26
`
`IEEEJOURNALOF SOLID-STATECIRCUITS,VOL.SC-20,NO.1,FEBRUARY1985
`
`Thermal
`
`INitridation of Si and Si02 for
`VLSI
`
`MEHRDAD M. MOSLEHI, MEMBER, IEEE, AND KRISHNA
`
`C. SARASWAT,
`
`MEMBER, IEEE
`
`reyiew of onr work on
`an extensive
`paper presents
`Abstract —This
`thermal nitridation of Si and Si02. High-quafity ultrathin films of silicon
`nitride and nitrided-oxide
`(nitroxide) haye been thermally grown in am-
`monia atmosphere
`in a cold-wall RF-heated reactor and in a lamp-heated
`system.
`‘f’ke growth kinetics and their dependence on processing time and
`temperature have been stndled from very short to long nitridation times.
`The kinetics of thermaf nitridation of Si02
`in ammonia ambient have also
`been studied.
`In nitroxide, nitrogen-rich layers are formed at the surface
`and interface at a very early stage of the nitridation. Then the nitridation
`reaction mainly goes on in the bulk region with the surface and near
`interface
`nitrogen content
`remaining
`fairly constant. Onr results
`afso
`indicate the formation of an oxygen-rich layer at the interface underneath
`the nitrogen-rich
`layer whose thickness
`increases
`slowly with nitridation
`time. The nitride and nitroxide films were analyzed using Auger electron
`spectroscopy,
`grazing angle Rutherford backscattering, and etch rate mea-
`surements. MIS devices were fabricated using these films as gate insulators
`and were
`electrically
`characterized
`using Z-V, C-V,
`time-dependent
`breakdown,
`trapping,
`and dielectric breakdown techniques. Breakdown,
`conduction,
`and C – P’ measurements
`on metal-insulator
`semiconductor
`(MIS)
`structures
`fabricated with these films show that very thin thermal
`silicon nitride and nitroxide films can be used as gate dielectrics
`for future
`highly scaled-dowm VLSI devices. The electrical
`characterization results
`also indicate extremely
`low trapping in the nitride films. The reliability of
`ultrathin nitride was observed to be far superior to Si02
`and nitroxide due
`to its much less
`trapping. Studies
`show that the interface transition from
`nitride to silicon is almost abrupt and the morphology and roughness of the
`interface are comparable to the Si02 -Si
`interfaces.
`
`I.
`
`INTRODUCTION
`
`H IGH-QUALITY
`
`very thin ( <100 A) gate insulators
`(VTGI) will be required
`for
`future
`scaled VLSI de-
`vices. To illustrate
`the trend of the larger
`scale of integra-
`tion and the need for
`thinner
`gate insulators, DRAM can
`be used as an example. Megabit DRAM and high-density
`FLOTOX E2PROM devices will need gate or tunnel
`insu-
`lators
`in the
`subhundred-angstrom
`region. The 256-kbit
`DRAM has
`a gate
`insulator
`thickness
`of about
`200 A
`whereas
`the projected
`l-Mbit DRAM using conventional
`one transistor
`per bit will require
`an insulator
`thickness
`less than 100 A.
`controllable
`an easily
`using
`be grown
`VTGI
`should
`process with minimum number
`of process
`control parame-
`ters. They
`should
`be uniform and amorphous
`with low
`
`of
`
`received July 19, 1984; rewsed October 9, 1984. This work
`Manuscript
`was Jointly supported by DARPA under Contract MDA 903-80-C-0432
`and by the Joint Serwces Electrorucs Program of the Defense Department
`under Contract DAA29-81-K-O057.
`The authors are with the Integrated Circuits Laboratory, Electrical
`Engmeenng Department, Stanford University, Stanford, CA 94305.
`
`high break-
`Furthermore,
`and conductivity.
`density
`defect
`distribution,
`clean interface,
`field,
`sharp breakdown
`down
`are important
`requirements.
`In
`and low trapping
`efficiency
`over electrical
`characteristics
`order
`to have better
`control
`of devices, VTGI
`should be effective masks against
`impur-
`ity diffusion
`films have many obvi-
`grown silicon dioxide
`Thermally
`and technology
`includ-
`ous applications
`in silicon devices
`ing gate
`insulators
`of
`IGFET’s
`and the tunnel
`insulators
`for nonvolatile
`memories
`such as E2PROM.
`There
`are,
`however,
`several
`technological
`and reliability
`problems with
`silicon
`dioxide
`in the very thin regime. The fact
`that very
`thin layers of thermal Si02 are poor masks
`against
`impur-
`ity diffusion
`places
`additional
`constraints
`on processing
`steps
`following
`the growth of gate or tunnel oxides. More-
`over,
`the growth
`of high-quality
`very thin layers of Si02 is
`rather
`difficult
`due to defect density,
`integrity,
`and yield
`problems.
`It
`is also known
`that high-energy
`radiation
`can
`generate
`a high density
`of
`interface
`states
`in the oxide,
`resulting
`in degradation
`of device performance.
`Very thin
`Si02 is not a good mask against
`impurity
`diffusion
`[1], [2]
`and has high defect density
`[3]. These poor properties
`of
`very thin Si02
`film and also its
`tendency
`to react with
`electrode material will
`limit
`its effective widespread
`appli-
`cation
`for gate,
`tunnel,
`or DRAM capacitor
`dielectrics
`of
`scaled VLSI devices. As a result of the technological
`and
`reliability
`problems with silicon dioxide
`in the very thin
`regime
`( <100 A),
`there is a demand
`for new higher quality
`ultrathin
`insulators
`to replace
`silicon
`dioxide
`in highly
`scaled
`down VLSI
`IGFET”S and memory
`devices
`such as
`DRAM and E 2PROM. Thermal
`nitridation
`of Si and thin
`Si02
`has
`been
`investigated
`in the past
`few years
`and
`appears
`to be an alternative
`to the oxidation
`process
`to
`grow good-quality
`films in the very thin regime. Thermally
`grown
`very thin films of silicon nitride
`have a number
`of
`advantages
`over silicon dioxide. Their growth is self-limited
`and,
`therefore,
`easily controllable.
`They have less number
`of process
`control
`parameters
`and high oxidation
`resis-
`tance. Moreover,
`they are effective barriers
`against
`impur-
`ity diffusion
`and their characteristics
`do not degrade during
`the VLSI processes
`[4], [5]. Devices
`fabricated with these
`films
`show a large
`transconductance
`and reduced
`unde-
`sired second-order
`effects
`[6]–[8].
`Thermal
`nitride
`films can be grown by high-temperature
`nitridation
`of silicon in pure ammonia
`[9], ammonia
`plasma
`[10], [11], nitrogen-hydrogen
`plasma
`[12], or low-tempera-
`
`0018-9200/85
`
`/0200-0026
`
`$01.00
`
`01985 IEEE
`
`SAMSUNG-1012.001
`
`

`
`MOSLEHIAND SARA,SWAT:THERMALNITRIDATIONFOR VLSI
`
`of
`nitridation
`[13]. Thermal
`nitridation
`ture laser-enhanced
`can
`furnaces
`quartz-tube
`hot-wall
`Si
`in resistance-heated
`of the grown films [5].
`lead to high oxygen contamination
`Most of the recent work on thermal nitridation
`has been on
`technological
`issues
`such as composition,
`growth
`kinetics,
`oxidation
`resistance,
`and interactions
`with impurity
`diffu-
`sion
`[9],
`[14]–[1 8].
`In a few cases,
`devices made with
`nitroxide
`and nitride
`have been electrically
`characterized
`[6],
`[8], [19] -[30]. A more detailed
`electrical
`characteriza-
`tion of
`thermal
`nitride
`films
`including
`trapping,
`however,
`had not been reported
`in the past.
`In addition
`to that, most
`of the work in the past has been on nitroxide. Nitridation
`of oxide
`results
`in the reaction
`of nitridation
`species with
`oxide
`at
`the surface
`and interface
`regions
`and to a lesser
`extent
`in the bulk region as will be described
`later.
`It has
`been
`reported
`that
`the generation
`of
`radiation-induced
`interface
`states
`in MOS devices
`is eliminated
`after proper
`high-temperature
`nitridation
`of oxide
`[31], [32]. This may
`be attributed
`to structural
`changes
`in the interface
`region
`of oxide
`after nitridation
`in ammonia. We have observed
`[19], as Ito et al. [33] have also reported,
`that
`the break-
`down
`characteristics
`of MOS structures
`can be improved
`by
`annealing
`the Si02
`films
`in ammonia
`gas
`at high
`temperatures.
`The
`effects
`of nitridation
`of Si02
`in am-
`monia
`on electron
`trapping
`has been investigated
`[34], [35]
`and it has been reported
`that ammonia
`nitridation
`results
`in a significant
`increase
`in electron
`trapping.
`The data
`presented
`so far, however, have not been sufficient
`to draw
`firm conclusions
`regarding
`trapping
`in nitroxide
`insulators.
`Moreover,
`effects
`of annealing
`of nitroxide
`insulators
`in
`nitrogen,
`oxygen, or argon ambients
`on trapping
`character-
`istics should be studied in more detail and correlated
`to the
`compositional
`parameters
`of nitroxide.
`of SiOz
`nitridation
`It has been
`observed
`that
`thermal
`or ni-
`films
`results
`in the formation
`of a nitrided-oxide
`troxide
`layer with
`composition
`profile
`that
`varies with
`depth.
`It has
`also been reported
`that both thickness
`and
`refractive
`index
`of
`thin
`oxide
`films
`are
`increased
`after
`high-temperature
`nitridation
`in ammonia
`[16],
`[36]. The
`changes
`in composition
`of
`the
`initial
`oxide
`film after
`nitridation
`in ammonia
`have been
`studied
`using
`several
`techniques
`sulch as Auger
`electron
`spectroscopy
`(AES),
`etch rate, Rutherford
`backscattering
`spectroscopy,
`infrared
`transmittance
`spectroscopy,
`and oxidation
`resistance mea-
`surements
`[16],
`[17]. Moreover,
`very thin ( <100 A)
`[9],
`nitroxide
`and reoxidized-nitroxide
`insulators were used as
`tunnel
`insulators
`of high-endurance
`E2PROM (nonvolatile
`memory)
`devices
`[20] –[27].
`was
`nitridation
`thermal
`here,
`In
`the work
`reported
`reactor
`accomplished
`in a closed-tube
`cold-wall RF-heated
`to minimize
`the oxygen
`contamination.
`Very short
`time
`( <5
`fin)
`nitridations
`were performed
`in a commercial
`lamp-heated
`rapid thermal
`annealing
`system, again in pure
`ammonia.
`The grown films have been characterized
`using
`grazing
`angle Rutherford
`backscattering
`(RBS), Auger
`electron
`spectroscopy
`(AES),
`cross
`sectional
`transmission
`electron microscopy
`(TEM),
`ellipsometry,
`and
`etch rate
`measurements.
`In addition,
`the electrical
`properties
`of the
`
`27
`
`v..TER-COOLED
`IN3uCTION
`COIL
`
`/
`
`‘ SUSCEPTOR
`
`Si wAFER
`
`\
`
`00
`
`8
`
`@
`
`8
`
`+++
`H2/N2
`
`Ar
`
`HC[
`
`v
`
`I
`
`E X HTAOUS T
`
`4
`NH3
`(HIGH PURITY)
`
`Fig. 1, The RF-heated thermal nit
`
`lor.
`
`by doing C-}
`films were determined
`dent
`breakdown
`and trapping measu.
`capacitors.
`
`time depen-
`.[s on Al-gate
`
`II.
`
`EXPERIMENTAL PROCEDIJRES
`
`.-
`
`~
`
`the long time ( >5 rein) expe, ments were conducted
`All
`in a horizontal
`RF-heated
`reactor operating
`at atmospheric
`pressure,
`as shown
`in Fig. 1. Thermal
`ritridation
`is per-
`formed
`in pure or diluted
`ammonia
`in argon. The oxygen
`contamination
`from outside
`and from
`quartz
`tube is
`insignificant
`because
`the tube walls
`,
`f ?t cold. The
`short
`time
`( <5
`rein) nitridations
`ormed
`in a
`lamp-heated
`rapid
`thermal
`annealing
`the range
`of 900–1200°C.
`;111) and (100)
`studies, both n- and p-ty~
`For kinetics
`from 0.002 to 3
`silicon wafers with resistivities
`ranging
`0. cm, and oxidized
`silicon wafers were used. The samples
`were first cleaned
`chemically,
`loaded
`nn a silicon carbide-
`coated
`graphite
`susceptor,
`and then inserted
`in the quartz
`tube. After
`an initial
`purging
`of
`the
`system with pure
`nitrogen,
`the
`flow of pure
`ammonia
`( <1 ppm contami-
`nants) was established. After
`flushing NJ out of
`the tube
`with ammonia,
`temperature
`was ramped
`up to the growth
`.
`temperature
`and ‘stabilized. Nitridation<
`e performed
`at
`temperatures
`ranging
`from 950 to 123f ’”. i.
`:>eriods of 30
`rnin to 4 h. For
`compositional
`stud:-:,
`,.~ ,, ~troxide,
`the
`oxide
`films were grown in dry 02 at 9’ J!’‘(.. :;> a thickness
`of 95 A and at 1000”C to thicknesses
`of .::’1, 405, and 1021
`A, on 5 –10 L?.cm (100)
`silicon wafers. The nitridations
`were performed
`for times from 15 s to 4 h at 900 to 1200° C
`in pure
`ammonia
`at atmospheric
`pres. u-e. To investigate
`the effect of heavy doping
`an experiment was performed
`where heavily
`phosphorus-doped
`silicon wafers with (100)
`and
`(111) orientations
`were nitrided
`at 1190° C for 2 h.
`Phosphorus
`doping
`had been done at 105O”C from POC1 ~
`resulting
`in a sheet
`resistance
`of 1.6 fl/sq.
`films were mea-
`The
`thicknesses
`of
`the growl.
`nitride
`ellipsometer with
`sured
`using a microcomputer-controlled
`dex of
`refraction
`a 6328-A laser
`source
`and substrate
`fixed
`at 3.85 –O.02i.
`In the measurements
`the
`refractive
`
`Ln
`
`SAMSUNG-1012.002
`
`

`
`28
`
`IEEEJOURNALOF SOLID-STATECIRCUITS,VOL.SC-20,NO.1,FEBRUARY1985
`
`NIT RI DATION KINETICS
`
`,001250
`
`1200
`
`1150
`
`T (“C)
`1100
`
`1050
`
`1000
`
`950
`
`20L78
`
`8.6
`llk T (eV-’)
`
`8.2
`
`9.0
`
`9.4
`
`(a)
`
`NITRIDATION
`
`KINETICS
`
`.
`
`A
`
`+
`x
`0
`
`4
`
`5’~ ~“
`
`/’”
`
`.
`
`1150
`
`1100
`
`1050
`1000
`T=950”c
`
`./”
`~,A/”
`
`—’
`+—+
`, —’
`x—
`q —.—”
`
`40
`
`25
`
`f-
`
`U3
`
`mwz~x
`
`IOL 2
`
`3
`TIME (hrs)
`(b)
`
`}
`om~,o
`
`8
`
`9
`{1/e V]
`
`1/kT
`
`(c)
`
`I
`
`!0
`
`Fig. 2. Silicon nitridation kinetics. (a) Arrhenius plot of nitride thick-
`ness versus nitridation temperature for 2-h nitridation and (100) and
`(111) substrate orientations. (b) Thickness of nitride grown on (100) Si
`versus nitridation time and temperature. (c) Temperature dependence
`of u and b in the power law expression,
`
`energy than
`activation
`silicon shows a higher
`tion of (111)
`as given in
`energies
`silicon. The
`activation
`that
`of
`(100)
`reported
`by
`than that
`Table
`I, are fairly small but
`larger
`et al. [5] who observed
`Murarka
`energy
`of
`an activation
`0.23 eV for 40 min nitridation
`in 5 percent NH3 and 16 h
`nitridation
`in 100 percent NH3. Hayafuji
`and Kajiwara
`[16] obtained
`an activation
`energy of 0.35 eV for 5 hour
`nitridation
`in ammonia
`partial
`pressures
`of 10-3
`kg/cm2
`and 1 kg/cm2.
`films
`nitride
`of thermal
`the thickness
`Fig. 2(b)
`illustrates
`grown on (100)
`silicon substrate
`versus nitndation
`time up
`
`SAMSUNG-1012.003
`
`of
`
`fixed at 2.0. This
`films was
`nitride
`silicon
`the
`index
`it
`is not possible
`to measure
`because
`is adopted
`method
`refractive
`index
`of very
`thin
`films
`thickness
`and
`both
`by ellipsometry.
`The variation
`of the
`( <100 A) accurately
`in the
`range
`between
`1.6 and
`2.0
`film refractive
`index
`exhibits
`little
`effect on the nitride
`thickness measurement
`results. The measured
`thickness
`values are almost
`indepen-
`dent of the film refractive
`index Nf for its value in between
`1.8 and
`2.0. All kinetics
`data
`presented
`in this paper,
`however,
`are for Nf = 2.0 which is the established
`value for
`thick CVD nitride.
`The nitride
`and nitroxide
`films were
`investigated
`using AES, RBS, TEM,
`etch rate, and electri-
`cal characterization
`techniques.
`In the etch rate measure-
`ments
`the thickness
`of remaining
`film was measured
`after
`step by step etching in 50:1 DI H ~0: HF using ellipsome-
`try.
`gate
`and nitride
`MIS devices with thin oxide, nitroxide,
`insulators
`were
`fabricated with aluminum-gate
`electrodes
`on n- and p-type
`Si substrates.
`The Si02 used in devices
`with oxide and nitroxide was grown to a thickness
`of about
`100 A in a furnace
`in oxygen at 900”C. The wafers were
`annealed
`in forming
`gas
`at 450° C for 20 min
`before
`patterning
`and after deposition
`of metal. Finally aluminum
`was patterned
`into dots. The gate areas were 2.01 X 10 – 2,
`5.03 x10-3,
`1.26 x10-3,
`3.14x 10-4, and 7.85 x10-5
`cm2.
`The gate
`electrode was Al-(l
`percent)
`Si deposited
`in a
`flash evaporator.
`
`III.
`
`COMPOSITION
`
`AND GROWTH KINETICS STUDIES
`
`A. Growth Kinetics
`
`silicon nitride
`thermal
`of
`thickness
`the
`shows
`Fig. 2(a)
`for nitrida-
`temperature
`nitridation
`of
`films
`as a function
`tion period
`of 2 h on (100) and (111) silicon wafers. Under
`identical
`processing
`conditions
`the films grown
`on (100)
`silicon
`are
`thinner
`than
`those on (111)
`silicon,
`and the
`orientation
`effect
`is more pronounced
`at higher
`tempera-
`tures. Below 1000”C, however,
`the effect of substrate
`orien-
`et al.
`tation
`on nitridation
`kinetics
`is negligible. Murarka
`[5] did not observe
`any orientation
`dependence
`of nitrida-
`tion kinetics. We believe
`that
`their observation
`could be
`due
`to large
`scatter
`in their
`thickness measurements
`and
`also very high oxygen contamination
`of their nitride
`films.
`It
`is evident
`that
`in contrast
`to the silicon thermal
`oxida-
`tion
`process,
`the nitridation
`kinetics
`has
`a rather weak
`dependence
`on the substrate
`orientation.
`This could be due
`to the fact
`that
`the latter process
`is diffusion-limited
`during
`most of the growth time except
`for the fast
`initial growth.
`For
`shorter
`times
`(like 30 rein),
`nitridation
`of silicon
`shows
`an approximately
`Arrhenius
`behavior with tempera-
`ture whereas
`for
`longer
`times
`(t >1 h) the behavior
`is not
`quite Arrhenius
`because more than one activation
`energy is
`observed
`in the
`temperature
`range
`of growth. Applying
`exponential
`least-squares
`curve
`fits of
`the
`form X~ =
`X~ exp ( – E. /k T’) to the experimental
`kinetics
`data,
`the
`energy E. and the coefficient X~
`values
`of the activation
`as T -+ m) were evaluated
`(The thickness
`as a functions
`of
`time with the substrate
`orientation
`as a parameter.
`‘Nitrida-
`
`

`
`MOSLEH1AND SARASWAT:THERMALNITRIDATIONFORVLSI
`
`29
`
`NITRIDATION TI!lE
`
`30 111!1.
`
`1 HOUR
`
`2 HOURS
`
`4 HOURS
`
`I
`I
`
`t
`
`667.4
`
`1052.9
`
`1183.9
`
`32!35.1
`
`TABLE I
`
`1
`
`(100)
`
`(111)
`
`0.342
`
`869.5
`
`0.390
`
`1525.2
`
`0.39s
`
`1705.0
`
`0.513
`
`7455.2
`
`I
`
`0.359
`
`0.%2Q
`
`0.43L
`
`0.602
`
`The Constants
`for
`.ar,
`ous
`(111) silicon.
`
`Xm and Ea (XN =Xm EXP(-Ea/LT))
`nitr]dation
`t>nws
`of
`(100)
`and
`
`thicknesses
`The
`as a parameter.
`to 4 h with temperature
`were measured with,
`the refractive
`index fixed at 2.0. The
`experimental
`data
`indicate
`a very fast
`initial growth
`in a
`short
`time ( <5 tin)
`followed by a self limiting growth. To
`express
`the longer
`time
`( >5
`rein)
`experimental
`kinetics
`results
`in a functional
`form, we have found several
`simple
`empirical
`relationships
`useful.
`[9]. One relationship
`which
`is X~ = a X t b as shown by
`fits the experimental
`data best
`in Fig. 2(b), where a and b are tempera-
`the solid curves
`ture dependent
`constants,
`and X~ and t are the silicon
`nitride
`thickness
`and nitridation
`time,
`respectively.
`The
`a and b are found
`constants
`by fitting
`the experimental
`data using the method
`of least squares.
`In the temperature
`range of interest
`they show nearly Arrhenius
`behavior with
`nitridation
`temperature
`with single activation
`energies. We
`can obtain
`the activation
`energies
`of 0.381 and 0.238 eV
`of a and b, respectively.
`for
`temperature
`dependence
`Fig.
`of a and b for
`2(c) illustrates
`the temperature
`dependence
`the power
`law expression. Their
`functional
`dependence
`on
`nitridation
`temperature
`can
`be
`expressed
`as
`a = 920.2
`~(– 0.3gl/~~) ~ and b = ().()183 e f02381~~J. In the power
`law
`t is normalized
`a and X~ are in angstroms,
`expression,
`to
`b is unitless. Reisman
`et al. [37] have
`1 tin,
`and
`also
`observed
`power
`law dependence
`of nitride
`thickness
`grown
`in Ar–NH3
`plasma
`on nitridation
`time (t 02 dependence).
`The power
`law relationship
`is only an empirical
`relation-
`ship which fits the experimental
`data
`fairly well. We have
`not
`investigated
`the physics
`of a growth
`that
`follows
`the
`X~ = a x t b kinetics. As will be described,
`at present we
`have
`a good physical
`understanding
`of the mechanism of
`nitridation
`of oxide;
`however,
`this knowledge
`is not yet
`available
`for nitridation
`of silicon. The major
`remaining
`question
`in the case of silicon nitridation
`is regarding
`the
`species
`(Si
`ions
`and/or
`nitrogen
`species)
`that
`diffuse
`through
`the film and react with silicon or nitrogen
`at
`the
`opposite
`interface
`(nitride/ambient
`or nitride/silicon).
`This
`question
`has not yet been answered.
`silicon
`for
`data
`Fig.
`3 illustrates
`the growth
`kinetics
`wafers nitrided
`at 1100° C for
`times
`from 30 to 120 s in a
`lamp-heated
`rapid
`thermal
`annealing
`system in pure
`am-
`monia. The data
`indicate
`that
`for very short nitridation
`times,
`the growth is linear with nitridation
`time.
`by
`studied
`The kinetics
`of nitridation
`have
`also been
`[16]
`some
`other
`investigators.
`-Hayafuji
`and Kajiwara
`at
`studied
`nitridation
`of
`bare
`Si
`and
`100 A Si02
`900–1150”C
`under
`ammonia
`partial pressures
`of 10”3 to 5
`
`I
`
`\
`
`40.0
`
`i
`
`0
`
`1100 ‘c, (111) 51
`
`50
`
`100
`
`150
`
`2<8
`
`Nitridati?n
`
`time
`
`(seconds)
`
`Fig. 3. The short
`
`time Si nitridation kinetics data for 11OO”Cnitrida-
`tion.
`
`and ammonia-nitrogen
`ammonia
`in high-purity
`kg/cm2
`in
`all the nitridation
`experiments
`mixture. They performed
`a resistively
`heated
`high-pressure
`system. For 5-h nitrida-
`tion at 900°, 1000°,
`and 1100° C,
`they studied
`the depen-
`dence
`of
`self-limited
`nitride
`thickness
`‘on NH3
`pressure
`from 10-3
`to 5 kg/cm2.
`For
`less than atmospheric
`pres-
`sures nitrogen was used to dilute
`the NH q. The pressure
`dependence
`of the nitridation
`growth kinetics was found to
`be rather
`small. The film thickness
`increased
`only by 20
`percent when the NH3 partial pressure was increased
`from
`2. The
`oxidation
`resistance
`data
`of
`10-3
`to
`5 kg/cm
`et al. [5], however,
`Murarka
`show more
`ammonia
`partial
`pressure
`dependence
`than those of Hayafuji
`and Kajiwara
`[16]. We believe
`that
`in the presence
`of oxidant
`impurities
`in the ambient
`the pressure
`dependence
`is more significant
`and lower partial
`pressure
`of NH3 results
`in higher oxygen
`contamination
`of
`the films. The nitride
`films of Hayafuji
`and Kajiwara, were much more oxygen free compared
`to
`et al. Chen et al. [38] also studied
`those of Murarka
`the
`high-pressure
`nitridation
`process
`and found that
`for high-
`-pressure nitridation
`of silicon in the range of 1-10 atm,
`the
`nitride
`thickness
`for a given nitridation
`time and tempera-
`ture
`increases
`slightly with increasing
`the ammonia
`pres-
`sure.
`at
`nitrogen
`in pure
`of nitride
`kinetics
`growth
`The
`indi-
`[39]. The results
`has been investigated
`1200–1300”C
`smooth and amorphous
`cated that
`for a given temperature,
`nitride
`films could be grown for short
`reaction
`times but
`they start
`to become
`rough and crystalline
`for long reaction
`The maximum
`reaction
`time
`for
`growth
`of
`times.
`amorphous
`uniform nitride
`is reduced
`by increasing
`the
`nitridation
`temperature.
`Contrary
`to ammonia
`nitridation,
`no self-limiting
`of growth could be observed
`for high-tem-
`perature
`nitridation
`in nitrogen.
`thermal
`for
`Recently
`two models
`have been proposed
`of silicon in ammonia. Wu et al. [40] developed
`nitridation
`
`SAMSUNG-1012.004
`
`

`
`30
`
`IEEEJOURNALOF sOLID-STATECIRCUITS,VOL.SC-20,NO.1,FEBRUARY1985
`
`95 ~ \nlt,al S102
`
`J
`
`,/
`
`,/
`
`/
`
`//
`
`/
`
`55 -
`
`* 4 hrs
`+ 2 hrs
`.lhr
`. 30 m,n
`
`40 -
`
`L
`
`x<
`
`.:
`
`‘5 -&&//>oN
`
`‘o~1000
`
`1100
`TEMPERATURE
`
`1200
`
`(“C)
`
`Fig. 4. The increase in the thickness of 95-A Si02 due to nitridation
`versus nitridation time and temperature. The actual thickness may be
`slightly lower because the index of refraction was fixed at 1.46.
`
`enhanced
`exhibited
`4 h), however,
`(e.g., 1200°C,
`periods
`experience
`1.7. Our
`indices
`up to more
`than
`refractive
`in the thickness measurement
`of very thin films
`shows
`that
`by ellipsometry,
`fixing the refractive
`index at a value lower
`than
`the
`true
`index
`value
`results
`in a measured
`value
`slightly
`higher
`than
`the real
`thickness.
`Therefore,
`in the
`foregoing
`high-temperature
`experiments
`for long processing
`times
`the measured
`thickness might
`be somewhat
`higher
`than the actual value.
`on whether
`in the literature
`There
`is some disagreement
`the thickness
`of oxide
`is increased
`or decreased
`after
`ex-
`et al. suggest
`posure
`to ammonia. Aucoin
`that
`the nitrida-
`tion of oxide
`causes
`a s~inkage
`in the film thickness
`[14].
`Hayafuji
`and Kajiwara
`[16]
`reported
`that
`nitridation
`of
`oxidized
`silicon resulted in a small
`increase
`in the thickness
`of
`the films. According
`to their data, nitridation
`of 1OO-A
`oxide
`at 1100”C for 5 h increased
`the
`thickness
`of
`the
`original
`oxide
`by about
`20 A. The
`change
`of
`thickness,
`however,
`can
`depend
`on nitridation
`conditions
`and
`the
`original
`oxide
`thickness. As we will explain,
`the growth of
`interracial
`oxide
`layer during nitridation
`of oxide tends
`to
`increase
`the thickness
`of
`the film and the exchange
`bulk
`reaction may result
`in a slight
`shrinkage
`of
`the bulk.
`In
`general
`the change
`in the film thickness
`due to nitridation
`of oxide
`is fairly
`small. Most
`of
`the thickness measure-
`ments
`are the data obtained with ellipsometry which can
`have
`some
`error due to the nonhomogeneous
`structure
`of
`nitroxide.
`Based
`on the results
`of our work on kinetics
`studies
`of nitridation
`of oxide, we have
`come up with a
`simple
`theoretical model which relates
`the final
`thickness
`of the nitroxide
`to the thickness of the initial oxide, and the
`bulk composition
`of nitroxide which is a function
`of
`the
`nitridation
`conditions
`[43]. According
`to this simple model,
`depending
`on the thickness
`of original Si02 and the nitri-
`dation
`conditions,
`the nitridation
`of oxide may result
`in
`increase
`or decrease
`in the film thickness. The ellipsometer
`data
`agree fairly well with this model.
`
`B. Auger Electron Spectroscopy and Grazing Angle RBS
`
`using
`samples were analyzed
`and nitroxide
`The nitride
`AES and some of
`the Auger depth
`profiles
`for
`them are
`
`of the continuity
`a model which is based on the solution
`equation.
`Except
`for
`the definition
`of an unrealistic
`diffu-
`sion length,
`their assumptions
`are similar
`to the Deal–Grove
`oxidation model
`[41]. Their
`assumptions
`that
`the grown
`film has a uniform composition
`and the reaction
`only takes
`place
`at
`the nitride}silicon
`interface
`are inconsistent
`with
`their
`result
`showing
`a variable
`nitridant
`flux through
`the
`film. Furthermore,
`the introduction
`of
`the characteristic
`diffusion
`length
`implies
`the annihilation
`of
`the nitridant
`species
`inside
`the grown
`film which
`they assume
`to be
`stoichiometric
`and uniform. We believe this model partially
`neglects
`the physics
`of the process. Our experiments
`show
`the
`existence
`of nonuniform
`(in depth)
`and nonstoichio-
`a
`metric
`films which
`should
`be considered
`in developing
`by
`physical
`kinetics model. A second model
`proposed
`Hayafuji
`and Kajiwara
`[16] assumes
`that nitridation
`occurs
`due
`to the
`reaction
`of ammonia
`with
`silicon which
`is
`diffused
`out
`to the surface of the film by a strong internally
`induced
`electric
`field. They suggest
`this model
`in order
`to
`explain
`the pressure
`independence
`of
`silicon
`nitridation
`kinetics. At present,
`the exact mechanism of thermal
`nitri-
`dation
`of silicon is unclear
`and there is no strong experi-
`mental
`evidence
`to show if nitridant
`species or silicon ions
`are the dominant
`diffusers
`through
`the film during nitrida-
`tion. Tan and Gosele
`[42] found
`the model
`proposed
`by
`Hayafuji
`and Kajiwara
`[16] acceptable.
`They
`suggested
`that
`the Si cations
`leaving the nitride–silicon
`interface
`are
`formed
`by interstitial
`diffused
`from the silicon interior
`to
`the interface.
`If true,
`this would imply that
`thermal nitrida-
`tion does not
`consume
`any silicon atoms
`taken from the
`nitride–silicon
`interface
`region and the interface
`is not
`a
`moving
`boundary. More
`experimental
`evidence
`is needed
`to conclude
`that
`the
`silicon
`consumed
`during
`thermal
`nitridation
`silicon
`is either
`provided
`by breaking
`the
`of
`Si–Si bonds
`the nitride-silicon
`interface
`or by the flow
`at
`of
`silicon
`interstitial
`from the
`substrate
`into
`the
`film.
`Whether
`the
`actual
`nitridation
`mechanism might
`be Si
`cation
`diffusion
`or nitridant
`diffusion
`(or a combination),
`we believe
`that
`the depletion
`of interstitial
`during nitrida-
`tion of silicon can not be used as a very strong evidence
`to
`come up with the actual nitridation mechanism.
`of
`effects
`In the
`experiments
`conducted
`to study
`the
`the
`that
`heavy doping
`on nitridation
`kinetics, we observed
`silicon
`nitride
`films grown on heavily phosphorous-doped
`had the same
`thickness
`as those grown
`on lightly doped
`silicon wafers.
`Therefore,
`no significant
`effect
`of heavy
`doping
`on nitridation
`growth kinetics was detected.
`in an
`resulted
`of’ = 100 A Si02
`Thermal
`nitridation
`increase
`in the film thickness. Oxide films 95 A thick were
`thermally
`nitrided
`at various
`temperatures
`for periods
`of
`30 rein,
`1 h, 2 h, and 4 h. The
`thickness
`increase
`as a
`function
`of processing
`temperature
`and time is shown in
`Fig. 4. The thickness
`change does not
`show an Arrhenius
`behavior with temperature
`because
`the film thickness
`in-
`creases more rapidly
`at higher
`temperatures.
`The thickness
`of all nitroxide
`films were measured
`using
`ellipsometer
`with
`the
`film optical
`refractive
`index
`fixed at 1.46. The
`oxide
`samples
`nitrided
`at very high temperatures
`for long
`
`SAMSUNG-1012.005
`
`

`
`MOSLEHIAND SARASWAT:THSRMALNITRIDATIONFOR VLSI
`
`31
`
`I
`
`,N
`
`1105”
`
`C, 4hrs,
`
`.“
`
`.
`
`.
`
`.“
`
`.
`
`.
`
`/1
`
`‘lLvv
`
`1098” C, 4hrs.
`
`~
`
`.p~,
`i--”-”-”
`/--”—”
`— x 100”/.
`
`,N,
`
`[N] + [0]
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`““” “k-
`
`‘A””
`
`”””””””
`,,~
`
`Normalized
`
`depth
`
`or sputtering
`
`time
`
`(a)
`
`1
`
`(
`
`Normalized
`
`depth
`
`or sputtering
`
`time
`
`(b)
`
`Fig. 5. The AES depth profiles for nitrides grown at (a) 1105”C for 4 h.
`(b) 11119C for 1 h.
`
`axis is the Auger
`in Figs. 5 and 6. The vertical
`illustrated
`peak-to-peak
`signal proportional
`to the actual
`derivative
`concentration.
`The horizontal
`axis is a measure
`of sputter-
`ing time or depth
`inside
`the film.
`In these figures,
`all
`the
`nitrogen
`intensity
`profiles
`should be multiplied
`by 1.9 for
`direct
`comparison
`with oxygen intensity
`profiles. The re-
`sults
`for
`thermal
`nitrides
`indicate
`that
`the nitrogen
`signal
`decreases
`slowly
`from the
`surface
`towards
`the interface,
`which may be partially
`due to the Auger broadening
`ef-
`fects. As shown in Fig. 5 for thermal
`nitrides
`grown under
`various
`nitridation
`conditions,
`the oxygen
`concentration
`builds
`up towards
`the surface with a maximum concentra-
`at
`tion
`the
`surface. As
`it was
`shown
`earlier,
`the initial
`kinetics
`of direct nitridation
`of silicon proceeds
`easily until
`it is limited by diffusion which significantly
`slows down the
`growth.
`The nitridation
`for
`longer
`times
`results
`in little
`increase
`in thickness
`but
`improves
`the stoichiometry
`of the
`nitride.
`The
`initial
`native
`oxide
`on the
`surface
`of
`the
`cleaned
`Si wafers
`and the residual
`oxidant
`impurities
`in
`the nitridation
`ambient
`cause the oxygen contamination
`of
`the grown
`films. Moreover,
`partial
`adsorption
`of oxygen
`and water
`at
`the
`surface
`of nitride
`after
`unloading
`the
`samples
`and during
`the storage
`time is probably
`the main
`source of oxygen concentration
`peak observed at the surface
`of thermal
`nitrides. The nitride
`film grown at 1105 ‘C for 4
`h shows
`an average
`relative
`atomic nitrogen
`concentration
`
`.
`
`.
`
`1
`
`‘
`
`Normalized
`
`depth
`
`or Sputtering
`
`time
`
`Fig. 6. The AES depth profile for 95-A Si02 nitrided at 1098”C for 4 h.
`
`from AES, depth
`as calculated
`85 percent
`than
`of more
`profile
`in Fig. 5(a). The relative
`atomic nitrogen
`concentra-
`tion ([N]/([N]
`+ [0]))
`is defined as the ratio of the nitrogen
`concentration
`to the total nitrogen
`and oxygen concentra-
`tion
`and
`is 1 for
`an oxygen-free
`film. This
`value
`(85
`percent)
`is about 25 percent higher
`than those for the films
`et
`grown in hot-wall
`resistance
`heated furnace by Murarka
`al. [5] and those grown in a closed tube hot-wall
`furnace by
`Hayafuji
`and Kajiwara
`[16]. To achieve reproducible
`growth
`results
`and good process
`control
`for high-quality
`nitride
`films,
`the
`amount
`of oxygen
`contamination
`should
`be
`minimized
`as was achieved
`in this work. Under
`identical
`growth
`conditions,
`due to oxygen contamination,
`the films
`grown in hot-wall
`furnaces
`are thicker
`than those grown in
`more oxygen-free
`systems
`such as the reactor
`used in this
`work. The AES depth profiles
`in Fig. 5 for thermal nitrides
`indicate
`that
`the free silicon concentration
`increases
`from
`the
`surface
`to the interface which is partially
`due to the
`Auger broadening
`effects. Moreover,
`the Auger data
`indi-
`cate
`that
`the films grown at higher
`temperatures
`and/or
`longer
`times have less oxygen contamination.
`For a given
`nitridation
`temperature
`such as = 1100~ C,
`the increase
`in
`nitridation
`time
`(e.g.,
`1 to 4 h) does
`not
`result
`in a
`significant
`increase
`in the thickness
`of the film. The com-
`parison
`of the Auger depth profiles
`(see Fig. 5(a) and (b)),
`however,
`indicates
`an increase
`in the nitrogen
`concentra-
`tion in the bulk as a result of increase
`in nitridation
`time.
`According
`to Fig. 6 for 95-A Si02 nitrided
`at 11OO”C for
`4 h,
`the fractional
`nitrogen
`concentration
`at
`the surface
`is
`more
`than
`50 percent.
`It
`also
`shows
`a fairly
`uniform
`nitrogen
`profile which vanishes
`close to the interface. The
`Auger
`depth
`profile
`for
`the nitroxide
`in Fig. 6 shows
`a
`minimum of relative nitrogen
`concentration
`profile close to
`the
`interface.
`The minimum arises
`from a peak
`in the
`oxygen
`concentration
`profile
`located
`close to the interface
`which is most possibly’ clue to the oxygen-rich
`layer grown
`at
`the interface
`by the liberated
`oxygen atoms provided
`by
`the bulk exchange
`reaction. This will be elaborated
`later
`in
`this paper. The relative
`surface nitrogen
`concentration
`for
`the oxides nitrided
`at 1100° and 1230”C were nearly equal.
`As a result,
`in this temperature
`range
`the surface nitrogen
`concentration
`is almost
`temperature
`independent.
`for 405-A
`Fig.
`7 illustrates
`the Auger
`depth
`profiles
`oxide
`samples
`nitrided
`at 1100”C for 30 s to 2 h of
`
`SAMSUNG-1012.006
`
`

`
`32
`
`5.00,
`
`IEEEJOURNALOF SOLiD-STATECIRCUITS,VOL.SC-20,NO.1,FEBRUARY1985
`
`16
`
`[.
`
`.—ioo
`
`A SIO*
`
`405; Si02
`.
`
`IIoo~2030
`
`90
`60
`Time (min. )
`
`Nitridation
`
`the surface of
`Fig. 8. The nitrogen to oxygen concentration ratio at
`nitroxide (1100”C nitridation of 405-A initial Si02 ) versus nitndation
`time.
`
`1.00
`
`.09
`
`5,00,
`
`0
`
`Oepth normal
`
`ized
`
`to
`
`the
`
`Si
`
`interface
`
`*
`
`Oata
`
`based
`
`on AES
`
`O: 405 ~ Si02
`._
`
`+
`
`9
`
`Data
`
`calculated
`
`with
`
`F= 0.30
`
`e N: 100 ~ Si02
`
`I
`
`2.00
`
`. F
`
`m
`0
`
`Fig. 9. The nitrogen and oxygen concentrations in the bulk of 405-A
`Si02 nitrided at 11OO”Cversus hitndation time.
`
`Nitridation
`
`time
`
`(minutes)
`
`0.30. The
`ratio is about
`concentration
`to oxygen
`nitrogen
`increase
`in the bulk
`nitrogen
`concentration
`de-
`rate
`of
`creases with nitridation
`time; however,
`there is no satura-
`tion
`after
`2 h. The monotonic
`decrease
`in the oxygen
`content
`of
`the bulk
`is due
`to the
`continuing
`exchange
`reaction
`of bulk
`Si02 with
`t