A Simple Evaluation of Grain Boundary Diffusion

in Liquid Metal Embrittlement

By Hiroyuki Ichinose*

The extents of grain-boundary diffusion of mercury into 70/30 brass and of 3% Zn amalgam into aluminiumwere evaluated by a simple tensile-test. Without knowing the penetration depth, the data on the diffusion-treated specimens can be analyzed by the equation

where ΔF is the difference in fracture load before and after the diffusion treatment, F0 is the fracture load before

the treatment, a and b are the breadth and thickness of the specimen, k is the parameter for the diffusion path,

D is the diffusion coefficient, and t is the contact time. The diffusion equations obtained are as follows:

The diffusion rate at room temperature, assessed from the results at elevated temperatures, is extremelysmall and the role of diffusion in the degree of embrittlement is not significant in these systems.

(Recieved November 17, 1965)

I. Introduction

Most works on liquid metal embrittlement up to1960 are compiled by Rostoker et al. (1) They statedthat grain boundary diffusion is not an origin ofembrittlement, since its rate, in general, is too small.However, Waterhouse et al. (2), performing theexperiment on 70/30 brass wetted with mercury,suggested that dissolution of zinc from grain-boundariesinto mercury accelerates the degree of embrittlement.From a work on the Cu-2% Be alloy in the presenceof a liquid 2% sodium amalgam, Rinnovatore et al. (3)found that the contact time of the liquid metalwith the solid is a factor affecting the degree ofembrittlement.

Then, the extent of grain boundary diffusion of aliquid metal has to be determined. If the embrittlingliquid is in the solid state at room temperature, thediffusion depth is easily determined by ordinarymetallographic techniques, as done by Bishop et al. (4)Since mercury which is often used as the embrittlingliquid is -still in the liquid state at room temperature,the ordinary techniques cannot be adopted.

The purpose of this paper is to determine the degreesof grain boundary diffusion of mercury into 70/30brass and of 3%Zn amalgam into aluminium by a

simple tensile test and to assess the role of diffusionin embrittlement.

II. Materials and Experimental Procedures

70/30 brass was prepared from 99.99% copper,and99.99% zinc. From a sheet of brass, 1mm thick,which had been cold rolled by 11%, ordinary tensilespecimens were obtained. The gauge section was 20mm in length and 15mm in breadth.

99.99% aluminium was also used. From a 3mmthick sheet of aluminium which had been cold rolled by88%, tensile specimens with the gauge section 20mmin length and 20mm in breadth were obtained.

To control the grain size, annealing was performedin vacuum at 500℃ for 1hr for brass and at 400℃

for 2hr for aluminium, respectively. In order toavoid thermal stress, the specimens were slowly furnace-

cooled. The grain sizes thus obtained were 0.053 and

0.20mm, respectively.The brass specimens were chemically polished in

a mixed solution of H2SO4, HNO3, and HCl for 1 minand the aluminium specimens were chemically polished

hl a solution of H3PO4 and H2O2 at 90℃ for 1 min.

These specimens were washed with water and alcohol,and were dried.

The gauge section of the brass specimen was wettedwith mercury, and that of the aluminium specimenwith 3%Zn amalgam. After that, the gauge sectionswere immersed in the mercury and amalgam pools inan Isolite brick, respectively. The assembly washeated at a temperature between 150゜and 300℃ for

2 to 72hr.After the diffusion treatment, the specimens as-

wetted with the liquids, were tested by an Olsen-

* Department of Metallurgical Engineering, YokohamaNational University, Yokohama, Japan.

(1) W. Rostoker, J.M. McCaughey, and H. Markus:Embrittlement by Liquid Metals, Reinhold (1960).

(2) R.B. Waterhouse and D. Grubb: J. Inst. Metals, 91(1962-63), 216.

(3) J.V. Rinnovatore, J.D. Corrie and H. Markus: Trans.ASM, 57 (1964), 474.

(4) G.H. Bishop, B.F. Addis, C.A. Steidel, and C.W.Spencer: Trans. AIMS, 224 (1962), 1299.

Trans. JIM 1966 Vol.7

A Simple Evaluation of Grain Boundary Diffusion in Liquid Metal Embrittlement 57

type tensile machine at a cross-head speed of 10.8

mm/min. that corresponds to the strain rate of 6.0×

10-3sec-1.

III. Experimental Results and Discussions

The effect of the contact time on the fracture stress

at 20℃ is shown in Fig.1. The fracture stress remains

constant. The fracture laods on the brass andaluminium specimens were 223 and 172Kg,respectively.

Fig. 1 Plot of the fracture stress as a function of the

contact time at 20℃.

Fig. 2 Plot of the fracture load as a function of time on

brass diffusion-treated.

The results on the brass specimens at elevated

temperatures are shown in Fig. 2. Each pointrepresents the average value of 5 specimens. The

decrease in fracture load is obvious with time. Such a tendency was also observed experimentally

by Rinnovatore et al(3). And, this was attributed

to the grain boundary grooving. However, thedecrease in fracture load with time can hardly be

explained only by grooving. On the other hand, inthe work on aluminium alloy wetted with amalgam,

Rostoker et al. (1) reported the absence of notch-sensitivity to fracture stress.

This conclusion may be extended to the absence ofnotch-sensitivity at grain boundaries, since some of

them are met with the root of the notch in the

polycrystalline specimen, if the grain size is not coarseenough.

Although the function of liquid metals in embrit-tlement has not been established, there are severalreasonable proposals: e.g., decreasing of the surfaceenergy of the solids(') and weakening of the bondstrength of the stressed solids(5)-(9). At any rate,it seems neccessary that the liquid metal atom doesnot bind strongly with the solid atom.

The present interpretation on the decrease infracture load with time is based on the following idea:1) The notch sensitivity at grain-boundaries may beneglected, 2) the mechanical strength of the diffusionzone is far weaker than the other parts of the specimen,and 3) the parabolic diffusion law can be applied.

Although the penetration depth depends largelyupon the degree of misorientation, our interest isconfined to the favourably oriented diffusion path,since it gives a maximum decrease in effective area tosupport the load. Then the fracture stresses before andafter the diffusion treatment may be equal and thecondition is expressed as follows:

(1)

where F0 and F are the fracture loads before and afterthe diffusion treatment, a and b are the breadth andthickness of the specimen, and x is the depth measuredin the direction perpendicular to the specimen surface.If x is not large compared with a or b, Equation (1) issolved with respect to x.

(2)

where ΔF=F0-F.

On the other hand, the parabolic diffusion law isexpressed in the following way:

(3)

where k is a parameter for the zigzag diffusion path inthe polycrystalline specimen. Strictly speaking, D inEquation (3) represents an order of magnitude, sincethe concentration of the liquid metal at the distanceof x perpendicular to the specimen surface is unknown.But the value of the activation energy obtained fromsuch diffusion coefficients at different temperaturesis absolute.Inserting Equation (2) into Equation (3), the followingexpression is obtained:

(4)

(5) N.S. Stoloff and T.L. Johnston: Acta Met., 11 (1963),251.

(6) A.R.C. Westwood and M.H. Kamdar: Phil. Mag., 9(1963), 787.

(7) E. Levine, H. So onion, and I. Cadoff: Acta Met.,12 (1964), 1119.

(8) E. Levine and I.B. Cadoff: Trans. AIMS, 230 (1964),1116.

(9) A.R.C. Westwood, D.L. Goldheim, and E.N. Pugh:Acts, Met., 13 (1965), 695.

58 Hiroyuki Ichinose

Fig. 3 Plot of the square of the decrease in fractureload against time for brass.

Fig. 4 Plot of the square of the decrease in fracture

load against time for aluminium.

Fig. 5 Plot of the diffusion coefficients in logarithmic

scale against the reciprocal of temperature.

Following Equation(4), the data on brass are shown in

a form of (ΔF)2 versus t in Fig.3. The expected

straight-line relation holds good. The slopes of thestraight lines are related to the diffusion constants ofthe liquid. The data on aluminium are shown in Fig.4. The values of k for the brass and aluminium speci-mens were determined from the photomicrographs tobe 1.6 and 1.3, respectively. The diffusion constantsobtained from the slopes and Equation (4) are repre-sented in Fig. 5, according to the conventional plot.The obtained activation energy and frequency factorare summarized in Table 1.

Table 1

Using: Equation (4) and the diffusion coefficients at

20℃, obtained by the egtraporation of the Arrhenius

equation, the contact time necessary to produce a

certain amount of decrease in fracture load at 20℃

can be calculated. Examples are listed in Table 2.

Table 2

In the case of brass, it requires 35 days to producean 1% decrease in fracture load, i.e., only 2.23Kg.This prediction is in good agreement with the observa-tion(10) that during the contact time of 90 days thefracture stress of brass wetted with mercury remainedconstant.

IV. Summary

The extent of grain boundary diffusion in thesesystems was evaluated by a simple tensile test.

The diffusion rate at room temperature, which wasobtained by the extraporation of the data at elevatedtemperatures, can be neglected. Then the role ofdiffusion in the magnitude of embrittlement is notsignificant in these systems.

AcknowledgementThe author wishes to acknowledge the earnest

assistance of Mr. T. Takeda and Mr. E. Iguchi in theexperiment.

Thanks are also due to Dr. Y. Hori of Nikko CopperRefining Plant, F rukawa Electric Co. for his kindhelp in preparing the specimens.

(10) H. Moore, S. Beckinsale and C.E. Mallinson: J. Inst.Metals, 25 (1921), 35.