568 The Liquidus Curves of the Ternary System:\#151 ; AluminiumCopper-Tin .
By J. H. Andrew , M.Sc .
, Dalton Scholar , and 0 .
A. Edwards , Besearch Fellow and Demonstrator in Metallurgy , Victoria University , Manchester .
( Communicated by J. E. Stead , F.R.S. Received May 3 , \#151 ; Read May 27 , 1909 .
) [ Plate 3 .
] The study of the constitution of alloys is of great theoretical interest , and the practical use of the knowledge so obtained can scarcely be overestimated ; in fact , it may be said , that the heat-treatment of a given series of alloys cannot be correctly accomplished without an accurate knowledge of the structural changes which occur with varying temperature and concentration .
So much work has been done on this subject , that we are now in possession of accurate data for a large number of the alloys containing only two metals , but with the exception of Mons. Charpy's* classical work on the lead-tin-bismuth alloys , comparatively little work has been published on mixtures of three or more metals .
This is not surprising when we consider the increased analytical and experimental difficulties , and also the possible number of alloys in a series containing three elements .
If relatively the same number of determinations are required for a ternary as for a binary series of alloys , then ny.\n will be the number required for the ternary when n is the number for the binary series .
In other words , if 30 alloys ( which is a moderate figure ) are necessary to determine the melting points and structural changes of a series of alloys containing two elements , 450 would be needed for a series containing three elements .
At first sight these figures seem rather overwhelming , and the carrying out of such an elaborate set of experiments may appear somewhat out of proportion to the interest of the results , but the writers wish to draw attention to the following points:\#151 ; 1 .
The properties which result from mixing two binary alloys having one metal common to each are seldom additive .
A large percentage of the non-ferrous alloys , used in the industries , contain more than two elements intentionally added for some particular purpose .
3 , The following question is often raised : Can one element be advantageously replaced by another , and if so , to what extent ?
* ' ]\#163 ; tude sir les Alliages Blancs , dits Antifriction .
' The Liquidus Curves of the Ternary .
4 .
Very little is known as to the ; effect of impurities on the non-ferrous alloys .
This is very important , as many alloys are useful only when in a metastable condition , e.g. , an alloy containing 90 per cent , of copper and 10 per cent , of aluminium , when cast in chill , gives the following mechanical properties :\#151 ; * Ultimate stress .
Reduction of area .
Elongation on 2 " .
38'0 tons 31*0 per cent. 29'0 per cent. If a bar of the same material is heated to 400 ' C. for one hour , the mechanical properties are radically changed , viz.:\#151 ; Ultimate stress .
Reduction of area .
Elongation on 2 " .
33'0 tons 3'0 per cent. 3'0 per cent. The structure of the alloy as cast consists of about equal proportions of a. containing 7'35 per cent , of aluminium , and / 3 containing 12 per cent , of aluminium .
The / ^-constituent is slowly decomposed between 400 ' and 500 ' C. into a and 7 , the latter being a very brittle body containing about 16 per cent , of aluminium .
There can be little doubt that the rapid deterioration of properties brought about by heating the above alloy is due to the formation of the 7-constituent .
This alloy is therefore practically ruined by heating for an hour at 400 ' C. , and it is quite possible that the presence of foreign elements may accelerate the + 7 conversion to such a degree as to render it mechanically unfit for use .
A systematic investigation of the constitution of alloys containing three elements is therefore a matter of some importance , for only by this means can the questions mentioned above be elucidated .
Before commencing work of this description , a knowledge of the three binary alloys is essential , e.g. , if A , B , and C are the elements , then the systems A + B , B + C , and A -f G should be known .
For this research the three metals aluminium , copper , and tin were chosen for the following reasons :\#151 ; 1 .
The constitutions of the three series : copper-tin , f copper-aluminium , { and aluminium-tin S have been investigated .
The liquidus curves are shown in figs. 2 , 3 , and 4 .
2 .
There is an increasing tendency to substitute aluminium for tin in brasses and other industrial alloys .
* ' Eighth Report to the Alloys Research Committee , Inst , of Mech. Engineers , p. 179 , Carpenter and Edwards .
t Heycock and Neville , 'Phil .
Trans. , ' vol. 202 , 1903 ; Shepherd , 'Journal of Physical Chemistry , ' vol. 8 , No. 6 , 1904 .
+ Carpenter and Edwards , ' Inst , of Mech. Engineers , ' 1907 .
S Shepherd , 'Journal of Physical Chemistry , ' vol. 8 , No. 4 , 1904 .
570 Messrs. J. H. Andrew and C. A. Edwards .
[ May 3 , The simplest way of graphically representing a ternary series of alloys is by means of an equilateral triangle.* Each corner of the triangle represents a pure metal .
The three sides correspond to the three binary mixtures A + B , A+C , and B + C , and any point inside the triangle indicates a certain mixture of the three .
A point on any of the vertical lines drawn from each base to meet the corresponding corners indicates the percentages of each metal , consequently the length of lines drawn from any point in the triangle to meet the three sides at right angles is proportional to the percentage of each metal , and the sum of the length of the lines so drawn is equal to the height of the triangle .
Alloys occurring on lines drawn from the sides of the diagram to the apex have two components relatively constant , while the third varies .
A line drawn parallel to one of the sides of the triangle represents a series of mixtures , in which the percentage of one component is constant .
The temperatures at which the liquid alloys begin to deposit solid are represented by drawing lines through identical freezing points , and thus obtaining isothermal lines for the whole series .
Preparation of the Alloys .
The metals from which the alloys were made had the following degree of purity:\#151 ; Aluminium ... ... ... . .
99*57 per cent. Copper ... ... ... ... ... 99*98 " Tin ... ... ... ... ... . .
99*98 " The methods of making the alloys , although essentially the same as those usually employed , viz. , of melting and mixing together the constituent metals , differed in some details , being modified according to the composition of the resulting mixture .
The copper-rich alloys were made by first melting the weighed quantity of copper , and then adding the necessary amounts of aluminium and tin , the aluminium being added always before the tin .
Since all alloys made in this manner were analysed , no special precautions were taken to prevent any slight oxidation likely to occur .
But for reasons to be mentioned later , alloys containing approximately more than 12 per cent , aluminium and 16 per cent , tin were made by mixing together a binary alloy of copper-aluminium or copper-tin with the third metal , every precaution being taken to prevent oxidation by carefully melting at a low temperature under a layer of charcoal .
Several alloys made in this way were duplicated , and were always found to * G. G. Stokes , 'Papers , ' vol. 5 , p. 226 ( 'Roy .
Soc. Proc. , ' 1891 ) .
1909 .
] The LiquidusCurves of the Ternary System .
571 give the same melting point within the limits of experimental error .
The value of this method , i.e. , of using binary alloys , may be expressed thus:\#151 ; 1 .
Analyses of the binary alloys can be made .
2 .
A lower temperature can be employed in making the ternary alloys , and in this way less oxidation takes place than in the mixing of three pure metals .
In all cases the melt was thoroughly stirred in order to ensure perfect mixing .
The crucibles used and found to be most suitable were the " Salamander " crucibles .
Chemical Analyses .
At the commencement of the research , it was intended to make a chemical analysis of each alloy .
Indeed , for the copper-rich alloys analyses were made and were quite satisfactory .
It was found , however , that analyses of the alloys containing more than 16 per cent , of tin and 12 per cent , of aluminium did not give concordant results .
For example , different samples of the same bars of two alloys showed the following variations in tin contents :\#151 ; Intended composition of the alloy .
Copper .
Tin .
Aluminium .
No. 1 .
65 -00 17 *50 17 -50 First result \#151 ; 16 -99 Second result \#151 ; 14 -90 Third result \#151 ; 16 -00 No. 2 .
52 -00 24 *00 24 -00 First result 18-00 Second result \#151 ; 15 -10 It was obvious that these widely different results must be due to the lack of homogeneity of the alloys .
In fact , this was quite apparent in alloys containing slightly greater quantities of tin and aluminium than the previous ones , the cast bars showing excessive segregation .
So great was this in many cases that the outer part was quite brittle , and could readily be separated from the soft core by mechanical means .
These mechanically separated samples were analysed , and although the results could only be taken as a rough measure of the segregation , they were convincing , showing that the inner core contained a large amount of tin , and the outer brittle layer was approximately a copper-aluminium alloy .
Messrs. J. H. Andrew and C. A. Edwards .
[ May 3 Experiments were made as to the possibility of producing a homogeneous solid bar for analyses , the idea being to freeze the whole mass before segregation had time to take place .
But even with this procedure segregation occurred .
Therefore , finding that chemical analyses were inapplicable synthetic mixtures made from binary alloys had to he relied upon .
Freezing Point Determinations .
In order to avoid any error due to reheating , the freezing points of the alloys were determined directly after mixing by means of a platinum -f 10-per-cent .
iridium thermo-junction , the wires being fused together in an oxy-hydrogen flame .
The free ends were soldered to copper leads , these junctions being kept at zero , and the wires connected with a mirror galvanometer and balancing arrangement similar to that described by Messrs. Carpenter and Keeling* in their work on the iron-carbon alloys .
With most of the alloys no difficulty was experienced in finding the beginning of solidification , but with those alloys containing approximately over 50 per cent , of tin the amount of solid deposited at the initial freezing temperature was so small that great care had to be taken in order to detect the evolution of heat at this temperature .
By cooling extremely slowly , and carefully timing the rate of fall of temperature of the melt , it was found possible to detect the slightest arrest of the movement of the galvanometer .
In many cases , determinations were repeated several times for verification .
The melt in all cases was cooled in the furnace , and except where the alloy froze at a very low temperature the furnace lid was not removed .
The freezing curves taken in the region H to J ( fig. 1 ) present a certain peculiarity , in the form of a slight arrest which takes place at approximately a constant temperature , viz. , 900 ' C. It was thought at first that this might be the freezing point arrest , but this idea is rendered improbable by the fact that such a value , if taken , would break the continuity of the isothermal curves , whilst the points actually taken conform to the general curvature of the lines .
The writers think this slight arrest is due to the mutual separation , in the liquid state , of the copper-aluminium mixture and the tin .
In order to locate the position of the isothermal curves , as shown m the plate , more than 400 alloys and melting point determinations were made .
* Carpenter and Keeling , " The Range of Solidification and the Critical Ranges of Iron-Carbon Alloys , " * Journal of the Iron and Steel Institute , ' vol. 1 , 1904 .
1909 .
] The Liquidus Curves of the Ternary System .
The Diagram .
( Plate 3 .
) The diagram is conveniently divided into areas , in which the direction of the isothermal curves is similar .
Alloys in each area may be classed as a group or family , in which the influence on the liquidus of a certain phase predominates .
As the compositions of the binary alloys only are known with any degree of certainty , the writers propose to speak of the alloys in Fig. 1.\#151 ; Key to Areas on Main Diagram ( Plate 3 ) .
groups , giving reference to the binary constituent which has the greatest influence on the liquidus of each group .
Alloys in the ABC area ( fig. 1 ) begin to solidify with the formation of the copper-tin a-crystals , which also contain a certain amount of aluminium .
The ACD group of alloys , on cooling , first deposit copper-aluminium a-crystals containing some tin .
574 Messrs. J. H. Andrew and C. A. Edwards .
[ May 3 , Mixtures in the area BEFC commence to freeze with the deposition of copper-tin ^-crystals , which also appear to contain some aluminium in solution .
The lower part of this area is not well defined , but seems to merge into the adjacent one , the liquidus of which is formed by the freezing out of a chemically strong compound .
In the area DFExyGHI , the constituent which is the first to separate is the copper-aluminium / 3 phase , which also holds tin in solid solution .
This \#166 ; 800PCRCENTAGE Fig. 2.\#151 ; Liquidus Curve of Copper-Tin Alloys .
constituent first appears in the binary alloys at 7'35 per cent , of aluminium , and is present in all alloys up to 16 per cent. An alloy with 12 per cent , of aluminium consists of pure / 3 and closely corresponds to the compound Cu3A1 .
Whilst there is no indication on the liquidus diagram to mark the precise position of the lower HI boundary line of this area , the writers consider the area shown as being approximately correct .
It was deduced by drawing 1909 .
] The LiquidusCurves oj the Ternary System .
575 a line from the point H , which corresponds to the disappearance of the / 3-phase."in the binary alloys , to the point I , which seems to be the extent to which this phase influences the formation of the curves .
It is interesting to note that the phase which is present in the copper-aluminium alloys , only over a range of 8*65 per cent. , viz. , from 7*35 to 16 per cent , of aluminium is the constituent which influences so markedly the freezing point of these alloys even when as much as 95 per cent , of tin is present .
Further , the copper-tin 7 and 77 phases which are deposited respectively from the liquid alloys containing from 32 to 57*5 , and 57*5 to 99 per cent , of tin , do not crystallise from any of the liquid ternary alloys .
This is obviously due to the added aluminium combining with the copper to form the copper-aluminium phase , thereby raising the freezing point of the alloy above that of either 7 or 77 .
Thus an alloy with 80 per cent , of tin and 20 per cent , of copper begins to freeze at 530 ' C. , but if 0*50 per cent , of aluminium is added to it , the mixture will commence to solidify at 730 ' C. Alloys in the area HIJ and JIG-K may be considered together : the first begins to deposit the copper-aluminium 7-phase , and the second the 7/ -phase , both of which are pure .
The tin separates and settles to the bottom in an almost pure state .
The area LGKMN encloses alloys which on cooling deposit pure crystals of the compound CuA12 .
The added tin has no influence on their melting points , but simply separates , forming a separate lower layer of tin .
Alloys in the area NMO on cooling deposit the solid solution of copper in aluminium which may also contain a small amount of tin .
The area LNO .
The solid first formed on cooling alloys in this area is the solid solution of tin in aluminium which may also contain a limited amount of copper .
Summary .
1 .
The character of the liquidus curves , as seen on the diagram , indicates that no ternary compound is deposited from any of the liquid alloys , and throughout the whole system no true ternary phase appears to form above the solidus .
The alloys around the point N cannot be said to consist of a ternary eutectic , but rather it would seem that the melting point of the copper-aluminium eutectic at M is slightly lowered by tin , which remains in the free liquid state .
This absence of ternary phases may be accounted for by the fact that the copper-aluminium phases , in alloys containing more than 12 per cent , of aluminium , have very little or no affinity for tin .
This behaviour of the copper and aluminium towards tin suggests that the system might almost be considered as a binary series of alloys .
In fact , if the Messrs. J. H. Andrew and C. A. Edwards .
[ May 3 , freezing points of alloys containing a constant percentage of tin are plotted , liquidus curves are obtained of a similar character to that of the copper-aluminium alloys .
( See figs. 5 , 6 , and 7 .
) Percentage Fig. 3.\#151 ; Copper-Aluminium Alloys .
Percentage Compose / t/ on .
Fig. 4.\#151 ; Aluminium-Tin Alloys .
The Liquidus Curves of the Ternary The liquidus curve of alloys containing 10 per cent , of tin , with the copper and aluminium varying from 0 to 90 per cent. , is shown in fig. 5 .
The general character of this curve is almost identical with that of the pure copper-aluminium alloys ( fig. 3 ) .
In both cases , the addition of aluminium first lowers the melting point from A to D , then produces a rise , to a M Ml , Fig. 5 .
Fig. 6 .
maximum at B , which corresponds to the compound Cu3A1 .
This is followed first by a gradual , and then by a rapid fall in the melting point to K , where the compound CuAl2 is in each case deposited from the liquid alloys .
With more aluminium the melting point is gradually lowered , in the case of the binary alloys to the minimum point M , but with those containing 10 per cent , of tin to a minimum range M , M ' .
From here to B , fig. 3 , 578 Messrs. J. H. Andrew and C. A. Edwards .
[ M^y 3 , pure aluminium on the one hand , and 90 per cent , aluminium with 10 per cent , tin on the other , B , fig. 5 , the melting points are raised ; in the case of the latter curve there is a slight break at S , this being the point where it cuts the line NO in the key diagram , fig. 1 .
Fig. 6 represents the liquidus curve of alloys containing 50 per cent , of tin and from 0 to 50 per cent , of copper and aluminium .
In this curve the branch AD , fig. 3 , which corresponds to the a-range , is absent , in other respects the curves are very similar .
Another curve of this character is shown in fig. 7 , for the alloys containing 80 per cent , of tin and from 0 to 20 per cent , of copper and aluminium .
On this curve it is seen that the 80-per-cent , tin and 20-per-cent , copper alloy begins to solidify at 530 ' C. , and with the addition of up to 1 per cent , of aluminium the freezing point is rapidly raised to 800 ' 0 .
Although there is 80 per cent , of tin in this series , the curve from 1 per cent , to 20 per cent , of aluminium is almost identical with fig. 3 from the point D to B. Considered in another way , the phases formed in the copper-aluminium series from D to B , which cover a range of more than 90 per cent. , are compressed into a range of only 20 per cent , in the presence of 80 per cent , of tin .
A curve plotted on a different system is shown in fig. 8 .
This curve is drawn through the melting points of alloys occurring on a line from the point B , ( CugAl ) through the maxima of the beak-like curves in the area DFExyGH to the point C which corresponds to pure tin .
From the curve it will be seen that the introduction of tin to the compound CU3AI steadily lowers the melting point of the latter to 945 ' C. at B ; from here to C the lowering is very slight , but afterwards there is a rapid fall in the melting point to D. 2 .
The remarkable stability of the compound Cu3A1 is clearly shown by the beak-like formation in the area DFExyG-IH , which shows that the affinity of tin for either copper or aluminium is not sufficient to overcome ( he affinity of those two elements for each other .
3 .
Alloys in the area KGLMN which contain a relatively constant proportion of copper and aluminium , i.e. , those which occur on lines drawn from the side of the triangle AB to the corner C , have identical freezing points .
This means that the introduction of tin to any of the copper-aluminium alloys between KM has no influence on the freezing points , which indicates that the two liquids , tin and the copper-aluminium mixtures , are mutually insoluble .
4 .
Tin is insoluble in by far the greater number of the alloys .
Thi$ Andrew and Edwards .
Roy .
Soc. Proc. A. vol. 3 .
ith the int tin the , her :r is 0 to the 5C .
, it is -_-r- cy \a_\#187 ; vXL !
rom from lium , , are cent. \#166 ; ve is i the area curve jadily C the elting wn by it the reome t pro-drawn reezing jopper-points , ainium This Temperature Decrees Cent/ crade Liquidus Curves OP Aluminium-Copper-Tin Alloys .
1909 .
] The Liquidus Curves of Ternary System:579 phenomenon is most evident in all those on the lower side of the line HI with the possible exception of a very limited number containing upwards of about 94 per cent , of aluminium .
Percentage T/ n. Fiq .
8 .
In conclusion , the writers wish to gratefully acknowledge the assistance given by Prof. H. C. H. Carpenter , in placing the apparatus in his laboratories , and the materials used , at their disposal , and for kindly criticism during the progress of the work .

