http://journals.cambridge.org Downloaded: 18 Mar 2015 IP address: 128.255.6.125

Noble metal silicide formation in metalyyySi structures during

oxygen annealing: Implications for perovskite-based memory

devices

K. L. Saenger, A. Grill, and C. Cabral, Jr.

IBM Research Division, T.J. Watson Research Center, Yorktown Heights, New York 10598

(Received 17 March 1997; accepted 26 June 1997)

This paper investigates the potentially undesirable noble metal silicide formation

reactions that may occur in noble metal electrodes deposited directly on silicon without

an intervening diffusion barrier. Metal (90–100 nm)ySi structures of PtySi, RhySi,

IrySi, and IryTiySi were annealed in oxygen or nitrogen ambients at temperatures of

640–700

C. Metalysilicon reactions and phase formation were studied by Rutherford

Backscattering Spectroscopy, x-ray diffraction, and electrical resistance measurements.

While complete silicidation was observed in the RhySi, PtySi, and IrySi samples after

640

Cy6 min anneals in nitrogen, some Pt and most of the Ir remained after equivalent

anneals in oxygen. More detailed studies of the IrySi samples indicated that some Ir is

left unsilicided even after a 700

Cy6 min anneal in O

, and that the iridium silicide

formed is the semiconducting IrSi

1.75

. The formation of this silicide can be delayed, but

not prevented, with the use of a 5 nm Ti adhesion layer between the Ir and Si.

I. INTRODUCTION

Noble metals are candidate electrode materials for

semiconductor memory devices incorporating ferroelec-

tric or high-epsilon dielectric materials, such as lead

lanthanum titanate, barium strontium titanate, etc., in part

because noble metals do not form insulating metal oxides

during the oxidizing conditions of dielectric deposition.

In most direct-contact device geometries, noble metal

electrodes contact silicon either directly, or through a

conductive diffusion barrier. This paper investigates the

potentially undesirable noble metal silicide formation re-

actions in metalySi structures not containing a diffusion

barrier layer. While some silicidation may be beneﬁcial,

leading, for example, to improved electrodeySi adhesion

and/or reduced contact resistance, completely silicided

noble metal electrodes are susceptible to oxidation with

attendant formation of a surface layer of low-epsilon

SiO

. An additional concern is the resistivity of the

formed silicide, which will depend on its stoichiometry

and crystalline phase. The silicides of Ir include Ir

and Ir

with room temperature resistivities in the

range

60–70 mV –cm, IrSi with a resistivity

of about

500 mV –cm, and IrSi

1.75

(believed to be the same

silicide described as Ir

in Refs. 2 and 3) which

is reported to be semiconducting with a resistivity

excess of 1 V –cm. A very high resistivity silicide, such

as IrSi

1.75

, at the noble metalySi interface would clearly

be undesirable.

Noble metal silicide formation in metalySi struc-

tures has been extensively studied, but most previous

work pertains to silicide formation during annealing in

inert gas environments, where conditions for silicide for-

mation are most favorable. Brieﬂy reviewed, the phase

formation sequences previously observed in vacuum or

inert gas for the silicides of Pt, Ir, and Rh are as

follows. In the case of Pt on Si, Pt and Si ﬁrst react to

form Pt

SiySi, which reacts further to form PtSiySi.

5–7

For Pt ﬁlms in the thickness range 50–260 nm, all

of the platinum is converted to Pt

Si prior to PtSi

formation, although the completeness of the subsequent

Si to PtSi reaction appears to depend on initial Pt ﬁlm

thickness.

In the case of Ir, the orthorhombic IrSi is the

ﬁrst silicide to form, at temperatures as low as 400

Further reaction at temperatures in the 400–550

range produces IrSi

1.75

. If the Ir ﬁlm is thick enough,

e.g., around 200–240 nm, all three phases—Ir, IrSi, and

IrSi

1.75

—can be present at the same time as distinct

layers parallel to the substrate surface.

9,10

The IrSi

1.75

can

further react with silicon to form the end-member silicide

IrSi

10,11

but this requires hours of high temperature

(,1000

C) annealing. Rh silicide formation shares a

number of similarities with Ir silicide formation, as well

as some important differences.

Both Ir and Rh ﬁrst

form the monosilicide, with the RhSi formation tempera-

ture typically 50

C lower than that for IrSi. A RhSi

phase may be observed with continued annealing at

the RhSi formation temperature, but higher temperature

annealing ﬁrst produces Rh

and then Rh

. The

equivalent Ir

and Ir

rhodium silicides are not

observed to form in thin ﬁlm reactions of Rh with silicon,

and the Rh equivalent to the semiconducting IrSi

1.75

apparently does not exist.

462 J. Mater. Res., Vol. 13, No. 2, Feb 1998  1998 Materials Research Society

HelpCommentsWelcome

Journal of

MATERIALS RESEARCH

http://journals.cambridge.org Downloaded: 18 Mar 2015 IP address: 128.255.6.125

K.L. Saenger

et al.:

Noble metal silicide formation in metalySi structures

There are many qualitative indications that the pres-

ence of oxygen in the annealing ambient interferes with

metal silicide formation. This has been seen in studies of

Ir (60 nm)ySi samples annealed in an ambient of oxygen-

contaminated Ar,

and observed in more detail for PtySi

in studies of Pt (50 nm)ySi annealed in 75 mTorr O

Most, but not all, observations are consistent with a

model

in which oxygen diffuses from the outside of the

sample into the metal layer until it reaches the reaction

front of the ﬁrst silicide to form (Pt

Si). If the oxygen

concentration exceeds a critical level, the oxide blocking

layer formed may be sufﬁcient to stop both further

consumption of Pt and formation of any additional Pt

Si.

Continued annealing then produces the stable PtSi phase

from further reaction of the Pt

Si layer with Si.

This paper compares noble metal silicide formation

in structures of PtySi, RhySi, IrySi, and IryTiySi annealed

in atmospheric pressure ambients of oxygen or nitrogen

at temperatures of 640–700

C. The thickness of the

metal ﬁlms was 90–100 nm. The oxygen annealing

conditions were chosen to approximate the worst case

oxygen exposure anticipated for processing of high ep-

silon dielectric materials. The results are compared with

existing models for the effects of oxygen on silicide

formation, and are discussed in the context of integrating

noble metal electrodes and perovskite-based memory

devices into silicon technology.

II. EXPERIMENTAL PROCEDURES

The RhySi and IrySi samples used in this study were

prepared by electron-beam evaporation on polycrys-

talline silicon (poly-Si) or k100l-oriented single crystal

silicon (c-Si) substrates heated to ,225–250

C. All sili-

con substrates were cleaned with dilute (10 : 1) HF prior

to deposition. The c-Si substrates were boron-doped

type Si with a conductivity of 1.6–2.4 mV –cm.

The poly-Si substrates consisted of 374 nm of undoped

polysilicon deposited by low pressure chemical vapor

deposition (LPCVD) on 350 nm of thermal oxide. Film

thicknesses were 90 nm for Rh and 100 nm for Ir.

The PtySi samples were prepared by electron-beam

evaporation on unheated c-Si and poly-Si substrates. The

c-Si substrates were lightly doped p type Si and the poly-

Si substrates consisted of 150 nm of undoped polysilicon

on 300 nm of thermal oxide. The Pt ﬁlm thickness was

100 nm.

The IryTiypoly-Si samples were prepared on poly-Si

substrates consisting of 374 nm of undoped polysilicon

on 350 nm of thermal oxide by two methods. In the ﬁrst

method, 10 nm of Ti and 100 nm of Ir were sequentially

deposited during a single electron-beam evaporation

on substrates kept approximately at room temperature.

These ﬁlms were also deposited directly on thermally

oxidized silicon wafers to examine Ir oxidation and Ti

diffusion on substrates less reactive than c- or poly-

Si. In the second method, the poly-Si substrates were

ﬁrst coated with 10 nm of sputter-deposited Ti. After

exposure to air, 100 nm of Ir was deposited by electron

beam evaporation onto the Tiypoly-Si substrates heated

to ,225–250

Rapid thermal anneals (AG Heatpulse) were per-

formed in oxygen or nitrogen on pieces placed on a

silicon susceptor plate containing an embedded thermo-

couple. Annealing conditions were typically 640

C for

6 min, although some samples were annealed at 700

for 1 or 6 min for comparison.

Metalysilicon reactions and phase formation were

studied by Rutherford Backscattering Spectroscopy

(RBS) and x-ray diffraction (XRD). RBS was performed

with 2.3 MeV

He ions. Ex situ XRD was performed

on the as-deposited and annealed samples using CuK

over the 2u range 20–70

. Some in situ XRD was

also performed on IrySi samples during rapid thermal

annealing in a He ambient, using a synchrotron

source as described elsewhere.

Electrical resistance

measurements were performed with a 4-point probe

apparatus.

III. RESULTS

A. RhyyySi

Figure 1 shows the XRD spectra of the Rh ﬁlms on

polysilicon substrates, as-deposited and after annealing.

The as-deposited Rh ﬁlms showed a mixture of k111l

and k200l crystal orientations, with k111l predominating

on both poly-Si [Fig. 1(a)] and c-Si (not shown). After a

640

Cy6 min anneal in N

, the Rh was completely con-

sumed to form a roughly randomly oriented orthorhom-

bic RhSi phase on both poly-Si [Fig. 1(c)] and c-Si

substrates. RBS data conﬁrmed the complete consump-

tion of Rh to form a silicide having RhSi stoichiometry

on both poly-Si (Fig. 2) and c-Si (not shown). Room

temperature resistance values of 8 Vysquare measured

on the insulating poly-Si sample substrates corresponded

to a resistivity of about 140 mV –cm (based on an

estimated 180 nm RhSi ﬁlm thickness produced from

a 90 nm Rh ﬁlm).

RhySi samples annealed in O

at 640

Cy6 min

produced a range of results. XRD of all oxygen-annealed

poly-Si and c-Si samples showed strong RhSi forma-

tion [Fig. 1(b)]. The corresponding RBS data [Fig. 2(b)]

indicated that all oxygen-annealed samples had small

amounts of oxygen in a surface layer having an Rh

content intermediate between RhSi and pure Rh. The

absence of any XRD peaks associated with oxides of Rh

suggests that any formed oxide products are amorphous.

On c-Si substrates, the Rh was completely consumed

by reaction with Si and/or oxygen, as indicated by

J. Mater. Res., Vol. 13, No. 2, Feb 1998 463

http://journals.cambridge.org Downloaded: 18 Mar 2015 IP address: 128.255.6.125

K.L. Saenger

et al.:

Noble metal silicide formation in metalySi structures

FIG. 1. X-ray diffractograms of Rh (90 nm)ypoly-Si: (a) as-depos-

ited; (b) O

-annealed at 640

C for 6 min; (c) N

-annealed at 640

for 6 min. (Symbols identify peaks originating from Rh, RhSi, and

the substrate.)

the absence of any Rh peaks in the x-ray diffraction

spectra. However, these spectra showed some additional

unidentiﬁed peaks. The strongest of these peaks (at 2u 

39.3

and d-spacing of 2.29

A) was previously observed

in an earlier study of Rh silicide formation

and may

possibly be associated with Rh

Si or the monoclinic

RhSi phase.

Rh consumption in the oxygen-annealed ﬁlms

showed some variability on the poly-Si substrates. The

Rh-rich surface layer seen by RBS on the Rhypoly-

Si samples was shown by XRD [Fig. 1(b)] to contain

some unreacted Rh in a ﬁlm having a resistivity of

R  8 Vysquare but no unreacted Rh in a ﬁlm with

R  13 Vysquare. Samples annealed in oxygen at

700

C for 6 min showed complete Rh consumption,

a resistance R  15 Vysquare, and RhSi peak intensity

comparable to that found for the completely reacted

samples annealed in oxygen at 640

FIG. 2. RBS of Rhypoly-SiySiO

ySi samples after different anneals:

(a) as-deposited; (b) O

-annealed at 640

C for 6 min; (c) N

-annealed

at 640

C for 6 min. Insets are 33.

B. PtyyySi

Figure 3 shows the XRD spectra of the Pt ﬁlms on

polysilicon substrates, as-deposited and after annealing.

The as-deposited Pt ﬁlms showed a mixture of k111l

and k200l crystal orientations, with k111l predominating

on both poly-Si [Fig. 3(a)] and c-Si (not shown). After

a 640

Cy6 min anneal in N

, the Pt was completely

consumed to form a strongly k002l-textured orthorhom-

bic PtSi phase on both poly-Si [Fig. 3(c)] and c-Si

substrates. RBS data conﬁrmed complete consumption

of Pt and the formation of a silicide having PtSi stoi-

chiometry on both poly-Si (Fig. 4) and c-Si (not shown).

Room temperature resistance values of 2.3 Vysquare

measured on the insulating poly-Si sample substrates

corresponded to a resistivity of about 45 mV –cm (based

on an estimated 197 nm PtSi ﬁlm thickness produced

from a 100 nm Pt ﬁlm).

XRD indicated the presence of unreacted Pt along

with strong PtSi formation in samples annealed in oxy-

gen at 640

C or 700

C for 6 min. As with the samples

annealed in nitrogen, the PtSi formed was orthorhom-

bic and strongly k130l-textured PtSi on both poly-Si

[Fig. 3(b)] and c-Si. Room temperature resistance meas-

urements for oxygen-annealed samples on the insulat-

ing poly-Si substrates indicated resistances somewhat

higher than those observed for the nitrogen annealed

samples (about 2.8 Vysquare, versus 2.3 Vysquare for

the sample annealed in nitrogen). RBS data for the

640

Cy6 min oxygen-annealed samples were consistent

with complete consumption of Pt to form PtSi, although

the shape of the Pt peaks suggested a nonuniform Pt

distribution in the PtSi layer, with local regions of Pt

enrichment both at the PtSi surface and also within the

PtSi. The thickness of the unreacted Pt layer detected by

464 J. Mater. Res., Vol. 13, No. 2, Feb 1998

http://journals.cambridge.org Downloaded: 18 Mar 2015 IP address: 128.255.6.125

K.L. Saenger

et al.:

Noble metal silicide formation in metalySi structures

FIG. 3. X-ray diffractograms of Pt (100 nm)ypoly-Si: (a) as-depos-

ited; (b) O

-annealed at 640

C for 6 min; (c) N

-annealed at 640

for 6 min. (Symbols identify peaks originating from Pt, PtSi, and the

substrate.)

XRD must therefore be very thin. Oxygen in the PtSi

layer was below detection limits for the poly-Si samples

(Fig. 4), and barely detectable in the c-Si samples. RBS

data for the 700

Cy6 min oxygen-annealed samples

were similar to those for the 640

Cy6 min samples,

except for the presence of a small surface oxygen peak

in the sample on the poly-Si substrate.

C. IryyySi

Figure 5 shows the XRD spectra of the Ir ﬁlms on

polysilicon substrates, as-deposited and after annealing.

The as-deposited Ir ﬁlms showed a mixture of k111l

and k200l crystal orientations, with k111l predominating

on both poly-Si [Fig. 5(a)] and c-Si (not shown). After

nitrogen annealing at 640

C for 6 min, the Ir was

completely consumed to form IrSi

1.75

on both poly-Si

[Fig. 5(c)] and c-Si substrates. RBS data conﬁrmed that

the N

-annealed samples were completely silicided on

both poly-Si and c-Si substrates. This can be seen in

FIG. 4. RBS of Ptypoly-SiySiO

ySi samples after different anneals:

(a) as-deposited; (b) O

-annealed at 640

C for 6 min; (c) N

-annealed

at 640

C for 6 min. Insets are 36.

Fig. 6 which compares the RBS data for the as-deposited

and annealed ﬁlms on poly-Si substrates. Room tempera-

ture resistance values of 120 3 10

Vysquare measured

on the insulating poly-Si sample substrates corresponded

to a IrSi

1.75

resistivity of about 4 V –cm, based on an

estimated 300 nm IrSi

1.75

ﬁlm thickness produced from

a 100 nm Ir ﬁlm. This value is consistent with the values

previously reported elsewhere.

The effects of oxygen annealing on Irypoly-Si

samples are presented in Figs. 5(b) (XRD) and 6(b)

(RBS) for the 640

Cy6 min anneals. RBS [Fig. 6(b)]

indicated that most of the Ir was left unreacted, except for

the formation of a thin silicide layer at the IrySi interface.

Similar results were observed for the c-Si substrates,

although these samples showed some additional IrO

the top surface of the Ir. XRD [Fig. 5(b)] indicated an

increased Ir k111l peak intensity and narrower half-width

after oxygen annealing, consistent with Ir grain growth.

Film resistance on the Irypoly-Si samples remained close

to the original 1.7–1.9 Vysquare value, which should

be compared to the 120 3 10

Vysquare value after the

nitrogen anneals and the 0.7–0.9 Vysquare values for

nitrogen-annealed Ir ﬁlms on the less reactive thermal

oxide substrates. Similar resistance values of about

2.0 Vysquare were also found for Irypoly-Si samples

after a 700

Cy6 min oxygen anneal, suggesting that

most of the Ir is left unreacted even after this anneal.

XRD conﬁrmed the presence of unreacted Ir on both

Irypoly-Si and Iryc-Si samples after the 700

Cy6 min

oxygen anneals, although these samples had a slightly

reduced Ir peak intensity and somewhat higher IrSi

1.75

peak intensity, relative to the samples annealed in

oxygen at 640

The IrO

layer indicated by RBS in the oxygen-

annealed c-Si substrate samples is consistent with about

J. Mater. Res., Vol. 13, No. 2, Feb 1998 465

http://journals.cambridge.org Downloaded: 18 Mar 2015 IP address: 128.255.6.125

K.L. Saenger

et al.:

Noble metal silicide formation in metalySi structures

FIG. 5. X-ray diffractograms of Ir (100 nm)ypoly-Si: (a) as-depos-

ited; (b) O

-annealed at 640

C for 6 min; (c) N

-annealed at 640

for 6 min. (Symbols identify peaks originating from Ir, IrSi

1.75

, and

the substrate.)

150

A of IrO

. Tetragonal IrO

peaks are deﬁnitely

present in the XRD spectra of Irypoly-Si and Iryc-Si

samples given oxygen anneals at higher temperatures

(700

C for 6 min), although these IrO

peaks are much

weaker than those seen in similarly annealed IryTi ﬁlms

deposited on less reactive SiO

substrates. However,

phase identiﬁcation of IrO

in the 640

C-annealed Ir

samples on silicon is only tentative due to overlap of the

IrO

peaks with those of IrSi

1.75

and the substrate. IrO

formation would be expected to somewhat increase the

sample resistance, since the reported bulk IrO

resistivity

of 30 mV –cm (Ref. 15) is higher than the 5.3 mV –cm

of bulk Ir metal. However, the use of ﬁlm resistance

values to estimate the relative amounts of IrSi

1.75

, Ir,

and IrO

is inadvisable due to the many factors acting

to make the resistivities of thin ﬁlms higher than their

values in bulk.

The iridium silicide phase formation sequence in

inert oxygen-free ambients, namely the transformation

FIG. 6. RBS of Irypoly-SiySiO

ySi samples after different anneals:

(a) as-deposited; (b) O

-annealed at 640

C for 6 min; (c) N

-annealed

at 640

C for 6 min. Insets are 35.

from IrySi to IrSiySi to IrSi

1.75

ySi was conﬁrmed by

in situ XRD in He with a synchrotron source for Ir

ﬁlms deposited on c-Si and poly-Si substrates. At a

heating rate of 3

Cys, the temperature midpoints for

IrSi formation from 100 nm of Ir on silicon were around

655

C for Ir on poly-Si and 665

C for Ir on c-Si.

Likewise, temperature midpoints for IrSi transformation

to IrSi

1.75

were around 739

C for IrSi on poly-Si and

743

C for IrSi on c-Si. Ex situ XRD of an Irypoly-Si

sample quenched from 715

C and having a resistance

of 8.3 Vysquare showed all three phases—Ir, IrSi, and

IrSi

1.75

—simultaneously.

D. Ir (100 nm)yyyTi (10 nm)yyySi

As in the case of the Ir ﬁlms deposited directly

on silicon, the as-deposited IryTiySi ﬁlms showed a

mixture of k111l and k200l Ir crystal orientations, with

k111l predominating on both poly-Si and c-Si substrates.

Figure 7 shows the XRD spectra for the Ir ﬁlms on

polysilicon substrates, as-deposited and after annealing,

for the case in which the Ir and Ti were deposited in

the same evaporation run without air exposure between

the two layers. The results obtained after an anneal at

640

C for 6 min in N

[Fig. 7(c)], are clearly different

from those obtained for the IrySi samples [Fig. 5(c)].

Not only is a substantial amount of the Ir still left, but

the silicide that has formed is the orthorhombic and

conductive IrSi phase, rather than the semiconducting

IrSi

1.75

. RBS data conﬁrmed the presence of unreacted

Ir in the N

-annealed samples, and showed some IrySi

interface mixing and dilution of Ti with Si. Similar

results were seen for the IryTi ﬁlms deposited in the two-

step deposition, although these ﬁlms showed slightly less

silicidation after the nitrogen anneals. Samples deposited

466 J. Mater. Res., Vol. 13, No. 2, Feb 1998

http://journals.cambridge.org Downloaded: 18 Mar 2015 IP address: 128.255.6.125

K.L. Saenger

et al.:

Noble metal silicide formation in metalySi structures

FIG. 7. X-ray diffractograms of Ir (100 nm)yTi (10 nm)ypoly-Si

samples (single-step deposition): (a) as-deposited; (b) O

-annealed at

640

C for 6 min; (c) N

-annealed at 640

C for 6 min. (Symbols

identify peaks originating from Ir, IrSi, and IrO

by both methods showed resistance drops of about

35% from their initial values of around 2–3 Vysquare,

probably due to Ir grain growth.

More silicide formation was observed after nitrogen

annealing at higher temperature (700

Cy6 min). XRD

of these samples showed mostly a mixture of IrSi and

IrSi

1.75

peaks, with some differences depending on

the IryTiySi preparation method. Samples deposited

at room temperature in a single deposition run showed

predominantly IrSi

1.75

plus some hints of IrSi, and a

resistance of 64 Vysquare. Samples deposited in the

two-step process showed less reaction progress, namely

a more even mixture of IrSi

1.75

and IrSi, along with a

weak peak of residual Ir (111). The resistance for this

sample was 7.4 Vysquare. In addition, there were some

weak but distinct unidentiﬁed peaks having d-spacings

(2u) of 2.445

A (36.8

), 2.400

A (37.5

), 2.359

(38.2

), and 2.140

A (42.2

). While some of these

peaks roughly agree with the peak positions expected

for a mixture of titanium silicides, the intensities seem

too strong for the small amount of titanium present.

Only one of these peaks (d  2.140

A) appeared

in the single-step deposited sample. We attribute the

reduced reactivity of the two-step deposited samples

to a titanium oxide layer formed at the TiyIr interface

during air exposure between the Ti and Ir depositions.

This layer must be extremely thin, since RBS detected

no oxygen in the as-deposited ﬁlms.

As expected, XRD showed substantial Ir remaining

after annealing in oxygen for 640

C at 6 min [Fig. 7(b)].

Somewhat surprising was the complete lack of any

silicide peaks, and the clear formation of IrO

. RBS

conﬁrmed the presence of unreacted Ir, and indicated the

presence of O in the Ti layer, interdiffusion at the SiyTi

interface, and IrO

at the top surface of the Ir. However,

the Ti appeared to be well localized at the IrySi interface.

This contrasts with our ﬁndings for the Ti in oxygen-

annealed IryTiySiO

ySi samples, which showed diffusion

of Ti to the top Ir surface. After higher temperature

oxygen anneals at 700

C for 6 min, XRD of the IryTiySi

samples showed stronger IrO

peaks, but at most a hint

of IrSi

1.75

in the one-step deposited samples and probably

none in the two-step deposited samples. More certainty

about the absence of IrSi

1.75

is not possible due to overlap

of IrO

XRD peaks with those of IrSi

1.75

IV. DISCUSSION

We now discuss these ﬁndings as they relate to

the choice of noble metal electrodes for perovskite-

based memory devices in silicon technology. Due to its

propensity for silicidation, Rh is clearly unsuitable as

an electrode material in geometries lacking a diffusion

barrier at the Rhysilicon interface. While Pt is more

resistant to complete silicidation than Rh, most of the Pt

layer still silicides. The volume changes and structural

instabilities associated with the expected amount of

PtSi formation also make Pt a poor choice for an

electrode material, unless a diffusion barrier is used.

Of the three metals examined, only Ir shows enough

resistance to silicide formation to merit further con-

sideration as an electrode material in direct contact

with silicon. However, the IrSi

1.75

phase formed has

such a high resistivity (1–5 V –cm) that its presence

in even small amounts can adversely affect the IrySi

contact resistance. The silicidation of even 30

Aof

Ir to form about 100

A of IrSi

1.75

would contribute a

resistance of about 3000 V for a round contact having

a diameter of 0.4 mm. While this additional resistance

might be tolerable, it is undesirable, especially given the

possibility of thicker silicide layer formation.

A thin Ti layer at the IrySi interface clearly delays

the iridium silicide formation sequence and shows some

promise as a quasi-diffusion barrier. However, the use-

J. Mater. Res., Vol. 13, No. 2, Feb 1998 467

http://journals.cambridge.org Downloaded: 18 Mar 2015 IP address: 128.255.6.125

K.L. Saenger

et al.:

Noble metal silicide formation in metalySi structures

fulness of Ti is expected to be limited by its propensity

for diffusing to the Ir top surface in IryTiySiO

structures

where the Ti may oxidize. Such effects were observed

in the present work and previously for Ti with Pt.

16,17

The differences in the relative amounts of noble

metal silicide and unreacted noble metal left after oxygen

annealing in the RhySi, PtySi, and IrySi samples reﬂect

differences in the competition between (i) silicidation

and (ii) oxygen diffusion to and pileup at the reaction

front between the silicided and unreacted noble metal.

Silicide formation is fastest with RhySi and slowest with

IrySi. Both Rh and Pt are expected to be permeable

to oxygen; Rh is reported

to absorb oxygen when

melted, and Pt is reported

to absorb oxygen when

heated in air. Surface IrO

formation rather than oxygen

absorption is reported for Ir.

Our data thus suggest

that the Rh silicides before enough oxygen is absorbed to

have an effect. Oxygen has more of an effect on the more

slowly siliciding PtySi, perhaps because platinum silicide

formation is still in progress at temperatures high enough

for Pt to start acting as a sponge for oxygen. Silicidation

in IrySi is even slower, but Ir is also less permeable to

oxygen. However, the relative abundance of unreacted Ir

versus the iridium silicide clearly indicates that oxygen

is reaching the reaction front in sufﬁcient amounts to

stop the silicidation reaction. It can thus be concluded

that the surface IrO

is either permeable to oxygen, or

does not become thick enough to interfere with oxygen

diffusion until after the silicidation reaction has stopped.

V. CONCLUSIONS

This paper compared noble metal silicide formation

in structures of PtySi, RhySi, IrySi, and IryTiySi annealed

in oxygen or nitrogen at 640

C for 6 min, the noble

metal having an initial thickness of about 100 nm. While

complete silicidation was observed in the RhySi, PtySi,

and IrySi samples after nitrogen annealing, some Pt

and most of the Ir remained after equivalent anneals

in oxygen. In contrast to the low resistivity silicides

formed from RhySi (orthorhombic RhSi, with a resistiv-

ity ,140 mV –cm) and PtySi (orthorhombic PtSi, with

a resistivity ,45 mV –cm), the iridium silicide formed

from IrySi is the semi-conducting IrSi

1.75

with a very

high resistivity (,1–5 V–cm).

The formation of the IrSi

1.75

silicide can be delayed,

although not prevented, with the use of a 5 nm Ti

adhesion layer. In this case, unreacted Ir remained after

both nitrogen and oxygen anneals, with XRD showing no

silicide formation after oxygen annealing and formation

of the conductive (,500 mV –cm) orthorhombic IrSi

after nitrogen annealing.

ACKNOWLEDGMENTS

The authors thank Roy Carruthers for the ﬁlm dep-

ositions, G. Coleman for the RBS measurements, and

F. d’Heurle for a helpful discussion.

REFERENCES

1. A. Grill and M.J. Brady, Integrated Ferroelectrics 9, 299 (1995).

2. C.E. Allevato and C. B. Vining, J. Alloys Compounds 200,99

(1993).

3. J. Schumann, D. Elefant, C. Glandun, A. Heinrich, W. Pitschke,

H. Lange, W. Henrion, and R. Grotzschel, Phys. Status Solidi A

145, 429 (1994).

4. S. Petersson, J. A. Reimer, M. H. Brodsky, D. R. Campbell,

F. d’Heurle, B. Karlsson, and P. A. Tove, J. Appl. Phys. 53, 3342

(1982).

5. C. Canali, C. Catellani, M. Prudenziati, W. H. Wadlin, and C. A.

Evans, Jr., Appl. Phys. Lett. 31, 43 (1977).

6. M. Wittmer, J. Appl. Phys. 54, 5081 (1983).

7. C. Harder, L. Hammer, and K. Muller, Phys. Status Solidi A 146,

385 (1994).

8. S.R. Das, K. Sheergar, D-X. Xu, and A. Naem, Thin Solid Films

253, 467 (1994).

9. T. Rodriguez, H. Wolters, A. Almendra, J. Sanz-Maudes, M. F.

DaSilva, and J.C. Soares, in Infrared Detectors—Materials, Proc-

essing, and Devices, edited by A. Appelbaum and L.R. Dawson

(Mater. Res. Soc. Symp. Proc. 299, Pittsburgh, PA, 1994), p. 313.

10. S. Petersson, J. Baglin, W. Hammer, F. d’Heurle, T. S. Kuan,

I. Ohdomari, J. de Sousa Pires, and P. Tove, J. Appl. Phys. 50,

3357 (1979).

11. M. Wittmer, P. Oelhafen, and K. N. Tu, Phys. Rev. B 35, 9073

(1987).

12. S. Petersson, R. Anderson, J. Baglin, J. Dempsey, W. Hammer,

F. d’Heurle, and S. LaPlaca, J. Appl. Phys. 51, 373 (1980).

13. M. Fernandez, T. Rodriguez, A. Almendra, J. Jimenez-Leube,

and H. Wolters, in Infrared Detectors—Materials, Processing,

and Devices, edited by A. Appelbaum and L.R. Dawson (Mater.

Res. Soc. Symp. Proc. 299, Pittsburgh, PA, 1994), p. 325.

14. R.A. Roy, L. A. Clevenger, C. Cabral, Jr., K. L. Saenger,

S. Brauer, J. Jordan-Sweet, J. Bucchignano, G.B. Stephenson,

G. Morales, and K. F. Ludwig, Jr., Appl. Phys. Lett. 66, 1732

(1995).

15. L. Krusin-Elbaum, M. Wittmer, and D. S. Yee, Appl. Phys. Lett.

50, 1879 (1987).

16. R. Bruchhaus, D. Pitzer, O. Eibl, U. Scheithauer, and W. H

osler,

in Ferroelectric Thin Films II, edited by A. I. Kingon, E.R. Myers,

and B. Tuttle (Mater. Res. Soc. Symp. Proc. 243, Pittsburgh, PA,

1992), p. 123.

17. R.E. Jones, Jr., P. D. Maniar, J. L. Dupuie, J. Kim, R. Moazzami,

J. Witowski, M.L. Kottke, N. C. Saha, and R. B. Gregory, in

Ferroelectric Thin Films IV, edited by B. A. Tuttle, S. B. Desu,

R. Ramesh, and T. Shiosaki (Mater. Res. Soc. Symp. Proc. 361,

Pittsburgh, PA, 1995), p. 223.

18. The Merck Index, 10th ed., edited by M. Windholz (Merck, Inc.,

1983).

468 J. Mater. Res., Vol. 13, No. 2, Feb 1998