Sie sind auf Seite 1von 11

JOURNAL OF APPLIED PHYSICS 106, 113717 共2009兲

Anodic vacuum arc developed nanocrystalline Cu–Ni and Fe–Ni thin film
thermocouples
S. K. Mukherjee,a兲 M. K. Sinha, B. Pathak, S. K. Rout, and P. K. Barhai
Department of Applied Physics, Birla Institute of Technology Mesra, Ranchi, Jharkhand 835215, India
共Received 14 May 2009; accepted 2 October 2009; published online 11 December 2009兲
This paper deals with the development of nanocrystalline Cu–Ni and Fe–Ni thin film thermocouples
共TFTCs兲 by using ion-assisted anodic vacuum arc deposition technique. The crystallographic
structure and surface morphology of individual layer films have been studied by x-ray diffraction
and scanning electron microscopy, respectively. The resistivity, temperature coefficient of
resistance, and thermoelectric power of as deposited and annealed films have been measured. The
observed departure of these transport parameters from their respective bulk values can be
understood in terms of intrinsic scattering due to enhanced crystallite boundaries. From the
measured values of thermoelectric power and the corresponding temperature coefficient of
resistance of annealed Cu, Ni, and Fe films, the calculated values of log derivative of the mean free
path of conduction electrons at the Fermi surface with respect to energy 共U兲 are found to be −0.51,
3.22, and −8.39, respectively. The thermoelectric response of annealed Cu–Ni and Fe–Ni TFTCs has
been studied up to a maximum temperature difference of 300 ° C. Reproducibility of TFTCs has
been examined in terms of the standard deviation in thermoelectric response of 16 test samples for
each pair. Cu–Ni and Fe–Ni TFTCs agree well with their wire thermocouple equivalents. The
thermoelectric power values of Cu–Ni and Fe–Ni TFTCs at 300 ° C are found to be 0.0178 and
0.0279 mV/ ° C, respectively. © 2009 American Institute of Physics. 关doi:10.1063/1.3257252兴

I. INTRODUCTION limitations such as low flux ionization that cannot be varied


significantly and low achievable deposition rates, to mention
Researchers have been developing and employing thin a few.
film thermocouples 共TFTCs兲 for temperature measurement We have employed anodic vacuum arc 共AVA兲 deposition
inside jet aircraft engines and in a variety of other applica- technique for the development of our TFTC samples. This
tions wherein measurement of surface temperature with technique has not been exploited fully and until date does not
minimal intrusion1 is a dire necessity. Their remarkable prop- enjoy the acclaim at par with its advantages over other depo-
erties include very fast response to temperature changes, sition techniques. AVA is a source of pure metal vapor
very high spatial resolution, and ability to form an integral plasma free from macroparticles that offers various advan-
part of the component whose temperature is being measured. tages such as high deposition rates, low thermal load on the
These are the reasons that make TFTCs useful in temperature substrate, and high degree of ionization with energetic
measurement applications for rapid thermal processing,2 in- ions30–32 over other deposition techniques. Flux ionization
stantaneous heat flux measurements inside combustion can be varied significantly from 1% to 30% while ion ener-
chambers,3,4 etc. Burger and Van Citter 共1930兲, and Harris gies range from 2 to 150 eV offering a better control over
and Johnson 共1933兲 were the first to report on the preparation the deposition process.33 Deposition rate up to 40 nm/ s is
and use of TFTCs.5 Subsequently, other known deposition achievable with input power in the range of 400– 900 W,
techniques were also employed for fabricating TFTCs. Mar- that is, much higher than those obtainable by using other
shall et al.5 reported the performance of various metal pair physical vapor deposition processes. For example, in sputter
TFTCs using vacuum evaporation technique and found their deposition even with much higher power inputs the sputter
properties very close to those fabricated using wires. Later, rates are quite low.34 The high deposition rate 共5 – 40 nm/ s兲
Kreider and his co-workers extensively studied various sput- of AVA coupled with macroparticle free stream of highly
ter deposited conventional and nonconventional TFTCs4,6–8 ionized energetic particles makes it industrially viable. In
Recently, different researchers in this area have attempted to this paper, we study the preparation and performance of
harness the advantages of a host of deposition techniques Cu–Ni and Fe–Ni TFTCs.
such as sputtering,9–17 vacuum evaporation,18–23 electro-
plating,24–26 chemical vapor deposition,27,28 and pulsed laser II. EXPERIMENTAL
deposition,29 for development of TFTCs and tailoring its
properties to suit one’s stringent requirement. Despite all the A. AVA: Construction, operation, and phenomenology
advantages that each of the aforesaid techniques offer with The schematic diagram of AVA deposition system is
regard to the film properties, they have their own inherent shown in Fig. 1. Both electrodes, the cathode, and the anode
are located in a vacuum chamber and connected to water-
a兲
Electronic mail: sanat_aphy@yahoo.co.in. cooled electrode supports being electrically insulated from

0021-8979/2009/106共11兲/113717/11/$25.00 106, 113717-1 © 2009 American Institute of Physics


113717-2 Mukherjee et al. J. Appl. Phys. 106, 113717 共2009兲

B. Measurement details
X-ray diffractometer 共XRD兲 共Philips, The Netherland兲
with Cu K␣ radiation 共0.154 06 nm兲 in the range 2␪
= 30° – 80° with a step of 2␪ = 0.02° and a scan rate of
0.05° / min has been used for the estimation of average crys-
tallite size and residual strain of the single layer films. The
average crystallite size was estimated from the diffraction
peak broadening using the Scherrer’s formula and Warren–
Biscoe correction factor for instrumental peak broadening.36
XRD patterns of the corresponding bulk components were
recorded for comparison. The surface morphology and thick-
ness of the films were obtained using field-emission scanning
electron microscopy 共FE-SEM兲 共Hitachi, Japan兲.

FIG. 1. 共Color online兲 Schematic diagram of AVA deposition unit.


JCPDS file 4-0836
(a)
as deposited

annealed

the chamber. The cathode consists of a graphite disk radially


shielded by a cylinder made up of stainless steel. A tungsten
basket containing the evaporating material represents the an-
ode. The electrodes are connected to a dc power supply
共100 V , 600 A兲 via an Ohmic resistor 共0.5 ⍀兲, improving the

(111)
arc stability. Prior to ignition, the chamber is evacuated to
1 ⫻ 10−3 Pa with the help of a diffusion pump backed by a
rotary pump. Arc is struck by briefly contacting the anode

(200)

(220)
and the cathode.
After contacting the electrodes shortly one or few bright
cathode spots appear on the cathode surface. These cathode
spots move slowly and show severe erosion of the cathode (b) JCPDS file 4-0850

surface. In this phase the arc is sustained by cathode material as deposited

共cathodic arc兲. This cathodic plasma expands in the direction annealed

of the anode and forms an electrical connection between the


anode and the cathode. The anode is heated up due to Ohmic
heating and particle bombardment from the cathodic plasma.
Eventually, the anode heats up sufficiently so that the anode
material is evaporated at which time the color of the plasma
changes to that of the most intense spectral line of the anode
(111)

material and the cathode spots transform into many, rapidly


moving spots that no longer erode the cathode. Thus the
(200)

(220)
anodic plasma is established, and the discharge mode
switches to the AVA. This anodic plasma expands into the
ambient vacuum and condenses at the wall of the chamber
and at the substrate surface.30 (c) JCPDS file 6-0696

Cu, Fe, and Ni wires of 99.98% purity were used as asdeposited

annealed
consumable portion of the anode for arc evaporation and
deposition of Cu–Ni and Fe–Ni TFTCs on sputter cleaned
good quality 75⫻ 25⫻ 1.3 mm3 microscope glass slides. The
film deposition parameters and the substrate cleaning proce-
dures are reported elsewhere.32 Metal mask prepared using
electrodischarge machining was used for obtaining the de-
(110)

sired linewidth and shape of the developed TFTCs. In order


to minimize size effects, the thicknesses of individual layers
(200)

are maintained between 250 and 350 nm.35 This is done by


actuating the substrate shutter. The substrates were exposed
to the plasma after the first 10 s of arc ignition, as the tem-
perature of the anode region rises rapidly during the first few 30 40 50 60 70 80

seconds. Normally, beyond 10 s, the temperature of the an- FIG. 2. 共Color online兲 XRD patterns of as deposited and annealed films of
ode region has been found to stabilize. 共a兲 Cu, 共b兲 Ni and 共c兲 Fe.
113717-3 Mukherjee et al. J. Appl. Phys. 106, 113717 共2009兲

The room temperature resistivity 共␳ f 兲 and temperature ⌬d ds − du


coefficient of resistivity 共␤ f 兲 of the single layer films were ␧= = , 共2兲
d du
measured by using Van der Pauw method.32 Temperature was
varied from 27 to 37 ° C for the determination of tempera- where du and ds are the unstressed and stressed lattice spac-
ture coefficient of resistivity. The thermoelectric output was ings, respectively. It may be noted that if ds ⬎ du, then ␧ is
recorded with the help of a 6 21 digit digital multimeter 共Kei- positive indicating the stress to be tensile and if ds ⬍ du, then
thley, model 2000兲 integrated with a personal computer ␧ is negative suggesting the generation of compressive stress.
through an IEEE-488 GPIB cable and a low thermal capacity Stress induced in the film changes the pattern by shifting the
furnace designed especially for the measurement. External peak positions. It can be seen from Fig. 2 that the peak po-
leads of materials same as those of the TFTC arms were sitions of thin films shift toward higher angle suggesting the
glued to the respective arms with the help of silver paste. The development of compressive strain. This strain is generated
thermoelectric output of all the thermocouples has been re- when two adjacent grains come in contact during their
corded by keeping the cold junction of the TFTCs at room growth. Such a strain is prominent, if the surface to volume
temperature 共27 ° C兲 with the help of water-cooled copper ratio of the grains in the film is large. Such compressive
blocks while the temperature of the hot junction was gradu- strain leads to a compact and dense film, however, excessive
ally raised up to 327 ° C. The temperature of the hot junction compressive strain developed in the film results in peel off. It
was measured by an electronic thermometer 共Omega Instru- is not out of place to mention that using the anodic arc depo-
ments, model HH506R兲 coupled with standard type K ther- sition technique and suitably adjusting the process param-
mocouples, with a resolution of 0.1 ° C and sampling rate of eters and in turn the deposition rate, it would be possible to
1 s−1. Measurement uncertainties were minimized by averag- have a control over the film compactness. The optimum val-
ing 1000 readings and using standard precautions in handling ues of the process parameters have been chosen based on our
low voltage signals, such as shielding and correcting for the earlier report.32 Although annealing relieves off the strain
voltmeter offset. The complete measurement setup was con- developed in thin films 共Table I兲, they do not become free
trolled by an in-house developed LABVIEW 7.1 program. from strain. This is very well reflected by the x-ray peaks
shift.
The average crystallite sizes for the as deposited and
III. RESULTS AND DISCUSSIONS annealed films have been calculated by using the Scherrer’s
formula,32
A. Structural analysis
1. Crystallographic structure k␭
Dc = , 共3兲
B cos ␪B
X-ray diffraction patterns of as deposited and annealed
films are shown in Figs. 2共a兲–2共c兲. All patterns show cubic where Dc is the crystallite size, ␭ is the x-ray wavelength, B
symmetry with polycrystalline structure and are in agreement is the full width at half maximum of the diffraction peak, and
with the JCPDS data available for Cu, Ni, and Fe. The lattice ␪B is the Bragg diffraction angle. The as deposited films
parameters have been calculated using the relation exhibited relatively low crystallinity. Due to substantial de-
crease in crystallite size the peak width increases and a sig-
ao = d共h2 + k2 + l2兲1/2 , 共1兲
nature of some amorphous nature can be seen. As AVA offers
and are found to be comparable with their corresponding very high deposition rates 共5 – 40 nm/ s兲 the entire process of
bulk values 共Table I兲. film deposition lasts for few seconds or, hardly, a minute for
The strain 共␧兲 is calculated from the relation36 thicker films. In this short duration the condensing atoms do

TABLE I. Structural data 共lattice parameter, crystallite size, grain size, d-value, and residual strain兲 of as
deposited and annealed Cu, Ni, and Fe films. Reported data 关Cu 共Ref. 69兲, Ni 共Ref. 70兲, and Fe 共Ref. 71兲兴 is
incorporated for comparison.

Lattice Crystallite Grain size


parameter size 共Dc兲 共Dg兲
共ao兲 共XRD兲 共SEM兲 d value Residual
Film 共Å兲 共nm兲 共nm兲 共Å兲 strain 共␧兲

Cu As deposited 3.605 26⫾ 5 ¯ 2.0813 −0.0032


Annealed 3.614 45⫾ 9 110⫾ 40 2.0868 −0.0006
Ref. 69 3.615 ¯ ¯ 2.0880 ¯

Ni As deposited 3.507 16⫾ 3 ¯ 2.0251 −0.0044


Annealed 3.518 40⫾ 8 125⫾ 45 2.0310 −0.0015
Ref. 70 3.523 ¯ ¯ 2.0340 ¯

Fe As deposited 2.855 19⫾ 3 ¯ 2.0190 −0.0038


Annealed 2.861 38⫾ 7 70⫾ 20 2.0232 −0.0018
Ref. 71 2.866 ¯ ¯ 2.0268 ¯
113717-4 Mukherjee et al. J. Appl. Phys. 106, 113717 共2009兲

not get sufficient time for participating in proper crystallite Figure 3 shows the SEM images of annealed single layer
growth process resulting in formation of nanosized crystal- nanocrystalline films of Cu, Fe, and Ni. Surface morphology
lites. However, annealing favors growth of crystallites to 关Figs. 3共d兲–3共f兲兴 of these films appears smooth and crack-
some extent 共Table I兲. free with nanosized grains in the range of 50– 150 nm. The
mean grain size 共Dg兲 is evaluated by the linear intercept
2. Surface morphology method.31 AVA offers flux ionization values as high as 30%
It is well known that the quality of the deposited thin that is advantageous in improving film properties. The grain
film resolutely depends on the deposition process used. In size or texture of the films and thus the smoothness of the
the case of low energy deposition, e.g., thermal evaporation, coated surface depends on the degree of the flux ionization;
the depositing atoms lack mobility and stick at the point of toward higher flux ionization the surface becomes smoother,
incidence on the surface of the growing film resulting in possibly due to a change in the mobility of absorbed atoms.30
films of poor density filled with voids and impurity
materials.37 The structure of the deposited film can drasti- As deposited (D = 26 nm; R = 0.52)
4.0 c c (a)
cally be affected by bombardment with energetic particles Annealed (D
c
= 45 nm; R
c
= 0.15)

from independent ion sources or from adjacent plasma dur- 3.5

ing film growth. These independent ion sources often use a


3.0 p = 0.01
process gas. In most of the cases the energetic ions used for
the bombardment of the growing films are, at least in part, 2.5

different from the coating material and, in fact, represent p = 1

impurity atoms that get incorporated into the deposited film 2.0

leading to undesired film properties such as brittleness.37 An 1.5


p = 0.01

ideal approach to avoid these difficulties is generation of


plasma containing solely the coating material. It must be 1.0
p = 1

stressed that this approach requires two simultaneous condi-


tions; first, to evaporate the coating material above a certain 3.2
(b)
p = 0.01
rate and, second, to produce and maintain a discharge solely
in the vaporized coating material. Moreover, plasma assisted 2.8

deposition techniques offer the advantage of adjusting the


ρ f / ρo

p = 1

crystallographic structures of the coatings to functional re- 2.4

As deposited (D = 16 nm; R = 0.64)


c c

quirements within certain limits and, generally, deposition Annealed (D = 40 nm; R = 0.51)
c c
2.0
from the plasma phase is known to produce coatings of high
quality. AVA deposition technique satisfies the above condi- p = 0.01

tions, i.e., firstly, no additional working gas 共such as argon兲


1.6

is necessary to drive the discharge and, secondly, the evapo-


ration rate 共and the deposition rate兲 is rather high.
1.2 p = 1

As deposited (D = 19 nm; R = 0.44)


2.8 c c (c)
(a) (d)
Annealed (D = 38 nm; R = 0.35)
c c

2.4 p = 0.01

2.0 p = 1

(b) (e) p = 0.01


1.6

p = 1
1.2

200 250 300 350

Thickness (nm)
(c) (f)
FIG. 4. 共Color online兲 Normalized resistivity 共␳ f / ␳o兲 of as deposited and
annealed 共a兲 Cu, 共b兲 Ni, and 共c兲 Fe films as a function of thickness. Symbols
in the upper section of individual graphs 共showing higher ␳ f / ␳o values兲
represent experimentally obtained normalized resistivity values of various
as-deposited films, whereas the similar symbols in the lower section of
individual graphs 共showing lower ␳ f / ␳o values are their respective normal-
ized resistivity for the annealed case. Experimental data are theoretically
fitted using the combined model of FS and MS 共taking p = 0.01 and 1兲. As
FIG. 3. FE-SEM micrographs of annealed thin films. First column 关共a兲–共c兲兴 the crystallite sizes of as deposited Cu and Ni films are less than their mean
shows cross-sectional views of 共a兲 Cu, 共b兲 Ni, and 共c兲 Fe, while the second free path 共␭o兲, the ␭o values used in Eq. 共4兲 for Cu and Ni are taken equal to
column 关共d兲–共f兲兴 shows the surface images of 共d兲 Cu, 共e兲 Ni, and 共f兲 Fe. their crystallite sizes.
113717-5 Mukherjee et al. J. Appl. Phys. 106, 113717 共2009兲

TABLE II. Electrical data 关resistivity 共␳兲, film temperature coefficient of resistance 共␤兲, thermoelectric power 共S兲, and mean free path at room temperature
共␭o兲兴 of bulk and annealed Cu, Ni, and Fe films. The crystallite boundary reflection coefficient 共Rc兲 and crystallite boundary limited mean free path 共␭c兲 of the
respective films are determined Eqs. 共4兲 and 共23兲.

␳ ob ␳f ␤ ob ␤f
E Fa ⫻10−8 ⫻10−8 ⫻10−3 ⫻10−3 S oc Sfd ␭ oe ␭c
Sample 共eV兲 共⍀ m兲 共⍀ m兲 共 ° C兲−1 共 ° C兲−1 共␮V / ° C兲 共␮V / ° C兲 共nm兲 Rc 共nm兲

Cu 7 1.67 2.13⫾ 0.31 3.93 3.12⫾ 0.19 1.85 1.74 39 0.15 30.6⫾ 4.4
Ni 4.91f 7.8 11.27⫾ 1.32 6.00 4.14⫾ 0.21 −19.24 17.35 10.9 0.51 7.5⫾ 0.9
Fe 11.1 10 14.68⫾ 1.46 5.00 3.81⫾ 0.18 16.20 14.88 19.6 0.35 13.4⫾ 1.3
a
Reference 42.
b
Reference 44.
c
References 45 and 46.
d
Reference 47.
e
References 32 and 38.
f
Reference 43.

AVA produces high energy depositing particles due to which is to enhance the diffused scattering within the crystallite
enhanced number of nucleation sites are produced leading to boundaries when the size of the crystallites is comparable to
smaller sized grains.31 However, these grains were observed mean free path of the conduction electrons. Upon annealing
to grow in size upon annealing at 300 ° C for 4 h 共figures not of Cu, Ni and Fe films, their respective resistivity values
shown for brevity兲. Annealing causes mobility of smaller reduced to around half of that obtained for as deposited
grains, which then coalesce with similar sized or with larger samples because of increase in crystallite size and decrease
grains. This decreases the grain boundary volume, thereby in defect density 共see Fig. 4兲. Decrease in electrical resistiv-
reducing the total energy of the film.38,39 Our results are in ity upon annealing of films appears to be due to the migra-
agreement with Ruan et al.,40 who have studied the mesos- tion of vacancies to the surface or to the sinks such as dis-
cale structures of electrodeposited nanocrystalline Ni–W al- locations or stacking faults.39 Crystallite growth occurs, and
loys and reported that micron sized grains 共what they have the disappearance of much of the crystallite boundary area is
termed as grain “colonies”兲 actually consist of few nano- probably the main cause of the decrease in resistance. The
meter sized crystallites. temperature at which the minimum in resistivity occurs co-
No columnar structure is observed as can be seen from incides with the Debye temperature of that metal, so that the
Figs. 3共a兲–3共c兲 that show cross-sectional views of the films. largest part of the lattice defects is removed at a temperature
Hence, it is clear that the films are dense and compact that at which the lattice vibrations have the maximum
makes them chemically more stable and resistant against cor- frequency.39
rosion as the chemical attack is mostly along the boundaries The observed resistivity values for Cu, Ni and Fe films
of the columnar structure. On the other hand films obtained are evaluated using Fuchs–Sondheimer 共FS兲 model for
by usual deposition processes, i.e., thermal evaporation and electron-surface scattering and Mayadas–Shatzkes 共MS兲
sputtering often show columnar grain structure. For many model for electron-crystallite boundary scattering. A simple
applications this columnar structure causes other undesired form combined of the approximate equations for both mod-
properties as well such as poor tensile strengths, optical els 共FS and MS兲 reads as32
properties, and barrier properties.30 AVA that is essentially an
ion-assisted deposition offers enhanced ion flux of high en-
ergy particles that improves the density and compactness of

␳ f = ␳o 1 +
3␭o
8t
共1 − p兲 +
共1.5兲␭oRc
Dc共1 − Rc兲
, 冎 共4兲

deposited films.31
where ␭o is the mean free path of electron, t is the thickness
of the film, and p is the specularity parameter ranging from 0
B. Thermoelectrical properties of individual metal
to 1 and is a measure of the size effect deviation from the
films
bulk behavior. p = 1 signifies perfect specular scattering that
Figure 4 shows that the resistivities of as deposited means all the conduction electrons are scattered elastically
samples of Cu, Ni and Fe are much higher than their bulk from both the film surface and the film-substrate interface
values. Due to the fast deposition rate that is inherent to with reversal of the velocity component, and p = 0 stands for
AVA, a large concentration of imperfections such as point completely diffuse scattering, that is, all the electrons are
defects 共mainly, vacancies兲, line defects 共mainly, disloca- scattered diffusely with complete loss of their drift velocity.39
tions兲 get frozen-in during the process of nucleation and Rc is the crystallite boundary reflection coefficient that takes
growth of thin films. The concentration of defects is expected on values between 0 and 1. It is the fraction of electrons that
to be determined by the size of the crystallites and the den- are scattered coherently at the crystallite boundaries and is
sity of defects at the crystallite boundaries. The crystallite also termed as intrinsic scattering. Considering specular re-
sizes of as deposited samples have been estimated and were flection coefficient 共p兲 equal to 0.01,32 the most reasonable
found to be comparable to the mean free path 共see Table II兲 fit to the experimental resistivity data for as deposited films
of conduction electrons in the respective materials. The ef- of Cu, Ni, and Fe could be obtained at Rc equal to 0.52, 0.64,
fect of a large density of defects at the crystallite boundaries and 0.44, respectively, while for annealed case the values of
113717-6 Mukherjee et al. J. Appl. Phys. 106, 113717 共2009兲

Rc are found to be 0.15, 0.51, and 0.35, respectively. In the 2.8


(a) p = 1
0.9

present investigation the thickness of films has been main- 2.6 0.8

tained much larger than the mean free path of the conduction 2.4
0.7
p = 0.01
electrons, and hence the extrinsic scattering at the film sur- 2.2
(2)

face is assumed to be negligible. Thus, the approximate re- 0.6

2.0

lation based only on intrinsic scattering due to crystallite 0.5

boundaries 共for the case, p = 1兲 has been used to explain en- 1.8

0.4

hanced values of transport parameters. As indicated by Fig. 1.6 (1)


p = 0.01
0.3
4, the approximate relation for the case p = 1 is in agreement 1.4

with the values of the combined model in higher thickness 1.2


0.2

p = 1
domain and also to the measured resistivity for the same p 1.0 0.1

and Rc. Temperature coefficients of resistance 共␤ f 兲 of an-


nealed Cu, Ni and Fe films are in agreement with the theo- 2.8
(b)
0.9

retical plots generated using the combined model for the 2.6
p = 1 0.8

same values of Rc 共see Fig. 5兲. Assuming the crystallite size, 2.4
0.7

p, and Rc do not change with temperature, the approximate

ρf / ρ o
2.2

relation for ␤ f based on combined model can be expressed

βf / β o
p = 0.01 0.6
(2)
2.0

as41 0.5

1.8
(1)
p = 0.01 0.4

再 冎
1.6

0.3

3␭o 共1.5兲␭oRc −1 1.4

␤ f = ␤o 1+ 共1 − p兲 + . 共5兲 0.2

Dc共1 − Rc兲
1.2 p = 1
8t
1.0 0.1

2.8 0.9
(c)
Thus, we can conclude that the films are structurally similar
2.6 0.8
to their bulk. Here it is noteworthy that several researchers fit p = 1

their experimental resistivity data in terms of grain size48 2.4


0.7

determined using SEM or atomic force microscopy, while 2.2


p = 0.01 0.6

others use crystallite boundaries32 determined using XRD. It 2.0


(2)

0.5
is very difficult to distinguish the size effects due to grain 1.8
(1)
boundaries and crystallite boundaries in the same film. Both 1.6
p = 0.01 0.4

these boundaries can be mathematically identified with scat- 0.3

tering potentials having certain height 共H兲 and width 共w兲


1.4

关creating a ␦-function potential S␦共x − xn兲, xn being its posi-


p = 1 0.2
1.2

tion and S being the “strength” of the potential兴. Thus, both 1.0

200 250 300 350


0.1

can be represented as N parallel partially reflecting planes,


Thickness (nm)
oriented perpendicular to the direction of a constant electric
field E 共MS theory兲.51 In support, taking average grain size FIG. 5. 共Color online兲 Normalized resistivity 共␳ f / ␳o兲 and normalized tem-
共Dg兲 in place of crystallite size 共Dc兲, bulk mean free path perature coefficient of resistance 共␤ f / ␤o兲 of annealed 共a兲 Cu, 共b兲 Ni, and 共c兲
values for Cu, Ni and Fe from Table II, and setting p = 0.01, Fe films as a function of thickness. Symbols in the lower section of indi-
the theoretical curve based on combined model 关Eqs. 共4兲 and vidual graphs 共having corresponding scale on left hand side兲 represent ex-
perimentally obtained normalized resistivity values of various annealed
共5兲兴 is plotted and found to agree with the measured ␳ f and films, whereas similar symbols in the upper section of individual graphs
␤ f at Rg = 0.30, 0.77, and 0.49 for annealed films of Cu, Ni, 共having corresponding scale on right hand side兲 are their respective normal-
and Fe, respectively 共Fig. 5兲. Rg is the grain boundary reflec- ized temperature coefficient of resistance values. Theoretical plots for ␳ f / ␳o
tion coefficient that takes on values between 0 and 1. It is the and ␤ f / ␤o, based on combined model of FS and MS 共taking p = 0.01 and 1兲
are generated using the same parameters as used in Fig. 4. The dots 共쎲兲 and
fraction of electrons that are scattered coherently at the grain the solid squares 共䊏兲 are the fitted ␳ f / ␳o values and ␤ f / ␤o values obtained
boundaries. by solving Eqs. 共4兲 and 共5兲 for the average grain size 共Dg兲 instead of average
Thermoelectric powers of annealed films of Cu, Fe, and crystallite size 共Dc兲. The curves fit the experimental data for Rg = 0.30, 0.77,
and 0.49 for annealed films of Cu, Ni, and Fe, respectively.
Ni are determined relative to their bulk wire and are found to
be less than the corresponding bulk values 共Table II兲. Among
various transport parameters, the thermoelectric power 共S兲 is
most sensitive to changes in the structure and composition of
a metal when they cause distortions at the Fermi surface,
So = −
3e dE

␲2k2T d ln ␴o
冊 E=EF
, 共6兲

particularly where the Fermi surface touches the Brillouin


zone, and the energy dependence of the mean free path at the
Fermi surface. According to Drude–Lorentz–Sommerfield- where EF is the Fermi energy, T the absolute temperature, e
free-electron model, the thermoelectric power of a bulk the absolute magnitude of the electronic charge, k the Bolt-
metal, So 共neglecting phonon drag contributions and assum- zmann’s constant, and ␴o the bulk conductivity given by
ing a spherical Fermi surface兲, is given by49
113717-7 Mukherjee et al. J. Appl. Phys. 106, 113717 共2009兲

␴o =
2 ␲ e 2␭ o
12␲3h
A, 共7兲
3
G共␣兲 = − ␣ + 6␣2 +
2
3␣3
1+␣
− 9␣3 ln 1 +
1

冉 冊
. 共15兲

where ␭o is the average mean free path of the electron and A For a thick polycrystalline film, its temperature coefficient is
is the Fermi surface area. In terms of the resistivity of a pure given by52
bulk metal 共␳o兲, its thermoelectric power can be expressed as
冋 G共␣兲

冉 冊
␤ f = ␤o 1 + , 共16兲
␲ k T d ln ␳o
2 2 C共␣兲
So = − . 共8兲
3eEF d ln E E=EF where ␤ f is the temperature coefficient of thick polycrystal-
line film, while ␤o is that of bulk metal. Substituting Eq. 共16兲
Substituting Eq. 共7兲 in Eq. 共6兲 and solving, one gets in Eq. 共12兲, ⌬S f can be expressed as

So = −
␲ 2k 2T
3eEF
共U + V兲, 共9兲 ⌬S f = −
3eEF
冉 冊
␲ 2k 2T ␤ f
U
␤o
−1 . 共17兲

where U and V are log derivatives of the mean free path of Since ⌬S f and ␤ f are measurable quantities, the values of U
conduction electrons at the Fermi surface and the Fermi sur- can be obtained by using Eq. 共17兲. The change in thermo-
face area, respectively, with respect to energy. They are given electric power of the thin films ⌬S f is measured with respect
by to their corresponding bulk sample. The measured values of
⌬S f and ␤ f are given in Table II. The values of U are calcu-
U= 冉 d ln ␭o
d ln E
冊 E=EF
共10兲
lated from Eq. 共17兲 and listed in Table III. Using the values
obtained for U and the standard thermoelectric power values
共Table II兲 reported for the pure metal, the log derivative of
and the Fermi surface area with respect to energy 共V兲 is calcu-
lated from Eq. 共9兲. The results obtained for the thin films
V= 冉 冊 d ln A
d ln E E=EF
. 共11兲
show a wide scatter of values. The value of U for Cu ob-
tained in this work is close to the values reported by Leonard
et al.59 however, it disagrees with some other reports.35,55–58
It is reported that surface scattering,21,35 grain boundary The results obtained by Gouault,56 Angus et al.,57 and Sug-
scattering,49 defects, and impurities50 greatly influence the awara et al.58 obtained positive values for U on unannealed
conduction process of thin films which, in turn, are the con- or insufficiently annealed samples. The value obtained by
sequences of the growth mechanism. Tellier and Tosser de- Chopra et al.35 might be due to enhanced size effects. The
duced a general expression for the thermoelectric power of observed negative value of V in Cu suggests that the area of
polycrystalline metallic films using the Mayadas and Shatz- the Fermi surface decreases for increasing energy.60 In accor-
kes conduction model.49 This model takes into account both dance with free electron model, the negative value of V sug-
the isotropic background and the crystallite boundary scatter- gests that the transport carriers are holes. However, the elec-
ing of electrons. Polycrystalline films are finely grained trical property of the metals strongly suggests that the
structures whose transport parameters are strongly domi- transport carriers are electrons. This discrepancy may be due
nated by crystallite boundary scattering.51 Mayadas and to the presence of higher concentration of crystallite bound-
Shatzkes conduction model uses a crystallite boundary re- aries and other defects, which act as scattering center for the
flection coefficient Rc 共between 0 and 1兲 to account the scat- electrons.60 This effect is prominent in the case of nanosized
tering at the crystallite boundaries of thin films having an crystallites as the number of crystallite boundaries is higher
average crystallite size 共Dc兲. Neglecting contributions of per unit volume. Due to presence of higher concentration of
electronic scattering at the surface and considering only crys- defects and crystallite boundaries, the Fermi surface is highly
tallite boundary scattering effects, the difference in thermo- distorted in nanocrystalline films, particularly at the places
electric power between a polycrystalline film 共SF兲 and a bulk where the Fermi surface touches the Brillouin zones. This
metal 共So兲 is given by49 distortion leads to different relaxation times for the neck re-
gion and the belly region of the Fermi surface. As a result,

⌬S f = SF − So = −
␲ 2k 2T
3eEF
U 冋 册
G共␣兲
F共␣兲
, 共12兲
the electron-electron scattering in the neck region is different
from that in the belly region, which ultimately affects the
mean free path of the conduction electrons. The energy de-
where pendence of the mean free path 共U兲 or the relaxation time at
the Fermi surface will change with the concentration of de-
␭o Rc fects. The changes in thermoelectric power may be due to
␣= , 共13兲 changes in energy dependence of mean free path. Thakoor et
D 1 − Rc
al.61 have shown that the energy dependence of conductivity

冉 冊
and, hence, the energy dependence of the mean free path or
3 1 relaxation time decreases exponentially with annealing tem-
F共␣兲 = 1 − ␣ + 3␣2 − 3␣3 ln 1 + , 共14兲
2 ␣ perature for Cu film of thickness of 400 nm assuming that
113717-8 Mukherjee et al. J. Appl. Phys. 106, 113717 共2009兲

TABLE III. Energy dependence of the mean free path of electrons at the Fermi surface and the Fermi surface
area for annealed Cu, Ni, and Fe films at 27 ° C.

Thickness ⌬S f
Film 共nm兲 ␤ f / ␤o 共␮V / ° C兲 U V References
a
Cu 300⫾ 50 0.794⫾ 0.061 −0.11⫾ 0.04 −0.51⫾ 0.22 −1.25⫾ 0.22 Present work
¯ ¯ ¯ 0 ¯ Reimerb
¯ ¯ ¯ 2.13 −3.53 Gouaultc
¯ ¯ ¯ −18.7 ¯ Chopra et al.d
¯ ¯ ¯ 2.3 −3.7 Angus et al.e
¯ ¯ ¯ 0.7 −2.1 Sugawara et al.f
240–85 0.868–0.793 0.0285–0.0430 −0.21 −1.43 Leonard et al.g

Nih 300⫾ 50 0.69⫾ 0.035 1.49⫾ 0.21 3.22⫾ 0.62 ¯ Present work
¯ ¯ ¯ 19 −5.8 Wedler et al.i

Feh 300⫾ 50 0.762⫾ 0.036 −1.32⫾ 0.17 −8.39⫾ 1.48 ¯ Present work
25–500 ¯ ¯ −49.4 ¯ Scarioni et al.j
a
Reference 53.
b
Reference 55.
c
Reference 56.
d
Reference 35.
e
Reference 57.
f
Reference 58.
g
Reference 59.
h
Reference 54.
i
Reference 64.
j
Reference 21.

carrier concentration and the effective mass remains con- theoretically expressed by Eq. 共17兲. The energy dependence
stant. of Fermi surface area 共V兲 will be determined accurately us-
In the case of Ni and Fe, the values of U are quite dif- ing Eq. 共18兲 and not by Eq. 共9兲 共as in case of noble metals
ferent from the one predicted by the free electron model such as Cu兲. The values of V for Ni and Fe are not tabulated
共U = 2兲. Ni and Fe being transition metals possess two over- 共Table III兲 due to the absence of data for corresponding den-
lapping bands at the Fermi level 共s and d兲. s-band of the free sity of states 关Nd共E兲兴 and its energy dependence
atoms is broad and roughly free electron like while the 关dNd共E兲 / dE兴. In the present study the values of U for Ni and
d-band is narrow and has a high effective mass and density Fe are closer to that of free electron model but differ sub-
of states at the Fermi level. Due to lesser effective mass the stantially from other reported values.64,21 Wedler et al.64 de-
transport mechanism is dominated by s-electrons. The high termined V on the basis of experimental data based on the
density of states at the Fermi level of the unfilled d-band has thermoelectric size effect and determined U using Eq. 共9兲
a profound effect on the scattering probability of the without considering the effects induced by the d-band. This
s-electrons. Thus, s-d scattering will predominate over s-s might be the reason for getting such a high value of U.64 The
scattering. For transition metals the relaxation time is in- anomalously high value of U for Fe obtained by Scarioni et
versely proportional to the density of d-states 关Nd共E兲兴. Since al.21 may be due to high concentration of defects leading to
Nd共E兲 is large, the relaxation time for the s-electrons are enhanced size effects.
correspondingly small leading to comparatively high resis- In order to account for enhanced resistivity and less ther-
tivities. Moreover, in transition metals, Fermi level lies at a moelectric power of nanocrystalline films relative to their
position in the density of states versus energy curve where bulk counterparts, an attempt is made to estimate the mean
Nd共E兲 changes rapidly with energy, giving rise to the large free path of the films. According to Matthiessen’s rule, the
thermoelectric powers of either sign. Thus, the complete ex- resistivity of a metal film can be written as
pression for thermoelectric power of transition metals62,63 is

So = −
3e 再
␲ 2k 2T 1
EF
共U + V兲 − 冋
1 dNd共E兲
Nd共E兲 dE
册 E=EF
冎 . ␳ f = 兺 ␳ j = ␳o + ␳i + ␳s + ␳c ,
j
共19兲

共18兲

The density of states 关Nd共E兲兴 and its energy dependence where the subscripts 共o, i, s, and c兲 refer to the contributions
关dNd共E兲 / dE兴 at the Fermi level are material dependent and, due to the lattice, imperfections 共dislocations, vacancies,
thus, assumed to be nearly the same for bulk and thin films etc兲, surface scattering, and crystallite boundary scattering,
共the s-d scattering is same for bulk and its thin film in a respectively. Since the thermoelectric power is proportional
particular metal兲. Based on the above assumption, the change to the energy derivative of the conductivity, the assumption
in thermoelectric power 共⌬S f 兲 of transition metals can be of Matthiessen’s rule yields59,61
113717-9 Mukherjee et al. J. Appl. Phys. 106, 113717 共2009兲

冒 S o␳ o S i␳ i S s␳ s S c␳ c 7 Cu - Ni
S f = 兺 ␳ jS j
j
兺 ␳j =
j ␳f
+
␳f
+
␳f
+
␳f
共20兲 6 Standard
45
Wire thermocouple
5 TFTC
for the thermoelectric power of a thin film. For film thickness
4
beyond 250 nm, the surface scattering can be neglected.
Also, for properly annealed sample, ␳i is reduced to a negli- 3

Thermocouple output (mV)


gible value. Assuming ␳i to be zero for annealed case, the 2
contribution of imperfections on S f reduces to zero. Thus, for
1
sufficiently thick and properly annealed films, ␳ f and S f takes (a)
the form as 0

␳ f = ␳o + ␳g 共21兲 10 Fe - Ni
72
Theoretical
and 8 Wire thermocouple
TFTC
S o␳ o S c␳ c
Sf = + . 共22兲 6
␳f ␳f
4
These observations reflect that the film resistivity and ther-
moelectric power are limited only by the crystallite bound- 2
aries. Using the resistivity values of thick and properly an-
(b)
nealed resistivity values of thin films and that of bulk, and 0
0 50 100 150 200 250 300
the bulk mean free path 共Table II兲, the crystallite boundary ο
∆Τ( C)
limited mean free path 共␭c兲 may be estimated from the
relation51 FIG. 6. 共Color online兲 Temperature dependence of thermoelectric output of

冉冊
共a兲 Cu–Ni TFTCs and 共b兲 Fe–Ni TFTCs.
␳o
␭c ⬇ ␭o . 共23兲
␳f
terial comprises of two components: electronic thermoelec-
The values of ␭c for the films are shown in Table II. ␭c is tric power 共So for bulk and S f for thin films兲, arising from the
found to be less than ␭o. Since the crystallite size is compa- nonequilibrium distribution of the conduction electrons and
rable to the mean free path, U cannot decrease with energy as phonon drag thermoelectric power 共Sph兲 共a conduction elec-
it would have done in the case of bulk material. This leads to tron contribution induced by nonequilibrium effects in the
a smaller thermoelectric power of films in comparison to phonon system兲. So or S f varies linearly with temperature. It
their bulk value resulting in reduced thermoelectric output. may be the phonon drag component 共Sph兲 which induces
nonlinear effects in the thermoelectric response of a material
C. Performance of TFTCs with increasing temperature. The change in slope at different
temperatures depends on the relative magnitudes of the
Figure 6 shows a plot of the thermoelectric output versus phonon-defect, phonon crystallite boundary, phonon-phonon
the temperature difference 共⌬T兲 between the hot and cold and electron-phonon scattering process.67,68 Sph has the same
junctions of annealed 共a兲 Cu–Ni and 共b兲 Fe–Ni TFTCs. Each sign as S f for normal electron-phonon processes, and the
point at a particular temperature difference is an average of opposite sign for the Umklapp electron-phonon processes.
16 similar samples prepared in eight different batches. For Thus, the total sign of the phonon drag contribution is con-
the sake of comparison, the experimentally determined ther- ditioned by the form of the Fermi surface, which governs
moelectric outputs of wire samples are shown in their respec-
tive figures. The thermoelectric outputs are fitted as per the
thermocouple equation, TABLE IV. Thermoelectric power of Cu–Ni and Fe–Ni TFTCs.

E = b⌬T + c共⌬T兲2 , 共24兲 S or ba c ⌬S f a


Thermocouple 共mV/°C兲 共mV/ ° C2兲 共mV/°C兲 References
where b and c are the thermoelectric power and the first
nonlinear coefficient of the respective thermocouples. These Cu–Ni wire 0.0193共0.0224兲 1.044⫻ 10−5 ¯ Present work
values are summarized in Table IV. The thermoelectric out- Cu–Ni wire 0.025 ¯ ¯ Marshall et al.b
put 共determined from linear fit data兲 of the TFTCs are in Cu–Ni TFTC 0.0178共0.0206兲 8.586⫻ 10−6 0.0015 Present work
共0.0018兲
agreement with those reported by Bullis65 and Marshall et
Cu–Ni TFTC 0.0023 ¯ ¯ Bullisc
al.,5 and are equally comparable with our respective bulk Cu–Ni TFTC 0.0023 ¯ 0.002 Marshall et al.b
values. The slight decrease in thermoelectric power of these Fe–Ni wire 0.0308共0.0283兲 −8.406⫻ 10−6 ¯ Present work
TFTCs relative to their corresponding bulk response is due to Fe–Ni wire 0.028 ¯ ¯ Marshall et al.b
the reduced thermoelectric power 共S f 兲 values of their indi- Fe–Ni TFTC 0.0279 −9.434⫻ 10−6 0.0029 Present work
vidual layers. It can be seen from Fig. 6 that with the in- Fe–Ni TFTC 0.027 ¯ 0.001 Marshall et al.b
crease in temperature, the thermoelectric output presents a a
Reference 66.
nonlinear behavior. Cu–Ni TFTCs are slightly more nonlin- b
Reference 5.
c
ear than their bulk. The total thermoelectric power of a ma- Reference 65.
113717-10 Mukherjee et al. J. Appl. Phys. 106, 113717 共2009兲

1.0
Cu -Ni TFTC the crystallite boundaries decreases appreciably, lowering the
annealed
defect induced thermoelectric output. The increase in ther-
0.8 as deposited

moelectric output of Cu–Ni and Fe–Ni TFTCs on annealing


is an indicative that the value of dislocation induced thermo-
0.6

electric power is more in Ni than in Cu and Fe.


Thermocouple Output (mV)

0.4

0.2
IV. CONCLUSION
(a)
0.0
共1兲 Good quality nanocrystalline Cu–Ni and Fe–Ni TFTCs
1.4
Fe-Ni TFTC
are developed using AVA. This technique offers advan-
annealed
1.2
as deposited tage of high throughput, less critical operation, and low
1.0
system cost.
共2兲 Fast deposition rate does not facilitate the long distance
0.8
diffusion of atoms on the film surface resulting in the
0.6
formation of nanosized crystallites.
0.4 共3兲 As deposited films with thickness ⬎250 nm deposited
0.2
using AVA are in compressive stress. Annealing up to
(b)
300 ° C for 4 h was found to relieve the films reducing
0.0

0 10 20 30 40 50
the strain to a low value.
Δ Τ (
o
C)
共4兲 The resistivity 共␳ f 兲, temperature coefficient of resistance
共␤ f 兲, and thermoelectric power 共S f 兲 of individual layer
FIG. 7. 共Color online兲 Temperature dependence of thermoelectric output of films have been studied extensively. As deposited films
as developed and annealed 共a兲 Cu–Ni TFTCs and 共b兲 Fe–Ni TFTCs.
show enhanced size effects due to the presence of
smaller crystallites and large concentration of frozen-in
whether the normal or the Umklapp process dominate.62 structural defects. Annealing causes considerable reduc-
Nanocrystalline thin films produced using AVA possess more tion in the size effects in these transport parameters of as
defect density and crystallite boundaries than bulk. This deposited films. The crystallite boundary reflection coef-
leads to a reduced phonon mean free path. As such, these ficient 共Rc兲 reduces from 0.52 to 0.15 for Cu, 0.64 to
scattering processes, particularly, the phonon crystallite 0.51 for Ni, and 0.44 to 0.35 for Fe.
boundary scattering become more pronounced in the entire 共5兲 Using intrinsic size effects in electronic transport theory
temperature range in nanocrystalline thin films resulting in a the difference in thermoelectric power 共⌬S f 兲 in film and
relatively more nonlinear thermoelectric response. bulk is related to normalized temperature coefficient of
The standard deviation in thermocouple outputs of resistance 共␤ f / ␤0兲 and the energy dependence of mean
Cu–Ni were 0.0631 mV at 100 ° C and 0.151 mV at 300 ° C. free path of conduction electrons. The values of U are
Reproducibility figures for 16 test samples of Cu–Ni TFTCs obtained and found comparable with the results of Le-
ranges around ⫾3 ° C at 100 ° C and ⫾6 ° C at 300 ° C. onard et al.59 for Cu and found closer to free electron
Also, the standard deviations in thermocouple output of theory based values for Fe and Ni than other reports.
Fe–Ni was 0.114 mV at 100 ° C and 0.2845 mV at 300 ° C. This departure from U values suggests strongly that
This shows that for the 16 test samples, reproducibility figure large concentration of crystallite boundaries causes dis-
for Fe–Ni TFTCs ranges around ⫾3 ° C at 100 ° C and tortion of Fermi surface and thereby reduces the thermo-
⫾8.5 ° C at 300 ° C. electric power of thin films by the randomizing effect of
For better understanding of the deposition technique, enhanced scattering.
four as-developed TFTCs were characterized with respect to 共6兲 Thermoelectric responses of Cu–Ni and Fe–Ni TFTCs
thermoelectric output upto ⌬T = 50 ° C and compared to that agree well with their wire thermocouple equivalents and
of annealed ones. The comparison is shown in Fig. 7. In the are fairly reproducible. The values of thermoelectric
temperature range studied the thermoelectric output of as de- power of 0.0178 and 0.0279 mV/ ° C at 300 ° C were
veloped TFTCs is lower than that of annealed ones. This is obtained for Cu–Ni and Fe–Ni TFTCs, respectively.
because, as deposited films possess greater concentration of However, the nonlinearity figures show a change for
structural defects such as dislocations, vacancies, etc., in- these TFTCs as compared to the corresponding wire
duced due to fast deposition process. Dislocations raise the thermocouples. The change is from 1.044
thermoelectric power of noble metals such as Cu, whereas ⫻ 10−5 to 8.586⫻ 10−6 mV/ ° C2 for Cu–Ni and from
vacancies lower the thermoelectric power. For Ni, the sign of −8.406⫻ 10−6 to − 9.434⫻ 10−6 mV/ ° C2 for Fe–Ni
the dislocation induced thermoelectric effect is positive, i.e., TFTCs. We propose that the observed nonlinearity is
the dislocations reduce the negative absolute thermoelectric due to enhanced phonon crystallite boundary scattering.
power of Ni.68 The interface between amorphous and nano- Moreover, annealing slightly increases the thermoelec-
crystalline phases strongly influences the thermoelectric tric response of Cu–Ni and Fe–Ni TFTCs that indicates
power of the films and, hence, lowers thermoelectric output reduction in defects, dislocations, and crystallite bound-
of as deposited pairs. On annealing the density of defects at aries in the films.
113717-11 Mukherjee et al. J. Appl. Phys. 106, 113717 共2009兲

A 8, 2160 共1990兲.
32
ACKNOWLEDGMENTS M. K. Sinha, S. K. Mukherjee, B. Pathak, R. K. Paul, and P. K. Barhai,
Thin Solid Films 515, 1753 共2006兲.
The authors wish to express their sincere thanks to the 33
M. Mausbach, H. Ehrich, and K. G. Muller, J. Vac. Sci. Technol. B 11,
administration of Birla Institute of Technology, Ranchi for its 1909 共1993兲.
34
continued support and financial assistance. The authors H. Ehrich, B. Hasse, K. G. Muller, and R. Schmidt, J. Vac. Sci. Technol. A
6, 2499 共1988兲.
gratefully acknowledge the support of Mr. Rajkishore Pra- 35
K. L. Chopra, S. K. Bahl, and R. Randlett, J. Appl. Phys. 39, 1525 共1968兲.
manik in fabrication and associated experimentation. One of 36
B. D. Cullity, Elements of X-ray Diffraction, 2nd ed. 共Addison-Wiley,
the authors 共S.K.M.兲 is thankful to Dr. R. L. Boxman, Direc- London, 1978兲.
37
tor, Plasma Division, Tel Aviv University, Israel for fruitful G. Musa, H. Ehrich, and M. Mausbach, J. Vac. Sci. Technol. A 12, 2887
共1994兲.
discussions, and to CSIR, New Delhi, India for the award of 38
L. I. Maissel and R. Glang, Handbook of Thin Film Technology 共McGraw,
Senior Research Fellowship. New York, 1970兲.
39
K. L. Chopra, Thin Film Phenomena 共Robert E. Krieger, Huntington,
1
J. F. Lei and H. A. Will, Sens. Actuators, A 65, 187 共1998兲. 1979兲.
2 40
H. M. Tong, G. Arjavalingam, R. D. Haynes, G. N. Hyer, and J. J. Ritsko, S. Ruan and C. A. Schuh, Scr. Mater. 59, 1218 共2008兲.
Rev. Sci. Instrum. 58, 875 共1987兲. 41
R. Suri, A. P. Thakoor, and K. L. Chopra, J. Appl. Phys. 46, 2574 共1975兲.
3 42
M. Emi, Y. Suzuki, Y. Yamada, H. Ishii, S. Kimura, H. Ogawa, and N. W. Ashcroft and N. D. Mermin, Solid State Physics 共Saunders, Phila-
Y. Enomoto, JSAE Rev. 23, 379 共2002兲. delphia, 1976兲.
4
K. G. Kreider, J. Vac. Sci. Technol. A 4共6兲, 2618 共1986兲. 43
A. S. Karolik and V. I. Sharando, J. Phys.: Condens. Matter 17, 3567
5
R. Marshall, L. Atlas, and T. Putner, J. Sci. Instrum. 43, 161 共1966兲 共ref- 共2005兲.
44
erences therein兲. C. D. Hodgman, Handbook of Chemistry and Physics, 31st ed. 共Chemical
6
K. G. Kreider, J. Vac. Sci. Technol. A 11, 1401 共1993兲. Rubber, Cleveland, OH, 1949兲, pp. 1979–1988.
7
K. G. Kreider, Sens. Actuators, A 69, 46 共1998兲. 45
I. S. Grigoriev and E. Z. Meilikhov, Handbook of Physical Quantities
8
K. G. Kreider, Thin Solid Films 376, 32 共2000兲. 共CRC, Boca Raton, 1997兲, pp. 692–694.
9
E. Castano, E. Revuelto, M. C. Martin, A. Gracia-Alonso, and F. J. Gracia, 46
So values are with respect to Pb or comparison with other results.
Sens. Actuators, A 60, 65 共1997兲. 47
S f = So + ⌬S f .
10
J. R. Ho, C. C. Chen, and C. H. Wang, Sens. Actuators, A 111, 188 共2004兲. 48
J. M. Camacho and A. I. Oliva, Thin Solid Films 515, 1881 共2006兲 共ref-
11
B. Revaz, R. Flukiger, J. Carron, and M. Rappaz, Sens. Actuators, A 118, erences therein兲.
238 共2005兲. 49
C. R. Tellier and A. J. Tosser, Thin Solid Films 41, 161 共1977兲.
12
W. M. Posadowski, Thin Solid Films 459, 258 共2004兲. 50
K. L. Chopra and A. P. Thakoor, J. Appl. Phys. 49, 2855 共1978兲.
13
M. Imran and A. Bhattacharyya, Sens. Actuators, A 121, 306 共2005兲. 51
A. F. Mayadas and M. Shatzkes, Phys. Rev. 1, 1382 共1970兲.
14
T. H. Kim and S. J. Kim, J. Micromech. Microeng. 16, 2502 共2006兲. 52
C. R. Tellier and A. J. Tosser, Thin Solid Films 43, 261 共1977兲.
15
H. Choi and X. Li, Sens. Actuators, A 136, 118 共2007兲. 53
Considering the So value used by Leonard et al., 共Ref. 59兲 V = −1.11.
16
A. Basti, T. Obikawa, and J. Shinozuka, Int. J. Mach. Tools Manuf. 47, 54
Values of V for Ni and Fe are not tabulated due to the absence of data for
793 共2007兲.
17 corresponding density of states 关Nd共E兲兴 and its energy dependence
A. Datta, X. Cheng, M. A. Miller, and X. Li, Thin Solid Films 516, 4307
关dNd共E兲 / dE兴.
共2008兲. 55
18 L. Reimer, Z. Naturforsch. A 12, 525 共1957兲.
T. M. Berlicki, E. Murawski, M. Muszynski, S. J. Osadnik, and E. L. 56
J. Gouault, J. Phys. 共France兲 28, 931 共1967兲.
Prociow, Sens. Actuators, A 50, 183 共1995兲. 57
19 R. K. Angus and I. D. Dalgliesh, Phys. Lett. A 31, 280 共1970兲.
F. E. Kennedy, D. Frusescu, and J. Li, Wear 207, 46 共1997兲. 58
20 H. Sugawara, T. Nagano, K. Uozumi, and A. Kinbara, Thin Solid Films
D. Debey, R. Bluhm, N. Habets, and H. Kurz, Sens. Actuators, A 58, 179
14, 349 共1972兲.
共1997兲. 59
21 W. F. Leonard and H. Y. Yu, J. Appl. Phys. 44, 5320 共1973兲.
L. Scarioni and E. M. Castro, J. Appl. Phys. 87, 4337 共2000兲. 60
22
H. Zou, D. M. Rowe, and A. G. K. Williams, Thin Solid Films 408, 270 S. F. Lin and W. F. Leonard, J. Appl. Phys. 42, 3634 共1971兲.
61
共2002兲. A. P. Thakoor, R. Suri, S. K. Suri, and K. L. Chopra, J. Appl. Phys. 46,
23
G. Golan, A. Axelevitch, B. Sigalov, and B. Gorenstein, Microelectron. 4777 共1975兲.
62
Reliab. 43, 509 共2003兲. R. D. Barnard, Thermoelectricity in Metals and Alloys 共Taylor and Fran-
24
I. Bakonyi, E. T. Kadar, J. Toth, T. Becsei, T. Tarnoczi, and P. Kamasa, cis, London, 1972兲.
63
J. Phys.: Condens. Matter 11, 963 共1999兲. F. J. Blatt, P. A. Schroeder, C. L. Foiles, and D. Greig, Thermoelectric
25
R. G. Delatorre, M. L. Sartorelli, A. Q. Schervenski, A. A. Pasa, and S. Power of Metals 共Plenum, New York, 1976兲.
64
Güths, J. Appl. Phys. 93, 6154 共2003兲. G. Wedler, H. Reichenberger, and H. Wenzel, Z. Naturf. 26, 1444 共1971兲.
65
26
Y. G. Kim, A. H. Kang, K. S. Gam, J. C. Kim, and J. H. Kim, Meas. Sci. L. H. Bullis, J. Sci. Instrum. 40, 592 共1963兲.
66
Technol. 15, 1266 共2004兲. The values within brackets are linear fit data of S. The linear fit values of
27
A. Giani, A. A. Bayaz, A. Boulouz, F. P. Delannoy, A. Foucaran, and A. S resemble more closely the referred values 共Refs. 5 and 65兲. It might be
Boyer, Mater. Sci. Eng., B 95, 268 共2002兲. possible that Bullis 共Ref. 65兲 and Marshall et al. 共Ref. 5兲 have presented
28
A. Balducci, A. D’Amico, C. D. Natale, M. Marinelli, E. Milani, M. E. the TEP values based on linear fit 关S = a + b共⌬T兲兴.
67
Morgada, G. Pucella, G. Rodriguez, A. Tucciarone, and G. V. Rinati, Sens. K. K. Choudhary, N. Kaurav, N. Gupta, and D. Varshney, J. Phys.: Conf.
Actuators B 111–112, 102 共2005兲. Ser. 92, 012146 共2007兲.
29 68
H. D. Bhatt, R. Vedula, A. B. Desu, and G. C. Fralick, Thin Solid Films A. S. Karolik and V. I. Sharando, J. Phys.: Condens. Matter 17, 3567
350, 249 共1999兲. 共2005兲 共references therein兲.
30 69
H. Ehrich, B. Hasse, M. Mausbach, and K. G. Müller, IEEE Trans. Plasma JCPDS Card No. 4-0836.
70
Sci. 18, 895 共1990兲. JCPDS Card No. 4-0850.
31 71
H. Ehrich, B. Hasse, M. Mausbach, and K. G. Müller, J. Vac. Sci. Technol. JCPDS Card No. 6-0696.

Das könnte Ihnen auch gefallen