recent developments in metals sealing into glass

RECENT DEVELOPMENTS IN METALS SEALING INTOGLASS .

BY

HOWARD SCOTT, A.B.

I. INTRODUCTION .

The rapidly expanding uses for sealed-off gaseous con-duction devices, such as small rectifiers, grid-glow tubes andIgnitrons has directed attention to the limitations of theglass-metal seals hitherto available . A vacuum tight metal-to-glass joint is required in these devices either to conductpower into a glass vessel or to insulate electrodes from metalwalled vessels . Tungsten and molybdenum serve admirablythe function of conducting power through a glass wall . Theyseal only into hard glasses having low coefficients of expansionand therefore high resistance to thermal shock. They are,however, not available except in very simple and small shapes .They are extremely difficult to fabricate or machine and aretoo costly to use as vessel walls .

Mercury is the usual source of a conducting vapor inmany of the devices mentioned . Its presence eliminates thepossibility of using copper in such forms as the sheath onDumet and cups or tubes for Housekeeper' seals . Someprotection against amalgamation may he afforded by electro-plating mercury resistant metals onto copper, but the per-manence of such protection is questionable and the expenseundesirable . High chromium-iron alloys seal into soft glass,but this glass, because of its high expansivity, has inadequateresistance to thermal shock . It is clear, therefore, that ametal is needed which seals into a hard glass, is resistant tomercury attack, can be readily machined or otherwise formedand is not unduly expensive . Such a metal will eliminatepresent restraints on size and capacity of sealed-off vacuumdevices .

The writer's investigation of the expansion properties of

1 7ourn. A . L E. E ., 42 : 954 (1923) .733

734

HOWARD SCOTT.

[J . F. 1 .

iron-nickel-cobalt alloys' suggested the possibility that alloysin this system might seal into hard glass and meet the otherrequirements just noted . While investigating this possibility,various obstacles were met and overcome with the result thatanew metal is now available for sealing into hard glass . It isfree from the major objections to the hitherto avilable sealing-in metals . The various factors involved in the developmentand use of the new seals are discussed in this paper .

U . PROPERTIES OF GLASSES .

It may be taken as axiomatic that, in order to make asuccessful seal, the expansion of the metal and glass mustbe substantially the same over the temperature range withinwhich the metal and glass are both elastic . A large differ-ence in expansion between metal and glass bonded by sealingproduces stresses which may cause the glass to crack whencooled to room temperature . Some degree of differentialexpansion, however, is tolerable or even desirable, as will heshown later, but for the moment a perfect match may beassumed to be preferable .

The relaxation characteristics of the glass determine theupper temperature limit to which matched expansion isessential . Obviously, at temperatures where the glass issubstantially plastic, equal expansion is no longer necessary .Since the iron-nickel-cobalt alloys whose expansion is com-parable with that of hard glasses have only a limited tempera-ture range of low expansivity, the temperature at which aglass becomes plastic is a vital characteristic in these con-siderations .

The relaxation of stress in glass under a given initial strainproceeds continuously and at a rate which rapidly increaseswith rising temperature . For purposes of exposition it isnecessary to select arbitrarily a temperature characteristic ofthe relaxation range of the glass . In glass technology twosuch temperatures, the strain point and annealing point, areused and define well the temperature range of rapid relaxation .Stresses in the glass will decay to an inappreciable value afterexposure for 15 hours at the strain point or 15 minutes atthe annealing point . Values of these characteristic tempera-

Trans . A . I . M. E ., Inst. of Metals Division, p . So6 , I9go .

Glas

s.

Soft

Thermometer

Clear Sealing

Nonex

(Pyr

ex b

rand

che

mica

lre

sist

ance glass)

TABL

E1.

Comp

osit

ions

and

Pro

pert

ies

ofSo

me G

lass

es I

nto

Whic

h Ap

prop

riat

e Me

tals

are

Sea

led

.

Code

No

.

001

752

705

772

774

1

3.2Ex

pansrvity

(25

to32

5°C

.).

Strain

Poin

t.

Anne

alin

gPo

int

. i

9.0X10

-61'C

.

389°C

.

425°C

.530°C

.56

6° C

.4

.6i

461

° C

.49

6' C

.3

.6

486° C

..

521

° C

.

503°

C.

550°

C.

%SiOs.

63 72 67 73 8o

Approximate Compositions

.

%Nas

.KsO

.

14 10 6.5

4.5

413

-* Low

est

temp

erat

ure

at which strains will decay to an in

appr

ecia

ble

valu

e af

ter

expo

sure

for

15

hour

s.

t Same for exposure of

15

minu

tes

.

7 BsO,

%PbO

.

2110 22 16

.5

6

%AIs

O,.

5 2 2

736

HOWARD SCOTT .

[7. F. 1 .

tures are given for several important glasses in 'fable I .They were obtained by the method of Littleton and Roberts Iat the Corning Glass Company and are given here throughthe courtesy of that concern .

FIG . i .

500

600°CTEMPERATUREExpansion curves of two hard glasses from room temperature to annealing points .

0 100 200

300

400

The values of expansivity given in Table I representheating from 25 to 325° C . over which range the expansion ispractically linear. In connection with scaling problems, how-

3 Journ. Optical Soc . Am ., 4 : 224 (1920) .

0 5 10

A

c~5

J

~

J

~--'y6P~\aG

ysG4P

1

~~

0 HEAT NGCOOLING

Dec., 1935 .1

ever; the expansion during cooling and over the temperaturerange from the annealing point to room temperature is ofchief interest . Figure i gives expansion curves of two im-portant hard glasses throughout that temperature range onboth heating and cooling . It maybe noted that the expansionfrom the strain point, about 470 0 C . for these glasses, to roomtemperature is much higher during cooling than on heating .Recording values of mean expansivity for the temperatureranges from room temperature to 325 and to 470 0 C. as inTable II, one may see that the effective expansivity on cooling

TABLE II .

Expansivity of Hard Glasses and Metals Sealing Into Them .

METALS SEALING INTO GLASS .

Expansivity .

737

is higher for the longer temperature range . Since the lowestfeasible annealing temperature of a glass is its strain point, avery useful single expansion characteristic of glass in thisconnection is the mean expansivity on cooling from the strainpoint to room temperature .

The expansion curve of a particular iron-nickel-cobaltalloy, Kovar, is given in Fig .. 2, together with that of twohard glasses for comparison. The expansion of this alloy isreversible, that is, the same on heating as on cooling . Valuesof its expansivity over two temperature ranges together withvalues for tungsten .and molybdenum are given in Table II .Its expansivity is lower than that of tungsten up to 450 0 C .and than that of molybdenum up to 500 0 C. These metalshave the lowest expansivity of any of the pure metals .Kovar contains about 29 per cent. Ni, 17 per cent . Co, 0.2per cent. Mn, balance iron .

It is apparent from Fig . 2 that the match in expansion

Material . Direction ofTemp. Change .

25 to 325° C . a5 to 470' C .

Nonex Glass Heating 3.5 X 10'/ ° C. 2.8X1n1 °/°t-Cooling 4.1

Clear Sealing Glass Heating 4 .2 3 .9Cooling 4.4 4.6

Kovar Metal Either 4.0 5.1Tungsten Metal 4.7 4.8Molybdenum Metal 5.5 57

738

0 100

HOWARD SCOTT .

[J. F. I

between Kovar and either glass is not perfect, but that thematch is much better in the case of the "Clear Sealing"glass than in that of "Nonex ." (Compositions of theseglasses are given in Table I .) Though the Clear Sealing glasshas a lower expansivity than Kovar at temperatures above435° C., the discrepancy is small so that seals practically free

FIG. 2 .

200

300

400TEMPERATURE

500°C .

11"xpansioa curve of Kovar compared with curves of two hard glasses observed during cooling fromannealing point .

from stress can be made with this combination . In fact,serviceable seals can be made of Kovar in Nonex glass also .Other hard glasses have a still lower expansivity and higherstrain point than Nonex, and, therefore, present a moredifficult sealing problem . Accordingly, it is desirable to studythe expansion characteristics of the iron-nickel-cobalt alloysfor information on their limitations and on the manner in

r

0-5X 10-3 ~~

/

CLEAR A NONEXGLASS

i

Dec, 1935 .]

which the composition of the particular alloy already con-sidered might be advantageously modified to meet otherconditions .

m . ALLOY COMPOSITION FACTORS .

For the study of composition effects, it. is desirable toexpress the expansion characteristics of the iron-nickel-cobaltalloys simply and compactly . The various reversible com-positions differ significantly only in the range of low ex-pansivity and the effective expansivity throughout thatrange . The upper limit of the range of low expansivity isgiven by the inflection temperature, Fig . 3 . For a repre-

6x10'

5

4

3

2


EXPANSIVITY25 TO T°C

INFLECTIONTEMPERATURE

10 -3

UNIT EXPANSIO25 TO T °C

739

0 0

100

200

300

40C

500°C (T)

Unit expansion and expansivity of Kovar showing method of evaluating inlleotian temperature .

sentative value of expansivity one need take only the meanexpansivity from room temperature to the inflection tempera-ture, which range is inferred by the term mean expansivityapplied to these alloys . So defined the mean expansivity andinflection temperature represent the best combination of

740

expansivity and temperature range of low expansivity avail-able in a particular alloy .

The properties just called to attention are plotted againstnickel plus cobalt content in Fig . 4 . Peculiarly, the effects

L lc . 4.IlaVKWn

N

ie OOvEz60rI-

'-"40Nw

n 10

MWaN_ 6

z0

2 6Zr

N 42aax

w 24

HOWARD SCOTT .

wa

°C a500M

aZw

404 ZorUw

300 z

42

44

46

48

50NICKEL PLUS COBALT CONTENT

[J. F . I .

Mean expansivity, ' . ftection temperature and resistivity of iron-nickel-cobalt alloys containingabout o .a per cent manganese.

of nickel and cobalt on the inflection temperature are in-distinguishable in the composition range considered, that is,cobalt acts as a substitute for nickel so far as this property isconcerned . Consequently, it is convenient to take the nickelplus cobalt content as a composition variable, and the curve

COBALT CONTENT----UNDER ONE PERCENT---UNRESTRICTED-OPTIMUM

RESISTIVITY

INFLECTIONTEMPERATURE '

MEANEXPANSIVITY

//

Dec, 1935 .1

for inflection temperatures of Fig. 4 is then not restricted toany particular relation of nickel to cobalt content .

In the absence of cobalt, that is, in plain iron-nickelalloys, low values of expansivity are also obtained, but for agiven inflection temperature they are considerably higherthan those obtained with optimum cobalt content, Fig. 4 .The improvement on substituting cobalt for nickel is greaterthe higher the amount of nickel replaced up to a certain

FIG. 5 .

r

Z 5axw2 4aW

e

3

50


300

350

400

450

500INFLECTION TEMPERATURE - ° C

550

741

Mean ex7ansivity of [me-nickel-cobalt alloys containing about 0.2 per cent manganese plottedagainst inflection temperature. The two individual observations designated "Clear Sealing" and" ,onex" represent contrLion of two glasses from their strain points to room temperature .

limit to be defined presently and the curve for mean ex-pansivity with optimum cobalt content represents substitutionto this limit. Since both the expansivity and inflectiontemperatures are functions of nickel plus cobalt content fora prescribed cobalt content, one may be plotted against theother as in Fig . 5 which shows clearly the best expansivityvalues available to a wide range of temperatures .

NO COBALT, PIE/ 0-CLEARSEALING

O-N N %~P

COBALTOPTIMUM

CONTENT

742

HOWARD SCOTT .

[J. F. I .

The limit to the benefit conferred by the substitution ofcobalt for nickel is set by the Ar, transformation of iron .The desired low and reversible expansivity of these alloys isobtained only when this transformation is depressed belowroom temperature . Increasing nickel content lowers it andcobalt content raises it so the net effect of substitution is arapid rise in the temperature at which this transformationstarts . Once it occurs above room temperature the expansioncurve of the alloy is no longer reversible nor is the expansivityas low as desired .

Practical considerations in alloy preparation dictate thatthe intended composition selected should be such that Ar 3starts at about - ioo° C . Accordingly, the optimum cobaltcontent is taken as that which, with nickel plus cobalt andother elements fixed, will cause the transformation to occurat this temperature . It is determined quite accurately forthe range of compositions considered by the relation :

%Ni+2.5(%%Mn)+18(%C)%n Fe

= 0.55 .

Expressing iron content by difference, the optimum cobaltcontent is given by :

%Co=1.55(%Ni+ Cc) +3.0(%Mn)+18(%aC)-55 .0 .

It may be seen from the preceding equation that theoptimum cobalt content is higher the higher the manganeseand carbon contents of the alloy and in fact the higher thecontent of any other elements that lower the Ar 3 transforma-tion . This consideration suggests that the expansion prop-erties can be improved still further by the addition of anysuch elements with a corresponding increase in cobalt content .Actually, with one known exception, the expansion propertiesare harmed more by the addition elements than they arebenefited by the greater cobalt addition permitted .

The exceptional element is carbon . Small additions havelittle direct effect on the expansion properties, but permitconsiderably higher cobalt contents . The effects of carbonand manganese on the mean expansivity for a given inflection

Dcc., 19351


temperature have already been determined quantitatively .'When the cobalt content is maintained at its optimum value,the mean expansivity is lowered by 0.12 X io-s/° C. by anincrease in carbon content of 0.1 per cent. and is raised by0 .04 X io ' C . by an increase in manganese content of o .iper cent .

Since the expansivity for a given inflection temperature islowered when carbon is added and a corresponding substitu-tion of cobalt for nickel is made, there is a real advantage tobe gained by so doing . The amount of the carbon addition,though restricted to about 0 .3 per cent ., is sufficient to lowerthe expansivity by 0.4 X io 6 C . Higher contents may notbe retained in solid solution in which case the excess doesnot have the desired effect . Advantage is not taken of thisexpedient in the case of the Kovar compositions consideredhere, however, because of an undesirable effect of carbon onthe glass-sealing characteristics of the alloy .

All other elements than carbon so far investigated, in-cluding manganese, do not afford even a balance between theirharmful effects on the expansion properties and the benefits ofthe corresponding cobalt substitution . Accordingly, in theabsence of carbon, these alloys should be made as pure aspossible to obtain the best expansion properties . The alloyKovar of Figs . 2 and 3 is so made, the melting stock beingpurified by annealing in hydrogen and melted and frozenunder a hydrogen atmosphere . A manganese addition of 0.2per cent. i s made as a concession to certain metallurgicalfactors though not really essential . This alloy has been pro-duced without any manganese addition in the form of a 150

pound ingot and successfully fabricated with no exceptionalforging trouble .

IV. STRESSES IN SEALS .

Clear Sealing glass into which Kovar has been sealed canbe obtained at room temperature practically free from stress .since the contraction of both materials on cooling from atemperature near the strain point is the same . This impliesthat an overnight anneal is necessary to achieve the condition

Loc. cit .

743

744 HOWARD SCOTT .

1J . 1' . I .

of stress absence. The expansion curve of the glass, however,is parallel to that of the alloy and not far distant at tempera-tures between the strain and annealing point, Fig . 2 . As aconsequence, complete stress release can be achieved byannealing for only a few minutes with little stress developmenton subsequent cooling .

In the case of other hard glasses, Nonex for example, nosuch close match in expansivity has been attained . Themean expansivity of this glass is lower by 25 per cent. thanthat of the best alloy of Fig. 5 . Nevertheless, successful sealsbetween Kovar and Nonex glass can be made because thestress relations in the glass are more favorable when the metalhas the higher expansivity than otherwise .

Consider for simplicity a conducting wire sealed into aglass vessel . If the major portion of the glass bead has acylindrical form and encloses a solid round wire, the stressesproduced by differential expansion can be calculated from theequations for a hollow cylinder shrunk on a solid cylinder asgiven by Timoshenko in his "Strength of Materials," p . 531 .The equations for radial and tangential stresses in theglass are :

Sr=P b 2 -a'(r2

a2

b2S

-P b 3 -a2 rz +I

where a is the radius of the metal wire, b of the overlying glass,r the radial distance from the wire axis to the point consideredand P the radial stress across the boundary between metal andglass . The value of P is, of course, determined by the differ-ence in expansion between metal and glass and their elasticconstants but these need not be evaluated since only relativemagnitudes are sought here . The radial stress in a sealcooled after annealing being tensional when the metal hasthe higher expansivity, the sign of P should be taken aspositive under this condition and negative when the glass hasthe higher expansivity, to give the stress sign as cooled afterannealing .

The radial and tangential stresses are a maximum at theboundary between metal and glass, Fig . 6. At this position,

Dec., 1935 .1

1 .0

Z,1z

0.5O

Q NJ Nt) ~?5 0

wiw¢ 2NN -0.5

-1 .0


the ratio of the tangential to the radial stress is :

b 2 + a 22b2- a2

Thus at small diameters of the metal the values of themaximum radial and tangential stresses approach equality,but for a thin layer of glass the tangential stress may be manytimes the radial .

Fic . 6.

0

----GLASS THICKNESS 2X METAL-GLASS THICKNESS 9X METAL

N

02

0.4

0.6

0 .8 DISTANCE FROM METAL SURFACE

THICKNESS OF GLASS

1 .0

745

Calculated stress distribution in cylindrical sheath of glass fused to metal rod of higher expansivitythan glass after annealing and cooling to room temperature .

Cracks form in glass as the direct result of tensionalstresses and afford evidence of the distribution of stress priorto cracking . If now one tabulates the maximum radial andtangential stresses as in Table III, it is clear that the glasswill crack in an entirely different manner when the metal hasthe higher expansivity than when the glass does . Assumethat the metal has the higher expansivity by a sufficientdegree to produce cracks though the development of thermal

RADIAL TR ES5ZOmZwH

i i

ZO

Ni TANGENT IASTRESS

L wa00v

746

HOWARD SCOTT .

TABLE III .Maximum Tangential and Radial Stresses in a Cylindrical Glass Bead as Sealed

Onto a Metal Rod, Annealed and Cooled to Room Temperature,The value of P is determined by expansivity and elastic constants of com-

ponents .

Material of HigherExpansivity.

Metal

Glass

S%gn .

CompressionTension

Tangential Stress .

Magnitude .

fb'+at` P`b'-a Jr

a is radius of metal rod .b "

" glass sleeve .

residual stresses is inhibited, then the cracks should be thedirect result of radial stresses and should form close to themetal surface and more or less parallel to it . This is indeedthe case, Fig . 7(b), for the crack that split the bead shown sostarted and spread . When the differential expansion is notso high, such cracks do not travel far from the metal face .

Assuming now that the glass has the higher expansivity,one would expectt from the analysis of stress distribution thatcracks would result from tangential or axial stresses . In fact,observed cracks are obviously due to a combination of ten-sional tangential and axial stress, since they take a helical paththrough the glass, Fig. 7(a) . The axial stress is materiallyhigher than the tangential for if the fracture were caused bytangential stress alone it would occur in a plane parallel to thewire axis . The axial stress could be calculated were the elasticconstants of the glass and metal known, but it suffices forthe present purpose to know that the axial stress is con-siderably higher than the tangential and of the same sign .

Since the axial stress is higher than the tangential and thetangential higher than the radial, particularly when the glassthickness is small, it is clearly preferable to have any unavoid-able tensional stresses in the radial direction . This conditionis achieved simply by choosing a metal of higher expansivitythan the glass into which it is sealed . The need for thisrelation of metal to glass becomes apparent when it is statedthat some residue of stress in practice is unavoidable, the

Sign.

TensionCompression

Radial Stress.

17 . F. I .

Magnitude.

P

Dec.. 1935 .]


reasons being : (i) imperfect uniformity of materials, (2) lackof metals that exactly match expansivity with glasses re-quiring a sealed-in conductor, and (3) limitations on timeavailable for annealing . It is undoubtedly on account of

Fm. 7 .

747

a

b

Cracks in beads of glass sealed onto wire 0 .072 inch diameter having a mean expansivity ofax X ,o5f . C., beads cooled in smoky flame. (a) Soft glass, expansivity from 25 to 325' C .Q .2 X 101/O C . (b) Hard glass, expansivity from 25 to 325° C . q .5 X r0-'/=C.

vo7.. 220, No. 132~57

748 HOWARD SCOTT .

[J r . t .

these considerations that sealing-in metals are consistentlychosen with higher expansivity than the glasses into whichthey are sealed .

Another factor favorable to the use of a metal havinghigher expansivity than the glass it seals is the residualstresses induced in the glass during cooling . Seldom is itfeasible to cool large seals slowly enough to prevent appre-ciable residual stress induction from temperature gradients inthe glass . Assuming that no great fraction of the heat flowduring cooling is along the wire axis, the residual stress inducedin the glass adjacent to the metal is a maximum and tensionalin the axial direction, presuming no differential expansionbetween metal and glass . Consequently, the axial corn-pressive stress from differential expansion tends to annul theresidual tensional stress due to temperature gradients whenthe metal has the higher expansivity .

The foregoing conclusions are based only on the case of asolid rod sealed into glass . In the case of a tube or cup seal,it is not necessarily desirable to have a metal of higher ex-pansivity than the glass . Cracking of the glass in this caseis generally due to bending stresses which are as likely toproduce as high local tensional stresses when the metal hasthe higher expansivity as when the glass does . Nevertheless,serviceable tubular seals can be made when the metal hasconsiderably higher expansivity than the glass, Kovar inNonex glass, for example . With so high a differential ex-pansion, however, a special annealing practice must be usedas will be described presently and a better match in ex-pansivity between metal and glass such as that betweenKovar and Clear Sealing glass is decidedly advantageous .

V . PREPARATION OF SEALS.

A metal must meet other requirements than expansivitybefore it can be rated as satisfactory for sealing into glass .For one thing it must be "wet" by glass applied at a suitabletemperature . Actually it is an oxide film on the metal thatis wet for the metals commonly sealed become more or lessoxidized during heating for sealing . Metal oxides are readilydissolved in fused glass so wetting appears to be simply thesolution of glass in the oxide film on the metal . Evidently

Dec., 1935 .1 METALS SEALING INTO GLASS . r' 49

some care must be exercised to control the oxidation for if thefilm is too thick it may be porous .

Iron-nickel alloys of appropriate compositions at one timewere sealed directly into soft glass, but were shortly dis-placed by "Dumet." The copper sheath on Dumet asnormally oxidized bonds with soft glass much better than therefractory oxide that forms on iron-nickel alloys at the lowtemperatures used in the blowing of soft glass . Substitutingcobalt for part of the nickel in iron-nickel alloys to producealloys such as Kovar improves materially the fusibility of theoxide formed. Alloys of high cobalt content are readily metby soft glass while Kovar bonds well with hard glass .

When the first seals between Kovar type alloys and hardglass were made, though no trouble was encountered ingetting the glass to wet the metal, bubbles formed in the glass,adjacent to the metal surface weakening the joint . Thesource of the gas forming the bubbles was undoubtedly re-action between carbon in the alloy and the oxide film on itssurface. In any event bubble formation was effectivelyeliminated by removing all carbon from the metal surfaceeither by annealing in hydrogen or by preparing the alloyfrom truly carbon free components .

With this preparation, no special precautions are necessaryto secure adhesion between Kovar and certain hard glasses .No "beading" is required unless the discrepancy in expansionis great and a long anneal is impracticable . By use of analloy and glass whose expansitivities are relatively close, theneed for a feather-edge on a tubular shape, such as is essentialfor sealing copper by the Housekeeper technique, is eliminated .In fact, it is preferable to have a heavy rounded edge on themetal to minimize concentration of stress in the glass . manylarge seals have been made between hard glasses and tubularshapes with a wall thickness of I /8 inch, an example of whichis shown in Fig . 8 .

VT. ANNEALING.With so perfect a match in expansivity as that between

Kovar and the Clear Sealing glass, the annealing procedure forleast development of stress is clearly indicated . The sealwhile hot from the blowing operations should be cooled to the

750 HOWARD SCOTT .

annealing point, held there 15 minutes and then cooled tothe strain point at a rate determined by the permissibleresidual stresses. Cooling from the strain point to roomtemperature can be quite rapid, the only limitation being

Fm. 8 .

Seal (of Clear Sealing glass tube) between cup and tube of Kovar 4 inches in diameter .

danger of cracking from temporary stresses due to tempera-ture gradients necessary to maintain cooling .

Our original experimental work was done, however, on thecombination Kovar and Nonex glass since it presented the

Dec . .

19351


751

most difficult of the problems which appeared susceptible ofpractical solution . From Fig. 2, it will he seen that the ex-pansivity matches most closely near 400° C. while the strainpoint is 486° C. If the seal be annealed overnight at 486° Cno stresses will remain at that temperature, but stressescorresponding to a unit differential expansion Of 0.5 X 10-'will develop when cooled therefrom to room temperature .

An alternative method for annealing this seal is to coolto 390° C . and hold at this temperature . Complete stressrelease could be achieved by holding at this temperature forsome days, but such a treatment is objectionable . Sufficientrelaxation may occur and certainly the major part does occurduring an overnight anneal . If the composition of the Kovarhas been modified to give an inflection temperature of 390° C .no additional stress will develop on cooling thereafter to roomtemperature .

To test the foregoing possibility, wires of Kovar weresealed into Nonex glass tubes, the glass being shaped so as togive three different degrees of stress concentration . Theresults of annealing overnight at several temperatures aregiven in Table IV and show definitely that annealing below

TABLE IV .

Occurrence of Cracks in Seats of o.rrS Inch Kovar Mire Into Nonex Glass .The seals were held overnight at the annealing temperature and cooled in

air . The stress concentration factor was determined by the contour of the glass .

the strain point is effective when severe stress concentrationis to be avoided . It is quite possible that relaxation of themetal also plays a part in relief of the stresses at annealingtemperatures below the strain point, particularly in tubularseals . Kovar loses 50 per cent . of the maximum strainapplied by bending when held 15 hours at 400° C.

AnnealingTemperature.

Stress Concentration .

Low. Intermediate. High .

465° C No cracks Cracked Cracked410

. . . . .. . . . . . . . . . 1 .

11

390 " No cracks375 . . . . . . . . . .

7 52

HOWARD Scorr .

[J . F. I .

Other factors contribute also to the success of the annealingprocedure described . By developing maximum stress at ahigh temperature there is little danger of cracking for theglass then still retains some degree of plasticity . In nocase have Kovar-Nonex glass seals cracked at the annealingtemperature with this practice despite the high differentialexpansion and fast cooling to the annealing temperature .Having a high stress at the annealing temperature, rate ofrelaxation is more rapid than otherwise for the rate increasesas the square of the stress .'

This treatment is effective when applied to tubular sealsas well as to rods . Tubes of Kovar 2 inches diameter, 1/8 inchwalls, have been sealed into Nonex glass therewith . Itshould be recognized, however, that a glass of higher ex-pansivity than Nonex is preferable . The Clear Sealing glassis particularly desirable for, in addition to its more favorableexpansivity, it does not devitrify under the extended heatingrequired to make large seals . This seal has also excellentresistance to thermal shock as proved by abuse in thelaboratory .

V[I. SUMMARY AND CONCLUSION .

An investigation of the effects of cobalt on the expansionproperties of low expansivity nickel steels revealed that agiven low expansivity can be obtained to much higher tem-peratures by the addition of cobalt to these alloys than with-out . This finding introduced for the first time a possibilityof sealing iron-base alloys into hard glass having an expan-sivity as low as 3 .6 X io s/°C. with a strain point up to470° C. A particular composition having optimum cobaltcontent and designated Kovar is most suitable for sealing intosuch hard glass .

Kovar is readily wet by hard glass, that is, the naturaloxide formed on it during heating fuses with the glass at glassblowing temperatures. The same is true of soft glass, par-ticularly when the cobalt content is higher than in Kovar .Bubbles may form in the glass adjacent to the metal if thecarbon content is not kept very low or other precautionstaken .

5 Adams, J . FRANK. INST ., 216 : 39 (1933) .

Dec ., 1935 .1 METALS SEALING INTO GLASS . 753

Kovar in the shape of either rod or tubes can be sealedinto Nonex glass without the use of a beading glass or gradedseal . Special precautions to avoid stress concentration and anunusual anneal are required, however . A more favorableglass for use with Kovar is Clear Sealing glass which contractsonly slightly less than Kovar on cooling from its annealingpoint to room temperature . Seals of Kovar into this glasshave a material advantage over one or the other of thestandard seals in that they have a high resistance to thermalshock, are not attacked by mercury, and the metal is readilyshaped hot or cold and is easily machined .

A study of stress relations and annealing showed that itis desirable for a metal to have somewhat higher meanexpansivity than the glass into which it is to be sealed,except possibly when the joint is tubular . When the metalhas a higher mean expansivity than the glass, successful sealscan be made even with tubular shapes and when there is a con-siderable discrepancy in expansivity by avoiding high tensilestress concentration in the glass . If, in addition to the con-ditions just mentioned, the strain point of the glass is abovethe inflection temperature of Kovar, annealing overnight at atemperature below the strain point may be required as is thecase with Nonex glass .

VIU . ACKNOWLEDGMENTS.

Material assistance in the development of the seals de-scribed here has been given by Messrs. Knowles and Bangratzof the Physics Division, Westinghouse Research Laboratories,who have used them in tube construction . Many of theexperimental data reported were observed by Mr . J . G. Hoopof the Metallurgical Division . The writer is indebted also toMr. P. H. Brace, Manager of the Metallurgical Division forsuggestions and support during the course of this work, andto Dr. Nadai of the Mechanics Division for a mathematicalanalysis of stresses in seals .

recent developments in metals sealing into glass

Documents