Non-Contact C-V Technique for high-k Applications.
Piotr Edelman, Alexandre Savtchouk, Marshall Wilson, John D' Amico,
Joseph N. Kochey, Dmitriy Marinskiy, and Jacek Lagowski
Semiconductor Diagnostics, Inc.; 3650 Spectrum Blvd. Ste. 130; Tampa, FL 33612 USA
Abstract. This non-contact high-k monitoring technique is based on a differential quasistatic C-V that is generated using timeresolved metrology combining corona charging and contact potential difference (CPD) measurements. The technique
incorporates transconductance corrections that enable measurements in the high field range (lOMV/cm) required for extraction of
large dielectric capacitance corresponding to ultra-low equivalent electrical oxide thickness (EOT) down to the sub-nanometer
range. It also provides a means for monitoring the flat band voltage, VFB, the interface trap spectra, DIT, and the total dielectric
charge, dQTOT- This technique is seen as a replacement for not only MOS C-V measurements but also for mercury-probe C-V.
EOT measurement by the differential corona C-V has a major advantage over optical methods because it is not affected by water
adsorption and molecular airborne contamination, MAC. These effects have been a problem for optical metrology of ultra-thin
dielectrics. The presented results illustrate the application of the technique to state of the art gate dielectrics, including Si-O-N
and HfO2.
surface voltage caused by the corona charge dQc on the
dielectric surface.
INTRODUCTION
A great deal of effort has been recently devoted to
improving non-contact corona metrology for electrical
monitoring of dielectrics on silicon wafers. This
technique involves charging a dielectric surface with
ions created by a corona discharge in air and measuring
of the surface potential with a vibrating capacitive
electrode [1,2]. No need for preparation of any test
structures results in cost and time saving advantages
over the commonly used MOS C-V technique. The
very fast feedback of results to processing has been a
driving force for implementation of non-contact
metrology on silicon microelectronic fabrication lines.
Fast data feedback is especially important for research
and development, R&D. Most recently, the interest in
corona metrology has increased further due to R&D on
new ultra-thin gate dielectrics and the realization that
monitoring of these films cannot be done solely
through optical technology [3]. Hydrocarbon deposits
from molecular airborne contamination and water
adsorbed from moisture in the ambient create a layer of
surface contamination that artificially increases optical
thickness, making it difficult to determine the correct
thickness of a dielectric underlayer. In corona
metrology the contamination layer behaves as an
electrical dipole layer covering the dielectric. It shifts
the voltage scale by AV^ie, but does not alter the
electrical thickness of the dielectric obtained from
capacitance C = dQc/dV, where dV is the change of
CORONA C-V MEASUREMENT
Electrical monitoring of ultra-thin dielectrics
requires overcoming two fundamental difficulties that
would be encountered in any C-V technique. First, the
measurement must be performed in the presence of
substantial leakage current flowing across a dielectric
[4,5]. This is a consequence of direct tunneling
promoted by the ultra-low thickness.
Second,
dielectric capacitance must be extracted from total
capacitance using high-field data in order to reduce the
substrate space charge contribution to capacitance
[6,7]. This further enhances the interference from
leakage. In this respect corona metrology has a
significant advantage over MOS-CV, namely, direct
tunneling currents are orders of magnitude lower than
that in MOS capacitors for the same dielectric field [4].
The electrical equivalent oxide thickness eEOT, or
simply EOT, is the key output parameter of a C-V
measurement. It is defined as the electrical thickness of
SiO2 that gives a capacitance the same as the dielectric
capacitance, Cd. Accordingly,
or
(La)
(l.b)
CP683, Characterization and Metrology for VLSI Technology: 2003 International Conference,
edited by D. G. Seiler, A. C. Diebold, T. J. Shaffner, R. McDonald, S. Zollner, R. P. Khosla, and E. M. Secula
© 2003 American Institute of Physics 0-7354-0152-7/03/$20.00
160
FLASK ANALOGY OF C-V EOT
MEASUREMENT
where s0 is the permittivity of free space, Td is the
thickness of a dielectric, and KSio2 and K are the
dielectric constants of SiO2 and of the measured
dielectric, respectively.
The corona C-V metrology is an extension of the
technique for measuring dielectric leakage current [5]
and the self-adjusting steady state method referred to as
SASS Tox [4]. However, in contrast to the single large
charge pulse used in these measurements, the corona
C-V uses a series of very quick charging with a small
dose, dQc. Each charging is followed by very precise
time resolved monitoring of the contact potential with a
time constant of 10ms and a sensitivity of 0.1 mV. The
latter enables the required precision in small
differential voltage measurement, dV. To achieve the
necessary corona charging precision, one must
precisely control several corona parameters, namely;
the discharge current value, the deposition time, the
corona gun geometry, its distance to the dielectric, and
a dynamic vacuum controlling the ambient within the
corona gun assembly. For very precise measurements,
a feedback between measured voltage and the corona
gun compensates for deposition rate changes caused by
electrostatic forces between ions approaching the
dielectric and ions already present on the surface. Such
feedback is essential for controlling dQc down to 0.1%
that is required to achieve similar accuracy in EOT
monitoring.
The basic ideas involved in time resolved corona CV metrology are discussed using the "Flask analogy"
depicted in Fig. 1. In this analogy, the dielectric is
represented by flask A. The goal is to determine the
flask cross section, S (analogous to EOT), by
measuring a change of the water level, Ah, caused by
filling the flask with a known amount of water. The
water dose is analogous to the corona dose, AQC, while
the change of water level, dh, is analogous to the
voltage change, dV.
Flask B represents the
semiconductor substrate of a dielectric film and is
connected with flask A. Its cross section decreases
with height, and for large heights it becomes much
smaller than that of flask A. This is analogous to the
"deep accumulation" condition. Flask B also connects
at different heights to additional flasks representing the
interface traps.
At the initial stages, water that is poured to flask A
fills predominantly the large flask B and the interface
traps. With increasing water level, the interface traps
become filled. In addition, the cross section of flask B
decreases.
This range, corresponding to deep
accumulation, provides a favorable condition for
measuring the cross section of flask A. However, in
this water level range, h > hieak, a leakage appears that
is analogous to the direct tunneling across the
dielectric.
MONITORING OF WATER LEVEL
(VOLTAGE)
FILLING At; MEASURING
DELAY
. MEASURING
INTERVALS
DOSE 1
t3
TIME
FIGURE 1. "Flask analogy" of time resolved Corona C-V metrology of dielectric capacitance Cd and EOT.
161
The most important feature in Fig. 2 is the elimination
of the leakage induced distortions in capacitance by the
time-resolved technique. It is also seen that in the
range of appreciable leakage the capacitance and oxide
thickness could not be measured correctly by the
standard corona method. The standard method [2]
relies on differentiation of the voltage vs. corona
charge curve, V-Q. This approach fails in the deep
accumulation and deep inversion ranges and, therefore,
cannot be used for extracting a dielectric capacitance.
As pointed out in our recent discussion of leakage
caused limitations [4], the standard corona
measurement approach is limited to SiO2 with
thicknesses above 2.5nm.
The corona C-V measuring cycle includes a
measurement of surface photovoltage, that is of the
illumination induced change of the voltage [1,8]. The
zero value of surface photovoltage corresponds to the
flat-band condition and is marked in Fig. 2 on the
voltage scale as VFB. If the voltage prior to any corona
charge application is V0, then a difference in total
corona charge placed on a dielectric surface between
V0 and VFB is the total charge, denoted as dQTOT [8].
This is an important parameter sensitive to electric
charges located in the dielectric (including the top
dielectric
surface)
and
also
at
the
dielectric/semiconductor interface. It is of interest to
note that the total charge is not obtainable in MOS C-V
measurements.
Further measurements are possible using a timeresolved procedure shown in the right side of the
figure. A quick dosing of water is followed by
monitoring of the water level transient. By
extrapolation of measured transients back in time, one
can determine the water level changes, Ah! and Ah2,
corrected for leakage which occurs during a small
delay between filling and measuring, At, that is a result
of the finite speed of the filling-measuring apparatus.
A limiting value of Ah that is obtained for large heights
contains a negligible contribution from flask B. In
corona C-V measurements performed in the high-field
range of deep accumulation, this is analogous to a
reduced contribution of space charge and interface
traps, CSc + Qt, to the total capacitance, C.
RESULTS AND DISCUSSION
C-V Characteristics
The corona C-V corresponding to a full range of
semiconductor space charge, from deep accumulation
to deep inversion, is shown in the upper portion of Fig.
2 for SiO2 on p-type silicon substrate with a dopant
concentration NA = lei5 cm"3. The lower portion of
Fig. 2 shows the corresponding leakage current density
obtained during the same measurement cycle. The
horizontal scale, for simplicity labeled "voltage",
represents the contact potential difference, VCPD,
measured with a vibrating probe as described in [1].
z.u
.0
LL
i
o
A
/
1.0
_
s'
\
t-Z.U
-1.0
1
9
*
VFB .•—.'* v0
. --.f . r,
on
-1.5
/
'••••"
I
•
• ••time resolved
" ——— standard
-2.0
"
INTERFACE
TRAPS
~ • • • • ••^ •0
s^""
**i
<D
0
(0
Q.
DEEP
DEEP
ACCUMULATION
^
0.5
1.0
1
1
1
0
0.5
1.0
-0.5
1.5
^
fQ1 woo
"w
0)
2
CO
fl)
_j
-20
-2.0
1
-1.5
1
-1.0
1
-0.5
*
1.5
Voltage [V]
FIGURE 2. Corona C-V and Leakage current measured on 2.2nm thick SiO2 on p-type silicon.
162
Extraction of Dielectric Capacitance CD
I
I
1
1
—— 1—.»- •
The static character of corona C-V and the
demonstrated sensitivity of C-V to interface traps
requires a modification of the extraction procedure for
dielectric capacitance in comparison to that used in
MOS C-V. In contrast to the entire MOS C-V curve
analysis [7], the extraction of Cd from corona C-V uses
only a high-field deep accumulation range of the
characteristic. In the corresponding high negative
corona limit, the dielectric field is about l.lMV/cm
and the (CSc + Qt) contribution approaches 100
fiF/cm2. This value is much larger than Cd of
3.5uF/cm2 corresponding to Inm SiO2, thereby creating
a favorable condition for extraction of Cd from the total
capacitance equation:
j— ..............
T^
| 3.0-
AAAAA
LL
"AA/AAaAAAA^AA
A
A
A
8 2.01
Drt~4e13A/
A
A
Csc
O
§. 1.0(0
o
nn -1.2
0
I
p—j |-—^
I__| |__I
..._,_|— .... — ...4
-1
-
-0.8
A
.
1
1
-0.6
-0.4
1
-0.2
.
A
A
1
.
0
0.2
Voltage [V]
FIGURE 3. The effect of flattening of corona C-V
observed for large interface trap density. Empty and
filled symbols correspond to nitrided oxides with
similar EOT of about 1.1 nm but with Dit - 4el3
q/cm2V and Iel2 q/cm2V, respectively.
c cd c<
Effect of Interface Traps
Interface traps, both of discrete energy level and
with a continuous energy distribution, are determined
in the corona C-V measuring cycle the same way as in
the COCOS technique, described in detail in [8]. One
shall notice, however, that the effect of traps on corona
C-V is different from that in MOS C-V. The first
difference stems from a true static nature of corona CV.
Thus, the corona capacitance contains a
contribution from all traps with a time response less
than the corona charging/measuring cycle - about 1 s.
In MOS measurement the capacitance is sensitive to
traps with a time constant limited by the capacitance
measuring frequency. For this reason, high frequency
MOS measurements are used for the purpose of
eliminating the contribution of traps to the measured
capacitance. The second very important difference is
due to a wafer treatment involved in the gate
preparation necessary for MOS C-V measurement.
The treatment (especially forming gas annealing)
reduces interface trap density. In contrast to corona
C-V wafers that are often measured just after oxidation
without any additional processing.
The static character and sensitivity of corona C-V
to interface traps is illustrated in Fig. 3 with data
obtained for nitrided oxides with similar EOT of about
1.1 nm but with very different interface trap density, Dit
of 4el3 q/cm2V and Iel2 q/cm2V, respectively. It is
seen that large Dit significantly flattens the corona C-V
curve. This is consistent with a static capacitance
equivalent circuit shown in the Fig. 3, where the oxide
capacitance, Cox, is in series with CSc + Qt.
163
(2,
Furthermore, CSc + Qt becomes a characteristic
power function of the net surface charge and is
independent of the dielectric thickness and the
semiconductor substrate doping. This behavior of
capacitance is similar to the high field MOS
capacitance recently reported by R. Clerc et.al [6] in
their study on characterization and modeling of MOS
with an ultra-thin oxide. From this reference, we
adopted the oxide capacitance extraction procedure and
we found it working exceptionally well for a large
range of dielectrics. We tested the procedure using a
series of wafers with SiO2 on p-type Si. The nominal
SiO2 thicknesses, given by the wafer supplier, were
based on "initial optical thickness" measured
immediately after oxidation in order to avoid effects of
contamination and adsorption that increase the optical
thickness. The example of the fitting curve together
with the corona C-V data and corresponding leakage is
shown in Fig. 4. The initial optical thickness of this
dielectric was 1.25nm (average of 5-site measurement).
This value compares very well with the EOT =
1.279nm extracted from corona C-V measured in the
center of the wafer. Figure 5 compares the average
values of 5-site corona C-V EOT with the
corresponding initial optical thicknesses. It is evident
that the adopted EOT extraction procedure gives
excellent one to one agreement with the nominal SiO2
thickness.
Effect of Moisture and MAC
Figure 6 shows results for SiO2 on p-type Si
measured for the first time within 1 day after oxidation
and then remeasured two weeks later. It is seen that
contamination of the dielectric with moisture and MAC
during wafer storage shifts the voltage by more than
0.2 V in the positive direction. This shift is consistent
with the dipole nature of adsorbed molecules [8]. It is
also seen from Fig. 6 that this surface contamination
has no effect on the oxide capacitance plateau in deep
accumulation and the extracted EOT values.
Note that both sets of data in Fig. 6 are fitted with
the same curve, but simply shifted in voltage. It is also
important to note that the initial optical elipsometric
measurements for this wafer agreed well with the EOT.
However, after two weeks of wafer storage the optical
thickness increased by 0.39nm, more than 30%, while
the corona EOT measurement was unchanged. It is
thus evident that corona C-V can bring a solution to
high-k monitoring problems discussed in ref. 3.
CD = 2.697 fj.Bcm2; EOT = 1.279 nm
2.5
•"X
•¥
£ *
^
\\
I1"5
S 1
• standard
—— fitting
n *; —————i—————i—————i—————i—————i—————
-1.6
-1.4
-1.2
-0.8
-0.6
-0.8
4
V-^
o 61 o
01 c
\
,
Leakage [nA/cm2]
-1.8
.8
-1.6
-1.4
-1.2
-1
Voltage [V]
FIGURE 4. Experimental data illustrating new timeresolved corona C-V and standard corona C-V in deep
accumulation with the corresponding leakage current
shown in the bottom portion. Solid line is the fitting
curve used to extract dielectric capacitance and EOT.
z
I
CD = 1 .68 nF/cm2;EOT = 2.054 nm
c$-
s
S.
3
e
c
s *
,
\
O
'
*
after storage
0
initial
\
^
\
\
—— fitting
1
\
*
icitance[(aFj
0
\
0.5 ————————|———————|———————|———————|———————
-1.3
-1.1
-0.9
-0.7
-0
-1 .5
2
Voltage [V]
Initial Optical Thickness [nm]
FIGURE 6. The effect of surface contamination during
wafer storage on corona C-V. Note the same dielectric
capacitance and EOT values are extracted from fitting
both C-V curves.
FIGURE 5. Correlation between EOT extracted with
the present corona C-V procedure and the optical
ellipsometric data measured just after oxidation. SiO2
films on p-type Si substrates.
164
research projects at Universities and non-industrial
laboratories.
o.u
Ta205
EOT=0.69nm
*r
5.0
Hf02
EOT=0.78nm
oooo
E
4.0
MA
EOT=1.05nm
« ®•
1
i
m
£
3
o
o
'* ^®
3.0
•
•
0 •
*•*
* »
SrTiO3
2.0
ACKNOWLEDGEMENT
%.
*•_
°o \
o
*
°
° •
o
*.o
EOT=2.10nm
o,*
.
*Sg$*
* ••
1.0
The authors would like to thank Dr. George Brown
from Sematech for enlightening discussions on
monitoring of advanced gate dielectrics.
.
m
REFERENCES
•*
*••••• X^...'
1.
nn
-3
-2
-1
0
+1
2.
3.
Voltage [V]
4.
FIGURE 7. Corona C-V characteristics measured for
high-k dielectrics on p-type substrates.
5.
High-K Dielectrics
The corona C-V technique and EOT extraction
method were applied to high-k dielectric films on ptype Si substrates. The results are shown hi Fig. 7.
EOT in the sub-nanometer range were measured for
Ta2O3 and HfO2; resulting in 0.69nm and 0.78nm,
respectively. Nominal physical thicknesses given by
suppliers were 4.0nm for Ta2O3, 4.6nm for HfO2,
3.0nm for AL2O3 and lOnm for SrTiO3.
The
comparison with corona EOT values gives the
corresponding values of dielectric constant of 22.6,
23.0, 11.1 and 18.6, respectively. These values are in
good agreement with literature data.
6.
7.
8.
CONCLUSIONS
The corona C-V method based on rapid corona
charging and time-resolved non-contact measurement
of dielectric potential with a vibrating contact potential
probe provides a reliable means for electrical
characterization of ultra-thin dielectrics. The method
enables measurement of differential capacitance in the
high-field range despite leakage neutralization of the
corona charge. This high field range is found suitable
for accurate extraction of dielectric capacitance and the
corresponding EOT. The methodology was tested
using SiO2 and subsequently applied to high-k
dielectrics with EOT extending to the subnanometer
regime. It is believed that the corona C-V EOT
measurement should prove extremely valuable as
electrical film characterization support for R&D of
new dielectrics. We also trust that this very accurate
and inexpensive metrology can be of interest for
165
Lagowski, J., Edelman, P., Inst. Phys. Conf. Ser. 160,
133-140 (1997).
Schroder, D. K.. Mat. Sci. and Eng. B91-92, 196 (2002).
Brown, A. E., Semiconductor International 25, 55
(2002).
Edelman, P., et al., Mat. Sci. and Eng. B91-92, 211
(2002).
Wilson, M., et al., "Study of Stress Induced Leakage
Current in Thni Oxides Stressed by Corona Charging in
Air: relationship to GOI Defects", in Structure and
Electronic Properties of Ultrathin Dielectric Films on
Silicon and Related Structures, edited by D.E. Buchanan
et al., MRS Symosium Proceedings 592, Warrendale,
PA, 2000, pp. 345-350.
Cleve, R., et al., Solid-State Electronics 46, 407 (2002).
Ahmed, K., et al., IEEE Trans. El. Devices 47, 1349
(2000).
Wilson, M., et al. "COCOS (Corona Oxide
Characterization of Semiconductor) Metrlogy: Physical
Principles and Applications", in Gate Dielectric
Integrity: Material, Process, and Tool Qualification,
ASTM STP1382, edited by D.C. Gupta and G.A.
Brown, American Society for Testing and Materials,
West Conshohocken, PA, 1999, pp. 74-90.