Stress in thermal Si02 during growth
`E. P. EerNisse
`Sandia Laboratories. Albuquerque. New Mexico 87185
`
`(Received 15 March 1979; accepted for publication 25 April 1979)
`
`Stress present in thermal Si02 at temperatures during growth in wet O2 has been measured as a
`function of growth temperature. During growth at 950·C and below, compressive stress on the
`order of7X 109 dyn/cm2 is generated in the Si02. During growth at 975 and l000·C, the Si02
`grows in a stress-free state. The results, which are consistent with a viscous flow point somewhere
`between 950 and 975 ·C, are of value in avoiding mechanical failure effects in integrated-circuit
`processing.
`PACS numbers: 73.40.Qv, 85.30.Tv, 46.30.Jv, 68.25. + j
`
`The mechanical properties of materials used in Si device
`processing, such as Si, thermal Si02, and deposited Si02
`and Si3N4, are rapidly becoming limiting factors in advanced
`integrated-circuit technology. Numerous workers have
`identified high-temperature dislocation generation in Si be(cid:173)
`neath the edges and/or corners of windows in Si02 and
`Si3 N4 films which seems related to mechanical stress fields.
`These dislocations lead to yield problems as device packing
`density increases. 1 Mechanical stress-induced cracking in
`Si3N~Si02 masking layers at discontinuities degrades
`yield. 2 The dependence of ionizing radiation effects in MOS
`devices on device processing variables can be explained by
`trap creation with viscous flow in thermal Si02 around
`965 .c. 3,4 Viscous flow of thermal Si02 , both steam- and
`dry-02 grown, has been observed around 965 .c. 4 Recent
`theoretical modeling of the growth process for thermal Si02
`suggests that the lack of free volume during growth to ac(cid:173)
`commodate the density differences between Si02 and Si
`should lead to significant stresses during growth and that
`viscous flow effects may play an important role in the growth
`process. 5
`Past works on stress in thermal Si02 involve measuring
`the stresses at room temperature after growth which arise
`from differences in thermal expansion between Si and Si02 .
`The present work demonstrates that large compressive
`stresses on the order of7X 10 9 dyn/cm 2 exist in wet-02-
`grown thermal Si02 during growth at temperatures from
`950 ·C downwards, but that this thermal Si02 grows in a
`stress-free condition from 975 to 1000 ·C, with possible
`small tensile stresses occurring above 1000 ·C. The stress
`present during growth are large enough to cause plastic de(cid:173)
`formation of Si at growth temperatures.
`Experiments were carried out on I-n cm P-doped (111)
`and (100) Si wafers. Cleaned wafers were steam oxidized to a
`4000-A thickness, back lapped to an approximately
`1.5 X 10 - 2_cm thickness, Si etched to remove polishing
`strains on the back surface, and coated on the back with
`approximately 750 A ofCVD Si3N4. Samples 0.5 X2.5 cm
`were scribed and the front surface oxide was removed, leav(cid:173)
`ing samples with bare Si on the front and Si3N4 on the back.
`Some of the samples were further treated to remove the
`Si3N4, grow 4000 A ofSi02 in wet 02 at l000·C on both
`surfaces, and then remove the front Si02 layer, thus leaving
`samples with bare Si on the front and thick Si02 on the back.
`
`In both sample geometries, the intent was to inhibit Si02
`growth on the back side so that stresses occurring during
`new Si02 growth on the front surface will cause bending of
`the Si sample.
`Bending during Si02 growth was monitored as in an
`earlier work 3 by cantilevering the sample inside a furnace
`and reflecting two parallel laser beams off the sample, one
`beam impinging near the clamped end, one impinging near
`the free end. The reflected light beams were directed around
`the room for 786 cm and then the spot separation was mea(cid:173)
`sured. Changes in spot separation ..18 are related to curva(cid:173)
`ture changes LlK by
`LlK = ..18 /2rS,
`(1)
`where S = 786 and r = 1.6 cm is the spot separation at the
`sample (LlK > 0 means convex on the front surface).
`Results for two (111) samples with Si02 back masks
`and for one (111) with a Si3N4 back mask are shown in Fig. 1
`as LlK versus time in a furnace with flowing O2 bubbled
`through water held just below boiling. In the case of the two
`Si02 back-mask samples, the large initial curvature at room
`
`0
`
`0
`
`0
`
`0
`
`0 000
`WOO C
`
`0
`
`0
`
`~
`
`e
`
`..
`•
`
`10000 C
`
`.. .. .... ....
`..
`..
`.. ..
`•
`• • •
`•• • • •
`
`900° C
`
`•
`
`• •
`• •
`• •
`..
`..
`..
`
`.. ..
`
`1111151
`
`0
`
`Sj3~4 BACK SIDE C.1ASK
`
`..... SiO? BACK" DE 1,1AI<
`
`-[
`
`·2
`
`>-
`
`~
`
`00
`
`~
`'" oc
`oc :::
`~
`:=
`z
`~
`~
`~ ·4
`
`10
`
`,0
`
`100
`
`){)[J
`
`TI1\1f IN FURNACE (minute~1
`
`FIG. 1. Observed sample curvature as a function of time during growth of
`thermal SiOz for two (\ II) samples with an SiOz back-side mask (two
`temperatures: 900 and 1000 "C) and for one (III) sample with an Si,N 4 back
`side mask (900 "C).
`
`8
`
`AppL Phys. Lett. 35(1). 1 July 1979
`
`0003-6951/79/010008-03$00.50
`
`© 1979 American Institute of Physics
`
`8
`
` This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP: 72.37.248.4
`On: Tue, 27 Jan 2015 22:40:19
`
`MICRON ET AL. EXHIBIT 1045
`Page 1 of 3
`
`

`
`16 r - - . - - - , - - - , - - - r - - , - - - , - - - , - - - r - - , - - - ,
`
`GROWTH
`TEMPERATURE
`10300 C
`1000
`975
`950
`900
`850
`1000
`950
`900
`
`11
`
`BACK SIDE MASK
`
`~01 ~4l
`
`o
`'V
`
`'
`I III I Si
`
`•
`
`0 o
`: i 1100) Si
`
`•
`
`•
`
`•
`
`w " u z
`0 '"
`
`-4
`
`..
`
`t><I
`t>
`
`t>t>
`<I
`<I
`
`t>
`
`t>
`
`<l
`
`<l
`
`400
`
`1100
`800
`THI CKNESS Of THERMAL Si01
`
`(~)
`
`1600
`
`1000
`
`FIG. 2. Sample curvature induced by stresses in thermal Si02 grown in wet
`O2 on (III) and (100) samples versus Si02 thickness. The line represents
`the curvature introduction expected for an Si02 stress level of 7 X 10 9
`dyn/cm 2,
`
`temperature (time = 0) is caused by the differences in ther(cid:173)
`mal expansion between Si and Si02 • The Si3N4 back-mask
`sample in Fig. 1 has less curvature at room temperature and
`is curved in the opposite direction to that of the Si02 back(cid:173)
`mask samples because the Si3N4 on the back is thin and in
`tension while the Si02 on the back ofthe other two samples
`is thick and in compression. As seen in Fig. 1, after entry into
`the furnace, the samples straighten up as they equilibrate
`with the furnace (which takes approximately 5-8 min). If
`there is stress in the growing oxide, the samples continue to
`bend as seen in the 900 ·C data, which shows continual bend(cid:173)
`ing for times greater than 10 min. Note that the 1000 ·C data
`shows no bending during Si02 growth (times greater than 10
`min, which leads to the conclusion that the Si02 is growing
`stress free at this temperature.
`Data such as that in Fig. 1 were obtained at several
`temperatures around the viscous flow point of 965 .c. Re(cid:173)
`sults were analyzed as follows to obtain a best estimate for
`L1Ko, the value of curvature at growth temperature for zero
`Si02 thickness, 1"SiO,' A plot of raw data such as in Fig. 1
`versus 1" SiO, exhibits an easily identified slope in the data for
`times after about 8 min. The intercept of this straight portion
`at zero 1"SiO, is taken as L1Ko for that sample. The value of
`L1Ko was generally different from zero (flat) because of a
`constant stress in the back mask. The value for L1Ko was
`subtracted from the data taken after 8 min and the resulting
`data replotted versus Si02 thickness 1" SiO,' Thickness 7' SiO, as
`a function of time was estimated by measuring the final oxide
`thickness and using published 7'sio,-vs-time plots for steam
`growth of Si02 shifted to match the final thickness-time
`
`2
`
`data point (a typical run at lOoo·C grew 3200 A in 3 h).
`Results are shown in Fig. 2 for various growth temperatures.
`Note first of all in Fig. 2 that there is a dramatic differ(cid:173)
`ence in going from 950 to 975 or lOoo·C for both the (111)
`and (100) samples. Also note that the results for either back(cid:173)
`mask choice are essentially the same for a given growth tem(cid:173)
`perature, thus assuring that the major effects are due to
`stress (or lack of stress) in the growing Si02 •
`The results for the Si3N4 back-mask samples tended to
`level off as 1"SiO exceeded 600 A. This is not understood at
`present, but the quality of our Si3N4 films and their contin-
`ued masking ability is suspect because their etch rates in
`room-temperature HF are far too high (500 A/min) even
`though the index of refraction (1.97) is consistent with low(cid:173)
`oxygen-content material. The Si02 back-mask results at
`growth temperatures of950 ·C and below show that the sam(cid:173)
`ples continue to curve as long as Si02 continues to grow .
`A line is included in Fig. 2 to indicate the level of stress(cid:173)
`es seen in the Si02 during growth. The curvature change L1K
`is proportional to the integral of the stress through the oxide
`thickness. If we assume that the oxide stress is constant
`through 7' SiO,' L1K is proportional to 1" SiO, and is related to the
`stress in the Si02 , SSiO" by
`
`(2)
`
`where 7'Si is the sample thickness and ESi and VSi are
`Young's modulus and Poisson's ratio, respectively. Letting 4
`ESi(1 - VSi) - 1 = 2.28 X 1012 dny/cm 2, the line in Fig. 2 is
`plotted for SSiO, = 7 X 109 dyn/cm 2. This value of stress is
`large compared to the room-temperature stresses of 3 X 10 9
`dyn/ cm 2 set up during cooldown 7 by the differences in ther(cid:173)
`mal expansion for Si and Si02 • Prediction of the existence of
`stress during growth by examination of the room-tempera(cid:173)
`ture stresses in our samples after coo/down has been incon(cid:173)
`clusive because there is apparently some occasional relief or
`compaction of the stressed films during cooldown. More
`work is needed here.
`Near discontinuities in the growing oxide, shear stress(cid:173)
`es will exist in the Si02 and Si near the SilSi02 interface.
`These shear stresses are of the order of the compressive stress
`in the oxide and are proportional to the integral over 7'SiO of
`the oxide stress. Since the recently measured 6 critical st:ess
`for shear flow in Si at lOoo·C is 3.4 X 107 dyn/cm 2, it is
`entirely plausible to see plastic deformation of Si beneath
`discontinuities in growing Si02 layers (near the edges of
`windows, etc.). In other words, the present results show that
`high stresses can exist in our Si02 at growth temperature
`where the Si is mechanically weak.
`The possibility of a tensile stress in the Si02 growing at
`1030 ·C is seen in Fig. 2. This result is tentative because the
`effect is small and has not been checked yet at higher
`temperature.
`Several runs using dry O 2 were attempted. The much
`reduced growth rate of thermal Si02 in dry O2 made the
`experiments more difficult. Preliminary results suggest that
`the stress levels are much reduced over those seen in Figs. 1
`
`9
`
`Appl. Phys. Lett., Vol. 35, No.1, 1 July 1979
`
`E.P. EerNisse
`
`9
`
` This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP: 72.37.248.4
`On: Tue, 27 Jan 2015 22:40:19
`
`MICRON ET AL. EXHIBIT 1045
`Page 2 of 3
`
`

`
`and 2, but further data must be taken before conclusions and
`interpretations can be made.
`In conclusion, the present results demonstrate that
`large compressive stresses exist in wet-02 -grown thermal
`Si02 during growth at 950·C and below. No stress is ob(cid:173)
`served at 975 and 1000·C with possible small tensile stresses
`seen above 1000 ·C. The results are consistent with viscous
`flow occurring between 950 and 975 ·C during Si02
`growth. A viscous set point around 925 ·C has been inferred
`from measurements of stress in thermal Si02 at the growth
`temperature 7 and was observed directly 4 as occurring
`around 965 ·C. The present results, which treat the stresses
`during growth at the growth temperature, should be of value
`in avoiding mechanical damage effects in VLSI or VHSI
`technologies by careful choice ofSi02 growth temperatures.
`Also, a minimum in stress level at 975-1000·C seems to
`correlate with the optimum conditions for radiation-hard
`dry-02 MOS technology, 8 which is probably not just a coin(cid:173)
`cidence because the viscous flow of both dry-02 - and steam(cid:173)
`grown thermal Si02 occurs around 965 ·C. 4
`
`R.E. Anderson contributed to this work with sample(cid:173)
`geometry suggestions, and J.B. Snelling and T.A. Plut con(cid:173)
`tributed to the experimental setup design and data taking.
`This work was supported by the United States Department
`of Energy (DOE) under contract No. DE-AC04-76-
`DPOO789.
`
`lAo Bohg and A.K. Gaind, Appl. Phys. Lett. 33, 895 (1978).
`'H. Miyoshi, N. Tsubouchi, and A. Nishimoto, J. Electrochem. Soc. 125,
`1824 (1978).
`'E.P. EerNisse and G.F. Derbenwick, IEEE Trans. Nucl. Sci. NS-23, 1534
`(1976).
`'E.P. EerNisse, Appl. Phys. Lett. 30, 290 (1977).
`'W.A. Tiller, J. Electrochem. Soc. (to be published).
`'S.M. Hu, J. Appl. Phys. 49,5678 (1978).
`'J.R. Patel and N. Kato, J. Appl. Phys. 44, 971 (1973).
`'G.F. Derbenwick and B.L. Gregory, IEEE Trans. Nucl. Sci. NS-22, 2151
`(1975).
`
`10
`
`Appl. Phys. Lett., Vol. 35, No.1, 1 July 1979
`
`E.P. EerNisse
`
`10
`
` This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP: 72.37.248.4
`On: Tue, 27 Jan 2015 22:40:19
`
`MICRON ET AL. EXHIBIT 1045
`Page 3 of 3