8/3/2019 969518 (1)

1/9

Electrocomponent Science and Technology1975, Vol. 2, No. 1, pp. 45-5 3 Gordon and Breach Science Publishers Ltd.Printed in Great Britain

FIELD ASSISTED GLASS SEALINGGEORGE WALLIS

P. R. Mallory & Co . Inc., Laboratory for Physical Science, Burlington, Ma O1803, U.S.A.(Received March 20 , 1974)

The paper reviews the experimental, theoretical and applications work by several authors on a glass sealingtechnique which was announced five years ago. In the process, an electrostatic field is utilized to promote bondingat relatively low temperatures. The process variables and seal properties are described in detail. A discussion ispresented of the mechanisms which are believed to play a role in bond formation. As an example of the utility ofthe technique, its highly successful application to the mounting of strain gauges is described.

INTRODUCTIONFive years ago, a new glass sealing technique wasannounced which has been variously known as "fieldassisted bonding", "anodic bonding" and "electro-static sealing". Since then, several groups haveworked on the process and a substantial number ofpapers has been published. This article is intended asa review of their work.From the start, the new technique was felt to beunique in several important respects: (1) The surfacesremained solid during the sealing process; (2)noforeign materials such as adhesives or fluxes orsolvents were utilized; and (3)the seals were made atrelatively low temperatures. Among the resultingadvantageous features are the following: since theglass remains solid during sealing, no macro-distortiontakes place; if the glass acts as a structural memberextremely close dimensional tolerances can berealized; if the glass has an optical function, an abilityto make distortionless seals is of particular impor-tance; in vacuum and in device applications it is adesirable feature that seals can be made without theuse of fluxes or solvents which might give rise tocontamination; the low sealing temperature broadensthe area of application to devices whose character-istics deteriorate at higher temperatures. Thus, thenew glass sealing technique appears to have broadapplications in optics and electron optics, in vacuumtechnology and in the encapsulation and mounting ofelectronic components.

2 DESCRIPTION OF GLASS-METAL SEALINGMETHODThe sealing method has been applied to glass-metal,

glass-semiconductor and, more recently, to glass-glass systems. In this section the first two, operation-ally identical cases are discussed using as an example aseal of an iron-nickel-cobalt alloy such as Kovar toa matching glass such as Corning #7052 or #7056. 2The faces that are to be sealed are first lapped andpolished so that they are both flat and smooth. Aftera thorough clean-up, they are placed on top of eachother and heated to a temperature that is typically inthe range 400C to 550C. A dc voltage of about1000 V is then applied between the alloy and the glasssuch that the glass is at a negative potential withrespect to the alloy. If this voltage is maintained fo rmore than a few seconds one finds that the glass andthe alloy form a strong hermetic bond.One of the important process parameters istemperature. In general, sealing at the annealingtemperature of the glass gives good results. In somespecial cases much lower bonding temperatures arepractical. For instance the annealing temperature ofCorning #7052 is 480C, yet good Kovar e7052 sealshave been made at 350C. Similarly, though theannealing point of Pyrex e7740 is 565C, seals ofPyrex to silicon are feasible at a temperature as low as300 C.

The voltage is generally in the range 50 to 1000 Vwith the optimum value depending on the temper-ature and the geometry. If the voltage is applied afterthe parts have reached a steady temperature, it isfound that a relatively large initial current decaysquite rapidly with time. As discussed below it issometimes preferable to apply the voltage at a lowtemperature and maintain it while the parts heat upand dwell at the bonding temperature.A sealing time of one to five minutes is typical. Ingeneral, the sealing time becomes longer as the sealingtemperature and voltages are reduced.

45

8/3/2019 969518 (1)

2/9

46 G. WALLISThe surface conditions of the parts are crucial.

Unless the parts mate very closely and are quitesmooth, they will not form strong hermetic bonds.Most sealing has been done with surfaces having afinish of about 1/ in. rms. Under these conditions nopreoxidation of the metal surfaces is required, butusually a very thin oxide layer forms during thesealing operation. In the case of preoxidized Kovargood seals have been found feasible to surfaces thathad a roughness of up to 6/in. A convenientcriterion fo r the preferred surface mating and surfacesmoothness is that, when the glass is placed on themetal, the parts should be sufficiently close so as togive rise to interference fringes. Furthermore, withthe glass lightly pushed against the metal, no morethan 4 or 5 fringes should be visible over a lineardistance of 25.4 mm. If the sealing temperature isreduced, the requirement for flatness and smoothnessbecomes more stringent.As in conventional glass-metal sealing, it isgenerally advantageous to select glass and metal partswith closely matching coefficients of thermal ex-pansion. However, when the parts are in the form ofthin foils or films, seals to non-matching materials arefeasible. Seals have been made in room ai r and also invarious controlled atmospheres such as oxygen,steam, nitrogen, hydrogen, argon and in vacuum.Many methods are available by which electricalcontact can be made to the glass. For smallgeometries, e.g. 6. 5 mm x 6.5 mm 2 contacting bythe point of a wire is adequate. For larger geometriesit is often convenient to make contact by means ofmetal rings or plates which are placed on the glass.These metal electrodes will not seal to the glass,because they are at a negative potential. In othersituations, painted-on electrodes have been used.Contact to glass tubing has been made by wrapping awire around the tubing.The process works for a large number of glass-metal systems. A partial list of materials is shown inTable I. In some cases, notably molybdenum, atendency has been observed for a fairly thick, poorlyadhering oxide to form during the bonding process.Though the oxide bonds strongly to the glass, theresulting seal is mechanically weak with failureoccurring at the metal-oxide interface. Two pro-cedures have been found to be useful in this case. Inone method, the growth of oxide is impeded by thesimple expedient of applying the bonding voltage atroom temperature and maintaining it while the partsare heated to the bonding temperature. As willbecome evident later, the voltage generates an electro-static force which pulls together the metal and glass

TABLESelected materials that have been sealed by thefield-assisted process.

Material Bonded to glassaKovar 70527056Molybdenum 7052Titanium 0080Tantalum 705 2Carpenter 49 Mosaic F/OComing F/ONiromet 44 Mosaic F/OCorning F/ONiSpan 7740Silicon 77407070Germanium 705 2A1 film on 7740/7052/0120 7740/7052/0120A1 film on CerVIT CerVITSiO film on 7740/7056 7740/7056Si film on quartz quartzNiCr film on 7740 7740SiO2 film on silicon SiO film on silicon

aNumbers refer to Corning glasses.

and obstructs the access of oxygen to the metalsurface. Since the mechanism becomes effective atrelatively low temperatures at which thermal oxi-dation is slow, the result is a substantial reduction ofoxide thickness. In a different method one depositson the polished molybdenum surface a metallic film(e.g. aluminium) which is known to bond easily toglass, thus producing a Mo-Al-glass sandwich ofquite satisfactory strength.Though most work has been limited to the sealingof glass to metals and semiconductors, it has beenfound that field-assisted glass sealing also has applica-tions in the sealing of glass to glass. With some glassto glass systems, bonding is achieved simply byplacing the polished faces on top of each other,heating them to the bonding temperature andapplying a suitable voltage. 3 If the two glasses havethe same composition, then the polarity is obviouslyirrelevent. On the other hand, if the glasses havedifferent compositions, the process usually worksbest if the glass with the higher sodium concentrationis at a negative potential.In another approach, excellent results may beobtained by coating one of the glasses with a thinmetal film (e.g. by evaporation or sputtering) andsealing the second glass to the film. This techniquehas, fo r instance, been applied to the bonding offused quartz parts. In this instance, the intermediatefilm is sputtered silicon and the seal is made at about

8/3/2019 969518 (1)

3/9

GLASS SEALING 47900C with 1000 V. As usual, the silicon should be ata positive potential; electrical contact can be madeeither directly to the film or, if more convenient, tothe coated quartz but within reasonable vicinity ofthe film. Seals of this type have been shown to bestrong and hermetic. Perhaps somewhat surprisingly,much the same results may also be obtained if a filmof silicon monoxide is substituted fo r the metal film.Finally, bonds of silicon to silicon have been madeby the following two methods. In our laboratorybonding wa s accomplished by providing the siliconfaces with fairly thick films of thermal oxide andthen using an electric field to seal the two oxide filmsto each other, thus producing a sandwich of silicon-silicon dioxide-silicon.* At the Research TriangleInstitute, Brooks, Donovan and Hardesty coated oneof the silicon surfaces with a glass layer to which theybonded the second silicon surface. Their methodundoubtedly is applicable to other semiconductorsand metals.

3 SEAL PROPERTIESThe strength of seals has been investigated in con-siderable detail for the system Kovar 47052 glass. 2 Adistribution of tensile strength is shown in Figure 1.Al l seals were fabricated from non-preoxidized Kovarparts with a surface finish of less than 1/ in. whichwere bonded to discs of optically polished 4g7052glass at 50 0C with 1000 V applied fo r 10 minutes.The figure shows that 80% of the seals broke in therange 1500-3000psi. The preponderant failuremechanism was fracture of the glass in a regionclosely adjacent to the glass-Kovar interface. Lessfrequently, failure occurred either within the oxide orat the alloy-oxide interface. Occasionally, failure wasobserved at the glass-oxide interface.

3o00250020001500ooo

to 500A IVot Preoxidize#)

2 3 4 5 6 7 8Surface roughness in Micro inches.

FIGURE 2 Effect of surface roughness on tensile strengthof Kovar #7052 bonds.

As a follow-up to the above experiments, theeffect of surface roughness on seal strength wa sinvestigated. The results, shown in Curve A of Figure2, indicate an unmistakable reduction in strengthwith increasing surface roughness. Concurrently, theyield of hermetic seals dropped sharply, and thefailure mechanism shifted from glass fracture for thesmoothest samples to failure at the interface fo rsamples with a surface roughness of greater than1.5/a in. (see Curve A, Figure 3). Rather surprisingly,the results were quite different when the alloy waspreoxidized by exposing it to air at 350C for 15minutes (oxide thickness of about 1000 )k). As shownin Curve B, Figure 2, the dependence of seal strengthon surface finish becomes much weaker in the range 1to 6/.t in. rms. Hermeticity could be obtained con-sistently, and failure occurred in the glass (Curve B,Figure 3) throughout the range.For purposes of our work, a seal was defined to behermetic if no leakage was indicated on the most

,,, 30AVG =2310 P.S.I.

o o o o o o o oo o o o o o o on o n o n o n 0 ,

IOO8o604O

.- 20

B (treoxidized)

A .(Not Preoxidized)2 3 4 5 6 7

Surface roughness in Micro inches.FIGURE Tensile strength distribution of Kovar #7052bonds.

FIGURE 3 Effect of surface roughness on failure mode ofKovar # 7052 bonds.

8/3/2019 969518 (1)

4/9

48 G. WALLISsensitive scale of a Helium Leak Detector, i.e. ifpossible leakage was less than 5 x 10 -1 o cc He/sec. Onthis basis a large number of glass-metal systems werecapable of producing hermetic seals, such as silicon to#7740 glass, fiberoptics to Niromet alloy, and #7052glass to Kovar alloy. In the latter case, special testswere also made which assured a leak rate of less than10 - cc He/sec.As previously stated, the field-assisted sealingtechnique produces strong, hermetic bonds withoutmacro-distortion in the glass and should therefore beuseful in the assembly of optical components. Anapplication to the sealing of Laser Brewster windowshas been recently described by Bernard Smith 6 whoessentially used the procedure outlined in Sec. 2 tobond elliptic quartz windows to a quartz systemwhile maintaining the Brewster angle within therequired degree of accuracy.The stability of the seals was assured by thefollowing tests:

1) As described2, Kovar #7052 seals, fabricatedat 400C in air, were heated to 700C fo r 30 minutesin atmospheres of air, nitrogen and forming gas. Itwas suspected that such treatments might producediffusion of the respective gases into the interfaceregion of a seal, followed by reaction of the gaseswith the thin oxide layer which invariably formsunder the above sealing conditions. If this were tooccur, it might affect the strength of the seals inaccordance with the predictions of various glasssealing theories. 7,8 In practice, the above heat treat-ments produced no visible evidence that the oxideunder the seal had changed. In tensile tests theheat-treated seals had the same strength as sealswithout heat-treatment and, as usual, the treated sealsfailed in the glass.

2) Subsequent to the application of the bondingvoltage, and while still at the bonding temperature, adc voltage of opposite polarity (metal negative) wasapplied to test whether the bond had any reversiblefeatures. 9 Using tensile bond strength as the criterion,no effect was detected.

3) Thermal cycling and thermal shock tests wereperformed on seals of fiber optics face plates offVeeco Helium Leak Detector MS-9, Veeco InstrumentInc., Plainview, N.Y.This estimate was derived from the characteristics of aphoto cathode in a vacuum tube which contained a field

assisted bond. After the tube was evacuated, it was sealed offand its cathode emission wa s monitored over a period ofseveral months.

o 90

3 7O0

,I2 4 6 8 I0Number of samples.

FIGURE 4 Distribution of failures of Kovar #7052 bondsduring thermal shock test.25.4 mm diameter to flanges of Niromet 44 alloy.The thermal cycling consisted of five cycles atalternately 125C and -25C, and had no detectableeffect on the seals. More instructive was a destructivethermal shock test in which seals were rapidlytransferred from hot to cold water. The temperaturedifference between the two water baths was increasedin 10C steps until a seal cracked or started to leak.Figure 4 shows a distribution of failures at theirrespective temperature differentials. This distributionis quite similar to what is found for conventional fritseals of similar geometry.In summary, all testing to date has indicated thatproperly designed field assisted seals are irreversibleand stable and that their mechanical and thermalproperties are similar to those of conventional seals ofthe same geometry.

4 THEORYA considerable amount of work has been carried outat this Laboratory to tr y and clarify the r61e of theapplied voltage during the bonding process. 1,1. Ithas been established that one function of the field isas follows: When the glass and the metal parts areplaced on top of each other, they contact at only afew points. Over most of their overlapping area theyar e separated by a gap which has a width of about1/. This gap results in interference fringes which arereadily observed. Before any bond formation can takeplace, the parts must be brought into intimatephysical contact over their whole sealing area. This isaccomplished by the applied voltage which sets up anelectrostatic field in the gap. The latter, in turn,generates an attractive force that pulls the parts

8/3/2019 969518 (1)

5/9

GLASS SEALING 49together, and holds them together under considerablepressure.

Since the initial gap between glass and metal isgenerally quite non-uniform, so also is the electro-static force. As a result the parts are usually notpulled into physical contact instantaneously over thewhole area; typically, contacting starts at a singlepoint and then gradually spreads over the remainingarea. This is believed to be advantageous in that ittends to prevent the formation of ai r pockets.

It should be stressed that the above mechanismwould require extremely high voltages at lowtemperatures when the glass resembles an insulatorsince most of the applied voltage would drop acrossthe glass rather than the gap. However, at elevatedtemperatures most glasses behave like electrolytes.In many glasses the mobile species are positive ions,frequently Na derived from the dissociation ofNa20. A voltage of 500 V, applied under theseconditions, deploys almost entirely across the gapbetween glass and metal. With a gap width of aboutone micron, electric fields of the order of5 x 106 V/cm could be expected.Although the temperature may be sufficiently highfo r the generation of electrostatic attraction, it maynot be high enough for rapid bond formation.Nevertheless, as noted before, application of the fieldat a low temperature may be desirable to impedeoxidation of the metal.

After the glass and the metal are brought intocontact, charge begins to pass through the glass. Withthe metal to be bonded having a positive polarity,sodium ions in the glass will migrate toward thenegative electrode. Plated out sodium wa s firstreported in field assisted bonds by DeNee 12 who alsopointed out that the free sodium can combine withadsorbed moisture on the glass to form NaOH. Thelatter will react with the glass and form sodiumsilicate and water.From electrostatic considerations it may bedemonstrated that fo r each Na ion that is plated outat the cathode, a corresponding Na ion is removedfrom a thin glass layer adjacent to the anode (metalto be bonded). At first sight, one might expect thatthis would produce a large charge imbalance whichshould give rise to enormous electric fields and resulteventually in dielectric breakdown. This, however, isnot what is usually observed. The following cases maybe differentiated:

1) Metal ions from the anode are pulled into theglass. If their mobility is comparable to that of thesodium ions, they simply replace the latter. Under

these conditions, the electric field at the interface willnot be greatly different from the field in the bulk ofthe glass, and the anodic metal surface, no longerstrongly attracted, will tend to separate from theglass. This case has been observed in the silver-Pyrexsystem.

1 No bonding could be achieved in thissituation.2) Metal ions from the anode enter the glass as in(1). However, in this case their mobility is orders of

magnitude lower than that of sodium so that fields ofabout 106 V/cm remain at the glass-metal interface.The high field region which typically has a width ofthe order of microns is essentially free of sodiumions. Most of the vacant sodium sites becomeoccupied by the foreign metal ions and the net spacecharge remains relatively small. Hence, the fieldremains below its breakdown value. This mechanismhas been identified in the aluminum-Pyrex systemby means of electrical and electron microprobemeasurements,1 and it is conducive to bond for-mation.A somewhat similar case has been observed byBorom 3 in an iron-sodium disilicate system.Electron microprobe examination clearly showed thatiron replaced sodium in the glass boundary layer.However, in this case only a small percentage of thesodium was removed from the high field region.

3) Under the influence of the high field, oxygenions in the glass drift toward the anode and chemi-cally combine with the metal. Experiments carriedout in a reducing atmosphere demonstrated that inthe Kovar #7052 system oxidation occurred under thebonding glass which covered a part of a Kovar disc.However, where the Kovar was not covered by glass itremained free of oxide. o Similar results were alsoreported by Borom.

4) Other mechanisms could be envisioned. Forinstance, DeNee2 suggested that adsorbed watercould dissociate permitting hydrogen ions to replacethe migrating sodium ions while the hydroxyl ionsmight react with the metal.In the presence of a high field region, as in cases 2,3 and possibly 4, the current flow is expected to beproportional not to the applied voltage but rather tothe exponential of the field. Thus a semilog plot ofcurrent versus field fo r the aluminum-Pyrex systemwas found to be a straight line (see Figure 5 andconsult Ref. 10 as to how the magnitude of the fieldwas measured).

For a given applied voltage, the field is inverselyproportional to the width of the boundary region

8/3/2019 969518 (1)

6/9

50 G. WALLISI00

0. I

X AI ELECTRODE0 AI ELECTRODE 2

2 3E (V/cm) I0 -s

FIGURE 5 Semilog plot of current versus field forAI-Pyrex system.from which the mobile Na* ions have been removed.Initially,, the depleted region is very thin and theresulting field, and hence, the current is large. Asmore Na ions are being removed, the depleted regionbecomes increasingly wide so that there is a steadydrop in current with time. This is illustrated in Figure6 which shows the decay of current over an extendedtime period for the aluminum-Pyrex system.

I00

I (H.a)I0

FIGURE 6system.

I0 I00TIME MIN.

Plot of" current versus time for AI-Pyrex

While the general features of Figures 5 and 6 arequite reproducible, the exact shape usually becomesmodified during the initial part of a bondingoperation because, as mentioned earlier, the parts donot come into contact instantaneously over the wholearea. Thus, the initial current flow is frequently quitenon-uniform.

While there is no doubt that ion migration takesplace during field assisted bonding, it remains to beseen whether it is instrumental in establishing thebond or whether it is merely incidental to it. In thisconnection it is interesting to note that in some ofthe most influential theories on glass-metalsealing,7,8 it is postulated that a satisfactory bondrequires the structure: metal-metal oxide-glassboundary layer saturated with metal oxide-bulkglass. In conventional seals, the glass is liquid duringbonding so that it can attain metal oxide concentra-tions close to saturation at the liquidus. This featureis considered to be essential fo r the formation ofstrong bonds. Bonds made by the field assistedprocess evidently have a similar structure. However,the concentration of metal oxide is determined bythe initial sodium concentration in the glass sinceonly ion exchange appears to be involved. As a result,the concentration is generally much lower than theequilibrium concentration at the liquidus of the glass.As stated above, large electric fields exist in theglass during the bonding process an d are responsibleboth fo r the electrostatic attraction of glass and metaland for the ion migration in the glass. It hassometimes been suggested that the fields persist aftera seal is cooled to room temperature, and byelectrostatic attraction contribute to the bondstrength. This speculation is not borne out byexperiments, which indicate that the fields dissipaterapidly during cooling. In any case, the electrostaticattraction could only make a very minor contributionto bond strength since its magnitude under optimumconditions was measured to be only of the order of35 0 ps i o as compared to the strength of 2000 to3000 psi for a typical glass-metal seal.

5 APPLICATION TO SILICON STRAIN GAUGESAlthough the bonding technique appears to havebroad applications in several, fields,9 this section willconcentrate on its application to the mounting ofsilicon strain gauges because this area has beenresearched most extensively and published data areavailable.

8/3/2019 969518 (1)

7/9

GLASS SEALING 51Conventionally, silicon strain gauges are mounted

to the structural elements by means of adhesives suchas the various expoxies. In some commerciallyavailable gauges the epoxy has been replaced by aglass frit which is far superior to the former in thatmechanical creep is greatly reduced and the tem-perature capabilities of the bond are significantlyincreased. A further improvement in mountingtechnique can be achieved by the substitution of aglass plate fo r the glass frit. The plate, typically 20/thick, is bonded by means of the field assistedtechnology to form an intermediate layer betweenthe strain gauge and the structural member. The glassplate inherently possesses the same advantages as thefrit but, in addition, is preferable in the followingrespects:

1) Glass frits which can be bonded at a relativelylow temperature, e.g. 300C to 350C, usually fail tomatch the silicon with regard to thermal expansion.By contrast, field-assisted sealing permits theselection of glasses that match silicon extremelyclosely.

2) Jigging and fixturing is easier, and alignmentmore accurate, if no liquid phase is present.3) Provided the surfaces are smooth and clean,

very uniform and reliable bonds, free of bubbles, areobtainable by the field assisted technique.

Preliminary experiments in our laboratory in-dicated that good results could be obtained providedthe three materials- silicon, glass, and substrate-had closely matched coefficients of thermal ex-pansion. The same approach could probably beapplied also to metallic foil gauges.In another approach which is particularly suitableto fluid pressure sensors, the strain gauge is anintegral member of a silicon diaphragm which isbonded to a silicon mount. An obvious virtue of thisscheme is the fact that the diaphragm and the mountare of the same material and thus constitute a perfectthermal match. However, this advantage is sometimesobviated by the use of epoxies as the bonding agent.With the use of field assisted bonding several superioroptions are available. In an extension of the previ-ously described approach, a thin glass plate can serveas the "cement" between the silicon parts. Aninteresting variation on this scheme has beendescribed by Brooks, Donovan and Hardesty.Instead of a glass plate, they employ a deposited glassfilm. The film consists of Corning #7740 ("Pyrex")borosilicate glass which is rf sputtered onto one ofthe silicon parts to a thickness of at least 4/. After

deposition, the films are annealed, preferably in anambient of steam, in the range 500C to 900C.Finally, a silicon diaphragm is placed on the coatedsubstrate and bonded by the field assisted process inthe temperature range 450C to 550C with thecoated part being at a negative potential. An unusualfeature of the process is that the dc voltage is appliedin steps as the current decreases with time. Amaximum voltage of 50 V was found to producesatisfactory hermetic seals.In yet another variation of this general approachthe "cement" between the silicon parts is SiO2 ratherthan Pyrex. An advantage of SiO2 is a far highertemperature capability and perhaps an even greaterabsence of creep than in Pyrex. A disadvantage is themarked mismatch in thermal expansion (about30x 10-Tcm/cm, C fo r silicon as compared to5 x 10 -7 cm/cm, C fo r SiO2). A brief description ofthe Si-SiO-Si bonding method follows.4 Thecarefully polished silicon faces that are to be bondedare first provided with an oxide layer having athickness in the range 1. 5 to 2.5/a. In our laboratorythis was accomplished by conventional thermaloxidation at 1200C in an atmosphere of wet oxygen.Presumably, the deposition of an oxide could also beachieved by pyrolytic methods or by sputtering.Whatever the method of deposition, it appears to beimportant that the oxide layers are as perfect andpure as possible so that they can, without dielectricbreakdown, support the large electric fields at hightemperatures that are required for successful bonding.After oxide deposition the silicon parts are placedagainst each other, with the oxide coated faces incontact, and heated to a temperature of about 900 C.A voltage is then applied between the parts for 5 to10 minutes. The procedure differs from the glass-metal sealing process in two respects.

1) In virtue of the symmetry of the situationthere is no preferred polarity.

2) While the magnitude of the applied voltage canvary over a wide range in glass-metal sealing, it iscritical in the bonding of the thin oxide layers, itspreferred value depending primarily on the thicknessof the layers and the perfection of the oxide. Foroxide layers that were 1.5/a thick, a voltage in theneighborhood of 150 volts was found to be appro-priate. If the two oxide films acted as perfectinsulators the resulting electrostatic field wouldamount to only 5 x l0 V/cm and could be expectedto be an order of magnitude lower than the break-down field. Nevertheless, breakdown sometimesoccurs under these conditions, especially if the

8/3/2019 969518 (1)

8/9

52 G. WALLIS

Bond

TABLE II"Span" outputZero shift with shift with temp.

temp. (%/38 C) (%[ 38 C)Compensated"span" output(mV/V)

Compensatedzero pressureoutput (mV/V)

Field-assisted -.098 (+100 C) -.18 (900 psi)-.140 (-40C) --.20W/Ca W/Ca+.52 (+100C) +.09 (40 psi)-.73 (-40C) +.08N/C a W/C

12.9 .42

Epoxy +.78 +-.12 (300 psi)W/C W/CaW/C with compensation N/C no compensation.

12.7 13.5

voltage is applied fo r an excessively long time. Thus,careful empirical definition of the bonding conditionsis required if good yields are to be attained/fTo date, no performance data have been publishedfor strain gauges that were assembled by the abovetechniques. However, some data are available fo r yetanother mounting scheme which has been thoroughlyexplored by Orth and Cannon. 14 Again, the sensingelements are piezoelectric resistors diffused into asilicon diaphragm (Yerman integral silicon trans-ducerlS). For purposes of comparison, the dia-phragms were attached to the polished ends of #7740Pyrex glass tubes which served as primary mountsusing either epoxy or the field assisted bondingprocess. In the latter case specially designed fixtureswere utilized to maintain the required geometry in ahorizontal tube furnace. Bonding was accomplishedat a voltage of kV and at relatively low tempera-tures. The devices were tested as fluid pressuresensors. The data in Table II is taken from the paperby Orth and Cannon and compares epoxy mountedtransducers with those bonded by the field assistedprocess.In addition, the paper presents data on hysteresisand on response to extended static and cyclic load.Transducers mounted by means of epoxy on Invar

This method of sealing silicon to silicon has alsointeresting applications in other areas. For instance, pro-cedures are available by means of which it is possible to thindown one of the silicon layers to almost any desired value. Asa result one obtains a thin silicon film on an insulating layer(SiO) which in turn is backed by a thick silicon wafer.Functionally, this structure is similar to the SOS structure(pyrolytically deposited silicon film on sapphire)which isfinding increasing acceptance in the manufacture of MOScircuitry. However, since the silicon layer in the formerstructure consists of bulk material it usually has superiorelectrical qualities as compared to the deposited film.

typically had hysteresis of 0.1% of full scale output at300 psi. The corresponding average value fo rdiaphragms mounted by the field assisted techniquewas 0.022% with a standard deviation of 0.030. Nobond failures were observed when 4 samples weresubjected to a static load of 1000 ps i for 1000 hoursand 6 samples were cyclically loaded (10 cycles in500 hours). The authors conclude that field assistedbonding of silicon strain gauges removes theuncertainties inherent in epoxy bonds and can beused in the manufacture of simple measurementdevices which have unusually good characteristicsespecially with regard to hysteresis, sensitivity andtemperature compensation.

6 CONCLUSIONSField assisted glass sealing offers a simple and rapidmethod of making reliable, strong hermetic bonds atunusually low temperatures. Of all process variables,only surface flatness and surface smoothness arecritical.

Apparently, the technique is applicable only if thefollowing conditions ar e met:

1) At least one of the parts that are to be bondedto each other should act as an essentially solidelectrolyte at the bonding temperature.

2) Under the influence of the applied dc voltage, ahigh electrostatic field should be se t up at theinterface of the electrolyte and the second part. Thisis equivalent to the requirement that the boundaryregion should be depleted of mobile ions.

3) Unless one of the parts is extremely thin andmalleable, the materials must have reasonablymatching thermal expansion to avoid cracking.

8/3/2019 969518 (1)

9/9

GLASS SEALING 5 3In principle, the technique might be expected towork also if an insulator is substituted for the solidelectrolyte. However, in practice we have not, so far,

encountered a material which behaved as a trueinsulator at the temperature at which bonding wasfeasible.It is probable that as a result of the high field atthe interface ions from on e part migrate into theother. In a few cases, such an exchange has beendocumented experimentally. It remains to be inves-tigated whether the ion migration is instrumental inthe establishment of a low temperature bond orwhether it is merely incidental to it.In the discussion of applications, this paper limiteditself primarily to the very successful mounting ofsilicon strain gauges. This area had received muchattention presumably because of a clearly perceivedneed for technical improvement and perhaps alsobecause the technique could be applied in a perfectlystraightforward manner. However, it is felt that fieldassisted glass sealing has broad applications in fieldsas diverse as optics, vacuum technology and elec-tronics.

REFERENCES1. D. I. Pomerantz, U.S. Patent No. 3 397 278. G. Wallis andD. I. Pomerantz, J. Appl. Phys., 40, 3946 (1969).2. G. Wallis, J. Dorsey and J. Beckett, Am. Ceram. Soc.Bull. 50, (12) 948 (1971).3. S. Wolsky, G. Wallis, D. Pomerantz and J. Dorsey, U.S.

Patent No. 3 589 9654. G. Wallis, Paper 23 9 RNP presented at Electrochem. Soc.Meeting, Atlantic City, Oct. 4-8, 19 70.5. A.D. Brooks, R.P. Donovan and C.A. Hardesty, J.Electrochem. Soc. 119 (4), 545 (1972).6. B. Smith, IEEE J. Quantum Electron, 9, 546 (1973).7. J.A. Pask and R. M. Fulrath, J. Am. Ceram. Soc. 45,592 (1962).8. B.W. King, H. P. Tripp and W. H. Duckworth, Z Am.Ceram. Soc. 42, 504 (1959).9. G. Wallis, J. Automot. Eng. 79, 15 (1971).

10. G. Wallis, J. Am. Ceram. Soc. 53,563 (1970).11. W.A. Weyl an d E.C. Marboe, The Constitution ofGlasses, vol. 11, pt . 2, (John Wiley and Sons, Inc. New

York, 1967) p. 893.12. P. B. DeNee, Z AppL Phys. 40,5396 (1969).13. M. P. Borom, J. Am. Ceram. Soc. 56,254 (1973).14 . E.D. Orth an d P. Cannon, Secretariat RangeCommanders Council, White Sands Missile Range, N. M. ,

Seventh Transducer Workshop, 4-6 April 1972, p. 119.15. A. J. Yerman, U.S. Patent No. 3 537 319.