Content uploaded by Peter De Wolf
Author content
All content in this area was uploaded by Peter De Wolf on Dec 05, 2016
Content may be subject to copyright.
Status and review of two-dimensional carrier and dopant profiling using
scanning probe microscopy
P. De Wolf,a) R. Stephenson, T. Trenkler, T. Clarysse, and T. Hantschel
IMEC, Kapeldreef 75, B-3001 Leuven, Belgium
W. Vandervorst
IMEC, Kapeldreef 75, B-3001 Leuven, Belgium and KU Leuven, INSYS, Kard. Mercierlaan 92,
B-3001 Leuven, Belgium
共Received 28 March 1999; accepted 21 September 1999兲
An overview of the existing two-dimensional carrier profiling tools using scanning probe
microscopy includes several scanning tunneling microscopy modes, scanning capacitance
microscopy, Kelvin probe microscopy, scanning spreading resistance microscopy, and dopant
selective etching. The techniques are discussed and compared in terms of the sensitivity or
concentration range which can be covered, the quantification possibility, and the final resolution,
which is influenced by the intrinsic imaging resolution as well as by the response of the investigated
property to concentration gradients and the sampling volume. From this comparison it is clear that,
at present, none of the techniques fulfills all the requirements formulated by the 1997
Semiconductor Industry Association roadmap for semiconductors 关National Technology Roadmap
for Semiconductors 共Semiconductor Industry Association, San Jose, CA, 1997兲兴. Most methods are
limited to pn-junction delineation or provide a semiquantitative image of the differently doped
regions. However, recent comparisons have shown that the techniques can provide useful
information, which is not accessible with any other method. © 2000 American Vacuum Society.
关S0734-211X共00兲01201-4兴
I. INTRODUCTION
The 1997 U.S. Roadmap for Semiconductors from the
Semiconductor Industry Association 共SIA兲defined the needs
for nanometer-scale measurement of carrier concentration
profiles for the next decade.1These needs are summarized in
Table I for a 0.25, 0.18, and 0.13
m technology. Clearly,
there is a demand for sub-10-nm resolution, combined with
sufficient sensitivity 共down to the 1e15 atoms/cm3level兲and
high-quantification accuracy over a dynamic range of
1e15– 1e20 atoms/cm3. The need for such an extreme spa-
tial resolution as well as the applicability towards standard
devices has spurred the development of numerous two-
dimensional 共2D兲carrier profiling tools. Until today, more
than 20 different methods have been developed for this pur-
pose. These techniques can roughly be divided into four cat-
egories: 共i兲2D techniques which are based on a widely used
one-dimensional 共1D兲technique such as secondary ion mass
spectrometry 共SIMS兲共Ref. 2兲关imaging SIMS,32D SIMS,4,5
tomography SIMS,6and lateral SIMS 共Ref. 7兲兴 or spreading
resistance profiling 共SRP兲.8共ii兲Electron microscopy-based
techniques including field-effect scanning electron micros-
copy 共FE-SEM兲共Refs. 9 and 10兲and electron holography.11
共iii兲Inverse modeling techniques.12 共iv兲Scanning probe mi-
croscopy 共SPM兲-based techniques. This article is limited to
SPM-based techniques. All SPMs are based on the ability to
position various types of probes in very close proximity with
extremely high precision to the sample under investigation.
These probes can detect electrical current, atomic and mo-
lecular forces, electrostatic forces, or other types of interac-
tions with the sample. By scanning the probe laterally over
the sample surface and performing measurements at different
locations, detailed maps of surface topography, electronic
properties, magnetic or electrostatic forces, optical character-
istics, thermal properties, or other properties can be obtained.
The spatial resolution which can be obtained is only limited
by the sharpness of the probe tip, the accuracy with which
the probe can be positioned, the condition of the surface
under study, and the nature of the force being detected. The
resolution can vary from a few angstroms to tens or hundreds
of nanometers. This extremely high spatial resolution makes
SPM the ideal candidate for a general applicable 2D carrier
profiling tool. SPM-based 2D dopant profiling methods in-
clude various scanning tunneling microscopy techniques
共STM兲, dopant selective etching, scanning capacitance mi-
croscopy 共SCM兲, Kelvin probe force microscopy 共KPM兲,
scanning spreading resistance microscopy 共SSRM兲, and
scanning surface harmonic microscopy 共SSHM兲. In general,
all SPM-based 2D profiling techniques are being applied on
the cross section of the semiconductor structure under inves-
tigation. Earlier reviews of this topic have been given by
Subrahmanyan in 1992,13 Vandervorst et al.14 and Dagata
and Kopanski in 1995,15 Yu in 1996,16 and Vandervorst and
co-workers in 1997 共Ref. 17兲and 1998.18 The basic charac-
teristics of the methods described in this article are summa-
rized in Table II. Table II displays the type of probe used and
the measured physical quantity.
II. PRINCIPLES OF THE DIFFERENT TECHNIQUES
A. Scanning tunneling microscopy „STM…
The scanning tunneling microscope is a very surface-
sensitive SPM technique which requires, however, a conduc-
a兲Present address: Digital Instruments, 112 Robin Hill Road, Santa Barbara,
CA 93117; electronic mail: dewolf@di.com
361 361J. Vac. Sci. Technol. B 18„1…, JanÕFeb 2000 0734-211XÕ2000Õ18„1…Õ361Õ8Õ$15.00 ©2000 American Vacuum Society
tive surface, such that STM measurements on Si can nor-
mally not be operated in air due to the presence of the native
oxide. It is clear that if the native oxide is removed through
sample heating to relative high temperatures 共⬎1000 °C兲, the
dopant distribution is disturbed. Therefore, one has to use in
situ cleavage to generate a fresh sample cross section, which
is, however, only possible along certain crystal directions.19
The structure of interest must thus be oriented exactly in that
direction, limiting the flexibility of the technique. Addition-
ally, it is not always possible to determine the carrier profile
with respect to the mask edge and mask shape since the STM
cannot image the oxide position 共due to the requirement of a
conductive surface兲. Despite these basic limitations, cross-
sectional STM offers several methods for high-resolution
dopant profiling.
1. Dopant atom counting
The most direct STM-based dopant profiling method is to
simply count the number of dopant atoms appearing in 共or
near兲the surface atomic plane. Examples are given by
Johnson and co-workers,20–23 who clearly resolved indi-
vidual Be 共and Zn兲dopant atoms in a cleaved 共110兲surface
of Ga. In this work, the negatively charged dopant atoms are
attractive to holes and appear in the STM image as protru-
sions. From the size and shape of the features, dopants at the
surface and one atomic layer below the surface could be
distinguished from each other and from dopants further be-
low the surface. So far, dopant atom counting has not been
reported for Si substrates. In this context it is worth noting
that the capability of imaging individual dopant atoms is
TABLE I. Most important requirements for 2D carrier profiling as given in the SIA Roadmap for Semiconductors
共Ref. 1兲.
Design rule 0.25
m 0.18
m 0.13
m
Accuracy of carrier
concentration 5% 5% 4%
Measurement repeatability 5% 5% 4%
Spatial resolution 5 nm 3 nm 2 nm
Carrier concentration range
共atoms/cm3兲
1e15– 1e20 1e15– 1e21 1e15– 1e21
TABLE II. Summary of the different scanning probe microscopy techniques which can be used for 2D carrier
profiling of semiconductor devices. The ‘‘mode’’ reflects the scanning mode which is being used to control the
movement of the probe (NC⫽noncontact; C⫽contact).
Technique Mode Probe Measured
quantity
Scanning tunneling microscopy/
spectroscopy
共STM/STS兲
STM Metallic
needle No. doping atoms
I–Vspectra
Selective etching⫹atomic force
microscopy NC-AFM Ultrasharp Si Topography
after chemical
etch
Scanning capacitance microscopy/
spectroscopy
共SCM/STS兲
C-AFM Metal-coated
Si
or metallic
Depletion
capacitance
C–Vspectra
Scanning spreading resistance
microscopy 共SSRM兲
C-AFM Diamond-
coated Si
or diamond
Electrical
resistance
I–Vspectra
Kelvin probe force microscopy
共KPM兲
NC-AFM Metal-coated
Si
or metallic
Electrostatic
potential
共electric field兲
Scanning surface harmonic
microscopy
共SSHM兲
STM Metallic
needle with
microwave
cavity
Depletion
capacitance
362 De Wolf
et al.
: Status and review of 2D carrier profiling 362
J. Vac. Sci. Technol. B, Vol. 18, No. 1, JanÕFeb 2000
associated with the special properties of cleaved GaAs共110兲
surfaces where the intrinsic surface states are outside the
bulk band gap.23 Although this method provides the ultimate
resolution, its sensitivity is rather poor. Since only the atoms
in the top two atomic layers are revealed, a high dopant
concentration is required to be able to determine the dopant
concentration adequately.
2. Scanning tunneling spectroscopy (STS) and
current imaging tunneling spectroscopy (CITS)
The work using STM as a dopant profiler has primarily
been focused on pn-junction delineation by detecting differ-
ences in tunneling current characteristics for n- and p-type
material. Feenstra and co-workers carried out the first de-
tailed imaging and spectroscopic studies of GaAs pn junc-
tions, observing electronically induced topography.24–26
Current–voltage spectroscopy allowed the n-type, p-type,
and depleted regions to be identified unambiguously. The
STM current dependence on dopant type and concentration
within semiconductors is due to tip-induced bandbending at
the surface.27
Kordic and co-workers performed the first cross-sectional
STM studies of Si pn junctions exposed by cleaving in air
and UHV.28–30 The location of the junctions could be re-
solved to within 30 nm using a potentiometric technique in
UHV 关scanning tunneling potentiometry 共STP兲兴: A forward
or reverse bias was applied across the pn junction, and dif-
ferences in tunneling current measured in p-type, depleted,
and n-type regions for various bias voltages applied to the pn
junctions then revealed the electronic structure of the biased
junctions.
Yu and co-workers31,32 used current imaging tunneling
spectroscopy 共CITS兲. In this technique a constant-current to-
pographic scan with the current stabilized at a fixed value I0
for a specific voltage V0is performed over the sample sur-
face and, at each point, a current–voltage spectrum is mea-
sured. Variations in electronic structure across the sample
surface produce corresponding variations in the current–
voltage spectra; these spatial variations can be revealed by
plotting the current measured at specific bias voltages other
than V0—so-called current images. CITS under UHV condi-
tions on cleaved, hydrogen-passivated cross-sectioned Si
metal–oxide–semiconductor 共MOS兲structures made it pos-
sible to image the source and drain junction profiles with a
spatial resolution on the order of 10 nm.32 Because the
current–voltage spectra differed significantly for the p-type,
n-type, and depleted regions, current images generated from
the spatially resolved tunneling spectra were able to reveal
the profiles of the pn junctions. A problem to extend this
work further to quantitative dopant profile information, is the
dependence of the tunneling current on the Fermi level and
bandbending rather than on the carrier concentration directly.
In addition, the surface disorder problems form a significant
limitation for the STM approach.
In conclusion, the different STM-based cross-sectional
carrier 共or doping兲profiling methods have low performance
on Si. This difficulty can be ascribed to the difficult cross-
section preparation, combined with the fact that the STM is a
very surface-sensitive technique. First, Si 共001兲wafers are
more difficult to cleave than III–V wafers; and second, the
as-cleaved 共110兲cross-sectional surface is atomically disor-
dered with its electronic structure dominated by dangling-
bond states. Despite these complications, 2D junction delin-
eation and qualitative 2D nanometer-scale carrier profiling of
cross-sectioned Si-based structures have been achieved by
several research groups. The applicability of STM 关or scan-
ning tunneling spectroscopy 共STS兲兴 as a quantitative carrier
profiling tool is complicated and, therefore, limited.
B. Dopant selective etching or staining
The method of chemical staining or etching of doped Si
layers has been known since the early days of semiconductor
processing. Staining techniques use selective deposition of a
metal such as Cu, Au, Ag, or Pt on one side of the junction
by an electrochemical displacement reaction from a metal-
ion-based solution. Etching techniques use mixtures of HF
and an oxidizing agent to preferentially etch regions with a
high carrier concentration. The impurity-sensitive etching so-
lutions typically consist of a mixture of HF, HNO3, and ei-
ther H2OorCH
3COOH. The sample topography after the
staining or etching step is a measure for the 2D carrier profile
and can be imaged using transmission electron microscopy
共TEM兲,32–34 SEM,35 STM,36,37 or atomic force microscopy
共AFM兲.38 In a final step, the measured topography profiles
have to be converted to electrical carrier distributions.
Among the advantages of using AFM compared with STM
in performing topographic measurements on selectively
etched cross sections is that AFM makes it possible to mea-
sure the surrounding structures 共oxides, metals, etc.兲since
this is not possible with the STM, which requires a conduct-
ing sample surface. Also, AFM is less sensitive to the de-
tailed electronic structure of the etched surface and to pos-
sible contamination of the sample or tip surfaces. Both the
STM and AFM method can suffer from tip–sample convo-
lution leading to an incorrect image of the etched region. The
regions with a high carrier concentration gradient are particu-
larly sensitive to this effect.
Independently of the technique which is used to measure
the sample topography, the accuracy of etching and staining
techniques is always limited by the reproducibility of factors
such as surface preparation, the concentration of the staining
or etching solution, etching time, the volume and the agita-
tion of the solution on the sample, temperature, and light
illumination. The effects of these factors can be minimized
by careful sample preparation and precise control of the etch-
ing factors. In this context, different approaches have been
used by several groups. A promising method to obtain longer
etching times and a better general control of the etching
makes use of an electrochemical etching procedure with po-
tentiostatic control.39
An important drawback of the etching techniques until
now is that analysis suffers from a poor understanding of the
etching process and the correlation between carrier concen-
tration and the observed topography. Two approaches have
363 De Wolf
et al.
: Status and review of 2D carrier profiling 363
JVSTB-MicroelectronicsandNanometer Structures
been employed so far for this calibration. The first calibra-
tion method uses a 1D doping profile adjacent to the 2D area
of interest. The 1D doping profile may be measured by SIMS
or SRP or be numerically simulated. After etching, the 1D
etched profile is characterized and related to the known 1D
carrier 共or doping兲distribution. A 2D doping map is obtained
by assuming that the dependence of the etching rate on car-
rier concentration obtained for the 1D region is also correct
for the complete 2D area. The second calibration method
includes direct measurement of the etching rate as a function
of carrier concentration using homogeneously doped bulk or
epitaxially grown Si samples. Again, it is assumed that under
similar etching conditions 共temperature, concentrations,
light, time, etc.兲the etching rate for a particular spot depends
exclusively on the carrier concentration at this point. How-
ever, it has been found that the etching also depends on the
carrier concentration gradient.40
In summary, the applicability of the selective etching
method to 2D carrier profiling is limited because of 共i兲poor
understanding of the etching process, 共ii兲poor control of the
etching conditions, and 共iii兲lack of a good general quantifi-
cation procedure. Despite these limitations, the method re-
mains valuable for fast qualitative analysis.
C. Scanning capacitance microscopy „SCM…
In the scanning capacitance microscope, the sample 共or
the metallic AFM tip兲is covered with a thin dielectric layer,
such that the tip–sample contact forms a metal–insulator–
semiconductor 共MIS兲capacitor, whose capacitance–voltage
(C–V) behavior is determined by the local carrier concen-
tration of the semiconductor sample. If no dielectric layer is
used, the tip–sample contact forms a metal–semiconductor
共MS兲structure and one has a so-called Schottky contact
SCM.41,42 By monitoring the capacitance variations as the
probe scans across the sample surface, one can measure a 2D
carrier concentration profile. Since the total tip–sample ca-
pacitance is large compared to the capacitance variations due
to different carrier concentrations, one usually measures the
capacitance variations and not the absolute capacitance val-
ues. Note that no signal is measured if the probe is posi-
tioned over a dielectric or metallic region since these regions
can not be depleted. Most SCMs are based on contact-mode
AFM with a conducting tip, and an essentially independent
capacitance measurement in parallel.43 The capacitance be-
tween the tip and sample is measured by using a high-
frequency capacitance sensor, based on a 915 MHz oscillator
driving a resonance circuit which is tuned in part by the
external capacitance to be measured. The capacitance detec-
tion limit is as small as 1e-19 F in a 1 kHz bandwidth 共trans-
lating into a noise level of 3e-21F/
冑
Hz). An extensive re-
view of SCM for 2D dopant profiling is given by Williams.44
The SCM is usually operated in one of the following two
modes.
1. Differential-capacitance (open-loop) mode
In the open-loop mode, an ac bias 共typically, 0.2–2 V,
10–100 kHz兲is superimposed on a dc sample bias 共⫺2–2
V兲, while the tip is at dc ground. The ac bias can alternately
deplete and accumulate the semiconductor surface region.
The modulated surface capacitance changes ⌬Cunder the
probe tip are registered using a lock-in technique, simulta-
neously with the topographical data, while the probe is
scanned across the surface. When using large ac voltages
共several volts兲, this setup measures ⌬Cacross the entire
C–Vcurve, and is almost independent on shifts in the flat-
band voltage caused by oxide or surface charges. When
smaller ac bias voltages are used, the differential capacitance
dC/dV is measured.
2. Closed-loop mode
In the closed-loop mode, the magnitude of the ac bias
voltage applied to the sample is adjusted by a feedback loop
to maintain a constant capacitance change as the tip is
scanned across the sample at constant force.45,46 The
feedback-controlled magnitude of the ac bias voltage is re-
corded. The main advantage of the closed-loop mode is the
fact that the capacitance and, consequently, the depletion
width is kept constant, whereas the depletion width might
become very large 共⬎1
m兲for lowly doped regions in the
differential capacitance mode, leading to a loss in spatial
resolution.
Since in SCM the measured capacitance signal is propor-
tional to the tip interaction area, shrinkage of the tip size will
improve the spatial resolution, but also reduce the sensitivity
of the capacitance measurement. A dynamic range of
1e14– 1e20 has been demonstrated on a special calibration
structure.47 The sensitivity is a function of the carrier con-
centration and can be increased by reducing the thickness of
the oxide layer. The response of SCM is not necessarily
monotonic with concentration but does depends on the ap-
plied experimental conditions as well.48 Several groups have
published qualitative 1D and 2D 共Refs. 49–51兲open- or
closed-loop SCM images, which were compared with TCAD
simulations and conventional 1D carrier profiling results.
The resolution of these images is on the order of 10–20
nm.52
Much of the recent work related to SCM is focused on the
theoretical interpretation and quantitative conversion of the
measured signals into carrier profiles. In an ideal 共flat兲MOS
capacitor, the dopant concentration is easily extracted from
the variation of capacitance with voltage. The situation is
more complex for SCM since there are stray fields between
the probe shaft and the sample surface, the sample has a
nonconstant carrier concentration, and the quality of the sur-
face is largely unknown 共surface charges, contaminants, ox-
ide quality, etc.兲. Several models were presented to set up a
quantification procedure. In the simplest model, the quantifi-
cation of the SCM images is achieved using SIMS or SRP
data in conjunction with contour mapping software. Hereby,
it is assumed that the dependence of the SCM signal on
carrier concentration obtained for the 1D line is also correct
for the complete 2D area.
In a quasi-1D analytical model, the tip is modeled as a
metallic sphere placed in an insulating dielectric medium
364 De Wolf
et al.
: Status and review of 2D carrier profiling 364
J. Vac. Sci. Technol. B, Vol. 18, No. 1, JanÕFeb 2000
near a Si surface with a sphere–Si gap just equal to the
experimental measured oxide thickness.49 The Si surface is
divided into annular regions. The insulator capacitance be-
tween the sphere and each annular region, and the Si deple-
tion capacitance for each annular region are calculated and
summed to give the total tip–sample capacitance. This ap-
proximate analytical model provides a means to rapidly cal-
culate the C–Vrelation as a function of tip radius, dielectric
constant, gap distance, and carrier density. Note that a con-
stant carrier density is assumed in these calculations, the ef-
fect of a carrier gradient 共in the lateral direction兲is not yet
included.
In an alternative model the C–Vrelation is calculated
using the solution of the three-dimensional 共3D兲Poisson
equations for various tip–sample bias voltages, tip–sample
gap distances, carrier concentrations, and oxide
thickness.49,53,54 The probe is modeled as a cone with a hemi-
spherical tip end to include stray-field effects. The quality of
the sample surface, although known to be an important pa-
rameter, was not included in the calculations. Algorithms
based on interpolation within the results of this database
have then been developed, to convert a measured SCM pro-
file into a quantitative carrier profile. Kopanski and
co-workers49,53 treat each measured capacitance point inde-
pendently and thus ignore the effect of carrier gradients,
whereas Yu, Griffin, and Plummer54 include the effect of the
dopant gradient in their simulations.
It needs to be pointed out that the behavior of the SCM
signal 共in the open- or closed-loop mode兲at pn junctions is
not well understood and is not yet included in the calcula-
tions, although pronounced effects have been observed.49
Calculations and experiments have shown that the tip–
sample bias 共ac and dc兲has a pronounced effect on the pres-
ence, position, and extent of extra contour lines in the vicin-
ity of the junction, complicating precise junction delineation
and data quantification at pn junctions.55,56 Several groups
have proposed and implemented promising approaches to
overcome this problem. Timp et al. use heavily doped Si tips
rather than metal-coated tips.57 In this case, not only the
sample, but also the tip itself, is being depleted and accumu-
lated. The resulting SCM images persist over a wide tip–
sample bias range. The delineation of the highly doped re-
gions 共such as the poly-Si gate兲and the metal regions
becomes now clearer and junction positions appear less in-
fluenced by the bias voltage used. However, now the exact
concentration of the tip material enters as an extra parameter
in the quantification 共in particular, in those cases where car-
rier concentration in the sample and tip are equivalent, inter-
pretation becomes less transparent兲. Edwards et al. present
another approach:58 scanning capacitance spectroscopy
共SCS兲. In SCS the C–Vcurve is measured for every position
of the probe in the 2D profile. The shape of the C–Vcurve
allows one to distinguish between n-type, p-type, and deple-
tion regions and facilitates data quantification. Both
approaches—Si tips, and SCS—result in easier data interpre-
tation, in particular, at pn junctions. However, quantification
of the data into carrier concentration values still requires 3D
simulations in order to determine the exact impact of carrier
concentration gradients and pn junctions on the observed
SCM images. In conclusion, the SCM is a promising tool for
quantitative 2D carrier profiling with nanometer resolution.
The spatial resolution 共10–20 nm兲and dynamic range
(1e15– 1e20 atoms/cm3) are good. The major challenges in
SCM include: surface preparation including the formation of
a good dielectric oxide,59 and the quantification and calibra-
tion methodologies. At present, the technique is widely used
in a qualitative manner and only limited by a lack of a gen-
erally applicable quantification routine.
D. Scanning surface harmonic microscopy „SSHM…
The SSHM consists of a STM with a microwave cavity,
in which a microwave signal is applied across the tip–sample
tunneling gap.60 The nonlinear tip–sample MIS capacitance
Cresults in higher-order harmonics in the tunneling current.
The second- and third-harmonic signals are proportional to
the first and second derivatives of C. The driving frequency
is chosen such that the second- or third-harmonic frequency
corresponds to the resonance frequency of the cavity 共typi-
cally, 2.7 GHz兲, making it detectable.61,62 The capacitance C
is a measure for the local active carrier concentration in the
semiconductor and the insulator quality in the same way as
in the SCM technique. Bourgoin, Johnson, and Michel have
applied the SSHM technique to delineate the qualitative car-
rier profile of Si pn junctions with 5 nm resolution.61 In
general, the signal-to-noise ratio is lower compared to the
SCM method and quantification of the data is faced with
similar problems as the SCM technique.
E. Kelvin probe force microscopy „KPM…
Kelvin probe force microscopy and the related scanning
Maxwell-stress microscopy, are high-resolution and highly
sensitive potential imaging methods.63 A conductive tip is
scanned in the noncontact AFM mode while an ac voltage
共frequency f兲is applied to it. This voltage results in an elec-
trostatic field which causes an oscillation of the cantilever at
the same frequency. This force disappears when the dc po-
tential difference between tip and sample is zero. Thus, by
observing the amplitude of the cantilever oscillation at fre-
quency fby a lock-in technique, and nulling it by changing
the dc bias voltage on the tip, the sample’s surface potential
can be measured. The maximum sensitivity is obtained when
fcorresponds to one of the resonance frequencies of the can-
tilever. In order to separate the height-control signal and the
voltage signal, the first cantilever resonance peak is usually
employed for the tip height control while fis taken equal to
the second resonance peak on a dual resonant probe.63 This
mode was used to measure the 2D potential distribution in-
side Si device structures.64–66 The measured electrochemical
potential difference between the probe tip and sample surface
is dependent on the carrier concentration-related work-
function difference, and can thus be used as a measure for
the local carrier concentration, although the sensitivity is
limited. This mode of KPM has been applied successfully for
qualitative 2D carrier profiling of Si structures.66,67 The tech-
365 De Wolf
et al.
: Status and review of 2D carrier profiling 365
JVSTB-MicroelectronicsandNanometer Structures
nique is sensitive to changes in carrier concentration from
1e15 to 1e20 atoms/cm3with a spatial resolution of about
100 nm. However, the sensitivity to small concentration
changes and the application towards quantitative profiling
are limited by surface charges on the sample and calibration
of the KPM technique against absolute doping concentration
standards remains to be demonstrated.
F. Scanning spreading resistance microscopy
„SSRM…
In SSRM the electrical resistance is measured between the
conductive probe tip and a large current-collecting back con-
tact while the probe is scanned in the contact mode across
the cross section of the Si device. When the applied force
exceeds a certain threshold force, the measured resistance is
dominated by the spreading resistance.68 As for the conven-
tional 共1D兲spreading resistance profiling method, the
spreading resistance depends inverse proportionally on the
local carrier concentration underneath the probe–silicon con-
tact. If one uses small bias voltages 共about 100 mV兲, the
displacements as observed in SCM imaging are minimized.
Contact potential differences between tip and surface can
possibly alter this situation. Junction positions can be as-
signed to a single point as their positions show up as a peak
in the resistance profile. On Si structures, high forces 共typi-
cally, a few
N兲are required in order to penetrate the native
oxide and to establish a stable electrical contact. As standard
AFM probes deform at these high forces, one needs to use
doped diamond or diamond-coated Si probes. If lower forces
are being used, the measured resistance is no longer domi-
nated by the spreading resistance but by the contact resis-
tance of the tip–sample contact. In this case, qualitative car-
TABLE III. Intercomparison of two-dimensional doping 共D兲and carrier 共C兲profiling methods (NA⫽not available).
Method Ref. Resol.
共nm兲
Range
共cm⫺3兲
Conc.
resol. D/
CQuanti-
fiable Comments and problems
SPM techniques
SCM 共43–59兲10 1e15– 1e20 Power C Limited Uncertainties at junctions, poor
quantification procedure
SSHM 共60–62兲5 NA Power C No No quantification procedure
STM-atom
counting 共20–23兲Atomic 1e18– 1e20 Linear D Yes Only on GaAs, not on Si
STM-STS/
CITS 共24–26兲
共31,32兲
10 NA Log. C Limited Only junction delineation and type 共n
or p兲identification
STM-STP 共27–30兲10 NA Limited C Limited Only junction delineation
KPM 共66,67兲100 1e15– 1 e20 Limited C Limited Poor quantification procedure, stray-
fields limit the resolution
SSRM 共68–73兲20 1e15– 1e20 Linear C Yes Availability diamond probes
Chemical etch
⫹AFM/STM 共37–39兲10–20 1e17–1e20 Limited C Limited Difficult to quantify, poor
reproducibility
1D-based techniques
Imaging SIMS 共3兲100 NA Linear D Yes Sensitivity limited by target volume
2D-SIMS 共4,5兲30–50 1e16– 1e21 Linear D Yes Special structures required
2D-tomography
SIMS 共6兲50 NA Linear D Yes Special structures required, complex
sample preparation
Lateral SIMS 共7兲5–10 Dose Linear D Yes Only the lateral dose distribution is
measured
2D-SRP 共8兲100 1e15– 1 e21 Linear C Yes Special structures required
EM techniques
Chemical etch
⫹SEM/TEM 共32–35兲20 1e17– 1e21 Limited C Limited Only qualitative
FE-SEM 共9,10兲10–20 4e16– 1e21 Limited Limited Robust model for quantification is not
available
E-holography 共11兲1–10 1e17–1e20 Limited C Limited
Inverse modeling techniques
Inv. modeling
with C–V
共12兲NA NA C Yes Resolution and accuracy are unknown,
long calculation times
366 De Wolf
et al.
: Status and review of 2D carrier profiling 366
J. Vac. Sci. Technol. B, Vol. 18, No. 1, JanÕFeb 2000
rier profiling remains possible, although the sensitivity to
small changes in the carrier concentration is lost.69
The quantification of the SSRM resistance data into car-
rier concentration values is possible using an n- and p-type
calibration curve in combination with an algorithm which
corrects for the effect of nearby layers with different carrier
concentrations.70 This quantification method is simple and
fast and does not require one to measure the 1D in-depth
profile by SIMS or SRP. The SSRM method has been ap-
plied on various InP 共Ref. 71兲and Si 共Refs. 68, 72, and 73兲
device structures. These measurements include fully quanti-
tative 2D profiles of both nMOSFETs and pMOSFETs
共metal–oxide–semiconductor field-effect transistors兲and al-
low extraction of values for effective channel lengths and
two-dimensional diffusion effects depending on gate
lengths.73 Reasonable agreement is also obtained when com-
pared to data extracted from device characteristics.74 The
data indicate a dynamic range of 1e15– 1e20 atoms/cm3,a
spatial resolution of about 25 nm, and also show that SSRM
can be applied on arbitrary device structures. At present, the
SSRM method is mainly limited by the availability of con-
ductive diamond or diamond-coated probes.71
III. INTERCOMPARISON
Various features of the methods described above are sum-
marized in Table III. The SPM-based 2D carrier profiling are
compared with alternative methods based on 1D techniques,
electron microscopy, or inverse modeling. The most impor-
tant features are spatial resolution 共in nanometers兲, dynamic
range 共in atoms/cm3兲, sensitivity 共defined here as the ratio of
the change in the instrument response to a corresponding
change in carrier concentration兲, quantification ability, and
applicability to real devices 共without needing special test
structures兲. The listed resolution and dynamic range corre-
spond to the best specifications found in the literature. Note
that different definitions for spatial resolution and sensitivity
共or dynamic range兲are used by different groups, complicat-
ing this intercomparison. If no value is found in the litera-
ture, the specification is labeled as ‘‘not available’’ 共NA兲.
The ability to transform the measured raw data values into a
fully quantitative 2D doping or carrier profile is labeled as
‘‘yes,’’ ‘‘no,’’ or ‘‘limited.’’ The concentration resolution is
labeled as ‘‘linear’’ if the instrument response is proportional
to the carrier concentration, ‘‘logarithmic’’ if the instrument
response is proportional to the logarithm of the carrier con-
centration, and ‘‘power’’ if there is a power-law relation be-
tween the carrier concentration and the instrument response,
or ‘‘limited.’’
Based on this intercomparison, it is clear that none of the
available techniques fulfills all the requirements formulated
in Table I and can at the same time be applied on any arbi-
trary semiconductor structure. Most techniques are limited to
pn-junction delineation or provide a qualitative image of the
differently doped regions. Whereas all SPM-based methods
can be applied on arbitrary devices, all 1D-based techniques
require a special test structure, making them less flexible as
compared to the SPM-based methods.
The authors believe that although the SSRM and the SCM
techniques are still in a development or optimization stage,
they are the leading candidates to become a general appli-
cable 2D carrier profiling tool. Both techniques have a reso-
lution close to the values asked for, can be applied on arbi-
trary structures, and have the required dynamic range.
Quantification accuracy and reproducibility are still issues
which need to be determined with much more detail.
IV. CONCLUSIONS
At present there are numerous SPM-based methods for
2D carrier profiling of Si 共and other兲device structures. These
include several scanning tunneling microscopy modes, scan-
ning capacitance microscopy, Kelvin probe microscopy,
scanning spreading resistance microscopy, and dopant selec-
tive etching. All of these methods can be directly applied on
the cross section of an arbitrary device and do not require
special test structures. The methods differ in the range of
carrier concentrations which can be mapped, the spatial reso-
lution, and the quantification possibility. The features of the
different methods can be summarized as follows:
共i兲The different STM-based methods have a rather low
performance on Si due to the difficult cross-section prepara-
tion, combined with the fact that the STM is a very surface-
sensitive technique. STM-based carrier profiling is, there-
fore, limited to 2D junction delineation and qualitative 2D
imaging. At present, the applicability of STM as a quantita-
tive carrier profiling tool is not possible.
共ii兲The applicability of the selective etching method 共fol-
lowed by AFM imaging兲is limited because of a poor under-
standing of the etching process, a poor control of the etching
conditions, and a lack of a general quantification procedure.
However, the etching method is a very valuable tool for fast
qualitative analysis of the 2D carrier profile.
共iii兲Both the SCM and SSRM are promising tools for
quantitative 2D carrier profiling with nanometer resolution.
The spatial resolution 共10–20 nm兲and dynamic range
(1e15– 1e20 atoms/cm3) are good. Some further research
and development is required such that these methods fulfill
the requirements set by the 1997 SIA roadmap for semicon-
ductors. This future work includes the development and fab-
rication of reliable probes for SSRM, and the development of
routines to interprete and convert the SCM images into quan-
titative carrier concentration values.
1National Technology Roadmap for Semiconductors 共Semiconductor In-
dustry Association, San Jose, CA, 1997兲.
2M. G. Dowsett, Proceedings of the SIMS XI Conference, edited by G.
Gillen, R. Larea, J. Bennet, and F. Stevie 共Wiley, Chichester, 1997兲,p.
157.
3S. R. Bryan, W. S. Woodward, R. W. Linton, and D. P. Griffis, J. Vac.
Sci. Technol. A 3, 2102 共1985兲.
4M. G. Dowsett and G. A. Cooke, J. Vac. Sci. Technol. B 10,353共1992兲.
5V. A. Ukraintsev, P. J. Chen, J. T. Gray, C. F. Machala, L. K. Magel, and
M.-C. Chang, J. Vac. Sci. Technol. B, these proceedings.
6S. H. Goodwin-Johansson, M. Ray, Y. Kim, and H. Z. Massoud, J. Vac.
Sci. Technol. B 10,369共1992兲.
7R. Van Criegern, F. Jahnel, R. Lange-Gieseler, P. Pearson, G. Hobler, and
A. Simionescu, J. Vac. Sci. Technol. B 16, 386 共1998兲.
8V. Privitera, W. Vandervorst, and T. Clarysse, J. Electrochem. Soc. 140,
262 共1993兲.
367 De Wolf
et al.
: Status and review of 2D carrier profiling 367
JVSTB-MicroelectronicsandNanometer Structures
9R. Turan, D. D. Perovic, and D. C. Houghton, Appl. Phys. Lett. 69, 1593
共1996兲.
10D. Venables, H. Jain, and D. C. Collins, J. Vac. Sci. Technol. B 16, 362
共1998兲.
11W. D. Rau, F. H. Baumann, H. H. Vuong, B. Heinemann, W. Hoppner, C.
S. Rafferty, H. Rucker, P. Schwander, and A. Ourmazd, Tech. Dig. Int.
Electron Devices Meet., 713 共1998兲.
12N. Khalil, J. Faricelli, C. L. Huang, and S. Selberherr, J. Vac. Sci. Tech-
nol. B 14,224共1996兲.
13R. Subrahmanyan, J. Vac. Sci. Technol. B 10, 358 共1992兲.
14W. Vandervorst, T. Clarysse, P. De Wolf, L. Hellemans, J. Snauwaert, V.
Privitera, and V. Raineri, Nucl. Instrum. Methods Phys. Res. B 96, 123
共1995兲.
15J. A. Dagata and J. J. Kopanski, Solid State Technol. 38,91共1995兲.
16E. T. Yu, Mater. Sci. Eng. 17,147共1996兲.
17W. Vandervorst, T. Clarysse, P. De Wolf, T. Trenkler, T. Hantschel, and
R. Stephenson, Future Fab. Int. 4,287共1997兲.
18W. Vandervorst, T. Clarysse, P. De Wolf, T. Trenkler, T. Hantschel, R.
Stephenson, and T. Janssens, in Proceedings of the International Work-
shop on Semiconductor Characterization: Present Status and Future
Needs, edited by D. Seiler 共AIP, New York, 1998兲.
19Y.-C. Kim, M. J. Nowakowaski, and D. N. Seidman, Rev. Sci. Instrum.
67, 1922 共1996兲.
20M. B. Johnson, O. Albrektsen, R. M. Feenstra, and H. W. M. Salemink,
Appl. Phys. Lett. 63, 2923 共1993兲.
21M. B. Johnson, H. P. Meier, and H. W. M. Salemink, Appl. Phys. Lett.
63, 3636 共1993兲.
22H. W. M. Salemink, M. B. Johnson, O. Albrektsen, and P. Koenraad,
Solid-State Electron. 37, 1053 共1994兲.
23K.-J. Chao, A. R. Smith, A. J. McDonald, D.-L. Kwong, B. G. Streetman,
and C.-K. Shih, J. Vac. Sci. Technol. B 16, 453 共1998兲.
24R. M. Feenstra, E. T. Yu, M. Woodall, P. D. Kirchner, C. L. Lin, and G.
D. Pettit, Appl. Phys. Lett. 61,795共1992兲.
25S. Gwo, A. R. Smith, K.-J. Chao, C. K. Shih, K. Sadra, and B. G. Street-
man, J. Vac. Sci. Technol. A 12, 2005 共1994兲.
26R. M. Silver, J. A. Dagata, and W. Tseng, J. Vac. Sci. Technol. A 13,
1705 共1995兲.
27R. M. Feenstra and J. A. Stroscio, J. Vac. Sci. Technol. B 5, 923 共1987兲.
28S. Kordic, E. J. Van Loenen, D. Dijkkamp, A. J. Hoeven, and H. K.
Moraal, J. Vac. Sci. Technol. A 8, 549 共1990兲.
29S. Kordic, E. J. Van Loenen, and A. J. Walker, IEEE Electron Device
Lett. 12,422共1991兲.
30M. B. Johnson and J.-M. Halbout, J. Vac. Sci. Technol. B 10, 508 共1992兲.
31E. T. Yu, M. B. Johnson, and J.-M. Halbout, Appl. Phys. Lett. 61, 201
共1992兲.
32E. T. Yu, K. Barmak, P. Ronsheim, M. B. Johnson, P. McFarland, and
J.-M. Halbout, J. Appl. Phys. 79,2115共1996兲.
33R. Alvis, B. Mantiply, and M. Young, J. Vac. Sci. Technol. B 14, 452
共1996兲.
34S. S. Neogi, D. Venables, Z. Ma, and D. M. Maher, J. Vac. Sci. Technol.
B16,471共1998兲.
35L. Gong, S. Petersen, L. Frey, and H. Ryssel, Nucl. Instrum. Methods
Phys. Res. B 96,133共1995兲.
36T. Takigami and M. Tanimoto, Appl. Phys. Lett. 58, 2288 共1991兲.
37M. Tanimoto and T. Takigami, Ultramicroscopy 42–44, 1381 共1992兲.
38G. Neubauer, M. Lawrence, A. Dass, and T. J. Johnson, Mater. Res. Soc.
Symp. Proc. 286,283共1992兲.
39T. Trenkler, W. Vandervorst, and L. Hellemans, J. Vac. Sci. Technol. B
16,349共1998兲.
40V. A. Ukraintsev, R. McGlothin, M. A. Gribelyuk, and H. Edwards, J.
Vac. Sci. Technol. B 16,476共1998兲.
41M. Morooka and S. Fukasako, Jpn. J. Appl. Phys., Part 1 35, 3686 共1996兲.
42Y. Li, J. N. Nxumalo, and D. J. Thomson, in Proceedings of the 4th
International Workshop on Ultra Shallow Junctions, Raleigh, NC, edited
by M. Current, M. Kump, and G. McGuire 共1997兲, p. 63.1.
43Y. Huang and C. C. Williams, J. Vac. Sci. Technol. B 12, 369 共1994兲.
44C. C. Williams, Annu. Rev. Mater. Sci. 29,471共1999兲.
45Y. Huang, C. C. Williams, and J. Slinkman, Appl. Phys. Lett. 66,344
共1995兲.
46Y. Huang, C. C. Williams, and M. A. Wendman, J. Vac. Sci. Technol. A
14, 1168 共1996兲.
47T. Clarysse, M. Caymax, P. De Wolf, T. Trenkler, W. Vandervorst, J. S.
McMurray, J. Kim, C. C. Williams, J. G. Clark, and G. Neubauer, J. Vac.
Sci. Technol. B 16,394共1998兲.
48R. Stephenson, A. Verhulst, P. De Wolf, M. Caymax, and W. Vander-
vorst, Appl. Phys. Lett. 73,2597共1998兲.
49J. J. Kopanski, J. F. Marchiando, D. W. Berning, R. Alvis, and H. E.
Smith, J. Vac. Sci. Technol. B 16, 339 共1998兲.
50J. S. McMurray, J. Kim, C. C. Williams, and J. Slinkman, J. Vac. Sci.
Technol. B 16, 344 共1998兲.
51W. Timp, M. L. O’Malley, R. N. Kleiman, and J. P. Garno, Tech. Dig.
Int. Electron Devices Meet., 555 共1998兲.
52Y. Huang, C. C. Williams, and H. Smith, J. Vac. Sci. Technol. B 14,433
共1996兲.
53J. F. Marchiando, J. J. Kopanski, and J. R. Lowney, J. Vac. Sci. Technol.
B16, 463 共1998兲.
54G. Yu, P. Griffin, and J. Plummer, Tech. Dig. Int. Electron Devices
Meet., 717 共1998兲.
55R. N. Kleiman, M. L. O’Malley, F. H. Baumann, J. P. Garno, and G. L.
Timp, Tech. Dig. Int. Electron Devices Meet., 671 共1997兲.
56C. J. Kang, C. K. Kim, J. D. Lera, Y. Kuk, K. M. Mang, J. G. Lee, K. S.
Suh, and C. C. Williams, Appl. Phys. Lett. 71,1546共1997兲.
57W. Timp, M. L. O’Malley, R. N. Kleiman, and J. P. Garno, Tech. Dig.
Int. Electron Devices Meet., 555 共1998兲.
58H. Edwards, R. Mahaffy, C. K. Shih, R. McGlothlin, R. San Martin, M.
Gribelyuk, R. S. List, and V. A. Ukraintsev, Appl. Phys. Lett. 72, 698
共1998兲.
59V. A. Ukraintsev, F. R. Potts, R. M. Wallace, L. K. Magel, H. Edwards,
and M.-C. Chang, Proceedings of the International Conference on Char-
acteristics and Metrology for ULSI Technology 共1998兲.
60B. Michel, W. Mitzutani, R. Schierle, A. Jarosh, W. Knop, H. Benedick-
ter, W. Bachtold, and H. Rohrer, Rev. Sci. Instrum. 63, 4080 共1992兲.
61J.-P. Bourgoin, M. B. Johnson, and B. Michel, Appl. Phys. Lett. 65, 2045
共1994兲.
62F. Bordoni, L. Yinghua, B. Spataro, F. Feliciangeli, F. Vasarelli, G. Car-
darilli, B. Antonini, and R. Scrimaglio, Meas. Sci. Technol. 6, 1208
共1995兲.
63M. Nonnenmacher, M. P. O’Boyle, and H. K. Wickramasinghe, Appl.
Phys. Lett. 58, 2921 共1991兲.
64A. Kikukawa, S. Hosaka, and R. Imura, Appl. Phys. Lett. 66,3510
共1995兲.
65A. Kikukawa, S. Hosaka, and R. Imura, Rev. Sci. Instrum. 67, 1463
共1996兲.
66M. Tanimoto and O. Vatel, J. Vac. Sci. Technol. B 14, 1547 共1996兲.
67A. K. Henning, T. Hochwitz, J. Slinkman, J. Never, S. Hoffmann, P.
Kaszuba, and C. Daghlian, J. Appl. Phys. 77,1888共1995兲.
68P. De Wolf, T. Clarysse, W. Vandervorst, L. Hellemans, Ph. Niedermann,
and W. Ha
¨nni, J. Vac. Sci. Technol. B 16,401共1998兲.
69J. N. Nxumalo, D. T. Shimizu, and D. J. Thomson, J. Vac. Sci. Technol.
B14, 386 共1996兲.
70P. De Wolf, T. Clarysse, and W. Vandervorst, J. Vac. Sci. Technol. B 16,
320 共1998兲.
71P. De Wolf, M. Geva, T. Hantschel, W. Vandervorst, and R. B. Bylsma,
Appl. Phys. Lett. 73, 2155 共1998兲.
72P. De Wolf, R. Stephenson, S. Biesemans, Ph. Jansen, G. Badenes, K. De
Meyer, and W. Vandervorst, Tech. Dig. Int. Electron Devices Meet., 559
共1998兲.
73P. De Wolf, W. Vandervorst, H. Smith, and N. Khalil, J. Vac. Sci. Tech-
nol. B, these proceedings.
74T. Trenkler, T. Hantschel, R. Stephenson, P. De Wolf, W. Vandervorst, L.
Hellemans, A. Malave
´,D.Bu
¨chel, E. Oesterschulze, W. Kulisch, P. Nie-
dermann, T. Sulzbach, and O. Ohlsson, J. Vac. Sci. Technol. B, these
proceedings.
368 De Wolf
et al.
: Status and review of 2D carrier profiling 368
J. Vac. Sci. Technol. B, Vol. 18, No. 1, JanÕFeb 2000