`
`12 A, Plé6l, G. Kréuter/ Materials Science and Engineering R25 (1999) I-88
`
` a
`
`b
`
`Fig. 9. Sequence of photographs taken with an infrared-sensitive camera, showing the propagation of the bonded area.
`
`energy. Bengstonet al. investigated the dependence of the bonding velocity on the wafer thickness
`[71]. Thicker wafers are less deformable and thus the bonded area spreads slower when thicker
`wafers are involved [71]. However, when using wafers of identical surface quality, no dependence on
`thickness could be observed. Instead the contact wave velocity was found to depend on the gas
`pressure in the interface [72]. The authors carry out bonding experiments at ambient pressure and
`under reduced pressure. The bonding velocity increases with decreasing pressure. The authors
`concludethat the bonding speedis determinedby pressing the gas out of a localised area just in front
`of the propagating bonding front [72].
`Bonding has also been carried out by contacting the wafers under ultra-pure water leaving a
`water film which is much thicker than the commonly observed few monolayers of water in the
`interface [73-75]. If the waferpair is stored at slightly elevated temperatures under vacuum the water
`slowly diffuses laterally out of the interface.
`
`3.5. Interface bubbles
`
`One problem frequently associated with wafer bonding is the formation of interface bubbles,
`sometimesreferred to as voids. In principle there are two different kind of interface bubbles: (a)
`bubbles which occur in the as-bonded interface at room temperature and (b) bubbles which are
`generated at elevated temperatures (typically at 200-800°C).
`Thefirst type is usually caused by surface irregularities, particulates or trapped air. Trapped air
`can be avoidedbyinitiating the bondingin the centre of the wafer pair or bonding in vacuo.Particles
`cause unbonded areas as they prevent
`the wafers from making close contact. The wafers are
`deformed around the particle upon bonding. The resulting bubbles or circular unbondedinterface
`areas are quite large comparedto the actual size of the particle. Tong et al. have investigated the
`bubble diameter as a function of the size of a particle trapped between two wafers [17,56].
`In Fig. 10 the deformation caused bya particle trapped in the interface in shown schematically.
`In Fig. 10(a) the radius h (or height H = 2h) of the particle is much smaller than the radius R of the
`unbondedarea resulting and much smaller than the wafer thickness ty. Using the simple theory of
`
`—OO 7ooo
`twi
`“
`twi
`2h
`twa
`tw2
`R
`G
`
`fT i
`
`Fig. 10. Schematic drawing of a particle leading to an unbonded area [17,56]. (a) Unbonded area with a radius R larger
`than the wafer thickness 4; (b) with a radius, R, smaller than the wafer thickness, f,.
`
`(a)
`
`(b)
`
`Page 1 of 259
`
`SAMSUNG EXHIBIT 1016
`Samsung v. Invensas Bonding
`
`Page 1 of 259
`
`SAMSUNG EXHIBIT 1016
`Samsung v. Invensas Bonding
`
`

`

`%
`
`LMNOPQROSTMUVWQX(YVZ[W(YV\WO]^_\Y`_YW`a&`b\`YYV\`bcdefggghfijj
`    
`
` 

    
`  



  
`
`  
`!"#$%&'(%)$*+$,-$#
`
`*! 
` 

) )

  
`&'!&$./0#1)

&
`
`2
`3
` 04
`
`3
`
`5
`)
`) 
`  

6
`
` 7
` 
`8 9 
` 

9 9    )  
` 
`
`5
`7   
`  :;  ))
`, 
`) 
`

)



6
` #; )  
`
`  
`7  : <
`
` 
`

6) 

) 
` 
`  
`) 

6
` 
`
` 
` 
`=
`
`9
`9   )


`
`) -9
`
`-!.()*$&'1$#>
`

 
`9  5:?



9
`

`


`
`
` 
`@ 

  
`


`
`5
`, 
`)  
`



-


`
` :: )   
` 

A 


` )
` 
`
` )



 
`:: 5
`
`
`

6 

 
`
`  



) )

`
` 

`

  
`
`
`)
`B



)
` )  7 



@
`9
`  
` 
`
` 5

`)


`6

)6
` C  
`?

  9   
`  @5
`9 

 
`@ 


` )






`

D 
`


`

  
`8
` 
` 
`
`

 
` @
`) 
`
`
` 9
`
`E
`   ) 
`EF
`    @ 9
` 
`E9
`

 
` 
`8
`
`

 )




`9
`
`  
` 
`
`
` 8

 
`


`

  
` 

 
`
`  
`)
`
`  
`
`
`7 
`
`  C
`8
`


`

` 
` 
` 
`

`)  
`
`
`
` 
`%

##:GHCI
`)9


` 
`
`  
`9 %::GH
`8 

  9

)




`)  


` 
`
`9:::GH?


` 
`
`   
` ) 
`<

 
`,,
`J



B?

 )

`


` 
`  ),,BC:I
`)9
`
`
` 
`)
`)


`
` )

 


` 
`
` 
`C9

 
`


`

` 
`) )






`)
`   
`    
` = 
`



 
`


`
`

 
`
`


` 
`
`   
`CC#K


` 
`
`


` 
`
`   

 C#?


`

`

 


`
`
`  
`)




` 

 
`
`


`)  8 

9
` )





7


` 


`
`
`Page 2 of 259
`
`SAMSUNG EXHIBIT 1016
`Samsung v. Invensas Bonding
`
`

`

`14
`
`A, Plé6l, G. Kréuter/ Materials Science and Engineering R25 (1999) I-88
`
`Oo
`
`numerical data
`
`
`lengthR[um]
`
`Crack
`
`®@
`
`
`experimental data
`bent beam model: R « h
`weeeee dislocation model:R e« vh
`
`seeeceenees elasto-mechanical instability
`
`Step height h [nm]
`
`[77]. (a) The high-voltage transmission electron
`Fig. 11. Investigation of crack length as a function of step height
`micrograph shows the long crack in the wake of the 18 nm step (marked with an arrow). (Figure reproduced with kind
`permission of the authors) (b) Comparison of theoretically predicted and experimentally observed crack lengths as a
`function of step height. The numerical data were calculated employing the boundary element method. The expected elasto-
`mechanical instability [17,56] is indicated (data taken from [76]).
`
`interface bubbles are nucleated. The bubbles grow by incorporating hydrogen molecules which
`diffuse along the interface [82].
`interface bubbles and the chemical
`temperature-dependent
`To study the formation of
`composition of the trapped gases arrays of cavities of the same size but with different areal
`densities were produced ona silicon wafer by anisotropic etching [83]. The test structure with the
`arrays of cavities is depicted in Fig. 12. At
`the bottom of each cavity only a thin membrane
`
`Page 3 of 259
`
`SAMSUNG EXHIBIT 1016
`Samsung v. Invensas Bonding
`
`Page 3 of 259
`
`SAMSUNG EXHIBIT 1016
`Samsung v. Invensas Bonding
`
`

`

`
`
`A. PléGl, G. Kréuter/ Materials Science and Engineering R25 (1999) 1-88 15
`
`cross section
`
`bonding
`
`—i
`
`density 1/16
`
`ceamtyl2
`
`y
`
`topview
`
`
`density 1/4
`
`density 1
`
`cavity layout on the wafer
`
`Fig. 12. Cavity structure: Test-wafer layout using various densities of cavities [83].
`
`remained. The expansion of the membrane is related to the pressure formed in the cavity. The
`structured wafer was bondedto a bare silicon wafer, with either hydrophilic or hydrophobic surfaces
`under high vacuum. Uponheating the pressure inside the cavities increases. The increase in pressure
`was measured as a function of cavity density, time and temperature applied. The results of this
`experiment are shown in Fig. 13. A higher pressure is observed in the cavities with a lower areal
`density. This can be explained by the larger bonding area around these cavities compared to the
`bonding area aroundthe cavities with a higher areal density. The gasses trapped in the cavities have
`
`100
`
`80+
`
`20
`
`© density 1/16
`x density 1/9
`+ density 1/4
`= density 1
`
`
`
`0—<S——
`0
`100 200 300 400 500 600 700 800
`
`Fig. 13. Pressure increase measured in cavities of bonded hydrophilic silicon wafers as a function of annealing temperature
`and areal density of the cavities (annealing time: 70/) [433].
`
`Temperature
`
`[°C]
`
`Page 4 of 259
`
`SAMSUNG EXHIBIT 1016
`Samsung v. Invensas Bonding
`
`Page 4 of 259
`
`SAMSUNG EXHIBIT 1016
`Samsung v. Invensas Bonding
`
`

`

`
`
`16 A, Plé6l, G. Kréuter/ Materials Science and Engineering R25 (1999) I-88
`
`been characterised by mass spectrometry [84]. The main content of the gas mixture trapped in both
`hydrophilic as well as hydrophobic silicon wafer pairs at a temperature range from room temperature
`to 700°C wasidentified as hydrogen. Minor amounts of water, hydrocarbonsand nitrogen have also
`been found [84]. These results indicate that at temperatures below 500°C hydrogen molecules diffuse
`along the interface until they find a cavity or form an interface bubble around a nucleus. Only above
`about 500°C an appreciable amount of hydrogen diffuses into bulk silicon.
`In hydrophilic silicon wafer bonding hydrogen is formed by the reaction between bulk silicon
`and water (Eq. (7)):
`
`Si + 2H20 — SiO, + 2H.
`
`(7)
`
`The water for this reaction originates from the few monolayers of water which are present on a
`hydrophilic silicon surface. Additional water is formed during the condensation reaction between
`silanol groups (Eq. (8)):
`
`|
`
`— §i-OH +
`
`|
`
`|
`
`|
`
`OH — Si— — — $i 0 — Si- + HO
`
`(8)
`
`The water diffuses to the bulk silicon where it reacts according to Eq. (7).
`The hydrogen in the interface of hydrophobic silicon wafers is formed by the desorption of
`hydrogen atoms which terminate the surface of hydrophobic silicon wafers (Eq. (9)):
`
`|
`
`|
`
`
`
`|
`
`|
`
`(9)
`Hp
`+
`— $i- Si-
`H— $i—
`+
`~ Si“H
`Theresults of the experimentdescribed aboveclearly indicate that interface bubbles formed upon
`annealing are caused by hydrogen. However, for interface bubbles to form the presence of hydrogen
`alone was foundinsufficient; hydrocarbonsas nucleation centres are also necessary [81]. Taking this
`in account methods can be devised to prevent the formation of interface bubbles. One step towards
`the prevention of interface bubbles would be the removal of any thermally unstable organic
`contamination prior to bonding. Mitani et al. could show that hydrophilic silicon wafers which have
`been exposed to oxygen or argon at elevated temperature prior to bonding do not form any interface
`bubbles after bonding and subsequent annealing [81]. A simpler approach to bonding without the
`occurrence of temperature-dependent
`interface bubbles is the treatment of hydrophilic silicon
`surfaces with the dihydrate of periodic acid (HIO,4-2H,O), a strong oxidising agent [50]. Both
`procedures remove organic contamination effectively. As a consequence no interface bubbles are
`generated upon annealing, even though plenty of water and, after annealing, subsequently hydrogen
`(Eq. (7)) is present in the interface. Interface bubbles can also be avoided by bonding wafers which
`are covered with a thermal oxide. The open structure of the thermal oxide is said to allow the
`hydrogen as well as the volatile organic contaminants to diffuse into the oxide, thereby greatly
`reducing the gas pressure at the interface [80].
`
`4, Examination of bonding quality
`
`4.1. Introduction
`
`There is a variety of parameters which characterise a bond interface. The relevance of specific
`properties of the interface depends on the application in mind. For each property, a particular
`
`Page 5 of 259
`
`SAMSUNG EXHIBIT 1016
`Samsung v. Invensas Bonding
`
`Page 5 of 259
`
`SAMSUNG EXHIBIT 1016
`Samsung v. Invensas Bonding
`
`

`

`;]
`
`JKE&LMENOP-,L)#"-QR,#"-%,E.S$%"'$",'+T'U%'""-%'UV WXBYYYZB[\\
`  
` 

    
`
`  
`


` 
`

 

     


`
`  


`  
`

  
`  
`

  

 
` 
`

 
`



 

`    
`
`

  
`

 
`  

 
` 

  
`


`

  

 
`
`

  
` !"#"$#%&'&()'*&'+"+,-",.
`/ 
`

   

00 
` 
` 

  
`

   

      1

2
`1 
`
`
` 2 1

 21

`

2 1


`

2 
`

    
` 



 
`

1

2  3 
`

    1

2    
`
`

  
` 





  


`

   

/ 
`

 
`

  
`

   
` 

 

 
` 

   

   

   4


`  


` 

        
` 5         0   

` 
`


`

  

 
`6789   
` 

 
` 



    


`

 

`   

  :;<
`    
`
`

:=<   

 
`

:>< 
`

    0


`  :?<  
`

`  

 
`

 

  
`

   



  

  
`  

`     
`@ 

 A   
` 



  
`


`

 


`

 0  
`

 

   4
`

  

`


`  
` 

 



   

 


` 
` 



  

  A
`
`
` 
`
`



   
`

   
` 
`

  



`     
` 

  

 
`



 

   
`0
`  

 
` BCD#%$,E#-,'.F%..%&'
`  

    0
`


`   

    6;=9 G2 

 
`

H  
` 

 0  0  
`

       

  67I9
`      
`

  0  A  
` 




`
` 

 

   
`


`
`Page 6 of 259
`
`SAMSUNG EXHIBIT 1016
`Samsung v. Invensas Bonding
`
`

`

`#)
`
`R/ST7UVTWX/Y<3U8:Z<[\3:Z<53T9125Z42Z34M]4654ZZ<546^0_`abbbcadee
`   
` 


`      
`      


` 
` 
`   
`    

`  
`

 
`   
` 
`  
`  

     

`  
` 
`  
`


`   
`

       
`
`    


`
`!  
` " 

  
`  
`
` 
`    
`


`
` 
`

!
`
`
` 




` 
`
`
`


` 
`   
` 
` ! 
`  

`
`


`
`# $#%&
`      


`  
` 
`  

` ' ( 
` 
` 
`       
` 

` 
`
`
` 
`   
`$)*&! 
`  
` 
` 

` 

`#+%, -
`
`
` 
`   


`
`
`
` 
`   
`


`   $))&
`./0/0/1234454632789:52;52<7927=>
`?
`
`
` 
`   
`
`
` 

 


`  
`  
`
`   
`


`
` 


`
`
` 
`  


`     
` 

    @ 
` 


`     
` 
`A 
`   $)B&
`       


`     
`
`
`  


`
`  

   
` 


`

   


`
` 
` 
`
`!
` 
`  

  
`
` 

 
`   
`
`#C #CCDEF@
`
`   
`    
` 
`  

 
` 
`
`
`
`
`  
`    !  
`


`  
`     
`
`
`! 

 
`      
`

 
`  
`  
`
` 
`  

 
`
`
`   


` 
` 
` 
`  
`  
`
`#G  
`  
`    @ @
`
` 
`
` 
`
`  
` 
`      
` 
` 
`
` 
` 
` 

    
`
`    #C  
` 
` 
`
`   
`      
` 
`   
` 
`
`


`

 


`
`
`  
`
` 


` 



 
` 
`  


`
`   
`
` 
`  

!   


` 
`
`   
`


` 
` 
` 
` 
`     
`    
` 
`
`  
`

  
` 

 
`


`
`    
`



  
` 
`! 
`  

` 


`   
`  

 

 
`  
` 
`

 
`
`  GCCDEF
`  

 
` 
`  

` 
` 

G, 
`  

),  %CCDEF
`  

 A
` 
` 

` 


`
`
`
`  HC
`$BC&! 




`  
` 
` 


`
` 
`

F   
` 
`   
` 


`I 
`  
`   
`   

   
`
` 
`  
` 
`
`


` 
` 


`


`
`   
`


`   

 
`

 
`   
` 
`
`
`  


`
` 


`


`
` 
` 
`  
` 
`     
`
` 
` 
`./0/J/KL<3>M5NN<32:574:7=76<3=O>
`P     
`
`
`Q  
`
`  $B#& 

 
`
`     
`  Q 
`        
`
` 
`
`
`
`Page 7 of 259
`
`SAMSUNG EXHIBIT 1016
`Samsung v. Invensas Bonding
`
`

`

`A. Plépl, G. Kréuter/ Materials Science and Engineering R25 (1999) I-88
`
`
`
`Totally reflected waves
`
`
`Partially reflected waves
`
`
`Specimen surface
`mm—— Transmitted waves
`
`Acoustic focal plane
`inside of specimen
`
`A Gate signal
`
`Surface
`
`
`
`reflectediwaves
`signals ——————————>__Time
`
`Intemal
`
`
`Input
`reflected
`
`waves
`
`Fig. 14. Principle of C-scan acoustic microscopy.
`
`planes in the investigated sample causelocal variations in diffracted X-ray intensities which then are
`represented in a two-dimensional image. Therefore, at least one of the bonding partners needs to be a
`single crystal.
`The bonded pair is aligned to meet the Bragg’s condition nA = 2d sin@ (where . is the X-ray
`wavelength, d the spacing of the diffracting planes, 0 the angle between the incident beam and
`diffracting planes and n the order of the diffraction), and while the sample is illuminated with a
`collimated beam of monochromatic X-rays, the ensemble of sample and X-ray recording film is
`scanned in front of the beam (Fig. 15). The elastic distortions in the wake of a void cause
`topographic contrast and thus reveal the presence of a bubble.
`In the absence of suitable X-ray lenses, there is no magnification involved in X-ray topography.
`The spatial resolution essentially is limited by the resolution of the detector, for X-ray film this
`amounts to ca.
`| um. Because of the need for scanning and the low X-ray intensities, long exposure
`times are necessary, making the technique time-consuming and expensive. As an example, an X-ray
`topogram of a 100 mm wafer usually takes several hours.
`For the investigation of bonded wafers, X-ray topography may be used in transmission or in
`reflexion [10,92,93].
`Of the common non-destructive void detection techniques discussed in this article, X-ray
`topography probably is the most sensitive technique; however,
`the equipment
`is not routinely
`available and the method is time-consuming and expensive.
`
`Page 8 of 259
`
`SAMSUNG EXHIBIT 1016
`Samsung v. Invensas Bonding
`
`Page 8 of 259
`
`SAMSUNG EXHIBIT 1016
`Samsung v. Invensas Bonding
`
`

`

`
`
`20 A, Plé6l, G. Kréuter/ Materials Science and Engineering R25 (1999) I-88
`
`bonded pair
`
`lead screen
`
`primary rays
`
`
`
`collimated x-rays
`photographic film
`
`N s
`
`—
`ynchronous
`x-y scanning
`
`Fig. 15. Principle of X-ray topography.
`
`4.2.4. Magic mirror topography
`topography is a simple technique for characterising the
`The magic mirror, or Makyoh,
`morphology of mirror-like surfaces [94,95]. The basic principle of the method is shownin Fig. 16.
`At a small inclination from normal
`incidence, a collimated beam of light illuminates the whole
`surface of the sample underinvestigation. The reflected light is projected into a camera or onto a
`screen or photographic film to record the topographic image. Any deviations from an ideal mirror
`plane can cause contrast in the reflected image. The method is very simple, fast and non-destructive.
`Particles enclosed in the bonding interface or a gas-filled delamination cause the wafer to
`locally bulge. In the case of a wafer with a mirror-polished backside, the convex surface deformation
`due to a bubble can be seen in the reflected image as a dark centre surrounded bya bright ring. As
`the radius of curvature concomitant with an interface bubble depends on the thickness of the wafers
`bonded, the ability of the method to detect voids is higher when a thinner wafer has been chosen as
`reflecting surface. To gain higher sensitivity,
`the wafer can even be thinned and polished after
`bonding [96].
`Okabayashi et al. compared X-ray topography, ultrasonic microscopy and magic mirror
`topography; they found that for a 350 um thick silicon wafer, the magic mirror topography had a
`slightly lower resolution than X-ray topography[94]. In addition to contrast from the delaminations,
`
`Light Source
`
` Sample
`Fim \
`
`Screen
`
`Page 9 of 259
`
`SAMSUNG EXHIBIT 1016
`Samsung v. Invensas Bonding
`
`Fig. 16. Principle of Magic Mirror[94].
`
`Page 9 of 259
`
`SAMSUNG EXHIBIT 1016
`Samsung v. Invensas Bonding
`
`

`

`LA
`
`cdefgehijZSfRZkYSZSeUl S m $ Z $noXpppqXrss
`    
`


`
`

` 



`  

 

 
` 

 
` 
`




`

 

`


` 
` !"# $
`%
`&
`

 
`
`'()  
` 

`
` '*) 
` 

`

  
`

+ & 

 '
` 



`,


`
` 



`

 
`-.
`



,  
`
`
`
` 

 



`
`

 )

`
`,
`

  
`/ 
`

 
` 





 
`

 

'

` ,'

`
`


`  , 
` 



 


`
`,
`
`


` 
` 



  
`


`
`



`0, 1234'356
`7

  
` 

'

  
`   

8  )

`

, 
`
`
`

 



&,

 

/ '

  






`






`)

`
` 
`
`
`




`

`/ 

 
`



 



`
` 
`

  



`

 
`

 
`
`',

`

  

,,, 

`


`  )

`
`
`+
`


`

2396 

 


`
`
`:
`
`,
`

 
`



-

`
` 


`  8
` ;<=>?   
`
`

,


`
`

`/ 
`

`
`

)&


` 
`

 
`
`


`
`<=> 


` 

` 

& 

`
`
`@
`ABCDE
`
`

<=>

`'






`
`


`  


`
`
`,
`
`


`

`  
`


` 
`>
`

2336
`  


`
`
`:
`
`    '
`

 <=>'


`FCB
`

 

G
`




`

 




`

 
`  

/ 
`  


`


`  


`
`


`     

` 
`

,


`
`,
`
`




`

 

H  
` ),  


`   
`




`
`


`
`
`

 '

  

 

, I
` 
`

`'

G
`




`

 
`- 
` 



-
`  
`  
`



<=>   J

  , &


`G
`



 

; -

`(=D'(;K=D?L'>M'>K=D'(>D(==>? ,
`
`


`

8
`
`

   
`


`M777CN  
` 


`

 'I ,
`


`


`  
`
`AOAOBE>LH=FO>L=LO>L=

`3E,

`



  
` 
` ,  
` 
` 2A36
`PQRS TS TS#U
`J
` 



 


`

  


`

` ,
` 
`'  &
`'


`0 

1 
`V 

`  
` ,- 

`
`
`  
`

 ' 

 
`@ W& 
` 

 ' , 




`PXYSURZ[ TURZTS Z$"
`7

 ,
`




` 
`   '

@ W& 
` ,
` 
`

 

 
` 

 
`-/ & 
`   
`'\A'\L'
` 

 
`
`] 
`^\A_\L`\ALa
`;AE?
`
`


`

 \AL/ 

  
`  

 
` 
`

 
` 
` 
`
` '8  
`b2AEECAEL6
`/ 

   

 

 

 
`,  
`  ,

`=,
`1  



   2D46 
`, ,O


`
`Page 10 of 259
`
`SAMSUNG EXHIBIT 1016
`Samsung v. Invensas Bonding
`
`

`

`
`
`22 A, Plé6l, G. Kréuter/ Materials Science and Engineering R25 (1999) I-88
`
`==:
`
`Fig. 17. Double cantilever beam test geometry under constant wedging conditions. The razor blade of thickness 2/ causes
`a crack of length c.
`
`double cantilever beam test geometry under constant wedging conditions, shown in Fig. 17. The
`elastic strain energy in the bent thin plate balances the work of adhesion Wag required to form two
`new surfaces throughthe extension of the crack. A wedge of a thickness of 2h is inserted at the rim of
`the beams into the bond interface so as to debond an area of crack length c. At equilibrium, the
`critical strain energy release rate G;. equals the work of adhesion
`
`Wap = 27 = Gee =
`
`3Eh*d?
`Act?
`
`(1)
`
`with d being the thickness of the beams and FE denoting Young’s modulus in the direction of crack
`propagation [102].
`The work of adhesion canalso berelated to thecritical stress intensity factor, K,; in the case of
`pure modeI loading, this can be expressed as [102-104],
`
`Wap = 27 = Gre =
`
`K7.(1 — v”)
`E
`
`:
`
`(12)
`
`with v denoting Poisson’s ratio. One of the advantages of the stress-intensity factors in the case of a
`known crack geometry is their relation to uniformly applied stress [102], o,:
`
`Kie = foVe,
`
`(13)
`
`with q a geometry term, tabulated for many crack geometries [105].
`Maszaraet al. probably werethefirst to apply the crack-opening characterisation method in the
`study of wafer direct bonding [106]. Following their example, the test customarily is applied to
`complete bonded wafer pairs, and,
`instead of the work of adhesion, most of the wafer bonding
`literature quotes the fracture surface energy yy as a measure for the bonding strength. The crack
`length usually is measured optically, allowing for the additional crack length shadowedbythetip of
`the wedgeandforthat part of the crack narrowerthan ca. \/4 of the probing wavelength. The double
`cantilever beam test applied to complete wafers is sometimes referred to as ‘razor blade’, ‘Maszara’
`or
`‘crack opening’
`test. When executed with the due circumspection,
`the values should be
`reproducible to approximately 10%. However, because of the deviation from the proper double
`cantilever beam geometry, the measurement on complete wafers systematically overestimates the
`fracture surface energy [107]; as shown by Bagdahnetal. through a finite elementanalysis, the error
`increases for increasing adhesion, in their example from about 20% to 80% (Fig. 18).
`With some caution, the test can also be applied to patterned wafers, where only a certain area,
`Ag, of the total wafer area A, is available for bonding [108]. In the case of stripes of bonding running
`
`Page 11 of 259
`
`SAMSUNG EXHIBIT 1016
`Samsung v. Invensas Bonding
`
`Page 11 of 259
`
`SAMSUNG EXHIBIT 1016
`Samsung v. Invensas Bonding
`
`

`

`
`
`A. Plépl, G. Kréuter/ Materials Science and Engineering R25 (1999) I-88 23
`
`—o-Wafer - IR uncorrected
`-G- wafer - IR corrected
`
`—@ Wafer - IR corrected, FEM
`-m-beam-like strips - |R-corrected
`
`energy/J/m?
`fracturesurface
`
`
`1.4
`
`1.3
`
`1.2
`1.1
`
`1.0
`
`0.9
`
`0.8
`
`0.7
`
`0.6
`
`100
`
`200
`
`300
`400
`500
`annealing temperature / °C
`
`600
`
`700
`
`Fig. 18. Fracture surface energy derived from various measurements on wafer and beam-like specimens, after Ref. [107].
`
`parallel to the direction of crack propagation, the apparent fracture surface energy is reduced; this
`was compensated by a multiplicative factor A,/A¢:
`
` _ 3ENa A,
`1 Ect
`A
`
`(14)
`
`Forstripes normal to the direction of crack propagation, the unmodified formula is applicable.
`Whenbonding wafers of different thickness or elastic properties, the work of adhesion could be
`calculated according to
`
`31E\d3End3
`(15)
`Wag = Ge =——1 >.
`2ct(E\d} + End3)
`AB
`I
`
`The case of a thin wafer bonded to a thick one, or the testing arrangement of Flemingetal. [109],
`essentially is the system studied by Obreimoff with
`
`3Eh*d?
`Was = 27 = Ge = ag
`
`(16)
`
`where only one beam is being bent.
`Without proper precautions, the fracture surface energy derived from the blade test cannot be
`identified with the intrinsic surface free energy of a solid. Frequently, the test is not completely
`reversible (cf. also Ref. [37]) and above all, environmental conditions like humidity are known to
`affect the observed fracture surface energy considerably, often causing a decrease with time [110].
`Thesensitivity of the fracture surface energy of silicon dioxide, for instance, towards humidity has
`long been recognised [111]; cracks in pure silicon, however, appear to be rather immune to chemical
`processes [102]. The main appeal of this blade test applied to complete bonded waferpairs is its ease
`of use,
`requiring no special
`sample preparation. When comparing data from different
`experimentators, or when comparing the values with other techniques, the short-cuts taken in the
`customary blade test should be borne in mind. In addition, there is sometimes some ambiguity in
`terminology. The fracture surface energy measured occasionally is referred to as interface energy
`whichis not to be confused with the interfacial energy mentioned above. Bond or bonding energyare
`other terms not unknownin the direct bonding literature. Those terms often makeit difficult to know
`precisely whether the fracture surface energy or the work of adhesion, twice the fracture surface
`energy, is meant.
`
`Page 12 of 259
`
`SAMSUNG EXHIBIT 1016
`Samsung v. Invensas Bonding
`
`Page 12 of 259
`
`SAMSUNG EXHIBIT 1016
`Samsung v. Invensas Bonding
`
`

`

`
`
`24 A, Plé6l, G. Kréuter/ Materials Science and Engineering R25 (1999) I-88
`
`interface or
`In principle, creating a new surface by advancing a crack through the contact
`advancing the bonding front should be equivalent. Suitably structured wafers may serve as tools for
`an ‘in situ’ determination of the work of adhesion. Horning et al. patterned silicon wafers with a
`sequence of parallel lines of known height which serve as wedges [112]. If the line height, line
`spacing and wafer-thickness were chosen appropriately for the work of adhesion, the wafers bonded
`betweena given pair of lines. A variation of line spacings across the wafer included a range of bond
`energies. The bond energy was then measured by a simple IR inspection. The technique was used to
`investigate the influence of surface treatments on the strength of adhesion. With a methodsimilarin
`spirit, the adhesion wasestimated with which two 10 nm thick platinum films, bonded immediately
`after sputter-deposition onto silicon, adhere to each other [113].
`The double cantilever beam test under constant
`loading was used by Lord Raleigh in his
`investigation of the bonding of glass [37]. For a rectangular crack geometry, the strain energy release
`rate is an increasing function of