Journal of Vntuum Stience 8. Technology A
`
`Etch mechanism in the reactive ion etching of silicon nitride
`
`J. Dulak, B. J. Howard, and Ch. Steinbriichel
`
`Citation: Journal of Vacuum Science & Technology A 9, 775 (1991); doi: 10.1116/1.577360
`
`View online: http://dx.doi.org/10.1116/1.577360
`
`View Table of Contents: http://scitation.aip.org/content/avs/journal/jvsta/9/3?ver=pdfcov
`
`Published by the AVS: Science & Technology of Materials, Interfaces, and Processing
`
`Articles you may be interested in
`Selective reactive ion etching of silicon nitride over silicon using CHF3 with N2 addition
`J. Vac. Sci. Technol. B 13, 2008 (1995); 10.1116/1.588124
`
`Reactive ion etching of gallium nitride in silicon tetrachloride plasmasa)
`Appl. Phys. Lett. 63, 2777 (1993); 10.1063/1.110331
`
`Reactive ion etching of plasma enhanced chemical vapor deposition amorphous silicon and silicon nitride:
`Feeding gas effects
`J. Vac. Sci. Technol. A 8, 1702 (1990); 10.1116/1.576833
`
`Selective reactive ion etching of silicon nitride on oxide in a multifacet (“HEX") plasma etching machine
`J. Vac. Sci. Technol. A 7, 1145 (1989); 10.1116/1.576243
`
`Reactive ion etching of silicon
`J. Vac. Sci. Technol. 16, 410 (1979); 10.1116/1.569962
`
`AVS’ Advance your technology or engineering
`career using the AVS Career Center, with
`hundreds of exciting jobs listed each month!
`
`http:/lcareers.avs.org
`
`INTEL 1010
`
`

`

`Etch mechanism in the reactive ion etching of silicon nitride
`J. Dulakf’ B.J. Howard,and Ch. Steinbriichel
`Materials Engineering Department and Centerfor Integrated Electram'cs, Rem'seZeerPoIyzeclmic Institute,
`Troy, New York 12180
`
`{Received 17 September 1990; accepted 9 December 1990)
`
`Reactive ion etching of silicon nitride with CHF3/O2 plasmas has been studied in a hcxode reactor
`and compared to silicon dioxide etching. Measurements of etch rates as a function of gas
`composition and pressure were combined with Langmuir probe data for the ion flux to the
`substrate to give etch yields (number of substrate atoms removed per bombarding ion). At low
`oxygen content, the etch yields of both materials are about 2 atoms/ion and are essentially
`independent of pressure between 10 and 60 mTorr. At higher oxygen content, the etch yield is
`considerably larger for silicon nitride, and for both materials the etch yields increase with
`increasing pressure. From these results it can be concluded that in a CHF3/O2 plasma at low
`oxygen content, the etch mechanism is mostly direct reactive ion etching for both silicon nitride
`and silicon oxide. On the other hand, at higher oxygen content the etching is ion enhanced for
`both materials, but to a much greater extent in the case of silicon nitride.
`
`E. ENTRODUCTEON
`
`IE. EXPEREMENTAL DETAil.,S
`
`Silicon nitride (Si3N4) is a frequently used material in mi-
`croelectronics technology. Its main applications are as a pas-
`sivation layer and as an oxidation mask.‘ There have also
`been proposals to use Si3N4 together with SiO2 as a multi-
`layer gate dielectric? In all of these applications Si3N4 needs
`to be patterned, for which operation the preferred technique
`now is dry etching in a plasma.
`Si3N4 is generally etched in a fluorocarbon plasma, i.e.,
`with a fluorine-based plasma chemistry? The main dill":-
`culty in etching Si3N,, is to achieve anisotropic etching with
`good selectivity relative to SiO1. The best selectivity is ob-
`tained in a barrel reactor‘ or an afterglow geometry5 with an
`F-atom-rich plasma, but this leads to isotropic etching. On
`the other hand, anisotropic patterning can only be done by
`reactive ion etching in an F-atom-poor, unsaturaterich plas-
`ma, where Si3N4 and SiO2 etch at comparable rates.3*5‘“ To
`‘circumvent these problems, a two~step process has been im-
`plemented, in which during most of the etch anisotropic
`plasma conditions are used but the etch is completed under
`conditions giving high selectivity.” The plasma etching of
`Si3N4 is sometimes said to be intermediate between SK); and
`Si,3"° but the underlying reasons for the similarities and dif-
`ferences between the behavior of Si3N4 and Si02 have not
`been identified. The details of the etching of Si3N4 also de-
`pend on whether the material was deposited by low-pressure
`chemical vapor deposition (LPCVD) or by a plasma-en-
`hanced process?
`in this work we reexamine the reactive ion etching of
`Si3N,, and Si02 in CHF3/O2 plasmas but, with the help of
`Langmuir probe data, we focus on comparing etch yields,
`rather than etch rates. This will allow us to clarify the etch
`mechanisms involved. In particular, we will be able to show
`to what extent direct reactive ion etching on the one hand
`and ion-enhanced chemical etching on the other hand are
`important for the two materials.
`
`All our experiments were performed in an Applied Mate-
`rials AME-8130 hexode reactor loaded with 23 Si wafers and
`
`one Si3N4 or SiO2 wafer to be etched. Standard plasma con-
`ditions were IOO scorn of total gas flow and l()0O W of radio
`frequency (rf) power, with the gas pressure variable from 10
`to 60 mTorr. Two gas mixtures were investigated:
`CHF3/l0% O2 and CHF3/20% 02 by flow. (Under these
`conditions, etching of the Si was very small, thus represent-
`ing a minimal load on the plasma) . The substrates were ther-
`mal SiO2 ( ~ 5000 13;, grown with O2 and H2 at 1000 °C) and
`LPCVD Si3N,, (~ l50O [1, grown with NH, and SiH2Cll at
`780 °C). Etch rates were measured interferornctrically on a
`substrate wafer in the top position on a hexagonal face,
`whereas Langmuir probe data were taken over a substrate
`wafer in the second position from the top.
`The Langmuir probe was L shaped, with the active part
`consisting of a tungsten wire of 0.5 mm diam and l0 mm
`length, and was mounted on a flange over one of the wafer-
`holding hexagon faces. The probe could be moved parallel as
`well as perpendicular to the substrate wafer surface. The
`sheath edge above the wafer was located from the probe posi-
`tion giving maximum ion current. lon density, electron tem-
`perature, and ion flux to the substrate wafer were deter-
`mined as described previously.“
`The etch yield Y, i.e., the number of substrate atoms re-
`moved per bombarding ion, was calculated from the etch
`rate R? and the ion fluxf, by
`/'
`: R¢,p]\ 9
`M}?
`
`( 1)
`
`where p is the density of the substrate film (3.4 g/cur’ for
`Si3N4 and 2.3 g/cm‘ for SiO2), Nis the number of atoms per
`molecular unit (N 2 7 for Si3N4 and N 2 3 for SiO2 ), and M
`is the molecular weight of the substrate material. The ion
`flux j", is given by”
`
`775
`
`J. Vac. sci. Technoi. A 3 (3), May/Jun 1991
`
`0734-2101/9‘!/030?75=-04$0‘l .00
`
`Ce) 1991 American Vacuum Society
`
`775
`
`
`
`
`
`
`
`

`

`776
`
`Dulak, Howard, and Steinbruchel: Etch mechanism in the HIE of Si3N.
`
`'
`
`776
`
`. . . . .
`
`.
`
`.
`
`. ,
`
`,
`
`,,
`
`€000 '
`
`30000 "
`
`lpz(uhz) 20000 -I
`
`
`
`Re(xlmin)
`
`
`
`
` 10000 " ' . '3
`
`'16!)
`"-140
`=-=l2O -100 -80
`"60
`-40
`-20
`
`V5: (volt)
`
`BOO '
`
`600
`
`400 '
`
`200
`
`10
`
`20 3 40
`Pressure (mT)
`
`so 5 70
`
`FIG. 1. Plot of square of probe current I E, vs probe voltage V,, in the ion
`saturation regime, where IF is due to collected ions. Slope ofcurve is propor-
`tional to N
`
`FIG. 3. Etch rates for Si, N4 vs pressurefl : CHF, -10% O2; 0: CHF3 -20%
`01.
`
`f,- =N,.<kT,./m,.>"2.
`
`(2)
`
`where M is the ion density, Te the electron temperature, and
`m, the mass of a CHFJ ion.
`
`III. RESULTS
`
`1 and 2 we give an example of Langmuir probe
`In Figs.
`data for the ion current and the electron current to the probe.
`One can see that the square of the ion current is a linear
`function of the probe voltage VP as is expected in a low-
`density plasma.” From this curve the ion density N, can
`readily be deduced." Also, the natural logarithm of the net
`electron current 1,. (probe current minus ion current) is a
`linear function of the probe voltage over about two orders of
`magnitude, so that an electron temperature Te can be asso-
`ciated with the electron energy distribution. (Note that we
`have always observed a slight s shape in the In [8 versus Vp
`curves). Typical figures obtained for the ion density and the
`electron temperature were N,» = 8 X 109 cm‘3 and T9 = 4.5
`eV under the given plasma conditions.
`Figures 3 and 4 display etch rates versus pressure. We
`point out that in pure CHF3 neither Si3N4 nor SiO2 etched.
`
`Furthermore, for Si3N4 at 40 and 60 mTorr of CHF3/10%
`02 the etch rate slowed down noticeably over time (of.
`Ref.8), so that in those two cases the initial etch rates were
`used in Fig. 3. In all other cases the etch rates were constant
`in time. Figures 5 and 6 present etch yields corresponding to
`the etch rates of Figs. 3 and 4.
`
`IV. DISCUSSION
`
`For the purpose of discussing the above results we men-
`tion first that increasing the 02 content in an CHF3/O2 plas-
`ma decreases the amount of fluorocarbon unsaturates but
`
`increases the amount of free F atoms in the plasma, whereas
`increasing the pressure has the opposite elfect, i.e., it in-
`creases the amount of unsaturates} It is then remarkable
`
`that neither 02 content nor pressure have a significant effect
`on the etch rate of SiO2 (Fig. 4) , but an increased 02 content
`leads to an increase in the etch rate of Si3N4 by a factor 2 or
`more (Fig. 3).
`The significance of these results becomes clearer if one
`focuses on the etch yieids. At 10% 02 the etch yield de-
`creases siightly with increasing pressure for Si3N4 and re-
`mains practically constant for SiO2 (Fig. 5). At 20% 0.2 the
`
`1000
`
`Inlo
`
`9;) (volt)
`
`
`
`Re(Xian...)
`
`800
`
`600 '-
`
`400 '-
`
`200
`
`0
`
`1 0
`
`20
`
`30
`
`40
`
`50
`
`60
`
`70
`
`Pressure (mT)
`
`FIG. 2. Natural logarithm of electron current I. vs probe voltage Vp. Slope
`of curve determines electron temperature '11,.
`
`FIG. 4. Etch rates for Sit), vs pressure. El: CHF3-10% O2; 0: CHF, -20%
`0..
`
`J. Vac. Sci. Technoi. A, Vol. 9, No. 3, May/Jun 1991
`
`
`
`
`
`
`
`

`

`777
`
`Duiak, Howard, and Stelnhruchel: Etch mechanism in the ME of Siam
`
`777
`
`,.
`
`f“\
`
`ZQ K EO
`
`l““'
`<\/
`
`Q.
`
`..J
`Li;
`is-i
`>-=
`3:
`(J
`I"
`
`fact that under those conditions the etching of Si, N4 does
`become ion enhanced and, apparently, ion»-enhanced chemi-
`cal etching requires a much smaller concentration of neutral
`reactants in the case of Si3N4 than in the case of Si02. This is
`consistent with the previous observation that the etching of
`SiO2 is strongly ion enhanced only with an excess of F atoms
`present in the plasma.”
`In order to show that the above mechanism of direct reac-=
`
`tive ion etching is plausible, we need to demonstrate that
`chemical reactions between plasma ions and the substrate
`give reasonable etch yields. Thus we note first that energetic
`molecular ions will dissociate into highly reactive atoms
`upon impinging onto the substrate. '5 in addition, mass spec-
`trometry and optical emission spectroscopy studies revealed
`that the important. etch products with Si3N4 are SiF2, N2,
`and CN (as a radical or as a fragment of CNF) . ""17 (Clarke
`at al. ‘6 also observed SiF in the plasma but argued that this
`was a gas phase reaction product from SiF2 and not a true
`etch product). It is then clear that the most eflicient ion
`reaction from the point of View of the etch yield is
`
`2 Si3N4 + 6 CHF[ ==->6 SiF2 + 6 CN + N2
`
`+ 3 H2,
`
`Y= 7/3,
`
`where Si and C are removed in the form of Sill‘, and CN.
`Other possible reactions such as
`
`2 Si3N,/, + 7 CHF[ ——>SiF2 + 5 SiF + 7 CNF + %N2
`
`+ QHZ,
`
`Y: 2,
`
`Si3N,, + 4 Cl-IF; ='*SiFg + 2 Si?’ + 4- CNF
`
`+ 2H2,
`
`Y: 7/4,
`
`Si3N4 + 6 CHF,,'‘' —>3 SiF2 + 2 N2 + 3H2
`
`—:— 6 CF,
`
`Y: 7/6,
`
`lead to lower etch yields. Furthermore, reactions producing
`SiF4 directly also give lower yields, for example
`
`Si3N4 + 12. CHF; —+3 Sift}, + 6 N2 + 6 H2
`
`+12CF,
`
`Y2?/12,
`
`or else they leave carbon on the substrate, for example
`
`Si3N4 + 6 CHF;‘ —>3 SiF4 + 4 CN + 3 H2 + 2 C,
`
`Y = 7/6,
`
`in which case the etch reaction would soon be suppressed. Of
`course, it is not possible to decide at this point which of the
`first four reactions for the etching of Si3 N, are the most
`important ones, but it is evident that the expected yields are
`consistent with the ones measured at 10% 02 content. One
`should keep in mind, though, that here as well as in other
`etching situations involving ion bombardment of the sub-
`strate, there is always a small contribution to the overall
`etching from simple physical sputtering.
`For SiO2 the etch yields measured here are somewhat
`higher than those found in a previous study. '2 This is prob-
`ably due to the fact that in the present experiment the ion
`energies are higher ( Vac ;: 500 V) than in the previous work
`
`,.
`
`l
`
`LLJ
`
`0
`
`W
`
`20 3 40
`
`so 5 7
`
`PRESSURE
`
`(MT?
`
`FIG, 5. Etch yields vs pressure with CHF, -10% 0, . O: Si, N4; 0: sio,.
`
`etch yield for Si3N,, is higher by a factor 2 or more than at
`10% O2 and increases markedly with increasing pressure.
`On the other hand, at 20% 02 the etch yield for SiO2 is
`slightly lower than at 10% O2, and it increases only slightly
`with increasing pressure (Fig. 6). Thus, the etch yield of
`SiO2 is largely unaffected by the presence of unsaturates in a
`CHF3 —basecl plasma as long as there is etching, in agreement
`with previous resu1ts,”"3 whereas the etch yield of Si_,N., is
`strongly enhanced under conditions of reduced unsaturates
`(and probably somewhat increased F atom content).
`We interpret these results as indicating that for Si3N4 at 10
`% 0, and for SiO2 at l0% 02 and to a large extent also at
`20% O2 , the etching depends only on the ions in the plasma.
`More specifically, ions may react directly with the substrate
`which, for a fixed reaction probability, leads to a constant
`etch yield independent of neutral reactants present. The in-
`creased yield for Si3N4 with 20% 02 may be ascribed to the
`
`(ATOMS/ION)
`ETCHYIELD
`
`10
`
`2o
`
`30
`
`40
`
`so
`
`so
`
`70
`
`FIG. 6, Etch yields vs pressure with CHF, ~20% 0;, . O: Si, N,; 0: SiO2.
`
`Psessuas
`
`(MT)
`
`J. Vac. Sci. Technol. A, Vol. 9, No. 3, May/Jun 1991
`
`
`
`
`
`

`

`778
`
`Dulak, Howard, and Stelnbruchelz Etch mechanism in the HIE of Si3N,
`
`778
`
`(Vdc z 150 V),‘2 where the reaction probability of the ions
`may well have been less than 1 and the contribution by sput-
`tering to the overall etching smaller. But for Si02 also, the
`numerical values of the etch yields are consistent with plau-
`sible ion reaction mechanisms and with observed etch prod-
`ucts.” The pressure dependence of the etch yield of SiO2 at
`20% 02 indicates that at that composition there is a slight
`ion-enhanced component to the overall etching.
`
`V. CONCLUSIONS
`
`The above results suggest to us that in a CHF3/O2 plasma
`at low 02 content, the etch mechanism for both Si3N,, and
`SiO2 is mostly direct reactive ion etching, with ions them-
`selves being the main reactants. On the other hand, at higher
`02 content the etching is quite strongly ion-enhanced for
`Si3N4 but only slightly ion-enhanced for SiO2 . Thus it is with
`respect to etch yields or, equivalently, with respect to the
`degree of ion enhancement under conditions of low concen-
`tration of unsaturate species that the etching of Si, N, in a
`fluorocarbon plasma can be said to be intermediate between
`SiO2 and Si. The strongest evidence for this general picture is
`that etch yields are practically independent. of pressure at
`low 02 content for both materials but increase to different
`degrees with increasing pressure at higher 02 content. In
`addition, the numerical values of the etch yields at low 02
`
`content can be rationalized on the basis of plausible ion-
`substrate chemical reactions.
`
`“’ Present address: IBM East Fishkiil, Hopewell Junction, NY 12533-0999
`‘C. J. Mogab, in VLSI Technology, ed. by S. M. Sze (McGraw-Hill, New
`York, 1983), p. 337.
`2C. G. Sodini, S. S. Wontedt, and T. W. Ekstedt, IEEE Trans. Electron
`Devices ED-29, 1662 (1982).
`"D. L. Flarnrn, in Plasma Etching: An Introduction, edited by D. M. Manos
`and D. L. Flamm (Academic, San Diego, 1989), p. 165.
`‘F. H. M. Sanders, J. Dieleman, H. J. B. Peters, and J. A. M. Sanders, J.
`Electrochem. Soc. 129, 2559 (1982).
`SN. Hayasaka, H. Okano, and Y. Horiike, Solid. State Technol. 4, 127
`(1988).
`“H. W. Lehmann and R. Widmer, J. Vac. Sci. Technol. 15, 319 (1978).
`7S. Matsuo, J. Vac. Sci. Technol. 17, 587 (1980).
`“T. C. Mele, J. Nulman, and J. 1’. Krusius, J. Vac. Sci. Technol. B 2, 684
`( 1984).
`9H. J. Stocker, J. Vac. Sci. Technol. A 7, 1145 (1989).
`‘"1’. E. Riley and D. A. Hanson, J. Vac. Sci. Technol. B 7, 1352 (1989).
`"Ch. Steinbriichel, J. Electrochem. Soc. 130, 648 (1983).
`“Ch. Steinbriichel, J. Vac. Sci. Technol. A 8, 1663 (1990).
`“Ch. Steinbriichel, H. W. Lchmann, and K. Frick, J. Electrochem. Soc.
`132, 180 (1985).
`“Ch. Steinbriichel and B. J. Curtis, Mat. Res. Soc. Symp. Proc. 76, 179
`(1987).
`“Ch. Steinbriiehel, J. Vac. Sci. Technol. B 2, 38 (1984).
`”’P. E. Clarke, D. Field, A. J. Hydes, D. F. Klemperer, and M.J. Seakins, J.
`Vac. Sci. Technol. B 3, 1614 (1985).
`“D. Field, 1). F. Klemperer, and 1. T. Wade, J. Vac. Sci. Technol. B 6, S51
`(1988).
`
`J. Vac. Sci. Technol. A, Vol. 9, No. 3. May/Jun 1991
`
`
`
`
`
`