Capacitance-voltage profiling: Research-grade approach

versus low-cost alternatives

Neal D. Reynolds, Cristian D. Panda, and John M. Essicka)
Physics Department, Reed College, Portland, Oregon 97202
(Received 21 July 2013; accepted 22 January 2014)
We describe an experiment that implements capacitance-voltage profiling on a reverse-biased
Schottky barrier diode to determine the density of impurity dopants in its semiconductor layer as well
as its built-in electric potential. Our sample is a commercially produced Schottky diode. Three
different experimental setups, one using research-grade instrumentation, the other two using low-cost
alternatives, are given and their results compared. In each of the low-cost setups, phase-sensitive
detection required to measure the sample’s capacitance is carried out using an inexpensive data
acquisition (DAQ) device and a software program that implements a lock-in detection algorithm. The
limitations of the DAQ device being used (e.g., restricted analog-to-digital conversion speed,
inadequate waveform generation capabilities, lack of hardware triggering) are taken into account in
each setup. Excellent agreement for the value of the doping density obtained by the all three setups is
found and this value is shown to be consistent with the result of an independent method (secondary
ion mass spectroscopy). VC 2014 American Association of Physics Teachers.
[http://dx.doi.org/10.1119/1.4864162]

I. INTRODUCTION II. THEORY

Semiconductor physics has proved itself a fertile field for A. Capacitance-voltage profiling of Schottky barrier
advanced instructional laboratory developers. The properties of
semiconductors engender natural student interest as they lie at In this section, we derive the theoretical relations that are
the heart of modern-day technologies such as computers, smart the basis of the capacitance-voltage profiling experimental
phones, and the internet, while, pedagogically, semiconductor technique.20,22,23,25 First, consider a Schottky barrier of
phenomena provide engaging applications of basic concepts in cross-sectional area A that consists of a metal layer in contact
quantum mechanics, electrodynamics, and thermal physics. with an n-type semiconductor (dielectric constant ; for sili-
Hence, many semiconductor-related advanced laboratory con  ¼ 11.7) and define an x-axis whose origin is at the
experiments have been developed. For example, electrical metal-semiconductor interface with its positive direction to-
measurements on forward-biased diodes and Hall devices have ward the semiconductor’s interior (Fig. 1). For the
been used to measure the semiconducting band gap,1–6
Schottky barrier height,7 charge carrier density,8–10 carrier
transport properties,11,12 and carrier statistical distributions.13
Additionally, optical experiments are available to determine the
band gap of bulk,14,15 thin film,16 and quantum-dot17 semicon-
ductor samples, the Maxwell-Boltzmann distribution of charge
carriers,18 and the vibration properties of nanomaterials.19
In this paper, we describe a newly developed instructional
lab experiment that implements capacitance-voltage (CV)
profiling on a reverse-biased Schottky barrier diode to deter-
mine the density of impurity dopants in its semiconductor
layer, as well as its built-in electric potential. In contrast to
the more complicated dual-semiconductor structure of a pn-
junction diode, a Schottky barrier diode consists of a single
semiconductor layer in contact with a metallic layer. Due to
the simplicity of its construction, the functioning of a
Schottky barrier diode can be explained theoretically using
only introductory electrodynamics and semiconductor con-
cepts. As with all semiconductor devices, fabrication of a
high-quality Schottky diode requires specialized expertise
and expensive deposition systems. To avoid this hurdle, we
use a commercially produced Schottky barrier diode in this
project. Additionally, in an effort to contain the cost of the Fig. 1. Band bending in a Schottky barrier (cross-sectional area A) under
capacitance characterization system, we describe several bias of –V0 creates a (shaded) depletion region of width W with space charge
density þeq due to ionized dopant atoms; q may be position-dependent.
options for the setup, ranging from one consisting solely of
When the reverse bias is increased by dVR, an additional charge of
stand-alone research-grade equipment to others that use dQ ¼ þeq(AdW) is created at the tail of the depletion region. The
low-cost op-amp circuitry and affordable computer-based metal-semiconductor interface is at x ¼ 0 and it is assumed each ionized
instrumentation. dopant atom has charge of þe.

196 Am. J. Phys. 82 (3), March 2014 http://aapt.org/ajp C 2014 American Association of Physics Teachers
V 196

This article is copyrighted as indicated in the article. Reuse of AAPT content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
134.10.101.160 On: Thu, 16 Oct 2014 23:22:59
 semiconducting material, we assume the following. First, Ð W¼ ðVR þ Vbi Þ and V(W) ¼ 0. Thus,
boundary conditions: Vð0Þ
positively charged dopant atoms are incorporated in its lat- from VðWÞ  Vð0Þ ¼  0 E  dx, we find
tice structure with a position-dependent volume number den- eq
sity (“doping density”) q(x). Second, its temperature is high VR þ Vbi ¼ 0 W 2 (1)
20
enough so that these dopant atoms are fully ionized, that is,
their extra electrons have all been promoted into the semi- and so the width of the depletion region required to screen
conductor’s conduction band. These negative conduction out the reverse bias VR is
electrons then perfectly compensate the dopant atom’s posi- sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
tive charge so that, in its bulk, the semiconductor is electri- 20 ðVR þ Vbi Þ
W¼ : (2)
cally neutral. Third, at large x the semiconductor is eq0
connected to external circuitry via an ohmic “back contact.”
We further assume that the type of metal is properly cho- If the reverse bias VR is increased by a small amount dVR,
sen so that when the two materials are joined, semiconductor (remembering Vbi is a constant) we find from Eq. (1) that
conduction electrons transfer to the metal’s surface, leaving dVR ¼ ðeq0 =0 ÞW dW, where dW is the increase in the
behind a positively charged layer of uncompensated dopant depletion region’s width. This increase in the depletion
atoms (called the “depletion region”) in the volume of the region width is due to flow of conduction electrons at the
semiconductor nearest the metal. This charge transfer takes edge of the depletion region into the semiconductor’s bulk
place until the resulting space charge produces an electric (and out of the back contact), creating the extra space charge
potential –Vbi at x ¼ 0 (Vbi  0 is termed the “built-in dQ ¼ þeq0 ðA dWÞ required to screen out the voltage change
potential”), which prevents further flow of charge (because –dVR at the metal-semiconductor interface. By definition,
the Fermi levels of the two materials have been equalized). this process produces a capacitive response given by
Then, if an externally applied voltage –VR (where VR  0 dQ eq0 ðAdWÞ 0 A
is called the “reverse bias”) is applied at the metal contact so C ¼ ¼ ; (3)
that the total potential at the metal-semiconductor interface dVR ðeq0 =0 ÞW dW W
is –V0 ¼ – (VR þ Vbi) and the total charge on the metal’s sur- or, using Eq. (2), the Schottky barrier’s capacitance as a
face is –Q, movement of semiconductor conduction electrons function of reverse bias VR is
away from the interface (and out the back contact at large x) rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
will extend the depletion region to a width W. Beyond the e0 q0
C¼A : (4)
depletion region, the semiconductor is neutral and the effect 2ðVR þ Vbi Þ
of –Q is not felt (i.e., the electric field is zero there). This
effect is called dielectric screening and W is the “screening This expression can be re-written as
length.” 1 2
Let’s now analyze the special case of uniform doping density ¼ ðVR þ Vbi Þ; (5)
C2 A2 e0 q0
q(x) ¼ q0 (a constant), where each dopant atom is assumed to
have a single charge þe. If the cross-sectional dimensions of the suggesting the following capacitance-voltage characteriza-
Schottky barrier are much greater than W, then by symmetry we tion method: For a Schottky barrier with uniform doping
can assume the electric field E in the depletion region points in density, a plot of 1/C2 versus VR will yield a straight line
the (negative) x-direction. Using the rectangular Gaussian sur- with slope m ¼ 2=A2 e0 q0 and y-intercept
2
face shown in Fig. 2, with one of its end caps in the neutral bulk b ¼ 2Vbi =A e0 q0 . Then, the doping density and built-in
region (where E ¼ 0) and the other end cap located a distance x potential are found from
from the metal-semiconductor interface, Gauss’s law gives 2
EA ¼ þeq0 AðW  xÞ=0 . For the electric potential V(x) as a q0 ¼ ; (6)
A2 e0 m
function of distance x into the depletion region, we have the two
and
b
Vbi ¼ : (7)
m
The general case of position-dependent doping density
q(x) can be solved as follows: We start with the identity
 
d dV dV d2 V
x ¼ þx 2 ; (8)
dx dx dx dx

and note from Poisson’s equation that

d 2 V þeqðxÞ
¼ : (9)
dx2 0
Fig. 2. Gauss’s Law applied to a Schottky barrier (cross-sectional area A) under
bias of –V0 ¼ –(VR þ Vbi) to determine the electric field E at location x, assum- Substituting Eq. (9) into Eq. (8) and integrating both sides of
ing constant doping density. By symmetry, the electric field E is directed along the resulting expression from x ¼ 0 to x ¼ W then yields
the (negative) x-axis. A (dashed) rectangular Gaussian surface is chosen with
one face in the depletion region, where the space charge density is þeq0; the ð
e W
other face is in neutral bulk region of the semiconductor, where the electric field VR þ Vbi ¼ xqðxÞdx: (10)
is zero. The total charge within the Gaussian surface is dQ ¼ þeq0 A(W – x). 0 0

197 Am. J. Phys., Vol. 82, No. 3, March 2014 Reynolds, Panda, and Essick 197

This article is copyrighted as indicated in the article. Reuse of AAPT content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
134.10.101.160 On: Thu, 16 Oct 2014 23:22:59
 Now if the reverse bias VR is increased by a small amount RL. Taking RS  0, ac circuit analysis then predicts that the
dVR, the depletion region width will change by dW, creating amplitude I of the total ac current flowing in this circuit is
the extra space charge dQ ¼ þeqðWÞðA dWÞ, where q(W) is  
the doping density at the edge of the depletion region when Vac 1
I¼ ¼ Vac þ ixC  Ix þ Iy : (18)
the reverse bias is VR. From Eq. (10), we have Z RL
ðW !
e Hence, relative to the applied ac voltage, the current will
dðVR þ Vbi Þ ¼ d xqðxÞdx ; (11)
0 0
have an in-phase component Ix proportional to 1/RL and a
90 out-of-phase (“quadrature”) component Iy proportional
or to xC. For a high-quality diode, the leakage current is small
(RL  1/xC) so that Ix Iy.
e
dVR ¼ WqðWÞ dW: (12)
0
B. Lock-in detection algorithm
Thus, the capacitive response is given by
As we have seen, for the circuit described by Eq. (18),
dQ þe qðWÞðA dWÞ 0 A phase-sensitive detection of current is required to measure (a
C ¼ ¼ : (13) signal proportional to) capacitance. In our experimental set-
dVR ðe=0 ÞW qðWÞ dW W
ups, this phase-sensitive detection will be accomplished by
Re-writing this expression as 1=C2 ¼ ðW=0 AÞ2 and differ- using a lock-in amplifier,26,27,32–35 which functions as fol-
entiating with respect to VR gives lows: Assume that in response to an ac modulation input
voltage Vac sinðxsig tÞ, an experimental system produces an
  “in-phase” (relative to the input) output signal Vsig sinðxsig tÞ,
d 1 1 dW
¼ 2W ; (14) where Vsig and xsig are the signal’s amplitude and angular
dVR C2 ð0 AÞ2 dVR frequency, respectively. If we construct an “in-phase refer-
ence” sinusoid 2 sinðxref tÞ of amplitude 2 and angular fre-
or, using Eq. (12), quency xref, and take the product of the output signal and
    reference, we obtain
d 1 1 0 2
2
¼ 2
2W ¼ 2 :
dVR C ð0 AÞ eW qðWÞ A e0 qðWÞ 2Vsig sinðxsig tÞsinðxref tÞ ¼ Vsig ½cosðxsig  xref Þt
(15)  cosðxsig þ xref Þt
; (19)
Equation (15) is called the “Profiler’s Equation” and can where we used the identity sin a sin b ¼ 1=2½cosða  bÞ
be used to characterize the spatial distribution of dopants in cosða þ bÞ
. Thus, the multiplicative result is two ac sinu-
the semiconductor as follows: Starting with data for the soids, each of amplitude Vsig, one with the “difference” fre-
Schottky barrier’s capacitance C as a function of reverse bias quency ðxsig  xref Þ and one with the “sum” frequency
VR, a plot of 1/C2 versus VR is constructed. At each value of ðxsig þ xref Þ. Note that for the special case xref ¼ xsig, the
reverse bias VR on this plot, the slope m ¼ dð1=C2 Þ=dVR is difference-frequency sinusoid is the dc voltage Vsig. Hence,
determined and the associated value of capacitance C noted. if the experiment’s output waveform consists of a collection
Then, each value of VR corresponds to probing the doping
density at the distance W from the metal-semiconductor
interface, which is given by Eq. (13) to be
0 A
W¼ : (16)
C
Using Eq. (15), the doping density at this distance W is deter-
mined by
2
qðWÞ ¼ : (17)
A2 e0 m

To carry out capacitance-voltage characterization of a

Schottky barrier diode, a negative dc voltage –VR is applied to
its metal contact with the back contact grounded, producing a
space-charge region of width W in the semiconductor. The
barrier’s capacitance C is then determined by adding a small
ac modulation of amplitude Vac and angular frequency x to
the applied voltage. To account for a small leakage current
through the barrier and the resistance of the semiconductor’s Fig. 3. Equivalent circuit of the CV characterization setup. The reverse-
neutral bulk region, the Schottky diode is modeled as the par- biased Schottky barrier diode is modeled as capacitance C in parallel with
leakage resistance RL. The semiconductor’s neutral bulk region (beyond the
allel combination of the capacitor C and a leakage resistor RL, depletion region) contributes a series resistance RS. For a high-quality diode,
in series with resistance RS (see Fig. 3). In most cases, the the impedance of RL is much greater than that of C and RS is negligible.
semiconductor doping density is large enough so that RS is Thus, the response of the diode to the small ac modulation voltage is pre-
negligible in comparison to the parallel combination of C and dominately due to C.

198 Am. J. Phys., Vol. 82, No. 3, March 2014 Reynolds, Panda, and Essick 198

This article is copyrighted as indicated in the article. Reuse of AAPT content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
134.10.101.160 On: Thu, 16 Oct 2014 23:22:59
 of component sinusoids of various frequencies, by multiply-
ing this waveform by 2sinðxref tÞ and then using low-pass fil-
tering to find only the resultant dc value, one can determine
the amplitude of the “in-phase” sinusoidal component within
the output waveform whose frequency equals that of the
reference.
In a similar way, if the experimental system produces a
“quadrature” output signal Vsig cosðxsig tÞ in response to the
input voltage Vac sinðxsig tÞ, by multiplying this output signal
by the “quadrature reference” 2 cosðxref tÞ and low-pass fil-
tering for the resultant dc value, this dc value will equal Vsig
(i.e., the quadrature signal amplitude).

III. EXPERIMENTAL SETUPS AND RESULTS

Our goal is to demonstrate both research-grade and low-cost
(but accurate) implementations of the capacitance profiling
technique for use in instructional laboratories. We first carry
out the capacitance-voltage (CV) method using research-grade, Fig. 4. Research-grade implementation of capacitance profiling method
stand-alone instrumentation in order to show how the technique using stand-alone instrumentation (Agilent 33210 A function generator, DL
works, and also to establish a baseline of precision by which to Instruments 1211 current preamplifier, Stanford Research System SR830
judge the accuracy of our lower-cost setups. Next, we show lock-in amplifier).
two inexpensive computer-based versions of this experiment,
each based on an affordable USB-interfaced data acquisition
we use, the resulting value for V is on the order of b 
device, which produce excellent results.
ðVac xCÞ ¼ 106 V=A  ½ð30 mV rmsÞ2pð1000 HzÞð109 FÞ

For each variant of the experiment, our sample is a com-

 200 mV rms. This voltage is then read by a
mercially produced Schottky diode. Because the measured
GPIB-interfaced Stanford Research System SR830 lock-in
signal is proportional to the capacitor’s cross-sectional area A,
amplifier. The function generator’s TTL sync output is used
we chose a Schottky diode with a large-area metallic contact
as the lock-in’s reference signal, with zero phase defined by
[Semiconix Semiconductor STPS20120D (Ref. 28)]. The con-
the moment of the negative-going zero crossing of the ac os-
tact is composed of a TiW alloy. We stripped the epoxy
cillation (that is, the moment at which the reverse bias begins
encapsulation from one of these devices29 to expose the diode
to increase). Since, at constant frequency, the lock-in meas-
and used a calibrated microscope to measure the square con-
ures a quadrature voltage that is proportional (via b) to the
tact30 to have a side length of (2.32 6 0.02) mm, yielding
circuit’s quadrature current, which in turn is proportional to
A ¼ 5.38 6 0.09 mm2. On this exposed diode, secondary ion
C [see Eq. (18)], a calibration capacitor36 of known capaci-
mass spectroscopy (SIMS)31 established that its n-type silicon
tance C0 is substituted in the circuit for the Schottky diode
layer has a phosphorous doping density of approximately
and the resultant quadrature voltage output V0y is measured.
3  1015 dopants/cm3, while no arsenic (another common
Then, with the diode replaced back into the circuit, its capac-
n-type dopant atom in silicon) was detected.
itance C in response to a particular reverse bias is determined
At zero applied bias, the capacitance of the STPS20120D
by measuring the quadrature voltage Vy and using the pro-
diode is on the order of 1000 pF. For all of the experiments
portionality relation: C=C0 ¼ Vy =V0y . Data taking is con-
described in this paper, the ac modulation amplitude and fre-
trolled by a LabVIEW program.
quency are Vac ¼ 30 mV rms and f ¼ 1000 Hz, respectively,
To account for small (on the order of a few degrees in most
and the reverse bias VR is scanned over the range from
cases) phase shifts due to other sources (e.g., cabling, ampli-
0.0–9.9 V in increments of 0.1 V. To allow for settling of the
fiers) than the capacitances of interest, a further refinement
measured signal, a wait of 10 lock-in time constants is taken
called “autophasing” can be included as follows: With the
before reading each data point during the scan. The modula-
calibration capacitor replacing the Schottky diode in the cir-
tion amplitude is chosen to be less than the reverse-bias in-
cuit, record Vx0 and Vy0, which are the lock-in’s in-phase and
crement so that each value of VR in our scan probes a unique
quadrature readings for the known calibration capacitor C0,
position in our sample; the choice of frequency is dictated by
respectively. Then, assuming the calibration capacitor has a
the maximum sampling rate of the USB-6009 DAQ device
purely capacitive impedance (that is, it has zero leakage cur-
(see discussion below).
rent), these two readings determine a vector in the complex
impedance plane that defines the direction of the sample’s
A. Research-grade implementation capacitive response. With the diode back in the circuit, its
in-phase and quadrature voltages Vx and Vy are measured.
A schematic diagram of our research-grade experimental
These two readings determine a vector describing the sam-
setup is shown in Fig. 4. A USB-interfaced Agilent 33210 A
ple’s ac response in the complex impedance plane. By mathe-
function generator applies the ac oscillation of small ampli-
matically finding the component of the sample’s response
tude Vac and frequency f along with a dc offset –VR to the
along the purely capacitive direction (via a dot product), the
Schottky diode’s metal contact. The other end of the diode is
relation for the sample’s capacitance is obtained as
connected to a DL Instruments 1211 current preamplifier,
which is a virtual electrical ground and converts the diode’s ðVx Vx0 þ Vy Vy0 Þ
current I to a voltage V ¼ bI, where the proportionality con- C¼ 2 þ V2
C0 : (20)
stant b ¼ 106 V/A. For the diode and data-taking parameters Vx0 y0

199 Am. J. Phys., Vol. 82, No. 3, March 2014 Reynolds, Panda, and Essick 199

This article is copyrighted as indicated in the article. Reuse of AAPT content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
134.10.101.160 On: Thu, 16 Oct 2014 23:22:59
 The uncertainties in our q and Vbi values obtained from
Eqs. (6) and (7) were estimated as follows. First, the influ-
ence of random measurement error in introducing uncer-
tainty in our straight-line fit of the 1/C2 versus VR plot was
gauged by comparing the results of ten identical runs of the
experiment. We found that such random errors contributed
uncertainty in m and b on the order of only 0.1%. Thus, in
using Eq. (6), the uncertainty in the contact area is the domi-
nant contribution to dq. With dA/A  0.02, we get
dq ¼ 0.1  1015 dopant/cm3. To determine Vbi via Eq. (7),
only the highly accurate m and b values are involved. With
dm=m  db=b  0:001, our model predicts dVbi  0:001 V.
However, one might question whether our model for the
Schottky barrier, which ignores secondary effects such as se-
ries resistance and small temperature corrections, describes a
Fig. 5. Experimental data (C vs VR and 1/C2 vs VR) taken at room tempera- real-life diode to this level of accuracy (literature values for
ture on a STPS20120D Schottky diode using research-grade instrumentation Vbi obtained from CV measurements are typically given with
(time constant of lock-in is s ¼ 300 ms). The latter plot determines the only two digits of precision).20,24
diode’s constant doping density and built-in potential to be 2.7  1015 dop- Using these same CV data, Eqs. (16) and (17) yield the
ants/cm3 and 0.622 V, respectively. Error bar for each data point is smaller q(W) versus W plot shown in Fig. 6. Also shown in this plot
than symbol used to represent point.
are the SIMS data for phosphorous dopant density versus dis-
tance from the metal-semiconductor interface taken on this
Commercial research-grade lock-in amplifiers commonly sample. The SIMS detection limit for phosphorous is 1  1015
offer autophasing with the push of a button. For software- dopant/cm3, so the SIMS data taken on our sample are some-
based lock-ins, autophasing can be accomplished by includ- what noisy. Averaged over the spatial region from W ¼ 0.5 to
ing Eq. (20) in the software program. 1.0 lm, the SIMS doping density is (3 6 2)  1015 dop-
The room-temperature capacitance versus reverse bias ant/cm3. Thus, the doping densities determined by the capaci-
data acquired on a STPS20120D diode, along with the result- tance and SIMS characterization techniques are consistent.
ing 1/C2 vs VR plot, are shown in Fig. 5. The straight-line Finally, we observe in Fig. 6 that the capacitance profiling
character of the latter plot indicates that the diode’s doping value for q(x) becomes more noisy as W increases. This
density is constant over the spatial region profiled. Using well-known effect21,38 is explained as follows. From
Eqs. (6) and (7), the slope and y-intercept of this plot deter- Eq. (17), the uncertainty dq in the doping density determined
mine that the doping density in this region and the diode’s at each location x ¼ W is given by dq/q ¼ dm/m, where dm is
built-in potential are 2.7  1015 dopant/cm3 and 0.622 V, the uncertainty in determining the slope m ¼ dð1=C2 Þ=dVR
respectively (in semiconductor physics, it is typical to give at that location (for this calculation, the same value for A is
doping density in units of dopant/cm3, rather than dop- used at all locations, so the uncertainty in contact area does
ant/m3). Our value for Vbi agrees with published values not contribute to the observed scatter of q-values). Writing m
obtained on titanium-tungsten silicide Schottky contacts.37 in terms of our measured quantities, we see that the slope at

Fig. 6. Spatial profile of phosphorous doping. The capacitance profiling result (solid circles) is found using Eqs. (16) and (17), where the x-axis is plotting W.
The minimum value of W is determined by Vbi. The SIMS result (open circles) gives an average of 3  1015 dopant/cm3 in the region that overlaps with that
profiled by the capacitance method.

200 Am. J. Phys., Vol. 82, No. 3, March 2014 Reynolds, Panda, and Essick 200

This article is copyrighted as indicated in the article. Reuse of AAPT content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
134.10.101.160 On: Thu, 16 Oct 2014 23:22:59
 Fig. 7. LabVIEW code to carry out two-phase lock-in amplifier algorithm. A fast Fourier Transform performs the required low-pass filtering to obtain the dc
value, which is the zero-index element in the array output by the FFT.vi icon.

x ¼ W is calculated by m  ð2=C3 ÞDC=DVR , where DC by a simple current-to-voltage (I-to-V) op-amp circuit with a
and DVR are the differences between neighboring values of 106 X feedback resistor so that b ¼ –106 V/A. In addition,
capacitance and reverse bias at that location, respectively. In the required phase-sensitive detection is carried out using a
our data scans, DVR ¼ 0.100 V at all locations and so the computer-based lock-in amplifier40–42 consisting of an inex-
uncertainty in this quantity cannot explain the increase in dq pensive data acquisition (DAQ) device and a LabVIEW soft-
at increasing W. Conversely, as shown in Fig. 5, the CV ware program. The central features of this computer-based
curve flattens out at larger biases and thus the difference DC lock-in are as follows: First, triggered by the negative-going
in neighboring capacitance values is smaller and more sus- transition of a function generator’s TTL (or square-wave)
ceptible to random measurement error as VR (and its associ- sync output, the DAQ device acquires N (a power of two)
ated W) increases. Keeping just this relevant contribution to samples of the I-to- V circuit’s voltage output. The sampling
uncertainty, we get dm=m  dðDCÞ=DC. Since DC is the dif- rate is chosen to be fsampling ¼ Npoint  f , where Npoint is the
ference between two neighboring capacitance p values,
ffiffiffi each number of samples to acquire during one reference cycle
with an uncertainty of dC, we find dðDCÞ ¼ 2dC. Also, and f is the reference frequency. A total of Ncycle reference
from Eq. (13), we have DC  ð0 A=W 2 ÞDW. Putting cycles are acquired so that the total number of acquired
Eq. (12) into this expression, we find DC  ð2 20 A=eqW 3 Þ voltage samples is N ¼ Npoint  Ncycle . Since the digitizing
DVR , and thus process is triggered at the our defined zero-phase angle, in
pffiffiffi software we create two copies of this acquired data wave-
dq dm dðDCÞ 2 dC form and multiply one copy by the “in-phase” reference
¼   2 sinð2pftÞ and the other copy by the “quadrature” reference
q m DC ð2 20 A=eqW 3 ÞDVR
pffiffiffi !  2 cosð2pftÞ. A fast Fourier transform is then taken of each
2e qdC of these arrays (hence the reason N is chosen to be a power
 2 2 W3: (21)
 0 A DVR of two) and the dc components of each picked out, resulting
in the in-phase and quadrature voltage amplitude at frequency
In our scans, dC and DVR are constant and q is found to be f in the original digitized waveform. Thus, the frequency
uniform in our sample. Thus, Eq. (21) predicts that dq will be bandwidth of our output signal is on the order of the FFT’s
proportional W3, explaining why the doping density becomes frequency resolution39 Df ¼ fsampling =N. We define the time
markedly more noisy as we profile deeper into the sample constant of our lock-in algorithm to be s  1=Df , so
(alternate profiling methods have been developed that use s ¼ N=fsampling ¼ ðNpoint Ncycle Þ=ðNpoint f Þ ¼ Ncycle =f . Figure 7
constant electric field increments, rather than constant reverse illustrates how this lock-in algorithm is programmed in
bias steps, which somewhat alleviate this noise problem at LabVIEW.43
larger profiling depths). From the data in Fig. 6 we determine We will describe how to carry out the above-described
that in the range W ¼ 2.00–2.25 lm, dq ¼ 4.4  1013 dop- scheme with two different commonly used low-cost DAQ
ant/cm3. Using this value, along with the other known quanti- devices: the USB-6009 and the myDAQ.44 In each case, the
ties in Eq. (21), we find that dC ¼ 0.02 pF is responsible for manner in which the scheme is implemented must be adapted
the observed scatter in q. to the limitations of the DAQ device.

B. Low-cost implementation 1. Setup using the USB-6009

We now demonstrate that similar results can be obtained The USB-6009 device performs 14-bit analog-to-digital
from two different low-cost experimental setups. In each of conversions of an incoming signal at rates up to 48,000
these setups, the commercial current preamplifier is replaced Samples per second (S/s). Each N-sample acquisition can be

201 Am. J. Phys., Vol. 82, No. 3, March 2014 Reynolds, Panda, and Essick 201

This article is copyrighted as indicated in the article. Reuse of AAPT content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
134.10.101.160 On: Thu, 16 Oct 2014 23:22:59
 Fig. 9. Experimental data (1/C2 vs VR) obtained at room temperature on a
STPS20120D Schottky diode using USB-6009 DAQ device (time constant
of lock-in is s ¼ 510 ms). This plot determines the diode’s constant doping
Fig. 8. Low-cost implementation of capacitance profiling using USB-60009 density and built-in potential to be 2.6  1015 dopant/cm3 and 0.613 V. Error
DAQ device. The op-amp (LF411) circuit serves as a current preamplifier bar for each data point is smaller than symbol used to represent point.
and the lock-in algorithm is carried out using a hardware-triggered DAQ de-
vice and LabVIEW software. Because the DAQ device has minimal wave- density over the region profiled and again is in excellent agree-
form generation capabilities, a stand-alone computer-interfaced Agilent
33210 A function generator is used.
ment with the research-grade results (Fig. 6). We note that, as
described by Eq. (21), the doping density determination
becomes more noisy as the profile probes deeper into the sam-
hardware triggered by a TTL signal. The device’s ple. This noise is more noticeable in Fig. 10 than in Fig. 6
programmable-gain amplifier allows for eight possible input because dC is larger for the USB-6009 setup. From the data in
voltage ranges over which to spread the 14-bit resolution. Fig. 10, we determine that in the range W ¼ 2.00–2.25 lm,
Since our input signal is on the order of a few hundred milli- dq ¼ 1.6  1014 dopant/cm3. Using this value, along with the
volts, we choose the (device’s most sensitive) 61 V range. other known quantities in Eq. (21), we find that dC ¼ 0.07 pF is
Additionally, given the 48 kS/s maximum sampling rate and responsible for the observed scatter in q.
that fact that we need the number of samples per cycle
(Npoint) to be a power of two, we chose our reference fre- 2. Setup using the myDAQ
quency to be 1000 Hz. Then, Npoint can be 32, the minimum
value we feel necessary to properly describe the 1000-Hz The myDAQ device performs 16-bit analog-to-digital con-
signal. Finally, this DAQ device possesses only modest digi- versions of an incoming signal at rates up to 200,000 S/s.
tal-to-analog conversion capabilities. With a maximum ana- The device’s programmable-gain amplifier allows for two
log output update rate of 150 Hz, the USB-6009 cannot possible input voltage ranges. We choose the (device’s most
produce the reference signal we require for our experiment. sensitive) 62 V range. However, the device offers no hard-
Hence, we retain the USB-interfaced Agilent 33210 A func- ware triggering capability for these digitizing operations.
tion generator for this setup, as shown in Fig. 8. Thus, the triggering for the voltage acquisitions must be
With f ¼ 1000 Hz, we chose Npoint ¼ 32 and Ncycle ¼ 512. done in software. Finally, this DAQ device can also perform
Then, fsampling ¼ 32,000 S/s, N ¼ 16,384 (¼214), and s ¼ 0.51 s.
Using these parameters, the room-temperature capacitance ver-
sus reverse bias data was acquired on a STPS20120D diode.
The resulting 1/C2 vs VR plot is shown in Fig. 9. The
straight-line character of this plot indicates that the diode’s
doping density is constant over the spatial region profiled.
Using Eqs. (6) and (7), the slope and y-intercept of this plot
determine that the doping density in this region and the diode’s
built-in potential are 2.6  1015 dopant/cm3 and 0.613 V,
respectively. These results are in excellent agreement with the
results obtained using research-grade instrumentation.
As with the research-grade setup, the uncertainties in our
q and Vbi values were estimated by first comparing the
results of ten identical runs of the experiment. We found that
such random errors contributed uncertainties in m and b on
the order of only 0.5%. So, again, the uncertainty in the con-
tact area is the dominant contribution to dq, yielding
dq ¼ 0.1  1015 dopant/cm3. With dm=m  db=b  0:005,
we predict dVbi  0.004 V. Fig. 10. Spatial profile of phosphorous doping obtained using a USB-6009
Using these same CV data, Eqs. (16) and (17) yield the q(W) DAQ device. A constant value for q(x) of about 2.6  1015 dopant/cm3 is
vs W plot shown in Fig. 10. This plot indicates constant doping indicated over the spatial range W ¼ 0.5–2.3 lm.

202 Am. J. Phys., Vol. 82, No. 3, March 2014 Reynolds, Panda, and Essick 202

This article is copyrighted as indicated in the article. Reuse of AAPT content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
134.10.101.160 On: Thu, 16 Oct 2014 23:22:59
 Fig. 13. Experimental data (1/C2 vs VR) obtained at room temperature on a
STPS20120D Schottky diode using a myDAQ DAQ device (time constant
of lock-in is s ¼ 130 ms). This plot determines the diode’s constant doping
density and built-in potential to be 2.6  1015 dopant/cm3 and 0.623 V. Error
bar for each data point is smaller than symbol used to represent point.
Fig. 11. Low-cost implementation of capacitance profiling using a myDAQ
DAQ device. The op-amp (LF411) circuit serves as a current preamplifier
and the lock-in algorithm is carried out using a software-triggered DAQ de-
vice and LabVIEW software. The waveform generation function of the
to suppress digital quantization. Figure 12 illustrates how
myDAQ device is used to create a modulated reverse bias. The op-amp low- software triggering is carried out in LabVIEW.
pass filter with f3dB  5 kHz suppresses digitizing steps on small-amplitude With f ¼ 1000 Hz, we chose Npoint ¼ Ncycle ¼ 128. Then,
ac modulation. fsampling ¼ 128,000 S/s, N ¼ 16,384 (¼214), and s ¼ 0.13 s.
Using these parameters, the room-temperature capacitance
digital-to-analog conversions at rates up to 200,000 S/s on versus reverse bias data was acquired on a STPS20120D
two analog output (AO) channels. We use these two AO diode. The resulting 1/C2 vs VR plot is shown in Fig. 13.
channels, each operating at 200 kS/s, to produce our required Again, the straight-line character of this plot indicates that
modulated bias voltage as well as a digital reference signal, the diode’s doping density is constant over the spatial region
whose transitions are in-phase with the bias voltage’s ac profiled. Using Eqs. (6) and (7), the slope and y-intercept of
modulation. As shown in Fig. 11, a 60.5 V square-wave ref- this plot determine that the doping density in this region and
erence signal generated by one of the analog output channels the diode’s built-in potential are 2.6  1015 dopants/cm3 and
is directly connected to, and read by, one of the analog input 0.623 V, respectively. These results are in excellent agree-
channels. The moment at which a transition of this square ment with those obtained using the other setups.
wave occurs is determined by searching the acquired wave- As before, the uncertainties in our q and Vbi values were
form in software, enabling lock-in detection. Also, the modu- estimated by first comparing the results of ten identical runs
lated bias voltage produced by the other AO channel is of the experiment. We found that such random errors con-
passed through an (inverting) op-amp low-pass filter in order tributed uncertainties in m and b on the order of 0.6%. Thus,

Fig. 12. LabVIEW code to carry out software analog triggering.

203 Am. J. Phys., Vol. 82, No. 3, March 2014 Reynolds, Panda, and Essick 203

This article is copyrighted as indicated in the article. Reuse of AAPT content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
134.10.101.160 On: Thu, 16 Oct 2014 23:22:59
 verified its results are consistent with another characteriza-
tion technique (secondary ion mass spectroscopy). Further,
we have shown that, in spite of their limitations such as re-
stricted analog-to-digital conversion speed, inadequate
waveform generation capabilities, and lack of hardware trig-
gering, inexpensive DAQ devices can be used to accurately
carry out capacitance profiling on semiconductor samples.
These low-cost solutions make the introduction of such
measurements an attractive option for advanced laboratory
projects.

ACKNOWLEDGMENTS
This work is dedicated to the memory of Professor J.
David Cohen.

a)
Electronic mail: jessick@reed.edu
1
Fig. 14. Spatial profile of phosphorous doping obtained using the myDAQ J. W. Precker and M. A. da Silva, “Experimental estimation of the band
DAQ device. The noise in smaller capacitance values (at greater depletion- gap in silicon and germanium from the temperature-voltage curve of
region widths) causes the evaluation of the derivative in the Profiler’s diode,” Am. J. Phys. 70, 1150–1153 (2002).
2
Equation to be inaccurate at W > 1.5 lm. H. Jaeger, M. J. Pechan, and D. K. Lottis, “Materials physics: A new con-
temporary undergraduate laboratory,” Am. J. Phys. 66, 724–730 (1998).
3
A. Sconza, G. Torzo, and G. Viola, “Experiment on the physics of the PN
dA is the dominant contribution to dq, yielding junction,” Am. J. Phys. 62, 66–70 (1994).
dq ¼ 0.1  1015 dopant/cm3. With dm=m  db=b  0:006, 4
L. Kirkup and F. Placido, “Undergraduate experiment: Determination of the
we predict dVbi  0.005 V. 5
band gap in germanium and silicon,” Am. J. Phys. 54, 918–920 (1986).
Using these same CV data, Eqs. (16) and (17) yield the B. D. Sukheeja, “Measurement of the band gap in silicon and germanium,”
Am. J. Phys. 51, 72 (1983).
q(W) vs W plot shown in Fig. 14. Again, we see that, per 6
P. J. Collings “Simple measurement of the band gap in silicon and
Eq. (21), the doping density determination becomes more germanium,” Am. J. Phys. 48, 197–199 (1980).
noisy as the profile probes deeper into the sample. This time 7
N. J. Pinto, “Two experiments in physics based on electrospun polymer
the noise is significantly larger than in Figs. 6 and 10, indicat- nanofibers,” Am. J. Phys. 76, 1163–1167 (2008).
8
ing that dC is larger for the myDAQ setup in comparison to J. Brody, Z. Dong, and T. Dennen, “Inexpensive fabrication of silicon Hall
the two other setups used. From the data in Fig. 14, we deter- devices,” Am. J. Phys. 74, 240–242 (2006).
9
D. W. Preston and E. R. Dietz, The Art of Experimental Physics
mine that in the range W ¼ 2.00–2.25 lm, dq ¼ 4.4  1014
(Academic Press, San Diego, 2003), Chap. 17, pp. 303–315.
dopant/cm3. Using this value, along with the other known 10
A. C. Melissinos and J. Napolitano, Experiments in Modern Physics, 2nd
quantities in Eq. (21), we find that dC ¼ 0.2 pF is responsible ed. (Academic Press, San Diego, 2003), Chap. 3, Section 3, pp. 441–442.
for the observed scatter in q. This larger uncertainty in the ca- 11
R. H. Chow, “Study on diffusion and recombination of minority carriers by
pacitance measurement (in comparison to the other two set- the method of photoconductive decay,” Am. J. Phys. 52, 842–844 (1984).
12
ups) most likely derives from the lower-level of precision A. Sconza, G. Galet, and G. Torzo, “An improved version of the Haynes-
inherent in the software triggering used here. Shockley experiment with electrical or optical injection of the excess
carriers,” Am. J. Phys. 68, 80–87 (2000).
13
M. D. Sturge and Song Bac Toh, “An experiment to demonstrate the ca-
nonical distribution,” Am. J. Phys. 67, 1129–1131 (1999).
14
IV. CONCLUSION J. M. Essick and R. T. Mather, “Characterization of a bulk semiconductors
band gap via a near-absorption edge optical transmission experiment,”
A summary of our results is given in Table I. Here, we Am. J. Phys. 61, 646–649 (1993).
15
see excellent agreement between the three setups, with pos- E. Redondo, A. Ojeda, G. Gonzalez Daz, and I. Martil, “A laboratory experi-
sibly slight systemic shifts due to the different lock-in ment with blue light-emitting diodes,” Am. J. Phys. 65, 371–376 (1997).
16
detection scheme, cabling, and breadboard wiring used in I. Martil and G. Gonzalez Daz, “Undergraduate laboratory experiment:
Measurement of the complex refractive index and the band gap of a thin
each case. film semiconductor,” Am. J. Phys. 60, 83–86 (1992).
In conclusion, using research-grade instrumentation, we 17
T. A. Laubach, L. A. Elizondo, P. J. McCann, and S. Gilani, “Quantum
have demonstrated the capacitance profiling method and dotting the “i” of Inquiry: A guided inquiry approach to teaching nano-
technology,” Phys. Teach. 48, 186–188 (2010).
18
Table I. Summary of results on STPS20120D Schottky diode. J. Brody, D. Weiss, and P. Young, “Observing the Maxwell-Boltzmann
distribution in LED emission spectra,” Am. J. Phys. 78, 933–935 (2010).
19
Setup q (1015 dopants/cm3) Vbi (V) dC (pF) A. Singha P. Dhar, and A. Roy “A nondestructive tool for nanomaterials:
Raman and photoluminescence spectroscopy,” Am. J. Phys. 73, 224–233
Research-grade 2.7 6 0.1 0.622 6 0.001 0.02 (2005).
20
USB-6009a 2.6 6 0.1 0.613 6 0.004 0.07 D. K. Schroder, Semiconductor Material and Device Characterization, 3rd
ed. (John Wiley and Sons, New Jersey, 2006), Chaps. 2 and 3.
myDAQb 2.6 6 0.1 0.623 6 0.005 0.2 21
D. K. Schroder, Semiconductor Material and Device Characterization, 3rd
SIMSc 362 – – ed. (John Wiley and Sons, New Jersey, 2006), p. 94.
22
B. Streetman and S. Banerjee, Solid State Electronic Devices, 5th ed.
a
Hardware triggering, audio-range 14-bit data acquisition, inadequate wave- (Prentice-Hall, New Jersey, 2000), pp. 202–205.
form generation. 23
S. M. Sze and K. K. Ng, Physics of Semiconductor Devices, 3rd ed.
b
Software triggering only, excellent 16-bit data acquisition and waveform (Wiley-Interscience, New Jersey, 2006), Chaps. 2 and 3.
24
generation rates. S. M. Sze and K. K. Ng, Physics of Semiconductor Devices, 3rd ed.
c
Performed by Evans analytical group. (Wiley-Interscience, New Jersey, 2006) p. 179–180

204 Am. J. Phys., Vol. 82, No. 3, March 2014 Reynolds, Panda, and Essick 204

This article is copyrighted as indicated in the article. Reuse of AAPT content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
134.10.101.160 On: Thu, 16 Oct 2014 23:22:59
 25 34
F. R. Goodman, D. L. Heald, and R. C. Neville, “Laboratory experi- J. H. Scofield, “Frequency domain description of a lock-in amplifier,” Am.
ment in semiconductor surface-field effects,” Am. J. Phys. 42, 572–579 J. Phys. 62, 129–133 (1994).
35
(1974). R. Wolfson, “The lock-in amplifier: A student experiment,” Am. J. Phys.
26
P. Horowitz and W. Hill, The Art of Electronics, 2nd ed. (Cambridge U.P., 59, 569–572 (1991).
36
New York, 1989), p. 1031. For the calibration capacitor, a 1-nF ceramic capacitor was used. Its actual
27
D. W. Preston and E. R. Dietz, The Art of Experimental Physics capacitance value was determined to 61% using an Agilent U1732B
(Academic Press, San Diego, 2003) Chaps. 17, p. 367. Handheld LCR Meter.
28 37
Available from Mouser P/N 511-STPS20120D. S. E. Babcock and K. N. Tu, “Titanium-tungsten contacts to Si: The
29
While wearing proper protective lab wear (including gloves, safety glass, effects of alloying on Schottky contact and on silicide formation,” J. Appl.
face shield) and working under a fume hood, the epoxy encapsulation of Phys. 53, 6898–6905 (1982).
38
an electronic component such as a Schottky diode can be dissolved by P. Blood, “Capacitance-voltage profiling and the characterisation of III-V
applying drops of fuming nitric acid to the component while at elevated semiconductors using electrolyte barriers,” Semicond. Sci. Technol. 1,
temperature (100  C). After the electronic component is exposed, the part 7–27 (1986).
39
should be rinsed in methanol. J. Essick, Hands-On Introduction to LabVIEW for Scientists and
30
Good results were also obtained on a smaller area Schottky diode (Vishay Engineers, 2nd ed. (Oxford U.P., New York, 2012) Chap. 10.
40
844-11DQ10) with cylindrical encapsulation. We found that the encapsu- R. Wolfson and D. Mullen, “Spreadsheet lock-in amplifier,” Am. J. Phys.
lation on this diode can be carefully cracked open using a vise and that, to 78, 1227–1229 (2010).
41
a good approximation, the Schottky contact area is equal to the cross- Y. Kraftmakher, “Lock-in detection with DataStudio,” Am. J. Phys. 74,
sectional area of the diode’s lead wire (0.49 mm2). 207–210 (2006).
31 42
Performed by Evans Analytical Group, Sunnyvale, CA. P. J. Moriarty, B. L. Gallagher, C. J. Mellor, and R. R. Baines, “Graphical
32
W. Yang, “Teaching phase-sensitive demodulation for signal condi- computing in the undergraduate laboratory: Teaching and interfacing with
tioning to undergraduate students,” Am. J. Phys. 78, 909–915 (2010). LabVIEW,” Am. J. Phys. 71, 1062–1704 (2003).
33 43
K. G. Libbrecht, E. D. Black, and C. M. Hirata, “A basic lock-in amplifier See supplementary material at http://dx.doi.org/10.1119/1.4864162 for
experiment for the undergraduate laboratory,” Am. J. Phys. 71, 1208–1213 the LabVIEW programs.
44
(2003). Both DAQ devices are available from National Instruments, Austin, TX.

205 Am. J. Phys., Vol. 82, No. 3, March 2014 Reynolds, Panda, and Essick 205

This article is copyrighted as indicated in the article. Reuse of AAPT content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
134.10.101.160 On: Thu, 16 Oct 2014 23:22:59