`Reg. No. 42,557
`
`UNITED STATES PATENT AND TRADEMARK OFFICE
`
`BEFORE THE PATENT TRIAL AND APPEAL BOARD
`
`MICRON TECHNOLOGY, INC., INTEL CORPORATION
`
`AND GLOBALFOUNDRIES U.S., INC.
`
`Petitioners
`
`V.
`
`DANIEL L. FLAMM,
`
`Patent Owner
`
`CASE lPR2017—00392
`
`US. Patent No. 5,711,849
`
`DECLARATION OF DANIEL L. FLAMM IN
`
`SUPPORT OF PATENT OWNER’S RESPONSE
`
`Mail Stop: PATENT BOARD
`Patent Trial and Appeal Board
`US. Patent & Trademark Office
`
`PO. Box 1450
`
`Alexandria, VA 22313—1450
`
`Exhibit 2003
`
`IPR2017—00392
`
`
`
`Inter Partes Review of US. Patent No. 5,711,849
`IPR2017-00392
`
`1, Daniel L. Flamm, Sc.D., hereby declare as follows:
`
`1.
`
`I worked in academia, research, and industry in various roles for more than 50
`
`years. My curriculum vitae, which includes a more detailed summary of my
`
`background, experience, and publication, is attached as Appendix A.
`
`2.
`
`I have been a leading researcher and educator in the fields of semiconductor
`
`processing technology, air pollution control, materials science, and other areas of
`
`chemical engineering. My research has been funded by NASA, National Science
`
`Foundation, Environmental Protection Agency, and AT&T Bell Laboratories.
`
`While a Distinguished Member of Technical Staff at Bell Laboratories, I led a
`
`semiconductor processing research group comprised of research colleagues,
`
`visiting university scientists, post-doctoral associates, and summer students.
`
`I
`
`have also served as a technical consultant to various semiconductor device and
`
`processing equipment manufactures.
`
`3.
`
`I have published over one hundred and fifty (150) technical journal articles
`
`and books, and dozens of articles in conference proceedings, most of them in
`
`highly competitive referred conferences and rigorously reviewed journals.
`
`I am
`
`an inventor listed in more than 20 US. patents, a number of which have been
`
`licensed through the industry, and most being in the general
`
`field of
`
`semiconductor processing technology.
`
`4.
`
`I had experience studying and analyzing patents and patent claims from the
`
`1
`
`
`
`Inter Partes Review of U.S. Patent No. 5,711,849
`IPR2017—00392
`
`perspective of a personal having ordinary skilled in the art (“PHOSTIA”) starting
`
`at least at the time of my employment at AT&T Bell laboratories in 1977. At
`
`AT&T Bell Laboratories, 1 served as a member of the patent licensing review
`
`committee where I was responsible for reviewing hundreds of patents for
`
`potential utility and licensing potential.
`
`I have also served as a technical expert
`
`in patent disputes and litigation.
`
`5.
`
`l was admitted to the patent bar as an Agent in 2003 and have been registered
`
`as a Patent Attorney since 2006.
`
`I am also a member of the California State Bar.
`
`6.
`
`I am the inventor of US. Patent No. 5,711,849,
`
`in the name of Daniel L
`
`Flamm and titled “(“the ‘849 Patent”).
`
`7.
`
`l have read the Petitioners Petition for Inter Partes Review in this matter and
`
`the various art cited therein, including, among others,
`
`8.
`
`Petitioner contends that the combination of Alkire and Kao discloses or
`
`suggests all of the elements of claim 1.
`
`I, however, respectfully disagree.
`
`9. Alkire does not teach “defining etch rate data comprising an etch rate and a
`
`spatial coordinate which defines a position within said relatively non—uniform
`
`etching profile on said substrate” as admitted by Petitioner. Additionally, Figure
`
`2 of Alkire shows a uniform etching profile.
`
`10. As shown in Figure 2, Alkire’s etching profile is completely flat except for
`
`edges of the film, which is typical before or after etching the wafer, and is not a
`
`2
`
`
`
`Inter Partes Review of US. Patent No. 5,711,849
`IPR2017-00392
`
`relatively non-uniform etch profile. Also, Alkire solves for “r” along the length
`
`of the wafer. Alkire emphasizes a “uniform” etching profile in its Abstract, and
`
`throughout the Alkire document. A PHOSITA would never be able to extract a
`
`surface reaction rate constant from this profile, even if the flat profile included
`
`etch data, since a flat profile would not yield a surface reaction rate constant using
`
`any of a model disclosed in Alkire or Kao. Additionally, as shown in Figure 2,
`
`the two films face each other in Alkire, which would give a different model than
`
`Flamm or even Kao. Moreover, Petitioner even acknowledges that “Alkire does
`
`not explicitly disclose measuring the etch rate at any spatial coordinates” in
`
`Petitioner’s paper.
`
`1]. Furthermore, Alkire’s model lacks any temperature dependence, express or
`
`otherwise,
`
`to extract
`
`the claimed surface reaction rate constant, which is
`
`temperature dependent. That is, Alkire assumes “operation is isothermal” [Alkire
`
`p. 649 second column] and excludes any temperature dependence from his model.
`
`At most, Alkire mentions in his conclusion section that because “chemical
`
`reactions are affected by temperature, the rate constant (k2) may not be a constant.”
`
`The extracted surface reaction rate constant that is claimed in the “849 Patent, on
`
`the other hand,
`
`is an Arrhenius expression depending on temperature. [‘849
`
`column 7, lines 10—14].
`
`(Note that the term “surface reaction rate constant” has
`
`a dual meaning, namely: a value of the constant corresponding to a selected value
`
`3
`
`
`
`Inter Partes Review of US. Patent No. 5,711,849
`lPR2017-00392
`
`of temperature, and the function “surface reaction rate constant" which can be
`
`written as an Arrhenius expression depending on temperature.) The claimed
`
`surface reaction rate constant, that is extracted is the function, as is explicitly
`
`taught at column 9, for example, of the “849 patent. The ‘849 patent expressly
`
`requires k(T) to select uniformity conditions for the fabrication of a device [’see,
`
`Figs. 3—5, for example]. Alkire’s model fails to teach the claimed surface reaction
`
`rate constant and would be useless to practice the ‘849 patent. Also, Kao focuses
`
`on non-uniformity of wafers along a reactor not within the wafer so solves for a
`
`completely different problem using a different model. Kao also fails to show or
`
`suggest any relatively non-uniform etch profile as claimed, but rather shows non—
`
`uniformity of wafers along spatial regions of the reactor. Kao also shows non—
`
`symmetric etching rates, which would not be useful in any of Alkire‘s models.
`
`Accordingly, Alkire and Kao lacks the fundamental elements of claim 1 of the
`
`‘849 Patent.
`
`12.
`
`It would have not been obvious for a PHOSITA to combine Alkire with the
`
`experimental measurement of reaction rate and the use of that data in modeling
`
`as taught by Kao. As shown above in Figure 2, Alkire relies upon a set up for
`
`an etching process having two wafer faces facing each other to formulate a model.
`
`Kao, however, uses the following radial flow reactor in Figure 1 for silicon based
`
`etching using a different combination of etching gases. Kao emphasizes solving
`
`4
`
`
`
`Inter Partes Review of US. Patent No. 5,711,849
`IPR2017-00392
`
`his model for the radius of the reactor, which is completely different from the
`
`model of Alkire, which solves for the radius of the wafer, as taught by Alkire.
`
`13. As shown, Figure 1 of Kao shows “r” as the radius of the reactor and solves
`
`the Kao model for “r” along the radius of the reactor. The twelve distinct etch
`
`rate measurements plotted in Figures 8 through 11 of Kao are not symmetrical
`
`across the wafer as shown. Accordingly, a PHOSITA would not use Kao’s
`
`empirical etch rate measurements at twelve distinct spatial coordinates due to the
`
`lack of symmetry for analysis of a plasma etching model to the plasma etching
`
`techniques and model of Alkire. Additionally, Kao focuses on different etch
`
`rates on different spatial regions of the reactor as shown. Specifically, column 1
`
`of page 955 of Kao discusses “[r]adial etch profiles were measured from the point
`
`closest to the reactor exit (point A) to the point closest to the reactor entrance
`
`(point B) for several wafer configurations” to determine etch conditions for
`
`different wafers along different spatial positions of the reactor. Alkire, however,
`
`provides a mathematical model for the reaction between a gas phase etchant and
`
`a substrate film, which faces another substrate film for an etching process, and
`
`completely lacks any experimental data to support or supplement the model.
`
`Additionally, Alkire assumes constant etch conditions for each of the wafers
`
`despite the location of the wafer within the reactor. A PHOSITA would not have
`
`been motivated to improve the model disclosed in Alkire by using any
`
`5
`
`
`
`Inter Partes Review of U.S. Patent No. 5,711,849
`[PR2017-00392
`
`experimental data for independent confirmation of the accuracy of the model as
`
`taught in Kao given that the etch rate data taught by Kao is not symmetrical, and
`
`would not yield the claimed invention.
`
`14.
`
`I disagree that Alkire teaches essentially the same mathematical model for
`
`the etch rate reaction that the 849 Patent uses and explains that when “the same
`
`mathematical model is used to describe the etch process, the solutions should be
`
`the same” as declared by Graves. Rather, as noted above,
`
`the Alkire’s
`
`mathematical model uses (1) a uniform film, rather than a relatively non-uniform
`
`film as claimed; (2) two uniform etchable films facing each other for an etching
`
`process, which would yield a different model from the “849 patent; (3) isothermal
`
`operation, rather than a temperature dependent model, as taught by the “849
`
`patent; and (4) a purely mathematical model, rather than use of etching data, as
`
`taught by the “849 patent.
`
`15.
`
`I also disagree that the two arguments of the Bessel functions in Alkire and
`
`the ’849 patent are the same, and that the models predict the same results, and
`
`that kS in the “849 patent is the same reaction rate constant as Alkire‘s k2. As
`
`noted, Alkire teaches an isothermal model while the “849 patent
`
`teaches a
`
`temperature dependent model, and ks in the ‘849 patent is temperature dependent,
`
`while k2 in Alkire is isothermal. A temperature dependent model of the ‘849
`
`patent is different from an isothermal model of Alkire.
`
`6
`
`
`
`Inter Partes Review of US. Patent No. 5,711,849
`IPR2017—00392
`
`16.
`
`l continues to believe that Alkire specifically teaches away from the use of
`
`‘purely empirical programs of development’, which would teach away from the
`
`use of etch rate data disclosed by Kao.
`
`I assert that Alkire’s statement that
`
`“purely empirical programs of development can be time consuming” does
`
`criticize, discredit, or disparage the use of empirical data to improve the
`
`fabrication of a device. Furthermore, Kao’s emphasizes that “the model predicts
`
`a larger effect of pressure on etch rate than observed,” which confirms
`
`incompatibility to discredit or discourage investigation. Kao tests “sensitivity of
`
`predicted etch rate” in a failed attempt to overcome the conflict between data and
`
`experiment. Kao finally admits that “the pressure dependence of ki is unknown
`
`and further work is warranted” tacitly acknowledging that a PHOSITA could not
`
`rely on his models or data and further discredits or discourages investigation.
`
`Additionally, Kao further supports the difficulty of using his etch rate data by
`
`way of the lack of correspondence in Figures 10 and 11, and concludes by
`
`emphasizing “discrepancies between the model and the experimental data for
`
`wafers located near the reactor exit are difficult to explain and could be the effect
`
`of plasma nonuniformities at the reactor center, or an incomplete description of
`
`the plasma chemistry” to further discredit or discourage investigation.
`
`Accordingly, a PHOSITA would not combine the etching data or teaching of the
`
`use of etching data from Kao with Alkire or any other reference in view of
`
`7
`
`
`
`Inter Partes Review of US. Patent No. 5,711,849
`IPR2017—00392
`
`established knowledge that Kao’s chemistries and reactor design would yield a
`
`wide range of variation.
`
`17.
`
`I disagree that Kao teaches a PHOSITA to conduct experiments to determine
`
`etch rate data and then extract a surface reaction rate constant
`
`from the
`
`experimental data, thus connecting the purely mathematical model of Alkire to
`
`empirical data for the etch rate data to extract the surface reaction rate for defined
`
`process conditions to test, validate, and improve Alkire’s model. That is.
`
`I
`
`disagree that a PHOSITA would have looked to Kao to use empirical analysis
`
`test and validate the model of Alkire with actual data of Kao. As discussed, the
`
`model of Alkire is for an isothermal process with each wafer having the same
`
`etching characteristic as other wafers within the reactor. The wafers are facing
`
`each other, which is uncommon an etching process. Alkire‘s model does not
`
`even use a relatively non-uniform etching profile, but rather uses a uniform
`
`etching profile, which is completely flat.
`
`In contrast, Kao focuses on different
`
`etching rates for different wafers spatially positioned within the reactor for a
`
`silicon etching process. Kao’s model and data do not even correlate with each
`
`other, and have severe shortcomings as discussed. Accordingly, a PHOSITA
`
`would not connect the purely mathematical model of Alkire to empirical analysis
`
`to test and validate the model of Alkire given that the data derived from Kao is
`
`for a completely different purposes and use.
`
`8
`
`
`
`Inter Partes Review of US. Patent No. 5,711,849
`IPR2017—00392
`
`18. A PHOSITA would not have been motivated to improve the model disclosed
`
`in Alkire by using experimental data to provide independent confirmation of the
`
`accuracy of the model as taught by Kao. Again, as discussed, the model of Alkire
`
`is
`
`for an isothermal process with each wafer having the same etching
`
`characteristic as other wafers within the reactor. The wafers are facing each other,
`
`which is uncommon for an etching process. Alkire’s model does not even use a
`
`relatively non-uniform etching profile, but rather uses a uniform etching profile,
`
`which is completely flat.
`
`In contrast, Kao focuses on different etching rates for
`
`different wafers spatially positioned within the reactor for a silicon etching
`
`process. Kao’s model and data do not even correlate with each other, and have
`
`severe shortcomings as discussed. The experimental data in Kao did not
`
`independently improve Alkire’s model or even Kao’s model. Accordingly, a
`
`PHOSITA would not have been motivated to improve the model disclosed in
`
`Alkire by using experimental data to provide independent confirmation of the
`
`accuracy of the model as taught by Kao.
`
`19. A PHOSITA would not find it obvious to use the empirical analysis of Kao
`
`in the similar plasma etching reaction of Alkire to test and validate the model of
`
`Alkire. As discussed, the etching processes in Alkire and Kao are not similar at
`
`all. Alkire taught an isothermal process with each wafer having the same etching
`
`characteristic as other wafers in the reactor. The wafers are facing each other,
`
`9
`
`
`
`Inter Partes Review of US. Patent No. 5,711,849
`lPR2017—00392
`
`which is uncommon, and for an etching process.
`
`In contrast, Kao focuses on
`
`different etching rates for different wafers spatially positioned within the reactor
`
`for a silicon etching process. The etching reactor design in Alkire is also different
`
`from Kao’s radial flow reactor. Accordingly, the etching processes in Alkire and
`
`Kao are not similar and therefore a PHOSITA would not combine them together.
`
`Even if they were combined, the combination would still not lead to the claimed
`
`invention.
`
`20.
`
`I further emphasize that Petitioner misinterprets the significance ofKao‘s ks,
`
`which is not a determined reaction rate constant but instead arises as one of three
`
`empirical fitting parameters, each of which alone or in combination, fails to
`
`reasonably fit Kao’s experimental results. As shown in the “849 patent,
`
`I
`
`explained in great detail on how k6 is established, and has also supplemented this
`
`Response with the Declaration from Dr. Daniel L. Flamm.
`
`I would like
`
`reconsideration of its arguments and does not agree with the testimony from Dr.
`
`Graves, which is not as detailed and complete as my analysis.
`
`21.
`
`In view of the foregoing, I would respectfully like for the Board to reconsider
`
`the Decision on claim 1 and challenge Petitioner’s arguments regarding all claims
`
`based upon the combined teachings of Alkire and Kao.
`
`22.
`
`I declare that a PHOSITA would not have combined Alkire with Kao.
`
`23. First, it would not have been obvious to a PHOSITA to improve Alkire with
`
`10
`
`
`
`Inter Partes Review of US. Patent No. 5,711,849
`IPR2017—00392
`
`the use of actual experimental data, such as disclosed in Kao, and there would be
`
`no motivation to do so. Alkire specifically teaches “purely empirical programs of
`
`development,” which does not require the use of etch rate data disclosed by Kao.
`
`Nothing in Alkire shows or suggests that Alkire intended to have Alkire’s
`
`model‘s supplemented in anyway by experimental data, was silent on the use of
`
`any experimental data. Kao’s, however, did attempt to validate Kao’s own
`
`models with experimental data, but failed miserably. Only empirical curve fitting
`
`seemed to accommodate the use of any of Kao’s data, rather than Kao’s models.
`
`In particular, Kao focused on the “analysis” of non-uniformities in the plasma
`
`etching of silicon, using only one specific feed gas composition, with a deficient
`
`reactor design which resulted in the “large degree of nonuniformity in etch rate”,
`
`“discrepancies between the model and experimental data,” and “unpredictable
`
`experimental results,” each ofwhich supports Kao’s failure in using experimental
`
`data to confirm a mathematical model, and a PHOSITA would not combine
`
`Kao’s use of experimental data with the theoretical model of Alkire.
`
`24. Additionally, a PHOSITA would never use the actual experimental data or
`
`related techniques reported by Kao in combination with any modeling technique
`
`or with any other rector design and related process.
`
`In particular, Kao was
`
`etching wafers in a tragically deficient geometry using a feed gas mixture known
`
`to generate a complex product composition that shifts widely responsive to
`
`ll
`
`
`
`Inter Partes Review of US. Patent No. 5,71 '1 ,849
`lPR2017—00392
`
`variations in flow, pressure, and electrical discharge parameters. Specifically, it
`
`was well known that the concentration of silicon etchant made from Kao’s
`
`mixture—thought
`
`to
`
`be mostly
`
`fluorine
`
`atoms—and the
`
`concomitant
`
`concentrations of species that inhibit silicon etching (e.g., oxygen atoms that can
`
`oxidize a silicon surface, and/or flurocarbon species that can inhibit etchant
`
`attack) vary greatly as wafer surface area, RF power, pressure, flow, temperature
`
`and other discharge variables are changed.
`
`(See David Edelson and Daniel L.
`
`Flamm, Computer Simulation ofa CF4 Plasma Etching Silicon, 56 J. APPLIED
`
`PHYSICS 1522 (1984), attached hereto as Exhibit 2001; see also l.C. Plumb and
`
`KR. Ryan,A Model ofthe Chemical Processes Occurring in CF4/02 Discharges
`
`Used in Plasma Etching, 6 Plasma Chem. and Plasma Processing 205 (1986),
`
`attached hereto as Exhibit 2002.) Accordingly, without any knowledge of Kao’s
`
`gaseous product composition, there was no known way to extract an extensible
`
`etch rate based on profile or point etch rate data measured using a CF4/O2 plasma,
`
`in Kao’s reactor, and the Kao reference did nothing to change this.
`
`25. Kao supports this conclusion. He explicitly states, “the model predicts a
`
`larger effect of pressure on etch rate than observed,” which confirms
`
`incompatibility. Kao tests “sensitivity of predicted etch rate" in a failed attempt
`
`to overcome the conflict between data and experiment. Kao finally admits that
`
`“the pressure dependence of ki is unknown and further work is warranted” tacitly
`
`12
`
`
`
`Inter Partes Review of US. Patent No. 5,711,849
`IPR2017-00392
`
`acknowledging that
`
`a PHOSITA could not
`
`rely on his models or data.
`
`Additionally, Kao further supports the difficulty of using his etch rate data by
`
`way of the lack of correspondence in Figures 10 and 11. Finally, Kao concludes
`
`by emphasizing “discrepancies between the model and the experimental data for
`
`wafers located near the reactor exit are difficult to explain and could be the effect
`
`of plasma nonuniformities at the reactor center, or an incomplete description of
`
`the plasma chemistry.” Most critically, Kao did not measure chemical
`
`composition or any products of his complex plasma chemistry,
`
`the identity,
`
`. relative concentrations, and spatial distributions of species interacting with his
`
`silicon wafers were unknown. Accordingly, a PHOSITA could not combine the
`
`etching data from Kao with Alkire or any other reference in view of established
`
`knowledge that Kao’s chemistries and reactor design would yield a wide range
`
`of variation. No PHOSITA would ever use results or data or the concept of using
`
`data from Kao’s analysis for the fabrication of any device with Alkire’s model.
`
`26. As further background, and reemphasized herein, this Petition is not the first
`
`initiated against the ‘849 patent. Lam Research Corp. initiated a petition for inter
`
`partes review ofthe '849 patent based upon Battey. Dr. Flamm overcame Lam’s
`
`invalidity arguments based upon Battey.
`
`(Lam Research Corp. v. Flamm,
`
`lPR2016—0466, Paper No. 7, at 8-10 (Jul. 19, 2016).) The Petitioners now argue
`
`that: “Kao, unlike Battey, explicitly discloses the calculation of a surface reaction
`
`13
`
`
`
`Inter Partes Review of US. Patent No. 5,711,849
`[PR2017-UU392
`
`rate constant (k6) based on the empirical measurements of an etch rate profile.
`
`Accordingly, Kao is materially different than Battey.”
`
`27. Petitioners, however, misinterpret the significance of Kao’s kc, which is not
`
`a determined reaction rate constant. In actuality, it arises as one of three empirical
`
`fitting parameters, each of which alone or in combination, fails to reasonably fit
`
`Kao’s experimental results, despite that they were for a rather narrow range of
`
`conditions using only one rf power level and a single temperature in the reactor.
`
`28. Kao discloses that what are tantamount to three fitting parameters, kd kg, kc
`
`(shown in Table 2), were calculated by iteratively using a general finite element
`
`numerical equation solver code (TWODEPEP) to solve greatly oversimplified
`
`transport/reaction equations (8a—b) they applied to represent a very complex
`
`system, then comparing the solutions obtained for various combinations of etch
`
`rate values to obtain the set that most closely representing the etch rates of silicon
`
`they observed. The single set of system parameters are listed in Table l and the
`
`radial flow reactor is shown in Fig. 1. Kao never discloses how such analysis
`
`was performed, explicitly or implicitly, but merely recites that TWODEPEP is
`
`available with reference “22. International Mathematical and Statistical Libraries
`
`TWODEPEP, 1983.” A PHOSITA would not have been able to understand or
`
`use a general finite element numerical equation solver code (TWODEPEP) in this
`
`manner.
`
`14
`
`
`
`Inter Partes Review of US. Patent No. 5,711,849
`IPR2017-00392
`
`29. Unsurprisingly, Kao’s fitted parameters do not work well, as shown by
`
`comparing the effects of pressure (at Fig. 9), on the etch rates amongst different
`
`wafers (at Figs 10, 11) or by comparing the parameter identified with “etch rate”
`
`from the Kao three parameter fit to the accepted etch rate constant for fluorine
`
`atoms (which Kao assumes is the only surface reacting species emanating from
`
`this complex chemistry).
`
`In fact, the authors admit as much: “At 70°C, kC from
`
`(20) has a value of 49 cm/s, which is a factor of six less than that found from our
`
`work.” Kao further states: “The three rate constants [i.e., the number used to fit
`
`the solver to their data] are assumed to be independent of pressure. However,
`
`when the model was use to predict the etch rates for the data of Fig. 4 (pressure
`
`effect), the model predicts a larger effect of pressure on the etch rate than was
`
`observed.
`
`.
`
`.
`
`. The reason for this pressure dependence of kt is unknown and
`
`further work is warranted.”
`
`30. Kao thereby acknowledges, as a PHOSITA would, that even using the exact
`
`same equipment and general process conditions (e.g.,
`
`the reactor parameters
`
`shown in Table I) which the fitted parameters (kd, ke, kt) were based on, these
`
`“rate constants” do not agree with their model predictions. Moreover, even if
`
`their original assumption that ki is as constant is arbitrarily changed to make kt
`
`proportional to pressure, things do not get much better.
`
`31. No PHOSITA would consider Kao’s data, modeling, or associated
`
`15
`
`
`
`Inter Partes Review of US. Patent No. 5,711,849
`IPR2017-00392
`
`techniques useful for anything, never mind obtaining a surface reaction rate
`
`constant that could be useful
`
`in the fabrication of a device or be useful
`
`in
`
`motivating a PHOSITA to combine with Alkire’s model.
`
`32. I declare under penalty of perjury under the laws of the United States of
`
`America that the foregoing is true and correct.
`
`Executed on this 6th day of September, 2017.
`
` 92M
`
`Daniel L. Flamm
`
`16
`
`