The document discusses sources of error in measuring minority carrier lifetime of silicon wafers during chemical passivation studies. It shows that placing wafers in liquid-filled containers for measurements can impact results due to changes in optical coupling and distance from the measurement coil. Thinner containers like plastic bags minimize these effects. The document also finds degradation of lifetime occurs over time for wafers immersed in passivation solutions, likely due to oxidation from air in the sealed containers. Accurately measuring lifetimes while wafers are in solution requires accounting for these container and degradation effects.
Minority Carrier Lifetime Measurement Errors in Chemical Passivation Studies
1. The Minority Carrier Lifetime Measurements: An Electrical Passivation
Study
Meixi Chen, James H. Hack, Abhishek Iyer, Robert L. Opila
Univeristy of Delaware, Newark, DE, 19716, US
Abstract—The measurement of minority carrier lifetime of
silicon in chemical passivation studies is often taken with the
wafer in solution. We show that variations in the optical constant
and inductive coupling to the wafer, as well as the presence of the
liquid solution, will lead to discrepancies in the measured lifetime
of the wafers. Continued deterioration of lifetime is observed
when wafers are in the passivation solution and is presumably
due to oxygen in the bag. N-type silicon (100) wafers appear to
have a limit of ten trials repeated use, after which carrier lifetime
decreases, likely caused by surface degradation.
Index Terms—charge carrier lifetime, degradation, organic
passivants, silicon.
I. INTRODUCTION
Room-temperature passivation of silicon wafers has been in-
vestigated for both in-process measurements and as a passivant
for future generation solar cells. Passivants like quinhydrone
(QHY), benzoquinone (BQ) and iodine in alcohol have been
shown to lead to an increase in the minority carrier lifetime
[1] - [5].
In monitoring the effects of organic surface passivants such
as benzoquinone on carrier lifetime, it is desirable to be
able to perform lifetime measurements while the specimen is
still immersed in solution. This allows for study of the time
evolution of the reaction and avoids atmospheric exposure
which may lead to further surface modification [6]. Carrier
lifetime measurements are performed using a WCT-120 Sinton
lifetime tester. Since the wafers are not being placed directly
on the measurement surface of the lifetime tester, but are
instead sealed within a liquid-containing bag, it is expected
that this modified experimental setup will have some effect on
the resulting lifetime measurements. The liquid filled bag may
act to modify the amount of light reaching the wafer and also
affect the coupling between the wafer and the inductive coil.
In this paper, the severity of these effects will be examined
and potential methods of accounting for them discussed.
Besides the error caused by testing setups, the change
of surface morphology can also cause uncertainties of the
lifetime. H-terminated silicon surfaces are prepared before
passivation. Normally, two silicon preparation procedures are
well accepted: wet-chemistry methods and vacuum passivation
with atomic hydrogen. Wet-chemistry involves a classic RCA
clean and HF-dip, which is easier to perform than atomic
hydrogen, but has less control of the surface morphology [7].
The wet-chemistry cleaning is a multistep process involves
sequential silicon oxidation and silicon removal reactions [8].
STM and polarized infrared absorption spectrum, as well as
Fig. 1. Structure of the passivants used in this paper. Chloranil, Tetrachloro-
1,4-benzoquinone (99% Sigma-Aldrich, used as is); BQ, p-Benzoquinone
(98+% Sigma-Aldrich, used as is); HQ, Hydroquinone (99% Acros Organics,
used as is)
Monte Carlo simulation have been used to study the etch
mechanism [9] - [10]. Presumably, the surface will be slightly
different every time after chemical cleaning procedure, which
might cause a uncertainty in passivation studies if the same
wafer is used multiple times.
In this paper, studies of the accuracy of lifetime measure-
ments in liquid filled container are carried out, as well as two
types of degradation of wafers lifetime: decrease of lifetime
after various etching and passivation trials, and continuous
deterioration during the time when the wafer is still immersed
in solution.
II. EXPERIMENTAL DETAILS
The wafers used in this work are double side polished n-
type silicon(100) wafer with a resistivity of 15-20 Ω-cm and
500 µm thickness.
The following cleaning procedure is followed for every trial:
The first step is a piranha clean consisting of a 5 minute bath
in 200 ml of a 4:1 solution of sulphuric acid and hydrogen
peroxide. Piranha solutions are freshly made and cooled at
room temperature for 10 minutes before use. Subsequently the
wafers are given a 5 minute submersion in 250 ml of Milli-Q
water. Following the Milli-Q water bath the wafers are placed
in a 2% HF solution for 2 minutes. The wafers are then briefly
rinsed with Milli-Q water and blown dry with N2.
The cleaned wafer is quickly placed in a sealed clear plastic
bag (acid-proof, polyethylene, thickness 4mil) containing a
0.01 mol/L passivation solution, including, for example benzo-
quinone in methanol, chloranil in methanol, or neat methanol.
Structures of passivants are shown in Fig. 1.
Once sealed, the bags are placed on the stage of a Sinton
Instruments WCT-120 wafer lifetime tool. The Sinton lifetime
tester operates by exciting carriers with a xenon flash lamp
2. and then performing eddy current measurements with a coil
inductively coupled to the sample wafer. The voltage from
the coil is measured in a bridge circuit, giving an indication
of the photoconductance, while the voltage from a reference
photodiode gives an indication of the incident irradiation.
Excess minority carrier lifetimes can then be calculated as
a function of minority carrier density [11].
Repeated measurements allow us to plot minority carrier
lifetime as a function of time as the reaction progresses. After
lifetime testing, the wafer is given acetone, methanol, iso-
propanol and DI water wash.
III. RESULTS AND DISCUSSION
A. Accuracy of Minority Carrier Lifetime Measurements in
Liquid Filled Containers
The Sinton lifetime tester operating in the quasi-steady state
mode calculates [11] the effective minority carrier lifetime of
a specimen based on (1).
τeff = σL/(Jph(µn + µp)) (1)
The value of the photoconductance, σL, is measured from
the bridge circuit voltage while the electron and hole mobilities
(µn + µp) are determined from tabulated values at known
average minority carrier densities ∆nav to satisfy (2). Here W
is the thickness of the sample and q is the elementary charge.
σL = q∆nav(µn + µp)W (2)
The value for the photogeneration, Jph, is related to the
incident light intensity, which is measured relative to a ref-
erence photodiode. From this measured irradiance the value
of Jph can be estimated by making assumptions about the
absorption properties of the sample. For a bulk silicon sample
of known thickness, the absorptivity is well documented and
can be used to estimate Jph. However modifications to the
reflectivity of the surface and other optical properties of the
laboratory setup may affect the amount of the incident light
which is transmitted to the bulk of the wafer. To account for
these factors, an optical constant representing the percent of
incident light absorbed is introduced. The Sinton lifetime tester
assumes for silicon samples of sufficient thickness and low
surface reflectivity an experimentally determined value of Jph
of 38mA/cm2
. The optical factor directly scales this value, for
instance, a constant of 0.7 represents reduced photogeneration
for a bare silicon wafer with high reflectivity, while higher
photogeneration with optical constants exceeding 1 are possi-
ble for thick samples with exceptional anti-reflection coatings.
The proper choice of optical constant will linearly affect the
inverse of the resulting measured minority carrier lifetime.
A second potential source of error arises from the determi-
nation of the value of the photoconductance σL. This value
is calculated from the voltage across an impedance bridge
inductively coupled to the sample wafer. The relationship
between the measured bridge voltage and the conductance of
a wafer is determined in the calibration of the instrument.
However this relationship may be modified due to changes
in the distance between the wafer and inductive coil, and the
presence of a liquid solution.
The instrument calibration fits a quadratic relation between
the bridge voltage, VB, and wafer conductance, however this
relation can generally be simplified to a linear relationship in
the measurement range of interest with reasonable accuracy
[12].
σL = KBVB (3)
When using the quasi-steady state photoconductance
(QSSPC) technique as described above, the measured lifetime
is dependent on the optical constant and the bridge voltage
scaling factor KB. Mismatch in the value of the optical
constant will lead to a linear change in the inverse of the
measured lifetime, while mismatch in the bridge voltage
scaling factor will have a nonlinear effect, due to a dependency
on average minority carrier concentration. When instead using
the transient photoconductance decay (PCD) analysis method,
the dependence on the optical constant is removed, and our
error depends only on the mismatch in KB . A third technique,
the generalized analysis method, is a combination of these
two methods dependent on the value of the optical constant at
low lifetimes while becoming largely independent of optical
constant values at high lifetimes. In the generalized analysis
mode a change in the optical constant value from 0.78 to 0.58
is seen to yield about a 30% increase in measured lifetime for
lifetimes around 20µs, a 20% increase for lifetimes around
200µs, and almost no increase at lifetimes of around 3000 µs.
To examine the magnitude of these effects on the measured
lifetime, repeat measurements were made on the same wafer in
various candidate containers (Fig. 2). Lifetimes were measured
in the generalized analysis mode for a minority carrier density
of 1 ∗ 1014
cm−3
. A clear decrease in measured lifetime was
observed with increased thickness of the container. The 4 mil
plastic bag, as the thinnest of the containers was seen to have
the smallest effect on the measurement. For the wafer in a
plastic bag, the decrease in lifetime is expected to be the
result of both the increase in distance from the inductive coil
and the decrease of light intensity reaching the sample. These
measurements were repeated on two additional wafers, which
showed similar nonlinear decreases in measured lifetime with
container thickness.
A further change in measured lifetime is expected for the
samples when placed in the liquid solution. An initial test
shown compares the difference in lifetimes recorded for wafers
when placed directly on the measurement surface and when
placed in a plastic bag containing 50 ml of methanol (ME)
solution. Changes of lifetime of -15%, -4%, -25% and -42%
were observed for lifetimes at a minority carrier density of
1∗1015
cm−3
using an optical constant of 0.7. A simple trend
is not clear from this data, and in one case the measurement
actually increases when placed in solution. Methanol, while
less effective than benzoquinone, is known to show some mild
passivation effect on silicon surfaces, so it is initially unclear
3. Fig. 2. Lifetimes measured for an n-type wafer in different sample-holding
conditions. Measured lifetime is seen to decrease with increased distance
from the instrument surface. Data is acquired for samples placed directly
on the surface (0mm), in an empty bag (.1mm), on a glass slide (1.0mm), a
plastic petri dish (1.3mm), a plastic wafer box (2.0mm) and a glass petri dish
(3.0mm).
whether this chemistry is having an additional effect on the
measurement [14]- [15].
To attempt to control for any chemical effects on the life-
time, a wafer (Sample N4) with a thermally grown oxide layer
of thickness 110nm, was tested. This thick oxide layer should
effectively block any passivating effects caused by organic
molecules from the solution. Lifetime measurements were
performed for this wafer in and out of solution when contained
in either a plastic petri dish or plastic bag. The data shown
in Fig. 3 show the combined effects of factors independent of
optical constant, such as placing the wafer in a petri dish, and
factors which affect only the optical constant, such as covering
the petri dish with a lid. It is expected that the methanol filled
bag represents a combination of both types of factors. Again
the bag was shown to cause a smaller decrease in the measured
lifetime than the petri dish, with further decreases in measured
lifetime for both containers when immersed in methanol. A
repeated measurement with the wafer placed directly on the
instrument surface was performed shortly after removal from
the methanol solution, which showed no change to the lifetime
of the wafer. The same tests were performed on a wafer with
only a thin natively formed oxide, which showed the same
trends with no increase in lifetime from the brief methanol
treatment.
The same tests were repeated using 50ml of a 0.01 M
solution of benzoquinone (BQ) in methanol as shown in Fig.
4. The wafer with a thermally grown oxide layer (Sample
N4) recorded very similar lifetimes as in the pure methanol
solution. By comparison, the wafer with native oxide layer
showed marked lifetime increases persisting after removal
from the BQ methanol solution. This data shows the usefulness
of the thermally grown oxide in separating the effects on
Fig. 3. Measurements of a wafer with thermally grown oxide (N4) and native
oxide (N3) (Note the logarithmic axis). Decreases in measured lifetime are
seen due to thickness of containers, light blocking lids, and immersion in
methanol. The last measurement is a repeat of the first measurement following
immersion in methanol and shows no significant change.
Fig. 4. Measurements of a wafer with thermally grown oxide (N4) and native
oxide (N5). The wafer with a thermally grown oxide layer shows similar
behavior in the methanol + BQ solution as observed in the pure methanol
solution. The wafer with native oxide shows an immediate increase in lifetime
when placed in methanol + BQ solution. This increased lifetime is shown to
persist immediately after removed from solution.
measured lifetime due to the physical experimental setup from
those due to chemical passivation.
The presence of a liquid solution is likely to affect both the
value of the optical constant and the bridge voltage scaling
factor KB. While these effects generally tend to decrease
the reported carrier lifetime, recalibration is recommended to
4. ensure accuracy of measurements. In most cases, the optical
constant can be estimated reasonably well for a particular type
of sample and environmental setup and can be kept fixed for
the duration of the experiment. At sufficiently high lifetime
values in the millisecond range, the transient photoconduc-
tance mode or generalized mode can be used for data analysis,
and the choice of optical constant becomes unimportant.
The more complicated source of error in measured life-
times comes from changes in the relationship between wafer
conductance and measured bridge voltage. An increase in
distance between the wafer and inductive coil will tend to
decrease the measured lifetime due to an underestimation
of the photoconductance and corresponding minority carrier
concentrations [13]. This effect is seen to be relatively strong
even for small displacements from the instrument surface.
Placing the wafer in a plastic petri dish was seen to decrease
the measured lifetime up to 42%. For samples that must
be placed in containers, thinner materials such as a plastic
bag minimize this introduced error. A better solution is to
recalibrate the instruments for the new environmental setup,
finding a new scaling factor KB for the bridge voltage, as
described by Lago-Aurrekoetxea et al. [12]
A wafer with a thick thermally grown oxide layer has been
shown to be useful in separating the physical and chemical
passivating effects of the liquid. It is expected that the use of
similarly chemically inert wafers will enable the removal of
the effect of the experimental setup on the measured lifetimes.
B. Lifetime Degradation in passivation solutions
Degradation of QHY/ME passivation, when the wafers are
taken out of passivation solution and exposed to air, has been
reported [6]. Surface oxidation is suggested to be responsible
for this degradation. In our experiment, continued deterioration
is also observed during the passivation experiments, while
wafers are still immersed in solution. Both the lifetime data
and the time-dependent behavior are reproducible.
The wafers were cleaned and immersed into BQ/ME solu-
tion in a zipper bag. The air was removed from the bag, but
a small amount of air is still present from air leaking into the
bag over time. Lifetime data was taken continuously for 15
hours while the wafers were kept in solution, shown in Fig.
5.
For the following reaction,
Si − QH + O2 → SiOx + HQ
R = −
d[SiQH]
dt
= k[SiQH]
m
[O2]
n
(4)
R is the degradation rate, t is the time of passivation treatment
and k is the rate constant. Lifetime is directly affected by the
amount of the oxide product, which is a function of time.
Polynomial fitting of the curve gives:
[Lifetime] = 233.7 − 0.033t + 2.078 ∗ 10−5
t2
(5)
Fig. 5. Degradation of BQ/ME passivation when substrates are still immersed
in solution.
Fig. 6. Passivation results of BQ/ME, HQ/ME and the control sample,
methanol.
the coefficient of determination r2
=0.95.
R =
d[Lifetime]
dt
= −0.033 + 4.15 ∗ 10−5
t (6)
The lifetime is at a maximum at time t = 0 and starts to
degrade immediately. It is also shown here that the degradation
rate is not constant. The change in degradation rate likely
corresponds with the change in the amount of oxygen in the
bag, since air is entering the system over time.
Although steady degradation is shown in BQ/ME passiva-
tion, no significant lifetime decrease is observed in HQ/ME
or ME control sample, illustrated in Fig. 6. However, the
possibility of a lifetime decrease in HQ/ME or ME control
cannot be ruled out because it may be hidden by the rise in
the lifetime.
5. Fig. 7. Lifetime of wafers after various trials (represented by three passivants).
C. Lifetime Degradation after a Number of Trials
For this section we carried out passivation repeatedly with
benzoquinone, methanol and chloronil on a n-type wafer.
The n-type wafer underwent passivation with benzoquinone,
methanol and chloronil several times with pirhana and HF
cleans between each run. Fig. 7 indicates the lifetime for this
n-type wafer after various trials.
As mentioned in the previous section, the lifetime of the BQ
passivated wafer decreased with time, so the average lifetime
data in the first 200 minutes is used in Fig. 7; for every trial
of BQ, the lifetime was measured for 200 minutes with one
reading every 15 minutes, and the average lifetime number
was recorded and presented. The same analysis was carried out
with chloranil, since it showed an identical curve to BQ. The
data of chloranil passivation on silicon surfaces as a function
of time will be published. For the case of benzoquinone,
which was tested for the 1st
, 5th
, 8th
, 11th
and 17th
trials,
the lifetime gradually decreased after yielding constant values
until the 11th
trial.
The same cleaning and passivation experiments were carried
out for the wafer in pure methanol and the lifetime increased
with time and reached a plateau around 700 minutes, as shown
in Fig. 7. This saturation level is adopted here to represent ME
passivation. It must be noted that passivation with methanol
occurred at the 10th
, 13th
, and 15th
trials, which means that
the wafer was already used several times before testing it with
methanol. Hence the wafer was already old before it was tried
for methanol. Fig. 7 shows that the lifetime for the n-type
wafer decreases with an increasing number of trials for ME
and chloranil.
It is likely that the wafer lifetime decreased with every
successive trial because the surface is sensitive to the etching
process resulting in damage to the surface with every trial.
After each trial, it is likely that the surface became rougher
resulting in less effective passivation and decreased lifetime
for the sample irrespective of the passivant.
Hence, it can be concluded here that the n-type silicon wafer
can give a reliable lifetime data up to about ten times of
repeated use, and in our experience then begins to degrade with
increasing trials. There is not one specific factor responsible
for the fall in lifetime. It can be roughness of the sample or
the accumulation of contaminants or a combination of both.
IV. CONCLUSIONS
Variations in the optical constant and inductive coupling of
the wafer, as well as the presence of a liquid solution, are
shown to lead to discrepancies in the measured lifetime of the
wafers. The accuracy of optical constant is more important
in the quasi-steady state photoconductance testing mode than
in the transient photoconductance decay or generalized mode.
Since optical constants typically range from around 0.63-
1.03 for standard test setups, even gross misjudgment of the
optical constant will still result in measurements accurate
within a factor of 2. The changes in the relationship between
wafer conductance and measured bridge voltage are more
complicated and can cause up to a 40% decrease of lifetime
(for example, if a petri dish is used). For samples that must
be placed in containers, thinner materials such as a plastic bag
minimize the introduced error.
Continued nonlinear deterioration of lifetime in BQ/ME
is observed during the passivation when wafers are still
immersing in solution. Although steady degradation is shown
in BQ/ME passivation, no significant lifetime decrease is
observed in HQ/ME or ME control sample. However, the
possibility of a lifetime decrease in HQ/ME or ME control
cannot be ruled out because it may be hidden by the rise in
the lifetime.
In addition, passivation of the n-type wafer begins to de-
grade after ten repeated trials in these experiments. There is not
likely to be a single specific factor responsible for the fall in
lifetime; surface roughness and/or accumulated contamination
can contribute to the observed decrease in carrier lifetime.
6. REFERENCES
[1] R. L. Opila, D. Yang, N. Kotulak, L. Costello, and B. Chhabra, “Mech-
anism of electrical passivation of Si surfaces with quinhydrone”, 38th
IEEE Photovoltaic Specialist Conference, pp. 2583, 2012.
[2] T. S. Horanyi, T. Pavelka, and P. Tutto, “In situ bulk lifetime measurement
on Si”, Applied Surface Science, 63, 306, 1993.
[3] H. Takato, I. Sakata, and R. Shimokawa, “Surface Passivation Effect
of Silicon Substrates due to Quinhydrone/Ethanol Treatment”, Japanese
Journal of Applied Physics, Part 2, 40, L1003, 2001.
[4] H. Takato, I. Sakata, and R. Shimokawa,“Quinhydrone/Methanol Treat-
ment for the Measurement of Carrier Lifetime in Silicon Substrates,
Japanese Journal of Applied Physics, Part 2, 41, L870, 2002.
[5] H. Msaad, J. Michel, J. Lappe, and L. C. Kimerling,“Electronic passiva-
tion of silicon surfaces by halogens, Journal of Electronic Materials, 23,
487, 1994.
[6] B. Chhabra, S. Bowden, R. L. Opila, and C. B. Honsberg, “High effective
minority carrier lifetime on silicon substrates using quinhydrone-methanol
passivation,” Applied Physics Letters, vol. 96, no. 6, p. 063502, 2010.
[7] F. Tian, D. Yang, R. L. Opila, A. V. Teplyakov, “Chemical and electrical
passivation of Si(111) surfaces, Applied Surface Science, 258, 3019, 2012.
[8] B. S. Aldinger and M. A. Hines, “Si(100) Etching in Aqueous Fluoride
Solutions: Parallel Etching Reactions Lead to pH-Dependent Nanohillock
Formation or Atomically Flat Surfaces,”, The Journal of Physical Chem-
istry C, 116, 21499, 2012.
[9] M. A. Hines, M. F. Faggin, A. Gupta, B. S. Aldinger, and K. Bao, “Self-
Propagating Reaction Produces Near-Ideal Functionalization of Si(100)
and Flat Surfaces”, The Journal of Physical Chemistry C, 116, 18920,
2012
[10] I. T. Clark, B. S. Aldinger, A. Gupta, and M. A. Hines, “Aqueous
Etching Produces Si(100) Surfaces of Near-Atomic Flatness: Strain
Minimization Does Not Predict Surface Morphology,”, The Journal of
Physical Chemistry C, 114, 423, 2010
[11] R. A. Sinton and A. Cuevas, “Contactless determination of currentvolt-
age characteristics and minority-carrier lifetimes in semiconductors from
quasi-steady-state photoconductance data”, Applied Physics Letters, vol.
69, p. 2510, 1996.
[12] R. Lago Aurrekoetxea, I. Tobias, C. del Cannizo, and A. Luque, “Life-
time Measurements by Photoconductance Techniques in Wafers Immersed
in a Passivating Liquid”, Journal of The Electrochemical Society, vol. 148,
no. 4, p. G200, 2001.
[13] J. Krausse, “Die abhangigkeit der tragerbeweglichkeit in silizium von der
konzentration der freien ladungstragerII”, Solid-State Electronics, vol. 15,
no. 12, pp. 1377, 1972.
[14] M. Chen et. al manuscript in preparation
[15] D. J. Michalak, S. Rivillon, Y. J. Chabal, A. Esteve, and N. S. Lewis,
“Infrared Spectroscopic Investigation of the Reaction of Hydrogen-
Terminated, (111)-Oriented, Silicon Surfaces with Liquid Methanol”, The
Journal of Physical Chemistry B, 110, 20426, 2006.