markedly lower (about 10%) for non-annealed thin blown samples as compared to annealed ones. It may be concluded that non-annealed blown samples are less resistant to indentation than annealed ones; this behaviour may be ascribed to the more open structure, which exhibits higher fictive temperature and lower viscosity and density, and which originated from the high cooling rate (and eventually from a solidification under stress) [15]. The hypothesis cannot be rejected that structural diversity alone, independently of the presence of stress, may affect the susceptibility to cracking of the samples. The S-shaped curves reported in fig. 3 indicate a markedly different susceptibility to cracking of non-annealed blown samples as compared to all

486

T. Toninato et al. / Crack nucleation in ultrathin glasses

other samples. The threshold load for crack formation is much higher for non-annealed blown glasses. It is interesting to observe that the threshold load appears to be higher for the plate sample cooled in air than for the annealed one; in this case, however, the presence of stresses also plays a role. The theoretical values of the critical threshold loads (table 2) calculated by the Lawn and Evans equations do not reflect, with as much clarity as experimental values, the behaviour of non-annealed blown glasses with respect to the other samples. Therefore, it can be concluded that the formulae available at present, even if fundamentally effective, do not yield values quite comparable with experimental ones. As regards the P / c 3/2 ratio, which is proportional to the critical stress intensity factor, it appears to be fairly similar for all samples, except for tempered olate samples, in which case the stress field originated by cooling in air must be taken into account. The tact that the P [ c ~z ratio for non-annealed blown samples closely approximates the values for annealed ones, while it is rather distant from the value of the tempered plate sample, may confirm that eventual mechanical stresses due to rough cooling are quite neglectable. The difference between the critical load values determined experimentally for annealed and non-annealed blown samples seems, therefore, ascribable mostly to the effects of the different structures. Hence, the conclusion can be drawn that the critical threshold load value is affected by the type of structure resulting from cooling methods. This influence seems instead to be less important as concerns crack growth phenomena.

References [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12]

B.R. Lawn and R. Wilshaw, J. Mater. Sci. 10 (1975) 1049. B.R, Lawn and M.V. Swain, J. Mater. Sci. 10 (1975) 113. C.M. Perrot, Wear 4a (1977) 293. D.B. Marshall and B.R. Lawn, J. Mater. Sci. 14 (1979) 2001. T. Haranoh, H. Ishikawa, N. Shinkai and M. Mizuhashi, J. Mater. Sci. 17 (1982) 1493. S. Chiang, D.B. Marshall and A.G. Evans, J. Appl. Phys. 53 (1982) 298. B.R. Lawn and A.G. Evans, J. Mater. Sci. 12 (1977) 2195. J.T. Hagan, J. Mater. Sci. 14 (1979) 2975. S.S. Chiang, D.B. Marshall and A.G. Evans, J. Aopl, Phys. 53 (1982) 312. R.J. Charles, J. Amer. Ceram. Soc. 45 (1962) 105. T.P. Dabbs and B.R. Lawn, J. Amer. Ceram. Soc. 65 (1982) C-37. M. Wada, H. Furukawa and K. Fujita, Proc. 10th Int. Congress on Glass (Ceram. Soc. Japan, Tokyo) 11 (1974) 39. [13] T. Toninato, L. Salvagno, G. Scarinci, G. Michelotto and R. Dal Maschio, Riv. Staz. Sper. Vetro 9 (1979) 1. [14] D.B. Marshall and B.R. Lawn, J. Amer. Ceram. Soe. 61 (1978) 21. [15] T. Toninato, G. Cogoni, V. Gottardi and G. Searinci, J. Amer. Ceram. Soc. 63 (1980) 180.