1076 IEEE TRANSACTIONS ON ELECTRON DEVICES. VOL. 36, NO. 6. JUNE IY89

A Lateral Magnetotransistor Structure with a Linear Response to the Magnetic Field

LJUBISA RISTIC, MEMBER, IEEE, TOM SMY, A N D HENRY P. BALTES, MEMBER, IEEE

Abstract-An experimental study and analytical model of a novel magnetotransistor is presented. This new device displays some very promising features. A linear response to the magnetic field is experi- mentally demonstrated and very high sensitivities are measured. Pre- vious comparable magnetotransistors have reported sensitivities of the order of 150 percent/T; the device presented in this paper shows sen- sitivities of up to 3000 percent/T. A theoretical explanation of the very high sensitivity is proposed, involving carrier deflection as the domi- nant operating principle.

I. INTRODUCTION NUMBER of tempting applications exist for a low- A cost batch-produced magnetic sensor. These appli-

cations include: contactless switching, displacement de- tection, and current detection as well as direct applica- tions such as measurement of the earth’s field and magnetic tape and disc reading [ 13. In particular, a solid- state magnetic field sensor (MFS) manufactured in a stan- dard semiconductor process and integrated into support- ing circuitry is suitable for these applications. A variety of magnetic sensitive semiconductor devices have been investigated; these include Hall devices, magnetodiodes, split-drain MOSFET’s, and magnetotransistors (MT) (see [ 11 and references therein).

Magnetotransistors, i.e., bipolar transistors optimized such that they respond to a magnetic field, can be divided into two major groups. Those MT’s that depend on the vertically flowing carriers for their magnetic operation are known as vertical magnetotransistors. Conversely, if the lateral flow of injected carriers is responsible for the mag- netic response, the devices are known as lateral mag- netotransistors (LMT).

The fundamental effect on which all solid-state MFS‘s are based is the Lorentz force. The operation of the Lor- entz force in an MT manifests itself in three basic forms, current deflection, emitter injection modulation, and mag- netoconcentration. Current deflection is the deflection of carriers in the neutral base or collector region of the de-

vice [2]-[4]. The emitter modulation model proposes that a Hall voltage is produced along the emitter-base junc- tion, which in turn causes a nonsymmetric injection of minority carriers [5]. Magnetoconcentration is the asym- metrical concentration of carriers within the device, caus- ing local modulation of the conductivity [6]. Of these three effects the last two involve a nonlinear magnetic response. Previously reported magnetotransistors that display a lin- ear response to the magnetic field have had relative sen- sitivities varying from 1 to 150 percent/T [7]. An MT displaying nonlinear behavior has been reported with a sensitivity of 400 percent /T [8]. However, this sensitiv- ity was reported for large fields of 1 T and very small collector currents of 20 nA.

In this paper we deal with a new MT that we call the Suppressed Sidewall Injection Magneto-Transistor (SSIMT). The SSIMT has a number of desirable charac- teristics: very high sensitivity (both relative and abso- lute), a linear response to the magnetic field, the possi- bility to work as a magnetic switch, and fabrication in standard CMOS technology. The SSIMT has a range of relative sensitivities varying from 50 to 3000 percent/T, determined by the choice of bias condition. The magni- tude of the collector currents at maximum sensitivity is of the order of 400 pA each. The large range of sensitivity and the structure of the device allow one to choose a bias point such that one of the collector currents can be driven to almost zero at any field between 30 mT and 1 T. Brief reports on some of the characteristics were presented re- cently [9]-[ 1 11. This paper provides a full analysis and a theoretical explanation for the high sensitivity.

In the next section we describe the SSIMT structure and provide a qualitative description of the basic principles of operation. In Section I11 an analytical model is developed as a means to understand the high sensitivity of the SSIMT. Experimental results are then presented in Sec- tion IV, followed by a conclusion in Section V.

Manuscript received June 3, 1988; revised February 7, 1989. This work was supported by the Natural Science and Engineering Research Council of Canada. The review of this paper was arranged by Associate Editor S. D. Senturia.

L. RistiC and T. Smy are with the Department of Electrical Engineering and the Alberta Microelectronic Center. The University of Alberta, Ed- monton, Alberta, Canada T6G 2G7.

H. P. Baltes is with the Physical Electronics Laboratory, Institute of Quantum Electronics, Swiss Federal Institute of Technology, ETH- Hoenggerberg HPT, CH-8093 Zurich, Switzerland.

11. OPERATING PRINCIPLES OF THE SSIMT STRUCTURE

The novel feature of the SSIMT is the two pf stripes placed along the edges of the emitter, parallel to the col- lectors; this is shown in Fig. 1. The stripes play a twofold role. First, they suppress carrier injection from the emitter in the lateral direction toward the ~f the p+ stripes are biased with a voltage V,., and V,. is less then or equal to the potential of the emitter V,, this effect is ob- IEEE Log Number 8927961.

0018-9383/89/0600-1076$01 .OO O 1989 IEEE

RISTIC er ai . : LATERAL MAGNETOTRANSISTOR STRUCTURE

N-SUBSTRATE

(a)

I077

N-SUBSTRATE

ib) Fig. 1. The SSIMT. (a) Cross section of the structure: components of injected electrons 1, 2, and 3 contribute to I,,, I C z . and

Is. respectively. (b' Top view. L , = 100 pm. W , = 50 pm, and W, = 8 pm.

vious. The suppression of laterally flowing electrons is a consequence of the reverse biasing of the junction be- tween the p+ stripe and the emitter; in fact, a small strip of the bottom of the emitter next to each stripe will also be reverse biased. The width of this reverse-biased strip will increase with increasing negative V,. The breakdown voltage of the n+ emitter p+ stripe junction limits the magnitude of V, . Carrier injection from the emitter is also confined to the vertical direction due to the formation of a potential hill (for electrons) around the p'. stripes. The second effect of biasing the pf stripes is the creation of a lateral electric field E, in the neutral base region. This field is due to the difference in potential between the base contacts B , and BZ. When the device is operated normally with V, 5 V, < V,, , this field will be oriented in such a way as to sweep the injected electrons laterally toward the collectors (Fig. l(a)).

The device structure was fabricated in a standard 4-pm CMOS process. The base region was formed by a p-well. The emitter and both collectors were realized using the standard doping procedure for the source and drain of a n-channel MOS transistor. The p+ stripes at the side of the emitter and the base contacts B , were formed using the standard doping procedure for a p-channel MOS Tran- sistor. The p-well diffusion depth is 10 pm, and the col-

lector and emitter diffusion depths are 1.2 pm. The p-well doping is 10l6 ~ m - ~ . The characteristic mask dimensions are shown in Fig. 1.

The biasing circuit of the SSIMT is shown in Fig. 2 . The values of Z b , V,, V,, V , , , and V,, determine the op- erating point of the device. In normal operation slightly different values for V,, and V,, are used for elimination of the collector current offset, Zc2 - I C , , and we define V, = (V,, + Vr2)/2 for convenience.

The device operation is a follows. If the emitter-base junction is forward biased, and electrons are injected from the emitter into the base region, the net effect of the p+ stripes, when biased with a small negative voltage, is to force the minority electron current down toward the sub- strate. It is then split into three distinct current flows. Two of these flows, I,-, and IC-, are collected by the collectors and the third, Is, by the substrate. The application of a larger negative potential to the p+ stripes amplifies the effect of the p+ stripes, pushing the minority current fur- ther into the device and then establishing a lateral flow of current out to the collectors. In the absence of a magnetic field, the collector currents Zclo and are equal because of the device's symmetry, and the device is balanced.

The application of a magnetic field B = B z , parallel to both the chip surface and the collectors, will produce an

1078 IEEE TRANSACTIONS ON ELECTRON DEVICES. VOL. 36. NO. 6. J U N E 1989

R C f Fig. 2 . Biasing circuit for the SSIMT. I,, denotes the base current, I,, and

I,? the two collector currents, V,y the substrate potential, V , the supply voltage to the collectors, and V,, and V,? the two stripe potentials.

imbalance in the two collector currents IC , and I C , due to the following “double-deflection’’ effect. The Lorentz force acts on all three current components I s , Zcl, and IC , . The I , component is deflected in the y direction, increas- ing I C , and decreasing Zcl. Moreover, the current com- ponents Icl and I,, are deflected in the “+x” and “-x” directions respectively, causing a further increase of IC , and a decrease of Zcl. These two deflections will collab- orate to cause one collector current to increase at the ex- pense of the substrate current and the other to decrease with a corresponding gain in the substrate current. The net effect on the substrate current should be zero.

The SSIMT is somewhat unique as the injected elec- trons flow first vertically down away from the emitter and then laterally out toward the collectors, (Fig. l(a)). How- ever, as most of the salient features of the device are due to the lateral flow of the electrons out to the collectors, we believe it should be classified as an LMT.

111. THEORY A theoretical model of the SSIMT can be developed

based on the assumption that a magnetic flux density B , causes a linear displacement of the minority carriers in the neutral base region. This deflection of the minority car- riers is due to the action of the Lorentz force. To obtain the magnetic response of the SSIMT we shall only con- sider the action of the magnetic field on one half of the device. The flow of the minority carriers within half of the device is shown in a cross section of right half of the structure; see Fig. 3. We analyze the part of the electron flow that contributes to the collector current; see Fig. 3(b). For simplicity we assume that this flow of electrons from the emitter to the collector consists of a current tube of constant cross-sectional area with a uniform current den- sity and that only the carriers that are inside this region will contribute to the collector current. We model this current flow as first a vertical flow and then a lateral flow; see Fig. 3(c). The vertical flow is a consequence of the negative potential applied to the pf stripes and the lateral flow is due to the strong lateral electric field in the neutral base region.

3(d). The corresponding current densities within these two boxes are J,, and Jny for the vertical and lateral directions, respectively. It remains to determine the boundaries of these boxes. The top and bottom edges of the horizontal box are defined by two planes, x = XI and x = X 2 . The first plane, x = XI, is defined so as to provide a means to analyze the effect of the negative potential V,. applied to the pf stripes. The negative potential applied to the stripes will block the injection of electrons from the emitter into the pf stripes and also the injection across a small region of the bottom of the emitter next to each stripe. This will have the effect of preventing movement of the injected electrons laterally and forcing the minority current to flow down into the device. The plane x = XI is positioned such that a majority of the laterally flowing electron current will flow below this plane. We also assume that OA = XI (penetration of the negative potential is equal in all direc- tions). The plane x = X 2 is used to analyze the effect of the substrate potential V, on the electron flow. The elec- tron current above this plane is presumed to be collected by the collector, conversely electron current below this plane is collected by the substrate. This definition is sim- ilar to the one presented in [7]. According to this model all of the electron current flow contributing to the collec- tor current will flow in the region

x, I x I x,. (1) The distance between these two planes XI and X , is de- fined as Weff = X 2 - XI and can be thought of as the effective width of the stream of laterally flowing elec- trons. We assume that all the electrons flowing in the re- gion defined by (1) that reach the plane y = WB are col- lected by the collector and form the collector current.

The length of the vertical box L1 can be expressed in terms of the two planes defined above as

With the assumption given above, L, can be expressed as

We, L, = w, + XI + - 2

or

(3)

L2 WB + Ll. (4b)

A . Analytical Model The total current collected by the right collector at zero

magnetic field Zczo can be expressed in terms of the cur- rent density of laterally flowing electrons .Inl.. If we pre- sume that a uniform current distribution exists in the re- gion between the two planes XI and x,, then we have

1 ~ 2 0 = J,Iy We, WE ( 5 ) To facilitate the analysis of the deflection of the car- riers, the L-shaped region can be further broken down into two boxes, one vertical and the other horizontal; see Fig. with WE denoting the emitter width.

RISTIC er U / . : LATERAL MAGNETOTRANSISTOR STRUCTURE 1079

I I

P-well

A n

Collector

p 0% Emitter N+ A 0

N - SUBSTRATE

(b) fl +I 4le

Fig. 3 . The SSIMT right-side cross section. (a) Minority current flow, (b) Portion of the minority carriers that contribute to the collector current flow. (c) L-shaped model of the collector current. (d) Vertical and horizontal box as equivalent for the L-shaped region.

To obtain the current change AIc2 due to a magnetic field B , we have to determine the effect of the magnetic field on both the vertical and lateral electron flows. Both flows will be deflected through an angle e,, which is de- fined by tan (@,) = PHB, where p, is the electron Hall mobility [4]. This is shown in Fig. 3(d). The change in the collector current Ic2 due to the deflection of the ver- tical component can be calculated by integrating the cur- rent density J,, in the plane x = Ll over the area defined by A Y and the emitter width W E . For the case of a uniform current distribution J,, , we obtain

AIcz,, = J , WEA Y (6)

AY = LIpHB. (7)

AIc,,, = J , WELIpHB. ( 8 )

where

Using (6) and (7) we obtain for AIc2,,

is an increase in the collector current due to elec- trons being deflected that would otherwise be part of the substrate current.

The change in the collector current Ic2 due to the de- flection of the lateral component can be calculated by in- tegrating the current density J,, in the plane y = W , over the area defined by AX and the emitter length W E . If we assume, as before, a uniform current distribution in the area of integration, we have

AIcg = J,, WEAX (9)

where

A x = L,.l,B. (10)

AIC21 = JnyWEL2PHB. ( 1 0

Substituting (10) into (9) we obtain

The electrons contributing to AIc2/ would also, without the deflection, be a part of the substrate current.

As our analysis presumed a continuous current flow of constant cross-sectional area and we have presumed uni- form current density in both boxes, the current density in both boxes must be equal, J,, = Jny. Using ( 5 ) to express J,, as a function of Ic20, we express (8) and (1 1) as

The total current change is then the sum of AI,,,, and AIC217

Defining the relative sensitivity as [7]

IO80 IEEE TRANSACTIONS ON ELECTRON DEVICES. VOL. 36. NO. 6. JUNE 19x9

we have using (14) expected that by altering these potentials it will be pos- sible to change the sensitivity of the device. It can also be expected that the applied base current Zb will also in- fluence the magnitude to Weff. We analyze the influence of each of these parameters when the others are held con- stant.

1) Influence of v r : The negative Potential Vi- applied to the P+ stripes can be expected to influence the positions of both planes XI and X,. The functional dependence of the two planes on V, should be linear, as the stripes make an ohmic contact with the neutral base region. This de-

s r = P H ~ Ll + L2 (Weff > 0). (16)

The above results for the magnetic response of the col- lector currents and the relative sensitivity were derived for current flow of half of the device. In the actual device there are, of course, two collector currents, I C , and Zc2, and the relative sensitivity is defined as

We,,

s = - . - l AZc I C , B pendente can be expressed as

where AZ, = AZ,, - Arc, and IC , = Zclo + ZC2,. The

-AZc2 and Zclo = ZC2,. The final expression for sensitiv-

XI = XI, - C I V ,

x, = x,, - c2 Vr

(24a 1 ( 24b 1

structure is symmetrical, and we can assume AZcl =

ity is therefore unaffected as both the current change and the total Current are increased by a factor of two. It is worthwhile to mention that the sensitivity could be de- fined differently. It could be normalized with respect to

where xl0 and x,, represent the position of the planes when V, = 0 and, cl and c2 are proportionality constants. The magnitudes of xI and x,, therefore, increase with an increase in I V, 1 ( V, is negative). using (24a) and (24b),

the emitter current, in which case we would have We, can be expressed as

We, = X2 - XI = We, + K r V r (25) 1 AZc (17b) S E = - . -.

where We,, = X,, - XI, and K, = cI - c2. From (25) it is obvious that, at some critical value, V,,, We,, will be reduced to zero, and the collector current shut off. In that case we have from (25)

I E

However, we use the definition (17a) in order to facilitate comparison with previously reported structures in CMOS technology [7].

We can express the collector currents Zcl and IC, as a (26)

Wefro V,, = - - Ki-

function of the magnetic field in terms of the zero field current and AZ,, and AZc2

I C , = I C , , + A k l (18) and can then express (25) in terms of V,,

Using AIc, = -AZc2 we have Using (27) and (16) we can express the sensitivity as

Equation (28) indicates that V, will have a dramatic effect on the relative sensitivity. It is necessary to note that the sum L, + L2 is also a function of V,, and using (4a) and (4b) as

(We, > 0). (21)

The last two equations suggest that both collector currents are linear functions of the magnetic induction and can be expressed as

as (24a) and (24b)3 it can be as

L1 + L2 = Lo - KLV, (29)

where S, is defined by (16). From (22) and (23) it is ob- vious the higher the sensitivity, the higher the change in the collector currents.

B. Parameters Equation (16) suggests that, in order to obtain high sen-

sitivities, We,, should be made as small as possible. The definition of planes XI and X2 used to derive We, imply that We, is a function of the applied potential V,., the stripe potential, and V,, the substrate potential. It is therefore

where Lo = W, + X , , + X,, and KL = cI + c2 . When V,. approaches V,, , We,, approaches zero and the sensitiv- ity tends toward infinity. This is of course physically un- sound and the maximum sensitivity attainable with V, = V,, will need to be determined by experiment. However, in the region where V, is slightly less than V,.,, the sensi- tivity will be very high and We,, very small, hence, we can assume that We, << L, + L,. In this region of V, , We,, will be reduced for increasing V, whereas LI + L2 will increase slightly. These two results allow us to as- sume L, + L2 is a constant. Using (4a) and (4b) the sum L, + L2 can be expressed as W, + 2Ll, and we assume for L, a likely value of one half of the p-well depth. We can now obtain V,, and K, from the experimental data.

RISTIC et id.: LATERAL MAGNETOTRANSISTOR STRUCTURE 1081

Once these two parameters are determined, we can obtain We,, from (26) and We,, as function of V, from (25).

A check of the above assumptions can be made by the following procedure. From (22) it is obvious that Zcl can be driven to zero if B is increased sufficiently to B,; it is only for B s B,. that the device will exhibit linear behav- ior. If B = B,, then I, , = 0 and we have from (22)

contacts are placed laterally, and we assume that I b affects both planes X1 and X 2 . A decrease in the base current will bring about an increase in XI and correspondingly a de- crease in X2 with the net result of decreasing w,,,. The sum LI + L2 will be a weak function of Ib if, as we as- sume, the change in both planes is nearly equal. This leads to the following expression for We, (see Appendix B)

1 - S,B, = 0 (30) = K B ( I b - zbc) (I6 > zbc) (36) or where KB is a constant and is the critical value of the

base current at which We, is reduced to zero. We have therefore that, in order to “open” up a lateral path for the electrons to the collectors, it is necessary to apply a min-

base current I b c . For all values of zb < zhc, the Col- lector currents Will be zero. Using (36) and (16) we can obtain for the sensitivity as a function of 1,

(31 )

where B,. is a the critical value of the magnetic induction at which ZC1 approaches zero. From (4b) it follows that

1 S r

B = -

L1 + L2 = WB f 2L1 (32) and using (32) and (16) we determine L1, as a function of Weff? to be

(33)

Once L1 is determined for a number of stripe potentials, L2 can be determined using (4b), and the position of the two planes XI and X 2 can be plotted as a function of V, .

2) Influence of Vs: V, is the substrate potential that provides for the reverse bias of the p-well junction. This junction can be approximated by a linearly graded junc- tion. The depletion width of a linearly graded junction is proportional to ( Vb; + Vs - Vn,)’/3 [12], where Vb; is the built-in junction voltage and VB, is the potential of the neutral base region at the edge of the depletion region. The assumption that the position of the planes XI and X2 are linearly related to the junction width (see Appendix A) leads to the following expression for We,,:

weff = K,[(Vsc - 1/B)1/3 - (Vs - Vn)’/3]

( I Vsl < I VscI). (34)

Using (16) we can express S, as

( I Vsl < I VscI) ( 3 5 )

where is a critical substrate voltage at which We, re- duces to zero, VB = VB, - v,; and K, is a constant. For this case the sum L1 + L2 can be expected to be relatively insensitive to changes in V, and the behavior of S, is dom- inated by the reduction in Weff. S, is an increasing func- tion of Vs. Ks can be determined from the experimental data, L1 from (33) and (34), and L2 from (4b).

3) Znfluence of Ih: A constant base current I b is used to forward bias the emitter-base junction of the device. It can be predicted that an increase in the magnitude of I b

will have an effect on We, that is equivalent to a decrease in the magnitude of the applied potential V,. The base

S, is a decreasing function of I b , Kn can be determined from the experimental data, L1 from (33) and (36), and L2 from (4b).

IV. EXPERIMENTAL RESULTS The experimental measurements were obtained using

the circuit shown in Fig. 2 . The circuit was operated in two configurations. Either the collector currents Zcl and Zc2 were measured directly and AZ, calculated, or the col- lector load resistors were placed between the supply volt- age and the collectors, and the potential difference be- tween the two collectors is measured, A V = VCl - Vc2. In all of the reported measurements the values of the two stripe potentials V,, and Vr2 were adjusted so as to make Zclo = ZC2,. V, is defined as ( V,, + Vr2)/2 and is used as a representative value for the stripe potential.

Two sensitivities were used to characterize the device: the relative sensitivity S, and the absolute sensitivity Sa. The relative sensitivity was defined by (17) and the ab- solute sensitivity is defined as [ 11

A V B

so = -

where A V is expressed as

A V = ( I C 2 - I C 1 ) R L (39)

and RL is the load resistivity.

A. Basic SSZMT Characteristics The basic characteristics of the SSIMT with respect to

bias conditions and the response of the collector currents to a magnetic field are presented in this section. Typical values for the emitter current and the n+ emitter/p+ stripe breakdown voltage are 15 mA and 4.5 V, respectively. We use for the Hall mobility of the electrons the value p H = 1200 cm2/V * s given that pH = rnp, , , p, = IO3 cm2/V - s at an acceptor level of 10l6 cm-3 [I21 and r,, = 1 . 1 ~ 3 1 , 1141.

IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 36. NO. 6, JUNE 1989 1082

Q Y

0 Q -

B(mT) Fig. 4. AI , = IC* - I , , as a function of fB, Ih = 8.5 mA, V , = 5 V, Vs

= 5 V. and V, = -250 mV. I , , , = 801 FA, I,,, = 795 FA.

In Fig. 4, the response of AZ, = Zc2 - Zcl to a magnetic field is shown. This figure clearly shows the symmetrical nature of the SSIMT, with the measured value of AZ, switching sign as the field is reversed from the positive I direction to the negative.

From the model of the SSIMT given above, we have from ( 5 ) that Z,,, is proportional to We, and also state in (27) that We, = K r ( V,. - Vrc) . We can therefore expect Zco = Zcl + Zc2 to be a linear function of V,. and to de- crease from a maximum at V,. = 0 to I,, = 0 at V,. = V,.,. In order to check this prediction I,, was measured as a function of V,. for three different base currents; these re- sults are presented in Fig. 5 . The linear relationship of Zco on V,. is confirmed by these result. In the immediate vi- cinity of Vrc the linear relationship no longer holds, which is a reflection of the increasing instability of the device as V, approaches Vrc. It is difficult to obtain V,., exactly from this plot due to this transition region, and the fact that one collector switches off at a slightly lower voltage due to the unavoidable asymmetries of a real device. However, an approximate value for V,., can be obtained by using the midpoint of the transition region as an indication of Vr,. For example, a value of approximately -200 mV is ob- tained from Fig. 5 for ZB = 7 mA, Vc = 5 V, and V, = 5 v.

B. S , as a Function of V,., V,, and Zb In Fig. 6 we can see the crucial role of V,. on the re-

sponse of the device to a magnetic field. Three bias points are shown here; in each case Zb and Vs are held constant at 7 mA and 5.0 V, respectively. The stripe potential is, however, different for the three bias points with V,. equal- ing -198, -193, and -150 mV. Fig. 6 clearly shows that increasing V,. dramatically increases the sensitivity of the device. The relative sensitivity increases from 500 percent/T at V,. = - 150 mV to 3050 percent/T at V,. = - 198 mV. This is the highest sensitivity reported for an

h

4 Y

- 8

0 ' ' ' I ' =-:, 0 200 400 600

lVrl (mv) Fig. 5 . Total collector current as a function of V, for B = 0 T, Vc = 5 V,

V, = 5 V. Curve 1: Ib = 7 mA. Curve 2: Ib = 8 mA. Curve 3: I , = 9 mA.

B (mT)

Fig. 6. AI , as a function of the magnetic induction, I,, = 7 mA, Vc = 5 V, Vs = 5 V. Curve 1: V , = -198 mV, S, = 3050 percent/T. Curve 2: V, = - 193 mV, S, = 1670 percent/T. Curve 3: V, = - 150 mV, S, = 580 percent/T. E , , = 30 mT, B,* = 60 mT, Bc3 = 185 mT.

MT and furthermore is achieved at low magnetic fields ( B < 40 mT). If definition (17b) is used to calculate the sensitivity at this operating point, we obtain 177 per- cent/T (using a typical value of 15 mA for ZE), a value still much higher then any other reported. Our investiga- tion shows that, as predicted by (20), one collector cur- rent can be driven to almost zero if B is increased suffi- ciently. The values of the magnetic field B, for which this occurs for the three operating points in Fig. 6 are 30, 60, and 185 mT. B, can be varied from 30 mT out well be- yond l .O T with the appropriate choice of biasing condi- tions; this means that the linear range of the device can be increased at the expense of lower sensitivity. This abil- ity of the SSIMT makes it particularly attractive as a de- vice for magnetic switching. It is interesting to note that the slope of Zc2 undergoes a reduction at the magnetic field

RISTIC et 01.: LATERAL MAGNETOTRANSISTOR STRUCTURE

r

X 2 -

1 -

1083

1

0 ~ " " ~ " " ~ " " ~ ' " ' ~ ' 0.00 0.05 0.10 0 15 0 20

lVrl (VI Fig. 7. S, as function of V r , Ih = 7 mA, Vc = 5 V, B = 30 mT, V.y =

5 v.

B,. It is obvious that a plot of AZ, will also undergo a change in slope at this point, as Zcl = 0 for any field greater than B, and therefore AZ, = AIc2 for B 1 B,.

The relative sensitivity S, as a function of V, is pre- sented in Fig. 7 and (28) is shown as a solid curve. As can be seen, the agreement between theory and experi- ment is good. The values of K, and V,, were 5.66 pm/V and -212 mV, respectively. It can be seen that V,, is far below the breakdown voltage of the n+ emitter/p+ stripe junction. These measurements clearly show the crucial role of V, for the attainment of high sensitivity. An ad- ditional insight into the effect of V, can be seen in Fig. 8 where results for We,,, L1 + L2 and the planes XI and X2 are presented. It can be seen that, as V, approaches V,,, We,, decreases to zero. At the same time L1 + L2 slightly increases. This conforms our assumption that the sum LI + L2 is almost constant and the reduction of We,, is the dominant influence on S, . Corresponding values for XI and X 2 are shown in Fig. 8(b). It is interesting to note that X 2 is almost unaffected by changes in V, and positioned at = 5 pm. The reduction of We, is therefore brought about only by a shift in the position of the plane X I .

The influence of V, on the relative sensitivity is pre- sented in Fig. 9. Experimental results are in good agree- ment with (35). It is of special interest to note the reduc- tion of V,, with the increasing of V,; this is expected as the plane XI is positioned deeper within the device for larger negative stripe potentials. For the case of V, = - 180 mV ( K , = 6.56 pm/( V)'I3), We, and L I + L2 are plotted in Fig. 10. As the sum L, + L2 changes only slightly, it is obvious that, as before with variations in V,, the dominant influence on S, is the reduction of We,.

Finally the influence of Zb on S, is presented in Fig. 11. These results are in agreement with (37) ( KB = 1 .O) , and confirm the existance of Ihc, with S, being at a maximum at Ib = Ibc and decreasing as Ib is increased. It is apparent that, as a larger negative potential is applied to the p+

h

Y E,

s c c

h

v E, N

- 1 I x. I

x I I

0 05 0 10 0 1 s 0 20 - . 0 00

lVr l (VI (b)

Fig. 8. (a) We,, and L , + L , as a function of V, . (b) X , and X z as a function of V,. lh = 7 mA, Vc = 5 V, Vs = 5 V.

stripes, it is necessary to have a larger Zh to open up a path for the lateral flow of electrons. We, and L1 + L2 are plot- ted in Fig. 12 as a function of Ib, and again there is little variation in the sum L I + L2, and We, governs the behav- ior of S, .

C . Absolute Sensitivity Very often the circuitry supporting an MFS uses a volt-

age signal produced on two load resistors rather than mea- suring the current change directly. The absolute sensitiv- ity Sa defined by (38) serves as a parameter used to characterize the device for this situation. It is apparent from the definition of S, that A V is proportional to AIc and RL. This of course implies that the absolute sensitivity can be chosen by an appropriate choice of load resistance. For magnetic sensors displaying small relative sensitivity, R, is sometimes chosen to be as high as the order of megaohms. However, when a sensor provides a much larger current change, R, can be much smaller.

1084

E- 8 rz v

1 2 3 4 5 6 7

VS(V) Fig. 9. S, as function of Vs, Vc = 5 V , I,, = 7 mA. Curve 1: V , = 0 mV.

Curve 2: V , = -50 mV. Curve 3: V , = -100 mV. Curve 4: V, = - 180 mV.

1 25

L1+ L 2 20

3 0

IEEE TRANSACTIONS ON ELECTRON DEVICES. VOL. 36. NO. 6. JUNE 1989

h

v E, 20-

P c c

1 0 - 1

: 1 5 h

E

- i o -I

3. v N

.- -I

5

vs (V) Fig. 10. We,, and L , + L2 as a function of Vs. I,, = 7 mA, Vc = 5 V, V,

= -180 mV.

E- $ & Y

Ib (mA) Fig. 11. S, as function of I,,, V , = 5 V, Vs = 5 V. Curve 1: V, = 0 mV.

Curve 2: V, = - 198 mV. Curve 3: V , = -460 mV.

4

3

h

v E,

P c U-

2

1

0

20 L l + L 2

Ib (mA)

Fig. 12. We, and L , + L , as a function of I,,. Vs = 5 V , Vc = 5 V , V. = - 198 mV

0.00 0.05 0.10 0 15 0

lVrl (VI Fig. 13. Sa as a function of V, for Vs = 5 V , Ib = 7 mA, R, = 500 Cl,

B = 30mT.

It can be seen that the absolute sensitivity increases with an increase in the negative potential of V,. This can easily be understood when Sa is expressed in terms of the rela- tive sensitivity using (22) and (23)

Sa = S,ICORL. (40)

S, is a sharply increasing function of V, and Ico linearly decreases with increasing V,. As a result, S, is an increas- ing function of V,, but does not increase as sharply as S,. This investigation is presented in Fig. 13.

V. CONCLUSION We have investigated a novel lateral magnetotransistor

structure realized in standard CMOS technology. An an- alytical model of the device has been presented to explain the high sensitivity of the device. This model would ap- pear to be confirmed and supported by the experimental analysis of the device. The results indicate that the con- finement of the minority carriers and the double deflection

RISTIC et U / : LATERAL MAGNETOTRANSISTOR STRUCTURE

in the neutral base region are primarily responsible for the high sensitivity of the device.

A number of important qualities of the SSIMT structure stand out in the experimental results presented above. The linear response to the magnetic field indicates that the un- derlying operating principle is carrier deflection. This lin- earity is itself a highly desirable quality of a magnetic sensor. A great choice of relative sensitivities are avail- able from 50 to 3000 percent/T. The last figure far ex- ceeds any reported sensitivities for linear MT’s. The com- bination of a linear response and the great range of sensitivity raise the possibility of magnetic switching. By correct choice of the operating conditions it is possible to choose at what magnetic field B, one of the collector cur- rents is driven to almost zero. The range of B, is from 30 mT to over 1 T. In practical circuitry applications the magnitude of the signal A V = AZcRc is important.

With the SSIMT structure we obtain very large absolute sensitivities of up to 9 V/T, at fields of 30 mT and with Rc = 500 Q. This is the highest reported absolute sensi- tivity for an MT using load resistors this small.

As the SSIMT is fabricated in a standard CMOS pro- cess, it can be readily manufactured. The overall behav- ior, the unique combination of high sensitivity, linear re- sponse, and a wide variety of dc working points suitable for switching applications make this device quite attrac- tive for different circuitry applications.

APPENDIX A We, AS A FUNCTION OF V,

The functional dependence of the two planes X2 and XI on V, can be expressed through a linear relation with the depletion region of the p-well/n-substrate junction

(Ala) WL

x2 = x2s - cs2 7 x, = XI, - c,, - WL

2

where X,, and XIS represent the positions of the two planes for an arbitrary value Vs = const., Cs, and C,, are pro- portionality constants, and WL/2 is the width of the de- pletion region on the p-well side of the junction. The p-well/n-substrate junction can be considered as a linearly graded junction and WL expressed as [ 121

where E is the dielectric permittivity for silicon, a is the impurity gradient at the junction, q is the elementary charge, Vbj is the built in junction voltage, and V,, is the potential in the neutral base region at the edge of the de- pletion region. Substituting VB = Vbj - V,, and C,,. = ( 1 2 ~ / q a ) I / ~ , WL can be expressed as

w, = c,( v& - VB)”3. (A3)

1085

Using (Ala), (Alb), and (A3) We,, can be expressed as

weff = x2 - XI = Weffr - Ks(Vs - VB)1/3 (A4)

where Weffs = X 2 , - XI, and Ks = (CS2 - Csl)( Cw/2) . Increasing Vs will decrease We, and at some critical value V,,, We, will be reduced to zero and the collector current shut off. We can therefore express Weffs as

(A51 1/3 WeffJ = Ks( Vsc - V B ) .

Using (A5) and (A4) we have for Weff, 1 / 3

= K S [ ( V S c - VB)1/3 - (‘S - ‘ E ) ] ( Vs V s c ) (‘46)

where V,, is a critical substrate value at which We,, re- duces to zero.

APPENDIX B we, AS A FUNCTION OF

The functional dependence of the two planes XI and X , of Zb should be a linear function as the base contact is an ohmic contact in the neutral base region. This dependence can be expressed as

XI = Xlb - cb l zb

x, = X2b - Cb2lb

where Xlb and X,, represent the positions of the two planes for an arbitrary value of I b , Zb = const., and Cb, and Cb, are proportionality constants. Using (Bla) and (Blb) We,, can be expressed as

(Bib)

we, = x, - XI = we,, + c,z, (B2)

where We,, = X 2 , - Xlb and C, = Cb2 - Cbl. Decreas- ing I b will decrease We, and at some critical value Z h c , We, will be reduced to zero and collector current shut off. We can, in that case, express We,,, as

= -cBzbc . (B3)

We, = C B ( z b - z b c ) ( I b > Zbc) (B4)

Using equations (B3) and (B2), We,, can be expressed as

where Zbr is a critical value of the base current at which effective width reduces to zero.

REFERENCES H . P. Baltes and R. S. Popovic. “Integrated semiconductor magnetic field sensors,” Proc. IEEE, vol. 74, p. 1107, 1986. L. W. Davies and M. S . Wells, “Magnetotransistor incorporated in a bipolar IC,” in Proc. lCMCST(Sidney, Australia), 1970, p. 34. J . B. Flynn, “Silicon depletion layer magnetoameter,” J . Appl. P h y s . , vol. 41, p. 2750, 1970. V. Zieren and B . P. M. Duyndam, “Magnetic-field-sensitive multi- collector n-p-n transistor,’’ IEEE Trans. EIectron Devices. vol. ED- 19, p. 83, 1982. A. W. Vinal and N . A . Masnari, “Magnetic transistor behavior ex- plained by modulation of emitter injection, not carrier deflection.“ IEEE Electron Device Lett., vol. EDL-3, p. 203, 1982. I . M. Mitnikova, T. V. Persiyanov. G. I. Recalova, and G. Shlyub- ner, “Investigation of the characteristics of silicon lateral magneto- transistors with two measuring collectors,” Sov. Phys. Semicond., vol. 12, p. 26, 1978.

1086 IEEE TRANSACTIONS ON ELECTRON DEVICES. VOL. 36. NO. 6. JUNE 1989

R. S . Popovic and R. Widmer, “Magnetotransistor in CMOS Tech- nology,” IEEE Trans. Electron Devices, vol. ED-33, p. 1334, 1986. R. S. Popovic and H. P. Baltes, “Dual-collector magnetotransistors optimized with respect to injection modulation,” Sensors and Actua- tors, vol. 4, p. 155, 1983. LJ. Ristic, T. Smy, H. P. Baltes, and I . Filanovsky, “A highly sen- sitive magnetic field sensor based on magnetotransistor action with suppressed sidewall injection,” in Proc. Yugoslav Conf. Microelec- tronics (Banja Luka, May 13-15, 1987), p . 25. LJ. Ristic, T. Smy, H. P. Baltes, and I . Filanovsky, “Suppressed sidewall injection magnetotransistor in CMOS technology,” in Dig. Transducers ’87, 4th Int. Con5 Solid-State Sensors and Actuators (Tokyo, June 2-5, 1987). p. 543. L. Ristic, H. P. Baltes, T. Smy, and I. Filanovsky, “Suppressed sidewall injection magnetotransistor with focused emitter injection and carrier double deflection,” IEEE Electron Device Let t . , vol. EDL-8, p. 395, 1987. S . M. Sze, Physics of Semiconductor Devices. New York: Wiley, 1981, p. 81. J. F . Lin, S . S . Li, L. C. Linares, and K . W. Teng, “Theoretical analysis of Hall factor and Hall mobility in p-type silicon,” Solid- State Electron., vol. 24, p. 827, 1981. P. Norton, T. Braggins, and H. Levinstien, “Impurity and lattice scattering parameters as determined from Hall and mobility analysis in n-type silicon,” Phys. Rev. B , vol. 8, p. 5632, 1973.

* LjubiSa RistiC (M’86) received the B.Sc., M.Sc., and Ph.D. degrees, all in electrical engineering from Nis University, Yugoslavia.

He served as a Development Engineer and a Chief Engineer in the Semiconductor Device Di- vision of the Electronics Industry Corp., Yugo- slavia, from 1975 to 1981. Between 1981 and 1985, he was involved in project management as an R&D Manager in the Semiconductor Device Division. From 1982 to 1985, he was also an Ad- junct Lecturer at Nis University. In 1985, he

joined the Department of Electrical Engineering at the University of Al- berta as an Associate Professor. His research interests have included the- oretical and experimental study of Zener rectifiers, power and Schottky diodes, Si-Si02 interfaces, and MIS structures. The phenomena related to planar transistors including passivation technology of planar structures, power transistors, and both bipolar and CMOS IC circuits and technology were among the subjects of his research interests. Also, a part of his re- search activity has been the failure mechanisms and reliability investiga-

tions of semiconductor devices. Currently, he is engaged in the research and design of microsensors and microactuators, micromachining pro- cesses, and integrated circuits. He is the author or co-author of more than 40 scientific papers and holds several patents.

Dr. RistiC was Chairman of the Yugoslav Conference on Microelec- tronics in 1980 and 1984.

Tom Smy was born in Vancouver, Canada, in 1962. He received the B.Sc. degree in electrical engineering from the University of Alberta, Can- ada, where as a National Science and Engineering Research Council scholar he is presently working toward the Ph.D. degree. His current research in- terest is in microelectronic magnetic sensors.

*

Henry P. Baltes (M’81) received the D.Sc. de- gree from ETH Zurich in 197 l .

Subsequently he was on the faculty of the Freie Univers~tat Berlin, the University of Dusseldorf, the University of Waterloo, Ontario, and the EPF Lausanne He joined Landis & Gyr Zug Corpo- ration in 1974, where he later became head of the solid-state laboratory involved in microsensor re- search. From 1975 to 1982, he taught laser phys- ics and microsensors at the EPF Lausanne. From 1983 to 1988, he held a Henry Marshall Tory

Chair at the University of Alberta, where he taught courses on microelec- tronic processing and sensor technology and directed a research program in microsensors. In 1986, as Acting President of the Alberta Microelec- tronic Centre, he directed the establishment of a microelectronic fabrica- tion facility. From 1986 to 1988, he was a Director of LSI Logic Corpo- ration of Canada, Inc. In 1988, he was appointed Professor of Physical Electronics at ETH Zurich.

Prof Baltes is a member of the OSA, the IOP, the EPS, the Swiss Phys- ical Society, and the Electrochemical Society In 1984, he was Chairman of the Canadian Conference on VLSI