Characteristics of high breakdown voltage Schottky barrier
diodes using p
+
-polycrystalline-silicon diused-guard-ring
Bor Wen Liou
a,*, Chung Len Lee
baDepartment of Electrical Engineering, Wu-Feng Jr. College of Technology and Commerce, 117 Chian-Ku Rd, Sec. 2, Chiayi, Taiwan,
Republic of China
bDepartment of Electronics Engineering and Institute of Electronics, National Chiao Tung University, 1001 Ta Hsueh Rd, Hsinchu,
Taiwan, Republic of China
Received 17 May 1999; received in revised form 14 July 1999; accepted 16 September 1999
Abstract
A new structure of Schottky diode using the p+-polycrystalline silicon (polysilicon) diused-guard-ring is
proposed. For 9508C with 30 min annealing condition, the diode gives a nearly ideal J±V characteristic with a high reverse breakdown voltage (148 V) and a low reverse leakage current density (8.4 mA/cm2). The guard-ring structure
prevents the premature breakdown; the polysilicon layer prevents the surface leakage. It was also found that the more the driving time of furnace, the higher the breakdown voltage of the Schottky diode. The breakdown characteristic of a Schottky diode was shown to be closely related with the diusion length of the boron ions being inside the Si wafer. The electrical characteristics of the p+-polysilicon diused-guard-ring Schottky device was
compared with those of the conventional diused-guard-ring sample. 7 2000 Published by Elsevier Science Ltd. All rights reserved.
1. Introduction
It is interesting that the Schottky diode, a majority carrier device, in which there is virtually no minority carrier storage, opens up a wide avenue of high-speed applications. It is essential to use the power recti®er with the high speed-switching capability. In compari-son, the conventional silicon recti®er has a low break-down voltage because of its excessive reverse leakage current and of its limiting forward conduction current-handing capability with increasing breakdown voltage
[1]. Hence, the epitaxial silicon wafer will be used for application of the high-power device in the future. Although a true reverse leakage current of the metal± semiconductor (Schottky) device has been dicult to interpret in theory, the diused-guard-ring structure can eliminate the premature breakdown and give a low leakage current [2,3]. Moreover, the overlap-metal structure [4] gives a near-ideal forward I±V character-istic and has a low leakage current density under the moderate reverse bias. Recently, many structures have been proposed [5] to prevent the electrode sharp-edge eect of the diode and give ideal forward and reverse characteristics. Similarly, the reverse direction of the polysilicon emitter (poly-emitter) diode of the p±n junction generally exhibits a lower leakage current and a higher breakdown voltage as compared to a conven-tional-Schottky diode [2,6]. Hence, we propose a new
0038-1101/00/$ - see front matter 7 2000 Published by Elsevier Science Ltd. All rights reserved. PII: S0038-1101(99)00258-0
* Corresponding author. Tel.: +886-5-226-7125-2531; fax: +886-5-226-4224.
E-mail addresses: [email protected] (B.W. Liou), [email protected] (C.L. Lee).
structure in this experiment, which can combine the advantage of a high reverse breakdown voltage of the poly-emitter structure with a fast response of the Schottky diode with the diused-guard-ring structure. The guard-ring structure of heavily doped polysilicon with a polysilicon-overlap structure onto the epitaxial wafer was used to demonstrate in this experiment. It was found that this new device not only took a low leakage current and high reverse breakdown voltage, but also maintained the advantage of the Schottky bar-rier. This fabricated device indeed demonstrates for the feature. In comparison, the conventional Schottky diode of diused guard-ring with the metal-overall structure and the poly-emitter diode was also investi-gated together. Furthermore, the native-oxide impact of polysilicon/mono-silicon interface was also studied together with this structure [7].
2. Experimental procedure
Since the investigation was mad to improve the understanding of the polysilicon overlap and guarding-ring structure can improve the breakdown voltage of
Schottky diode. Four dierent types of diodes were studied together and are depicted in Fig. 1. In Fig. 1(a), a conventional-planar-Schottky barrier employing the metal overlap is referred to as diode A. In Fig. 1(b), the p+diused-guarding diode without
polysili-con overlap is referred to as diode B. Fig. 1(c) shows a poly-emitter diode employing a polysilicon layer with polysilicon±metal overlap, which is referred to as diode C [6]. Finally, in Fig. 1(d), the proposed diode by using the guard-ring diode of p+-polysilicon-diused
with the 25 mm polysilicon guarding-ring and 30 mm polysilicon-overlap length, hereafter referred to as diode D. The wafer used is a n-type 0.01 O cm h100i wafer with a 20 mm thick epitaxial layer h100i orien-tation whose resistivity of the epi-layer was 10 O cm. Fig. 2 (a) is a process-¯ow chart of four-diode fabrica-tion and Fig. 2(b) is a sequential process step of pro-posed diode D. The fabricated-process sequence is as follows: ®rst, the wafer was cleaned by RCA process sequence and the SiO2 layer of approximately 7000 AÊ
was grown at 11008C, wet oxygen for 1 h; then de®ned the guard-ring pattern of the oxide ®lm. However, diode C did not de®ne the guard-ring pattern and used the metal ®lm to directly contact onto the polysilicon
®lm surface. Next, the undoped polysilicon (or amor-phous) ®lm of 3000 AÊ was deposited in a LPCVD (lower pressure chemical vapor deposition) reactor at 6258C (or 5508C) with SiH4 gas under a pressure of
100 mTorr [6]. The H2SO4 sample was cleaned by
H2SO4:H2O2(3:1) solution before the polysilicon layer
was deposited; likewise, the HF sample was dipped in dilute HF (1:50) solution for 40±50 s before the polysi-licon layer was deposited [8,9,16]. Thus, the H2SO4
sample exhibited a thin native oxide ®lm (<20 AÊ) at the polysilicon/mono-silicon interface [9]. Following this, the conventional diode of Fig. 1 (a) was implanted by BF2+at 40 keV with a dosage of 1 1016
cmÿ2. The structures of the diodes depicted in Fig.1(b±
d) received the BF2+ implantation at 120 keV with a
dosage of 1 1016 cmÿ2. After implantation, all the
samples were annealed at 9508C with 30±210 min in N2gas. Then diodes B and D were used for the
stan-dard photo-resist technique to form the polysilicon-guard-ring structure. Before the metal deposition, all the wafers were chemically cleaned by a RCA sequence and ®nally rinsed in a dilute HF solution. Finally, all the samples were formed by thermal evaporation of the Al ®lm (with 1% Si), sequentially sintered at 4008 C for 30 min in N2gas and the devices were patterned
to de®ne. The Keithley 237 and Keithley 590 equip-ment measured the current±voltage characteristic and the capacitance±voltage curve, respectively. The junc-tion depth of boron ions was measured by the SRP (spreading resistance pro®ling) technique of the SSM (Solid State Measurements) Company.
3. Results and discussion 3.1. Forward J±V characteristics
Fig. 3 shows the measured forward J±V character-istics of four dierent structures (diodes A±D) with HF and H2SO4, respectively. In this ®gure, several
facts can be observed: (1) the current density of diode A was less than that of diode D and diodes A and D exhibited a three logarithmic-decade dierence; (2) the conventional diode exhibited the lowest eective bar-rier height fbn10:73 eV as compared to the other
diodes; (3) the poly-emitter diode C had a good ideal-ity factor n11:002 for the 10 logarithmic-linear dec-ades; (4) the guard-ring structure of diode D exhibits a near-ideal forward characteristic and a high driving current density. The ideality factor, n, was about 1.07 for seven logarithmic decades and the forward current density was about 30 A/cm2at 0.6 V; and (5) the
eec-tive barrier height of the Schottky diode with naeec-tive oxide was lightly greater than that of the without-native-oxide sample. The eective barrier height, ideal-ity factor and breakdown leakage current densideal-ity Jbd
of these diodes are given together in Table 1. Because the native oxide of polysilicon/mono-silicon could block the minority carrier into the polysilicon ®lm, so that the eective barrier height of the HF diode (0.75 eV) is slightly smaller than that of the H2SO4(0.76 eV)
sample; likewise, the forward current of all diodes with the H2SO4 sample was also slightly smaller than that
of the HF sample. Similarly, the forward current of
Fig. 4. 1/C2versus applied voltages for fabricated diodes. The
polysilicon ®lm was of ADP structure. The furnace tempera-ture was 9508C and drive-in time 30 min.
Fig. 3. Forward current density versus applied voltage of the fabricated diodes and poly-emitter diodes. The polysilicon ®lm was ADP structure. The furnace temperature was 9508C and drive-in time was set at 30 min.
diode D was lightly smaller than that of diode B under the same forward bias. On the other hand, this for-ward current of the diode D consists of two com-ponents: one is a pure Schottky barrier current and another is a p+/n guard-ring junction current.
More-over, the forward current of the poly-emitter diode was much smaller as compared to the Schottky barrier diode, as shown in Fig. 3. Hence, we can conclude that the forward current of the Schottky diode with p+
-polysilicon-diused guard-ring dominated by the Schottky barrier current.
3.2. Capacitance±voltage characteristics
The capacitance±voltage characteristics were measured at 100 kHz. Fig. 4 shows the data of Cÿ2
versus the reverse bias. The dopant concentration of the Schottky diode was also measured using the slope of the Cÿ2±V curve. From Fig. 4, the obtained value
of the eective carrier concentration is very consistent with the initial-growth doping level (34.8 1014cmÿ3)
of epi-wafer. Furthermore, it was due to the eect of the barrier height lowering of the J±V measurement technique; as expected, the eective barrier height of diode D (H2SO4 sample) was lightly smaller than the
value of 0.76 eV, as determined by C±V measurement [10]. The measured C±V barrier heights of the other diodes were also given in Table 1. For example, fbnof
diode A was about 0.74 eV; fbnof diode B was about
0.75 and 0.74 eV for the H2SO4 and HF samples,
re-spectively. Likewise, fbn of diode D was about 0.76
eV. It can be observed that the eective barrier height and the reverse-breakdown voltage of the H2SO4
sample were lightly greater than those of the HF sample. This result was consistent with the previous mentioned, where the reverse-breakdown voltage of diode is depended upon the barrier height of the metal±semiconductor (MS) contact.
3.3. Reverse J±V characteristics
The reverse characteristics of p+-polysilicon
Schottky barrier diodes are given in Fig. 5, which shows the 9508C, 30 min annealing condition. Fig. 5 (a) is the as-deposition polysilicon (ADP) diode and Fig. 5 (b) is the stacked-amorphous silicon (SAS) diode [6]. As can be seen in Fig. 5 (a), the breakdown voltage of the conventional p+-n diused guard-ring
diode is about 25 V and its reverse leakage current density about 0.33 mA/cm2. A low breakdown voltage
and a great reverse leakage current may be due to the premature edge breakdown of diode. Next, for the p+
-polysilicon diused-guard-ring diode without the over-lap metal (diode B), the breakdown voltages are 68 and 73 V for the HF and the H2SO4sample,
respect-ively. For the p+-polysilicon diused-guard-ring diode
with the polysilicon-overlap structure (diode D), the breakdown voltage of HF sample is about 148 V and its reverse leakage current density about 19.5 mA/cm2.
Likewise, the H2SO4 sample had a higher breakdown
voltage (3149 V) and a lower leakage current (38.4 mA/cm2) as compared to the HF sample. This
re-duction can be attributed to the presence of the native oxide layer, which blocks the minority-carrier injection into the polysilicon layer. As a result, the electron± hole pair recombination at grain boundaries of the polysilicon ®lm and in the interior of the polysilicon grains would be reduced. Bravman [11] has already shown that the interfacial native oxide layer could avoid the breaking of the polysilicon ®lm. This exten-sion of epi-regrowth at the poly-Si/mono-Si interface increases with the increase of annealing temperature and drive-in time [12±15]. This result might cause the interface of poly-Si/mono-Si to look very rough [14]. Then the native oxide of the H2SO4sample could
inhi-bit more epi-regrowth of this interface as compared to the HF-dip sample [11]. Their breakdown voltage and leakage-current density are also given in Table 1. From this result, we can conclude that the polysilicon diused-guard-ring can improve the electrical ®eld
Table 1
The electrical characteristics of diode A±D. It shows the reverse current density, reverse breakdown voltage, barrier height and ide-ality factor Jr (mA/cm2) at V r 50V Jr (mA/cm2) at V r 65V Vbd
(V) J(mA/cmbd 2) f(eV) by J±Vbn f(eV) by C±Vbn n
Diode A (C.S.) ± ± 25 332 0.734 0.740 1.100 Diode B (H2SO4) 4.68 6.61 73 24.2 0.742 0.746 1.080 Diode B (HF) 2.59 3.18 68 6.04 0.741 0.744 1.090 Diode C (H2SO4) 1.85 10ÿ3 2.27 10ÿ3 159 3.72 10ÿ3 ± ± 1.002 Diode C (HF) 2.25 10ÿ3 2.67 10ÿ3 156 4.47 10ÿ3 ± ± 1.004 Diode D (H2SO4) 1.96 2.32 149 8.4 0.757 0.760 1.070 Diode D (HF) 2.59 3.18 148 19.5 0.753 0.758 1.070
limitation, the overlap-polysilicon sample could sup-press the edge electrical ®eld and the breakdown vol-tage of the H2SO4 sample is slightly greater than that
of the HF sample.
From Table 1, for the H2SO4 sample, the
break-down voltage of diode A was 25, of diode B 73.5 and of diode D about 149 V. The reverse leakage current could increase with the decrease of contact area of overlap metal or polysilicon layer. Thus, this could
cause the diode to exhibit a low breakdown voltage. Therefore, the overlap-metal and/or -polysilicon sample exhibited a higher reverse breakdown voltage and a lower leakage current as compared to non-over-lap-metal or -polysilicon sample, respectively.
Comparing the breakdown characteristic of the ADP and SAS structures, one can ®nd that the break-down voltage of the ADP structure is slightly greater than that of the SAS structure and the leakage current of the former is smaller as compared to the latter. For the 9508C annealing condition of furnace, the reverse breakdown voltage and boron-diusion depth in terms of the drive-in time are given in Fig. 6. For example, for 30 min drive-in, the boron-diused depth of the ADP structure is about 0.25 mm and for SAS about 0.21 mm [6]. On the other hand, the deeper the diused length of the p+/n guard-ring junction, the higher the
reverse breakdown voltage. From the previous discus-sion of breakdown voltage of diode, it could be observed that the breakdown phenomenon occurred at the edge of the guard-ring of heavily doped polysilicon ®lm. Thus, another method of improving the break-down voltage was the increase of breakbreak-down voltage of the p+/n guard-ring junction. In addition, the
reverse leakage current density of the p+-polysilicon
diused-guard-ring Schottky diode was much smaller than that of the conventional Schottky sample. This result can be explained by the fact that the polysilicon-guard-ring structure could trap the mobile ions and the shallow-diused junction could smooth the edge-electrical ®eld. This eect made the guard-ring Schottky diode with the overlap-polysilicon structure
Fig. 6. Reverse J±V characteristics and boron diusion length in terms of the drive-in time of furnace. It shows the ADP and SAS sample, respectively.
Fig. 5. (a) Reverse current density versus applied voltage of fabricated ADP-guard-ring diodes and conventional diode. The furnace temperature was 9508C and drive-in time 30 min. (b) Reverse current density versus applied voltage of fabri-cated SAS-guard-ring diodes. The furnace temperature was 9508C and drive-in time 30 min.
to inhibit the reverse leakage current and give a high breakdown voltage.
Fig. 7 shows the reverse characteristic of the poly-emitter diode with the ADP structure. The breakdown voltage of the H2SO4 sample is about 159 V and its
reverse leakage current density about 3.72 nA/cm2;
similarly, the breakdown voltage of the HF sample is about 156 V and its leakage current density about 4.47 nA/m2. This phenomenon can be similar to the result
of a heavily doped polysilicon diode with polysilicon overlap that the breakdown voltage of H2SO4 sample
was lightly greater as compared to the HF sample. The punch-through-breakdown voltage of 1021.5 O cm, 20 2 3 mm epitaxial-layer wafer was about 174 V. No matter what the poly-emitter diode or p+polysilicon
diused-guard-ring Schottky diode with polysilicon overlap, their breakdown voltages were very high and approached the ideal punch-through voltage of this raw epi-wafer substrate.
3.4. The statistical plots of breakdown voltages
Fig. 8 shows the statistical plots of the breakdown voltages of the ADP structure for three types of diode for each of the 30 diodes. For these data, the break-down conditions were: furnace temperature 9508C and drive-in time 30 min. The conventional p+±n
guard-ring diode had a low breakdown voltage occupation; for example, the occupation of 25 V was about 60%. The guard-ring diode of heavily doped polysilicon had
a higher breakdown voltage statistic as compared to the conventional diode. It is clearly shown that the polysilicon-overlap sample tended to have a higher breakdown voltage occupation as compared to a non-polysilicon-overlap sample. In Fig. 8(a), for diode D, the occupations of 150 V: the H2SO4 sample was 50%
and the HF sample 20%; likewise, the occupation of 140 V: the H2SO4sample was 15% and the HF sample
35%. This result can help us to explain that diode D
Fig. 8. (a) Statistical plots of reverse breakdown voltages for fabricated diodes with the polysilicon-overlap structure. (b) Statistical plots of reverse breakdown voltages for fabricated diodes without polysilicon/overlap and conventional struc-tures.
Fig. 7. Reverse current density versus applied voltage of poly-emitter diode. The polysilicon ®lm was ADP structure. The furnace temperature was 9508C and drive-in time 30 min. It shows the H2SO4and HF samples, respectively.
was a higher barrier-height device as compared to the other diodes, as is previously mentioned. On the other hand, it can endure a strong electric ®eld between the metal contact and the silicon substrate, which leads to a tough breakdown voltage. Therefore, the guard-ring structure of heavily doped polysilicon could prevent premature electrical breakdown around the periphery of the diode contact. The breakdown voltage phenom-enon of the non-polysilicon-overlap structure and the conventional p+±n guard-ring diode is shown in Fig.
8(b): the 60 V occupation of the HF sample is 65% and the 70 V occupation of the H2SO4sample 40%. It
is further observed that diode D had a greater break-down voltage occupation as compared to diodes B and A. On the other hand, by joint applying the polysili-con-overlap, metal-overlap and diused-guard-ring structure, a high performance Schottky diode with a high breakdown voltage and low leakage current can be achieved. Therefore, the fabricated device was found to demonstrate the above future.
4. Conclusions
This work has shown nearly ideal forward and higher reverse breakdown voltage characteristics of the proposed diode which obtains to use the heavily doped-polysilicon-guard-ring structure. A heavily doped polysilicon ®lm can be applied to a Schottky barrier diode to form a polysilicon emitter guard-ring junction. Due to the high breakdown voltage of the polysilicon emitter diode of the guard-ring, a high breakdown voltage Schottky diode can be obtained. By comparing J±V characteristics of the conventional sample and the proposed sample, the following general behaviors can be observed:
1. For most of the practical Schottky diode, the domi-nant reverse current component is the edge-leakage current around the periphery of the metal plate. Therefore, leakage current decreases with the increase of the guard-ring area of heavily doped-polysilicon ®lm. For the doped-polysilicon-overlap sample, the shallow-diused depth can smooth the edge elec-tric ®eld, and this result makes the breakdown vol-tage to be high and leakage current to be small. 2. The polysilicon-overlap diode D has a near-ideal
forward J±V characteristic and a low leakage cur-rent at moderate reverse bias, in comparison, diode B exhibits a high reverse current density and a low breakdown voltage. Thus, the polysilicon-overlap structure can suppress the electrode sharp edge eect of diode.
3. The presence of native oxide layer at the
polysili-con/mono-silicon interface can avoid the polysilicon ®lm to break and block minority carrier injection. Thus, this eect makes the trapped density to decrease and improves the breakdown voltage of diode.
4. For the 9508C temperature annealing, the longer the drive-in time of furnace, the deeper the boron diu-sion length, which causes the breakdown voltage of the diode to be high
.
Acknowledgements
The authors would like to acknowledge the technical support of the Vanguard International Semiconductor Corporation and ®nancial support of National Science Council of R.O.C. through contract NSC-83-0404-E009-017 for this research.
References
[1] Henisch HK. Rectifying semiconductor contacts. Oxford Clarendon press, 1957.
[2] Liou BW, Lee CL, Lei TF. IEEE Electron Lett 1995;31:1950.
[3] Chang CY, Sze SM. Solid-State Electron 1970;13:727. [4] Saltich JL, Clark LE. Solid-State Electron 1970;13:857. [5] Andrews JM, Lepselter MP. Solid-State Electron
1970;13:1011.
[6] Wu SL, Lee CL, Lei TF, Chang HC. IEEE Trans Electron Devices 1993;40:1797.
[7] Neugroschel A, Arienzo M, Komem Y, Isaac RD. IEEE Trans Electron Devices 1985;32:807.
[8] Chu TL. J Cryst Growth 1977;39:45.
[9] Ashburn P, Soerowirdjo B. IEEE Trans Electron Devices 1984;31:853.
[10] Yang ES. Microelectric devices. McGraw international editions, 188, 1988.
[11] Bravman JC, Patton GL, Plummer JD. J Appl Phys 1985;57:2779.
[12] Jorgensen N, Barry JC, Booker GR., Ashburn P. Wolstenholme GR, Wilson MC and Hunt PC. In: Microscopy of semiconductor materials Conf. Dig. (Oxford, UK), 1985.
[13] Patton GL, Bravman JC, Plummer JD. IEEE Trans Electron Devices 1986;33:1754.
[14] Wolstenholme GR, Jorgensen N, Ashburn P, Booker GR. J Appl Phys 1987;61:225.
[15] Liou BW, Lee CL, Lei TF, Wu YH. Jpn J Appl Phys Part 1 1997;36:3389.
[16] Chu TL, Chu SS, Vanderleen GA, Lin CJ, Boyd JR. Solid-State Electron 1978;21:781.
