An Improved Method For Differential Conductance Measurements
An Improved Method For Differential Conductance Measurements
PA P E R
Introduction
As modern electronics continue to shrink, researchers are increasingly looking
to nanotechnology as the basis for the next breakthrough in device size and
power consumption. Indeed, as semiconductor structures are made smaller and
smaller, the distinction between small silicon geometries and large molecules
becomes blurred. Approached from either direction, the consequences are the
same. Quantum behavior such as tunneling begins to play an important role in
the electrical characteristics. In the macroscopic world, conductors may have
obeyed Ohm’s Law (Figure 1a), but in the nanoscale, Ohm’s definition of
resistance is no longer relevant (Figure 1b). Because the slope of the I-V curve
is no longer a fundamental constant of the material, a detailed measurement
of the slope of that I-V curve at every point is needed to study nanodevices.
This plot of differential conductance (dG = dI/dV) is the most important
measurement made on small scale devices, but presents a unique set of
challenges.
(1) Perform a current-voltage sweep (I-V curve) and take the mathematical derivative,
or
I-V Technique
The I-V sweep technique has the advantage that it is easier to set up and control. It only
requires one source and one measurement instrument, which makes it relatively easy to
coordinate and control. The fundamental problem is that even a small amount of noise
becomes a large noise when the measurements are differentiated.
A G r e a t e r M e a s u r e o f C o n f i d e n c e
Figure 2a. I-V curve Figure 2b. Differentiated I-V Figure 2c. 100 curves, averaged
curve together
Figure 2a shows an I-V curve, a series of sourced and measured values (V1, I1), (V2,
I2), etc. Several techniques can be used to differentiate this data, but the simplest and most
common uses the slope between every pair of consecutive data points. For example, the first
point in the differential conductance curve would be (I2–I1)/(V2–V1). Because of the small
differences, a small amount of noise in either the voltage or current causes a large uncertainty
in the conductance. Figure 2b shows the differentiated curve and the noise, which is
unacceptably large for most uses. To reduce this noise, the I-V curve and its derivative can be
measured repeatedly. Noise will be reduced by , where N is the number of times the curve
is measured. After 100 repetitions, which can take more than an hour in a typical application,
it is possible to reduce the noise by a factor of 10, as shown in Figure 2c. While this could
eventually produce a very clean data set, researchers are forced to accept high noise levels,
because measuring 10,000 times to reduce the noise 100× would take far more time than is
usually available. Thus, while the I-V curve technique is simple, it forces a trade-off between
high noise and very long measurement times.
The AC Technique
The AC technique superimposes a low amplitude AC sine wave on a stepped DC bias,
as shown in Figure 3. The problem with this method is that, while it provides a small
improvement in noise over the I-V curve technique, it imposes a large penalty in terms of
system complexity.
A G r e a t e r M e a s u r e o f C o n f i d e n c e
Applied
Stimulus
Time
Figure 3. The AC technique measures the response to a sine stimulus while sweeping the DC bias through
the device’s operating range.
A G r e a t e r M e a s u r e o f C o n f i d e n c e
so the series resistors used to combine the AC and DC must be higher impedance than the
device, which is unknown until the measurement is made. An example system is shown in
Figure 4. Assembling such a system requires, in addition to time, extensive knowledge of
electrical circuitry. Time and technical resources often challenge material researchers and
students of other disciplines who are interested in characterizing devices.
AC Technique
T
R1 R2
DC AC
V
Voltage V
Voltage
DC
DUT Lock-In
V
Voltmeter
VDC dV
DC Lock-In
Meter
IDC dI
Figure 4. Block diagram of the AC technique, which reduces noise, but is very expensive (in equipment and
time) and complicated.
A G r e a t e r M e a s u r e o f C o n f i d e n c e
synchronized with the nanovoltmeter via the Trigger Link cable. After measuring the voltage
at each current step, the nanovoltmeter calculates the delta voltage between consecutive steps.
Each delta voltage is averaged with the previous delta voltage to calculate the differential
voltage, dV. The differential conductance, ΔG, can now be derived using dI/dV. The math
for this is covered in more detail in the white paper titled “Achieving Accurate and Reliable
Resistance Measurements in Low Power and Low Voltage Applications,” which is available
online at www.keithley.com.
AC Technique
T
R1 R2
DC AC
V
Voltage V
Voltage
VDC dV
DUT
DC Lock-In
Meter
IDC dI
Meas
V5 Meas
Meas V6
V3 Meas
Meas V4
V1 Meas
V2
2182A
Applied V-Meas
Stimulus Delay ∆I ∆I
Figure 6a. Applied current during Keithley Figure 6b. Detail of applied current and
Differential Conductance measurement measured device voltage
A G r e a t e r M e a s u r e o f C o n f i d e n c e
Why Choose the Keithley Method
The new Keithley method for differential conductance provides low noise results at least
10× faster than previous methods. There are many other benefits of selecting the Keithley
6220/6221 Current Source and 2181A Nanovoltmeter to make differential conductance
measurements. One is that significantly fewer instruments are required (see Figure 5).
The Keithley method requires two instruments (the 6220/6221 and 2182A) while the AC
technique requires eight instruments and external circuitry.
Devices must be characterized over a user-defined set of currents (or voltages). When
the user-defined currents are small, the low current capability of the 622X is unmatched by
any user-made system in source accuracy, source noise, and capability to do guarding, which
reduces DC leakage and improves system response time. It is also able to source current
accurately well below 10pA. Whether or not the current levels are low, the voltage levels are
almost always low. The nanovoltmeter’s sensitivity, which is superior to lock-in amplifiers, its
low 1/f noise, and its ability to compensate for offsets and drift make it a superior solution for
differential conductance measurements.
The Keithley method uses four-wire connections, which are desirable because
they eliminate voltage errors due to lead resistance or contact resistance. This is of
especial concern when the device has regions of low or moderate impedance, because lead
resistance may be significant compared to the actual resistance of the device. In a four-wire
configuration, there is no current flowing through the sense leads and thus no voltage drop.
Essentially, the voltmeter is placed directly across the DUT.
Another key benefit of sourcing current and measuring voltage is that in areas of
highest conductance, more data points are collected by sourcing the sweep in equal current
steps. Often, these areas are exactly the areas in which researches are most interested. Thus,
the data collected with the Keithley method is most detailed around these regions of interest.
As discussed before, the I-V technique can take tens of thousands of measurement
sweeps to attain sufficiently low noise. Because of its inherently low sourcing and measuring
noise, the Keithley technique requires only one pass, shortening hours of data collection to a
few minutes.
In the AC method the speed limitations are a direct result of cable capacitance. Device
impedance multiplied by cable capacitance (R × C) is the fundamental system time constant.
With three separate instruments connected directly to the device, this capacitance can be
extremely high. This means that stimulus cannot be applied to the device at a frequency faster
than this time constant. This forces the AC technique to be used at lower frequencies, which
results in higher noise due to 1/f effects. With lock-in amplifiers, the lower the frequency
A G r e a t e r M e a s u r e o f C o n f i d e n c e
used, the greater the measurement noise. The Keithley method, with two instruments
connected to the device instead of three as in the AC technique, has an inherently lower cable
capacitance. Further, the 6220/6221 provides an active guard, which eliminates the slowing
effects of cable capacitance. This greatly improved device settling time results in higher speed
and greater accuracy.
Special Cases
Some devices have non-monotonic I-V curves. This behavior is classified into two categories.
Both categories are sometimes referred to as NDC (negative differential conductance).
(1) Current Hop – A given voltage may correlate to more than one possible current
(Figure 7)
(2) Negative Differential Conductance – A given current may correlate to more than
one possible voltage (Figure 8)
Let’s look at how the source current/measure voltage differential conductance method
can be used with devices that exhibit these types of behaviors.
Current Hop
Some devices exhibit an I-V curve where the current is a multi-valued function of voltage
(Figure 7a). The negative differential conductance region cannot be characterized by applying
a voltage source, because any regulated voltage source is unstable into negative resistance
loads. Instead, a voltage source would produce a hysteresis curve that never traces out the
NDC region (dashed lines in Figure 7a). Interestingly, a current source is not stable over this
NDC region either, but adding a series resistor in the HI lead presents a composite device to
the current source, so that it does not see any NDC region. This resistance must be at least as
large as the largest negative resistance throughout the NDC region of the device.
Without any change to the 622x setup, the entire differential conductance curve can
be measured because the four-wire configuration connects the nanovoltmeter directly across
the device and not the series resistor, whose voltage drop is rejected along with all other lead
resistance. This lead resistance, which is normally considered a problem, actually makes full
characterization possible with these devices. With Keithley’s source-current architecture there
is no additional measurement required. Other methods require measurements of both device
current and voltage in these cases.
A G r e a t e r M e a s u r e o f C o n f i d e n c e
Current Negative
egative differential reg
region: Current Series resistor
both I and V sources eliminates
unstable, no direct NDC region.
measurements possible.
V
Voltage V
Voltage
Figure 7a. Devices with negative differential Figure 7b. The blue curve shows the device voltage
conductance regions require special as seen by the nanovoltmeter. The pink
treatment. curve shows the voltage across the device
plus the series resistor as seen by current
source – no NDC region to create
instability.
Negative
egative differential reg
region: Resistor in parallel
both I and V sources with DUT eliminates
unstable, no direct instability from NDC.
Current measurements possible. Current
V
Voltage V
Voltage
Figure 8. Even when the device shows multiple possible voltages for some currents, differential conductance
can easily be obtained with the new method.
A G r e a t e r M e a s u r e o f C o n f i d e n c e
Conclusion
Slow, noisy, and complex are how the old I-V sweep and AC techniques have always
been described. This is even more true now that Keithley has developed a new technique.
Keithley’s technique can:
• Replace the need for thousands of measurement sweeps to only one sweep,
shortening hours of data collection to a few minutes.
A G r e a t e r M e a s u r e o f C o n f i d e n c e
A G r e a t e r M e a s u r e o f C o n f i d e n c e
Specifications are subject to change without notice.
All Keithley trademarks and trade names are the property of Keithley Instruments, Inc.
All other trademarks and trade names are the property of their respective companies.
Keithley Instruments, Inc. 28775 Aurora Road • Cleveland, Ohio 44139 • 440-248-0400 • Fax: 440-248-6168
1-888-KEITHLEY (534-8453) • www.keithley.com
A G r e a t e r M e a s u r e o f C o n f i d e n c e