nd
tro
r Hu
lesex
has b
ique
nce of
of the
s at lo
e ac im
es. Th
facial
minat
s sign
test m
563� A
ceived
tween electrodes. The SIR technique cannot distinguish these In Fig. 1, the relationship between the real and imaginary parts of
al So
3.00 ©
individual processes. In contrast, the ac impedance technique, using
small ac voltages, has been shown to access different electrochemi-
cal reactions.4,5 By scanning across a wide frequency range, differ-
ent parameters associated with basic conduction processes can be
quantified separately due to their different frequency dependency.
In this paper, ac impedance is measured on different flux types
and compared with SIR results. Using standard circuit analysis,
equivalent inductance-capacitance-resistance circuits are constructed
to investigate the effect of contaminants and their concentration.
Using this approach, specific features from the impedance spectrum
can be attributed to electrochemical parameters, such as the double-
layer capacitance and the charge-transfer resistance. The dominant
conduction processes can be estimated by ac impedance spectrum
analysis, and it is shown that dendrite formation can be predicted.
Theory
AC impedance measurement.— AC impedance spectra made
over a wide frequency range provide mechanistic information, be-
cause the various conduction processes have different frequency de-
pendencies. Circuit impedance, Zˆ , has a real and imaginary part and
is given by Zˆ = Z� + jZ�, where Z� is the real impedance, Z� is the
imaginary impedance, and j is the square root of −1. The three
common conduction processes characterized by the ac impedance
impedance for a parallel resistance–capacitance �RC� circuit is
shown. The absolute impedance is given by �Z� = �Z�2 + Z�2, with
the phase angle � = tan−1�Z�/Z��. The locus of Zˆ reflects the fre-
quency dependency. At very low frequencies Zˆ has a small imagi-
nary component; as the frequency increases, Zˆ includes an increas-
ing imaginary component due to the capacitance, and at high
frequency the capacitance acts as a short and Zˆ falls to zero. The
phase angle gives the delay of impedance relative to the real part
and hence the balance between capacitance and resistance.
Z
Z
-Z
Low
frequency
High
frequency
|Z|
φ
|Z| - Impedance φ- Phase angle
Characterization of the Co
Electrolyte Layers on Elec
Impedance
Ling Chunxian Zouz and Christophe
National Physical Laboratory, Teddington, Midd
The reliability impact of flux residues on electronic assemblies
�SIR� measurement with dc voltages. An ac impedance techn
conduction mechanisms and electrode reactions in the prese
fluxes using a standard comb pattern the relative applicability
assessing the reliability of electronic boards. Impedance value
produce similar predictions of reliability. More importantly, th
although the technique itself does not actually promote dendrit
resistance of the thin water layer and impedance from inter
solution resistance, between the copper-comb electrodes, do
impedance from interfacial electrochemical processes become
predictive capability could be developed into a nondestructive
ization and indication of future reliability.
© 2008 The Electrochemical Society. �DOI: 10.1149/1.3005
Manuscript submitted August 20, 2008; revised manuscript re
The surface insulation resistance �SIR� technique measures the
resistance of an adsorbed moisture film between two metal elec-
trodes on a substrate surface using dc voltage under elevated tem-
perature and humidity conditions within a chamber. Typical standard
conditions are 40°C/93% relative humidity �RH� or 85°C/85% RH,
which results in a moisture layer of approximately 100 nm. This
technique is used to qualify flux residues, e.g., ISO 9455 part 12, or
to assess the effect of contaminants on the assembly reliability, IEC
61189 part 10.2 and Ref. 1-3. The SIR measurement technique itself
is uncomplicated, but the science behind the measurement is com-
plex. Although the SIR technique entails a straightforward ohmic
measurement, albeit in the nanoamp range, the intertrack region
does not behave as a simple ohmic element, as there are a number of
conduction processes involved. Electrochemical processes at the
metal-track/electrolyte interfaces, in the presence of the flux resi-
dues, must take place to translate electronic conduction in the elec-
trodes to ionic conduction in the thin, aqueous electrolyte film be-
Journal of The Electrochemic
0013-4651/2008/156�1�/C8/8/$2
C8
z E-mail: lz@npl.co.uk
Downloaded 16 Aug 2010 to 121.8.210.35. Redistribution subject to EC
uction Mechanisms in Adsorbed
nic Boards Using AC
nt
TW11 0LW, United Kingdom
een traditionally evaluated using surface insulation resistance
has been investigated to provide detailed information on the
flux residues on electronic boards. By evaluating different
ac impedance and SIR techniques has been made in terms of
w frequencies, �1 Hz, are close to the SIR results and hence
pedance spectrum can be used to predict dendrite formation,
e ac impedance method can distinguish between ionic solution
electrochemical processes. At low contamination levels the
es the overall impedance. At high contamination levels the
ificant and increases the potential of dendrite formation. This
ethod to provide a more detailed electrochemical character-
ll rights reserved.
September 25, 2008. Published October 31, 2008.
measurement are ohmic conduction, represented by resistance R,
dielectric displacement, represented by capacitance C, and diffusion
of electroactive species, represented by the Warburg impedance.6,7
Ohmic conduction can be either electronic or ionic and is defined as
the real impedance, Z� = R, with the current and voltage in phase.
Displacement currents are defined by an imaginary impedance, Z�
= j /2�fC, where f is the frequency of the applied voltage. The
current leads the voltage by a phase angle of 90°. The Warburg
impedance is complex but can be described by a frequency-
dependent term Zˆ = ��2�f�−1/2�1 − j�, where � is the Warburg co-
efficient, a constant that depends on the mobility and concentration
of diffusing species.7 Interpreting impedance measurements using an
equivalent circuit can be complex; however, as all circuit compo-
nents have varying frequency dependencies, the main components
can be extracted from a simplified equivalent circuit, and other less
significant components can be eliminated.
ciety, 156 �1� C8-C15 �2009�
The Electrochemical Society
Z Real Z - Imaginary
Figure 1. AC impedance for a parallel RC circuit.
S license or copyright; see http://www.ecsdl.org/terms_use.jsp
al So
Figure 2 shows the ac impedance spectra of four simple circuits.
The second row gives the relationship between Zˆ and frequency,
which are known as Bode plots. In all the plots where the behavior
is purely resistive, an R is marked on the figure. The third row
shows the phase angle as a function of frequency. The last row plots
−Z� against Z�, which is known as a Nyquist plot. The arrow shows
how Zˆ changes from low to high frequency.
For a serial RC circuit, column three in Fig. 2, the impedance is
expressed in Eq. 1 and evaluated using Eq. 2
Zˆ = R − j 1
2�fC �1�
�Z� = �R2 + 1
�2�fC�2 �2�
For a parallel RC circuit, the impedance is expressed in Eq. 3 and
evaluated using Eq. 4
Zˆ =
R
1 + �2�f�2C2R2 − j
2�fCR2
1 + �2�f�2C2R2 �3�
�Z� =
R
�1 + �2�f�2C2R2 �4�
In the Nyquist plot, the frequency dependency of Zˆ describes a
semicircle for a parallel RC circuit; at a certain frequency, fz�max, at
the top of the semicircle the imaginary impedance reaches a maxi-
mum value, Zmax� , and is equal to the real impedance. This frequencyfz�max can be calculated using Eq. 5 or can be used to evaluate the
capacitor and resistor values in the circuit
fz�max =
1
2�RC
�5�
These equations are used later in analyzing the response from im-
pedance measurements of fluxes on comb patterns.
Conduction model for parallel electrode system.— An equiva-
lent circuit representing each significant conduction process for a
comb pattern at an elevated temperature and humidity is presented
in Fig. 3, along with a simplified version for SIR. The electrochemi-
cal process initiates with the adsorption of a thin water layer from
R C
R
C R
|Z|
f
1/2πfC
φ
-Z
-90°
f
Z
f
|Z|
R
φ
0° f
-Z
R
Z
|Z|
f
φ
R
-90°
-Z
R
f
Z
φ
R
f
-90°
-Z
R
f
Z
|Z|
Figure 2. AC impedance spectra and their equivalent circuits.
Journal of The Electrochemic
the high-humidity atmosphere, and the conduction process is sup-
ported by the presence of flux residues dissociating to form ionic
charge carriers. The impedance to ionic conduction is dominated by
the space between the metal tracks of the comb and is represented
Downloaded 16 Aug 2010 to 121.8.210.35. Redistribution subject to EC
by the terms RC and CC, where RC is the ionic resistance of the thin
water layer with contaminants across the interelectrode gap of the
comb and CC is the capacitance of the comb pattern. The circuit
Cu Cu
Water with
contaminants
SIR
RcRct Zw
Electrode reaction
AC impedance
Rc
Rct
Cdl
Zw
Cc
Electrode reaction
Figure 3. Equivalent circuit for ac impedance and SIR measurement on a
comb pattern.
C9ciety, 156 �1� C8-C15 �2009� C9
elements in the dotted box are a representation of electrochemical
processes at the interfaces of the contaminant layer and electrode;
only one equivalent circuit is presented in the figure, as the two
electrode interfaces are identical. These reactions represent the tran-
S license or copyright; see http://www.ecsdl.org/terms_use.jsp
al So
sition from electronic conduction in the electrode to ionic conduc-
tion in the contaminant layer. The equivalent circuit for these elec-
trochemical processes includes the charge-transfer resistance RCT,
associated with a faradaic reaction at the electrode surface; CDL is a
double layer or blocking capacitance at the interface, which may
allow ionic current to flow in the absence of a faradaic reaction;8
and ZW is the Warburg impedance, detectable when a faradaic reac-
tion occurs under diffusion control near the interface. When these
processes are dominant, the equivalent-circuit model in Fig. 3 can be
used to evaluate the impedance parameters by measuring and ana-
lyzing the impedance spectrum. In contrast, the SIR technique can
only measure all three parameters in combination, and it cannot
distinguish between each process.
Experimental
Test flux.— Four fluxes were selected as representative of a wide
range of flux chemistries, all using carboxylic acid activators, and
these are listed in Table I. SB1, SB2, and WB1 are commercial
fluxes, and WB2 is the flux developed for solderability tests by
National Physical Laboratory.9 Fluxes were diluted using isopropyl
alcohol �IPA� or DI water, depending on the flux base, to achieve
SIR values that were above 106 � and did not cause dendrite for-
mation. These flux concentrations are also listed in Table I. One
further test was conducted using a more concentrated version of flux
SB1 �80%� to examine the capability of ac impedance measurement
to predict dendrite formation.
Sample preparation.— An NPL-designed board consisting of
four identical SIR comb patterns with Cu finish was used as shown
in Fig. 4. The size of each comb pattern was 25 � 25 mm, with
Table I. Test fluxes details.
Flux Base Solid
Acid
no.
Contains
resin
Concentration
�%�
SB1 IPA 1.6 15.5 No 50
SB2 IPA 6.0 41.0 Yes 30
WB1 Water 4.6 37.1 No 3
WB2 Water 4.8 41.1 Yes 3
C10 Journal of The ElectrochemicC10
Figure 4. Test board for SIR and impedance measurement.
Downloaded 16 Aug 2010 to 121.8.210.35. Redistribution subject to EC
0.60 mm pitch and 0.20 mm gap. The boards were cleaned in an
Ionograph 500M with 75% IPA/25% deionized water �DI� solution
at 45°C. The cleaned boards were fluxed using 50 �L fluxes to
cover each comb pattern. Fluxed boards were dried in a 100°C dry
oven for 5 min.
SIR and impedance measurement.— Five experiments �A, B, C,
D, and E� were conducted for each of the four fluxes, diluted as
follows; 50% SB1, 30% SB2, 3% WB1, and 3% WB2. In these
experiments, A–C were impedance measurements. Sample imped-
ance spectra were measured periodically during the 24 h environ-
mental exposure in response to a small 50 mV peak–peak ac signal,
with frequencies ranging from 10−1 to 105 Hz. Impedance spectra
were determined using an ACM Instruments Gill AC. Between mea-
surements the samples were conditioned with either 0 V, 5 V at
50 Hz, or 5 V dc for experiments A, B, and C, respectively. The
conditioning was applied after the first measurement. Experiment D
followed a continuous SIR measurement schedule, with measure-
ments every 10 min using a Gen3Systems Auto-SIR. A bias voltage
of 5 V dc was applied during the 24 h exposure. This instrument has
a 106 � limiting resistor on each test channel to protect dendrites
from burning out; hence, the minimum SIR measurement value was
106 �. For fluxes 3% WB1 and 3% WB2, experiment E, a periodic
SIR measurement was performed. Here, there were four measure-
ments during the 24 h exposure, and there was no conditioning bias
between measurements. Concentrated SB1 �80%� was also tested to
simulate conditions where dendrite formation is likely. The environ-
mental condition for all experiments was 40°C/93% RH.
Results and Discussion
AC impedance spectra and modelling without dendrite form-
ation.— The results from the three ac impedance experiments, A, B,
and C, with diluted fluxes could all be interpreted using the same
model equivalent circuit. In Fig. 5, results are presented as a Bode
plot for the SB2 flux at 30% dilution and for increasing exposure
times in the chamber. The other fluxes produced similar results.
With these flux dilutions, dendrites would not occur with SIR test-
ing.
In Fig. 6, again for the SB2 flux, and after 3 h exposure, the data
from the same experiment are presented, giving the Bode �a�, Ny-
quist �b�, and phase angle �c� plots. Referring to Fig. 2 the results in
Fig. 6 are clearly consistent with a simple parallel RC and CC net-
work. At these flux concentrations, the impedance response has no
component that can be attributed to the charge-transfer or diffusion
components, RCT, CDL, or ZW, the impedance contribution expected
from the electrode electrochemical processes. Figure 6 also reveals
1.0E+04
1.0E+05
1.0E+06
1.0E+07
1.0E+08
1.0E+09
1.0E+10
1.0E-01 1.0E+01 1.0E+03 1.0E+05
Frequency (Hz)
Im
pe
da
nc
e
(o
hm
).
@1hour
@2hour
@3hour
@3hour
@4hour
@6hour
@7hour
@23hour
@24hour
Figure 5. Body plots for 30% SB2 flux with experiment A.
ciety, 156 �1� C8-C15 �2009�
that at high frequencies the capacitive behavior of the metal comb
dominates, and impedance is independent of the nature of the sur-
face contaminant layer, as can be seen by the phase angle tending to
−90°. At low frequencies, impedance is more sensitive to the con-
S license or copyright; see http://www.ecsdl.org/terms_use.jsp
al So
tamination layer, with the effect that RC is more pronounced and the
phase angle tends toward 0°. So, the equivalent circuit presented in
Fig. 3 can be simplified, as shown in Fig. 7, because the electrode
reactions do not contribute significantly to the overall impedance.
For 30% SB2 flux at 3 h, as shown in Fig. 6, RC can be evaluated
using the Nyquist plot. Because at frequency fz�max, Zmax� = Zmax� , so
RC = 2Zmax� = 2.0 � 107 �, then CC can be calculated using Eq. 5,
CC = 1/2�fmaxRC = 7.5 � 10−11 F. This capacitance value is near
the sensitivity limit for the instrument and includes stray capacitance
from edge connector and leads in the experimental arrangement. The
1.E+04
1.E+05
1.E+06
1.E+07
1.E+08
1.E+09
1.0E-02 1.0E+00 1.0E+02 1.0E+04
Frequency (Hz)
Im
pe
da
nc
e
(o
hm
).
Measured
Modelled@ 3 hour
a
0.0E+00
4.0E+06
8.0E+06
1.2E+07
1.6E+07
0.0E+0
0
6.0E+0
6
1.2E+0
7
1.8E+0
7
2.4E+0
7
Z' (ohm)
-Z
''(
oh
m
)
Measured
Modelled
@ 3 hour
b
0
30
60
90
1.0E-02 1.0E+00 1.0E+02 1.0E+04
Frequency (Hz)
P
ha
se
an
gl
e
(°
).
Measured
Modelled
_
_
_
@ 3 hour
c
Figure 6. Measuring and modelling Bode �a�, Nyquist �b�, and phase-angle
�c� plots for 30% SB2 flux with experiment A at 3 h.
Journal of The Electrochemic
capacitance from edge connector and leads was measured, and the
value was 2.9 � 10−11 F. So, the capacitance for this comb line
geometry was 4.6 � 10−11 F. Based on these values, and using Eq.
3 and 4, the solid lines were plotted in Fig. 6 following an optimi-
Downloaded 16 Aug 2010 to 121.8.210.35. Redistribution subject to EC
zation process. There was good agreement between measured and
modelled spectra, with the optimized values within 10% of the start-
ing values.
Comparison of SIR and impedance results.— As mentioned
above, RC can be estimated using the impedance value at low fre-
quency, because here the capacitive property of the comb is insig-
nificant. Hence, the impedance at 0.1 Hz should be similar to the
SIR value, and this is evaluated in Fig. 8-11, where these properties
are plotted as a function of exposure time. The plotted data are an
average from three measurements. There are some salient points to
note from these figures. For two solvent-based fluxes, SB1 and SB2,
the impedance responses with time were similar with different con-
ditioning and were close to the SIR value as well. This probably
suggests that the increase in impedance and SIR values with time
was most likely caused by flux volatilization and not the electrical
conditioning. For the two water-based fluxes, WB1 and WB2, how-
ever, the impedance and SIR with dc conditioning were significantly
higher than that with no conditioning and that with ac conditioning.
RcRct Zw
Electrode reaction
RC
CC
Figure 7. Equivalent circuit for four fluxes in Table I.
1.E+06
1.E+07
1.E+08
1.E+09
1.E+10
Im
pe
da
nc
e
(o
hm
)
.
Experiment A
Experiment B
Experiment C
Experiment D
C11ciety, 156 �1� C8-C15 �2009� C11
0 10 20 30
Time (hour)
Figure 8. Impedance �A, B, and C� and SIR �D� results for 50% SB1 flux.
S license or copyright; see http://www.ecsdl.org/terms_use.jsp
al So
This is not surprising, as the dc conditioning results in nonreversible
local-ion-concentration polarization1 and the formation of corrosion
products �oxides or hydroxides�. This was confirmed by only apply-
ing voltage periodically in the SIR test, as shown by the violet curve
in Fig. 10 and 11. Here, the conditioning with dc and ac produced
results that were in agreement.
Dendrite formation with ac and dc conditioning.— Dendrite
formation is an important failure mechanism and can lead to cata-
strophic failure of electronic boards. In a further set of tests, the
concentration of the SB1 flux was increased to 80% to deliberately
promote dendrite formation in the SIR test, and results are presented
in Fig. 12 showing two typical SIR responses from dendrite forma-
tion. The rapid change on SIR in Fig. 12 indicates dendrite forma-
tion on the board. This is because when metallic dendrites start to
grow from the cathode, the SIR is dramatically reduced. The SIR
value drops to zero if the dendrite reaches the anode and results in a
short circuit. The high current due to the short circuit can burn out
the dendrites, and then the SIR value rapid increases again. In Fig.
12, the SIR approaches a minimum value of 106 �. This is because,
in the Auto-SIR measurement system, a 106 � resistor on each mea-
surement channel is installed to preserve dendrite formed; therefore,
the SIR value should be no less than 106 � even when the board is
short-circuited. The ac impedance was similarly measured on boards
with 80% SB1 flux after 1 and 24 h exposures. During these imped-
ance tests, two conditioning modes were applied, one using 5 V dc
and the second using 5 V ac at 50 Hz. These results are shown in
Fig. 13, where the impedance was slightly higher wit