Cao Z. (Ed.) Thin Film Growth: Physics, materials science and applications

Подождите немного. Документ загружается.

188 Thin ﬁlm growth

which allows the fabrication of microscopic metallic links with maskless laser

or electron beam writing (Nosaka et al. 2001, Maruyama and Morishita 1996).

It can also be used as a sacrice layer for the fabrication of free-standing

layers of other functional materials, and in write-once optical recording. A

density over Mb/cm

in a 1 ¥ 1 mm dot array has been realized on Cu

thin lm using electron beam irradiation, and the write rate can reach 3.3

Mb/s

(Cremer et al. 2000).

The cubic Cu

N lattice is a rather open structure, noting that the center

of the unit cell is empty; hence it is expected to be a good host material.

In fact, there are both theoretical and experimental investigations into the

metallic Cu

N variant where one extra Cu atom resides at the center of the

unit cell (Moreno-Armenta et al. 2004). There is also report of the synthesis

of ternary Cu

NPd

, where the Pd atoms are believed to occupy some of

the unit cell centers (Hahn and Weber 1996). Remarkably, the interposition

of another metal atom at some cell centers of the Cu

N lattice will result

in semimetals or totally metallic compounds due to the modication of the

energy bands. This may provide an exemplar material system for the study

of semiconducting-to-metallic transition via doping.

Due to the thermal instability of Cu

N, the physical properties for

this material reported in literature are quite inconsistent, sometimes even

controversial. The temperature for the initiation of compound decomposition

is believed to lie in the range from 100°C to 470°C according to different

authors (Asano et al., 1990, Maruyama and Morishita 1996, Wang et al.

1998, Nosaka et al. 2001). As to the band-gap of the semiconducting Cu

experimental values generally fall within the range of 0.8–1.9 eV (Kim

et al., 2001, Borsa et al., 2002), but a value as small as 0.23 eV was also

declared (Hahn and Weber 1996, Ma et al., 2004). The large discrepancy in

the aforementioned data arises obviously from the unstable nature of copper

nitride – in the growth stage the thermal instability results in Cu-rich samples

which have a deteriorated structure dispersed with Cu nanoparticles, while in

the characterization stage the energetic light or electron beams may destroy

the samples to some extent. For both characterization and application, it is

desirable to obtain some Cu

N thin lm samples that are strictly stoichiometric

and of a uniform microstructure. We will see that due to the weak bonding

between the Cu and N atoms, this is, however, a challenge that is not easy

to meet.

8.1.3 Reactive magnetron sputtering for Cu

Since nitrogen molecules do not combine with metal atoms, by reactive

magnetron sputtering of Cu target with nitrogen, the incorporation of nitrogen

to Cu occurs neither in the growing surface, nor on the transport path, but

solely on the target surface via plasma sputtering processes. In fact, on the

ThinFilm-Zexian-08.indd 188 7/1/11 9:42:04 AM



189Thin ﬁlm growth for thermally unstable noble-metal nitrides

transport path and in the growing lm what is of concern is the nitrogen

re-emission from the clusters or the lm. (We have no evidence for the

former, but for the latter it is quite obvious from the morphology studies on

the deposits. See below). Therefore, the key factor for successful deposition

of stoichiometric Cu

N thin lms requires a deliberate adjustment of the

working parameters, including the gas pressure, gas composition and applied

power. Bearing in mind that, while a sufciently high substrate temperature

is favorable for obtaining a satisfactory crystallinity of the lms, it may,

however, foster the re-emission of nitrogen from the deposits, hence the

substrate temperature has to be carefully chosen based on the evaluation of

its inuence on the lm quality.

The lm growth for Cu

N by RF reactive magnetron sputtering was rst

tried by Terada et al. in 1989, where the working gases are Ar and N

in a

ratio of 3:2 (unclearly specied). The success of the method for obtaining

crystalline Cu

N thin lms was judged by the Cu

N (001) and (002) reection

on the X-ray diffraction patterns. It was claimed that the resulting deposits

are ‘almost perfect insulators’, but no data of electrical resistivity were

specied in the literature.

By using dc magnetron sputtering, and selecting a substrate temperature

from room temperature (RT) to 150°C, Reddy et al. obtained (111)-oriented

lms, where the partial pressure of nitrogen of 1 ¥ 10

–3

mbar is only one ftieth

of the total sputtering pressure (Reddy et al. 2007). The electrical resistivity

of the lms has been reduced from 8.7 ¥ 10

–1

to 1.1 ¥ 10

–3

Wm, which we

will see obviously comes from the deviation from chemical stichiometry,

i.e., the deposits suffer from deciency of N. The same research group has

also investigated the inuence of applied power, the partial pressure on the

characters of the deposited lms. They claim that single phase lms of copper

nitride were obtained at a sputtering power of 75 watt with an electrical

resistivity of 5.8 ¥ 10

–2

Wcm and an optical band-gap of 1.84 eV (Reddy et

al. 2007). In a more recent publication, Dorranian et al. found that the N

partial pressure inuenced the structural, electrical and optical properties of

the deposited lms. The X-ray diffraction measurement showed the change

of the preferred orientation of the Cu

N samples from Cu-rich (111) planes

to N-rich (100) planes (Dorranian et al. 2010).

In the past few years it has gradually become clear that the deposition

of high-quality – by quality we refer to stoichiometry, structure, and other

lm features such as orientation, morphology – Cu

N lms by reactive

magnetron sputtering using nitrogen gas is far from trivial. Various working

parameters including the power supply, total pressure of the working gas

and the partial pressure of nitrogen therein, and the substrate temperature

should be balanced on the basis of careful characterization of the deposits,

to which a complementary set of analytical tools should be employed.

ThinFilm-Zexian-08.indd 189 7/1/11 9:42:04 AM



190 Thin ﬁlm growth

8.2 Deposition of stoichiometric Cu

We adopted the radio frequency (27.12 MHz) reactive magnetron sputtering

method to grow Cu

N thin lms on Si (100) wafers in a custom-designed

system, with a gas mixture of argon and nitrogen. The high-purity (5N) Cu

target is 60 mm in diameter. In order to suppress the nitrogen re-emission

by heat, the substrate was held at 60°C just to achieve adequate adhesion

for the deposits. During reactive sputtering growth, the RF power was tuned

to a maximum of 150 W, and the working pressure was maintained at 9.0 ¥

–1

Pa with a total gas ow rate of 4.0 sccm. The proportion of N

in the

working gas is another variable for the optimization of lm composition,

and consequently the structure for this material. Prior to each deposition,

the target was sputtered for 30 minutes to obtain a steady material ow to

the substrate.

The low substrate temperature provokes less deteriorating effects in the

deposits, and consequently samples of prescribed compositions can be well

reproduced. A low substrate temperature and higher working pressure can

effectively suppress the formation of pure metal phase in the deposits. In

order to avoid the destructive effect of the probing particles upon the samples,

repeated measurements have been conducted on the equally fresh samples

for the analysis of the unstable materials considered here.

With a xed power supply of 150 W, the copper content in the deposits

declines steadily with increasing nitrogen proportion, and at a nitrogen

proportion of 90% it approaches the value 76.9% (Fig. 8.2). Further increase

of nitrogen in the working gas failed to obtain an expected stoichiometric

0.3 0.4 0.5 0.6 0.7 0.8 0.9

Flow rate ratio [N

]/([N

]+[Ar])

Cu content

0.95

0.90

0.85

0.80

0.75

8.2 Cu content in the copper nitride thin ﬁlms deposited with varied

nitrogen proportion in the working gas.

ThinFilm-Zexian-08.indd 190 7/1/11 9:42:05 AM



191Thin ﬁlm growth for thermally unstable noble-metal nitrides

N deposit, even with pure nitrogen at this power supply. This failure in

obtaining a strictly stoichiometric copper nitride deposit has nothing to do

with the substrate temperature because nitrogen re-emission from compound

is unlikely to occur at 60°C, but rather lies in the nature of magnetron

sputtering. Since the reaction with nitrogen can only take place in the

region of the plasma ring on the surface of the Cu target, it is the energetic

sputtering there that prevents the formation of stoichiometric Cu

N clusters.

The unwanted presence of elemental copper precipitates in the deposits

provides an explanation for the discrepancy or even controversy in many

characterizations of the lm properties, as will be made clear later.

X-ray diffraction conrmed the nanocrystallinity of the deposits containing

N and Cu biphases. With a Cu content below 78.8%, the reections

from the cubic Cu

N phase become dominant, as can be seen from Fig. 8.3.

The reections assigned to the {001}- and {002}-planes of the Cu

N phase

remain unchanged, while the reection from the {111}-plane of the Cu

phase shifts irregularly through the sequence (a)–(e). This is quite reasonable

since the nitrogen re-emission most preferably occurs on the {111}-planes,

thus their spacing is susceptible to the variation of the overall composition

of the deposits. The lattice constant calculated from this reection oscillates

between 3.868 and 3.797 Å.

Intensity (arb. units)

N (001)

N (111)

Cu (111)

N (002)

Cu (002)

25 30 35 40 45 50 55 60

2q (deg)

8.3 X-ray diffraction patterns for the copper nitride thin ﬁlms with

a Cu content of (a) 76.9%, (b) 78.8%, (c) 84.5%, (d) 89.2%, and (e)

94.0%.

ThinFilm-Zexian-08.indd 191 7/1/11 9:42:05 AM



192 Thin ﬁlm growth

For Cu

N, the {111}-planes consist exclusively of either Cu atoms or N

atoms, hence nitrogen loss most preferably occurs on these planes, which sets

a limit on the size of the Cu

N crystallites (to be discussed below). At the

grain boundaries, the Cu atoms agglomerate to form nanocrystals, generally

being ~5 nm in size as revealed by transmission electron microscopy. The

existence of juxtaposed Cu-rich {111}-planes of Cu

N and the Cu nanocrystals

implies an uncertain factor in determining the optical and, in particlar,

electrical properties. By measuring the electrical resistivity, a sudden drop

was observed when the content of Cu was over 78.8% (Fig. 8.4), suggesting

that a new mechanism for the electrical conductance was now switched on.

Writing the composition of this sample in the form Cu

15.2

–Cu

63.6

21.2

, and

recalling that the excessive Cu atoms exist both as Cu nanocrystals and on

the Cu-rich {111} planes of Cu

N crystallites, an electrical conduction path

solely via Cu chains can be established through the sample, i.e. the percolation

mechanism comes into play. The resistivity for the sample with 76.9% Cu is

4.04 ¥ 10

–5

Wm, just with a little bit more excess Cu it drops to 1.0 ¥ 10

–6

Wm, falling already into the range for good conductors. The sensitivity of

the electrical resistivity to the content of Cu explains the large inconsistency

in the reported values of electrical resistivity for copper nitride deposits.

The lms obtained above at 150 W seriously deviate from the stoichiometry

for Cu

N. In order to obtain a nearly stoichiometric sample, the combination

of power supply and the pressure of working gas (now we employ pure

nitrogen gas) should be tested. Figure 8.5(a) and (b) display the XRD patterns

for the samples deposited with various RF powers, and under two working

pressures of 0.7 Pa and 0.9 Pa for comparison. It can be seen that the lms

0.92 0.88 0.84 0.80 0.76

Cu content

Resistivity (Wm)

1E-4

1E-5

1E-6

1E-7

8.4 Electrical resistivity of the copper nitride deposits at room

temperature as a function of Cu content.

ThinFilm-Zexian-08.indd 192 7/1/11 9:42:05 AM



193Thin ﬁlm growth for thermally unstable noble-metal nitrides

always display the (001) and (002) reections, and the (111) reection

appears when the supplied power is over 100 W. The presence of pure Cu

crystallites is disclosed by the minor Cu-(111) reection at 2q = 43.32°.

Clearly, the higher pressure and the lower RF power favor the formation of

deposits of pure Cu

N phase in a competitive way. At 0.7 Pa, the Cu-(111)

reection is visible in the deposit prepared with an RF power of 100 W; at

0.9 Pa, however, it remains absent when the RF power was raised to 150 W.

This is quite easy to understand since both the higher pressure and lower RF

Intensity (arb. units)

N(001)

N(111)

Cu(111)

N(002)

150 W

100 W

70 W

20 30 40 50 60

2q (deg)

(a)

Intensity (arb. units)

N(001)

N(111)

N(002)

150 W

100 W

70 W

50 W

20 30 40 50 60

2q (deg)

(b)

8.5 XRD patterns for the copper nitride deposits prepared at different

RF powers under a working pressure of (a) 0.7 Pa and (b) 0.9 Pa,

respectively.

ThinFilm-Zexian-08.indd 193 7/1/11 9:42:06 AM



194 Thin ﬁlm growth

power can reduce the energy of nitrogen ions bombarding the Cu target. By

reactive sputtering of Cu target with nitrogen plasma, a very large projectile

energy (large power and low pressure) is unfavorable for the formation of

Cu–N bonds in the sputtered particles. Generally speaking, to obtain nearly

stoichiometrical copper nitrides at a preset low substrate temperature, the

gas pressure and the power should be chosen in such a way that the density

of nitrogen ions is sufciently high and the ion energy has to be low so as

not to spoil the Cu–N bonded species emitted from the bombarded target. A

strict specication of the processing parameters demands a thoughtful mass-

resolved analysis of the bombardment products. Remarkably, in addition to

the [001]-orientation, the larger RF power also provokes the occurrence of

the [111]-orientation for the Cu

N crystallites. From Fig. 8.5 one can say

with condence that with a power supply at 70 or 50 W, [001] oriented thin

lms of copper nitride without any trace of the Cu(111) reection on the

XRD pattern can be obtained.

Copper nitride even in the pure Cu

N phase still reveals a minor deciency

of nitrogen. This is to say that in the stoichiometrical samples we presented

here, the Cu content is still somewhat larger than 75.0%. In fact, from the

energy-dispersive X-ray analysis and photoelectron spectroscopic data, the

lowest Cu content is about 75.6%, to the accuracy of the methods. The

nearly stoichiometrical copper nitride sample should be a typical decit

semiconductor, as seen from the temperature dependence of the electrical

resistivity (see Fig. 8.6(a)). A slight increase in the content of Cu to ~76.0%

can totally destroy this behavior for a decit semiconductor, as the resistivity

rst drops irregularly with the decreasing temperature, and below 50 K it

turns to increase steeply (not shown). In the Cu-rich sample with 78.8% of

Cu, due to the presence of distinct Cu particles (Du et al. 2005), it shows a

characteristic metallic behavior that the electrical resistivity increases linearly

with the temperature rising from 5 K to 300 K (Fig. 8.6(b)). According to the

equation r (T) = r

[1 + a (T – T

)], the temperature coefcient a is determined

to be 7.54 ¥ 10

–4

, much less than the value 1/232 for pure Cu. This can be

explained by the fact that in the Cu-rich copper nitride, the conduction path

is made of both Cu nanoparticles and the Cu

N nanoparticles enclosed by the

Cu-terminated {111}-planes. On the contrary, the stoichiometric Cu

N sample

manifests a typical semiconductor character, the temperature dependence of

electrical resistivity (Fig. 8.6(b)) can be approximately tted by the equation

r(T) = C exp (–E

/kT), with the activation energy E

being approximately

8.59 × 10

–2

eV. For comparison, E

is 1.47 ¥ 10

–2

eV as reported by Wang

et al. (1998).

The host material Cu

N in the cubic anti-ReO

lattice is a typical

semiconductor with an indirect band-gap of ~1.9 eV (Moreno-Armenta et

al. 2004, Nosaka et al. 2001, Maruyama and Morishita 1996). Due to the

weak Cu–N bonding that facilitates nitrogen re-emission from the compound,

ThinFilm-Zexian-08.indd 194 7/1/11 9:42:06 AM



195Thin ﬁlm growth for thermally unstable noble-metal nitrides

the copper nitride samples generally show some deciency of nitrogen;

consequently they display electron-like conductivity as conrmed by the

negative Hall coefcient. By careful control of the processing parameters,

nearly stoichiometric, single-phased Cu

N thin lms can be obtained with

a carrier density brought down to below 3.0 ¥ 10

/cm

With the stoichiometrical Cu

N sample at hand, a high-resolution TEM

image can be obtained, which requires enormous expertise since the energetic

electron beam in a TEM can readily destroy the Cu

N lattice. Figure 8.7

displays the rst TEM image for the stoichiometric Cu

N, exhibiting well-

aligned {001} fringes, which conrms the high crystallographic quality of the

Resistivity (10

–1

Wm)

160 200 240 280

Temperature (K)

(a)

Resistivity (10

–6

Wm)

0 50 100 150 200 250 300

Temperature (K)

(b)

1.45

1.40

1.35

1.30

1.25

1.20

8.6 Temperature dependence of electrical resistivity for a

stoichiometric Cu

N sample (a) and for a Cu-rich sample (b).

ThinFilm-Zexian-08.indd 195 7/1/11 9:42:06 AM



196 Thin ﬁlm growth

deposit. The planar distance for the {001}-planes, i.e., the lattice constant,

measures 0.383 nm in the TEM image, which is in good agreement with the

XRD measurements.

Due to the nitrogen deciency, the deposits consist of distinct nanosized

particles instead of being a single crystalline as claimed by some authors.

This can be judged easily from the line width of the XRD patterns. From

the SEM images (Fig. 8.8) we see that the sample with 76.9% Cu is sharply

contrasted, where the size of the crystallites ranges from 40 to 60 nm. For

the sample with 75.6% Cu, the crystallites are roughly 40 nm in size, and the

image seems smeared. This is obvious since the nearly stoichiometric sample

is an insulator – it causes a serious charging effect under the illumination

of the electron beam in an SEM.

With the nearly stoichiometrical samples, many inconsistencies concerning

the physical properties of Cu

N can be resolved. First, the pure, nearly

stoichiometric Cu

N deposit is a typical wide-gap semiconductor, whose

electrical resistivity at room temperature measures 2 ¥ 10

–2

Wm. Smaller

values for this quantity originate in the incorporation of Cu particles or Cu-

rich {111} surfaces of the Cu

N crystallites. It has a band-gap larger than 1.8

eV, as determined from the photoreectance spectrum even for the slightly

substoichiometric copper nitride thin lm with 76.9% Cu (Du et al. 2005).

Another important feature is the decomposition temperature. To study the

d = 0.383 nm

2 nm

8.7 TEM image of the stoichiometric Cu

N with a Cu content of

75.6%.

ThinFilm-Zexian-08.indd 196 7/1/11 9:42:06 AM



197Thin ﬁlm growth for thermally unstable noble-metal nitrides

thermal stability of the as-deposited lms, the samples were annealed in the

same chamber under the nitrogen environment at 0.9 Pa for 20 minutes. in a

temperature range from 220°C to 420°C. The temperature at which Cu(111)

reection becomes noticeable on the XRD pattern of the post-annealed

samples is taken as the decomposition temperature. Aannealing for 20

minutes. at 220°C and 300°C does not provoke any discernible changes to the

diffraction pattern. Only at 350°C do two tiny peaks emerge at 2q ª 43.32°

and 2q ª 50.55° (arrowed in Fig. 8.9) which can be assigned to the (111)

and (002) reections of the pure Cu phase (Fig. 8.9). The primarily absent

Cu (111) and (002) reections unambiguously signify the onset of thermal

decomposition of Cu

N. At 420°C, only strong Cu peaks were observable,

the Cu

N completely dissolved. Although from the current experimental

data the precise value of the decomposition temperature cannot be xed, we

(a)

200 nm

(b)

200 nm

8.8 Surface morphology of copper nitride thin ﬁlms with a Cu

content of (a) 76.9% and (b) 75.6%.

ThinFilm-Zexian-08.indd 197 7/1/11 9:42:07 AM

