Barrier Height Voltage Reference

Typical solid-state voltage reference devices generally obtain a reference voltage from a predictable breakdown voltage or from a known forward biased current-voltage relationship of a junction device. Temperature dependence of one device is often compensated for by combination with another device also having a temperature dependence. A constant-current source is usually required for the best precision. The present invention is a voltage reference comprising a pair of field-effect devices each substantially identical to the other except for gate-to-channel potential barrier characteristics. Each field-effect device is biased with a substantially identical drain current to produce a reference voltage corresponding to the difference in barrier potentials at the gate-to-channel interface of the two field-effect devices. The invention provides a voltage reference which is substantially independent of operating point, supply potential, and temperature.

FIG. 1 is a cross-sectional drawing of a junction field-effect transistor.

FIGS. 2A, 2B and 2C show energy band diagrams of two substantially identical semiconductor substrates wherein the first has in combination therewith, a metal gate of a first type and the second substrate has in combination therewith, a metal gate of a second type.

FIGS. 3A, 3B and 3C show energy band diagrams of two semiconductor substrates having gates of the same material as the substrates, the gates being doped with different concentrations and forming P-N junctions with their respective substrates.

FIGS. 4A, 4B and 4C show energy band diagrams of a first semiconductor substrate of a first type and a second semiconductor substrate substantially identical to the first semiconductor substrate, the first semiconductor substrate having a semiconductor gate of a second type and the second semiconductor substrate having a semiconductor gate of a third type.

FIG. 5 is a diagram of a voltage reference made in accordance with the invention.

FIGS. 6A, 6B, 6C and 6D are diagrams illustrating the processing steps for fabricating a voltage reference monolithically in accordance with one embodiment of the invention.

FIG. 7 is a diagram of another voltage reference made in accordance with the invention.

FIG. 8 is a diagram showing a method for combining a plurality of voltage references made in accordance with the invention to achieve larger reference voltages.

FIG. 9 shows a preferred voltage reference wherein surface area required for integrated fabrication is minimized.

FIG. 10 is a diagram of a preferred embodiment utilizing voltage variable capacitors in accordance with the invention.

FIG. 11 is a diagram of another preferred embodiment utilizing voltage variable capacitors.

The electrical characteristics of an N-channel field effect transistor, FET hereafter, operated above turn-on and below pinch-off are described, for example by A. S. Grove, Physics and Technology of Semiconductor Devices, John Wiley and Sons, Inc., New York, 1967: ##EQU1## where I.sub.D is the drain current, Z is the channel width, q is the magnitude of the electronic charge, .mu..sub.n is the electron mobility, N.sub.D is the density of donors in the channel region, d is the channel thickness, L is the channel length, V.sub.D is the drain voltage which is measured between the drain and the source, .epsilon..sub.s is the semiconductor dielectric constant, .phi..sub.B is the built-in potential between the gate and the channel and referred to herein as the barrier potential, and V.sub.G is the gate voltage which is measured between the gate and the source.

A FET operated in the pinch-off condition is also described by Equation 1 wherein the value for V.sub.D is constant and V.sub.D is given by (see, for example, Physics and Technology of Semiconductor Devices as cited above): ##EQU2##

Equations 1 and 2 are applicable to a junction FET having geometry similar to that of FIG. 1. Equations similar to equations 1 and 2 are also applicable to metal gate FETs as well as to heterojunction FETs.

For two FETs, identical except for their gate-to-channel potentials and biased with equal I.sub.D 's in pinch-off or equal I.sub.D 's and V.sub.D 's below pinch-off, the quantity (.phi..sub.B - V.sub.G) is identical. A reference voltage can therefore be developed, which is:

where V.sub.Ref is the reference voltage, .DELTA. V.sub.G is the difference in gate voltages, and .DELTA. .phi..sub.B is the difference in barrier potentials. This reference voltage is developed whether or not the FET's are operated above or below pinch-off. It is also developed when the FET's are operated below turn-on so long as identical I.sub.D 's and V.sub.D 's are maintained.

Referring to FIG. 1, there is shown an FET having an n-channel 802 connected to a back gate 804 of p-type semiconductor material, a front gate 806 also of p-type semiconductor material, a source contact 808, and a drain contact 810. While FIG. 1 shows a junction FET having both a top gate 806 and a back gate 804 wherein both form step homojunctions, the only requirement for making a voltage reference in accordance with this invention is that two FET's be substantially identical except for the gate-to-channel built-in potential of at least one gate-to-channel junction.

A difference in .phi..sub.B 's is obtained, for example, by utilizing two different metals for gates which make rectifying contacts to the substrates but which result in different gate-to-channel built-in potentials. FIG. 2(a) shows the energy band diagrams of two substantially identical semiconductor substrates 402 and 404 which are, for example, n-type, the first semiconductor substrate 402 being connected to a first gate metal 406, and the second semiconductor substrate 404 being connected to a second gate metal 408. Gate metals 406 and 408 are chosen so as to make rectifying contact with the semiconductor substrates 402 and 404 and so that the barrier potential .phi..sub.B1 of the first FET is greater than the barrier potential .phi..sub.B2 of the second FET. Since .phi..sub.B2 is smaller than .phi..sub.B1' the channel width is larger in the second semiconductor substrate 404 than in the first semiconductor substrate 402. As shown in FIG. 2(b) biasing the second gate metal 408 with respect to the second semiconductor substrate 404 by an amount .phi..sub.B2 - .phi..sub.B1 results in the channel width of the second semiconductor substrate 404 being the same as in the first semiconductor substrate 402. A reference voltage V.sub.Ref = .phi..sub.B2 - .phi..sub.B1 is thereby produced. In FIG. 2(c) gate metals 406 and 408 are biased by different voltages from those shown in FIG. 2(b). However, the channel width in the second semiconductor substrate 404 is the same as in the first semiconductor substrate 402, resulting in the same reference voltage (V.sub.Ref = .phi..sub.B2 - .phi..sub.B1) as in FIG. 2(b).

The difference in .phi..sub.B 's is also obtained by utilizing two semiconductor materials for the gates which are the same as the semiconductor material of the two substrates and utilizing a first gate semiconductor material which has a first doping density different from the doping density of the second gate semiconductor material. FIG. 3(a) illustrates an energy band diagram for two substantially identical semiconductor substrates 502 and 504. The first semiconductor substrate 502 is connected to a first semiconductor gate 506, and the second semiconductor substrate 504 is connected to a second semiconductor gate 508. The doping species of semiconductor gates 506 and 508 are chosen so that the majority carrier of the semiconductor substrates 502 and 504 is of different polarity from that of the semiconductor gates 506 and 508. The magnitude of the doping density in the first semiconductor gate 506 is different from that of the second semiconductor gate 508 thereby producing different gate-to-channel built-in potentials. Since .phi..sub.B2 is smaller than .phi..sub.B1, the channel width is larger in the second semiconductor substrate 504 than in the first semiconductor substrate 502. As shown in FIG. 3(b), biasing the second semiconductor gate 508 with respect to the second semiconductor substrate 504 by an amount .phi..sub.B2 - .phi..sub.B1 results in the channel width of the second semiconductor substrate 504 being the same as in the first semiconductor substrate 502. A reference voltage (V.sub.Ref = .phi..sub.B2 - .phi..sub.B1) is thereby produced. In FIG. 3(c) semiconductor gates 506 and 508 are biased by different voltages from those shown in FIG. 3(b). However, the channel width in semiconductor substrate 504 is also the same as in the semiconductor substrate 502, resulting in the same reference voltage (V.sub.Ref = .phi..sub.B2 - .phi..sub.B1) as in FIG. 3(b).

The difference in .phi..sub.B 's is also obtained by utilizing two different semiconductor materials for the gates. FIG. 4(a) shows two substantially identical semiconductor substrates 602 and 604 which are of a first semiconductor material which is, for example, n-type. The first semiconductor substrate 602 is connected to a first semiconductor gate 606, which is of a second semiconductor material and which is, for example, p-type. The second semiconductor substrate 604 is connected to a second semiconductor gate 608 which is of a third semiconductor material and which is, for example, p-type. The materials and doping species of semiconductor gates 606 and 608 are chosen so that rectifying contact is made to the semiconductor substrates 602 and 604 and so that .phi..sub.B1 is not equal to .phi..sub.B2. Since .phi..sub.B2 is smaller than .phi..sub.B1, the channel width is larger in the second semiconductor substrate 604 than in the first semiconductor substrate 602. Biasing the second semiconductor gate 608 with respect to the second semiconductor substrate 604, as shown in FIG. 4(b) by the amount .phi..sub.B2 - .phi..sub.B1 results in the channel width of the second semiconductor substrate 604 being the same as in the first semiconductor substrate 602. A reference voltage V.sub.Ref = .phi..sub.B2 - .phi..sub.B1 is thereby produced between the second semiconductor gate 608 and the first semiconductor gate 606. In FIG. 4(c) semiconductor gates 606 and 608 are biased by different voltages from those shown in FIG. 4(b). However, the channel width in the first semiconductor substrate 602 is also the same as in the second semiconductor substrate 604, resulting in the same reference voltage (V.sub.Ref = .phi..sub.B2 - .phi..sub.B1) as in FIG. 4(b).

Combinations of gate materials shown in FIGS. 2-4 can also be used to fabricate a voltage reference in accordance with this invention. Such voltage references include the following combinations: (1) a homojunction and a schottky barrier, (2) a homojunction and a heterojunction, and (3) a schottky barrier and a heterojunction.

Referring to FIG. 5, there is shown a first preferred embodiment made in accordance with the invention. A first n-channel FET 5 and a second n-channel FET 7 are substantially identical except for different barrier potentials. A bias supply 1 is connected between source contact 13 and drain contact 17 of FET 5 and a bias supply 3 is connected between source contact 15 and drain contact 19 of FET 7. The potentials of gate contact 35 of FET 5 and gate contact 37 of FET 7 are adjusted by gate biasing means not shown in FIG. 5. The gate biasing means not shown in FIG. 5, the supply source 1 and the supply source 3 are adjusted so that the drain voltage of FET 5 is substantially equal to the drain voltage of FET 7 and the drain current of FET 5 is substantially equal to the drain current of FET 7. A reference voltage, V.sub.Ref, is produced by the difference between the potentials of the gate contact 35 of FET 5 and the gate contact 37 of FET 7.

The matching of the two FET's is enhanced by fabricating them in close proximity to each other on the same substrate. The substrate doping densities and the mobilities are then substantially the same in both FET's. FIG. 6 shows processing steps to fabricate monolithically two schottky barrier FET's which are identical except for their gate metals. Referring to FIG. 6(a), a high resistivity substrate 702, such as, for example, silicon, is first cleaned and polished. A low resistivity epitaxial layer 704 is then grown on the substrate 702. The field oxide 706 is grown, and the isolation diffusion windows 708 opened, followed by the isolation diffusion 710. As shown by FIG. 6(b), source and drain windows 712 and 714 are etched, followed by source and drain diffusions 716 and 718. In FIG. 6(c) gate window 720 and source and drain contacts 722 and 724 are etched, followed by deposition of a first metal 725, as, for example, by RF sputtering platinum. The first metal is removed except over the source and drain contacts 722 and 724 and over gate window 720. Referring to FIG. 6(d), second gate window 726 is etched and a second metal 727 is deposited, as, for example, by electron-beam evaporating titanium. The second metal 727 is removed everywhere except over the previously deposited first metal 725, except for any necessary interconnecting metal and bonding pads, and except over the second gate window 726. Modifications to the above processing steps are possible. One such modification, for example, is to fabricate the two FET's either on the same or different substrates using the same processing steps until after the contact mask step. The FET's are separated and different gate field plate metals and interconnecting metals are deposited on each.

V.sub.Ref is temparature dependent only to the extent of mismatches in the FET's and to the extent of the difference in temperature coefficients of their respective gate-to-channel barrier heights.

Referring to FIG. 7, there is shown another preferred embodiment made in accordance with the invention. A first p-channel FET 2 and a second p-channel FET 4 are substantially identical except for different gate-to-channel barrier heights. An amplifier 6 is coupled to drive gate 44 of FET 4. Amplifier 6 has a non-inverting input connected to source contact 20 of FET 2 and a resistor 8 and an inverting input connected to source contact 40 of FET 4 and a resistor 10. Amplifier 6 has a larger input impedance as compared with the resistances of resistors 8 and 10. A negative supply voltage is coupled across terminal 12 and ground 14. Drain contacts 22 and 42 are connected to terminal 12. Gate contact 24 of FET 2, resistor 8 and resistor 10 are connected to ground 14. The drain currents and drain voltages of FET 2 and FET 4 are maintained substantially equal by selecting matching resistances 8 and 10 and by the action of amplifier 6 driving the gate contact 44 of FET 4. A reference voltage is obtained across terminals 16 and 14.

Referring to FIG. 8, there is shown a combination comprising two of the voltage references depicted in FIG. 7. The reference voltage is obtained across terminals 16 and 14. A plurality of references may be similarly combined to provide larger reference voltages.

Referring to FIG. 9, matched FET's 28 and 30 are substituted for the matched resistors 8 and 10 in the embodiment depicted in FIG. 7. This substitution allows conservation of substrate area should large resistance values be required for the resistors R8 and R10.

Other .phi..sub.B dependent parameters, as, for example, the AC (alternating current) conductance and transconductance, of the FETs can also be detected and used to adjust the bias conditions so that a reference voltage is obtained.

The invention may also be fabricated using two voltage variable capacitors. Gate-to-substrate junctions are made, as described above for the FETs. Referring to FIG. 2, it is seen that the depletion layer in semiconductor substrate 404 is substantially the same as in semiconductor substrate 402 if gate 408 is biased with respect to semiconductor 404 at a voltage V.sub.Ref more negative than gate 406 is biased with respect to semiconductor substrate 402. For such a bias condition the capacitance between gate 406 and substrate 402 will be substantially equal to that between gate 408 and substrate 404. The semiconductor junctions of FIG. 3 and the heterojunctions of FIG. 4 may also be used so long as the major part of the depletion layer extends into the substrates.

Referring to FIG. 10, a preferred embodiment made in accordance with the invention using voltage variable capacitors is shown. Voltages are applied to each of the two voltage variable capacitors so that the same depletion layer width is obtained in each voltage variable capacitor, thereby resulting in the same capacitance for each capacitor. The two capacitors are essentially identical except for their built-in potential. A reference voltage V.sub.Ref is produced, which is given by:

where V.sub.G1 and V.sub.G2 are gate-to-substrate voltages for first and second voltage-variable capacitors 202 and 204 and .phi..sub.B1 and .phi..sub.B2 are the respective built-in potentials for the first and second voltage variable capacitors 202 and 204. The gate-to-substrate voltages are varied to obtain matching capacitances.

An alternating current (AC) voltage source 206 is coupled by coupling capacitors 208 and 209 to gates 210 and 212 of the voltage variable capacitors 202 and 204, respectively. Substrate contacts 214 and 216 of the voltage variable capacitors 202 and 204 are coupled to ground 200 by resistors 218 and 219, respectively. Resistors 218 and 219 are of substantially identical resistance and are coupled respectively to rectifiers and filters 220 and 222. Outputs 228 and 230 of the rectifiers and filters 220 and 222 are coupled to differential inputs of amplifier 224. The amplifier 224 has an output coupled to the gate 210 of the voltage variable capacitor 202.

The resistance of resistors 218 and 219 is selected to be less than the magnitude of the reactance of voltage variable capacitors 202 and 204. Biasing means 226 adjusts the capacitance of voltage variable capacitor 204 to a convenient value and biasing means 226 has an impedance of magnitude greater than the magnitude of the reactance of voltage variable capacitor 204. The magnitudes of reactances of coupling capacitors 208 and 209 are chosen to be less than the magnitude of reactances of voltage variable capacitors 202 and 204 at the frequency of AC voltage source 206.

Alternating current voltages having magnitudes substantially proportional to the capacitances of voltage variable capacitors 202 and 204 are developed at substrate contacts 214 and 216. These voltages are converted to direct current (DC) voltages by rectifiers and filters 220 and 222 to produce signals at outputs 228 and 230 which drive differential amplifier 224. Amplifier 224 drives the voltage at the gate 210 of voltage variable capacitor 202 until the capacitance of voltage variable capacitor 202 is substantially equal to the capacitance of voltage capacitor 204. The reference voltage V.sub.Ref is developed between gate 210 and gate 212 and is substantially equal to the difference in built-in potentials of field-effect capacitors 202 and 204.

Referring to FIG. 11, there is shown another preferred embodiment made in accordance with the invention. Identical AC current sources 305 and 306 drive identical AC currents into voltage variable capacitors 302 and 304 which are substantially identical except for their built-in potentials. Rectifying and filtering means 320 and 322 are connected to voltage variable capacitors 302 and 304 respectively.

A DC voltage is produced at input terminals 328 and 330 of amplifier 324 which forces the output of amplifier 324 to bias voltage variable capacitor 302 such that its capacitance is equal to that of voltage variable capacitor 304. A reference voltage V.sub.Ref is produced between terminals 350 and 352 which is equal to the difference in the built-in potentials of voltage variable capacitors 302 and 304 (see Equation 4). Bias source 326 is adjusted to obtain a convenient value of capacitance.

Differences in voltage variable capacitor built-in potentials are produced, for example, by the techniques disclosed herein with respect to field-effect transistors.

Given below are tables of approximate voltage reference values produced by the embodiments discussed hereinabove for various metal combinations on the substrate material types indicated. In general, the values given below are processing dependent and are affected by the degree of substrate doping as well as species. A V.sub.Ref obtained in practise can therefor vary from the V.sub.Ref 's given below.