See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/256545067

An Experimental Investigation on the Nature of Reverse Current of Silicon

Power pn-Junctions

Article in IEEE Transactions on Electron Devices · January 2002

DOI: 10.1109/TED.2002.1291850

CITATIONS READS
34 928

1 author:

Vasile Obreja
National Institute for Research and Development in Microtechnologies - IMT Bucharest
67 PUBLICATIONS 908 CITATIONS

SEE PROFILE

All content following this page was uploaded by Vasile Obreja on 29 May 2014.

The user has requested enhancement of the downloaded file.

IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 49, NO. 1, JANUARY 2002 155

An Experimental Investigation on the Nature of

Reverse Current of Silicon Power pn-Junctions
Vasile V. N. Obreja

Abstract—At this time, little importance is given to the surface is the charge carriers intrinsic concentration. Power diodes in-
component of reverse leakage current of silicon pn-junctions al- volved in new developed applications [9], [10] may experience
though reliability issues reveal device blocking weakness when per- high surge values of the current in reverse direction with high
forming at high voltage and temperature. Junctions which have
almost the same perimeter but different area have been realized reached peaks of the junction temperature.
and their reverse current has been measured at room temperature Fewer experimental results about behavior at reverse bias
and high temperature both for standard recovery and fast recovery voltage and high operation temperature of silicon power
(gold-doped or electron irradiated) pn-junctions. It is shown that junctions have been published. In [1], are shown reverse
for standard recovery junctions the surface component of the re- current-voltage ( - ) characteristics from room temperature
verse current is the primary component from room temperature
up to above 150 C and has influence on reaching high permissible up to 220 C for a silicon rectifier diode of 16.5 mm junction
working voltages. For gold-doped or electron irradiated junctions, diameter. The characteristics are shown up to 200 V reverse
the bulk component is dominant at high junction temperature, but voltage. of about 7 mA at 10 V reverse voltage and around
it is shown that a comparatively negligible surface component can 10 mA at 100 V are evidenced as typical values for 200 C
impose lower reverse working voltages or lower junction operation
junction temperature. The analysis carried out concludes that
temperature. The surface component may be a cause of limitation
on the operation of power silicon diodes at high reverse voltage the surface component of reverse current is negligible in
above 175–200 C junction temperature. comparison with the bulk component. Later in [4], results about
Index Terms—High temperature, high voltage, junction passiva- the level of at 2000 V reverse voltage and 175 C junction
tion, leakage current, semiconductor device. temperature were reported for a power rectifier diode whose
structure dimensions do not differ much from the corresponding
ones for the diode used in [1]. Statistical data indicated that
I. INTRODUCTION
for polymide passivated diode structures, was above 10

A T this time, the junction operation temperature

icon devices is limited to 175–200 C. For silicon fast
recovery rectifiers, the maximum permissible junction temper-
of sil- mA up to 30 mA, while most of the structures passivated with
a double layer of silicon dioxide/silicon nitride, exhibited
values lower than 10 mA. It was mentioned that significant
ature is even lower (125–150 C). Attempts of operation of reduction of by changing the passivation dielectric was due
silicon junctions at higher temperature than above may lead to the surface component. After two years in [2], electrical
to device failure caused by a loss of junction blocking capa- characteristics are shown at room temperature and 125 C for
bility and finally a device definitive electrical shortcircuit. The power thyristors and it was stated that the surface component
experience reveals that the electrical shortcircuit is usually lo- is negligible. In [3], a reverse electrical characteristic for a
cated on the junction peripheral surface of the device structure. power diode of 50 mm diameter is shown at room temperature.
For devices with a higher junction operation temperature than The level of is approximately 0.2 A at 10 V reaching
200 C, other semiconductor materials have been considered, approximately 2 A at 4000 V near the breakdown voltage
e.g., SiC or GaAs. Nevertheless, commercial devices based on value. - characteristics at high are not shown for the
such semiconductors demonstrating better performance in ap- diode, but at 120 C, high-voltage thyristor (50 mm di-
plications than silicon devices have not appeared yet. ameter) characteristics are shown indicating increasing voltage
Despite the common opinion that the reverse leakage current dependence of from about 10 mA at 1500–2000 V up to
of a silicon pn-junction would be determined mainly by the 50 mA and even more at 4000 V. A similar increased voltage
bulk component [1]–[3], there have been works reporting that dependence is revealed in [2]. Results regarding behavior of
the influence of the junction peripheral surface is not of low im- at high junction temperature have also been reported in
portance [4]–[8]. An experimental study reported in [5] for mesa [11] and [12]. To the our knowledge, other significant details
SiO passivated power silicon junctions, concluded that “the in- about the reverse current at high junction temperature which is
crease of 25 C A to 100 C A is too normal for device operation have not been published. This may
small to be explained by diffusion controlled currents be due to the competition power device companies face, which
and too large to be explained by generation currents .” determines that proprietary relevant results be kept unknown.
A possible influence of the surface component of on device
Manuscript received December 4, 2000. The review of this paper was ar- performance is practically ignored in the literature.
ranged by Editor M. A. Shibib. An analysis of technical specifications for power silicon de-
The author is with the Institute for Microtechnology (IMT), Bucharest 72996,
Romania (e-mail: vasileo@imt.ro; vnobreja@hotmail.com). vices manufactured by leading companies in the field shows that
Publisher Item Identifier S 0018-9383(02)00213-7. no significant progress has been achieved in the reduction of
0018–9383/02$17.00 © 2002 IEEE
156 IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 49, NO. 1, JANUARY 2002

leakage current, e.g., of silicon rectifiers in the last 20 years.The

purpose of this paper is to present results of an experimental in-
vestigation related to the nature of leakage current of both stan-
dard recovery and fast recovery silicon pn-junctions.

II. EXPERIMENTAL PROCEDURE

To find out if the reverse leakage current is determined by
bulk or junction surface phenomena, usually a correlation of its
level with the area or perimeter of the junction is made. Previous
reported experiments are related to structures of the same area
of the junction but a very different perimeter on the same wafer.
In other experiment, structures of different area and perimeter
have been used. The drawback of such experiments consists
in inaccurate results when a dominant surface reverse current
is not uniformly distributed around the junction perimeter or
when local insufficiencies exist at the junction peripheral sur-
face where most of the reverse current may be concentrated.
The experiment for this paper uses junctions that have almost
the same perimeter, but very different area. As shown in Fig. 1,
the three plane junctions existing on the same diffused silicon
structure have dimensions so that the perimeter ratio is quite
close to unity and the junction area ratio is very different from
unity. Such an experiment has been made on silicon standard
diffused wafers and on wafers of the same type excepting an
additional gold doping at a temperature higher than 800 C or
electron irradiation has been performed. Fig. 1. Three junctions on the same silicon diffused structure (one central
To obtain structures like that shown in Fig. 1, the silicon dif- junction and two outer half-ring junctions) of almost the same perimeter but
fused wafer has been masked with acid resistant wax on both very different area.
sides. Then, on the side, two concentric circles of close radii
are scratched in the wax layer and from two points of the inner entire perimeter of the central junction. After removing the wax
circle, which are diametrically opposed, further scratching in from structures with the three separated junctions, as shown in
the wax layer is made toward the edge of the wafer. The uncov- Fig. 1, no careful cleaning of the junction peripheral surface
ered wax path obtained in the wax layer has a width of approxi- followed by dielectric deposition and treatments (passivation)
mately 0.2 mm. Some obtained structures do not have identical was performed. Thus, without a protective dielectric layer, the
outer half-ring junctions. This is because in the wax layer were pn-junctions have been exposed to contamination from the sur-
formed two external half-circles with different radii so that in rounding ambience. In such a way, a large surface component
Fig. 1, the edge of the structure is determined by two half-circles of has intentionally been favored so that its properties can be
with little difference between their radii. Thus, a different area better revealed. Also, the junction termination is plane almost at
is obtained for the half-ring junctions. a right angle with the peripheral surface. This was obtained by
The wafer then was introduced in a HF HNO CH COOH using an acid etching solution with a low content of hydroflu-
acid solution where the silicon was etched along the path existed oric acid.
in the wax. After the depth of the groove reached 200–250 m Commercially available high-power silicon diodes and
(the thickness of the diffused wafers was of 350–400 m with thyristors have a proper bevel angle junction termination (mesa
a P layer of 70 m and layer of 50 m), the path in the junction) obtained after careful surface cleaning followed by
wax formed by the inner circle and the two segments via the a dielectric layer deposition and suitable heat treatments to
second circle was covered with wax, and the wafer was again reduce as much as possible the level of the surface component
kept in acid solution until the wafer peripheral region delin- of Organic dielectrics, such as silicone rubber, are still used
eated by the outer scratched circle is completely separated. We for junction passivation for large area structures, but for lower
have used this simple technique to obtain structures like the area structures, inorganic dielectric, such as glass passivation
one shown in Fig. 1, which has the advantage of not requiring is used. Other devices use terminations related to the planar
specific equipment. The width of the obtained groove is higher technology to avoid the effect of the junction curvature on the
than that of the initial path in wax due to some lateral etching. maximum reachable voltage, an effect which is similar to mesa
Nonetheless, other advanced methods can be used, e.g., plasma junctions with negative bevel angle. Nonetheles, as the experi-
etching, laser cutting, sandblasting, etc. Planar junctions with ence reveals, only a positive bevel angle for plane junctions or
advanced termination can also be realized on the same wafer diffused rings or field plates for planar junctions are not enough
with area/perimeter features as in Fig. 1, having the advantage to prevent a large surface leakage current component. Only a
that the outer half-ring junctions include in their perimeter the proper realized silicon dielectric interface around the junction
OBREJA: NATURE OF REVERSE CURRENT OF SILICON POWER pn-JUNCTIONS 157

Fig. 2. Reverse electrical characteristics at room temperature for silicon Fig. 4. Reverse electrical characteristics for silicon junctions of approximately
junctions of approximately the same perimeter but very different area. Solid the same perimeter but very different area at high junction temperature 150 C
line: the same surface treatment. Dashed and dotted line: different surface (the same structures used for Figs. 2–3).
treatment. The dimensions according to the structure shown in Fig. 1.

Fig. 5. Electrical characteristics at high junction temperature for three silicon

junctions of approximately the same perimeter but very different area (standard
Fig. 3. Reverse electrical characteristics at 100 C for the same silicon recovery diode structure also used for Figs. 2–4).
junctions used in Fig. 2 of approximately the same perimeter but very different
area, according to the structure from Fig. 1.
atoms/cm for the starting N-type silicon have been
used. The initial diameter of wafers was 50 mm. On several
can keep the surface component at an acceptable reduced level
diffused wafers, an additional gold diffusion at 900 C has been
for good device performance.
made. The reverse recovery time measured for wafers without
gold diffusion was 15–20 s, whereas for gold diffused wafers,
III. EXPERIMENTAL RESULTS it was less than 0.5 s. The central junction has a diameter of
One group of experimental results is illustrated in Figs. 2–5. 38 mm. The outer half-ring junctions were etched at a diameter
Diffused P NN wafers with an impurity doping of about of 40–43 mm, as specified in Figs. 2–5, and some difference
158 IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 49, NO. 1, JANUARY 2002

exists between their areas so that the perimeter ratio of the central
and a half-ring junction is not lower than 0.85, but the area
ratio is of 10 or 14. The measurements have been performed on
nonpassivated junctions and, as mentioned earlier, without any
measure to minimize the surface reverse current component. In
Figs. 2–4, reverse current–voltage – characteristics at room
temperature, 100 C, and 150 C are shown for standard diffused
and gold-doped structures with the three junctions separated, as
shown in Fig. 1. As expected, the reverse current is dominated by
the surface component in the case of standard diffused junctions
and by the bulk component in the case of fast recovery junctions
obtained by gold doping. After the gold doping at 900 C, the
increase of reverse current may exceed two orders of magnitude.
From Fig. 2, one can see that an additional etching of the central
junction may lead to visible change in the level of reverse current
for the standard recovery junction, but not for the fast recovery
junction (the dashed line characteristic is shown only for the
standard diode because for the fast diode, the new electrical char-
acteristic is identical with the previous characteristic). In the case
of fast recovery junction, some increase of on behalf of the
surface component becomes visible after the central junctions
were electrochemically contacted by a layer of nickel and then Fig. 6. Electrical characteristics at room temperature and high junction tem-
again etched for a short time (the dotted line characteristics). For peratures for three gold diffused junctions at 850 C (approximately the same
the standard junction, this surface treatment leads to a significant perimeter but different area). Solid line—the same junction surface treatment.
Dotted line—different junction surface treatment.
increase of of about one order of magnitude. At higher
junction temperature, as shown in Figs. 3–4, the corresponding
electrical characteristics for the mentioned surface treatments Fig. 7 shows current–voltage characteristics for electron irra-
of the central junctions (dotted line or dashed line) have been diated junctions. The structure with the three junctions has been
not shown anymore because for the fast diode there is no visible obtained this time by using a nonpassivated electron irradiated
change in the level of reverse current and for the standard diode mesa structure of 9.5 mm diameter with slight positive bevel
the change of is lower than for the room temperature case angle which is not significantly different from a right angle.
(under one order of magnitude). Fig. 5 shows current–voltage Only an inner circle of 7 mm diameter was necessary to obtain a
characteristics for the standard recovery junctions at tempera- structure like that shown in Fig. 1. Before irradiation, a value
tures higher than 100–150 C. In comparison with the situation of about 35 s was measured. After irradiation, a value of about
at 100 C, the reverse current of the central junction becomes 1.5 s was found. A storage for 3–5 h of the irradiated structures
higher than the corresponding level for the outer half-ring at 300 C has led to a decrease of the reverse leakage current and
junctions. Nevertheless, the difference remains under one order an increase of to about 12 s. As seen from Fig. 7, no signif-
of magnitude at a junction temperature of 270 C so that the icant difference exists between the reverse current of the central
surface component is not negligible in comparison with the junction and the two half-ring junctions, from room temperature
bulk component as in the case of gold-doped junctions. up to 150 C. Such a difference is visible at temperatures higher
If the gold diffusion is made at a lower temperature, e.g., than 150 C. For electron irradiated junctions, it has been found
850 C, then values of around 1.0 s are obtained and the that even after their passivation [13], the surface component of
level of the reverse current is lower than the corresponding one the reverse current may be dominant up to 100–125 C.
for a diffusion at 900 C. – characteristics for junctions dif-
fused at 850 C are shown in Fig. 6. This time, an influence
IV. DISCUSSION
of the surface component on the current–voltage characteristics
can be seen at room temperature. At low reverse voltage, the The above mentioned experimental results show that un-
leakage current ratio is lower than 6 (the junction area ratio) and controlled electrical phenomena taking place at the junction
higher than 6 toward 100 V. A short etching after nickel contact peripheral surface can seriously limit the permissible device
metallization of the junction peripheral surface of the central working reverse voltage due to strong voltage dependence of
structure reveals an influence on the reverse current from low , appearing at much lower reverse voltage than for the case of
reverse voltage (the dotted line characteristic) at room temper- similar passivated devices. Our target has been to passivate the
ature but not at higher temperature. For of 3–5 s (obtained structures used for Figs. 2–7 in order to minimize the surface
by gold doping at a temperature lower than 850 C), compa- component of and to reach reverse voltages of 1000 V
rable values of for the three junctions have been found up to and even higher, as is the case for standard manufactured
100–125 C, but at higher temperature the reverse current of the diodes when using silicon diffused wafers with the same
central junction is dominant. characteristics. Unfortunately, to date, the experiment could
OBREJA: NATURE OF REVERSE CURRENT OF SILICON POWER pn-JUNCTIONS 159

certain value of the reverse voltage, the surface component still

can limit the junction permissible working voltage, in a similar
way as illustrated in Figs. 2–7, at a higher value still well
below the expected bulk breakdown value. For diodes with low
carrier lifetime in their base, although the surface component
may be negligible in comparison with the bulk component, its
increased voltage dependence from some reverse bias is not
visible because of the dominant bulk component. When of
the fast diode becomes more voltage dependent, its surface
component has already reached a comparable level with the
bulk component. At junction temperature above 100 C, in
such a situation, risk of failure may appear. By passivation of
lower area mesa junction structures mm separated
from similar wafers used for the results shown in Figs. 2–7,
reverse voltages higher than 1000 V have been obtained for
most of the structures but stable breakdown voltage values of
1600–2000 V corresponding to the resistivity of initial starting
silicon could hardly be reached. It has been found that ,after
another passivation for the same structure, a lower or a higher
reverse voltage may be reached both for standard recovery and
fast recovery junctions [11], due to the fluctuations in the level
of the surface component. At junction temperature higher than
Fig. 7. Electrical characteristics at room temperature and high temperature for 80–100 C, lower permissible values have been obtained
three electron irradiated junctions of about the same perimeter but different area
situated on the same structure (see Fig. 1).
for gold-doped diodes than for undoped ones.
As it is known, the pn-junction current–voltage characteristic
theory provides a quantitative relation for the bulk component
not be completed because of difficulties encountered with of In the case of a one-sided junction, as is the case in the
the passivation procedure utilized for junction devices with above mentioned experimental results, the bulk component of
only one peripheral surface. It is difficult to use it to multiple is given by
junction structures like those shown in Fig. 1. To discover
whether the results shown in Figs. 2–7 are not particular results (1)
corresponding to a structure with groove, as shown in Fig. 1,
similar diode structures with a diameter close to that of the
where
central junction but with positive junction bevel angle have and constants related to the junction physical param-
been observed after acid etching and a rinse in DI water of eters;
their peripheral surface (according to standard device manufac- carrier intrinsic concentration;
turing). They were left a long time to contact the air ambience effective doping density in the diode base;
without dielectric layer deposition. It was discovered that some minority carrier lifetime in the junction base;
of them manifested – characteristics revealing almost the effective lifetime for the space charge region
same behavior regarding the level of the leakage current and its formed in the low-doped side;
voltage dependence, as shown in Figs. 2–7. For other structures, applied reverse voltage.
better reverse characteristics were found but very few reached The exponent “ ” for the above analyzed junctions should have
1000 V. By depositing a silicone rubber layer on the junction a value of one-half because at higher reverse voltage, the junc-
peripheral surface of such diode structures after a fresh etching tions can be considered abrupt. Only at low values the junc-
and rinse in DI water, followed by a high-temperature treatment tion may be considered linearly graded so that in this case the
at 175–200 C, significantly lower values of and higher exponent can have a value of one-third. The first term in (1) is
reverse voltages reaching 1000 V and above were obtained for related to the diffusion component and the second one to the
most of the standard recovery junctions. The reduction of generation current component.
in comparison with the initial level before passivation occurred From Fig. 2, one can see that a dominant surface compo-
from reverse voltage was as low as 0.1 V up to high values. nent of can vary approximately as for the standard
For the fast recovery junctions, higher reverse voltages toward recovery diode at room temperature (solid line for the central
1000 V have been obtained but visible reduction of level was junction) or as (dashed line). Further excessive increase
observed above 10–50 V reverse voltage. Other experimental of the surface component (the dotted line for the central junc-
data suggest that, in spite of the fact that the surface component tion in Fig. 2) leads to a reverse voltage dependence of ap-
of is significantly reduced by passivation, it continues to proximately . Almost the same voltage dependence is also
have an important contribution to the level of for long carrier exhibited by the gold-doped junctions where the bulk compo-
lifetime diodes. Due to its increased voltage dependence from a nent is dominant. Consequently, a voltage dependence found
160 IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 49, NO. 1, JANUARY 2002

according to the second term in(1) even for standard fabricated and takes place after irradiation, but the reverse current level
high-voltage diodes cannot be considered sure experimental evi- does not change significantly up to 100–125 C [13], as is the
dence for a dominance of the generated current in the bulk space case for gold-doped junctions. As one can see from Fig. 7 after
charge region because a dominant surface component of can high-temperature annealing, due to a reduction of the genera-
have a similar dependence [11]. The – characteristic with tion-recombination centers leading to increase of about an
unusual voltage dependence for the one of outer half-ring junc- order of magnitude, lower values of the reverse current are ob-
tions (Fig. 2) has been caused by possible mechanical damage tained. Above 150 C this current decrease takes place mainly
to the crystalline lattice near the junction perimeter which acci- on behalf of the bulk component but below 150 C lower values
dentally has occurred in measurements. Such a behavior is due of reverse current are due to a reduction of the surface com-
to an excessive high surface component. As can been seen from ponent. At 24 C, from Fig. 7, one can even see a tendency
Figs. 3–5, this anomaly disappears at high temperature. While of increase of after carrier lifetime increase because of the
at high junction temperature (see Figs. 3 and 4) almost the same fluctuations in the level of the dominant surface component for
voltage dependence for the room temperature case is manifested nonpassivated junctions. From Fig. 7, it can be seen that the in-
by the heavy gold-doped junctions; for the long carrier lifetime troduction of generation-recombination centrers by irradiation
junctions less voltage dependence is exhibited. A saturation ten- may also increase the level of the surface component in compar-
dency is visible, for example, at 150 C at low voltage, although ison with the situation when these centers are not present. The
the surface leakage current is still the primary component. The same behavior may be valid for irradiation with light ions. A sig-
behavior at 200 C and 270 C (see Fig. 5) could be caused nificant change of the effective doping density in the diode base
by the bulk diffusion component becoming noticeable above after irradiation has not been found for diodes with an avalanche
150 C. A saturation tendency of the current, especially for the breakdown voltage of 800–1000 V.
central junction, could support such a supposition. Nevertheless, As a consequence, the generation current component given
such a saturation tendency has been observed in other experi- by (1) is not very useful for explaining the behavior at reverse
ments when the surface component is dominant. On the other bias voltage of power junctions.According to [11], a variation
hand, for another experiment using standard recovery wafers such as or 0.6 eV activation energy has also been found for a
with a of 40–45 s, no significant difference between the dominant surface component. Concerning behavior at low for-
reverse current of the central junction and of the half-ring junc- ward voltage of 0.1–0.3 V, it has been found that the forward
tions has been found at 200–250 C. Taking into account that, current of the three junctions from the standard diffused wafer
for the junctions used in Fig. 5 has a value of about 20 s, (Figs. 2–4) does not differ by more than a factor of two. Conse-
it is possible that above 150–200 C, the diffusion bulk compo- quently, the forward current at low voltage may be dominated by
nent will become visible. A significant surface component still the surface component instead of the bulk recombination com-
present at 250–300 C may explain why present-day manufac- ponent, as was also revealed in [11]. Almost the same difference,
tured silicon standard recovery diodes cannot have a reliable op- by a factor at least ten, has been found for the low-voltage for-
eration at such temperature values and high reverse voltage. To ward current as for the reverse current in the case of gold-doped
reach such junction operation temperatures for silicon devices, junctions. On the other hand, fast diodes obtained by electron
further advancement in the junction passivation techniques is irradiation may have the same value of near room tempera-
necessary to enable a lower level of the surface reverse current ture as the similar nonirradiated ones, although at low forward
component in comparison with the present state of the art. voltage the direct current is significantly different. Such a be-
An excessive high level for the of gold-doped junctions havior at low forward bias voltage could be attributed to the
in comparison with similar undoped junctions could be under- recombination current component in the depletion layer of the
stood according to (1) by significant decrease of and as a gold-doped or electron irradiated junction. However, this is not
consequence of generation-recombination centres presence. A supported by a visible voltage dependence of the reverse cur-
dominance of the first term component from (1) for gold-doped rent due to the generation of carriers in the depletion layer for
junctions seems to be supported by less voltage dependence of gold-doped junctions or bulk component change at room
reverse current. Also, some change of the effective doping den- temperature after irradiation.
sity due to the presence of the gold atoms in the base may In [12], this behavior was explained by a drastic reduction of
contribute to the level of if in (1) decreases by com- the junction built-in voltage in the case of the fast recovery diode
pensation of some donor centers when heavy doping is per- due to the low carrier lifetime. For this reason, at low forward bias
formed. Nonetheless, for gold doping at 850 C, no significant voltage the fast recovery junction is more open for the flowing
change in the effective doping concentration in the base has current than a similar standard recovery junction. Consequently,
been found [12], because about the same avalanche breakdown a significant forward current flows through the gold-doped junc-
voltage value of approximately 1300 V has been found before tion after little increase of the direct bias voltage in comparison
and after the gold doping of low-power rectifier diodes. Experi- with the similar undoped junction. For this reason, at reverse bias
mental observations may indicate visible difference in the level voltage for the gold-doped junction, the current in our opinion,
of for standard recovery junctions of the same area but with is mainly determined from carriers generated in the low-doped
significant different values. Nonetheless, such a behavior is neutral region (diffusion component), because the density of
caused by the surface component[12]. charge carriers near the border of the depletion layer is much
What cannot be explained by (1) is the behavior of electron higher than for the standard recovery junction. A gradient dis-
irradiated junctions. A significant change of and hence of tribution of the carriers along the neutral low-doped region
OBREJA: NATURE OF REVERSE CURRENT OF SILICON POWER pn-JUNCTIONS 161

starting from the edge of the depletion layer at reverse bias, al- using the planar process. Special junction terminations, such as
most does not exist for the gold-doped diode as in the case of the multiple rings and other methods, enabled planar junctions to
similar standard recovery diode, due to much lower carrier life- overcome the junction curvature effect and to reach permissible
time. A mismatch of activation energy value found experimen- working reverse voltages closer to the values obtained by using
tally under 1.0 eV, despite the fact that the first term of (1) indi- mesa technology (plane junctions). Inorganic dielectrics, such as
cates proportionality with , could be attributed to the fact that SiO and other suitable junction protective layers for the planar
some suppositions admitted for (1) may be not valid (e.g., low- process, are used instead of organic dielectrics in the mesa tech-
level injection) in the case of gold-doped junctions. nology. The question now arises: “Does the surface component
For standard recovery junctions it has been shown [11] that, of have any influence on the – reverse electrical character-
after the passivation process, significant lower values of reverse istic of advanced junctions?” As demonstrated in [11], the sur-
leakage current are obtained both at low and high reverse voltage face component of and its related effects do not disappear by
due to the reduction of the surface component. Nevertheless, the using advanced developed processes. As an example, we could
surface current component can still remain the primary compo- mention that N P junctions fabricated by using an n-channel
nent. For gold-doped junctions, a reduction of the reverse cur- MOSFET process exhibited excessive high leakage reverse cur-
rent after passivation may take place at higher reverse voltage rent when some change was made in the nitrous oxide passiva-
but not at low voltage. A reduction of the surface component tion layer [14]. Such a junction surface treatment, which is com-
after passivation is not visible at low reverse voltage because pletely different from the treatment applied to the central junc-
the bulk component is dominant even when excessive surface tion of the standard recovery diode in Fig. 2 (dotted curve) leads
leakage current is present. At high reverse voltage where the to a similar effect. Furthermore, as demonstrated in [8], advanced
surface component may become comparable with the bulk com- power diodes manufactured by using a junction termination ex-
ponent, this reduction is detectable. tension exhibited serious reduction of the reverse voltage (ini-
A negligible surface component in comparison with the tially of 3500 V) to 2000 V only after a high-temperature re-
bulk component for gold-doped diodes does not mean that this verse bias stress of 1000 V at 165 C for 7000 s. Such a behavior
component cannot influence their performance. As shown in was attributed to charge changes at the silicon nitrous oxide cov-
Figs. 2–4 for the standard recovery junctions, their – char- ering the junction peripheral surface. A similar effect of lowering
acteristics could not be explored above 100–300 V because of the initial measured reverse voltage well above 1000 V has been
device failure imminence due to the surface breakdown caused found for mesa junction devices with passivation insufficiencies
by excessive high surface reverse current. The same behavior when they are stored a long time without applied reverse bias
is valid for gold-doped junctions. Such a limitation is still valid voltage at a temperature of 150–175 C. These are not the only
for higher voltages above 1000 V, available after passivation examples which prove that the stability of the reverse current of
when the maximum permissible working voltage cannot reach today’s power device junctions still depends on electrical phe-
values close to the theoretical bulk breakdown voltage. Even nomena taking place at the semiconductor-dielectric interface
in the case where the surface component for gold-doped which is in contact with the pn-junction.
diodes is at the same level as for the similar standard recovery
ones, this hidden component may lead to an imminent device
V. PN-JUNCTION PHYSICAL MODEL FOR
failure without the possibility of its monitoring by means of
THE SURFACE COMPONENT
the noticeable increase of the reverse current. Unlike the bulk
component, the surface component is nonuniformly distributed We know from semiconductor surface physics, that the
over the junction area. It flows only through a very thin surface charge due to surface states or fixed ions induces a
layer at the junction peripheral surface. Moreover, the current space charge layer in the semiconductor surface region. In
distribution around the junction perimeter may be nonuniform. our opinion, although considerable advancements have taken
Consequently, a surface current lower than the bulk dominant place in the silicon junction manufacturing processes, this
current may cause a local overheating in comparison with surface space charge layer is still unavoidable. When little
the rest of the silicon structure. This overheating can lead the importance is given to the contamination of the junction
reverse electrical characteristic instability through local thermal peripheral surface from the surrounding ambience, as is the
runaway, resulting in catastrophic device failure. Because the case with the results shown in Figs. 2–7, then a strong surface
surface component is not very noticeable, it is more difficult for accumulation layer (see Fig. 8) or a strong inversion layer
fast recovery diodes to reach the same high working voltages (see Fig. 9) adjacent to the semiconductor-dielectric interface
as standard recovery diodes. As a consequence, the task of a is present. Such semiconductor surface layers are favored by
tight control by passivation of the surface component for a silicon of higher resistivity. These surface layers are first shown
gold-doped diode that can be negligible in comparison with its at zero bias voltage both for a mesa and a planar junction.
exhibited current is important for the standard recovery diodes. A deposition of dielectric layer on an already contaminated,
The results and discussions presented thus far in this paper noncleaned junction peripheral surface has almost no effect on
have been related to mesa technology mostly used in the past for the – characteristics. When reverse bias voltage is applied,
the manufacture of power silicon devices. The availability of the the formed junction bulk depletion layer is affected near the
planar technology that has contributed to significant advance in junction peripheral surface. A lower width is possible at the
the development of power MOS and IGBT power devices has junction peripheral surface than in its bulk. Consequently, a
determined that even simple junction diodes can be realized by higher intensity for the electric field is favored although positive
162 IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 49, NO. 1, JANUARY 2002

layer adjacent to the dielectric layer acts as some kind of shunt

for the reverse electrical current, and in this way a surface
component is still favored for present-day manufactured power
junctions. At high junction temperature normally encountered
in the operation of high-voltage devices, an overheating near
the junction peripheral surface is possible which, over time,
can induce structural changes at the silicon-dielectric interface
and hence in the surface space charge layer not existing at the
initial phase of operation. This may explain reverse current
instabilities during device operation. Consequently, a higher
level of reverse current and even reverse voltage reduction can
appear leading to device failure. For this reason, operation
of devices at high voltage is not usually permitted above
150–175 C junction temperature.

VI. CONCLUSION
Fig. 8. Strong surface accumulation layers for mesa and planar junctions (1:
passivation dielectric layer; 2: surface space charge layer). (a) Mesa junction, An experiment for mesa silicon junctions of about the same
zero bias voltage. (b) Planar junction, zero bias voltage. (c) Mesa junction under
reverse bias voltage. (d) Planar junction under reverse bias voltage.
perimeter but significant different area has revealed a possible
contribution of the surface component to the level of reverse
leakage current of present-day manufactured pn-junction silicon
power devices. For gold-doped diodes known for their high
excessive leakage reverse current, it has been shown that, in
spite of a dominant bulk component, even an initial negligible by
comparison surface component may limit the maximum permis-
sible reverse working voltage or maximum permissible junction
operation temperature. For electron irradiated junctions, it has
been found that the surface component may be the primary
component up to 100–150 C. For standard recovery junctions
(high carrier lifetime in their base) the surface component
of reverse current may be dominant from room temperature
up to 250–300 C junction temperature. A limitation of the
permissible reverse working voltage or of the junction operation
temperature of present-day manufactured silicon power devices,
is due to junction reverse electrical characteristic instability
leading to catastrophic device failure through thermal runaway
Fig. 9. Strong surface inversion layers for mesa and planar junctions (1:
caused by the surface component. A pn-junction rough physical
passivation dielectric layer; 2 and 3: surface space charge layer formed by model taking into consideration electrical phenomena near the
inversion and depletion layers). (a) Mesa junction, zero bias voltage. (b) Planar silicon-dielectric interface provides at least a qualitative expla-
junction, zero bias voltage. (c) Mesa junction under reverse bias voltage. (d)
Planar junction under reverse bias voltage.
nation of the presented results. Advancement in the performance
of power silicon devices is possible by further reduction of the
surface reverse current component through suitable junction
bevel angle or proper junction termination for a planar process passivation techniques leading to optimization of the surface
is achieved. Due to higher electrical field, surface breakdown space electric charge layer both at low and high temperature.
may appear at lower voltages. Solid line and dashed line
characteristics in Fig. 2 can be attributed to an accumulation ACKNOWLEDGMENT
layer whereas dotted line characteristics can be attributed to
a strong inversion layer. By junction passivation consisting The author wishes to thank Dr. E. Lakatos for providing the
of proper cleaning treatments and dielectric covering of the gold diffused silicon wafers processed at Baneasa S.A. Com-
junction peripheral surface, strong accumulation or inversion pany; colleague G. Dinoiu, and E. Iliescu from IFTAR Bucharest
layers are eliminated and improvement in the reverse electrical for the electron irradiation of junction structures. Thanks are also
characteristics is obtained. Nevertheless, a surface space due to the anonymous reviewers whose pertinent comments have
charge region, usually as a surface depletion region or even contributed to the improvement of this paper.
weak inversion or accumulation layers, is present. An ideal
desired case without any surface charge layer is not possible. REFERENCES
Changes in the surface space charge layer may take place when [1] F. Dannhauser, “Analysis of bulk reverse current in diffused silicon
the temperature increases. At high junction temperature, a power rectifiers,” Solid-State Electron., vol. 10, p. 361, 1967.
[2] J. Cornu, S. Schweitzer, and O. Kuhn, “Double positive beveling a better
surface depletion layer may be present instead of inversion or edge contour for high-voltage devices,” IEEE Trans. Electron Devices,
accumulation layers. Consequently, this silicon surface charge vol. ED-21, pp. 181–189, Jan. 1974.
OBREJA: NATURE OF REVERSE CURRENT OF SILICON POWER pn-JUNCTIONS 163

[3] K. Platzoder and K. Loch, “High-voltage thyristors and diodes made of Vasile V. N. Obreja was born in Gorban, Romania, on
neutron irradiated silicon,” IEEE Trans. Electron Devices, vol. ED-23, May 3, 1946. He received the B.S.E.E., M.S.E.E., and
pp. 805–905, Apr. 1976. Ph.D. degrees from the Department of Electronics
[4] R. R. Verdeber et al., “SiO =Si N passivation of high-power recti- and Telecommunications Engineering, Bucharest
fiers,” IEEE Trans. Electron Devices, vol. ED-17, pp. 797–806, Apr. Polytechnic Institute, Bucharest, Romania, in 1969,
1970. 1980, and 1989, respectively. His thesis work in-
[5] R. E. Blaha and W. R. Fahrner, “Technology and reliability of thermal volved an innovative passivation technique for mesa
SiO layers as used for passivation of silicon power devices,” Phys. silicon devices with application in a high-volume
Status Solidi (A), vol. 50, p. 551, 1978. semiconductor fabrication and studies on the leakage
[6] B. J. Baliga and E. Sun, “Comparison of gold, platinum, and electron ir- reverse current at high junction temperature.
radiation for controlling lifetime in power rectifiers,” IEEE Trans. Elec- In 1969, he joined the R&D Center of the Radio
tron Devices, vol. ED-24, pp. 685–692, Mar. 1977. Components and Semiconductors Factory (IPRS Baneasa), Bucharest, where
[7] M. Bakowski and K. I. Lundstrom, “Depletion layer characteristics at he contributed significantly to prototyping and fabrication introduction of new
the surface of beveled high-voltage pn-junctions,” IEEE Trans. Electron types of silicon devices: Zener diodes, controlled-avalanche rectifier diodes,
Devices, vol. ED-20, pp. 550–558, Mar. 1973. thyristors, and rectifier bridges (modules). In 1974, he joined the newly founded
[8] J. R. Trost, R. S. Ridley, M. K. Khan, and T. Grebs, “The effect of charge Research Institute for Electronic Components (ICCE), where he was involved
in junction termination extention passivation dielectrics,” in Proc. 11th in device design, process, and packaging development, as well as characteriza-
Int. Symp. Power Semiconductor Devices and ICs, 1999, p. 189. tion and failure analysis of silicon diodes and thyristors manufactured by the
[9] M. Domeij, B. Breitholtz, L. M. Hillkirk, J. Linnros, and M. Ostling, neighbor factory (IPRS Baneasa). In the first half of 1979, he pursued graduate
“Dynamic avalanche in 3.3 kV Si power diodes,” IEEE Trans. Electron studies related to second breakdown in bipolar transistors and novel diffusion
Devices, vol. 46, pp. 781–790, Apr. 1999. techniques at the Liverpool University and Middlesex Polytechnic, U.K., on an
[10] Y. Tomomatsu et al., “An analysis and improvement of destruction im- UNIDO Fellowship. In 1982, he rejoined ICCE to work on manufacturing tech-
munity during reverse recovery for high-voltage planar diodes under nology development and small volume fabrication of new device types, such as
high dIrr=dt condition,” in Proc. 8th Int. Symp. Power Semiconductor silicon transient voltage suppressors and silicon microwave mixer and detector
Devices and ICs, 1996, p. 353. diodes required by home customers. He has performed studies on solid state
[11] V. Obreja, “On the leakage current of present-day manufactured semi- phenomena, such as second breakdown of silicon junctions, surface effects, and
conductor junctions,” Solid-State Electron., vol. 44, p. 49, 2000. behavior at high junction temperature and high avalanche currents. Since 1996,
[12] V. Obreja, G. Dinoiu, and E. Lakatos, “Effect of bulk carrier generation- he has been with the newly founded National R&D Institute for Microtech-
recombination centers upon the I -V characteristic of silicon pn-junc- nology (IMT), Bucharest, in which ICCE was included. He has been working
tions,” in Proc. 1998 Int. Semiconductor Conf., 1998, p. 293. on the development of different types of power modules (surge protection, recti-
[13] V. Obreja and G. Dinoiu, “On the nature of leakage current of silicon fier bridges, high-voltage rectifiers, MOS/IGBT, and solid state relays). He has
fast recovery rectifiers,” in Proc. 7th Europ. Conf. Power Electronics written more than 70 research or technical internal reports. He holds five Ro-
and Applications, 1997, p. 2.082. manian patents and has authored more than 30 papers in the solid-state device
[14] W. K. Mathews et al., “Degradation of junction leakage in devices sub- area in refereed journals/conferences.
jected to gate oxidation in nitrous oxide,” IEEE Electron Device Lett., Dr. Obreja’s biography has been selected for the 6th Edition of Who’s Who
vol. 13, pp. 648–650, Dec. 1992. in Science and Engineering (2002–2003).

View publication stats