PERGAMON                                  Microelectronics Reliability 39 (1999) 133±137
Research note
       Investigation of MOSFET operation in bipolar mode
  V.S. Pershenkov *, V.V. Belyakov, S.V. Cherepko, I.N. Shvetzov-Shilovsky,
                               V.V. Abramov
         Microelectronics Department, Moscow Engineering Physics Institute, Kashirskoe shosse 31, Moscow 115409, Russia
                                    Received 16 July 1997; received in revised form 7 July 1998
Abstract
  The work is concerned with the properties of conventional MOSFET in bipolar mode of operation. It is shown
that the base current can provide useful information about interface trap density at the Si±SiO2 interface. The new
device characteristics are found promising for use in low-voltage low-power logic circuits. # 1999 Elsevier
Science Ltd. All rights reserved.
1. Introduction                                                      not be achieved in a conventional transistor [3]. With a
                                                                     MOSFET operating in a bipolar mode the appropriate
   Years ago it was noticed, that the simple MOSFET                  gate±substrate voltage provides the control of eective
is quite similar to a lateral bipolar junction transistor            base doping and brining the base properties close to
(BJT) with a gate upon the base region. Source, drain                the intrinsic Si. It results in the increase of source
and substrate terminals of the MOSFET are the emit-                  region emitting eciency and, at the same time, in the
ter, collector and base of the lateral bipolar transistor,           decrease of recombination due to the very little
respectively (Fig. 1). In an integrated circuit a                    amount of the major carriers in the base. The recombi-
MOSFET can be used in bipolar mode only if the sub-                  nation is also decreased by pushing diusing carriers
strates of dierent devices are separated. The sub-                  to the surface. The same eect also decreases noise
strate-source junction must be forward biased to                     level in the transistor. The practical application of
provide bipolar transistor operation, which is dierent              such high gain bipolar device was suggested in ref. [3].
from the MOSFET mode where this bias should be                       It was used in a precise current mirror and low-noise
zero or reverse.                                                     low-frequency ampli®ers, but the main goal of ref. [3]
   The fact that the dierence between these two                     was to use the advantages of combining bipolar and
devices resides only in an operation mode has been                   ®eld-eect devices on the same chip in the convention-
used in scienti®c researches. Using the bipolar mode of              al CMOS technology.
MOSFET provides the reliable model for some speci®c                     The progress in BiCMOS technology discarded this
eects, including the breakdown of short-channel                     main advantage of the device under consideration.
transistor [1]. Vice-versa, MOSFET can serve as a                    However, it can still be very promising from several
model device for the investigation of low-dose-rate                  points of view. First, the device can be promising for
radiation response of bipolar transistor [2].                        the applications in the low voltage logical circuits, due
   With respect to conventional lateral BJT the                      to its ability to operate at extremely low collector
MOSFET used in bipolar mode has an extra terminal,                   (drain) voltages. Second, its gain can be additionally
the gate. It can be used for the additional control of               increased by the combining of bipolar and subthres-
the device, providing the operation modes which can-                 hold MOSFET mode of operation. Third, it can be
                                                                     used for physical investigations of MOS structures. It
                                                                     means that the physical parameters of oxide-semicon-
 * Corresponding author. Tel.: 007 095 324 0184; Fax: 007            ductor interface, e.g. interface traps density, can be
095 321 2111; E-mail: pershenkov@d408.micro.mephi.ru.                extracted by appropriate processing of device current±
0026-2714/99/$ - see front matter # 1999 Elsevier Science Ltd. All rights reserved.
PII: S 0 0 2 6 - 2 7 1 4 ( 9 8 ) 0 0 1 5 7 - 7
134                         V.S. Pershenkov et al. / Microelectronics Reliability 39 (1999) 133±137
                                                                      Output characteristics of MOSFET in bipolar mode
                                                                   are aected by the speci®c features of n + -i-n + struc-
                                                                   ture. The contact potential of n + -i junction is 0.3±
                                                                   0.35 V lower as compared to the conventional n + -p
                                                                   junctions. That is why collector saturation voltage of
                                                                   MOSFET in the bipolar mode is low. This saturation
                                                                   voltage is less than 50 mV, that is 2±4 times lower
                                                                   than the one for BJT.
                                                                      The output I±V curve can be derived from the
                                                                   analysis of output characteristic of conventional
                                                                   MOSFET in the subthreshold mode. The equation for
                                                                   drain current of MOSFET leads to:
                                                                              V                                                                                    
                                                                              ÿ jbe
                                                                   Id 0 1 ÿ e     t   ;                                  1
                                                                   where Id is drain current; Vds is drain±source voltage;
                                                                   jt is thermal potential.
Fig. 1. The lateral bipolar npn transistor based on the              A logical circuit based on the MOSFET in bipolar
nMOSFET.                                                           mode could be able to handle potentials from about
                                                                   50 mV (logical `0') to about 0.4 V (logical `1'). This
voltage characteristics. It can give additional important          makes it possible to design IC powered with as low as
information compared with the values obtained by                   0.5 V supply. The presence of additional gate terminal
other approaches. The current work deals with these                provides additional ¯exibility in using the device.
three items.                                                         The bipolar transistor thus formed has certain draw-
                                                                   back. The dynamic characteristics are expected to be
                                                                   poor due to the low currents ¯ow during switching.
2. Analysis                                                        However, the device under investigation was not opti-
                                                                   mised for operation in bipolar mode. A special design
   The main feature of the device under consideration              may lead to improvement of dynamic characteristics.
resides in the fact that its collector current consists of
the electrons injected by the emitter under the gate. It
can be assumed to constitute only a small part of the              3. Circuitry
emitter current because the lateral surface of the emitter
at the collector side is only a minor part of the total               In recent works some circuits using MOS transistor
emitter surface. However, by the proper gate bias the              in bipolar mode were presented [3]. One of the very
region under the gate can be depleted or weakly                    interesting applications is the use of such a structure in
inverted. For injection conditions it corresponds to the           digital circuitry. The logical invertor is presented in
emitter±base junction with an extremely low doped base             Fig. 2. A couple of complementary MOS transistors is
region. In conventional BJT the decrease of base doping
is limited by the increase of base resistance. In bipolar
mode MOSFET only a very shallow layer of the base is
i-type. Due to the fact that the injection in the i-region
is several orders higher than in the MOSFETs substrate,
one can obtain an operable device.
   With the injection level increase, the i-region is
modulated and the gain coecient of the device
decreases. The injection from the bottom part of emit-
ter±base junction increases with the junction forward
bias that also decreases current gain. The reasonable
limit of emitter±base bias is thus found to be about
0.5 V. The devices used in this study were conventional
n- and pMOSFETs.
   The high current gain expected for relatively wide
base region can be explained by the fact that the hole
concentration in the i-region is low and, therefore,               Fig. 2. Logical invertor based on substrate controlled n- and p-
recombination is negligible.                                       MOS transistors and biasing element (surrounded by dashed line).
                              V.S. Pershenkov et al. / Microelectronics Reliability 39 (1999) 133±137                        135
                                                                     biases were chosen to provide Vsup/2 threshold voltage
                                                                     of invertor and initial currents Ibias n and Ibias p of ap-
                                                                     proximately 0.1 mA (Vg = 0.4 V for nMOSFET and
                                                                     Vg = ÿ0.7 V for pMOSFET). The transfer regions of
                                                                     characteristics are quite sharp and it is evident that
                                                                     such a structure can be used as a base of logical el-
                                                                     ements. The supply current for such a structure in sta-
                                                                     tic condition is not higher than a maximum of Ibias n
                                                                     and Ibias p. However, the output current can be much
                                                                     higher during switching and is demonstrated in Fig. 4.
                                                                     The varying of substrate (base-emitter) bias in the (0±
                                                                     0.5) V range leads to huge variations (from 100 to
                                                                     1000 times) in output current. Drain current saturates
                                                                     at a very low drain voltage which allows one to use
                                                                     supply voltage as low as to (0.1±0.2) V. An essential
                                                                     feature of substrate controlled invertor is a low sensi-
                                                                     tivity to the transistor's parameters variations: 0.2 V
Fig. 3. Transfer characteristics of the invertor for dierent        threshold voltage shift leads to only about 50 mV shift
supply voltages.
                                                                     of transfer characteristics.
connected serially between `Ground' and ` + Vsup'
terminals. An input signal Vin, is applied to the sub-               4. Interface trap characterisation
strates of transistors, and the output signal Vout, is
measured at the common point of drains as in the                        The base current of a lateral bipolar transistor can
CMOS invertor. The gates of the invertor's transistors               be at least partially attributed to the recombination via
are biased to provide initial conductivity which corre-              fast interface traps. The recombination is most ecient
sponds to Ibias n and Ibias p values. The biasing element            through the traps with energy levels close to the middle
(marked by a dashed line in Fig. 2) consists of two                  of the Si bandgap. Therefore, the surface recombina-
halves of `current mirrors', the second half being the               tion component of the base current is most sensitive to
transistors of an invertor. One biasing element can                  the interface trap density in the near-midgap region.
provide gate bias for a number of invertors.                            The base current of lateral bipolar transistor
   Experimental transfer characteristics of the invertor             (nMOSFET in bipolar mode) vs gate voltage is pre-
for dierent supply voltages are shown in Fig. 3. Gate               sented in Fig. 5 for irradiated and non irradiated
Fig. 4. Output I±V characteristics of nMOS (a) and pMOS (b) transistors for dierent substrate biases. Gate voltages are ®tted to
provide 0.1 mA drain current at the zero substrate bias.
136                               V.S. Pershenkov et al. / Microelectronics Reliability 39 (1999) 133±137
                                                                         the surface, ni is intrinsic electron concentration in Si
                                                                         and vth is thermal velocity. According to Eq. (2) the
                                                                         most ecient recombination occurs when the electron
                                                                         and hole concentrations are equal. In terms of electron
                                                                         and hole quasi-Fermi levels it means, that Si midgap
                                                                         energy Ei at the surface falls in the middle between Efn
                                                                         and Efp. In Fig. 5 this occurs when gate voltage is
                                                                         equal to Vg = ÿ0.5 V. Since the dierence Efn ÿEi is
                                                                         known (Ube/2), the value of peak base current can be
                                                                         obtained from (2):
                                                                                             Efn ÿEi      Ei ÿEfp
                                                                                         ni e  qjt     ni e   qjt    ÿ n2i
                                                                         Js  svth Nit    Efn ÿEi             Ei ÿEfp
                                                                                      ni e  qjt   ni e        qjt       2ni
                                                                                          Vjbe      
                                                                             svth Nit ni e t ÿ 1
                                                                           
                                                                                 2          Vbe
                                                                                           e 2jt  1                                                                                                   
Fig. 5. Dependencies of nMOS transistor substrate current at                 svth Nit ni V2jbe
                                                                                          e t ÿ1 :                             3
0.4 V substrate bias vs. gate voltage before and after ir-                       2
radiation.
                                                                         The surface recombination current Is then could be
devices. The base±emitter junction bias was kept con-                    obtained as Js times the gate area S: Is = JsS.
stant during the measurement. As it can be seen from                       Provided the peak base current is known, the inter-
the ®gure the base currents have the peaks. These                        face trap density (Nit) could be expressed from (3). For
peaks indicate maximum surface recombination due to                      example, from the data shown in Fig. 5 one could esti-
the presence of both electrons and holes in the chan-                    mate:
nel. Radiation induced shift of peak gate voltage (DVg)                                            ÿ1
is equal to MOSFET threshold voltage shift due to                                     svth ni V2jbe
                                                                         Nit  Is S          e t ÿ1      11:1  1012 cmÿ2 ;
oxide trapped charge DVot. The magnitude of the peak                                    2
is correlated with the number of fast interface traps                    where s = 2.510 ÿ 15cm2 was taken from [7].
introduced during irradiation.
   The surface recombination current is reverse pro-                     5. Summary and conclusion
portional to the carriers lifetime. This lifetime is corre-
lated with the trap recharge time constant. It means                        Characteristics of the MOSFET in the bipolar mode
that only the traps with small recharge time constant                    were investigated. Experiments show that the lateral
will contribute signi®cantly to the surface recombina-                   bipolar transistor can be used as a low voltage low
tion current. Therefore, the base current of MOSFET                      power logical element. Certain drawbacks also exist
in bipolar mode can be used to extract information                       due to the speci®cs of the MOSFET design which is
about fast interface traps with energies close to the Si                 not optimised for bipolar operation.
midgap. This is dierent from the interface trap den-                       The expression for surface recombination current is
sity extracted via subthreshold technique [4] because                    derived. This current dominates the base current of
the subthreshold technique gives the density of traps                    MOSFET in bipolar mode. The peak of base current
recharged during measurement time, thus resulting in                     is correlated with the number of midgap fast interface
the sum of the border traps [5] and fast interface traps.                traps at the Si±SiO2 interface, these traps being dier-
This is also dierent from the interface trap density                    ent from that measured by subthreshold and charge-
extracted via charge pumping technique [6] because the                   pumping techniques.
charge pumping is sensitive to the fast interface traps                     The logical inverter is proposed utilising substrate
with energy levels over the whole Si bandgap.                            controlled complementary MOSFETs. The invertor is
   The peak value of the base current could be easily                    experimentally shown to be operable within (0.2±0.4)
expressed in terms of interface traps density (Nit), car-                V supply voltage.
riers capture cross-section (s) and base-emitter bias
Vbe. The surface recombination current density can be
estimated in terms of conventional SRH theory:                           References
                  ns ps ÿn2i
Js  svth Nit                 ;                                2         [1] Hsu F-C, Muller RS, Hu C. A simpli®ed model of short-
                ns  ps  2ni
                                                                              channel MOSFET characteristics in the breakdown
where ns and ps are electron and hole concentrations at                       mode. IEEE Trans on Electron Dev 1983;30:571±6.
                             V.S. Pershenkov et al. / Microelectronics Reliability 39 (1999) 133±137                        137
[2] Belyakov VV, Pershenkov VS, Shalnov AV, Shvetzov-                [5] Fleetwood DM, Winokur PS, Reber RA, Jr,
    Shilovsky IN. Use of MOS structures for the investi-                 Meisenheimer TL, Schwank JR, Shaneyfelt MR, Riewe
    gation of low-dose-rate eects in bipolar transistors.               LS. Eect of oxide traps, interface traps, and `border
    IEEE Trans Nucl Sci 1995;42:1660±6.                                  traps' on metal-oxide-semiconductor devices. J Appl
                                                                         Phys 1993;73(10):5058±74.
[3] Vittoz EA. MOS transistors operated in the lateral bipo-
                                                                     [6] Groeseneken G, Maes HE, Beltran N, DeKeersmaecker
    lar mode and their application in CMOS technology.
                                                                         R. A reliable approach to charge-pumping measurements
    IEEE Journal of Solid-State Circuits 1983;SC-18(3):273±
                                                                         in MOS transistors. IEEE Trans Electron Dev
    9.                                                                   1984;31:42.
[4] Benedetto JM, Boesch HE. MOSFET and MOS capaci-                  [7] Kavalieros JT, Sah CT. Separation of interface and non-
    tor responses to ionizing radiation. IEEE Trans Nucl Sci             uniform oxide traps by the DC current±voltage method.
    1984;31:1461±6.                                                      IEEE Trans Electron Dev 1996;43:137±41.